Datasets:

griffin

/

ChemSum

like 12

Tunable Circularly Polarized Luminescence from Molecular Assemblies of Chiral AIEgens

Circularly polarized luminescence (CPL) is important to chiral photonic technologies. In the molecular systems, besides their intrinsic chemical structures, the architectures of molecular assemblies at the mesoscopic scale also account for the final macroscopic CPL properties. Herein, tunable CPL responses can be induced through architectural regulation of these molecular assemblies in suspension and solid states. A liquid crystalline assembled system of DPCE-ECh exhibiting smectic C* phase with a high dissymmetry factor (gCD = -0.20 and glum = +0.38) is reported. The intense and apparent CD and CPL of the film stem from the intrinsic helical structure of the molecular assembles with weak contribution of Bragg reflection, where the helical axis is perpendicular to the optical axis and is parallel to the direction of the glass substrate. To the best of our knowledge, this large glum factor is very rare for organic compounds even in the assembled state formed by annealing at smectic liquid crystalline temperature. Interestingly, strong CPL signal with glum value of +0.18 is still recorded when DPCE-ECh is annealed at chiral isotropic liquid (Iso*) state. On the other hand, DPCE-ACh can form two coexistence phases of chiral hexagonal and smectic liquid-crystalline phases due to intermolecular hydrogen bonding. The non-periodic molecular orientations of DPCE-ACh break itself helical structure to give a weak negtive CPL signal in 10 -3 order. This work thus provides a new insight for developing efficient chiroptical materials in the aggregate state and profound implications in highperformance CPL-based device.

tunable_circularly_polarized_luminescence_from_molecular_assemblies_of_chiral_aiegens

Introduction<!>Synthesis and characterization<!>Please do not adjust margins<!>Chiroptical in solution and suspension<!>Chiroptical in condensed phase<!>Please do not adjust margins<!>Conclusions<!>Experimental Section<!>Synthesis of 3-(3,4-bis(dodecyloxy)phenyl)-2-(4hydroxyphenyl)acrylonitrile (2).<!>Table of Contents

<p>Development of circularly polarized luminescent (CPL) materials has gained increasing interest owing to their potential applications in stereoscopic optical information storage and processing, [1] optical recognition sensor, [2][3][4][5] quantum computing, [6] and circularly polarized electroluminescence for 3D displays. [7][8][9][10][11][12][13][14][15][16][17][18][19][20] The CPL response of a molecular system is quantified by the dissymmetry factor (glum), [21][22][23][24][25][26][27] where glum = 2(IL-IR)/(IL+IR) and IL and IR denote the emission intensitiy of left-and right-CPL, respectively. The common strategy to achieve CPL is to synthesis molecules with a specific chiral configuration. [28][29][30][31][32] However, the CPL response of synthetic advanced materials not only relies on chiral functions on the molecular level, but also depends on the mesoscopic architectures of the molecular assemblies. Through a self-assembly approach, nanostructured chiral materials are able to transfer and amplify the molecular functions to an amplified CPL property at a specific length scale. [33][34][35][36][37][38][39][40][41] Therefore, investigation on the relationship between hierarchical structure of molecular assembles and their corresponding CPL properties is still an important issue to achieve efficient CPL materials.</p><p>Normally, the luminescence normalized dissymmetry factor of organic system ranges between 10 −4 to 10 −2 . [3,5,[16][17][18][19][20][21][22][23][24][25][26][27] In rare cases, extremely high g-values exceeding 0.2 or even up to 1 have been reported for polyfluorene thin films [8][9][10][42][43][44] or cholesteric organic system [45][46][47][48][49][50][51] . In polyfluorene system, the circular polarization is largely determined by the anisotropy of the cholesteric dielectric medium. The glum value is thickness dependent and strong CPL effect originates from the selective CP reflection due to the long-range cholesteric ordering (Bragg reflection). [10,44] The helical axis of this system is perpendicular to the direction of the substrate. In cholesteric films, hierarchical chiral mesoscopic structures were found in this system. Strong CPL response can be arisen from the sum of two main contributions, including the inherent chiral supramolecular structure and birefringence pattern (Bragg reflection). However, these doped cholesteric systems often suffer the problems of incompatibility and instability. Thus the pursuance of strong chiroptical signal from pure organic compounds remains challenging. Akagi's group reports a glum of +0.29 in chiral bithiophene-phenylene copolymer film annealed in chiral nematic state. [45] They also reports a high glum of -0.23 in chiral disubstituted polyacetylene without no chiral dopant. [46] Recently, chiral molecular assemblies with aggregation-induced emission (AIE) effect become the focus of attention [52][53][54][55] . Benefiting from the enhanced emission intensity upon aggregation of AIEgens, efficient CPL response can be generated in solid state to realize their applications in devices. [17] Although the significant progresses have been achieved to access efficient glum value increasement, the approaches to controlling the mesoscopic structure and the ensuing CPL properties are still limited. [41][42][43][44][45][46][47] Therefore, the CPL properties of chiral luminogens in condensed matter state might have profound implications for the high performance CPL-based device at the macroscopic scale. Herein, two rod-like aggregation-induced emission luminogens with a rigid core containing ester or amide linkage and a cholesterol moiety at one end and long aliphatic chains at the other end, namely DPCE-ECh and DPCE-ACh, are presented and illustrated in Figure 1A. In solid state, DPCE-ECh self-assembles into supramolecular liquid-crystalline smectic C* (SC*) phase and shows an impressive high positive CPL response with glum of +0.380±0.011 and gCD of -0.20. The intense and apparent CD and CPL of the film stem from the intrinsic helical structure of the molecular assembles with small contribution of Bragg reflection, where the helical axis is parallel to the direction of the glass substrate. To the best of our knowledge, this large glum factor is very rare for organic compounds even in the assembled state formed by annealing at smectic liquid crystalline temperature. On the other hand, DPCE-ACh can form two coexistence phases of hexagonal and smectic liquidcrystalline phases with a weak negative CPL response. The glum falls in the range of -0.61 × 10 -3 to -5.96 × 10 -3 . Such non-periodic molecular orientations give a weak CPL signal in 10 -3 order. The large different |glum| is attributed to the amplified artifact induced by the birefringent domains of the thick film.</p><!><p>The synthetic procedures of DPCE-ECh and DPCE-ACh are outlined in Scheme S1. Their structures were confirmed by NMR and high resolution mass spectroscopies (Figure S1-S15). Thermogravimetric analysis (TGA) revealed that two compounds have high decomposition temperatures (Td) which could up to 300 o C (Figure</p><!><p>Please do not adjust margins S16), suggesting that they are thermally stable. In dilute THF solution, DPCE-ECh and DPCE-ACh show adsorption band centered at 364 and 360 nm, respectively (Figure S17A). DPCE-ECh and DPCE-ACh exhibit a similar fluorescence spectrum with a peak maximum centered at 430 nm (Figure 17B). As shown in Table S1, two compounds are weekly emissive in THF solutions with quantum yield ΦF, soln. of 0.005 and 0.003 and emissive in the solid powders with ΦF, solid of 0.12 and 0.114. Their αAIE (ΦF, solid/ΦF, soln.) values were calculated to be 24 and 38, suggesting a typical AIE feature of these compounds.</p><!><p>The weakly-emissive THF solutions of DPCE-ECh and DPCE-ACh become progressively emissive upon addition of water (Figure 2A, S18-19), demonstrating an AIE phenomenon. Such chiral AIEgens with cholesterol moieties are promising candidates for chiral induction, which might be capable to take supramolecular helicity with assistance of the long alkoxy chains. The chiroptical properties of the aggregates generated in THF/water mixtures with different H2O fractions (fw) were then investigated (Figure 2B-E, Figure S20-35). The aggregates of DPCE-ECh with a ester linkage are CD silent regardless of water fraction variation (Figure S20-28). In contrast, at fw ≥ 40%, aggregates of DPCE-ACh with an amide linkage exhibit obvious CD signals with negative and positive Cotton effects at wavelengths between 300 nm and 400 nm (Figure 2B). It is noted that at fw = 60%, an induced positive split-type Cotton effect consisting of a positive Cotton effect at 375 nm and a negative Cotton effect at 338 nm was observed in DPCE-ACh aggregates, suggesting the formation of organized helical superstructures in solution [56][57][58] . The maximum absorption anisotropy factor (gCD) reaches 2.78 × 10 -3 at 375 nm (Figure 2C). In addition, the UV-Vis absorption intensities at long wavelength region began to increase substantially when fw = 60% (Figure S25), this could be attributed to the production of large aggregates with strong light scattering. [59] The above CD and UV data further proved the large-size helical aggregates formation with the addition H2O into THF solvent (fw = 60%). The above CD and UV data reveal the formation of large-sized helical aggregates in THF/H2O mixture with fw of 60%. However, the continuous increment of fw to 90% leads to a dramatic decrease in gCD by an order of magnitude (+2.3 × 10 -4 at 375 nm, Figure 2C), suggesting the dissociation of the chiral helical aggregates. Analog to CD spectroscopy, CPL reflects the chiroptical properties of the luminescent materials upon excitation. Consistent with those of CD results, the isolated species in THF solution and aggregates in THF/H2O mixture of DPCE-ECh are all CPL silent (Figure S29-35). However, the DPCE-ACh aggregates suspended in THF/H2O mixture show positive CPL signal at fw = 40% with glum of 2.0 ×10 -4 (Figure 2D). The detailed glum values of DPCE-ACh aggregates in THF/H2O mixture are depicted in Figure 2E. The maximum glum reaches ~6.0 ×10 -4 at fw between 50% and 70% (Figure 2E). This journal is © The Royal Society of Chemistry 20xx</p><p>Please do not adjust margins Please do not adjust margins To better understand the origins of the chiroptical properties of DPCE-ECh and DPCE-ACh aggregates, scanning electron microscopy (SEM) was employed to study their assembled structures in THF/H2O mixtures with fw varying from 40%-90% (Figure S36-40). For DPCE-ECh, its aggregates keep spherical structure regardless of the water fraction variation (Figure S36). Such a symmetrical morphology leads to a silent CPL response. For DPCE-ACh, its aggregates show an obvious morphological evolution from intertwined network (fw = 40%, Figure S37) to left-handed helical nanofibers (fw = 60%, Figure 3A,B). Benefited from the helical fibrous morphology, DPCE-ACh aggregates show higher dissymmetry factors compared than those of the spherical aggregates. On the other hand, M-helices and Psupra-helices are found in DPCE-ACh suspension (Figure 3A, B). According Michael D. Barnes' report [60], they thought the measured g value in disperse phase represents a weighted average of all possible orientations and interaction with the host. As for DPCE-ACh in suspension, the |glum| value detected in this system can be attributed to cancellation effects in ensemble measurements of a randomly oriented (M-helices and P-suprahelices) bulk sample. As seen in Figure 3A, M-helices sturctrues accounts for most of the proportio, thus leading to positive CPL signal. Careful examination of the chemical structures of the AIEgens may account that intermolecular hydrogen bonding between the amide linkages and the chiral nature of cholesterol moiety of DPCE-ACh serve as the external driving forces for forming the helical self-assembled structure to generate CD and CPL signals. To gain further insight into the dynamic nature of hydrogen bonding within the induced helical fibrous structure [56] , temperature-dependent CD spectra are monitored on the DPCE-ACh-based aggregates (fw = 60%, Figure 3C,D, Figure S41). At the low temperature of 5 o C, the CD spectrum of DPCE-ACh aggregates shows an obvious positive Cotton effect with absorption peak at 375 nm and gives a corresponding gCD value of + 2.58 × 10 -3 . The peak at 375 nm is ascribed to the achiral aromatic rigid core of DPCE-ACh, which is induced by the molecular helical arrangement. It is found that the CD signals of helical aggregates were sensitive to the temperature. The CD signals gradually decrease and completely disappear upon heating to 49 o C (Figure 3C). Such temperature-dependent gabs factor variations are summarized and plotted in Figure 3D. It is noted that the gCD of 375 nm was only + 2.5 × 10 -4 at the high temperature of 49 o C, indicating that the hydrogen bonding became very weak and thus lead to the dissociation of helical assembled structure of DPCE-ACh.</p><!><p>The rod-like molecular structures of two AIEgens with a cholesterol moiety and flexible tails at two ends, makes them promising to form liquid-crystalline phase in a chiral fashion. DPCE-Please do not adjust margins Please do not adjust margins ACh was first explored considering its capability to form helical fibers mentioned above. The phase transition temperatures of DPCE-ACh in solid film are shown graphically in Figure 4A (top, Figure S42). Upon cooling the isotropic liquid of DPCE-ACh to 210 o C, a smectic liquid-crystalline phase with a fan-shaped texture followed by a columnar liquid-crystalline phase with a mosaic texture are observed (Figure 4A, bottom; Figure S43). The molecular orientations in the liquid-crystalline phases are revealed by 1D wide angle X-ray diffraction (1D WAXD, Figure S44-46). The 1D WAXD pattern at 210 o C shows a sharp peak at 2θ = 2.97° and a high-order diffraction peak at 2θ = 5.66°. These two diffraction peaks are associated with a smectic phase structrue [60,61] . with a layer thickness of 3.05 nm (Figure 4B 110) and ( 200) planes of the hexagonal columnar liquid crystals [34,37,63,64] (Figure 4B, S46). Analysis of small-angle X-ray scattering (SAXS) confirms the lamellar and columnar organization of DPCE-ACh with a lamellar thickness of 3.05 nm and a columnar diameter of 4.33 nm, respectively (Figure S46). Electron density reconstruction was caculated according to the method of previous published paper. [60] The electron density map (Figure 4C) of the phase based on the XRD result shows that the high electron density (red) is concentrated at the center of columns and low-density areas (green and blue) are located at the column periphery associated with the alkyl chains and the cholesterol side chains, respectively. On the other hand, as DPCE-ACh is hexagonal columnar mesophase, the average number (n) of molecules per slice of the column could be obtained by the following formula. [65] n = (ɑ 2 )(√3/2)(hρNA/M) where the notation "ɑ" is the hexagonal lattice parameter, NA is Avogadro's number, M is the molecular mass of the compound and the density (ρ) of these samples is set as 1 g/cm 3 . After calculation, the number of molecules (n) in one disk is approximately 2 for DPCE-ACh. Thus, the possible molecular stacking mode for the hexagonal columns is suggested as Figure 4D, in which a slice is composed of two molecules based on the hydrogen-bonding action between N-H and C=O groups. These results certainly support that the hydrogen bond plays the crucial role for inducing the columnar mesophase of the asymmetrical diphenylacrylonitrile derivatives.</p><p>As illustrated in Figure 4E, the film exhibits a negative signal at 415 nm with a gCD of -(1.92 ± 0.063) × 10 -3 after annealing at 180 o C. Moderate profile change in the CD spectrum was obtained by rotating the sample at different angles around the optical axis (Figure 4F, S47-49), suggestting that the LDLB effect [35] (birefringent phenomenon) contributes the final CD. Because DPCE-ACh forms both the coexistence of hexagonal and smectic phases, the nonperiodic molecular orientations break itself helical structure to give a weak CD signal in 10 -3 order. On the other hand, we also investigated the CPL spectra at different angles in both sides in a 7 μm of liquid crystal cell. However, a large difference in |glum| was observed at different angles in both sides and the |glum| falls in the range from -0.61 × 10 -3 to -5.96 × 10 -3 (Figure 4F, S50-51). This phenomenon is attributed to the amplified artifact (Bragg reflection) induced by the birefringent domains of the thick film (7 μm). This journal is © The Royal Society of Chemistry 20xx</p><p>Please do not adjust margins Please do not adjust margins The ester linkage of DPCE-ECh offers only weak the intermolecular interaction and the relationship between the molecular orientations and the chiroptical properties are also investigated. The DSC trace of DPCE-ECh recorded during the first cooling cycle shows three exothermic transitions at around 121, 86 and 30 o C upon cooling from 180 o C (Figure 5A). The POM image shows oily-streak textures that are typical of the sematic phase in the liquid crystalline state at 85 o C (Figure S52). 1D WAXD measurements were then carried out to monitor the structural evolution. It is found that a sharp peak at 2θ = 2.35 o appears at 80 o C, indicating the formation of an ordered structure In addition, a high-order diffraction peak at 2θ = 4.78 o is also observed. The ratio of the scattering vectors of the two peaks is approximately 1:2, indicating the formation of a smectic structure (Figure 5B, S53, 54). Such an ordered structure was retained at the temperature range of 86-33 o C. The transition recorded by DSC at 86 o C corresponds to the clearing point (Figure 5A). In the temperature range of 86-121 o C (DSC), a broad and weak peak compared with that of smectic phase was observed (Figure S55) and we identified it as isotropic liquid phase. To further prove the smectic structure of DPCE-ECh, 2D SAXS and wide-angle X-ray scattering (WAXS) were carried out (Figure 5C,D). The oriented sample for the measurements was prepared by mechanically shearing the melted film at 85 o C. As shown in the illustration in Figure S56, the point-focused X-ray beam was aligned perpendicular to the shear direction. It is noted that the diffuse peaks at smaller (Figure 5C) and larger angles (Figure 5D, marked with dash line) are not orthogonal to each other. This suggests that the director makes a tilt angle with respect to the smectic layer and the angle rotates from layer to layer to form a smectic C phase. CD experiments were conducted to analyse the chiroptical activities of DPCE-ECh solid film on glass slides. No matter rotating or flipping samples, strong and consitent CD signals were obtained (Figure 5E, S57). It implies that the long helical molecular stacking axis was perpendicular rather than parallel to the optical axis (Figure S58A). In addition, the molecular orientation in-plane was supposed to be aligned randomly. In this sense, the LDLB effect in such solid film could be neglected and genuine chiroptical signals from chiral supramolecular structure were resulted. A gCD value of -0.20 at 404 nm was achieved (Figure 5F). Similarly, we also faricated a film in a 7 μm thick liquid crystal cell for CPL measurement. CPL spectra were also obtained by rotating the sample at different angles in both sides (Figure 5G). Strong CPL responses with positive signals were observed in this annealled film with a maximum glum value (average, 70 o C) of +0.380±0.011. However, different glum ranging from 0.342 to 0.438 were observed with vaired angles and sides. The strong CPL response is arisen from the chiral supramolecular structure in which the helical axis is perpendicular to the optical axis and is parallel to the direction of the glass substrate. The large difference in glum (about 0.1) at different angles in both sides is attributed to the birefringence pattern. Because the film thickness is 7 μm for CPL detection which is much thicker than film for CD detection (50 nm), the artifact induced by the birefringent domains is amplified in such a thick film. To the best of our knowledge, such large glum value (+0.380±0.011) with weak contribution of Bragg reflection is very rare for organic compounds.</p><p>With increasing the annealling temperature (>70 o C), the dissymmetry factor (gCD and glum) of the thin film decreases (Figure 5F,H), indicating the dissociation of smectic C phase. Interestingly, when the isotropic liquid of DPCE-ECh was annealed at temperature range of 90-120 o C followed by CPL measurent, strong CPL signal with gCD value of -0.11 and glum value of +0.18 was still recorded (Figure S59-71). The CPL response of the isotropoic state of DPCE-ECh indicates that a twisted orgainzation is still retained in aggregates and such twisting is still sufficient for CPL induction. Thus, we identify this chiral isotropic state as chiral isotropic liquid (Iso*) [66][67][68] in Figure S55 (inset), which was recently discovered new phase.</p><p>Normally, the measured g value in a disperse phase represents a weighted average of all possible orientations. [60] In THF solution, DPCE-ECh and DPCE-ACh (10 -5 mol L -1 ) are soluble and dispersed isolatedly in this solution. Therefore, the measured g value in dilute solution comes from single molecule itself. For single molecule, chiral function is mainly focused on cholesterol unit and luminescent function is mainly focused on diphenylacrylonitrile unit. Hence, no CPL signal is observed in these single molecules. On the other hand, the aggregates of DPCE-ECh keep spherical structure regardless of the water fraction variation. Such a symmetrical morphology leads to a silent CPL response. Meanwhile, in the solid state, combining the X-ray results and chiroptical activity of DPCE-ECh, a semctic C* phase was identified finally. Moreover, SEM textures of the fracture plane of DPCE-ECh with layered and arched structures further support our hypothesis (Figure S72). Such the smectic C* state leads to a giant CPL response. For DPCE-ACh, positive CPL signal is observed due to M-helical nanofibers formation in suspension. Negative CPL signal is observed due to the complex liquid crystalline (H + S) orientations in solid state. These findings demonstrate that the CPL response (intensity and orientation) of synthetic advanced materials not only relies on chiral functions on the molecular level, but also depends on the mesoscopic architectures of the molecular assemblies.</p><!><p>Please do not adjust margins</p><!><p>In summary, two AIEgens with rigid cores containing different linkages are developed. These chiral AIEgens show silent CPL response when existed as species in THF solution. In contrast, tunable CPL response is achieved through regulating their aggregated structures in solution and solid states. Driven by the intermolecular hydrogen bonding in DPCE-ACh, opposite CPL responses with glum in 10 -3 order are obtained from M-helical nanofibrous structure and complex liquid crystalline (H + S) orientations. Meanwhile, DPCE-ECh exhibits a liquid crystalline assembled system (smectic C*) with a high dissymmetry factor (gCD = -0.20 and glum = +0.38). The intense and apparent CD and CPL of the film stems from the intrinsic helical structure of the molecular assembles with weak contribution of Bragg reflection, where the lone helical molecular stacking axis is perpendicular to the optical axis and is parallel to the direction of glass substrate. To the best of our knowledge, this large glum factor is very rare for organic compounds even in the assembled state formed by annealing at smectic liquid crystalline temperature. This path opens new capabilities for structural control of molecular assemblies to generate versatile CPL responses that are inaccessible from isolated AIEgen alone. These findings demonstrate that the CPL response (intensity and orientation) of synthetic advanced materials not only relies on chiral functions on the molecular level, but also depends on the mesoscopic architectures of the molecular assemblies. We hope that the present strategy for constructing CPL-active materials in the condensed matter states will open numerous opportunities for applications in photonic devices.</p><!><p>Chemicals and Methods. All chemicals were purchased from Sigma-Aldrich, J&K Chemical Co. and used as received without further purification unless otherwise specified. Anhydrous THF and CH3CN were used for fluorescence property investigation. Deionized water was used throughout this study. Pre-coated glass plates were used for TLC analysis. Column chromatography was carried out by using silica gel (200-300 mesh) as adsorbent.</p><p>1 H and 13 C NMR spectra were measured on a Bruker ARX 400 NMR spectrometer and reported as parts per million (ppm) from the internal standard TMS. High-resolution mass spectra (HR-MS) were obtained on a Finnigan MAT TSQ 7000 Mass Spectrometer System operated in a MALDI-TOF mode. Thermogravimetric analysis (TGA) was performed on a TA TGA Q5000 under nitrogen at a heating rate of 10 °C min −1 . Differential scanning calorimetry (DSC) analysis was performed on a TA Instruments DSC Q1000 at a heating rate of 5 °C min −1 . The sample size was about 2 mg and encapsulated in hermetically sealed aluminum pans, and the pan weights were kept constant. The temperature and heat flow were calibrated using standard materials such as indium and benzoic acid. Polarized optical microscopy (POM) was carried out to observe the liquid crystalline textures of the samples on a Leitz Laborlux 12 microscope with a Leitz 350 hot stage.</p><p>The morphological structures of the aggregates were investigated by a HITACHI-SU8010 scanning electron microscope (SEM) at accelerating voltages of 200 and 5 kV. Stock solutions of DPCE-ECh and DPCE-ACh in THF (10 -3 mol L -1 ) were prepared. A certain volume (30 μL) of such stock solutions was transferred to small glass vials (5 mL). After addition of appropriate amounts of THF, This journal is © The Royal Society of Chemistry 20xx</p><p>Please do not adjust margins Please do not adjust margins distilled water was added dropwise under vigorous stirring to afford 5 × 10 -5 mol L -1 of DPCE-ECh and DPCE-ACh solutions. The mixtures were dropped on silicon wafers, the solvents were removed under reduced pressure at room temperature, and the SEM images of the aggregates on silicon wafers were taken.</p><p>To identify the liquid crystalline structure of DPCE-ECh and DPCE-ACh, 1D XRD experiments were performed on a Philips X'Pert Pro diffractometer equipped with a 3 kW ceramic tube as the X-ray source (Cu Kα), an X'celerator detector, and a temperature control unit of Paar Physica TCU 100. The sample stage was set horizontally. The diffraction peak positions of the 1D XRD were calibrated with silicon powder for wide-angle region and silver behenate for smallangle region, respectively. The data was collected by a Mar165 detector and calibrated by CeO2 powder. The sample temperature was controlled by a Linkman THMSE600 hot stage. The heating and cooling rates in the experiments were 5 °C /min. The data were collected using an exposure time of 120 s. The 2D SAXS and WAXS data of DPCE-ECh were collected on Xeuss 2.0 (Xenocs, France), and the measurement details are listed in Table S2.</p><p>Absorption spectra were measured on a Milton Roy Spectronic 3000 Array spectrophotometer. Steady-state photoluminescence (PL) spectra were measured on a Perkin-Elmer spectrofluorometer LS 55. The lifetime and the absolute luminescence quantum yield were measured on a Edinburgh FLSP 920 fluorescence spectrophotometer equipped with an integrating sphere (0.1 nm step size, 0.3 second integration time, 5 repeats).</p><p>Circular dichroism (CD) spectra were recorded with a Chirascan spectrometer (Applied Photophysics, England). Circularly polarized photoluminescence (CPPL) spectra of the films and solution were recorded at 50 nm min -1 scan speed with a commercialized instrument JASCO CPL-300 at room temperature with the resolution of 15 nm. The film samples for CD and CPL measurement were prepared by drop-casting on the quartz substrate from the CHCl3 solution (5 mg/mL) of DPCE-ECh and DPCE-ACh, subsequently by volatilization of CHCl3 solvent at room temperature. Samples were subsequently thermally annealed for 45 min at the indicated temperatures. Preparation took place under inert atmosphere in a nitrogen filled glove box. To freeze temporarily the phase of the DPCE-ECh and DPCE-ACh, the film sample was quenched from the indicated temperatures to liquid nitrogen. The CD and CPL response of the quenched sample was recorded over the same time interval (per 3 min) at room temperature. The magnitude of circular polarisation in the excited state is defined as glum = 2 (IL -IR)/(IL + IR), where IL and IR indicate the output signals for left and right circularly polarized luminescence, respectively. Experimentally, the value of glum is defined as ΔI/I = [ellipticity/(32980/ln10)] / (unpolarized PL intensity) at the CPL extremum [69] .</p><p>Electron density reconstruction was caculated according to the method of previous published paper [64] . The diffraction peaks were indexed on the basis of their peak positions, and the lattice parameters and the space groups were subsequently determined. Once the diffraction intensities are measured and the corresponding space group determined, three dimensional (3D) electron density maps can be reconstructed, on the basis of the general formula E(xyz) = Σhkl F(hkl) exp[i2π(hx+ky+lz)] (Eqn. 1) Here F(hkl) is the structure factor of a diffraction peak with index (hkl). It is normally a complex number and the experimentally observed diffraction intensity I(hkl) = KF(hkl)F*(hkl) = K|F(hkl)| 2 (Eqn. 2) Here K is a constant related to the sample volume, incident beam intensity, etc. In this paper we are only interested in the relative electron densities, hence this constant is simply taken to be 1 (Eqn. 4) As the observed diffraction intensity I(hkl) is only related to the amplitude of the structure factor |F(hkl)|, the information about the phase of F(hkl), ϕhkl, cannot be determined directly from experiment. However, the problem is much simplified when the structure of the ordered phase is centrosymmetric, and hence the structure factor F(hkl) is always real and ϕhkl is either 0 or π.</p><p>This makes it possible for a trial-and-error approach, where candidate electron density maps are reconstructed for all possible phase combinations, and the "correct" phase combination is then selected on the merit of the maps, helped by prior physical and chemical knowledge of the system. This is especially useful for the study of nanostructures, where normally only a limited number of diffraction peaks are observed.</p><!><p>A mixture of 4hydroxyphenylacetonitrile (1.62 g, 12.0 mmol), compound 1 (5.7 g, 12.0 mmol) and NaOH (0.96 g, 24.0 mmol) in 60 mL of EtOH and 30 mL THF mixture solution was refluxed for 24 h. After the cooling to room temperature, 24 mL of HCl solution (1 M) was poured into the reaction mixture, then the solvent of C2H5OH was removed by a rotary evaporator. And the water (60 mL) was added. The mixture was extracted with ethyl acetate (3 × 50 mL). The combined organic layers were dried with anhydrous Na2SO4 and evaporated under reduced pressure to obtain the crude product. The residue was purified by silica-gel column chromatography using hexane/ethyl acetate (15:1) as an eluent. Compound 2 was obtained as a light brown powder with 35% yield. (2.48 g, 4.2 mmol). 1 (2.45 g, 17.75 mmol) was stirred and refluxed in 60 mL MeCN and 30 mL THF for 12 h at 90 °C. After cooling the room temperature, the solvent was removed by a rotary evaporator. And then the water (60 mL) was added. The mixture was extracted with CH2Cl2 (3 × 40 mL). The combined organic layers were dried with anhydrous Na2SO4 and evaporated under reduced pressure to obtain the crude product. The residue was purified by silica-gel column chromatography (hexane/ethyl acetate =20:1) to yield 2.01 g (84%) of the product as a yellow powder after removal of the solvent. 1 H NMR (400 MHz, CDCl3) δ (ppm): 7.63 (s, 1H), 7.59 (d, J = 9.2 Hz, 2H), 7.34-7.32 (m, 2H), 6.97 (d, J = 8.8 Hz, 2H), 6.91 (d, J = 8.4 Hz, 1H), 4.67 (s, 2H), 4.30 (q, J = 6.8 Hz, 2H), 4.10-4.05 (m, 4H), 1.89-1.82 (m, 4H), 1.59-1.45 (m, 4H), 1.39-1.27 (m, 35H), 0.89 (t, J = 6.0 Hz, 6H); 13 C NMR (100 MHz, CDCl3) δ (ppm): 168.0, 157. 6, 150.5, 148.4, 140.3, 127.9, 126.5, 126.1, 123.3, 118.1, 114.5, 112.4, 112.2, 107.1, 68.6, 68.4, 64.8, 60.9, 31.3, 29.1, 29.06, 29.01, 28.82, 28.79, 28.77, 28.55, 28.49, 25.44, 25.38, 22.1, 13.56, 13.52. MALDI-TOF-MS (C43H65NO5) Calcd.for m/z = 675.9950, found: m/z = 675.4874 (M + ).</p><!><p>Tunable CPL response is achieved through regulating their aggregated structures in solution and solid states. Driven by the intermolecular hydrogen bonding in DPCE-ACh, opposite CPL responses with glum in 10 -3 order are obtained from M-helical nanofibrous structure and complex liquid crystalline (H + S) orientations. Meanwhile, DPCE-ECh exhibits a liquid crystalline assembled system (smectic C*) with a high dissymmetry factor (gCD = -0.20 and glum = +0.38). The intense and apparent CD and CPL of the film stems from the intrinsic helical structure of the molecular assembles with weak contribution of Bragg reflection, where the lone helical molecular stacking axis is perpendicular to the optical axis and is parallel to the direction of glass substrate.</p>

ChemRxiv

Dithiocarbamates effectively inhibit the α-carbonic anhydrase from Neisseria gonorrhoeae

AbstractRecently, inorganic anions and sulphonamides, two of the main classes of zinc-binding carbonic anhydrase inhibitors (CAIs), were investigated for inhibition of the α-class carbonic anhydrase (CA, EC 4.2.1.1) from Neisseria gonorrhoeae, NgCA. As an extension to our previous studies, we report that dithiocarbamates (DTCs) derived from primary or secondary amines constitute a class of efficient inhibitors of NgCA. KIs ranging between 83.7 and 827 nM were measured for a series of 31 DTCs that incorporated various aliphatic, aromatic, and heterocyclic scaffolds. A subset of DTCs were selected for antimicrobial testing against N. gonorrhoeae, and three molecules displayed minimum inhibitory concentration (MIC) values less than or equal to 8 µg/mL. As NgCA was recently validated as an antibacterial drug target, the DTCs may lead to development of novel antigonococcal agents.

dithiocarbamates_effectively_inhibit_the_α-carbonic_anhydrase_from_neisseria_gonorrhoeae

Introduction<!>Enzymology and CA activity and inhibition measurements<!>Chemistry<!>Bacterial strains and media<!>Antibacterial activity of DTCs against N. gonorrhoeae strains<!>Results and discussion<!><!>Results and discussion<!><!>Results and discussion<!><!>Conclusions<!>Disclosure statement

<p>A decase ago, prokaryotic carbonic anhydrases (CAs, EC 4.2.1.1) were proposed as drug targets for development of novel antibacterials1. CAs catalyse the interconversion between CO2 and bicarbonate, which generate a pH imbalance; CAs are widespread in bacteria and play an important role in various metabolic functions2,3. Bacteria encode at least four genetic families of CAs, including the α-, β-, γ-, and ι-CAs, with many species containing more than one class and more than one CA isoform; however the functions of these different CAs have only recently started to be understood in detail1–3. Although comprehensive in vitro inhibition studies of bacterial CAs are available1,2, these results have only recenlty been validated in vivo. Seminal reports of Flaherty's and Seleem's groups showed that in some bacteria, such as in vancomycin-resistant enterococci (VRE) or Neisseria gonorrhoeae, clinically used sulphonamide CA inhibitors (CAIs) possess potent antibacterial activity4,5. N. gonorrhoeae is a sexually transmitted pathogen that is becoming a global health concern due to increased resistance to a wide range of antibioticsincluding next generation cephalosporins6,7. Acetazolamide, the CAI par excellence, and some of its newly designed derivatives were recently shown to be bacteriostatic against N. gonorrhoeae with minimum inhibitory concentration values as low as 0.25 μg/mL and no toxicity obseved to host cells5. Sulphonamides, of which acetazolamide belongs to, are one of the main classes of CAIs, and their interaction with bacterial CAs from various pathogens has been extensively studied in the last decade8–11. As there is an urgent need for novel antibacterials, including antigonococcal agents, a deeper investigation of CA and profiling various classes of CAIs may be of great interest. A previous study of anion inhibitors found interesting inhibitory effects of N,N-diethyl-ditiocarbamate [5b], which was as a low micromolar inhibitor of the α-CA N. gonorrhoeae (NgCA). Based upon this previous study, we investigated dithiocarbamates as inhibitors of NgCA.</p><!><p>An Applied Photophysics stopped-flow instrument was used to assay the CA- catalysed CO2 hydration activity12. Phenol red (0.2 mM) was used as a pH indicator, working at the absorbance maximum of 557 nm, with 10 mM HEPES (pH 7.4) as a buffer, and in the presence of 10 mM NaClO4 to maintain constant ionic strength, in order to follow the initial rates of the CA-catalysed CO2 hydration reaction for a period of 10–100 s. The CO2 concentrations ranged from 1.7 to 17 mM for the determination of the kinetic parameters and inhibition constants. For each inhibitor, at least six traces of the initial 5–10% of the reaction were used to determine the initial velocity. The uncatalyzed rates were determined in the same manner and subtracted from the total observed rates. Stock solutions of inhibitors (10–20 mM) were prepared in distilled-deionized water, and dilutions up to 0.01 µM were done thereafter with the assay buffer. Inhibitor and enzyme solutions were preincubated together for 15 min at room temperature prior to the assay, in order to allow for the formation of the E-I complex. The inhibition constants were obtained by non-linear least-squares methods using Prism 3 and the Cheng-Prusoff equation, as reported earlier13,14, and represent the mean from at least three different determinations. The NgCA concentration in the assay system was 6.3 nM. The NgCA used was a recombinant enzyme obtained in-house, as described earlier5,15,16.</p><!><p>DTCs 1–30 were previosuly reported by one of our groups17,18 and were of > 99% purity. DTC 31, acetazolamide, buffers and other reagents are commercially available from Sigma-Aldrich (Milan, Italy).</p><!><p>Strains and media used in this study were previously reported by our group5,19. N. gonorrhoeae strains used in the study were clinical isolates obtained from the Centres for Disease Control and Prevention (CDC). Media and reagents were purchased commercially: brucella broth, IsoVitaleX, and chocolate II agar plates (Becton, Dickinson and Company, Cockeysville, MD, USA), yeast extract and dextrose (Fisher Bioreagents, Fairlawn, NJ, USA), protease peptone (Oxoid, Lenexa, KS, USA), haematin, pyridoxal, and nicotinamide adenine dinucleotide (NAD) (Chem-Impex International, Wood Dale, IL, USA), and phosphate buffered saline (PBS) (Corning, Manassas, VA, USA).</p><!><p>The (MICs of DTCs compounds were carried out using the broth microdilution method as described previously5,19. Briefly, bacterial strains were grown for 24 h on GC chocolate agar II, at 37° C in presence of 5% CO2. Then a bacterial suspension equivalent to 1.0 McFarland standard was prepared and diluted in brucella broth supplemented with yeast extract, protease peptone, haematin, pyridoxal, NAD, and IsoVitaleX, to achieve a bacterial concentration of about 1 × 106 CFU/mL. Test agents were added in the 96-well plates and serially diluted along the plates. Plates were then, incubated for 24 h at 37° C either aerobically or in the presence of 5% CO2 before determining the MICs as observed visually.</p><!><p>Sulphonamide-type CAIs were first used to inhibit growth of N. gonorrhoeae in vitro in the 1960s; however, it was not untill the 1990s that Carter's group reported the presumed presence of CAs in N. gonorrhoeae by using a monospecific antibody prepared against the purified Neisseria sicca enzyme15. This enzyme was thereafter purified and characterised in 1997 by Lindskog's group16, who showed that NgCA is an α-class enzyme that possesses a high catalytic activity, with a kcat for the CO2 hydration reaction of 1.7 × 106 s−1 17. The same group showed that NgCA was inhibited by metal complexing anions such as cyanide, cyanate, thiocyanate, and azide (as determined by using the esterase actvity of the enzyme with 4-nitrophenyl acetate as a substrate16) as well as by the sulphonamide acetazolamide (5-acetamido-1,3,4-thiadiazole-2-sulphonamide)16. Thereafter, we reported a comprehensive anion inhibition study of NgCA [5b], which found that the most effective inhibitors were sulfamide, sulphamic acid, and N,N-diethyl-dithiocarbamate. This compound possesses the CS2− zinc-binding group (ZBG), also present in trithiocarbonate (TTC)17, which has been shown via X-ray crystallography on human CAs (hCAs) to bind in a monodentate fashion to the metal ion from the enzyme's active site to displace the nucleophile (water or hydroxide ion) that is essential in the catalytic process17. The X-ray structure of TTC bound to hCA II led thereafter to the discovery of DTCs and their derivatives (monothiocarbamates and xanthates) as potent CAIs18,20. X-ray crystallography of some DTCs bound to hCA II demonstrated that their ZBG is coordinated in a monodentate fashion to the metal ion whereas the organic scaffold participates in a range of favourable interactions with the active site amino acid residues18 – Figure 1.</p><!><p>(A) Surface representation of hCA II active site in adduct with superimposed trithiocarbonate (cyan, PDB 3K7K) and the DTC morpholinocarbodithioate 23 (magenta, PDB 3P5A). The hydrophobic half of the CA active site is shown in red, and the hydrophilic one in blue; the proton shuttle residue His64 is shown in green. Cartoon view of hCA II active site in complex with B) trithiocarbonate and C) DTC 23.</p><!><p>Thus, we decided to investigate a series of previously reported DTCs18, types 1–30 together with the N,N-diethyl derivative 31, for their interaction with NgCA (Table 1). The following structure-activity relationship (SAR) may be observed from the data presented in Table 1:</p><!><p>Inhibition constants (KIs) of DTC inhibitors 1–31 against hCA I, II, and NgCA by a stopped flow CO2 hydration assay, using acetazolamide (AAZ) as the standard drug12.</p><p>aMean from three different assays, determined using a stopped flow technique (errors were in the range of ± 5–10% of the reported values); bfrom ref. [5b].</p><p>The most effective NgCA inhibitors among the investigated DTCs were compounds 1, 20 and 29, which showed KIs in the range of 83.7–136 nM. It is interesting to note that both 20 and 29 possess the same scaffold of piperazine-dithiocarbamate. However, in the case of 29 a second DTC function is incorporated, whereas for 20, a bulkier cyclohexyl-aminocarbonylmethyl moiety is present. This leads to an increased inhibitory effect in the case of 20 compared to 29 (84.4 versus 136 nM, Table 1), probably due to favourable contacts between the bulky tail and amino acid residues from the active site. The second observation pertains to compounds 1 and 2. Derivative 1 incorporated two ZBGs, the DTC and the sulphonamide ones, whereas the second structurally related derivative (2) lacks the sulphonamide moiety. It is likely in the case of 1 that sulphonamide is the dominant interacting group and participates in the enzyme inhibition process by binding to the zinc ion in the active site. This is however impossible for 2, which exhibited 3.1 times weaker NgCA inhibitory activity compared to 1. However, derivative 2 still significantly inhibited the NgCA CO2 hydrase activity with a KI of 259 nM.</p><p>Another small group of DTCs, including 2, 9, 13, and 28 showed KIs in the range of 242 – 297 nM, which indicates that they are effective NgCA inhibitors. The next most effective inhibitors showed KIs between 300 and 500 nM and included 4, 5, 10–12, 14, 15, 21, 23, 25, and 30. These compounds incorporated a variety of diverse aliphatic, aromatic, and heterocyclic scaffolds, and are derivatives of both primary and secondary amines. This proves that many diverse chemical entities may lead to the development of efficient DTC inhibitors of NgCA (Table 1).</p><p>The least effective inhibitors were 3, 6–8, 16–19, 22, 26, and 27, which showed KIs in the range of 514–827 nM. Finally, 31, the lead compound was the least effective DTC inhibitor, with a KI of 5100 nM. In contrast, acetazolamide, a sulphonamide derivative, was an effective NgCA inhibitor, with an activity in the same range as the most effective DTCs mentioned above (Table 1).</p><p>Many of the investigated DTCs were much more effective as inhibitors against hCA II than NgCA, whereas their activity on hCA I was in the same range as against the bacterial enzyme, i.e. in the high nanomolar range.</p><!><p>A subset of DTCs were selected for antibacterial testing against three clinical strains of N. gonorrhoeae. It has previously been established that bacteria will become less susceptible to CAIs in conditions that contain elevated levels of CO221. Molecules were assayed in both ambient air conditions as well as conditions containing 5% CO2 to assess for activity at the proposed intracellular NgCA. The three strains tested displayed reduced susceptibility towards the molecules under elevated CO2 conditions suggesting that inhibition of NgCA is, at least partially, responsible for the antimicrobial activity of these molecules. The control antibiotic azithromycin, which has a different mechanism of action, did not display differential activity based on the culture conditions. This result provides confidence that the difference in CO2 levels did not have unintended effects on the bacteria that would result in non-specific reduced susceptibility to the test agent.</p><p>It was observed that in this cohort, three DTCs, 1, 22, and 24 exhibited moderate antigonococcal activity. DTC 1 was the most potent molecule with a MIC value of 1–2 µg/mL against N. gonorrhoeae (Table 2). This was followed by 22 (MIC = 2–4 µg/mL) and 24 (MIC =4–8 µg/mL). DTCs 23 and 25 each displayed weak antibacterial activity against N. gonorrhoeae with MIC values ranging from 8 to 32 µg/mL. It is interesting to note that while 1 was the most potent molecule against both NgCA and N. gonorrhoeae, the DTCs that exhibited moderate potency against N. gonorrhoeae (22 and 23) were among the weaker analogues versus NgCA (KIs > 500 nM). Moreover, the weakest DTCs, in terms of antigonococcal activity, were 23, 25, 28, 29, and 30 with MIC values > 8 µg/mL; however, these molecules were more potent inhibitor of NgCA with activities in the range of 136 − 460 nM. Several of these molecules contain polar functional groups such as morpholine (23), piperazine (28) and Di-DTC (29) moieties that may have an adverse effect on molecule accumulation within the Gram-negative bacterial cell, thus leading to reduced antigonococcal activity. As for DTC 25, this molecule contains hydrophobic linear alkyl chains that give rise to additional rotatable bonds that also may have an adverse effect on accumulation into Gram-negative bacterial cells22,23. In summary, while the DTCs displayed moderate-to-weak antibacterial activity against the N. gonorrhoeae strains tested, the data does suggest that the DTC functionality may be a useful modification to incorporate into a drug design campaign for development of new anti-gonococcal agents.</p><!><p>Minimum inhibitory concentrations of DTCs versus N. gonorrhoeae clinical isolates.</p><p>aIndicates incubation in presence of 5% CO2. bIndicates in ambient air.</p><!><p>NgCA, a high-activity α-CA present in the genome of N. gonorrhoeae, was investigated for potential inhibition by a series of 31 DTCs derived from both primary and secondary amines. NgCA was inhibited by all investigated derivatives, with KIs in the range of 83.7 nM − 5.1 µM. The most effective NgCA inhibitors were contained piperazine-dithiocarbamates that showed activity with KIs < 140 nM; however, these molecules did not display antibacterial activity in vitro against N. gonorrhoeae. Conversely, DTCs containing more hydrophobic amines did exhibit moderate antibacterial activity even though these analogs possessed reduced NgCA activity. This data suggests that DTCs could be incorporated as the zinc-binding groups in place of sulphonamides, into more traditional CAI molecular scaffolds. Since antibiotic resistance is well documented against many N. gonorrhoeae strains worldwide, finding alternative chemotypes to presently used drugs is relevant. Our study provides interesting steps regarding developing these types of enzyme inhibitors.</p><!><p>The authors have no relevant affiliations of financial involvement with any organisation or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. CT Supuran is Editor-in-Chief of the Journal of Enzyme Inhibition and Medicinal Chemistry. He was not involved in the assessment, peer review, or decision-making process of this paper. The authors have no relevant affiliations of financial involvement with any organisation or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.</p>

PubMed Open Access

Cytochrome aa3 oxygen reductase utilizes the tunnel observed in the crystal structures to deliver O2 for catalysis

Cytochrome aa3 is the terminal respiratory enzyme of all eukaryotes and many bacteria and archaea, reducing O2 to water and harnessing the free energy from the reaction to generate the transmembrane electrochemical potential. The diffusion of O2 to the heme-copper catalytic site, which is buried deep inside the enzyme, is the initiation step of the reaction chemistry. Our previous molecular dynamics (MD) study with cytochrome ba3, a homologous enzyme of cytochrome aa3 in Thermus thermophilus, demonstrated that O2 diffuses from the lipid bilayer to its reduction site through a 25-\xc3\x85 long tunnel inferred by Xe-binding sites detected by X-ray crystallography.1 Although a similar tunnel is observed in cytochrome aa3, this putative pathway appears partially occluded between the entrances and the reduction site. Also, the experimentally determined second-order rate constant for O2 delivery in cytochrome aa3 (~108 M\xe2\x88\x921s\xe2\x88\x921) is 10 times slower than that in cytochrome ba3 (~109 M\xe2\x88\x921s\xe2\x88\x921). A question to be addressed is whether cytochrome aa3 utilizes this X-ray inferred tunnel as the primary pathway for O2 delivery. Using complimentary computational methods including multiple independent flooding MD simulations and implicit ligand sampling calculations, we probe the O2 delivery pathways in cytochrome aa3 of Rhodobacter sphaeroides. All of the O2 molecules that arrived in the reduction site during the simulations were found to diffuse through the X-ray observed tunnel, despite its apparent constriction, supporting its role as the main O2 delivery pathway in cytochrome aa3. The rate constant for O2 delivery in cytochrome aa3, approximated using the simulation results, is 10 times slower than in cytochrome ba3, in agreement with the experimentally determined rate constants.

cytochrome_aa3_oxygen_reductase_utilizes_the_tunnel_observed_in_the_crystal_structures_to_deliver_o2

Introduction<!>System preparation<!>Simulation protocols<!>Production runs<!>Flooding simulations<!>Implicit ligand sampling (ILS)<!>O2 delivery pathway to the reduction site<!>Comparisons of O2 delivery pathways in cytochrome aa3 and cytochrome ba3<!>Delivery rate of O2<!>Constrictions along the O2 delivery pathway in cytochrome aa3<!>Conclusion

<p>A-family heme-copper oxygen reductases (HCOs) including the aa3-type cytochrome c oxidases (cytochrome aa3) are the terminal respiratory oxygen reductases of all eukaryotes and many bacteria and archaea.2–5 These enzymes oxidize cytochrome c and reduce O2 to water, and use the free energy from the O2 reduction reaction to power the biosynthesis of ATP.6–12 The atomic structures of cytochrome aa3 from Rhodobacter sphaeroides (R.s),13,14 Paracoccus denitrificans15 and bovine heart mitochondria16–19 have been determined by X-ray crystallography. The bacterial enzymes comprise 3–4 subunits (Fig. 1A), whereas the mitochondrial enzymes contain at least ten additional accessory subunits.20,21 The smaller bacterial enzymes are widely used as models for the mitochondrial enzymes. A homologue of cytochrome aa3 is cytochrome bo3 of E. coli,5,22,23 which uses ubiquinol instead of cytochrome c as the electron donor.</p><p>The reduction of O2 takes place in Subunit I (SI), a 12-TM helical subunit, containing heme a and the O2 reduction site composed of heme a3 and CuB (Fig. 1A). The reaction requires four electrons and eight protons, of which four are for the reduction of O2 and four are translocated across the membrane. O2+8HN,chem,pump++4e−→2H2O+4HP,pump+</p><p>The four electrons are provided by the sequential oxidation of four cytochrome cred molecules by the one-electron redox center CuA located in the periplasmic domain of Subunit II (SII). Electrons are sequentially shuttled to heme a and then to the reduction site. All of the protons come from the N (electrically negative) side of the membrane and are transferred via two proton-conducting input channels, the D and K channels.8,20 The D channel transfers two of the chemical protons and all of the pumped protons. It is 25–30 Å long and comprises a continuous hydrogen-bonded network formed by water molecules and conserved polar amino acids linking residue D132 (numbering for the R. sphaeroides enzyme) anchored on the surface of the N side and residue E286 located near heme a and the reduction site. The K channel, which transfers two chemical protons, begins at residue E101 in SII and includes residue K362 near the SI–SII interface. The two channels are not redundant, each providing chemical protons during different steps in the catalytic cycle. Starting with a fully reduced enzyme, the binding of O2 to heme a3 initiates the O2 reduction chemistry and reaction steps associated with proton pumping. The O2 delivery pathway, however, has not been well characterized and is the topic of the current work.</p><p>A previous MD study from this laboratory with a B-family HCO, cytochrome ba3 from Thermus thermophilus1 demonstrated that O2 diffuses from the lipid bilayer to the reduction site through a hydrophobic tunnel that is the same as that defined by Xe-binding sites observed in the crystal structure of the protein (Fig. 1D).24,25 This tunnel is also apparent in the absence of bound Xe in all of the available crystal structures of cytochrome ba3.26–28 The O2 delivery pathway in cytochrome ba3 has two entrances enabling O2 access to the protein from the lipid bilayer. Both entrances begin at SI-lipid interfaces (Fig. 1D); one (termed TM1–3) is spanned by TM1, TM2 and TM3 helices, and other (termed TM4–5) is spanned by TM4 and TM5 helices. This structural feature is a "static" tunnel and is referred to as the "X-ray inferred" pathway.</p><p>An equivalent tunnel with the same two entrances (Fig. 1A–C) is observed in the crystal structures of the A-family oxygen reductases including cytochrome aa3 from different sources.13,19,24,29–31 However, in the A-family enzymes, these tunnels appear to be partially occluded (Fig. 1B–C), consistent with the experimental apparent second order rate constant for O2 diffusion from the aqueous solution to the reduction site of cytochrome aa3 (108 M−1s−1) being 10 times slower than for cytochrome ba3.32,33</p><p>The A-family oxygen reductases have a 7-TM helical subunit III (SIII) that is not present in cytochrome ba3 or other B-family HCOs, and the TM4–5 entrance of the putative O2 pathway in SI is adjacent to SIII. However, there are no data that indicate a functional role of SIII in O2 delivery to the reduction site that might result from its proximity to the tunnel entrance. Although the elimination of SIII has deleterious effects, the A-family enzymes remain functional.8,34–37 X-ray structures of A-family HCOs without SIII indicate no structural perturbations to SI or SII or to the O2 pathways14,15,38–40 (Fig. 1B–C).</p><p>Two MD studies have, however, reached opposite conclusions about the use of the X-ray inferred tunnel by O2 to reach the reduction site in the A-family HCOs. An early (1998) MD study by the Schulten group41 applied the locally enhanced sampling (LES) technique42 to simulate O2 diffusion in cytochrome aa3 oxygen reductases from P. denitrificans (P.d) and bovine. They observed an O2 molecule initially placed at the reduction site exiting via the entrance of the X-ray inferred tunnel in each case. However, these LES simulations41 were performed in vacuum and lasted only several picoseconds. Because LES accelerates O2 diffusion by softening interactions between O2 and its surroundings, it might not correctly capture the dynamics that are functionally relevant for O2 delivery.</p><p>More recently, an MD study of the A-family R.s. cytochrome aa3 by Oliveira et al43 probed the O2 delivery pathway by performing five 100-ns flooding simulations, in which multiple copies of O2 were explicitly included and simulated with the rest of the system (i.e. protein, lipids, water and ions). Unlike LES, flooding simulation is based on equilibrium, conventational simulations with no perturbation and rescaling of interactions between atoms in the system. However, no O2 entry was observed during these simulations. They performed energy estimations using the implicit ligand sampling (ILS) technique, a post-simulation analysis used in probing high affinity O2 sites,44 and the results suggested two alternate dynamically formed tunnels as preferred O2 pathways over the X-ray inferred tunnel.</p><p>Experimental mutagenesis data which support the use of the X-ray inferred tunnel as the O2 delivery pathway in the A-family HCOs are based on amino acid residues, such as V279 of P.d. cytochrome aa345 and V287 of E. coli cytochrome bo346 (equivalent to V287 R.s. cytochrome aa3), that are located very close to the reduction site rather than within the tunnel. Also, although an X-ray crystallographic experiment detected two Xe binding sites in the R.s. enzyme,13 those sites are located near the entrances of the pathway and therefore do not necessarily define a clear pathway for O2 delivery.</p><p>Since the two previous MD simulations for O2 diffusion to the reduction site of A-family oxygen reductases arrived at different conclusions, the present study re-examines the issue by independently performing both ILS analyses and an extended set of flooding simulations of the diffusion of O2 to the reduction site of the R.s. cytochrome aa3. The X-ray model used for the simulations includes SI and SII (Fig. 2). To improve the statistics on the O2 delivery pathways compared to previous studies, twenty independent flooding simulations with different starting points were performed and each was extended to 150 ns. ILS analyses were performed on an independent 200-ns MD simulation. The results clearly show that the only pathway used by O2 to reach the reduction site is the X-ray inferred tunnel, similar to what has been shown for the B-family cytochrome ba3 from T. thermophilus (T.t.1 Constrictions along the pathway result in a rate of O2 delivery that is at least 10 times slower than the rate observed32,33 and calculated1 for cytochrome ba3.</p><!><p>The membrane-embedded model of R.s. cytochrome aa3 was prepared using the 2.0-Å crystal structure (PDB 2GSM) as the protein model. The structure comprises SI (catalytic subunit), which binds low-spin heme a, high-spin heme a3 and CuB cofactors, SII, which binds the CuA complex, and 282 water molecules. Hydrogen atoms were added using PSFGEN in VMD.48 Histidine residues except H102, H333, H334, H411, H419 and H421 of SI were in the HSE tautomeric form (Nε atom of the imidazole ring carrying proton). H102 and H421 are ligated to the Fe atom of heme a, H284, H333 and H334 are ligated to CuB, H419 is ligated to the Fe atom of heme a3, and H411 forms a hydrogen bond with the propionate A of heme a3. The carboxylate side chain of residue E286, for which pKa has been experimentally estimated to be >9,49 was assigned to be protonated.</p><p>The first principal axis of the protein was aligned with the z axis (membrane normal) using the OPM (Orientations of Proteins in Membranes) database.50,51 The protein was then inserted into a patch of POPE (1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine) bilayer. Lipids that overlapped the protein were removed, keeping 132 lipids in the periplasmic and cytoplasmic leaflets. The membrane-embedded enzyme complex was then solvated with water. Water molecules that were in the membrane (−18 < z < 18), except those from the crystal structure, were removed, keeping 20,637 water molecules. Finally, 0.2 M NaCl (78 Na+ and 79 Cl− ions) was added to neutralize and ionize the system, resulting in a fully solvated model of 104,710 atoms.</p><p>The CHARMM22 force field with ϕ/ψ corrections52,53 was used to describe the protein and the heme cofactors, CHARMM3654 for the lipids, and the TIP3P model55 for water molecules. The cofactors were treated in the reduced state. The partial charge of 0.383 was used for CuB according to Hofacker and Schulten.41 The vdW parameters of Cu with ε = 0.19 kcal/mol and Rmin =1.4 Å were from Fuchs et al.56 The force field parameters for the hydroxyethylfarnesyl side chain, attached to the porphyrin ring of hemes a and a3, were not available in the CHARMM force field library; they were constructed by analogy using parameterized alcohol, aldehyde, alkene, and alkane fragments;57 the complete structure of the hemes with the atomic partial charges is shown in Fig. S1. For the CuA complex, the partial charges of its Cu atoms and its ligated amino acids were assigned according to Hofacker and Schulten.41</p><!><p>MD simulations performed to prepare the systems consisted of the following steps: (1) 0.5-ns melting of lipid tails during which only the lipid tails were allowed to move in order to achieve better packing of lipids around the inserted protein; (2) 0.5-ns simulation with restraints (k = 1 kcal/mol/Å2) applied to heavy atoms of the protein and cofactors (all lipid atoms and water moving) and with harmonic potentials (k = 0.1 kcal/mol/Å2) applied to keep water out of the membrane; (3) 0.5-ns simulations with only backbone atoms of the protein and heavy atoms of the cofactors restrained (k = 1 kcal/mol/Å2); (4) 1-ns simulation with only Cα atoms of the protein and heavy atoms of the cofactors restrained; and (5) 20-ns unrestrained relaxation. Energy minimization (1,000 steps) was performed at the beginning of Steps 1, 2 and 3 using the conjugate gradient algorithm. To maintain the ligation of CuB to H284, H333 and H334 and thereby the structure of the reduction site, the His-CuB, and heme a3 Fe-CuB connections were described by bonded interactions (k =200 kcal/mol/Å2 for bonds and k =50 kcal/mol/rad2 for angles).</p><p>All simulations were performed using NAMD258 with a time step of 2 fs and with the periodic boundary condition (PBC). All bonds involving hydrogen atoms were kept rigid using the SHAKE algorithm.59 To evaluate long-range electrostatic interactions in PBC without truncation, the particle mesh Ewald (PME) method60 with a grid density of 1/Å3 was used. The cutoff for van der Waals interactions was set at 12 Å. All of simulation steps except the melting of lipid tails (Step 1) were performed in a flexible cell, which allows the system to change its dimensions independently as an NPT ensemble. The temperature was maintained at 310 K by Langevin dynamics61 with a damping coefficient γ of 1 /ps. The Nosé-Hoover Langevin piston method61,62 with a piston period of 200 fs was used to maintain the pressure at 1 atm.</p><!><p>The production runs consist of a 200-ns apo simulation performed in the absence of O2 molecules (used for ILS analysis and starting flooding simulations), and a set of twenty 150-ns flooding simulations in which a large number of O2 molecules were added. The apo simulation started from the 20-ns time point of the relaxation simulation. The set of twenty flooding simulations started from snapshot taken at different time points from the 200-ns apo simulation described above: 4 snapshots (at 25-ns, 100-ns, 150-ns, and 200-ns time points, respectively) were selected from the apo simulation and each one were used to seed 5 indepedent simulations, resulting in an ensemble of twenty simulations.</p><!><p>Flooding simulations were carried out to probe potential O2 delivery pathways and the dynamics associated with the O2 delivery process. To maximize the sampling of O2 delivery pathways within a limited timescale (150 ns), 130 O2 molecules, corresponding to a concentration of 210 mM with respect to the volume of the simulation box, were added to the equilibrated structure of the membrane-embedded cytochrome aa3. At the start of the simulation, 70 O2 molecules were placed in the membrane and 60 molecules in the aqueous phas generating a concentration of 185 mM in the aqueous phase with all O2 molecules occupying space outside the protein (Fig. 2). Twenty 150-ns simulations, adding upto a total of 3,000 ns, were carried out to probe for potential O2 pathways. The simulated O2 molecules are described by the standard CHARMM force field.52 The partial charges for the oxygen atoms of O2 are +0.021 and −0.021, respectively. Its intramolecular interactions are described by the bond distance of 1.23 Å and the spring constant of 600 kcal/mol/Å2. The vdW parameters of the oxygen atoms are with ε = −0.12 kcal/mol and Rmin =1.7 Å. We note that the O2 model with the partial charges of ±0.021 and the one with the partial charges of 0 used in the ILS calculations, in which no electrostatic terms are included, show negligible differences in O2 solvation in the aqueous solution and O2 partitioning in the membrane (Fig. S3–S4 and Table S1). Moreover, it is important to note that although the phrase "O2 ligand for heme" is tagged in the CHARMM topology file containing both O2 and hemes, these charges of ±0.021 are far away from the strongly polarized O2 molecule when ligated to the heme iron. A calculation by Daigle et al63 showed that oxygen atoms of an O2 molecule ligated to heme are strongly polarized with the partial charges of −0.18 and −0.32, clearly indicating that the slightly charged model (±0.021) used in the flooding simulations is not representing a heme-bound O2. The very small charges of ±0.021 used in the present study (and in other simulation studies64,65) are used to take into account a portion of the polarization that O2 would experience when it approaches to strongly charged portions of the protein. In such cases, the rotation of O2 can orient the small introduced dipole with the surrounding field, thereby giving a small portion of polarization effects.</p><p>To identify whether O2 delivery events took place during the simulations, we used the 6-Å distance cutoff from CuB as the criterion. Then, we obtained the identity of the delivered O2 molecules and examined the simulation trajectories whether they were indeed localized in the reduction site.</p><!><p>Complementary to the flooding simulations in which ligand diffusion is explicitly probed, ILS was employed to identify potential regions and pathways for O2 insertion that may not be sufficiently sampled by flooding simulaitons. ILS calculates ligand-interaction energies (Ei) in any position inside the protein over an ensemble of protein conformations and ligand orientations,44 which estimate a 3D free energy map of inserting an O2 molecule at position i (ΔGi). ΔGi=−RTlnpip0=−RTln<e−Ei/RT>p0where p0 (in vacuum) = 1 and pi is the probability of inserting an O2 molecule at position i.</p><p>Following the assumption that small, hydrophobic gases weakly interact with proteins and therefore do not affect the protein structure and dynamics, the 200-ns apo trajectory (10,000 frames) was analyzed for O2 pathways. O2 molecules were sampled in a 55 × 50 × 75 Å3 grid with spacing of 1 Å, covering the entire cytochrome aa3 enzyme. Ten orientations of O2 were sampled in each subgrid, which contained 3×3×3 interaction sites. The solvation free energy of O2 (ΔGsol) was used as the reference for calculating the partitioning free energy of O2 (ΔGi,sol). ΔGi,sol=ΔGi−ΔGsolΔGsol was independently calculated over a 30×30×30 Å3 using of NaCl solution ILS and free-energy perturbation (FEP).66 Both techniques yield ΔGsol values of 2.1 kcal/mol, which is consistent to a previous calculation by Cohen et al.44</p><!><p>To unequivocally probe for pathways used by O2 to diffuse to the reduction site of cytochrome aa3, flooding simulations with 130 O2 molecules were performed. Twenty simulations were performed using different starting points taken from the equilibrated system of the apo simulation: at t = 25 ns, 100 ns, 150 ns or 200 ns. The O2 molecules were initially placed outside the protein, 70 molecules in the membrane and 60 molecules in the aqueous solution. Each simulation lasted 150 ns, but it took only ~20 ns to achieve steady distributions of O2 in both the membrane and aqueous phases (Fig. 2), where ~110 O2 molecules resided in the membrane and ~20 resided in the aqueous solution. In 15 of 20 simulations, O2 molecules were observed to enter the reduction site; the number of this event ranged from 1 to 7 with an average of 2 events per simulation, corresponding to an average of O2 entry in every 75 ns. Based on the 6-Å distance cutoff from the center of Fe of heme a3 and CuB, the residence times of O2 in the reduction site ranged from ~1 ns to ~90 ns (Fig. 3).</p><p>O2 molecules that reached the reduction site were identified and their diffusion dynamics was visually examined in order to locate the O2 delivery pathways. Their trajectories are shown in Fig. 4; each colored line represents an individual O2 molecule. The results of flooding simulations showed that the delivered O2 molecules enter the reduction site via a pathway, which contains three entry branches (entrances) accessible from the membrane. One of the entrances begins at a lipid-protein interfacial region formed by TM1, TM2 and TM3 (TM1–3) helices, one begins at the TM4–TM5 (TM4–5) interface, and one begins at the TM5–TM6 (TM5–6) interface. In some of the simulations, O2 is observed to enter via all three entrances, while in others, the entries occur via only one or two entrances (Fig. 4). The location of the pathway in cytochrome aa3 is similar to the one previously defined for O2 delivery in cytochrome ba3, which resembles a Y-shaped tunnel (shown in Fig. 1C). The O2 delivery pathway in cytochrome ba31 corresponds to a two-branched hydrophobic tunnel as was also determined to bind Xe by X-ray crystallography.24,25 The TM1–3 entrance of cytochrome aa3 is equivalent to Branch A of cytochrome ba3, so it is referred to as Branch A. The TM4–5 entrance is equivalent to Branch B, so it is referred to as Branch B. The TM5–6 entrance was not found in cytochrome ba3 and is denoted as Branch C.</p><!><p>The slower O2 delivery rate in cytochrome aa3 (1×108 M−1s−1) compared to the one in cytochrome ba3 (1×109 M−1s−1)32,33 is probably due to the presence of diffusions barriers within the protein. However, the inability to observe O2 delivery in some of the flooding simulations makes it difficult to compare our current results with cytochrome aa3 and our previous ones with cytochrome ba3 using the results of flooding simulations. Therefore, ILS analyses were performed to calculate thermodynamically favorable O2 regions within cytochrome aa3. The results of ILS were used to map out potential O2 delivery pathways, which were then compared to the pathways obtained from the flooding simulations and to the results of cytochrome ba3 obtained from our previous study.1 The comparison between cytochrome aa3 and cytochrome ba3 is shown in Fig. 5A–B as 3D free energy isosurfaces, in which the colored surfaces represent free energy states of O2 insertion.</p><p>For cytochrome ba3, the entire O2 delivery pathway is energetically favorable as indicated in Fig 5B by the Y-shaped red surface corresponding to ΔG of −3.3 kcal/mol. For cytochrome aa3, the pathway identified by the flooding simulations are found to be less favorable for O2 insertion relative to the one in cytochrome ba3. The pathway contains several higher energy regions between the entrances and the reduction site illustrated in Fig 5A by the discontinuity of the magenta surfaces corresponding to ΔG of −1.8 kcal/mol forming diffusion barriers. Here, O2 molecules diffusing from any of the three entrances to the reduction site encounter a barrier located in the vicinity of residues F172 and G283.</p><p>In both cytochrome aa3 and cytochrome ba3, the results of flooding simulations are correlated to the results of ILS calculations. In our previous study of O2 delivery in cytochrome ba3, although flooding simulations were performed for only 50 ns using the O2 concentration of 210 mM, we were able to observe O2 delivery on an average of 20 events.1 For cytochrome aa3, the simulations lasted much longer (150 ns), but only two delivery events occurred on the average, which represent a 30-fold decrease compared to cytochrome ba3. Fewer O2 delivery events indicate that the pathway in cytochrome aa3 is less accessible to O2 than the one in cytochrome ba3.</p><p>The study by Oliveira et al43 with the same R.s. cytochrome aa3 presented five 100-ns flooding simulations but found no O2 entry. These results are consistent with the present study insofar we also found free energy barriers that limit the rate of O2 diffusion to the reduction site. Five of the 20 simulations performed in the present study exhibited no O2 entry events while the rest observed only few events (Table 1). Based on their ILS analyses, however, Oliveira et al43 concluded that O2 prefers to use alternate pathways rather than the X-ray inferred pathway to reach the reduction site43 although no passage of O2 through such alternative pathways was observed.</p><p>Different free energy isosurfaces, especially the ones at higher free energy contours (e.g., at ΔG = 1 kcal/mol, shown in Fig. 5C), were used to test for the existence of potential O2 pathways besides the X-ray inferred pathway. However, despite the free energy barriers within the X-ray inferred pathway, the results of flooding simulations show that all of the 39 O2 molecules reaching the reduction site in cytochrome aa3 during the flooding simulations from the solutions (Table 1) use this X-ray inferred pathway, suggesting that the X-ray inferred pathway is the primary O2 delivery pathway in cytochrome aa3.</p><!><p>To directly connect the results of the present study to the ones experimentally determined from time-resolved absorption spectroscopy,32,33 we provide an approximation of the second order rate of O2 delivery (kobs) to cytochrome aa3. Since flooding simulations provide dynamic details of O2 diffusion, the obtained data can be used to semi-quantitatively describe steps associated with O2 delivery. The delivery of O2 to the reduction site involves two major steps: 1) the diffusion of O2 from the solution to the entrance(s) of the pathway and 2) the migration of O2 from the entrance(s) to the reduction site. Assuming that the consumption of O2 is 100% efficient once in the reduction site, kobs can be calculated by using the following steady-state kinetics model: O2+E⇌k−1k1E(O2)→k2E(O2)catwhere E(O2) is the species in which O2 has arrived at one of the entrances and E(O2)cat is when O2 is in the reduction site. kobs is defined as: kobs=[O2]k1k2[O2]k1+k−1+k2.k1 is the rate constant of O2 reaching the entrance(s) of the pathway describing the diffusion step of O2 from the solution to the reduction site. It was calculated as the reciprocal of the product of the time taken to observe the first O2 molecule diffusing from the solutions to the entrance(s) of the pathway during the simulation (tent) and the aqueous concentration of O2 ([O2]). The average tent calculated from all 20 simulations is 2.7 ns and the average [O2] in the aqueous solution is 67 mM (Table 1), so k1 is ~5.5×109 M−1s−1.</p><p>k−1 is the dissociation rate constant of O2 from the entrance of the pathway to the membrane. It is the product of k1 and the standard concentration (1 M) over the partitioning ratio of O2 at the entrance and the membrane defined as Pent,mem). Pent,mem is inversely proportional to exponent of the substraction of ΔGent from ΔGmem. ΔGmem is ΔG of O2 in the membrane with respect to the aqueous solution and is −2 to −1.5 kcal/mol.1,67 ΔGent is ΔG of O2 at the entrances of the pathway with respect to the aqueous solution. Because the ΔG contour with ΔG = −3.3 kcal/mol is found at all of the entrances according to ILS calculations (Fig. 5A), ΔGent is approximated to be −3.3 kcal/mol, indicating that O2 is 1.3 to 1.8 kcal/mol or 10–20 folds more favored to partition at the entrance of the delivery pathway than in the membrane. Hence, k−1 is approximated to be ~2.8–5.5×108 s−1.</p><p>k2 is the rate constant of O2 migration from the entrance(s) to the reduction site and is the reciprocal of the time of O2 to diffuse into the reduction site after reaching the entrance(s) of the pathway (tcat−ent). tcat−ent is obtain by subtracting the time taken to observe the first event of O2 to the catalytic site or tcat, which is ~57 ns, from tent. Although the average tcat−ent is 54 ns, O2 delivery occurred only in 15 out of 20 simulations or 75% of the total number of simulations. To account for this, the average value of tcat−ent was scaled to 72 ns. Thus, k2 is ~1.39×108 s−1.</p><p>Under physiologically relevant O2 concentrations, it is reasonable to assume that [O2] k1 ≪ k−1 and k2, so kobs≈[O2]k1k2k−1+k2.This leads to the estimated kobs/[O2] of 1.1–1.8×109 M−1s−1, which is 8–13 fold slower than the one estimated by our previous MD study with cytochrome ba3 (15×1010 M−1s−1).1 Although this estimated kobs/[O2] is 10-fold faster than the experimentally determined second-order rate constant of 1×108 M−1s−1,32,33 the experimental rate constant was determined from the time required to convert the enzyme from the fully reduced non-O2 bound state to the ferrous-oxy (O2 bound) state. Hence, the experimental measurement includes the chemical ligation of O2 to the heme a3, the reaction that is beyond to scope of classical MD simulation. The experimental rate constant for O2 to form the heme Fe adduct in T.t. cytochrome ba3 is 1×109 M−1s−1,32,33 which is 10-fold faster than that determined for R.s. cytochrome aa3.</p><p>The computationally estimated rate constant for O2 delivery to the reduction site of cytochrome aa3 is 1.1×109 M−1s−1, which means that even when the O2 concentration is as low as 10 µM, the rate of O2 diffusion into the reduction site (~104 s−1) will be considerably faster than the rate of O2 catalysis (kcat~102 s−1). Because of the specific pathway allowing rapid diffusion of O2 from the membrane to the reduction site of the enzyme, the rate of catalysis will not be limited by the ambient concentration of O2 until that concentration in solution is in the low micromolar range.</p><!><p>Although the O2 delivery pathway in cytochrome aa3 contains an additional entrance (Branch C) when compared to cytochrome ba3, its presence probably has little or no effect on O2 migration to the reduction site. This is concluded based on previous simulations on cytochrome ba3 with in silico mutants designed to block the two O2 entrances (Branches A and B in cytochrome ba3). Those mutants, however, did not appreciably impair the passage of O2.1</p><p>Three O2 entrances of cytochrome aa3 merge at 12–15 Å distant from the reduction site. At this location, the flooding simulations identified a low O2 sampling region, which can be seen in the left panel of Fig. 6A. The presence of this barrier region is correlated with a broad range of residence times (1–90 ns) for O2 in the reduction site (Fig. 3). This region was also characterized by ILS calculations as a diffusion barrier for O2. The previous study by Oliveira et al43 also observed this kinetic barrier and, using a similar approach, calculated the energetic cost of ~9.5 kcal/mol for O2 to pass through this region, thereby excluding the X-ray inferred pathway as the primary O2 delivery pathway.43 In the present study, the height of this barrier is suggested to be much smaller since O2 molecules crossed this region in 15 of 20 150-ns simulations. These results support the role of the X-ray-inferred tunnel as the main, if not the only, pathway used for the substrate O2 to diffuse to the reduction site.</p><p>The region that coincides with the kinetic barrier is surrounded by bulky amino acids M107, W172, F282 and E286 (Fig. 6B, left). This structural feature is common in A-family oxygen reductases. Residue M107 is within TM2, W172 is in the loop connecting TM3 and TM4, and F282 and E286 are both within TM6. W172 and F282 have been suggested to restrict the access of O2.24 In cytochrome ba3, the equivalent region contains amino acids with smaller side chains (Fig. 6B, right), and the barrier height has been estimated to be much less, around 1.5 kcal/mol (Fig. 5C).1 For example, M107Rs and F282Rs are equivalent to A77 and T231 of cytochrome ba3, respectively. W172Rs is equivalent to Y133 of cytochrome ba3, and replaceing this tyrosine by a tryptophan by site-directed mutagenesis resulted in a 5-fold decrease in the O2 delivery rate.68 E286Rs is the terminal amino acid of the D channel, a proton-delivery pathway that is present in the A-family HCOs but is absent in B-family HCOs;26,69 in T.t cytochrome ba3, it is equivalent to I235 (Fig. 6B).</p><p>The X-ray inferred pathway in cytochrome aa3 is not a static pathway, and requires dynamics to change from a closed to an open form to allow for O2 to migrate across the barrier (Fig. 7 and Fig. S2). Conformational dynamics of the bulky residues of cytochrome aa3 were assessed by measuring minimum pairwise distances of F282-E286, F282-F108, W172-E286 and M107-E286 (Fig. 7A). The inuences of this dynamics on O2 migration were examined by ILS calculations over selected periods (Fig. 7B). When the side chains of F282, E286 and M107 protrude into the pathway, they approach each other as well as F108 and W172, significantly interfering with O2 migration. For example, during the 171–175 ns period of the 200-ns apo simulation in which all of the four distances came close to 4 Å, the free energy for O2 to migrate through the kinetic barrier can be greater than 4 kcal/mol (Fig. 7B). As indicated in Fig. 7A, the occurrence of these conformational changes, however, is relatively infrequent and might not be observed using shorter simulation times.</p><p>Finally, it is noted that, in the A-family HCOs, both pumped and chemical protons pass from the D channel to the periplasmic surface and the reduction site via hydrogen-bonded water chains. The terminal region of the O2 delivery pathway is also overlapped to that of the D channel. Since chemical protons and O2 use the same delivery route to the reduction site, the presence of transiently localized water molecules may interfere with the passage of O2. However, in the simulations performed in the present study as well as those reported in the recent study by Oliveira et al,43 the region connecting the terminus of the D channel and the reduction site remained dehydrated. Therefore, we can only conclude that the observed restriction of O2 migration is related to the presence of amino acids with bulky side chain. The transient presence of water in this region could also affect the rate of O2 delivery in the A-family HCOs which possess the D channel.</p><!><p>This MD study characterizes the X-ray inferred tunnel as the primary delivery pathway for O2 in R.s. cytochrome aa3. This conclusion is opposite to the one made by Oliviera et al, likely related to shorter simulation times used in the previous study43 due to slow conformational dynamics of amino acids lining the pathway. Because there are multiple constriction regions along the pathway indicated by both flooding simulations and ILS calculations, the simulation time needs to be sufficiently long to observe the migration of O2 to the reduction site of R.s. cytochrome aa3. We conclude that both R.s. cytochrome aa3 and T.t. cytochrome ba3 use similar pathways to effectively deliver O2, and this pathway also appears to be conserved in other HCOs.</p>

PubMed Author Manuscript

Emerging Role of Ferrous Iron in Bacterial Growth and Host-Pathogen Interaction: New Tools for Chemical (Micro)Biology and Antibacterial Therapy

Chemical tools capable of detecting ferrous iron with oxidation-state specificity have only recently become available. Coincident with this development in chemical biology has been increased study and appreciation for the importance of ferrous iron during infection and more generally in host-pathogen interaction. Some of the recent findings are surprising and challenge long- standing assumptions about bacterial iron homeostasis and the innate immune response to infection. Here we review these recent developments and their implications for antibacterial therapy.

emerging_role_of_ferrous_iron_in_bacterial_growth_and_host-pathogen_interaction:_new_tools_for_chemi

Background<!>Bacterial Iron Homeostasis<!>Ferrous Iron at the Interface of Host-Pathogen Interaction<!>The First Reactivity-Based Probes of Ferrous Iron

<p>Antibiotics were the formative products of the early pharmaceutical industry, and their importance to human health has only increased in recent years [1]. The antibiotic 'golden age' of the mid 20th century saw unprecedented productivity in terms of numbers of new agents brought to market, but the focus on natural product scaffolds for which resistance mechanisms were already extant in bacterial populations, combined with rampant overuse of some agents, has led to the current situation where ~70% of bacterial pathogens exhibit resistance to the most widely used antibiotics [2,3]. It is estimated that by 2050, if no corrective action is taken, these infections will cause over 10 million deaths per year, with economic costs of 100 trillion USD in lost global production [4].</p><p>The dearth of effective treatment options is especially concerning for multidrug-resistant (MDR) Gram-negative bacteria such as carbapenem-resistant Acinetobacter, Pseudomonas and Enterobacterales [5]. These MDR pathogens possess efflux transporters and outer membrane permeability barriers, which act synergistically to increase their resistance to diverse antibacterial classes [6]. Horizontal gene transfer furthermore enables these pathogens to rapidly acquire new resistance mechanisms that further limit antibiotic efficacy [7]. This situation has led to calls for novel targets and new approaches that could lead to a more robust and less vulnerable antibiotic pipeline for the new century [8]. We argue here that increasing understanding and appreciation for the role of low-molecular mass ferrous iron species ('labile iron') in bacterial growth, pathogenicity, and host interaction could provide inroads into novel antibiotic strategies. The recent development of chemical probes of labile iron should aid such efforts but have to date been employed primarily in studies of eukaryotic iron homeostasis and speciation. Here we review the emerging biology surrounding ferrous iron in bacterial growth and pathogenicity and describe the new chemical biology tools and therapeutic approaches being explored to exploit ferrous iron at the interface of infection and immunity.</p><!><p>Transition metals such as iron, manganese, copper, and zinc, are essential micronutrients for most pathogenic bacteria, with an estimated 30% of proteins in the bacterial proteome using metal cofactors to enable their cellular functions. Bacterial life evolved in a reducing, anoxic environment where reduced forms of iron were prevalent and were apparently co-opted by early life for the redox chemistry they enabled. The 'great oxidation' that coincided with the evolution of photosynthesis presented early bacterial life with an existential challenge, since ferrous iron readily converts oxygen to toxic reactive oxygen species, whilst yielding water insoluble ferric hydroxides. Continued utilization of iron thus required the evolution of iron chaperones (siderophores) that could bind and solubilize the ferric ion, as well as new biochemical pathways to detoxify reactive oxygen species. While the study of bacterial iron homeostasis has historically focused on transport and storage of the ferric ion, there has been increasing study and appreciation over the past decade that bacteria have not lost the means to utilize ferrous iron [9], enabled by biochemistry that is evolutionarily ancient [10].</p><p>The 'labile iron pool' (LIP) of a bacterial cell consists of low molecular mass ferrous (Fe2+) iron species bound by still poorly defined cellular ligands (Figure 1). Mössbauer and electron paramagnetic resonance (EPR) spectroscopy of live E. coli puts the concentration of LIP in the low-mid micromolar range [11], but LIP levels can be significantly higher than this during exponential-stage growth as detailed below [12]. Bacteria employ various iron regulatory proteins to maintain the LIP and can utilize this system to sense iron depletion as a marker of host tissue during infection [13]. These proteins modulate LIP by activating the expression of metal-acquisition systems under iron limiting conditions or by the expression of efflux transporters and storage proteins under iron replete conditions [14]. The regulation of metal homeostasis involves both transcriptional and post-transcriptional control, as recently reviewed [15]. The Fur (Ferric uptake regulator) family of iron regulatory proteins are ubiquitous in bacteria and regulate the binding of Fe2+ (or mismetallation with Mn2+ in some instances) from the LIP in order to initiate the transcription of iron-regulatory genes [16]. Genetic knockouts of the fur gene in several bacterial species have shown variable and species-specific effects including decreased virulence [17] and impairment of the growth of bacterial colonies [18]. New players in bacterial iron homeostasis continue to be identified, including a family of Fe2+ transporters (IroT/MavN) residing within a specialized vacuole in the bacteria Legionella pneumophilia [19].</p><p>Metal-sensing regulatory elements called riboswitches have recently been shown to modulate bacterial gene expression in response to different metal ions including Ni2+/Co2+ [20], Mn2+ [21–23], Mg2+ [24] and Fe2+ [25]. The riboswitch czcD (NiCO) employed by the human gut microbe Erysipelotrichaceae bacterium, was shown to preferentially bind Fe2+ (and to a lesser extent Mn2+) under physiological conditions [25]. Previous studies of czcD performed under aerobic conditions had erroneously suggested it was unresponsive to iron, highlighting the importance of considering oxygen tension when studying an oxidation-prone analyte like iron. Studies in the Gram-negative pathogen Treponema denticola revealed the presence of a cytoplasmic oxygen-binding di-iron (ODP) metalloprotein that functions as a chemoreceptor to sense both oxygen and ferrous iron for various bacterial signal transduction pathways [26].</p><p>Bacteria possess two ferritin-like iron storage proteins, the bacterial ferritin (FtnA) and bacterioferritin (Bfr) as recently reviewed [27]. In P. aeruginosa under iron limiting conditions, cytosolic BfrB binds to the bacterioferritin-associated ferredoxin (Bfd) and accepts electrons from its [2Fe-2S] cluster, causing Fe3+ stored within BfrB to be reduced and mobilized as Fe2+ [28]. The essentiality of the BfrB:Bfd interaction for P. aeruginosa has been demonstrated by genetic deletion of the bfd gene [28] and with small-molecule inhibitors [29]. However, other recent studies have questioned whether these bacterial proteins serve the same essential iron storage function as eukaryotic ferritins. Using Mössbauer and EPR spectroscopy, Lindahl and co-workers found significant elevation of LIP concentrations (as high as 500 μM!) in E. coli during exponential growth and conversely found little spectroscopic evidence for FtnA or Bfr bound iron [12]. Moreover, the major ferric signature observed during stationary growth was present even in FtnA or Bfr deletion strains, and ascribed to Fe3+ oxyhydroxide nanoparticles. Based on these findings, it was suggested that a 'respiratory shield' comprising membrane-bound enzymes of the electron transfer chain serve to maintain a microaerobic cytoplasm in which even high-μM concentrations of LIP can persist during exponential growth without causing cellular damage. If correct, these findings imply that bacterial cells should be easily distinguishable from eukaryotic cells in their iron speciation and invites strategies to target bacteria on this basis, as will be discussed later.</p><!><p>The concept of "nutritional immunity" [30] – the limitation of systemic iron and other nutrient resources from a pathogen – lies at the heart of the innate immune response as has been well reviewed [9,31,32]. During the mammalian response to bacterial infection, interleukin-6 (IL-6) acts as the main pro-inflammatory cytokine, mediating the release of the peptide hormone hepcidin from the liver. Direct binding of hepcidin to the cellular iron exporter ferroportin then leads to its internalization and degradation, promoting iron accumulation in phagocytic cells [33]. The expression of mammalian iron-binding proteins such as ferritin, transferrin and lactoferrin also plays a crucial role in restricting the iron resources available to invading pathogens [32]. The host protein lipocalin 2 is another effector of innate immunity, recognizing bacterial siderophore-bound iron (primarily enterobactin) [34]. In addition to withholding of Fe and Mn to limit bacterial proliferation, recent evidence has also suggested that copper and zinc can be weaponized to poison intracellular bacteria via phagocyte-mediated delivery to the phagosome [35–38].</p><p>To overcome iron withholding by the host, pathogenic bacteria employ diverse tactics to acquire the iron they need for growth and virulence. Best studied among these strategies is the secretion of Fe3+-chelating small molecules called siderophores (Figure 2a). Upon binding of the ferric ion, a soluble Fe3+-siderophore complex is formed that can be transported into a bacterium and unloaded in either the periplasm or cytoplasm depending on the specific siderophores and transporters involved [39,40]. Recent studies in P. aeruginosa have identified complex networks of interacting proteins and transporters that enable, for example, the reduction of ferric iron bound to pyoveridine (PVD, Figure 1). This reduction step takes place in the bacterial periplasm and liberates Fe2+ from the PVD-Fe3+ complex, where it can bind to chaperone proteins and be shepherded into the cytoplasm [40–43]. Siderophores such as PVD in P. aeruginosa, have also been shown to function as signaling molecules in the production of bacterial virulence factors [44,45]. Proteomics experiments investigating the iron starvation response in P. aeruginosa have revealed a variety of novel compensatory responses made by the pathogen in order to survive iron-limitation, which includes the upregulation of proteins required for iron-sulfur cluster biogenesis [46].</p><p>The host protein calprotectin (CP) is a metal-sequestering protein released from host neutrophils as part of the metal-withholding innate immune response (Figure 2b). While the ability of CP to bind Zn2+ and Mn2+ has been recognized for some time, work from Nolan and co-workers over the past decade has revealed a high affinity of the His6 site in CP for Fe2+ [47], and this has uncovered a hitherto unappreciated role for CP in limiting Fe2+ [48] during host-pathogen interaction, as recently reviewed [49]. The high affinity of the CP His6 site for the Fe2+ ion can even shift the iron redox equilibrium in favor of the ferrous ion under aerobic conditions, an effect that is further facilitated by pyocyanins, redox-cycling phenazines produced by P. aeruginosa [50,51]. Further emphasizing the complex and multifaceted role of CP in innate immunity is a recent report [52] that CP can inhibit the growth of Borrelia burgdorferi, the pathogen responsible for Lyme disease, by physical interaction with the bacterium, absent metal sequestration.</p><p>Under conditions where the Fe2+ ion predominates over Fe3+ (e.g., in highly acidic, reducing and/or anaerobic conditions), bacteria utilize the ferrous iron transport system FeO (Figure 1) as the primary means to import Fe2+ into the cytoplasm [53]. Like PVD siderophores, the FeO system has been identified as a major contributor to the virulence of P. aeruginosa, helping the bacterium colonize hypoxic environments [54] such as in the lungs of cystic fibrosis (CF) patients [55,56]. The production of aforementioned pyocyanin metabolites by P. aeruginosa is implicated in chronic CF infection where these redox-cycling small molecules are proposed to liberate Fe2+ from host proteins in the extracellular environment [57], leading to Fe2+ uptake via the transporter FeoB [58]. Phenazine production has also been reported to promote biofilm formation [55,57] and to increase tolerance to clinically relevant antibiotics [59].</p><!><p>Increased appreciation for the role of Fe2+ in bacterial growth and innate immunity, as described in the prior sections, has coincided with the development of new chemical probes to study iron with oxidation-stage specificity. While chelation of the electron-poor ferric iron is dominated by σ-donation from electron-rich ligands (e.g., catechol and hydroxamate siderophores), the more electron-rich ferrous ion can engage in back-donation from metal d-orbitals to heterocyclic ligands such as the imidazole ring of histidine (e.g., as in the His6 site in CP). While some measure of selectivity can be achieved in chelation-based iron probes [60], the true breakthrough in oxidation-state specificity has been realized through reactivity-based sensing, as recently reviewed [61]. First generation reactivity-based probes Rho-Nox1 [62] and IP-1 [63] gave early indications that this approach might improve oxidation-state selectivity as compared to chelation-based probes like PhenGreen SK [64]. Further development of the RhoNox-family of probes by Hirayama and co-workers (Figure 3a) has yielded new tools to study Fe2+ in the endosome-lysosome system [65], at the plasma membrane [66], in hypoxic tumor spheroids [67], in cellular models of the blood-brain barrier [68] and neural vascular barrier [69], and in mammalian cells undergoing ferroptosis [70], a form of non-apoptotic, iron(II)-dependent cell death [71]. A separate class of reactivity-based probes inspired by the antimalarial trioxolane (TRX) arterolane was described by Renslo and co-workers [72,73] and further developed by Chang [74,75] and Zhang [76] (Figure 3b). The TRX-based probes TRX-PURO [72] and FIP-1 [74] were able to detect an increase in LIP in cancer cell lines and revealed a previously unappreciated link between GPCR signaling and epigenetic regulation by mononuclear Fe2+-dependent TET enzymes [77,78].</p><p>These new tools have only just begun to be applied to the study of Fe2+ in bacterial iron-homeostasis and in bacterial infection. Thus, after first establishing that ICL-1 (Figure 3c) was highly selective for reaction with the Fe2+ ion over other biologically relevant metals and reductants, this caged form of luciferin was used to detect the mobilization of Fe2+ in luciferin-expressing mice (FVB-luc+) infected with the Gram-negative pathogen A. baumannii [75]. Whereas ICL-1 treatment of mock-infected control animals showed only weak bioluminescence near the intra-peritoneal site of ICL-1 administration, infected mice treated administered ICL-1 showed more dramatic bioluminescence in organs that were later shown by ex vivo analysis to also be major sites of infection (Figure 3d). Analysis of total iron load by ICP-MS in the infected tissues showed a substantial elevation of total iron only in liver, an observation consistent with canonical nutritional immunity and iron sequestration by that organ. The bioluminescent signal in the other infected tissues was not correlated with total iron load however, and thus indicates that Fe2+ specifically is elevated during infection. This mobilization of Fe2+ might reflect LIP (Fe2+) expansion by a pathogen undergoing exponential growth, as described by Lindhal in the spectroscopic studies described above [12]. Alternatively, these changes might reflect the pathogen's utilization of extracellular Fe2+ in the infection microenvironment [53,55], or the canonical unloading of Fe2+ from siderophore–Fe3+ within the pathogen. Further studies will be required to address the questions posed by these recent studies and new tools, which are challenging some long-held assumptions about iron utilization and sequestration during infection.</p><p>Antibiotic therapy exploiting iron has been well explored for siderophore-antibiotic conjugates ('sideromycins') designed to undergo active uptake via bacterial Fe3+-siderophore transporters [79,80]. More recent findings concerning Fe2+ in infection, as detailed herein, invite study of a new therapeutic approach comprising the reactivity-based activation of antibiotics selectively at infection sites, in response to Fe2+. To explore this notion, our laboratory recently described [81] the design and synthesis of an Fe2+-activatable form of the LpxC inhibitor PF-5081090 (i.e., TRX-PF508) and its study in P. aeruginosa infection models (Figure 3e). Whereas our previously described Fe2+ sensors [73,82] were designed to cage amine-bearing payloads, the TRX sensor employed in TRX-PF508 enables caging of hydroxamate-based metalloenzyme inhibitors, potentially mitigating the well-known [83] toxicological and pharmacokinetic liabilities of the hydroxamate function. Intriguingly, whereas TRX-PF508 showed only weak activity in MIC experiments, the compound was highly efficacious in an acute P. aeruginosa lung infection model, reducing bacterial CFU counts in a dose-dependent fashion and being well tolerated even at the highest dose of 64 mg/kg. By contrast, mice administered the parent drug PF-5081090 showed only a modest (non-significant) reduction in lung CFU counts at a dose equimolar to the lowest effective dose of TRX-PF508 (16 mg/kg). Overall, the findings of this study were consistent with a drug concentrating effect of TRX-PF508 in the lung, presumably resulting from selective activation by Fe2+ in this (infected) tissue.</p><p>Here we have summarized recent studies of bacterial acquisition and utilization of ferrous iron during infection and in the context of host-pathogen interaction. New chemical probes of ferrous iron emerging contemporaneously with these microbiological studies invite future opportunities to apply these new tools to advance a more nuanced understanding or iron acquisition and specification, both in vitro and in vivo. Early studies of this type include the imaging of Fe2+ in a murine infection model with ICL-1 and the development of antibiotics conjugates like TRX-PF508 that are activated in the infection microenvironment by reactivity-based sensing of Fe2+. These new tools and emerging biology suggest a bright future for the field of chemical (micro)biology.</p>

PubMed Author Manuscript

Active-Loop Dynamics within the Michaelis Complex of Lactate Dehydrogenase from Bacillus stearothermophilus

Laser-induced temperature-jump relaxation spectroscopy was used to study the active site mobile-loop dynamics found in the binding of the NADH nucleotide cofactor and oxamate substrate mimic to lactate dehydrogenase in Bacillus stearothermophilus thermophilic bacteria (bsLDH). The kinetic data can be best described by a model in which NADH can bind only to the open-loop apoenzyme, oxamate can bind only to the bsLDH\xc2\xb7NADH binary complex in the open-loop conformation, and oxamate binding is followed by closing of the active site loop preventing oxamate unbinding. The open and closed states of the loop are in dynamic equilibrium and interconvert on the submillisecond time scale. This interconversion strongly accelerates with an increase in temperature because of significant enthalpy barriers. Binding of NADH to bsLDH results in minor changes of the loop dynamics and does not shift the open\xe2\x80\x93closed equilibrium, but binding of the oxamate substrate mimic shifts this equilibrium to the closed state. At high excess oxamate concentrations where all active sites are nearly saturated with the substrate mimic, all active site mobile loops are mainly closed. The observed active-loop dynamics for bsLDH is very similar to that previously observed for pig heart LDH.

active-loop_dynamics_within_the_michaelis_complex_of_lactate_dehydrogenase_from_bacillus_stearotherm

<!>Samples<!>Temperature-Jump Measurements<!>Fluorescence T-Jump Measurements on bsLDH\xc2\xb7NADH Binary Complexes<!>Analysis of the Fluorescence T-Jump Data of bsLDH\xc2\xb7 NADH Binary Complexes<!>Fluorescence Experiments on bsLDH\xc2\xb7NADH\xc2\xb7Oxamate Ternary Complexes<!>CONCLUSION

<p>An enzymatic reaction involves the diffusion-controlled formation of an encounter complex between the protein and its substrate followed by the appropriate structural and dynamical arrangements, which produce a Michaelis complex capable of product formation. During the formation of the Michaelis complex, the binding pocket is substantially rearranged. Specifically, protein flaps or loops often close over the bound ligand; the binding pocket is desolvated, and catalytically important residues are brought into contact with the bound substrate. It is clear from earlier studies that enzymes (proteins) exist in an ensemble of conformations, some of which can competently bind their ligands while others bind poorly or not at all.1–3 It is also found that proteins bind their ligands in a rather complex dynamical pathway.4–9 Moreover, it has been shown directly that conformational changes occur within the ensemble of the enzyme/substrate Michaelis complex(es) on various time scale from femto- to picoseconds through milliseconds and slower.10 Here we investigate the dynamics of the active site-loop dynamics of Bacillus stearothermophilus lactate dehydrogenase (bsLDH) and compare the results with those from mammalian lactate dehydrogenase.</p><p>LDH catalyzes the NADH-dependent conversion of pyruvate to lactate. The active site structures of bsLDH and its mammalian counterparts (human or pig heart LDH) are the same. The substrate binding pocket is sequestered inside the protein ~10 Å from the surface.11,12 However, computational studies suggest that there are interesting differences in how bsLDH and mammalian LDH accomplish the chemical step of the catalytic mechanism. In human heart LDH, the hydride transfer appears to be directly correlated with the dynamics of the axis residues, which are known as the promoting vibrations. In bsLDH, the hydride transfer is predicted to show a wider range of possibilities that might be correlated with the dynamics of the axis residues, and the promoting vibration is not well-defined.13 The computed differences in the enzyme dynamics occur at fewer than tens of picoseconds around the chemical bond breaking/forming event.</p><p>In this study, the goal is to determine how slower nanosecond to millisecond conformational fluctuations are used by bsLDH to search the energy landscape for ligand binding pathway(s) to reach catalytically competent conformations. From there, the fast vibrations are predicted to be effective in promoting passage over the transition state.</p><p>To preclude the occurrence of enzyme-catalyzed chemistry that would complicate a kinetic analysis, we used a substrate analogue, oxamate, rather than the natural substrate, pyruvate. Oxamate is isoelectric and isosteric to pyruvate and has been shown to have binding kinetics very similar to that of pyruvate. Thus, oxamate is generally a very close mimic of the natural substrate. On the basis of several X-ray crystallographic data, oxamate is placed near the nicotinamide ring of the NADH and the following key active site protein residues: His195, Arg109, and Arg171 (Scheme 1). The C2=O bond of oxamate forms hydrogen bonds with His195 and Arg109, while C1OO− forms a salt bridge with Arg171.14,15 As the substrate approaches the catalytic site, a catalytically key surface loop (residues 98–110) closes over the ligand, bringing residue Arg109 into hydrogen bonding contact with ligand, water leaves the pocket, and the pocket geometry rearranges to allow for favorable interactions between the cofactor and the ligand, which facilitates on-enzyme catalysis.16,17</p><p>We probed the transient events associated with the binding of NADH to bsLDH and of oxamate to the bsLDH·NADH complex, using temperature-jump (T-jump) relaxation techniques. T-Jump relaxation monitors the re-equilibration of a chemical system following an instantaneous increase in temperature induced by a laser pulse tuned to an infrared water band. The re-equilibration results in changes in the concentration of the species involved, and the transient changes are characterized using spectroscopic probes. First, we probed the fluorescence emission of NADH in bsLDH to report on the time evolution of the changes within the NADH environment over the microsecond to millisecond time scale. Such analyses should yield data about the dynamics of binding and associated conformational binding. Second, emission of the intrinsic tryptophan in bsLDH is measured to report on conformational changes, such as loop motion. We probed changes in the NADH and the intrinsic tryptophan fluorescence for the bsLDH·NADH binary complex as well as the bsLDH·NADH· oxamate ternary complex. A comprehensive picture of the dynamics of ligand binding and Michaelis complex formation in LDH is obtained from the various structural reports.</p><!><p>All required reagents were purchased from Sigma-Aldrich Co. (St. Louis, MO) except for NADH and oxamate, which were purchased from Roche Diagnostic Corp. (Indianapolis, IN).</p><p>The preparation of bsLDH has previously been described in detail.15,18 In brief, the bsLDH gene, obtained from Genomic DNA from Geobacillus stearothermophilus ATCC 12980D, was subcloned into the pET3a vector and transformed into C43(DE3) competent Escherichia coli cells. The growth conditions of the transformed cells and the protein purification procedures followed a published procedure.14 Some studies were performed on a mutant bsLDH containing a single Trp residue, G106W, in which the three naturally occurring Trp residues have been mutated to tyrosines.18</p><!><p>A custom-built fluorescence T-jump instrument was used to measure relaxation kinetics, based on the same principles described previously.19 Sizable temperature jumps are induced by exposing a volume of water to a pulse of infrared light (1.56 μm wavelength from a D2 gas Raman shifter pumped by a YAG laser, typically 100–200 mJ/pulse, 1.5 mm diameter spot on the sample). Water absorbs the laser energy, and the temperature of the exposed volume is increased in approximately 7 ns. Typical T-jump values ranged from 5 to 10 °C in the study presented here. With a sample cell thickness of 0.5 mm, diffusion of heat out of the interaction volume takes approximately 25 ms. Hence, the apparatus generated a T-jump within 10 ns that remained nearly constant until approximately 5 ms. The sample temperature after the jump is monitored over time by probing changes in the water IR absorption at 1460 nm.</p><p>The T-jump instrument is designed to probe either tryptophan or NADH fluorescence. To probe changes in the fluorescence intensity of the tryptophan, the sample was irradiated by the 300 nm (300.3 and 302.4 nm) emission lines of an Innova 200-25/5 argon ion laser (Coherent, Palo Alto, CA), while the 360 nm (351.1 and 363.8 nm) emission lines of the same laser were used to stimulate NADH fluorescence. To prevent photodamage to the sample, the excitation light was modulated using a shutter that allowed an only 10 ms exposure for every T-jump pulse. Also, the power of the excitation beam was attenuated by a neutral-density filter; the typical intensity was 15–20 μW. The incident excitation beam is focused onto a 0.05 mm diameter spot on the sample in the center of the beam path of the 1.56 μm pulse. Fluorescence emission, detected at 50° to the excitation beam, was passed through a narrow band filter (340 ± 12 nm to probe tryptophan fluorescence and 450 ± 20 nm for NADH emission) and monitored using a R4220P photomultiplier tube (Hamamatsu, Bridgewater, NJ). A lab-written program using LabVIEW software (National Instruments, Austin, TX) was used for instrument control and data collection. Data were normalized to the average fluorescence intensity taken before the T-jump. Curve fitting was done with Igor Pro version 4.0 (Wavemetrics, Inc., Lake Oswego, OR). The uncertainties in the reported values of relaxation rates were determined from the scatter in values from four or five replicate runs and were approximately ±10% of the measured relaxation rate. Errors in the response amplitudes were analyzed in the same manner and were estimated to be ±10–15%.</p><p>Samples for T-jump measurement were dissolved in 5 mM fructose 1,6-bisphosphate (FBP), 0.1 M triethanolamine (TEA) HCl buffer at pH 6. FBP assures that the protein forms tetrameric uniformity and serves as an allosteric activator of the enzyme.20 The T-jump studies typically employed 50–80 ± 10 μN (μN indicating the active site concentration) bsLDH solutions. Different amounts of NADH and oxamate were added as appropriate to form the "binary" and "ternary" complexes.</p><!><p>The fluorescence of NADH is strongly enhanced upon binding to dehydrogenase enzymes, as its quantum yield is increased 2–3-fold,21 thus making it a good direct probe of NADH binding. It is also sensitive to the local NADH environment and can report on conformational changes around the active site. Additionally, protein tryptophan fluorescence is partly quenched by bound NADH because of Förster resonance energy transfer and can be used as an independent probe of NADH binding. Depending on the actual value of the tryptophan quantum yield, the Förster radius for this donor–acceptor pair is approximately 15–25 Å,22 and protein tryptophan fluorescence intensity significantly decreases when NADH is bound to the enzyme. Tryptophan fluorescence can also report on changes in the environment of tryptophan residues, i.e., on conformational changes that might be different from those reported by NADH. bsLDH contains three Trp residues (at positions 80, 150, and 203 roughly spread over the protein; cf. ref 18), and this study reports the emission from all three equally depending on their unknown emission efficiency. Additionally, some studies were performed on a mutant bsLDH containing a single Trp residue, G106W, where the three naturally occurring Trp residues have been mutated to tyrosines.18</p><p>NADH and tryptophan fluorescence T-jump kinetics were measured within the time span of a few nanoseconds to 15–30 ms for a number of samples with different concentrations of NADH and bsLDH. NADH fluorescence kinetic profiles (Figure 1, top graphs) show an intensity drop from 1 μs to a few milliseconds. This drop can be attributed to the release of NADH from its complex with the protein, because bound NADH has a fluorescence quantum yield much higher than that of free NADH in solution. Multiexponential curve fitting of these kinetics resolves three relaxation times in addition to the cooling component, which is ~25 ms and not shown in the figure because a time cutoff of ~10 ms has been imposed. The relaxation times represented, as evaluated from the kinetic profiles, are as follows: a strong midrange component (100–1000 μs), a weaker slow component with a 1–3 ms relaxation time, and a very weak fast component in the time range of 1–20 μs. The fast component, which is roughly independent of protein and ligand concentrations, is usually observed in similar systems. It can be partly determined by protein dynamics and may also depend on NADH internal photophysical processes involving the triplet state.</p><p>Tryptophan fluorescence kinetic profiles (bottom graphs in Figure 1) exhibit an intensity rise in the same time range where the main NADH intensity drop is observed. This correlation implies that both kinetic events are dominated by the same process of unbinding of NADH from bsLDH, disrupting energy transfer from tryptophan to bound NADH and enhancing the tryptophan fluorescence. Tryptophan kinetics can be fitted with a single-exponential component with a relaxation time very close to that of the midrange NADH component. Minor evidence of an additional slower component in the millisecond time range was observed for tryptophan kinetics at extreme concentrations and temperatures. To help elucidate the origins of these fast and slow transients, we measured the tryptophan fluorescence T-jump traces for the G106W mutant bsLDH· NADH complex (data not shown), which directly measure structural changes associated with the mobile loop.18 Not surprisingly, the trace yielded a fit to a double exponential with a set of fast and slow transients that have rate constants of ~2000 and ~10000–20000 s−1, respectively. This is significant because the structural changes as reported by the modulation of the NADH signal reflect a change in the loop tryptophan fluorescence. The slow component at ~2000 s−1 is likely the bimolecular step associated with the binding/release of NADH. Given the concentrations of free LDH and NADH under the conditions used in this study, the ~2000 s−1 transient leads to a second-order rate constant of 150 mM−1 s−1, which is quite similar to that found for the pig heart protein in previous studies.23 Moreover, the observed decrease in intensity with time is consistent with the release of NADH from the protein, given that the NADH fluorescence is 4 times higher in the binary complex than in solution. The ~10000–20000 s−1 transient represents then a unimolecular structural change within the LDH·NADH complex.</p><p>Increasing NADH and enzyme concentrations shift the main features of both NADH and tryptophan kinetics to shorter times. The left graphs of Figure 1 present the kinetic profiles for a sample with stoichiometric amounts of bsLDH and NADH at a low concentration level of 10 μM, while the right graphs show the kinetics for a sample with an excess of 207 μM NADH over 50 μM bsLDH (enzyme subunit concentrations are shown everywhere). Figure 2 displays the effect of an increasing NADH concentration on an 80 μM sample of bsLDH with regard to the NADH and Trp emission T-jump profiles.</p><p>The observed rates of tryptophan relaxation and the midrange (major) component of NADH relaxation are plotted in Figure 2 as a function of the total concentration of free NADH and free enzyme, for three after-jump temperatures. Free concentrations were calculated using NADH dissociation constants obtained from steady-state fluorescence data.18 Two distinct patterns emerge at low (<20 μM) and high (>100 μM) concentrations. At low concentrations, there is a sharp rise in the observed tryptophan relaxation rate and midrange NADH relaxation rate, and at higher concentrations, a much slower, nearly linear rise is observed. Total concentrations of bsLDH active sites in this series varied from 5 to 80 μM, and total NADH concentrations varied from 5 to 670 μM. Within the range of the initial sharp rise, NADH and bsLDH were present in approximately stoichiometric amounts. The slow rise, meanwhile, corresponds to excess NADH concentrations. The observed character of the rate concentration dependence indicates two different binding patterns for the low and high NADH concentrations. This assumption is confirmed by the analysis of kinetic data, which shows that an additional, weak binding type for NADH is required at higher concentrations. The photophysical properties (fluorescence quantum yield and energy transfer efficiency) of weakly bound NADH must be close to those of strongly bound NADH, because for both NADH and tryptophan fluorescence midrate kinetics, the concentration dependence (Figure 2) is essentially the same throughout the concentration range while the fraction of weak binding changes a lot. Indeed, NADH unbinding decreases the NADH fluorescence quantum yield and increases the tryptophan quantum yield because of the disruption of energy transfer. Therefore, the relative change in each quantum yield is approximately the same for both types of NADH binding. Free NADH in solution mainly has a folded conformation, while bound NADH takes on an extended conformation, resulting in a higher fluorescence quantum yield.21 Because nearly equal quantum yields are observed for the two NADH binding types, the weakly and strongly bound NADH must both have an extended conformation.</p><p>Figure 3 shows the observed rates of the slow NADH relaxation component at 20 and 30 °C. At higher temperatures, this component is much weaker with high fitting errors. The slow component at each temperature is nearly independent of concentration. This suggests that binding of NADH to bsLDH is followed and/or preceded by a unimolecular process that likely involves some conformational rearrangement.</p><!><p>The observed concentration-dependent kinetic rates shown in Figures 2 and 3 are first analyzed using a minimal reaction model to simulate the NADH binding process: E + NADH ↔ E1·NADH ↔ E2· NADH. Here, a conformational change occurs after the NADH binds. This problem can be solved analytically to determine two concentration-dependent observable kinetic rate functions in terms of microscopic kinetic constants.7 The higher rate function shows minimal concentration dependence at low concentrations and then increases linearly with concentration. The lower rate function shows a linear increase at low concentrations and then gradually becomes saturated when the concentration becomes higher. Furthermore, the slopes of the linear regions in these two functions are the same (cf. Figure 4 in ref 7). Because the observed kinetic rates in Figure 2 show a sharp linear increase in the low-concentration region and a slower linear increase in the higher-concentration region, neither of the two rate functions derived from the minimal model can fit the data accurately in both concentration regions. Our data analysis suggests that when the observed rate shows a linear dependency on the concentration, this rate is typically associated with the bimolecular ligand binding step.</p><p>To test other reaction models, we used the kinetic simulation procedure based on the biochemical reaction modeling program, GEPASI,24,25 as was previously implemented in our earlier studies.19,26 Briefly, the following algorithm was used. The prejump equilibrium concentrations of all the reaction components required by the tested model are first obtained from the GEPASI run, using some initial guesses for the prejump rate constants. In the next GEPASI run, using the obtained prejump equilibrium concentrations and guessed after-jump rate constants, the time dependence is obtained for each transient concentration on the approach to the new equilibrium after the T-jump. The simulated NADH fluorescence kinetic profile is calculated using previously determined values for relative quantum yields and some reasonable guess values for the less reliably known yields. The rate constants are adjusted to obtain the best fit to the experimental kinetic profile. Rates derived from the exponential fits of the experimental and simulated kinetic profiles should approximate one another. The resulting optimal rate constants are extrapolated to find the projected values for the next pair of pre- and postjump temperatures, and the process is repeated for all available temperature pairs. Then the whole sequence is repeated iteratively. In the early iterations, the quantum yields are also adjusted within their experimental uncertainties, as needed, to obtain the most consistent fit to the experimental kinetics.</p><p>The results for several reaction systems with different concentrations of NADH and bsLDH were preliminarily analyzed this way in the initial attempts to establish a consistent reaction model. The final detailed analysis was performed for two reaction systems: a nearly stoichiometric 50 μM bsLDH/48 μM NADH system and a 50 μM bsLDH/207 μM NADH system with excess NADH. First, it was found that the kinetics for low-concentration stoichiometric systems could be satisfactorily described by the simple two-step reaction model E + NADH ↔ E1·NADH ↔ E2·NADH in which a conformational change follows NADH binding, and unbinding is not possible after this change. Another two-step model in which a conformational change in the enzyme would precede NADH binding and binding would be possible to only one enzyme conformation was less representative of the data. For the systems with a large excess of NADH, kinetic reconstruction was possible only with two NADH binding steps (strong binding and weak binding), plus a conformational change after the strong NADH binding step. After this rearrangement, the weak NADH binding would not be possible. However, the reaction rate constants required for the kinetic reconstructions with this model were considerably different for low- and high-concentration systems. This discrepancy was eliminated by the addition of a conformational change in the apoenzyme that makes strong NADH binding possible. The best fit, with nearly perfect reconstruction of all the kinetic profiles (using GEPASI) for both low- and high-concentration systems, corresponds to approximately the same equilibrium constants for the two conformational changes before and after strong NADH binding. This suggests that all three conformation changes have the same nature. The X-ray crystallography data,27,28 from Protein Data Bank entries 1LDN and 2LDB, show that the NADH binding pocket in bsLDH is partly covered by the active site loop (amino acid residues 98–110) when it is in the closed conformation. Therefore, closing of this loop can block NADH from entering its binding pocket and block the substrate from binding to the active site. Apparently, the conformational changes required by the model are the opening and closing motions of the active site mobile loop, and NADH binding is possible to only the open-loop conformation.</p><p>The minimum model for NADH binding reaction (Scheme 2) includes four steps: (1) opening the active site loop in apo-bsLDH, (2) strong NADH binding with formation of a binary complex between open-loop bsLDH and NADH (bsLDHOPEN· NADH), (3) closing of the active site loop in the binary complex, and (4) weak binding of an additional NADH molecule to the binary complex in the open-loop conformation.</p><p>The microscopic rate constants resulting from the kinetic reconstructions, along with other derived parameters, are listed in Table 1. The errors for the absolute values of rate constants were estimated as their variances in the last iterations (30–50%); however, the forward and back rates vary mainly in parallel, and their relative changes for different temperatures and forward/back ratios have smaller errors. This evaluation of errors is rough, but a more rigorous estimate is not currently possible.</p><p>The fraction of open-loop state that determines binding competence for the apoenzyme [Capo = k1/(k1 + k−1)] and binding or unbinding competence for the binary complex [Cbin = k−3/(k3 + k−3)] significantly drops with an increase in temperature (Table 1). While its value of 0.5–0.6 at 20 °C corresponds to nearly equal populations of competent and noncompetent conformations, the value is 2 times lower at 40°C, signifying a strong domination of noncompetent states. The pattern of temperature dependence of NADH binding competence, with an equilibrium shift to the closed-loop noncompetent state with an increase in temperature, is similar to what was previously observed for pig heart LDH (phLDH).19,26</p><p>In an experiment in which the concentration of bsLDH was held constant at 80 μM (Figure 4 and Table 2), exponential fits of the experimental profiles possibly reveal that as we increase the NADH concentration in the binary complex, we observe a different kinetic event. A close inspection of the rate kinetics indicates that at low NADH concentrations of 30–300 μM, the transient (the observed relaxation rate, k1_binary_trp) derived from the tryptophan T-jump trace approximates the fast transient (k1_binary_NADH) derived from the fit of the NADH T-jump trace. However, at a NADH concentration of 400 μM, that no longer is the case; the transient (k1_binary_trp) from the tryptophan trace matches that of the slower transient (k2_binary_NADH) derived from the NADH trace. It is even worse at 500 μM NADH, as k1_binary_trp no longer matches any of the transients derived from the NADH T-jump kinetics. The change in the kinetic profile of the tryptophan T-jump seems to be gradual as it remains fairly constant at 30–50 μM NADH and starts to increase to 1 order of magnitude larger at ≥300 μM. It is fairly interesting that the two limiting values approximate the transients derived from the biexponential fit of the tryptophan T-jump trace for the single-loop-containing Trp mutant, G106W bsLDH. Perhaps at low NADH concentrations, the tryptophan T-jump trace prominently detects the NADH unbinding event. Then at higher NADH concentrations where unbinding is expected to be negligible as the system is fairly saturated with NADH, the conformational change, which is loop motion, is evident.</p><!><p>Binding of a substrate to dehydrogenase enzymes requires preliminary binding of the NADH cofactor.29–31 Without a cofactor, the binding of substrate or its mimic, oxamate, to apo-LDH is very weak and does not noticeably occur under our conditions. We do not consider this further. We also do not consider the level of unbinding of NADH from a ternary bsLDH·NADH·oxamate complex that is known to be negligibly low under conditions close to ours.32</p><p>Bound oxamate strongly quenches the fluorescence of bound NADH. Therefore, binding of oxamate to the bsLDH·NADH binary complex, much like NADH binding, can be monitored directly by measuring changes in NADH fluorescence intensity. NADH fluorescence T-jump kinetics for a ternary sample with stoichiometric amounts of bsLDH, NADH, and oxamate (50 μM of each) are shown in the top graph of Figure 5, and green curves on the same graph show kinetics for the corresponding binary sample without oxamate. In these kinetic events, the dominating features are (a) unbinding of NADH from bsLDH resulting in the intensity drop for both systems in approximately the same submillisecond time range and (b) oxamate unbinding that produces an intensity rise around 1 ms.</p><p>At high oxamate concentrations (Figure 5, bottom panel), the NADH emission kinetics at 20 °C exhibit a minor drop of fluorescence intensity in submilliseconds. At higher after-jump temperatures, however, there is a large increase in intensity throughout the whole microsecond to millisecond time range. In the multiexponential fits of NADH emission kinetics for ternary samples, in addition to the known cooling component (~25 ms, which is only partially shown in the figure), three exponential components can be resolved: fast (20000–30000 s−1), mid (1500–5000 s−1), and slow (250–2000 s−1). The dependence of the free species concentration on the observed relaxation rates (Figure 6) displays complicated patterns that imply a multistep reaction mechanism.</p><p>The traces in Figure 5 represent measurements involving a "ternary complex" of bsLDH, NADH, and oxamate taken at low oxamate concentrations and final temperatures. Exponential profiles were used to fit the kinetic data to yield three time constants within the observational time range of 1 μs to 5 ms. In a less complete earlier study exploring binding of oxamate to phLDH,5,7,8 using a combination of stopped flow-based and laser T-jump spectroscopies, results suggested a reaction that starts with a bimolecular step followed by unimolecular transformations within the phLDH·NADH·oxamate complex. After the association of the LDH·NADH complex with oxamate and formation of a bimolecular encounter complex, the observed kinetics show key hydrogen bond formation between the ligand and His195 (~500 μs), followed by the closure of the surface loop (~2 ms). Using these results as a guide in interpreting the data presented here, the slowest observed transient (k1 in Table 3) can be attributed to a unimolecular transformation involving loop motion while the fastest (k3), which scales linearly with the oxamate concentration (see below), is primarily the bimolecular on rate between LDH/NADH and oxamate for forming an encounter complex (see Scheme 3). A third weakly observed transient, which is more distinct in complexes with a high oxamate concentration and a low final temperature (not observed in the phLDH system7), with a relaxation rate, k2, of ~1500 s−1 and an amplitude of opposite sign was first thought to involve the formation of the binary complex (see Figure 1). The simplest kinetic model that could account for the T-jump NADH emission studies for oxamate complexes is shown in Scheme 3.</p><p>Linearization of the kinetic equations describing the model in Scheme 3 yields the following provided the rearrangement after oxamate binding is by far the slowest step:33,34</p><p> (1)k1=kr+kf1+KEC[LDH/NADH]+[oxamate]Kbin+1 (2)k2=kon_binary+koff_binary1+[oxamate][LDH/NADH]+KECKloopKloop+1 (3)k3=koff_ternary+kon_ternary([LDH/NADH]+[oxamate]) where Kbin = koff_binary/kon_binary, KEC = koff_ternary/kon_ternary, and Kloop = kr/kf. They represent the (linearized) analytical solutions for each of the observed transients.</p><p>Kinetic simulations using GEPASI and initial guesses based on values observed for phLDH7 helped associate k2 with concomitant concentration changes of the binary complex. In phLDH, k2 was not observed because the rate at which the LDH/NADH concentration changes occurs in the millisecond time regime, which is much slower than the rate of ternary complex formation. Another distinction between phLDH and bsLDH systems is that the change in the LDH·NADH·oxamate complex concentration is 1 order of magnitude slower in bsLDH (Table 2) than in phLDH.7 Furthermore, while the rates for phLDH decrease with an increase in temperature, the opposite was observed for the bsLDH system. This behavior may be consistent with the thermophile bsLDH having an optimal temperature higher than that of the pig heart protein.35</p><p>While the structural similarities for phLDH and bsLDH in the active site are apparent, the structural differences remote from the active sites are also obvious. For example, two cofactor fructose-bisphosphate molecules are required to bind two bsLDH dimers to form a tetramer, and this feature is believed to be a critical component for the rigidity of bsLDH to resist heat-induced denaturation. Our previous and current T-jump studies on phLDH and bsLDH binary and ternary complexes revealed that the more rigid bsLDH structure results in several interesting differences in the ligand binding dynamics in these two isozymes as discussed below.</p><p>For the binding of oxamate to LDH·NADH binary complex, the bimolecular step (step 1 in the model) can be observed in the time range of tens of microseconds in both isozymes. However, in phLDH, the oxamate molecule can reach its final position in this step, as shown by the observation that the NADH fluorescence is already quenched in this time range.7 In bsLDH, the NADH fluorescence quench by oxamate is no more than approximately one-third in this step, and additional enzyme conformational changes on the slower time scale are required to bind oxamate to its final position to complete the quench (Figure 5).</p><p>For the binding of NADH to phLDH, the observed concentration-dependent rate for the bimolecular step (the concentration-dependent rate) is in time range of hundreds of microseconds and shows a well-defined linear relationship, and a single binding site is sufficient to describe the binding process.23 Interestingly, in both phLDH and bsLDH, concentration-independent rates faster than the rate associated with the binding are observed. However, concentration-independent rates slower than the rate associated with binding are observed only in bsLDH (Figure 1).23 One possible explanation is that in phLDH, the faster motions, including unfolding of NADH and active site local residue rearrangements to accommodate the extended conformation of NADH, are in sync with the bimolecular step and no additional slower environmental changes to the nicotinamide are required to anchor NADH to its final position. This is possible because of a special feature of the T-jump experiment: the faster kinetic processes can be observed regardless of whether they occur before, after, or at the same time as the slower processes. On the other hand, the slower additional protein conformational changes are required in bsLDH, as in the case of binding of oxamate to the LDH·NADH complex.</p><p>The analysis described above is adequate for samples with low oxamate concentrations where the contribution from the binary complex equilibrium may be significant but not for samples with high oxamate concentrations where the ternary bsLDH·NADH·oxamate species dominates the reaction mixture. At high oxamate concentrations, a different kinetic scheme could be in place. Indeed, in a recently published paper, we proposed the plausible scheme depicted in Scheme 4.18</p><p>In this kinetic scheme, which is in agreement with various results,7,8,15 the encounter complex, presumed to have an "open"-loop conformation, leads to two other conformations of the Michaelis complex, labeled inactive and active, which are probably of "closed"-loop conformations with varying loop structure and an atomic arrangement in the active site. The two closed conformations do not appear to directly interconvert but do so via the encounter complex. The observed increase in the magnitude of emission with an increasing oxamate concentration and a lower final T-jump temperature (Figure 7) supports this contention. Again, the higher the oxamate concentration, the smaller the expected contribution from the binary complex equilibrium. Because the "bump" in the ~100 μs region of the kinetic profile becomes prominent at a higher oxamate concentration (green traces in Figure 7), this kinetic event cannot be due to the binding of NADH to LDH. In Figure 7, there is no hint of the "bump" at a lower oxamate concentration (blue traces); more evident is a decrease in the signal in the ~100 μs region reminiscent of a similar decrease in intensity observed in the kinetic traces for the "binary complex" (Figures 1 and 4). Although the changes in the concentration of the binary complex are very small, its effect on the final T-jump trace is magnified as the NADH emission quantum yield of this species is 12 times that of the ternary complex. The "bump" in the T-jump traces (Figure 7) occurs only at high oxamate concentrations and low temperatures when the rate of ternary complex formation is faster than the rate of formation of the binary complex.</p><!><p>T-Jump studies on phLDH (previous work) and bsLDH (this work) binary and ternary complexes revealed that the more rigid bsLDH structure results in several interesting differences in the ligand binding dynamics in these two isozymes. The kinetic data can be best described by a model in which NADH can bind only to the open-loop apoenzyme, oxamate binding is possible only to the bsLDH·NADH binary complex in the open-loop conformation, and oxamate binding is followed by closing of the active site loop preventing oxamate unbinding. The loop open and closed states are in dynamic equilibrium and interconvert on the submillisecond time scale and faster, strongly accelerating with an increasing temperature due to significant enthalpy barriers. Binding of NADH to bsLDH results in minor changes of the loop dynamics and does not shift the open–closed equilibrium, but oxamate substrate mimic binding shifts this equilibrium to the closed state. At high excess oxamate concentrations when all active sites are nearly saturated with the substrate mimic, all active site mobile loops are mainly closed.</p><p>The closing of an active site mobile loop that follows substrate binding and brings the atomic groups required for the catalysis into proper contacts is a ubiquitous pattern of key importance in modern enzymology and is addressed by different experimental and molecular modeling approaches.3,9,26,36–41 The loop dynamics for bsLDH initiated by substrate binding was found to be in the submillisecond time range38,42,43 and is assumed to be rate-limiting. Interestingly, this work suggests that the key surface loop of LDH does not undergo a rigid body motion upon closure. Very typically, these studies show at least two motions for the loop, one on the ~300 s−1 scale and another 5 times or so faster (see the G103W data in Table 2). Also, motions of the surface loop appear to be coupled to motions of internal loops.44 Our laser-induced T-jump studies of binding of the substrate or substrate mimic to phLDH and the number of other proteins3,9,26,39–41,45 showed similar temporal characteristics of active site-loop dynamics.</p><p>However, quite little is known about the loop motion in free enzymes and their complexes with cofactors. In this study, we applied the laser-induced T-jump method to investigate the active site-loop dynamics in all binding states of bsLDH on the way to the Michaelis complex. The reaction model that describes the reaction and the microscopic kinetic parameters for individual reaction steps were determined from the experimental kinetics using computer simulation. We found that the time scale of active site-loop (residues 98–110) motion is quite similar for both the apoprotein and the binary complex (Table 1) and in turn is quite similar to that found in the ternary complex; all proceed with an ~300 s−1 motion. We found that the conformation of the active site loop (residues 98–110) affects both NADH cofactor and oxamate substrate mimic binding and that binding is possible only when the loop is open.</p><p>We also found that binding of the second NADH molecule to the bsLDH·NADH binary complex occurs at high excess NADH concentrations and is also only possible for the open-loop complex. This dimerization of NADH at the bsLDH binding pocket can affect the loop dynamics and needs to be taken into account in LDH studies if high NADH concentrations (>0.1 mM) are to be used to ensure enzyme saturation with the cofactor.</p>

PubMed Author Manuscript

Odorant receptors from Culex quinquefasciatus and Aedes aegypti sensitive to floral compounds

Mosquitoes rely heavily on the olfactory system to find a host for a bloodmeal, plants for a source of energy and suitable sites for oviposition. Here, we examined a cluster of eight odorant receptors (ORs), which includes one OR, CquiOR1, previously identified to be sensitive to plant-derived compounds. We cloned 5 ORs from Culex quinquefasciatus and two ORs from Aedes aegypti, ie, CquiOR2, CquiOR4, CquiOR5, CquiOR84, CquiOR85, AaegOR14, and AaegOR15 and then deorphanized these receptors using the Xenopus oocyte recording system and a large panel of odorants. 2-Phenylethanol, phenethyl formate, and phenethyl propionate were the best ligands for CquiOR4 somewhat resembling the profile of AaegOR15, which gave the strongest responses to phenethyl propionate, phenethyl formate, and acetophenone. In contrast, the best ligands for CquiOR5 were linalool, PMD, and linalool oxide. CquiOR4 was predominantly expressed in antennae of nonblood fed female mosquitoes, with transcript levels significantly reduced after a blood meal. 2-Phenylethanol showed repellency activity comparable to that of DEET at 1%. RNAi experiments suggest that at least in part 2-phenylethanol-elicited repellency is mediated by CquiOR4 activation.

odorant_receptors_from_culex_quinquefasciatus_and_aedes_aegypti_sensitive_to_floral_compounds

Introduction<!>Insect preparations and sample collection<!>OR cloning<!>Electrophysiology<!>dsRNA synthesis<!>dsRNA microinjection<!>Quantitative analysis of transcription levels<!>Surface-landing and feeding assay<!>Phylogenetic analysis of mosquito ORs<!>Graphic preparations and statistical analysis<!>Deorphanization of clustered Culex and Aedes ORs<!>Tissue expression analysis<!>Repellency assays<!>Attempts to link odorant reception with repellency activity

<p>Mosquitoes rely on the olfactory system to find plants as a source of carbohydrates, hosts for bloodmeals, and oviposition sites. Because pathogens might be transmitted during a bite by an infected mosquito, there is understandably a great deal of interest in unraveling the olfactory aspects of human-mosquito interactions (Bradshaw et al., 2018) to explore ways of reducing mosquito bites. However, plant nectar sources are often essential for mosquitoes because they increase mosquito life span and reproductive capacity (Lahondère et al., 2019) and long-living mosquitoes are more dangerous (Tan et al., 2019). Therefore, understanding how mosquitoes find plants/flowers is also important for reducing the transmission of vector-borne diseases.</p><p>Previously, we have identified generic and plant kairomone sensitive odorant receptors (ORs) from the Southern house mosquito, Culex quinquefasciatus (Xu et al., 2013). One of these ORs, CquiOR1, belongs to a cluster of 6 ORs from the Southern house mosquito and 2 ORs from the yellow fever mosquito, Aedes ( = Stegomyia) aegypti (Fig. S1). Specifically, CquiOR2, CquiOR4, AaegOR14, AaegOR15, CquiOR5, CquiOR84, and CquiOR85. Of note, CquiOR2 (VectorBase, CPIJ0000542) is not the previously reported oviposition attractant-detecting OR2 (Pelletier et al., 2010), which has been renamed CquiOR121 (CPIJ014392) (Leal et al., 2013). We cloned CquiOR2 (Genbank, MG280964) and the other ORs in this cluster (CquiOR4, MG280965, AaegOR14, MN227017, AaegOR15, MN227018, CquiOR5, MG280966, CquiOR84, MN227015, and CquiOR85, MN227016). We then deorphanized these receptors using the Xenopus oocyte recording system and a panel of odorants with physiologically and behaviorally relevant compounds, including oviposition attractants, mosquito repellents, and plant-derived compounds (Hughes et al., 2010; Leal et al., 2013, 2017a; Pelletier et al., 2010; Xu et al., 2013, 2014, 2019; Zhu et al., 2013). Here, we report that these receptors, particularly CquiOR4, CquiOR5, and AaegOR15, are very sensitive to plant-derived compounds, including repellents. CquiOR4, for example, which is very specific to female antennae, with high and low transcript levels in nonblood fed and blood-fed mosquitoes, respectively, showed a robust response to the natural repellent 2-phenylethanol. Repellency activity elicited by 2-phenylethanol reduced significantly in CquiOR4-dsRNA-treated mosquitoes, but it was unchanged when these mosquitoes were tested against DEET, which is detected with another receptor (Xu et al., 2014).</p><!><p>Mosquitoes used in this study were from a laboratory colony of Cx. quinquefasciatus originating from adult mosquitoes collected in Merced, CA in the 1950s and kept at the Kearney Agricultural Research Center, University of California, Parlier, CA. Specifically, we used mosquitoes from the Davis colony, which was initiated about eight years ago with mosquitoes from the Kearney colony. In Davis, mosquitoes were maintained at 27 ± 1 °C, 75 ± 5% relative humidity, and under a photoperiod of 12:12 h.</p><!><p>Total RNA samples were extracted from one thousand 4-7 day-old Culex female antennae with TRIzol reagent (Invitrogen, Carlsbad, CA). Antennal cDNA was synthesized from 1 μg of antennal total RNA from each species using iScript cDNA synthesis kit (Bio-Rad) according to the manufacturer's instructions (Clontech, Mountain View, CA). Total RNA was extracted from Aedes mosquitoes provided by Dr. Anthon J. Cornel. This colony was established in 2016 from eggs that were laid by females collected in BG-sentinel traps (Biogents, Regensburg, Germany) in the City of Clovis, California (Xu et al., 2019). Aedes mosquitoes were kept at 25 ± 2 °C, 75 ± 5% relative humidity, and under a photoperiod of 12:12h.</p><p>To obtain full-length coding sequences of CquiOR2, CquiOR4, CquiOR5, CquiOR84, CquiOR85, AaegOR14, and AaegOR15, PCRs were performed using the following gene-specific primers containing restriction endonuclease sites (XmaI and XbaI) and Kozak motif (acc):</p><p>InFu-CqOR2-F: AGATCAATTCCCCGGGaccATGAGGTTCGCCCCGCTC</p><p>InFu-CqOR2-R: TCAAGCTTGCTCTAGATCAAATGCTATCCTTTAAAATCACA</p><p>InFu-CqOR4-F: AGATCAATTCCCCGGGaccATGAAATCCCACAGTCCCCTCAA</p><p>InFu-CqOR4-R: TCAAGCTTGCTCTAGA TTACAACCTCTCCTTCAGCACGACA</p><p>InFu-CqOR5-F: AGATCAATTCCCCGGGacc ATGAAATTCTACGAGCTCCGCG</p><p>InFu-CqOR5-R: TCAAGCTTGCTCTAGA TTATGAATGCATCAATCGCTCCCT</p><p>InFu-CqOR84-F: GATCAATTCCCCGGGaccATGGAGTTCCTGGCCGC</p><p>InFu-CqOR84-R:CAAGCTTGCTCTAGATTAGTTCACTCCCTGCAGTCG</p><p>InFu-CqOR85-F: AGATCAATTCCCCGGGaccATGGAGTTCCTGGCCGCC</p><p>InFu-CqOR85-R: TCAAGCTTGCTCTAGATTAGATCACTCCCTGCAGTCGTTC</p><p>InFu-AaOR14-F: GATCAATTCCCCGGGaccATGAACTACTTTGAGCTA</p><p>InFu-AaOR14-R: CAAGCTTGCTCTAGATTACAAGTAATCCTTAAGCACC</p><p>InFu-AaOR15-F: GATCAATTCCCCGGGaccATGAAGTACTTTGAGCT</p><p>InFu-AaOR15-R: CAAGCTTGCTCTAGATTACAACTGATCCTTTAGTACAACGT</p><p>PCR products were purified by a QIAquick gel extraction kit (Qiagen) and then subcloned into pGEMHE vector with In-Fusion HD cloning kit (Clontech, Mountain View, CA). After transformation, plasmids were extracted using the QIAprep Spin Miniprep kit (Qiagen) and sequenced by ABI 3730 automated DNA sequencer at Davis Sequencing (Davis, CA) for confirmation.</p><!><p>Two-electrode voltage-clamp technique (TEVC) was performed as previously described (Xu et al., 2014). Briefly, the capped cRNAs were synthesized using pGEMHE vectors and mMESSAGE mMACHINE T7 Kit (Ambion). Purified OR cRNAs were resuspended in nuclease-free water at 200 ng/mL and microinjected with the same amount of CquiOrco (or AaegOrco) cRNA into Xenopus laevis oocytes in stage V or VI (purchased from EcoCyte Bioscience, Austin, TX). Then, the oocytes were kept at 18 °C for 3–7 days in modified Barth's solution [NaCl 88 mM, KCl 1 mM, NaHCO3 2.4 mM, MgSO4 0.82 mM, Ca(NO3)2 0.33 mM, CaCl2 0.41 mM, HEPES 10 mM, pH 7.4] supplemented with 10 mg/mL of gentamycin, 10 mg/mL of streptomycin. Odorant-induced currents at holding potential of −80 mV were collected and amplified with an OC-725C amplifier (Warner Instruments, Hamden, CT), low-pass filtered at 50 Hz and digitized at 1 kHz. Data acquisition and analysis were carried out with Digidata 1440A and pCLAMP 10 software (Molecular Devices, LLC, Sunnyvale, CA). The panel of odorants included the following compounds: 1-butanol (acquired from Sigma-Aldrich/Fluka, St. Louis, MO; hereafter Fluka, 99% purity), 1-pentanol (Sigma-Aldrich, St. Louis, MO; hereafter Sigma, 98%), 1-hexanol (Sigma, 99%), 1-heptanol (Fluka, 99%), 1-octanol (Fluka, 99%), 1-nonanol (Fluka, 98%), 1-do-decanol (Fluka, 98%), 2,3-butanediol (Sigma-Aldrich, St. Louis, MO; Aldrich, 99%), 2-butoxyethanol (Aldrich, 99%), 3-methyl-1-butanol (Sigma, 99%), 2-hexen-1-ol, (E)-3-hexen-1-ol (Aldrich, 96%), 1-hexen-3-ol (Aldrich, 98%), 1-hepten-3-ol (Aldrich, 97%), 3-octanol (Aldrich, 99%), 1-octen-3-ol (Fluka, 98%), 2-butanol (Aldrich, 99%), 2-nonen-1-ol (Aldrich, 97%), 2-pentanol (Aldrich, 98%), 4-methylcyclohexanol (Fluka, 98%), 1-hexadecanol (Aldrich, 99%), 3-pentanol (Aldrich, 98%), 3-methyl-2-butanol (Aldrich, 98%), 3-methyl-2-buten-1-ol (Aldrich, 99%), 2-methyl-3-buten-2-ol (Aldrich, 98%), propanal (Aldrich, 97%), butanal (Fluka, 99%), pentanal (Fluka, 97%), isovaleraldehyde (Sigma, 85%), hexanal (Aldrich, 98%), (E)-2-methyl-2-butenal (Sigma, 98%), heptanal (Sigma, 95%), octanal (Aldrich, 99%), nonanal (Aldrich, 95%), decanal (Sigma, 99%), undecanal (Aldrich, 97%), 1-do-decanal (Acros Organics, Morris Plains, NJ; hereafter ACROS, 98%), (E)-2-hexenal (ACROS, 95%), (Z)-8-undecenal (Aldrich, 96%), (E)-2-heptenal (Aldrich, 97%), (E)-2-nonenal (Aldrich, 97%), phenylace-taldehyde (Aldrich, 90%), 2,4-hexadienal (Aldrich, 95%), furfural (Fluka, 99%), benzaldehyde (Fluka, 99.5%), α-hexylcinnamaldehyde (Millipore-Sigma, Burlington, MA; hereafter Millipore, 95%), methyl acetate (Sigma, 99.5%), ethyl acetate (Sigma, 99.8%), propyl acetate (Sigma, 99.5%), butyl acetate (Fluka, 99%), pentyl acetate (Aldrich, 99%), hexyl acetate (Fluka, 99.7%), heptyl acetate (Sigma, 98%), octyl acetate (Aldrich, 99%), nonyl acetate (Millipore, 97%), decyl acetate (Millipore, 95%), methyl propionate (Fluka, 99.8%), ethyl propionate (Aldrich, 99%), methyl butyrate (Fluka, 99.8%), ethyl butanoate (Aldrich, 99%), methyl hexanoate (Aldrich, 99%), ethyl hexanoate (Aldrich, 99%), ethyl 3-hydroxyhexanoate (Aldrich, 98%), ethyl 3-hydro-xybutanoate (Fluka, 97%), ethyl linoleate (Millipore, 99%), phenyl propanoate (Tokyo Chemical Industry, Tokyo, Japan; hereafter TCI, 95%), phenethyl propionate (Sigma, 98%), ethyl 2-(E)-4-(Z)-dec-adienoate (Fluka, 97%), (E)-2-hexenyl acetate (Aldrich, 98%), (Z)-3-hexenyl acetate (Aldrich, 98%), (E)-2-hexenyl acetate (Aldrich, 98%), ethyl lactate (Aldrich, 98%), phenyl isobutyrate (TCI, 95%), eugenyl acetate (Sigma, 98%), methyl salicylate (Sigma, 99%), ethyl stearate (Sigma, 99%), methyl myristate (Millipore, 99%), isopropyl myristate (Aldrich, 98%, palmitic acid methyl ester (Sigma, 99%), 1-octen-3-yl acetate (Aldrich, 97%), isopentyl acetate (Fluka, 99.7%), ethyl pheny-lacetate (Fluka, 99%), geranyl acetate (Fluka, 98%), octadecyl acetate (Millipore, 99%, discontinued), acetylacetone (Millipore, 99%), 2-butanone (Sigma, 99%), 2-heptanone (Sigma, 99%), geranylacetone (Fluka, 98%), 6-methyl-5-hepten-2-one (Aldrich, 99%), 5-methyl-2-hexanone (Fluka, 98%), 2,3-butanedione (Fluka, 99%), 3-hydroxy-2-butanone (Aldrich, 96%), 2-pentanone (Millipore, 99.5%), 2-hexanone (Sigma, 98%), 2-octanone (Millipore, 98%), 2-undecanone (Aldrich, 99%), 2-tridecanone (Fluka, 99%), 2-nonanone (Fluka, 99.5%), 1-octen-3-one (Aldrich, 96%), cyclohexanone (Fluka, 99.9%), acetophenone (Fluka, 99.9%), γ-valerolactone (Aldrich, 99%), γ-hexalactone (Millipore, 98%), γ-octalactone (Millipore, 97%), γ-decalactone (Millipore, 98%), γ-dodecalactone (Millipore, 98%), p-coumaric acid (Sigma, 98%), isovaleric acid (Aldrich, 99%), dodecanoic acid (Millipore, 98%), ( ± )-lactic acid (Sigma, 90%), ethanoic acid (Sigma, 99%), propanoic acid (Fluka, 99%), butanoic acid (Aldrich, 99%), isobutyric acid (Millipore, 99%), 2-oxobutyric acid (Fluka, 97%), pentanoic acid (Millipore, 99%), 2-oxovaleric acid (Fluka, 98%), myristic acid (Aldrich, 95%), palmitoleic acid (Fluka, 99%), oleic acid (Aldrich, 99%), hexanoic acid (Aldrich, 99.5%), (E)-2-hexenoic acid (Aldrich, 99%), 5-hexanoic acid (Aldrich, 98%), (E)-3-hexenoic acid (Millipore, 97%), heptanoic acid (Fluka, 99%), octanoic acid (Aldrich, 99.5%), nonanoic acid (Fluka, 97%), decanoic acid (Fluka, 98%), n-tridecanoic acid (Millipore, 98%), linoleic acid (Sigma, 99%), ammonia (Sigma, 25% ammonia in water), trimethylamine (Fluka, 45% in water), propylamine (Fluka, 99.5%), butylamine (Fluka, 99.5%), pentylamine (Fluka, 99.5%), hexylamine (99.5%), heptylamine (Fluka, 99.5%), octylamine (Fluka, 99%), 1,4-diaminobutane (Fluka, 99%), cadaverine (Fluka, 97%), 1,5-diaminopentane (Aldrich, 97%), phenol (Fluka, 99.5%), 2-methylphenol (Aldrich, 99%), 3-methylphenol (Aldrich, 99%), 4-methylphenol (Aldrich, 99%), 4-ethylphenol (Aldrich, 99%), 3,5-dimethylphenol (Aldrich, 99%), 2,3-dimethylphenol (Aldrich, 99%), 2,4-dimethylphenol (Fluka, 99.6%), 2,5-dimethylphenol (Millipore, 99%), 2,6-dimethylphenol (Fluka, 99.8%), 3,4-dimethylphenol (Fluka, 99.6%), guaiacol (Sigma, 99%), 2-methoxy-4-propylphenol (Aldrich, 99%), 2-phenoxyethanol (Fluka, 99.9%), 1,2-dimethoxybenzene (Aldrich, 99%), benzyl alcohol (Fluka, 99%), 2-phenylethanol (Fluka, 99%), 1-phenylethanol (Fluka, 99%), phenylether (Fluka, 99%), (S)-(−)-perillaldehyde (Aldrich, 92%), fenchone (Aldrich, 98%), thujone (Fluka, 99%), camphor (Aldrich, 98%), α-terpinene (Fluka, 99%), γ-terpinene (Fuji Flavor Co. Japan, 95%), (—)-menthone (Aldrich, 90%), menthyl acetate (Aldrich, 97%), limonene (TCI, 95%), linalyl acetate (Sigma, 97%), α-humulene (Aldrich, 98%), linalool oxide (Fluka, 97%), geraniol (Fluka, 99%), nerol (Fluka, 97%), thymol (Sigma, 99.5%), ( ± )-linalool (Bedoukian Research Inc., Danbury, CT, 95%), eucalyptol (Sigma, 99.5%), citral (Fluka, 95%), eugenol (Sigma, 99%), α-pinene (Aldrich, 98%), ocimene (Sigma, 90%), ( ± )-citronellal (Sigma, 85%), α-phellandrene (Sigma, 75%, stabilized), nerolidol (Aldrich, 98%), jasmone (Fluka, 99%), menthol (Sigma, 99%), carvone (Aldrich, 98%), cymene (Aldrich, 99%), terpinolene (Fluka, 97%), β-myrcene (Sigma, 95%), ( + )-δ-cadinene (Fluka, 97%), ( + )-limonene oxide (Aldrich, 97%), (E,E)-farnesol (Aldrich, 96%), (E,E)-farnesyl acetate (Aldrich, 95%), farnesene (Sigma, stabilized mixture of isomers, purity unspecified), α-methylcinnamaldehyde (Sigma, 97%), cinnamyl alcohol (Aldrich, 98%), α-terpineol (Millipore, 96%), citronellol (Millipore, 95%), (E)-cinnamaldehyde (Aldrich, 99%), (−)-caryophyllene oxide (Sigma, 95%), β-caryophyphyllene (Sigma, 80%), carvacrol (Sigma, 98%), terpinen-4-ol (Fluka, 95%), 7-hydroxycitronellal (Sigma, 95%), pyridine (Sigma, 99.8%), pyrrolidine (Millipore, 99%), 2-pyrrolidinone (Aldrich, 99%), indole (Aldrich, 98%), 3-methylindole (Aldrich, 98%), isoprene (Aldrich, 99%), 5-isobutyl-2,3-dimethylpyrazine (Aldrich, 97%), dibutyl phthalate (Aldrich, 99%), dimethyl phthalate (Fluka, 99%), phenethyl formate (Millipore, 96%), benzyl formate (Millipore, 95%), 2-acetylthiophene (Aldrich, 98%), methyl disulfide (Millipore, 98%), 2-ethyltoluol (Aldrich, 99%), 2-methyl-2-thiazoline (Aldrich, 99%), methyl anthranilate (Millipore, 99%), 4,5-dimethylthiazole (Aldrich, 97%), p-mentane-3,8-diol (PMD) (Bedoukian Research, Inc. 95%), N-(2-isopropyl-phenyl)-3-methylbenzamide, and N,N-diethyl-3-methylbenzamide (DEET) (Sigma, 97%).</p><!><p>Double-strand RNA (dsRNA) of CquiOR4, CquiOR5, and β-galactosidase were synthesized by in vitro transcription from PCR product using the MEGAscript T7 transcription kit (Ambion, Austin, TX). PCR was performed using plasmids containing the target genes as DNA template with the following gene-specific primers that included T7 promoter sequence (underlined);</p><p>Cquiβ-gal-F: 5′-TAATACGACTCACTATAGGGAATGGTTCAGGTCGAAAACG-3′ and Cquiβ-gal-R: 5′-TAATACGACTCACTATAGGGCCGCCTCGTACAAAACAAGT-3′.</p><p>CquiOR4-F: 5′ AATACGACTCACTATAGGGGTGCGTGAAACTGTTCGGA-3′</p><p>CquiOR4-R: 5′ TAATACGACTCACTATAGGGGCGAGTGTCCAGCCCGTA-3′</p><p>CquiOR5-1-F: 5′TAATACGACTCACTATAGGGATGAAATTCTACGAGCTC</p><p>CquiOR5-1-R: 5′TAATACGACTCACTATAGGGAGTAATAAGCATGACATG</p><p>CquiOR5-2-F: 5′TAATACGACTCACTATAGGGCCGATGCACTTGCTGTTAG</p><p>CquiOR5-2-R: 5′TAATACGACTCACTATAGGGTTATGAATGCATCAATCGCT</p><p>Large scale dsRNAs were purified with MEGAclear kit (Ambion, Austin, TX) and precipitated with 5M ammonium acetate to yield 4–5 μg/μl of CquiOR4&5-dsRNA.</p><!><p>Female pupae (0-d-old) were collected in plastic cups filled with distilled water and kept on ice for 15 min. The sharp end of a yellow pipette tip was cut diagonally to make a stage to hold a pupa. Forty-five nanograms of dsRNAs in 9.2 nL volume were injected in the dorsal membrane close to the base of the trumpet using a NanoLiter 2000 inject (World Precision Instruments). The injected pupae were put in new plastic cups with distilled water and kept at 27 °C. Newly emerged adults were supplied with sugar water (10% wt/vol), and newly emerged males (ratio 1:1) were released into the cage for mating.</p><!><p>For tissue expression pattern analysis, each type of tissue (antennae, maxillary palps, proboscis, and legs) from three hundred non-blood-fed female mosquitoes (4–5 days old) and antennae from three hundred blood-fed female mosquitoes were dissected and collected in TRIzol reagent (Invitrogen, Carlsbad, CA) on ice using a stereomicroscope (Zeiss, Stemi DR 1663, Germany). For dsRNA treated mosquitoes, thirty pairs of antennae were dissected from each group and collected in 50% (vol/vol) ethanol diluted in DEPC-treated water on ice using a stereomicroscope. Total RNAs were extracted, and cDNAs were synthesized using iScript Reverse Transcription Supermix for RT-qPCR according to the manufacturer's instructions (Bio-Rad). Real-time quantitative PCR (qPCR) was carried out by using a CFX96 Touch™ Real-Time PCR Detection System (Bio-Rad) and SsoAdvanced SYBR Green Supermix (Bio-Rad). CquiRPS7 gene was used as a reference. The following pairs of detection primers were designed with Primer 3 program (frodo.wi.-mit.edu/):</p><p>CquiRPS7-Fw: 5′- ATCCTGGAGCTGGAGATGA -3′;</p><p>CquiRPS7-Rv: 5′- GATGACGATGGCCTTCTTGT -3′;</p><p>CquiOR4-qPCR-Fw: 5′- CTTTACCTCGGAGACCACCA-3′</p><p>CquiOR4-qPCR-Rv: 5′ -ACCGGTAGCTTGATCCAGTG-3′</p><p>CquiOR5-qPCR-Fw: 5′ -TTACTCCGATGCACTTGCTG-3′</p><p>CquiOR5-qPCR-Rv: 5′ -TCTCGCAAAGTTGATCCAGA-3′</p><p>qPCR was performed with three biological replicates, and each of them was replicated three times (three technical replicates per biological replicate); data were analyzed using the 2−ΔΔCT method.</p><!><p>The bioassay arena was modified from our surface-landing assay (Syed and Leal, 2008) initially designed to mimic a human arm without odors or humidity. CO2 at 50 mL/min was added to activate female mosquitoes, and blood was provided as both an attractant and a reward. In short, two 50-mL Dudley bubbling tubes, painted internally with a black hobby and craft enamel (Krylon, SCB-028), were held in a wooden board (30 × 30 cm), 17 cm apart from each end and 15 cm from the bottom. The board was attached to the frame of an aluminum collapsible field cage (30.5 × 30.5 × 30.5 cm; Bioquip). Two small openings were made 1 cm above each Dudley tube to hold two syringe needles (Sigma-Aldrich, 16-gauge, Z108782) to deliver CO2. To minimize the handling of mosquitoes, test females were kept inside collapsible field cages since the latest pupal stage. These female cages had their cover modified for behavioral studies. A red cardstock (The Country Porch, GX-CF-1) was placed internally at one face of the cage, and openings were made in the cardboard and cage cover so the cage could be attached to the wooden board with the two Dudley tubes and CO2 needles projecting inside the mosquito cage 6 and 3 cm, respectively. Additionally, windows were made on the top and the opposite end of the red cardstock for manipulations during the assays and a video camera connection, respectively. The mosquito cage housing 30–50 test females was connected to the platform holding the Dudley tubes at least 2h before bioassays. At least 10 min before the assays, water at 38 °C started to be circulated with a Lauda's Ecoline water bath, and CO2 at 50 mL/min was delivered from a gas tank just at the time of the behavioral observations. Sample rings were prepared from strips of filter papers 25 cm long and 4 cm wide and hung on the cardstock wall by insect pins to make a circle around the Dudley tubes. Cotton rolls (iDental, 1 × 3 cm) were loaded with 100 μL of defibrinated sheep blood purchased from the University of California, Davis, VetMed shop and placed between a Dudley tube and a CO2 needle. For each run one paper ring was loaded with 200 μL of hexane (control) and the other with 200 μL tested compounds at a certain concentration in hexane. The solvent was evaporated for 1–2 min, blood-impregnated cotton plugs, and filter paper rings were placed in the arena, CO2 was started, and the assays were recorded with a camcorder equipped with Super NightShot Plus infrared system (Sony Digital Handycam, DCR-DVD 810).</p><p>During the assay, the arena was inspected with a flashlight whose lens was covered with a red filter. After 5 min, the number of females that landed and continued to feed on each side of the arena was recorded. Insects were gently removed from the cotton rolls, and the assays were reinitiated after rotation of sample and control. Thus, repellency for each set of test mosquitoes was measured with the filter paper impregnated with the same sample at least once on the left and once on the right side of the arena. After three runs, filter paper strips and cotton plugs were disposed of, and new loads were prepared.</p><!><p>Amino acid sequences of ORs from Cx. quinquefasciatus, Ae. aegypti, and An. gambiae were aligned by Clustal Omega (https://www.ebi.ac.uk/Tools/msa/clustalo/) with the generated FASTA file being used as an entry for phylogenetic analysis in MEGA7 (Kumar et al., 2016). Parameters were set as following: bootstrap method: 1000; p-distance; pairwise deletion. The generated tree was imported into iTOL (https://itol.embl.de/).</p><!><p>Graphic illustrations were prepared with Prism 8 (GraphPad, La Jolla, CA). The number of mosquitoes in the treatment (T) and control (C) side of the arena were transformed into % protection, P% = (1-[T/C]) × 100, according to WHO (WHO, 2009) and EPA (EPA, 2010) recommendations. Data that passed the Shapiro-Wilk normality test were analyzed with the two-tailed, unpaired t-test; otherwise, data were analyzed with the Mann-Whitney test. All data are expressed as mean ± SEM.</p><!><p>We cloned the cDNAs for CquiOR2, CquiOR4, CquiOR5, CquiOR84, CquiOR85, AaegOR14, and AaegOR15, which are clustered (Fig. S1) along CquiOR1 (Xu et al., 2013). Then, each OR was separately coexpressed along with its coreceptor in Xenopus oocytes for deorphanization. We used a panel of odorants containing oviposition attractants, repellents, plant-derived compounds and other physiologically or behaviorally relevant compounds. We expected that the odorant profiles for these receptors would resemble that of CquiOR1 (Xu et al., 2013), but found marked differences. For example, three floral compounds elicited robust currents on CquiOR4/CquiOrco-expressing oocytes. Specifically, these oocytes were very sensitive to 2-phenylethanol, phenethyl formate, and, propionate (Fig. 1). To obtain more information about the sensitivity of the CquiOR4 + CquiOrco receptor, we performed concentration-response analyses for the three best ligands. Currents elicited by 2-phenylethanol were already saturated at the normal screening dose of 1 mM. Thus, we performed these analyses with concentrations in the range of 0.1 μM to 0.1 mM (Fig. 2). 2-Phenylethanol was indeed the most potent of the compounds in our panel, activating CquiOR4 + CquiOrco with EC50 of 28 nM.</p><p>CquiOR5 + CquiOrco receptor showed a quite different profile, with the three best ligands being linalool, p-methane-3,8-diol (PMD) and linalool oxide (Fig. 3). Although the responses were not as robust as those elicited by the best ligands for CquiOR4 + CquiOrco, they were dose-dependent (Fig. 4). Linalool activated CquiOR5/CquiOrco-expressing oocytes with an EC50 of 574 nM. By contrast, no compounds in our panel elicited relevant currents in CquiOR2/CquiOrco-expressing oocytes, except for isopentyl acetate (Fig. S2A). On the other hand, CquiOR84/CquiOrco-expressing oocytes responded with large currents when challenged with N-(2-isopropyl-phenyl)-3-methyl-benzamide or PMD (Fig. S2B). Similar responses were recorded with CquiOR85 + CquiOrco receptor.</p><p>The two receptors from Ae. aegypti in the cluster, ie, AaegOR14 and AaegOR15, gave remarkably different responses in terms of profile and sensitivity. AaegOR14/AaegOrco-expression oocytes elicited only weak currents when challenged with 4-methylphenol and 4-ethylphenol (Fig. S2C). By contrast, AaegOR15/AaegOrco-expressing oocytes generated dose-dependent, robust currents when challenged with phenethyl propionate, phenethyl formate, and acetophenone (Fig. 5). As far as the two major ligands are concerned, AaegOR15 profile resembles that of CquiOR4 + CquiOrco.</p><!><p>The best ligands for CquiOR4 and CquiOR5 have been previously reported to have repellency activity, particularly the commercially available PMD and 2-phenylethanol (USDA, 1947). To get a better insight into the possible role of these receptors in repellency behavior, we performed qPCR analyses. We surmised that transcript levels of the receptor for repellents and host attractants would decrease after a blood meal because blood-fed mosquitoes cease host finding. By contrast, transcripts of receptors involved in the reception of oviposition attractants would increase given that gravid females use their olfactory system to locate suitable sites for oviposition. Transcript levels of CquiOR4 decreased significantly (P = 0.0122) after a blood meal (Fig. 6A). Additionally, this receptor was highly enriched in female antennae, with only basal levels in maxillary palps, proboscis, and legs (Fig. 6A) thus further supporting a possible role in hosting finding or repellency activity. By contrast, transcript levels of CquiOR5 did not change significantly after a blood meal (Fig. 6B). Moreover, this receptor is not specific to or enriched in female antennae. We found high transcript levels in the proboscis, maxillary palps, and legs (Fig. 6B).</p><!><p>Previously, 2-phenylethanol has been tested against the yellow fever mosquito (USDA, 1947) and reported to be active between 61 and 120 min. Using our surface landing and feeding assay (Leal et al., 2017b; Xu et al., 2014), we tested repellency activity against the Southern house mosquito. In our preliminary assessment, 0.01% 2-phenylethanol gave 34.5 ± 6.1% protection (n = 8), whereas at 0.1% it provided significantly higher protection 66.8 ± 2.7% (n = 7, Mann-Whitney test, P = 0.0003). We then compared 2-phenylethanol-elicited protection with that obtained with DEET, both at 1% and they showed comparable activity: DEET, 98.7 ± 0.9%; 2-phenylethanol, 96.1 ± 2.1% (n = 10 each; Mann-Whitney test, P = 0.6313). It is noteworthy mentioning that 2-phenylethanol is more volatile than DEET; thus, it may appear as strong as DEET but may lose activity more rapidly, consistent with the earlier tested against Ae. aegypti (USDA, 1947). One of the main features of DEET is a long complete protection time.</p><!><p>We performed RNAi experiments to test whether reducing transcript levels of CquiOR4 would affect repellency activity. Similarly, we tested CquiOR5 vis-à-vis the best ligands. First, we compared the transcript levels of these genes in water-, β-galactosidase-dsRNA- and CquiOR4-dsRNA-injected mosquitoes. Transcript levels of CquiOR4 decreased significantly (P = 0.0003, unpaired, two-tailed t-test) in knockdown mosquitoes (Fig. S3A). Similarly, CquiOR5-dsRNA-treated mosquitoes had significantly lower transcripts of CquiOR5 than mosquitoes injected with water (P = 0.0013, unpaired, two-tailed t-test) or β-galactosidase-dsRNA.</p><p>Next, RNAi-treated mosquitoes were used to compare repellency activity. Because RNAi treatment reduced transcript levels only by ca. 60%, it is important to test repellency at low doses, otherwise, a possible link between reception and behavior may be overlooked. When tested at 0.01%, protection conferred by 2-phenylethanol, albeit low, was significantly reduced (n = 12–15, Mann-Whitney test, P = 0.0062) (Fig. 7A). At a higher dose (0.1%), 2-phenylethanol-elicited protection is lower in CquiOR4-dsRNA-treated than in β-galactosidase-dsRNA-treated mosquitoes, but there was no significant difference (n = 6–7, Mann-Whitney test, P = 0.2191). By contrast, DEET-elicited repellency in these two groups of mosquitoes was not significantly different (n = 4, Mann-Whitney test, P > 0.999) (Fig. 7A).</p><p>Measuring linalool-elicited repellency activity at 0.1% showed a non-significant reduction of protection in CquiOR5-dsRNA-treated mosquitoes as compared to β-galactosidase-dsRNA-treated mosquitoes (n = 11–16, Mann-Whitney test, P = 0.567) (Fig. 7B). Likewise, protection by PMD was also reduced, but not significantly different (n = 12–16, Mann-Whitney test, P = 0.1411) (Fig. 7B). There was no significant difference in repellency (protection) elicited by DEET (n = 12, Mann-Whitney test, P = 0.2571) (Fig. 7B). Attempts to test repellency at lower doses were unrewarding. At 0.05% linalool, 56 ± 5.1% of mosquitoes responded to the control side of the arena, whereas 44 ± 5% responded to treatment (n = 6). Likewise, PMD at 5% provided no protection: control, 51.7 ± 2%; treatment, 48.2 ± 2%. Thus, we were unable to test the effect of RNAi treatment on repellency activity at lower than 0.1% dose.</p><p>In conclusion, repellency activity mediated by linalool and PMD might involve multiple receptors. Here, we show that PMD activates not only CquiOR5 but also CquiOR84/85. Previously, we showed that PMD activated a DEET receptor in the Southern house mosquito, CquiOR136 (Xu et al., 2014). The reduction in protection observed with knockdown mosquitoes, albeit not statistically significant, suggests that CquiOR5 might be one of the receptors mediating linalool and PMD repellency activity. On the other hand, we cannot rule out the possibility that other receptors are involved in a combinatorial code reception of 2-phenylethanol, but the significant reduction in protection in CquiOR4-dsRNA-treated mosquitoes suggests it may play a significant part in 2-phenylethanol-mediated repellency activity.</p>

PubMed Author Manuscript

Noncollinear Relativistic DFT+U Calculations of Actinide Dioxide Surfaces

A noncollinear relativistic PBEsol+U study of the low-index actinide dioxides (AnO 2 , An = U, Np, Pu) surfaces has been conducted. The surface properties of the AnO 2 have been investigated and the importance of the reorientation of magnetic vectors relative to the plane of the surface is highlighted. In collinear nonrelativistic surface models, the orientation of the magnetic moments is often ignored; however, the use of noncollinear relativistic methods is key to the design of reliable computational models. The ionic relaxation of each surface is shown to be confined to the first three monolayers and we have explored the configurations of the terminal oxygen ions on the reconstructed (001) surface. The reconstructed (001) surfaces are ordered as (001)αβ < (001)α < (001)β in terms of energetics. Electrostatic potential isosurface and scanning tunneling microscopy images have also been calculated. By considering the energetics of the low-index AnO 2 surfaces, an octahedral Wulff crystal morphology has been calculated.

noncollinear_relativistic_dft+u_calculations_of_actinide_dioxide_surfaces

Introduction<!>Magnetic Structure<!>Computational Methodology 2.1 Calculation Details<!>Low-Index Surface Models<!>The HIVE Code<!>Wulff Reconstruction<!>Model Constraints 3.1.1 Surface Energetics<!>Ionic Relaxation<!>Surface Properties 3.2.1 Electronic Structure<!>Magnetic Deviation<!>Scanning Tunneling Microscopy<!>Electrostatic Potential Isosurface<!>Crystal Morphology

<p>The surface chemistry of the actinide dioxides (AnO 2 , An = U, Np, Pu) is key to understanding corrosion mechanisms, 1-8 which impacts the design of long-term storage facilities and the industrial reprocessing of nuclear fuels. [9][10][11][12][13][14][15] An oxide layer is inexorably formed on actinide metal surface, which affects the chemistry of the underlying actinide metal. 1-5, 7, 16 The rapid onset of corrosion has resulted in thermal excursions, failure of containment vessels, and the resulting dispersal of nuclear materials. To reduce the risk of nuclear proliferation and assist in nuclear decommissioning, the controlled oxidation of actinide metals offers a means of converting classified nuclear material to simple ingots. 7 In terms of fuel fabrication, the surface energetics of the AnO 2 impact on fuel sintering and particle morphology. 17 As a result of their inhomogeneous and radioactive nature, few AnO 2 experimental surface studies have been completed. 9,13,[18][19][20][21][22][23][24] . To circumvent experimental issues, computational methods offer an attractive alternative and complementary mode of study. However, a computational investigation of heavy-fermion systems is extremely challenging. To investigate the complex electronic structure by computational methods, we must consider exchange-correlation influences, relativistic contributions, and noncollinear magnetic behaviour. Only a limited number of studies have considered relativistic contributions (spinorbit interaction, SOI), which is, however, important in the treatment of actinide systems. [25][26][27] In addition, the actinides often have complex (noncollinear) magnetic structures, and thus far no investigation of AnO 2 surfaces has incorporated noncollinear magnetic behavior into the models.</p><p>The actinides are highly-correlated f-electron systems for which conventional DFT methods calculate an incorrect electronic structure. To model highly-correlated materials correctly, a number of methods have been developed: the self-interaction correction (SIC) method, 28 modified density functional theory (DFT+U), [29][30][31][32][33] dynamic mean field theory (DMFT), 34 and hybrid density functionals. [35][36][37] . As a computationally tractable method, DFT+U offers a means of study in which the electronic structure can be computed. In the Liechtenstein DFT+U formulism, where independent Coulomb (U) and exchange (J) terms treat the on-site Coulomb repulsion of the An f-electrons. The values are derived from higher level ab-initio methods or obtained through semi-empirical analysis. 25 In this study, the low-index AnO 2 (An = U, Np, Pu) surfaces have been investigated by computational methods. The electronic structures of the AnO 2 are heavily influenced by changes in magnetic order [26][27] and the importance of magnetic vector reorientation is underlined. The effect of transverse 3k AFM behavior on the properties of the UO 2 surface is unknown, whereas investigations on NpO 2 and PuO 2 surfaces are even less common. 9,11 Surface energetics, the degree of ionic relaxation, electrostatic isosurfaces, scanning tunneling microscopy (STM) images and crystal morphologies have been calculated and the impact of oxygen ion reconstruction on the inherently unstable (001) surface is also considered.</p><p>©British Crown Owned Copyright 2018/AWE This manuscript has been authored by UT-Battelle, LLC, under contract DE-AC05-00OR22725 with the US Department of Energy (DOE). The US government retains and the publisher, by accepting the article for publication, acknowledges that the US government retains a nonexclusive, paid-up, irrevocable, worldwide license to publish or reproduce the published form of this manuscript, or allow others to do so, for US government purposes. DOE will provide public access to these results of federally sponsored research in accordance with the DOE Public Access Plan (http://energy.gov/downloads/doe-public-access-plan).</p><!><p>The magnetic structure of the AnO 2 is highly complicated. A discontinuous first-order magnetic phase transition (T N = 30.8 K) 38 in UO 2 has been established by heat capacity, [39][40] magnetic susceptibility 41 and neutron diffraction [42][43][44] measurements. A transverse 3k antiferromagnetic (AFM) ground-state has been identified (Figure 1). 25,[45][46] The ground-state corresponds to an internal Pa 3 (No. 205) crystallographic distortion synonymous with magnetic order (the displacement of the O 2ions is 0.014 Å). [44][45][47][48] Figure 1: The longitudinal 3k AFM and transverse 3k AFM phases for the AnO 2 crystal structure.</p><p>The magnetic structure of NpO 2 remains unresolved. In the absence of interactions that break time-reversal symmetry conditions, the Np 4+ ion (a Kramers ion, one with an uneven number of valence electrons) should order magnetically at low-temperature. 49 A first-order paramagnetic (PM)-AFM phase transition (T = 25.4 K) has been inferred by: magnetic susceptibility 50 and specific heat capacity measurements. [51][52] In-spite of an exhaustive search, a measurable local magnetic moment has not been identified by: low-temperature Mossbauer (T = 1.5 K), 49 neutron diffraction (12 K < T < 30K), 53 and muon spin rotation (0.3 K < T < 25.4 K) measurements. [54][55] In terms of the crystal structure, no evidence has been found of an external distortion, which would indicate noncollinear 3k AFM order. 52 An internal O 2ion distortion (indicative of transverse 3k AFM behaviour with Pa 3 (No. 205) crystal symmetry) can be inferred from: the small broadening of Mossbauer spectroscopic lines, 49 and inelastic neutron scattering (INS) (5 K < T < 25 K) measurements. [56][57] An internal O 2ion distortion of 0.02 Å has been calculated, which is, however, below the experimental resolution. 52,55,58 In contrast, a longitudinal 3k AFM ground-state has been indicated by: resonant X-ray scattering 59 (10 K < T < 17 K) and 17 O NMR measurements (T = 17 K). 60 The transverse 3k AFM state, relative to the longitudinal 3k AFM state, is 0.002 eV•formula unit -1 lower in energy (as calculated by HSE06 incorporating SOI). 26 An experimental singlet Γ1 diamagnetic (DM) PuO 2 ground-state has been inferred from: magnetic susceptibility (T = 4 K), inelastic neutron scattering (T > 30 K), and nuclear magnetic resonance (T > 4 K) measurements. However, a number of inconsistencies have been identified, and an ordered magnetic ground-state can be assumed. In contrast to experimental measurements, a longitudinal 3k AFM ground-state has been calculated. 18, 22-24, 26, 37, 61-71 It is thought that PuO 2 could be a small-moment insulator (similar to NpO 2 ) for which DFT overestimates the magnetic moments. 26 In this study, transverse 3k AFM order (UO 2 , NpO 2 ) and longitudinal 3k AFM order (PuO 2 ) have been used to describe the crystals.</p><p>To model noncollinear magnetic behavior, it is imperative that relativistic effects are considered. A significant number of studies ignore the SOI (important in heavy-fermion systems) to reduce the computational cost. 11,[72][73][74] A limited number of studies on UO 2 9 and PuO 2 23 consider relativistic contributions to the total energy. The importance of SOI on modeling UO 2 by DFT initially seemed to be inconsequential. 9 In a nonrelativistic treatment of other actinide systems, the study has often been cited to justify the absence of SOI. 9,23,72 The importance of SOI on the PuO 2 (111) surface energies has now been highlighted by hybrid DFT, 23 but all studies have limited themselves to a discussion of collinear 1k AFM order.</p><p>A major limitation of scalar calculations is the inability to orient the magnetic moments relative to the direction of the surface. In this manner, the magnetic moments are directed orthogonal to the surface plane, which leads to notable inconsistencies within the electronic structure. If not corrected, the orientation of the magnetic field is also directed orthogonal to the surface plane, because the principal axis differs between the surfaces. Consequently, the electronic, magnetic, and crystal structures differ between the bulk crystal structure and individual surfaces. If the magnetic vectors are not reoriented, the energetics and structural relaxations derived by this approach are incomplete. [26][27] This is particularly concerning when calculating the surface energy, which is derived from the bulk structure and it is therefore important that the magnetic vectors relative to the surface are carefully reoriented. In past studies where this essential transformation has been omitted, the energies of the bulk and surface are therefore often incomparable, which introduces a significant error when calculating the energy of the surface. In this study, the magnetic vectors are reoriented relative to the surface plane, which ensures that we preserve the noncolinear 3k AFM structure. In addition, the reduction of cubic symmetry associated with collinear 1k AFM states (used in past calculations) is avoided. [26][27] Figure 2: The surface magnetism of a two-dimensional material. The direction of the magnetic moments for the respective surfaces are shown for the first two layers of the bulk crystal structure. The highlighted (01) (green) and ( 11) (blue) surfaces correctly emulate the magnetic structure in the bulk crystal. In contrast, the (11) (red) surface illustrates an incorrect depiction where the magnetic moments are aligned orthogonal to the surface.</p><p>The magnetic structure is commonly defined by the principal axis. The principal axis of the AnO 2 (111), (011) surface differs from that of the bulk crystal and the final magnetic, electronic and crystal structures are therefore inequivalent. However, this is not the case for the AnO 2 (001) surface which shares the same axes. To illustrate in a two-dimensional material, we consider the first two layers of a collinear 1k AFM material (Figure 2). The (01) surface and the crystal share the same principal axis and the magnetic structures are therefore directly related. In the (11) surface, the principal axis differs from that of the crystal, which results in an unrelated magnetic and electronic structure. It is therefore critical to orient the magnetic vectors to emulate the initial crystal structure.</p><!><p>All calculations have employed the Vienna Ab-initio Simulation Package (VASP) 28,34,75 using a plane wave basis set, relativistic effective core potentials (ECPs), and the frozen-core projector-augmented wave (PAW) method. The cut-off energy of the plane wave basis set is 500 eV. The uranium (6s 2 , 7s 2 , 6p 6 , 6d 2 5f 2 ), neptunium (6s 2 , 7s 2 , 6p 6 , 6d 2 5f 3 ), plutonium (6s 2 , 7s 2 , 6p 6 , 6d 2 5f 4 ) and oxygen (2s 2 , 2p 4 ) valence electrons are implicitly considered. The integration over the Brillouin zone was performed using the Blöchl tetrahedron method. 76 The influence of the SOI 72 and noncollinear magnetic wave-vectors are considered.</p><p>The on-site Coulomb repulsion of the An 5f electrons is treated by the Liechtenstein et al. DFT+U [31][32][33] formulism. 32 In the Liechtenstein et al. formulism, the Coulomb (U) and exchange (J) modifiers are treated as independent variables. 32 The Coulomb modifier for each ion is written in the parentheses: uranium (U = 3.35 eV), neptunium (U = 4.25 eV) and plutonium (U = 6.00 eV). In the past, the influence of J on noncollinear magnetic materials has been investigated. [25][26][27]77 The introduction of J increased the anisotropic nature of the fstates, 78 and it is therefore not considered in this study. The selected conditions offer an accurate representation of the electronic structure. The integration of the Brillouin zone is performed with a Γ-centered k-point grid. 79 The exchange-correlation energy is evaluated by the revised Perdew-Burke-Ernzerhof for solids (PBEsol) functional. [29][30]80 The iteration threshold for electronic and ionic convergence is set at 1x10 -5 eV and 1x10 -2 eV Å -1 , respectively. As the crystal and electronic structures of AnO 2 are highly dependent on the magnetic state, it is imperative to correctly reorientate the magnetic vectors with respect to the surface plane. Ionic relaxation is a common mechanism by which the surface energy is minimized with respect to the unrelaxed surface. The surface energy (γ) is a measure of the surface stability and is defined by:</p><p>The number of formula units (N), the total energy of the surface slab (E tot (N)) and the total energy per formula unit ( E AnO 2 ) are defined in the parentheses. In our calculations, all ions are relaxed while the dimensions of the unit cell are fixed. The conjugate gradient method is employed in the relaxation of the ions. Images are visualized by the Crystal Maker 81 and VESTA codes. 82 The density of states have been illustrated by the SUMO code, a commandline plotting tool for ab-initio calculations.</p><!><p>The low-index AnO 2 (111), ( 011), (001) surfaces are generated by the METADISE code (Figure 3) 83 from the ionically relaxed bulk material. The nonpolar (111) surfaces are comprised of repeat O-An-O unit layers. In the (111) surface, the individual monolayers are charged, but the surface is characterized by the absence of a dipole moment perpendicular to the surface plane. In contrast, the (011) surface is comprised of nonpolar and charge neutral planes. The polar (001) surface (formed of dipolar An-O layers) is inherently unstable 17,[84][85] as the electrostatic energy diverges (caused by the formation of an electric dipole) with increasing number of monolayers. [86][87] In nature, the surface undergoes a reconstruction to prevent the formation of an electrostatic dipole. The reconstruction is influenced by environmental conditions. 85,88 In this study, the dipolar perpendicular to the surface is removed by transposing half of the charged oxygen anions from one surface to the other (Figure 4), which involves the formation of an oxygen-terminated surface with half-filled oxygen vacancies. The result is a non-polar reconstructed (001)r surface, which in a (1•1) unit cell can be either the (001)α or (001)β reconstruction. Although numerous configurations are possible in a (1•2) unit cell, the (001)αβ reconstruction offers a hybridization between the two (1•1) reconstructions and in this study we have calculated the relative stabilities of these three surface configurations.</p><p>The surface energy is converged with respect to the k-point grid to under 0.05 J m -2 (Figure 5). The (111) surface is calculated from a 5•5•1 Γ-centered k-point grid recommended for hexagonal structures, whereas the (011) and (001) surfaces are calculated from a 4•4•1 Γcentered k-point grid. 84 To minimize potential aliasing errors, the initial bulk structure (from which the surfaces are derived) is calculated with both a 4•4•4 and a 5•5•5 Γ-centered k-point grid for direct comparison with the surface in the surface energy calculations. Finally, the (001)αβ surface is calculated form a 4•2•1 Γ-centered k-point.</p><!><p>In the scanning tunnelling microscopy (STM) HIVE code, 89-90 the Tersoff-Hamann model is considered, where the tunnelling-current is equivalent to the local density of states. 91 A point source at a constant height of 2.5 Å and a Fermi energy sample bias of -2.50 eV is used. Topographies calculated by HIVE include: copper, 92 germanium, 89-90 gold, 93 iron oxide, 94 thorium dioxide. 95</p><!><p>According to the Gibbs thermodynamic principle, the equilibrium crystal morphology is influenced by the total surface energy of the medium interface. An equilibrium crystal morphology that minimises ∆ G i has been calculated as follows (Equation 3):</p><p>The terms in the parentheses describe the total crystal-medium interface free energy ( ∆ G i ), the surface Gibbs free energy ( γ j ) and the surface area ( A j ).</p><!><p>As a function of the number of formula units used, the energy of the low-index AnO 2 surfaces has been calculated (Additional Information, Figure A1). The ions are fully relaxed while keeping the relative dimensions of the unit cell fixed. In this study, the surface energy is converged to within 0.01 J•m -2 when 12 or more formula units are used. The surface energy increases across the series as (111) < (011) < (001)α < (001)β (typical of fluorite-structured materials) (Table 1). 86,95 The energy difference between the (001)α and (001)β terminations are relatively small in UO 2 (0.08 J m -2 ) and NpO 2 (0.06 J m -2 ), compared to PuO 2 (0.19 J m -2 ) .</p><p>If one uses a (1•1) unit cell model, the (001)α surface relative to the (001)β surface is energetically favourable, which is confirmed independently by an interatomic potential-based investigation on UO 2 . 88 Compared with past DFT-based methods, the calculated surface energies are considerably greater for each surface. 9, 11, 17-18, 23, 74, 96 although interatomic potential models 88 and relativistic hybrid calculations 15 of UO 2 have resulted in even higher surface energies. In addition, interatomic potential models of UO 2 have calculated lower-energy (001) surface reconstructions, which are formed using a larger unit cell. 88,97 In the reconstruction of the (001) surface in our (1•1) unit cell, only the (001)α and (001)β configurations can be generated, whereas the surface energy of the (001)αβ configuration from a (1•2) unit cell (calculated using 28 formula units) relative to the (001)α and (001)β configurations, is considerably lower in energy (Table 1). This implies a limitation of the DFT (1•1) unit cell model and it is clearly possible that other configurations, in even larger cells, could be more stable. However, increasing the size of the cell increases the computational cost of the system significantly, and a systematic fully relativistic DFT study of bigger simulations cells is currently computationally intractable.</p><!><p>The low-index AnO 2 surfaces are characterized by the changes in the interlayer spacings (Figure 6-7), which enables a quantitative analysis of the structural relaxation between layers. The interlayer relaxation (Δd interlayer ) is calculated by:</p><p>where (d i,i+1 ) relaxed is the average interlayer separation of ions in the relaxed surface and d unrelaxed is the average interlayer separation of ions in the unrelaxed surface. The interlayer relaxation is reminiscent of studies on the isostructural CeO 2 material with similar results found for the (111) and (011) surfaces. 98 In the context of An-An relaxation, the (111) surface is marginally distorted. The major difference is confined to the oxygen separation in the second interlayer space. The (011) surface undergoes the greatest overall interlayer relaxation, with the first surface layer experiencing a marked contraction, where the first An layer contracts significantly more than the first O layer. The contraction of the first layer is countered by a slight expansion of An ions in the second layer, but the bulk structure is regained by the fifth layer. The terminal O ions in the (001)α and (001)β surface undergo a significant contraction, although the remainder of the structure is relatively unaffected. In general, the interlayer relaxation is confined to the first 5 Å, indicating that for investigations of surface reactivity, a slab of minimally 10 Å thick should be used. Our results are similar to those found in studies of CeO 2 and ThO 2 . 95 In the context of interlayer O-O relaxation, the distortion of the surface is primarily confined to the first three to four monolayers and the degree of ionic relaxation is generally identical in the AnO 2 surfaces, with the exception of the PuO 2 (001)β surface. In the PuO 2 (001)β surface, the relaxation of the oxygen ions is significantly less relative to the UO 2 and NpO 2 (001)β surfaces. Thus, of the (001)r surfaces, the UO 2 and NpO 2 (001)β surfaces undergo the greatest surface relaxation, whereas in PuO 2 , the (001)α surface undergoes the greatest surface relaxation, which is a result of magnetic order and the relaxation in the xy-plane.</p><p>No significant structural distortion in the xy-plane occurs in the AnO 2 (111), ( 011) or (001)α surfaces, possibly as a result of preserving the Pa 3 (No. 205) or Fm 3 m (No. 225) cubic symmetry from the use of noncollinear 3k AFM order. [26][27] In contrast, the oxygen ions in the UO 2 and NpO 2 (001)β configuration are shifted from their initial positions by the use of transverse 3k AFM ordering (Figure 8). This distortion is not observed in the corresponding PuO 2 surface in which the ions are relatively fixed, although there is a minor distortion of the surface plutonium ions, potentially as a consequence of using either transverse 3k AFM or longitudinal 3k AFM behavior. By comparison, the oxygen ions in the (001)αβ configuration are relatively static and, instead, the actinide ion is partially shifted toward the terminal oxygen ions.</p><!><p>The electronic structure of the AnO 2 surfaces has been calculated (Figure 9). The covalent nature of the AnO 2 materials (a consequence of An (f) and O (p) mixing) is seen to increase along the series. The Mott-Hubbard insulating nature of UO 2 is characterized by transitions primarily occurring across the An f-bands. Compared to relativistic hybrid DFT calculations of UO 2 , the calculated band gaps for the low-index surfaces are considerably greater. 15 The charge-transfer insulating nature of PuO 2 is characterized by transitions primarily between the valence Pu f-band and conduction O p-band. In NpO 2 , both Mott insulating and chargetransfer characteristics are shown in the surface. In general, the electronic structure is only partially perturbed between surfaces.</p><p>In addition, the electron affinity and ionization potential of the AnO 2 surfaces has been calculated (Table 2). This information fills a significant gap in the literature where X-ray photoelectron spectroscopy (XPS) and Kelvin probe microscopy studies have yet to be performed. The electron affinity and the ionization potential increases along the (011) < (111) < (001)β < (001)α series. Of the AnO 2 (An = U, Np, Pu) materials, UO 2 is the least reactive, whereas PuO 2 is the most reactive.</p><!><p>The magnetic structure of the low-index AnO 2 surfaces has been investigated. A complete analysis of the actinide ions can be found in the Additional Information. The localized magnetic normalized root-mean-square deviation (nRMSD) of the first three monolayers has been calculated for each surface (Figure 10). As the monolayer surface depth increases, the magnetic distortion decreases. The total magnetic moment of the U (1.37 µ B •ion -1 ), Np (2.70 µ B •ion -1 ), and Pu (3.80 µ B •ion -1 ) ions remains constant. 011), (001)α surfaces for the first three monolayers. The initial magnetic vector (silver), relaxed magnetic vector (green), actinide (blue) and oxygen (red) are shown.</p><p>The localized magnetic deviation in NpO 2 for identical surfaces is relatively high. A number of competing low-temperature (T < 25.4 K) magnetic states could cause the distortion. 26 For instance, the transverse 3k AFM state, relative to the FM (111) ground-state, is 0.002 eV per formula unit higher in energy; however, no experimental evidence of a FM (111) ground-sate, which results in a R 3 m (No. 166) crystallographic distortion, exists. 26 In addition, the localized magnetic deviation of the (001)α series can be ascribed to the surface instability. In the first three monolayers of the (001)α surface, a FM and an AFM domain are formed. The lowest RMSD is found for the PuO 2 (011) surface.</p><!><p>The surface energies of UO 2 are extremely sensitive to stoichiometry, defect chemistry, and environmental conditions. [99][100][101] Low-energy electron diffraction (LEED) measurements of the UO 2 (111) surface have identified over 16 individual patterns. 102 . To assist experimental analysis, low-index AnO 2 STM images have been calculated (Figure 11) and the resulting images are analogues of experimental STM studies of AnO 2 surfaces. 85,[103][104] However, in an STM experiment, ionic positions are influenced by perturbations of the electric field caused by the probe and the calculated resolution is therefore considerably greater compared to that of an experimental study.</p><p>The terminal O 2ions are observed in white, whereas the An 4+ ions area considerably darker. The individual AnO 2 (An = U, Np, Pu) (111), ( 011) and (001)α surfaces patterns are indistinct. In the (111) surface, the O 2ions result in a hexagonal structure, whereas in the (011) surface, a series of darker channels is observed in one direction. In the (001)α surface, the alignment of the O 2ions results in a diamond pattern. As a means of differentiating between compounds, the (001)β surface is influenced by the magnetic state. In the transverse 3k AFM state for UO 2 and NpO 2 , the O 2channels oscillate continuously, whereas in the longitudinal 3k AFM state for PuO 2 , the O 2channels are perfectly linear. In other words, the structures can be differentiated by the transverse 3k AFM state of UO 2 and NpO 2 or by the longitudinal 3k AFM state of PuO 2 which is useful information for comparison with future experimental patterns to deduce the magnetic states.</p><!><p>The electrostatic potential isosurface for the low-index AnO 2 surfaces has been calculated using the PBEsol+U functional (Figure 12</p><!><p>Low-voltage scanning electron microscopy (SEM) of UO 2 has shown a truncated octahedral Wulff crystal morphology, 105 which to our knowledge is the only experimental study concerning the morphology. The truncated octahedral Wulff crystal morphology of UO 2 is inconsistent with studies of other CaF 2 -type crystal structures and may be the result of environmental influences and the method of sample preparation. The crystals were formed under high pressure (400 MPa) and temperature (1700 °C). A truncated octahedral morphology for a fluorite-structured material has not as yet resulted from any computational approach.</p><p>In this study, an octahedral Wulff crystal morphology has been calculated (Figure 13) from the surface energies of the low-index (111), (011) and (001)αβ surfaces only. As a result of their relative instabilities, the (001)α and (001)β surface are omitted. Indeed, other high-index surfaces are considerably greater in energy, and their influence on the Wulff crystal morphology is assumed to be negligible. In terms of computational theory, calculations have shown that the crystal structure is influenced by the magnetic state. [25][26][27] In theory, the low-temperature octahedral Wulff crystal morphology is linked to the noncollinear 3k AFM state, whereas the high-temperature truncated octahedral Wulff crystal morphology is linked to the PM state. In contrast, the octahedral Wulff crystal morphology of the AnO 2 materials is consistent with fluorite-based materials. The octahedral morphology in the present study is consistent with that calculated by interatomic potentials 88 and with previously reported morphologies for PuO 2 22 and ThO 2 95 calculated by DFT. The (111) surface dominates the morphological features of the particle.</p><p>Interatomic potential models of the UO 2 (001) surface have indicated surface configurations of lower energy in a (2x2) unit cell, however this energy is not sufficiently low enough to result in a truncated octahedron. 88 In the calculation of (001) surface energetics, the major limitation is the size of the unit cell and there is therefore a possibility that larger cells may result in a configuration of sufficiently low energy to result in a truncated octahedron. In this study, we have used a (1x1) unit cell with either the (001)α or (001)β configuration, although additional configurations are possible in larger supercells. In theory, one of these surfaces may possess sufficiently low energy to affect the morphology. However, a systematic investigation of the (2x2) surface is computationally unfeasible, because of the large number of compute-intensive configurations that must be explored.</p><p>In another scenario, the experimental sensitivity of UO 2 resulted in a crystal morphology influenced by environmental conditions. It is known that the interaction of oxygen with the AnO 2 surfaces influences the composition range of the solid and the formation of superficial structures. 102 In the past, DFT+U studies have indicated that the truncated crystal morphology is the result of oxygen-rich conditions at 300 K. 106 In addition, interatomic potentials indicate that the AnO 2 (001) surface energy is reduced by hydroxylation, 12,17 which also results in a truncated octahedron. Other models which use interatomic potentials have obtained an octahedral morphology at thermodynamic equilibrium. However, these studies concluded that the truncated morphology is the result of kinetic limitations. 107 Finally, numerous experimental investigations have shown that the surface energies are temperaturedependent. 99,108 4 Conclusions PBEsol+U has been used to investigate AnO 2 surfaces. In the past, collinear 1k AFM states have been used to model surface structures, but these models predominately use scalar approximations of the crystal electric field which causes an inability to reorient the magnetic vectors relative to the plane of the surface. Therefore, the magnetic structures differ across surface indices. This study considers non-collinear 3k AFM behavior and SOI contributions to the surface energetics of the low-index AnO 2 (111), ( 011) and (011) surfaces. The magnetic field is carefully re-oriented relative to the plane of the surface for a complete description of the magnetic surface structure. Localized magnetic distortions have been identified.</p><p>The interlayer relaxation of the (111), ( 011) and (001)α surfaces is confined to the first 5 Å.</p><p>In contrast to past DFT investigations, our surface energies are considerably higher, 11,74 which illustrates the important contribution of the SOI 72 to the calculated surface energetics. Our surface energies suggest that the chemical reactivity of the surface has previously been underestimated. The surface stability increases across the (001)β < (001)α < (011) < (111) series, which is typical of CaF 2 -type structures. From our Wulff reconstruction, the octahedral crystal morphology is completely dominated by (111) facets. As stated, this is consistent with previous calculations of fluorite-type structures. Thus, we have developed a computationally tractable method to model the low-index AnO 2 surfaces with improved energetics, which may serve as the basis for future studies 5 Acknowledgements</p>

ChemRxiv

Modulation of conformational changes in helix 69 mutants by pseudouridine modifications

Centrally located at the ribosomal subunit interface and mRNA tunnel, helix 69 (H69) from 23S rRNA participates in key steps of translation. Ribosome activity is influenced by three pseudouridine modifications, which modulate the structure and conformational behavior of H69. To understand how H69 is affected by the presence of pseudouridine in combination with sequence changes, the biophysical properties of wild-type H69 and representative mutants (A1912G, U1917C, and A1919G) were examined. Results from NMR and circular dichroism spectroscopy indicate that pH-dependent structural changes of wild-type H69 and the chosen mutants are modulated by pseudouridine and loop sequence. The effects of the mutations on global stability of H69 are negligible, however, pseudouridine stabilizes H69 at low pH conditions. Alterations to induced conformational changes of H69 likely result in compromised function, as indicated by previous biological studies.

modulation_of_conformational_changes_in_helix_69_mutants_by_pseudouridine_modifications

1. Introduction<!>2. Materials and methods<!>3.1 Pseudouridine modifications modulate pH-dependent structural changes of the H69 wild-type sequence and A1912G mutant<!>3.2 Pseudouridine modifications and A1919G mutation work synergistically to modulate the pH-dependent structural changes of \xce\xa8\xce\xa8\xce\xa8-A1919G<!>3.3 The U/\xce\xa81917C mutation affects pH-dependent structural changes of H69<!>3.4 The pH-dependent structural changes of H69 and mutants are not correlated to alteration in global stability<!>4. Conclusions

<p>As the universally conserved machinery for protein synthesis, ribosomes play an indispensible role in the survival of all living organisms [1]. The ribosome is composed of two subunits, small and large, associated through multiple intersubunit bridges [2–3]. At the subunit interface, the mRNA molecule snakes through the ribosome, in concert with binding of tRNAs at the aminoacyl (A), peptidyltransferase (P), and exit (E) sites [4]. Within this region, helix 69 (H69) (Fig. 1), a 19-nucleotide hairpin located in domain IV of the 23S rRNA (large subunit RNA), establishes intersubunit bridge B2a with helix 44 (h44) of the 16S rRNA (small subunit RNA), and participates in the formation of intersubunit bridge B2b with helix 24 (h24) of the 16S rRNA [5]. Due to its central location in the ribosome, H69 is involved in almost every step of translation, in addition to ribosome rescue and antibiotic binding [6].</p><p>X-ray crystallography and solution-state nuclear magnetic resonance (NMR) spectroscopy revealed that bacterial (Escherichia coli) H69 assumes a hairpin structure, with a stem composed of five base pairs and a single-stranded loop of nine nucleotides [7–8], which is also observed in eukaryotic H69 [9]. Similarity in the 3D structures was deduced by high conservation and covariance of H69 sequences throughout phylogeny (Fig. 2a) [10]. In addition, pseudouridine (Ψ) modifications at loop positions 1911, 1915, and 1917 (E. coli numbering) are also conserved across the three kingdoms [11–15]. In this work, pseudouridylated H69 and the corresponding unmodified (uridine-containing) RNA will be referred to as ΨΨΨ and UUU, respectively. Uridine is isomerized to Ψ by replacing the N1-C1′ glycosidic bond with a C5-C1′ linkage (Fig 2b). In E. coli, a single pseudouridine synthase, RluD, catalyzes this conversion at all three nucleotides in H69 [16]. The Ψ modifications have been shown to modulate structure and conformational behavior of H69 by stabilizing the loop-closing base pair and promoting base stacking in the 3′ half of the loop [8]. In E. coli H69, no net effects on hairpin stability due to Ψ modifications were observed at neutral pH. Previous studies showed that slight destabilizing effects of Ψ1915 and Ψ1917 canceled the stabilizing effects of Ψ1911 [17].</p><p>Conformational modulation by Ψ was discovered by examining pH sensitivity of the H69 structure [18–19]. More specifically, an A1913 stacked structure with protection from solvent (referred to as the "closed" conformation) was observed only in pseudouridylated H69 at low pH (5.5). In contrast, an alternate structure occurred at high pH (7.0), with increased exposure of A1913 ("open" conformation) [20–21]. These structures were proposed to be correlated with different conformational states of H69 during translation and establishment of intersubunit bridge B2a in complete ribosomes [18–19]. Pseudouridine modifications in H69 are not essential for bacterial growth under normal conditions [22]; however, loss of these modifications is disadvantageous for the growth of yeast under certain environmental challenges [23].</p><p>Extensive mutagenesis studies were done to identify functionally important residues of H69, especially within the loop region [24–29]. Mutations A1912G, A1919G, U/Ψ1917C were lethal to E. coli [24]. Furthermore, A1912, U1917, and A1919 were included in the functional sequences selected from randomized rRNA libraries [25]. Mutations at A1912 and A1919 did not affect pseudouridylation of H69 [26], but A1912G and A1919G caused compromised ribosome assembly, lower growth rates in vivo, and decreased in vitro protein synthesis activity of ribosomes [24], which correlated with reduced processivity of translation [30]. Similar effects were observed with the U/Ψ1917C mutant [24], which did not block pseudouridylation at positions 1911 and 1915 [26].</p><p>This work is focused on the biophysical properties of small RNAs representing the wild-type sequence and mutants (A1912G, U/Ψ1917C, and A1919G) of bacterial (E. coli) H69 with and without Ψ modifications. For the wild-type sequence and mutants with Ψ modifications, pH-dependent structural changes in the loop region are observed by NMR spectroscopy. These structural changes are not correlated with global RNA stability. However, at low pH conditions (5.5), a modest stabilizing effect from Ψ modification is observed in all cases, suggesting that pseudouridylation facilitates the folding of H69 and its mutants into energetically favorable conformational states. In some cases, the structural effects of the loop mutations propagate into the stem regions. These effects on structure are unique to each mutation, and influenced by Ψ modifications to different extents. These data suggest that Ψ modifications are capable of modulating the induced conformational changes of wild-type and mutant H69 RNAs, with some loop mutations causing additional conformational effects in the stem region, therefore explaining how loop mutations in conjunction with Ψ modifications result in functionally compromised ribosomes and decreased cell viability.</p><!><p>The wild-type sequences (referred to as UUU and ΨΨΨ for the unmodified and modified constructs, respectively) and the modified mutant RNA oligonucleotides (ΨΨΨ-A1912G, ΨΨC-Ψ1917C, and ΨΨΨ-A1919G) representing H69 were purchased from Dharmacon® (Thermo Scientific) and purified by HPLC as described previously [8]. Unmodified mutant RNA oligonucleotides (UUU-A1912G, UUC-U1917C, and UUU-A1919G) RNA oligonucleotides were synthesized by in vitro T7 RNA polymerase transcription, and treated with calf-intestinal phosphatase (CIP) to remove the 5′ triphosphate moiety on the transcripts [31]. Twenty percent (w/v) denaturing polyacrylamide gels were used to purify the RNA transcripts and CIP-treated samples. The final RNA oligonucleotides were characterized by MALDI-TOF mass spectrometry. The sequences are listed in the Supplementary Information (Table S1).</p><p>A Bruker AVANCE-AQS 700 MHz NMR spectrometer equipped with a TXI cryoprobe was employed to obtain the NMR spectra of all samples at 15 °C. Each RNA sample was dissolved in a buffer consisting of 250 μL of 10 mM phosphate (pH 5.5 or7.0), 50 mM KCl in 9:1 H2O/D2O to a final concentration of 50 μM. The water suppression was achieved using WATERGATE 5 with a gradient pulse sequence [32]. Digital Quadratic detection for 16,000 data points was used to acquire the one-dimensional proton (1D1H) NMR spectra, and 512 scans were collected to improve the signal-to-noise ratio.</p><p>Circular dichroism (CD) experiments were carried out on a Chirascan™ CD spectrometer from Applied Photophysics. A buffer consisting of 15 mM cacodylic acid (pH 5.5 or 7.0), 70 mM ammonium chloride, and 30 mM potassium chloride was employed. The extinction coefficients for pairs of unmodified and modified RNA oligonucleotides are as follows: 187,000 Lmol−1cm−1 (wild-type), 184,800 Lmol−1cm−1 (A1912G), 184,800 Lmol−1cm−1 (U/Ψ1917C), and 184,900 Lmol−1cm−1 (A1919G), at 260 nm [33]. The concentrations of the samples were about 13 μM. The CD spectra were collected from 220 to 320 nm at 23 °C in quadruplicate.</p><p>The thermal melting experiments were performed on a Beckman Coulter DU® 800 spectrophotometer equipped with temperature controller, multi-cell cuvette holder, and high-performance transport. The RNA samples (2 to 16 μM) were dissolved in a buffer containing 20 mM cacodylic acid (pH 5.5 or 7.0), 15 mM NaCl, and 0.5 mM Na2EDTA. The absorbance at 280 nm of each sample was measured from 10 to 95 °C. The concentration of each sample was calculated using the UV absorbance (λ = 260 nm) at 90 °C and corresponding extinction coefficient. Meltwin 3.5 was used to derive the thermodynamic parameters assuming a two-state model [34].</p><!><p>Crystal structures of complete ribosomes and individual subunits reveal that H69 adopts different conformational states during the translation process [35–39]. The conformational flexibility of H69 is believed to be important for optimal ribosomal activity. While exploring how H69 structure adapts under different solution environments and the role of Ψ modifications in this modulation, it was discovered through CD spectroscopy that pH variation could be used to promote Ψ-dependent conformational changes of this RNA motif [18]. Results from chemical probing experiments on ribosomal 50S subunits were consistent with those findings [20]. Modulation of the H69 conformational behavior by pseudouridylation is believed to result from a combination of enhanced base-stacking and hydrogen-bonding interactions unique to the Ψ-containing RNA, as determined by direct comparison of the solution structures of unmodified (UUU) and modified (ΨΨΨ) H69s [8]. Fully modified H69 containing N3-methylpseudouridine (m3Ψ) at position 1915 (referred to as Ψm3ΨΨ) was used in the initial pH-dependence studies, in which a two-state equilibrium model was proposed [18]. To confirm that pseudouridylation in the absence of methylation at position 1915 has the capability to modulate the pH-dependent structural changes, a modified H69 (ΨΨΨ) containing only the Ψ modifications at residues 1911, 1915, and 1917 was employed in CD and NMR experiments.</p><p>First, the CD spectra of ΨΨΨ at pH 7.0 and 5.5 (Supplementary Information, Fig. S1a) were compared. Both spectra have a peak maximum around 263 nm and a local peak minimum at 230 nm, with a crossover point around 240 nm, indicating a general A-form RNA structure (Table 1) [17–18]. Characteristic shifts of the CD profile are also observed when the pH is lowered from 7.0 to 5.5. An increase in molar ellipticity occurs from 235 to 251 nm with a concomitant decrease between 251 and 265 nm, which gives rise to an isosbestic point at 251 nm. The crossover point is shifted from 241 to 239 nm, and the peak maximum moves from 262 to 264 nm. The trends observed in this study are identical to those observed previously with Ψm3ΨΨ [18], suggesting that the ΨΨΨ construct undergoes pH-dependent structural changes similar to those of the fully modified RNA.</p><p>To explore more details of the pH-dependent structural changes of ΨΨΨ in comparison to the corresponding unmodified RNA, UUU, imino proton regions of the 1D1H NMR spectra of UUU and ΨΨΨ at pH 7.0 and 5.5 were compared (Supplementary Information, Fig. S1b–e). The stem imino proton resonances (low field) are not shifted by more than 0.1 ppm with different modification status (UUU and ΨΨΨ) or pH conditions [8, 40]. Compared to the spectra of UUU, an extra resonance corresponding to Ψ1911N1H (10.2 ppm) is apparent in the ΨΨΨ spectrum at pH 7.0 [8, 17], as well as a peak from Ψ1915N1H (10.6 ppm) at pH 5.5 (Supplementary Information Fig. S2a, b) [40]. The resonance patterns in the ΨΨΨ spectra at different pH conditions are very similar to those of Ψm3ΨΨ [18], indicating that Ψ modifications are capable of modulating the pH-dependent structural behavior of the H69 loop in the absence of m3Ψ1915.</p><p>Pseudouridine modifications have similar effects on the A1912G mutant H69 RNA. As shown in Fig. 3, the CD spectra of UUU-A1912 at pH 5.5 and 7.0 are almost completely overlapping. In contrast, an increase in molar ellipticity between 261 and 305 nm is observed with decreasing pH for the corresponding pseudouridylated mutant ΨΨΨ-A1912G. In this case, the peak maximum is red shifted from 262 (pH 7.0) to 266 nm (pH 5.5) (Table 1), compared to a 2 nm red shift for the pseudouridylated wild-type sequence (ΨΨΨ) with decreasing pH (Supplementary Information, Fig. S1a). These observations suggest that Ψ modifications are necessary for pH-dependent structural changes of the A1912G mutant, and the nucleotide composition also plays a role in this event. In support of this hypothesis, no change in chemical shifts or peak intensities is observed in the NMR spectra of UUU-A1912G when the pH condition is changed (Fig. 3c, d). In the NMR spectra of ΨΨΨ-A1912G (Fig. 3e, f), resonances corresponding to the ΨN1H protons (Supplementary Information Fig. S2c, d) are clearly observed, with peak patterns similar to those of ΨΨΨ (Supplementary Information, Fig. S1d, e). A difference worth noting is that the relative peak intensities of the ΨN1H protons compared to the stem imino proton resonances in the spectra of ΨΨΨ-A1912G are higher than those of the ΨΨΨ RNA. This observation suggests that G1912 may have specific effects on the Ψ- and pH-dependent structural changes in the H69 loop region that differ slightly from those in ΨΨΨ. More specifically, hydrogen-bonding interactions involving the corresponding ΨN1H protons appear to be enhanced. These pH-dependent differences in the CD and NMR spectra of ΨΨΨ and ΨΨΨ-A1912G might appear small; however, unique conformational states or alterations in H69 dynamics have the potential to impact the RNA function such that the A1912G mutation is lethal to E. coli [24, 30].</p><!><p>In previous studies, pseudouridylation at position 1911 was shown to have an impact on the thermodynamics, structure, and conformational behavior of H69 by establishing a stable Watson-Crick base pair with A1919 and a hydrogen-bonding interaction specific to RNAs with a Ψ1911 residue [8, 17, 41]. These results suggest that the H69 loop structure and conformational behavior are sensitive to the modification status of the loop-closing base pair, which may in turn impact ribosomal function.</p><p>Besides the modification status, nucleotide composition of the loop-closing base pair may also play a role in modulating the H69 loop conformational behavior. We therefore tested the A1919G mutants for pH-dependent structural changes. As shown in Fig. 4a and b, the global peak maxima for both UUU-A1919G and ΨΨΨ-A1919G are red shifted by 3 nm (Table 1) when the pH is lowered from 7.0 to 5.5, and there is a concomitant increase in molar ellipticity between the peak maxima and 310 nm. These observations suggest that the A1919G mutation plays a role in the pH-dependent structural changes of these RNAs, even in the absence of Ψ.</p><p>Structural effects of the A1919G mutation are further revealed in 1D1H NMR spectra (Fig. 4c–f). For UUU-A1919G (Fig. 4c, d), the same imino proton resonances as with UUU are observed (Supplementary Information, Fig. S1b, c), along with two extra imino proton peaks at 11.1 and 11.7 ppm. These peaks are assigned as U1911N3H and G1919N1H in a U·G wobble pair, based on NOE difference spectroscopy [17]. In the same region (Fig. 4e, f), two peaks from Ψ1911N3H and G1919N1H (Supplementary Information Fig. S3) are observed in the ΨΨΨ-A1919G spectra. Formation of a Ψ1911·G1919 wobble pair results in a 0.3 ppm downfield shift of the Ψ1911N1H resonance (Supplementary Information Fig. S2e–g) in both spectra (pH 7.0 and 5.5), compared to those from ΨΨΨ (Supplementary Information, Fig. S1d, e). Even though no difference in either chemical shift nor peak pattern is observed in the UUU-A1919G spectra at different pH conditions (Fig. 4c, d), the possibility should not be excluded that there is a pH-dependent structural change in the UUU-A1919G loop region, as suggested by the CD spectra (Fig. 4a). Structural effects in the UUU-A1919G loop (e.g., base stacking or solvent exposure of nucleotide bases) do not necessarily alter hydrogen-bonding interactions. In contrast, when the pH is decreased from 7.0 to 5.5, changes in the chemical shifts of the Ψ1911N3H and G1919N1H imino protons and extra imino proton resonance in the loop region are observed for ΨΨΨ-A1919G (Fig. 4e, f). These spectral changes reveal that the presence of Ψ extends the pH-dependent structural changes of A1919G beyond the loop single-stranded residues to include the loop-closing base pair Ψ1911·G1919. The mutated 1919 residue (G) and Ψ modifications appear to work synergistically to modulate the pH-dependent structural changes of ΨΨΨ-A1919G, which differ from ΨΨΨ. Such changes could be detrimental to the translation steps that require proper conformational-state adjustments, especially in the single-stranded loop and loop-closing base pair of H69.</p><!><p>The U1917 residue of H69 shows 100% conservation across phylogeny [10], and Ψ modification at this position has been detected in all species whose rRNA pseudouridylation sites are currently known [11]. The important functional role of Ψ1917 is also revealed in ribosome assays. A Ψ1917C mutation was shown to decrease cellular polysome levels and cause a strong growth defect [24], even though Ψ modifications at sites other than 1917 were unlikely affected [26]. The importance of Ψ1917 can be attributed to its structural roles. As observed in NMR solution studies on UUU and ΨΨΨ, Ψ1917 mediates base-stacking interactions in the 3′ half of the ΨΨΨ loop region [8]. Residue Ψ1917 is also suggested to participate in formation of the B2a intersubunit bridge by hydrogen bonding to A1912 in complete ribosomes [2]. To reveal the effects of U/Ψ1917C on H69 structure and pH-dependent structural changes, CD and NMR experiments were employed.</p><p>As shown in Fig. 5a and b, a decrease in molar ellipticity between 275 and 310 nm is observed in both unmodified (UUC-U1917C) and modified (ΨΨC-Ψ1917C) mutant H69 RNAs, when the pH is lowered from 7.0 to 5.5. No shift in the peak maxima is demonstrated in either UUC-U1917C or ΨΨC-Ψ1917C. Similar spectral features shared by the unmodified and modified U/Ψ1917C mutants, along with differences identified between ΨΨC-Ψ1917C and ΨΨΨ spectra (Supplementary Information, Fig. S1a), indicate that nucleotide composition of the loop may play a dominant role in determining the pH-dependent structural changes in U/Ψ1917C H69 RNAs.</p><p>The hypothesis that nucleotide composition at position 1917 plays a role in mediating H69 structural changes is supported by NMR data (Fig. 5c–f). At pH 7.0, all five signature imino proton resonances from the H69 stem region are observed (Fig. 5c, e). In contrast, only four sharp peaks are visible when the pH is lowered to 5.5 (Fig. 5d, f). These observations suggest that pH-dependent structural changes of the U/Ψ1917C mutants are extended into the stem regions, and the G1910·C1920 stem-closing base pair is destabilized at pH 5.5. Accompanying the loss of G1910·C1920 base pairing, the proton resonance corresponding to G1921N1H is downfield shifted by 0.1 ppm, and the peak intensities are decreased. This structural change at lower pH common to both U/Ψ1917C mutants may account for the similar trend in pH-induced changes of the CD spectra (Fig. 5a, b). The Ψ1917C mutation also causes changes in the H69 structure at neutral pH. The peak most likely corresponding to Ψ1911N1H in the ΨΨC-Ψ1917C spectrum (Fig. 5e) is upfield shifted by 0.1 ppm compared to that of ΨΨΨ (Supplementary Information, Fig. S1a). The resonance pattern in the upfield (ΨN1H) imino proton region of ΨΨC-Ψ1917C at pH 5.5 is also very different from that of ΨΨΨ. Since Ψ1917 and Ψ1911 do not have direct interactions in either the solution structure of ΨΨΨ [8] or X-ray crystal structures of complete ribosomes [36], the difference in spectral features between ΨΨC-Ψ1917C and ΨΨΨ at both pH conditions potentially results from an altered structure of the ΨΨC-Ψ1917C loop region caused by the mutation. Long-range effects of the U/Ψ1917C mutation on the stem-region structure, which are independent of the Ψ modifications, suggest stem-loop crosstalk [19, 41]. Mutations at key loop residues, such as 1917, involved in mediating this crosstalk, may also be responsible for monitoring interactions with the P-site tRNA and other translational factors, e.g., A-site tRNA or release factors (RFs), in complete ribosomes [27–29, 42], and therefore responsible for lethality in E. coli.</p><!><p>A change in stability of an RNA molecule may be correlated to either a structural adaptation or altered conformational flexibility, although not all conformational changes lead to altered RNA stability because of the enthalpy-entropy compensation [43]. In a previous study with the Ψm3ΨΨ construct, the RNA was stabilized by −0.5 kcal/mol when the pH was lowered from 7.0 to 5.5, which was accompanied by CD spectral changes [18]. In thermal melting experiments with ΨΨΨ, the difference in stability at the two pH values is only −0.2 kcal/mol (Table 2), negligible when the expected standard error (0.2 kcal/mol) is considered. As discussed in Section 3.1 (CD experiments), the conformational changes of ΨΨΨ with pH variation closely resemble those of Ψm3ΨΨ. The minor differences could be attributed to N3-methylation on Ψ1915 in Ψm3ΨΨ, contributing to the modestly enhanced stabilizing effects [18]. Taken together, these results suggest that the Ψ modifications are responsible for the pH-dependent structural changes of ΨΨΨ, but have minimal or no effects on the global stability of ΨΨΨ at different pH conditions. Similar small differences in Gibbs free energy between unmodified and modified RNAs are observed with the H69 mutants (Table 2), indicating that Ψ modifications confer conformational effects without altering the global stability as the solution pH is changed. The thermodynamic parameters for the mutants are listed in the Supplementary Information (Table S1).</p><p>Even though modest or negligible differences in the thermal stabilities of the RNA constructs are observed with varying pH conditions (ΔG°37pH 5.5 − ΔG°37pH 7.0), Ψ modifications are shown to slightly stabilize the global folding of H69 and mutants at pH 5.5 compared to pH 7.0 (Table 2). A ΔΔG°37 (ΔG°37ΨΨΨ − ΔG°37UUU) value of −0.5 kcal/mol is observed for H69 (Supplementary Information Fig. S4a) and mutants A1912G and U/Ψ1917C. This value matches that of a previous thermal melting study with Ψm3ΨΨ [18]. These results suggest that Ψ modifications dominate the stabilizing effects of certain conformational states of H69 and the loop mutants. The hypothesis is supported by fluorescence studies employing modified H69 with 2-aminopurine (2AP) at position 1913, in which the fluorescence intensity from 2AP in ΨΨΨ decreased at pH 5.5 relative to 7.0 due to enhanced base stacking of 2AP1913 with neighboring bases [19]. A similar change was not observed with 2AP-containing UUU, suggesting that the modification plays a role in the conformational switch. As discussed previously, A1913 plays important roles in establishing the intersubunit bridge B2a [2, 5]. Therefore, the conformational response of A1913 in Ψ- and pH-dependent structural changes of H69 may contribute to the growth advantage of living organisms [23].</p><p>Stabilizing effects of pseudouridylation are also observed for the A1919G mutant, with ΔΔG°37 (ΔG°37ΨΨΨ − ΔG°37UUU) values of −0.9 kcal/mol (pH 5.5, Supplementary Information Fig. S4b) and −0.4 kcal/mol (pH 7.0, Supplementary Information Fig. S4c), respectively (Table 2). The larger ΔΔG°37 values for A1919G compared to other H69 mutants suggest that the sequence alteration in combination with Ψ modifications impacts the H69 structure. Furthermore, the largest free energy difference is observed at pH 5.5, implying that G1919 plays a different role than A1919 in modulating H69 conformational states. Position 1919 is unique in mediating the structural effects of Ψ modifications. In the NMR structure of ΨΨΨ, A1919 forms a Watson-Crick base pair with Ψ1911, as well as mediates base-stacking interactions on the 3′ side of the H69 loop from Ψ1915 to A1919, which is promoted by modified nucleotides Ψ1915 and Ψ1917 [8]. Similar to A1919 in ΨΨΨ, G1919 may also mediate structural effects of the Ψ modifications within ΨΨΨ-A1919G. As shown by NMR spectroscopy, pH-dependent structural changes of ΨΨΨ-A1919G involve the formation, as well as changes in imino proton chemical shifts, of a Ψ1911·G1919 wobble pair. The synergistic modulations of G1919 and Ψ modifications, resulting from the sequence and structural contexts in which they reside, appear to cause a more significant impact on conformational-state modulation of H69, which in turn elicit biological effects that differ from ΨΨΨ or the other H69 mutants.</p><!><p>Helix 69 plays important roles in the assembly of ribosomes and protein synthesis process [6]. Conservation of Ψ modifications at positions 1911, 1915, and 1917 are suggested to be correlated to their functions in modulating the structures and conformational behaviors of H69 [8, 17–21, 41, 44]. Results from the work presented here reveal that both Ψ modifications and the lethal mutations, A1919G and U/Ψ1917C, are capable of modulating pH-dependent structural changes of H69 independently. In the modified H69 lethal mutants, the mutated residue may also work synergistically with the Ψ modifications to introduce additional structural effects. For example, A1912G appears to alter hydrogen-bonding interactions at lower pH, but only in the presence of loop Ψ residues. Within a different sequence and structural context, the same mutation of A to G has an altered role. The A1919G mutation leads to formation of a loop-closing Ψ1911 G1919·wobble pair that appears to modulate the H69 loop structure with pH and modification status, but does not impact the stem structure. In contrast, the Ψ1917C mutant expands the pH-dependent structural changes from the loop region into the stem G1910·C1920 base pair. These pH-dependent structural changes do not affect the global thermal stability of H69 RNAs, but modest stabilizing effects of Ψ modifications are consistently observed at pH 5.5. These results suggest that stabilization of certain conformational states of H69 is inherent to Ψ modifications, possibly through alterations in hydrogen-bonding and/or base stacking interactions involving Ψ residues. As with the wild-type H69 construct and full-length 23S rRNA, the changes in conformational states of the RNA can be induced by lowering the solution pH to 5.5. These changes occur due to protonation of nucleotides that have been yet to be identified, but are likely to be conserved within the loop of H69. Furthermore, NMR spectral evidence of the loop-stem crosstalk within H69 is directly observed in pH-dependent structural changes of U/Ψ1917C, consistent with suggestions in the literature that H69 may not simply form simultaneous contacts with the A- or P-site tRNAs [38] or RFs [39, 45], but instead plays a role in coordinating the activities of the tRNAs and RFs [27–29, 42].</p><p>Given the essential functions of H69 in translation and the multiple conformational states observed in structure studies, alterations in the structural changes of H69 by mutations may lead to the phenotypes observed in the functional analyses. Knowledge gained in this project from pH-induced structural changes of H69 extends our understanding of the effects of post-translational modifications and interplay with mutations on the structure and conformational behavior of these RNAs. In addition, modulation of the structure and pH-dependent structural changes of H69 by ligand binding may serve as an effective pathway to design and screen antibiotics targeting bacterial protein synthesis.</p>

PubMed Author Manuscript

Lignin-Rich PHWE Hemicellulose Extracts Responsible for Extended Emulsion Stabilization

Wood hemicelluloses have an excellent capacity to form and stabilize oil-in-water emulsions. Galactoglucomannans (GGM) from spruce and glucuronoxylans (GX) from birch provide multifunctional protection against physical breakdown and lipid oxidation in emulsions. Phenolic residues, coextracted with hemicelluloses using the pressurized hot water (PHWE) process, seem to further enhance emulsion stability. According to hypothesis, phenolic residues associated with hemicelluloses deliver and anchor hemicelluloses at the emulsion interface. This study is the first to characterize the structure of the phenolic residues in both GGM- and GX-rich wood extracts and their role in the stabilization of emulsions. PHWE GGM and GX were fractionated by centrifugation to obtain concentrated phenolic residues as one fraction (GGM-phe and GX-phe) and partially purified hemicelluloses as the other fraction (GGM-pur and GX-pur). To evaluate the role of each fraction in terms of physical and oxidative stabilization, rapeseed oil-in-water emulsions were prepared using GGM, GX, GGM-pur, and GX-pur as stabilizers. Changes in droplet-size distribution and peroxide values were measured during a 3-month accelerated storage test. The results for fresh emulsions indicated that the phenolic-rich fractions in hemicelluloses take part in the formation of emulsions. Furthermore, results from the accelerated storage test indicated that phenolic structures improve the long-term physical stability of emulsions. According to measured peroxide values, all hemicelluloses examined inhibited lipid oxidation in emulsions, GX being the most effective. This indicates that phenolic residues associated with hemicelluloses act as antioxidants in emulsions. According to chemical characterization using complementary methods, the phenolic fractions, GGM-phe and GX-phe, were composed mainly of lignin. Furthermore, the total carbohydrate content of the phenolic fractions was clearly lower compared to the starting hemicelluloses GGM and GX, and the purified fractions GGM-pur and GX-pur. Apparently, the phenolic structures were enriched in the GGM-phe and GX-phe fractions, which was confirmed by NMR spectroscopy as well as by other characterization methods. The frequency of the main bonding pattern in lignins, the β-O-4 structure, was clearly very high, suggesting that extracted lignin remains in native form. Furthermore, the lignin carbohydrate complex of γ-ester type was found, which could explain the excellent stabilizing properties of PHWE hemicelluloses in emulsions.

lignin-rich_phwe_hemicellulose_extracts_responsible_for_extended_emulsion_stabilization

Introduction<!>Hemicelluloses<!>Reagents<!>Centrifugation of Hemicelluloses<!>Purification of Rapeseed Oil<!>Preparation of Emulsions<!>Accelerated Storage Test<!>Droplet-Size Distribution<!>Determination of Peroxide Value<!>Quantitative Analysis of Carbohydrates<!>Structural Characterization of Starting Materials GGM and GX (Non-acetylated) and Phenolic Samples GGM-Phe and GX-Phe (Acetylated) by 2D HSQC and HSQC-TOCSY NMR, and Evaluation of Diffusion Constants by 2D DOSY NMR (Acetylated or Partially Acetylated Samples)<!>Analysis of Molar Masses<!>Determination of Phenolic Content by Pyrolysis GC-MS<!>Quantitative Determination of Extractable Phenolic Residues by UHPLC-DAD-FLD and Identification by LC-MS<!>Fractionation of Hemicelluloses and Preparation of Emulsions<!>Physical Properties and Stability of Emulsions<!><!>Physical Properties and Stability of Emulsions<!><!>Oxidative Stability of Emulsions<!><!>Oxidative Stability of Emulsions<!>Carbohydrate Composition of Starting Hemicelluloses and Fractionated Materials From Centrifugation<!><!>Carbohydrate Composition of Starting Hemicelluloses and Fractionated Materials From Centrifugation<!>Structural Characterization of Starting Hemicelluloses GGM and GX and Precipitated Fractions GGM-Phe and GX-Phe by 2D HSQC NMR Spectroscopy<!><!>Structural Characterization of Starting Hemicelluloses GGM and GX and Precipitated Fractions GGM-Phe and GX-Phe by 2D HSQC NMR Spectroscopy<!>Molar Mass Analysis of Starting Hemicelluloses GGM and GX, Purified Fractions GGM-Pur and GX-Pur, and Precipitated Fractions GGM-Phe and GX-Phe<!><!>Molar Mass Analysis of Starting Hemicelluloses GGM and GX, Purified Fractions GGM-Pur and GX-Pur, and Precipitated Fractions GGM-Phe and GX-Phe<!>Evaluation of Diffusion Constants by DOSY NMR (Acetylated or Partially Acetylated Samples)<!><!>Evaluation of Diffusion Constants by DOSY NMR (Acetylated or Partially Acetylated Samples)<!>Analysis of Phenolic Contents of Starting Hemicelluloses and Phenolic Fractions GGM-Phe and GX-Phe Using Pyrolysis GC/MS<!><!>Analysis of Phenolic Contents of Starting Hemicelluloses and Phenolic Fractions GGM-Phe and GX-Phe Using Pyrolysis GC/MS<!>Analysis and Quantitation of Small Extractable Phenolic Compounds From Phenolic Fractions GGM-Phe and GX-Phe—Vanillin and Syringaldehyde as Indicators in Lignin Participating in Formation of Emulsions<!><!>Analysis and Quantitation of Small Extractable Phenolic Compounds From Phenolic Fractions GGM-Phe and GX-Phe—Vanillin and Syringaldehyde as Indicators in Lignin Participating in Formation of Emulsions<!>Properties of Hemicelluloses Affecting the Physical Properties and Stability of Emulsions<!>Conclusions<!>Data Availability Statement<!>Author Contributions<!>Conflict of Interest

<p>The sustainable use of natural resources requires the development of new functional materials from side streams of industrial processes. Woody material is renewable biomass, which contains unexploited components that can be used for valorized products. The main components of wood are cellulose (33–51%), lignin (21–32%), and hemicelluloses (23–31%), and it contains minor amounts of other compounds, such as extractives and minerals (Fengel and Wegener, 1983; Sjöström, 1993). The processes of pulp mills are optimized and efficient for the production of cellulosic fibers from wood, which are still needed for many traditional products, such as paper and board materials (van Heiningen, 2006). New technologies are also under development for cellulosic fibers, for example as textile fibers, as reinforcing structures in composite materials, and in the form of nanocellulose (Faruk et al., 2012; Kim et al., 2015; Sixta et al., 2015). However, there is also a general need and will to produce valorized products and processes for the currently underutilized parts of wood as well as other sources of biomass, namely, hemicelluloses and lignin (Faruk et al., 2012; Sainio et al., 2013).</p><p>Hemicelluloses can be extracted from biomass prior to other processing steps by using pressurized hot water (Kilpeläinen et al., 2014). During pressurized hot water extraction (PHWE), hemicelluloses and sulfur-free lignin are released from woody material at temperatures of 160–170°C. Hemicelluloses and lignin are partially separated, and extracts enriched with hemicelluloses can be further purified from lignin using other methods, such as ultrafiltration or precipitation with ethanol, if necessary (Bhattarai et al., 2019). We have recently developed a method of using differing centrifugal forces to separate hemicellulose- and lignin-rich fractions of PHWE extracts (Valoppi et al., 2019a).</p><p>The chemical structure of wood components reflects their material properties. The role of hemicelluloses is to provide flexibility to the cell wall. In contrast to cellulose, hemicelluloses and lignin are heterogeneous materials with a complex chemical structure, which is further dependent on the type of wood (Sjöström, 1993). Thus, the main hemicelluloses in softwoods and hardwoods are glucomannans and glucuronoxylans, respectively, although softwoods also contain minor amounts of glucuronoxylans and hardwoods have some glucomannans (Sjöström, 1993).</p><p>In spruce, the main hemicelluloses are galactoglucomannans (GGM) (Lundqvist et al., 2002; Hannuksela and Hervé du Penhoat, 2004; Xu et al., 2007). The backbone of GGM consists of (1 → 4)-linked β-D-mannopyranosyl (Manp) units and (1 → 4)-linked β-D-glucopyranosyl (Glcp), and some α-galactopyranosyl (Galp) units are (1 → 6)-linked to the backbone as single-unit side groups. The proportions of the carbohydrates in GGM are 4:1:0.5 for Manp:Glcp:Galp (Willför et al., 2003). In birch, the main hemicelluloses are 4-O-methylglucuronoxylans (GX) (Teleman et al., 2002). The backbone of GX consists of (1 → 4)-linked β-D-xylopyranosyl (Xylp) units, to which α-4-O-methylglucuronic (MeGlcA) substituents are (1 → 2)-linked. The proportions are 0.5–1:10 for MeGlcA:Xylp (Teleman et al., 2002). Both GGM and GX contain O-acetyl (OAc) groups at C-2 and/or C-3 positions of Manp and Xylp units. The reported degrees of acetylation are about 15–20% for spruce GGM and 25% for birch GX (Lundqvist et al., 2002; Xu et al., 2007; Du et al., 2013).</p><p>Lignin is an aromatic polyphenol composed of phenylpropane units (Sjöström, 1993; Boerjan et al., 2003; Vanholme et al., 2010). The role of lignin in plants is to provide mechanical strength, enable efficient liquid transportation, and provide protection against microbial attack (Fengel and Wegener, 1983; Boudet, 2000). The building blocks of lignin are the monolignols p-coumaryl alcohol (minor component of wood plants), coniferyl alcohol (major component of softwoods), and sinapyl alcohol (major component of hardwoods). During lignin biosynthesis, the monolignols are oxidized to form phenoxy radicals, which leads to radical polymerization and the formation of different types of bonding patterns. The most frequent structure in lignin is the β-aryl ether type (β-O-4), which is prone to react with different chemicals, leading to degradation of lignin, for example in kraft pulping process (Sjöström, 1993). The other frequent linkage types include phenyl coumaran type (β-5) and resinol type (β-β).</p><p>Hemicelluloses and lignin are closely associated in the wood cell wall. In addition to non-covalent interactions, the presence of covalently bound lignin-carbohydrate complexes (LCCs) was suggested decades ago (Fengel and Wegener, 1983). However, even today, unequivocal proof of different types of LCCs is lacking, because these structures have low frequencies and because structural modifications may occur during their isolation for characterization (Giummarella et al., 2019). Recently, isolation and characterization of the α-ether type LCC was successfully performed (Nishimura et al., 2018) and was made possible by the development of a methodology for enriching LCCs and characterizing by nuclear magnetic resonance (NMR) spectroscopy. To date, the structure of phenolic residues and their associations with hemicelluloses in PHWE extracts are largely unknown.</p><p>Hemicellulose- and phenolic-rich PHWE extracts exhibit excellent emulsifying ability and physical emulsion stabilization capacity, which gives them great potential in both bulky and specialized industrial applications, such as food, paints, cosmetics, and pharmaceuticals (Mikkonen et al., 2016, 2019; Valoppi et al., 2019b). Furthermore, PHWE hemicelluloses containing phenolic co-components offer excellent oxidative stability in rapeseed oil-in-water emulsions (Lehtonen et al., 2016, 2018). For comparison, emulsions prepared from rapeseed oil, which has been purified from natural antioxidants of the oil, tocopherols, and stabilized with Tween 20 or gum Arabic, are oxidized in a few days (Heinonen et al., 1997; Lehtonen et al., 2016). The presence of tocopherols retards the oxidation, which is, however, modest compared to PHWE GGM, which improves the oxidative stability for several weeks in an accelerated storage test (Lehtonen et al., 2018). The strong, previously developed hypothesis was that phenolic residues associated with hemicelluloses would be responsible for improved physical and oxidative stability, but their exact chemical structure was still unclear (Lehtonen et al., 2018).</p><p>The aim of this study was to reveal the structure of phenolic residues responsible for the efficient emulsion stabilization capacity of PHWE spruce GGM and birch GX. In this study, aqueous GGM and GX were centrifuged in a parallel manner to separate hemicellulose-rich supernatants and lignin-rich pellets. For the first time, the fractions were characterized in detail using complementary chemical analyses to investigate the role of the lignin-rich fraction in the physical and oxidative stability of emulsions. Furthermore, characterization of both GGM and GX using various methods enabled a comparison of softwood and hardwood hemicelluloses and different analytical methods. The results explain the structural elements in hemicellulose-rich wood extracts that are responsible for their excellent performance in emulsions.</p><!><p>A pressurized hot water flow-through extraction (PHWE) system was used to obtain GGM-rich extract from spruce and GX-rich extract from birch (Kilpeläinen et al., 2014). Spruce sawdust (from Herralan Saha, Finland, 96.9 kg, 43.5 kg on dry basis) was extracted at 170°C for 60 min at a rate of 20 l min−1, and 1,000 l of the extract was collected. Birch sawdust (from Haka-Wood, Finland, 103.7 kg, 54.2 kg on dry basis) was extracted at 170°C for 60 min at a rate of 20 l min−1, and 700 l of the extract was collected. Both the spruce and birch extracts were ultrafiltrated to obtain concentrated extracts, as described previously (Bhattarai et al., 2019). The concentrated extracts were finally spray dried to powdered form using a Buchi Mini Spray Dryer B-290 (Buchi, Switzerland), which has the evaporation capacity of 1 l h−1 for water. The conditions for the spray drying used were as follows: inlet temperature 170°C, outlet temperature 65°C, and drying air flow rate 667 l h−1. The moisture contents of the materials were 6.0% for GGM and 3.9% for GX after storage in the dark at room temperature.</p><!><p>Rapeseed oil (Keiju, Bunge Finland Ltd, Raisio, Finland) was purchased in a local supermarket. Monosaccharides used for the analysis of carbohydrate content were L(+)-arabinose (Ara), D(+)-xylose (Xyl), D(+)-galactose (Gal), D(+)-glucose (Glc), D(+)-mannose (Man) from Merck (Darmstadt, Germany), L(+)rhamnose monohydrate (Rha) from Sigma (St. Louis, MO, USA), D(+)-sorbitol, D(+)-galacturonic acid monohydrate (GalA) from Fluka (St. Louis, MO, USA), and D(+)-glucuronic acid sodium salt monohydrate (GlcA) from Aldrich. The reagents used for silylation were bis(trimethylsilyl)trifluoroacetamide (BSTFA) from Merck (Darmstadt, Germany) and trimethylsilyl chloride (TMSCl) from Fluka (St. Louis, USA). Pullulan kit standards (Z-POS-pulkith) were used for size-exclusion chromatography (SEC; Postnova Analytics, Landsberg am Lech, Germany). Vanillin (Sigma-Aldrich) and syringaldehyde (Aldrich) were used for the quantification of phenols.</p><p>The solvents used for NMR analysis were D2O and d6-DMSO, which were purchased from Eurisotop (Saint-Aubin, France). All other solvents used were HPLC or LC-MS grade. Milli-RO water was used in centrifugal separation, and Milli-Q was used as a solvent in chemical analysis.</p><!><p>The 10% solutions of GGM and GX were dissolved in Milli-RO water and stirred for 2 h at room temperature (total amount 150 ml). The solutions were then centrifuged at 18,677 g at room temperature for 20 min. The supernatants (GGM-pur and GX-pur) and pellets (GGM-phe and GX-phe) were collected and freeze-dried separately. The recovered yields were 91.0% for GGM-pur, 4.6% for GGM-phe, 86.6% for GX-pur, and 5.8% for GX-phe (based on dry material).</p><!><p>Rapeseed oil (Keiju, Bunge Finland Ltd, Raisio, Finland) was purchased from a supermarket and stripped of tocopherols using a previously described method (Lampi et al., 1999). The composition has been determined in earlier publications (for example Lehtonen et al., 2016, 2018).</p><!><p>The amount of oil and emulsifier, and the applied buffer and pH used, were based on optimizations performed in previous studies (Mikkonen et al., 2016). Emulsions containing hemicelluloses (1 w-%), GGM or GX, or their supernatants, GGM-pur and GX-pur, in 25 mM Na-citrate buffer, pH 4.5, and stripped rapeseed oil (5 w–%) were prepared by high-pressure homogenization using a previously described method with some modifications (Lehtonen et al., 2016). The total weight of each emulsion was 80 g. First, carbohydrate was dissolved in buffer by stirring overnight at room temperature. After the addition of oil, the coarse emulsion was prepared by stirring the resulting mixture with Ultra-Turrax (T25 basic, IKA, Staufen, Germany) at 22,000 rpm for 2 min. The mixture was further homogenized by passing it continuously through a high-pressure homogenizer for 32 s at a pressure of 800 bar (Microfluidizer 110Y, Microfluidics, Westwood, MA, USA). The homogenizer was configured with 75 μm Y-type F20Y and 200 μm Z-type H30Z chambers in series.</p><!><p>For the accelerated storage test, emulsions were stored in glass bottles (100 ml) at 40°C in the dark for 3 months. For all the determinations, emulsions were gently mixed by turning their containers upside down 10 times before sampling. The properties of emulsions were monitored on the day of preparation, after 1 and 2 weeks of preparation, and after that approximately every 2 weeks. The properties, which were monitored during the storage test, were droplet-size distribution and peroxide value. At the end of the storage period, optical microscopy was used to visualize the morphology of emulsions (AxioScope A1, Carl Zeiss Inc., 203 Oberkochen, Germany). For microscopic imaging, the 100x objective was used, with a Zeiss Phase Contrast condenser with a Ph3 port.</p><!><p>The droplet-size distribution was determined by static light scattering technique using a Mastersizer Hydro 3000 (Malvern Instruments Ltd, Worcestershire, UK). The refractive indexes used were 1.33 for water and 1.47 for rapeseed oil (Rumble, 2018–2019). The emulsions were added directly into the dispersion accessory, which allowed dilution to avoid multiple scattering effects. The rotor speed during measurement was 2,400 rpm. Each sample was measured three times.</p><!><p>Peroxide values (PVs), as an indicator of primary oxidation of emulsions, were determined by a previously reported method, in which lipids are first released and extracted and then analyzed using the ferric thiocyanate method (Lehtonen et al., 2011, 2016). Analytical samples of extracted lipids were prepared in duplicate, and from both samples of extracted lipids, two samples were withdrawn for the determination of PVs. Thus, the results were calculated as averages and standard deviations of four measured values.</p><!><p>The carbohydrate content of GGM and GX and of their centrifuged fractions, GGM-pur, GGM-phe, GX-pur, and GX-phe, were analyzed by GC using the acid methanolysis and silylation method described previously (Sundberg et al., 1996). The instrumental details of the analysis were described previously as well (Chong et al., 2013). External calibration of five levels of concentration was used to calculate the amount of each monosaccharide in the samples. Methyl glucuronic acid (MeGlcA) was quantified based on the two major signals and the D-glucuronic acid standard (Chong et al., 2013). All samples were analyzed in triplicate (n = 3).</p><!><p>For structural characterization of the starting hemicelluloses, GGM and GX, and the fractions enriched with phenolic compounds, GGM-phe and GX-phe, 2D Heteronuclear Single Quantum Coherence (HSQC) spectra, 2D Heteronuclear Single Quantum Coherence—Total Correlation SpectroscopY (HSQC-TOCSY) spectra and 2D Diffusion Ordered SpectroscopY (DOSY) data were acquired using a Varian Unity Inova 500-MHz spectrometer equipped with a 5-mm pulsed-field-gradient triple resonance probehead (1H, 13C, 15N) capable of delivering z-gradient amplitudes up to 20 G/cm. The pulse sequences used in this study were readily available in Varian VNMR 6.1C spectrometer operating software.</p><p>All samples were analyzed in DMSO-d6 at 27°C. Of the starting hemicelluloses, GGM and GX, 30 mg was first dispersed in D2O (0.7 ml), freeze-dried, and finally dissolved in DMSO-d6 (0.7 ml). Of the phenolic fractions, GGM-phe and GX-phe, 40 mg was acetylated in pyridine/acetic anhydride (1:1), and the remaining reagents/solvents were removed by evaporating the mixture with ethanol twice, with toluene four times, and finally with chloroform under reduced pressure twice. The acetylated sample was then dissolved in DMSO-d6 (0.7 ml) and analyzed.</p><p>All HSQC spectra were recorded with a standard, phase-sensitive, gradient-selected HSQC sequence using echo-antiecho acquisition mode in the indirectly detected dimension. The hard, rectangular 90° pulse widths were 6.7 and 11.5 μs for 1H and 13C, respectively. The spectral width was 5,573 Hz for 1H (carrier at 5.47 ppm) and 25,133 Hz for 13C (carrier at 90.01 ppm). The relaxation delay was 1 s, and the acquisition time was 0.128 s. Experiments were acquired using 64 steady-state scans, 64 transients, and a data matrix size of 713 (1H, complex points) × 200 (13C, complex points). The data matrices were apodized by a Gaussian function (gf = 0.032) in 1H-dimension and a Gaussian function (gf = 0.004) in 13C-dimension and zero-filled up to 1,024 (1H, complex points) x 1,024 (13C, complex points) prior to Fourier transformation.</p><p>For HSQC-TOCSY, a standard phase-sensitive, gradient-selected (echo-antiecho) pulse sequence was applied. The hard, rectangular 90° pulse widths were 6.7 and 11.5 μs for 1H and 13C, respectively. The spectral width was 5,573 Hz for 1H (carrier at 5.47 ppm) and 25,133 Hz for 13C (carrier at 90.01 ppm). Relaxation delay was 1 s and acquisition time was 0.184 s. The TOCSY mixing was performed using the windowed MLEV-17 spin-lock scheme (Griesinger et al., 1988) to suppress possible ROESY correlations. TOCSY mixing was applied for 100 ms at an RF power of 7.9 kHz. Experiments were acquired using 64 steady-state scans, 64 transients, and a data matrix size of 1,024 (1H, complex points) × 200 (13C, complex points). The data matrices were apodized by a Gaussian function (gf = 0.037) in 1H-dimension and a Gaussian function (gf = 0.003) in 13C-dimension and zero-filled up to 1,024 (1H, complex points) × 1024 (13C, complex points) prior to Fourier transformation.</p><p>2D DOSY spectra were measured using Bipolar Pulse Pair Stimulated Echo sequence with convection compensation (BPPSTE-cc) (Wu et al., 1995; Jerschow and Müller, 1997). The spectral width of 8,000 Hz in 1H-dimension was covered by the acquired 8,002 complex points, resulting in 1-s acquisition time. The relaxation delay was 1 s. In order to map diffusion coefficients (Dc), 20 spectra were acquired with increasing amplitudes of rectangular diffusion gradient pulses (from 0.5 to 20 G/cm). The diffusion gradient pulse duration was 2 ms, and the diffusion delay was 600 ms. The eddy-current recovery delay was 150 μs. A total of four steady-state scans and 32 transients were used to collect all 20 of these spectra. The free induction decays (FIDs) were apodized using an exponential weighting function (10-Hz line broadening) and zero-filled up to 8,192 complex points before the Fourier transform. The 2D DOSY plots were calculated using the dosy macro (a monoexponential fit on the peak tops) incorporated into VNMR 6.1C software. The final size of the diffusion dimension in 2D DOSY was 256 data points. The diffusion coefficients of the macromolecule and the residual DMSO signal were estimated from each DOSY spectrum (see Figures S2A–D); the horizontal line shows the estimated values for Dc(GGM), Dc(GGM-phe), Dc(GX), and Dc(GX-phe). Moreover, the Dc(DMSO) value for each sample is shown. In order to compensate the effects of possible sample viscosity differences in the diffusion coefficient results, the estimated diffusion coefficients of the macromolecules were corrected using the measured diffusion coefficient values for residual DMSO signal of the solvent (Kavakka et al., 2009). In the correction procedure, the Dc(DMSO) value in the GGM sample was selected as the reference [Dc(DMSOref)], and the DOSY results for each sample were multiplied by Dc(DMSOref)/Dc(DMSO); that is, after the correction, Dc(DMSO) was the same for all four 2D DOSY spectra.</p><!><p>Molar masses of GGM, GX, and their centrifuged fractions were analyzed by SEC (GPCmax, Viscotek Corp., Houston, TX, USA). The instrumental details were described in a previous study (Pitkänen et al., 2011). The samples were dissolved in 0.01 M LiBr in DMSO overnight to a concentration of 10 mg ml−1 and filtered through a 0.45 μm syringe filter (GHP Acrodisc 13 mm, Pall Corp., Ann Arbor, MI, USA). The volume of the sample injected was 100 μl. DMSO, containing 0.01 M LiBr, was used as the eluent, with a flow rate of 0.8 ml min−1. The molar mass of samples was estimated using pullulan standards for calibration (342, 1,320, 5,900, 11,800, and 22,800 Da). The elution data were processed using the OmniSEC 4.5 software (Viscotek Corp.).</p><!><p>Pyrolysis of the starting hemicellulose samples GGM and GX and the centrifuged pellets GGM-phe and GX-phe was performed using a foil pulse-type Pyrola 2000 MultiMatic pyrolyzer (Pyrol AB, Lund, Sweden). The pyrolysis unit was connected to an Agilent GC model 7890B equipped with an HP5-MS column (25 m x 0.20 mm, film thickness 0.33 μm), coupled with an Agilent 5977B quadrupole-MSD with EI ionization (Agilent Technologies, Santa Clara, CA, USA). Approximately 100 μg of dry sample and a drop of acetone was applied to the Pt filament and pyrolyzed at 600°C for 2 s. Conditions for the GC analysis were as follows: gas flow (helium) 0.8 ml min−1, injector temperature 300°C, split 1:20; the column oven temperature was 50°C for 1 min, then heated with a rate of 8°C min−1 to 320°C, which was maintained for 5 min; the transfer line temperature was 250°C.</p><p>Compounds were identified by comparing acquired spectra with spectra in the Laboratory of Wood and Paper Chemistry, Åbo Akademi, Finland (own database) and with Wiley 10th/NIST2012. The results were calculated as the relative abundance of each pyrolysis product (peak area-% of total peak area).</p><!><p>Phenolic residues of the pellets were extracted and quantified based on a previously described method, which was slightly modified (Lehtonen et al., 2016). For extraction, 10 mg of GGM-phe or GX-phe was dissolved in 80% ethanol (1 ml) and centrifuged three times. The supernatants were combined and evaporated under reduced pressure. The ethanol-soluble phenolic residues were then extracted with ethyl acetate (3 × 500 μl), after adjusting the pH by adding 400 μl of 6 M HCl, and finally the ethyl acetate was evaporated under nitrogen stream. The ethanol-soluble phenolics were analyzed after extraction (neutral) or after acid or base hydrolysis. The pellets containing remaining carbohydrates from GGM-phe were also hydrolyzed after extraction with acid or base, as described previously (Lehtonen et al., 2016). No pellet remained from GX-phe after extraction of phenol. All treatments were performed in triplicate (n = 3).</p><p>For the analysis, all samples were redissolved in 10% MeOH (1 ml), filtered through a 0.2-μm PTFE syringe filter (VWR International, Radnor, PA, USA), and separated with an ACQUITY UPLC system (Waters, Milford, MA, USA), as described previously (Kylli et al., 2011; Lehtonen et al., 2016). The injection volume for all samples was 10 μl. In addition, the sample containing the ethanol-soluble phenols hydrolyzed with base was diluted to 1/10 for quantification at the concentration levels explained below.</p><p>For the identification of the main phenolic compounds extracted, the same UPLC system equipped with a Waters Synapt G2-Si high definition mass spectrometer with a LockSpray Exact Mass Ionization Source was used. The LC-MS spectra were processed with MassLynx 4.1 software, which uses an m/z lockmass value of 556.2771 for leucine eukephalin (equal to m/z of [M+H]+ + e−). Based on identification with MS, the main phenolic compounds extracted were quantified using vanillin and syringaldehyde as the reference compounds, with three levels of concentration in the range of 10–45 ng/injection. LC-MS spectra are presented in the Supplementary Material. ESI-MS vanillin: m/z 153.0552 [M + H]+ (C8H9O3 + e− requires 153.0552), syringaldehyde m/z 183.0657 [M + H]+ (C9H11O4 + e− requires 183.0657).</p><!><p>The starting hemicelluloses, spray-dried hot-water-extracted GGM and GX, were fractioned by centrifugation. As will be explained in the characterization of materials, supernatants consisted of hemicelluloses partially purified from phenolic compounds (GGM-pur and GX-pur fractions), and pellets contained mainly lignin and other phenolic residues (GGM-phe and GX-phe). This solvent-free fractionation method takes advantage on the low solubility of lignin in water, which enables partial separation of precipitated lignin and water-soluble hemicelluloses (Valoppi et al., 2019a). In a recently published study (Valoppi et al., 2019a), the effect of using different centrifugal treatments on the degree of purification and properties of GGM-rich PHWE extracts was evaluated in detail. In the present study, we used high centrifugal forces, optimized in the previous study (Valoppi et al., 2019a), on both GGM and GX to compare softwood and hardwood hemicelluloses for the first time by this fractionation method and to reveal the structure of phenolic residues coextracted with hemicelluloses.</p><p>Oil-in-water emulsions were then prepared from rapeseed oil stripped from tocopherols, using the starting hemicelluloses (GGM and GX) and the purified fractions (GGM-pur and GX-pur) as emulsifiers. The resulting emulsions were characterized to investigate the effect of removed phenolic residues on the physical and oxidative stability of emulsions. During the accelerated storage test of emulsions, the droplet-size distribution was measured periodically to monitor the physical stability, and the morphology was further confirmed by microscopy.</p><!><p>The droplet-size-distribution curves of all emulsions are presented in Figure 1, and values from selected time points of measurements are presented in Table 1 (All other values for droplet-size measurements are found in Table S1). Only the fresh emulsion stabilized with GGM-pur had unimodal droplet-size distribution, and the more bimodal distribution observed previously for GX emulsions (Mikkonen et al., 2016) was most evident for fresh emulsion stabilized with GX-pur. The surface average droplet size D[3,2] for all fresh emulsions was in the range of 120–150 nm, which is similar to the previous result for PHWE GGM (Lehtonen et al., 2018). However, the D[3,2] value of GGM-pur and GX-pur (120 nm) emulsions was smaller than that of emulsions with the starting GGM and GX (140–150 nm). The more pronounced bimodal droplet-size distribution of GX-pur compared to GX emulsion also increased the volume average droplet size D[4,3], which is affected more by the larger droplets compared to D[3,2]. During the storage test, the droplet size increased for all samples, and the change was more apparent for the emulsions stabilized with the purified hemicelluloses GGM-pur and GX-pur. This was most clearly observed in the D90 values, which take into account 90% of the oil droplets, which are equal or smaller than D90. In conclusion, it seems that the fraction that was removed from both of the purified hemicelluloses GGM-pur and GX-pur by centrifugation was responsible for a slightly larger droplet size of fresh emulsions, but on the other hand, it enhanced the long-term physical stability of emulsions, in agreement with previously published results for GGM (Valoppi et al., 2019a).</p><!><p>Droplet-size distribution of emulsions using the starting hemicelluloses, GGM and GX, and purified fractions, GGM-pur and GX-pur, as the emulsifiers. The physical stability of emulsions was observed during the storage test at 40°C by measuring droplet-size distribution periodically. The values for selected measurements are presented in Table 1.</p><p>Average droplet-size values for fresh o/w emulsions.</p><p>The D[3,2] and D[4,3] are the surface and volume weighted mean diameters. The D10, D50, and D90 values mean, respectively, that 10, 50, and 90% of droplets are smaller compared to that value.</p><!><p>The microscopic images obtained at the end of the 3-month storage period (Figure 2) confirm the results from the droplet-size-distribution measurements. The average droplet size D[3,2] for GGM and GGM-pur was still fairly low, 210–220 nm, at the end of the storage period, and these droplets were hardly visible by optical microscopy with the magnification used. In the image of GGM-pur in Figure 2c, there are possibly a couple of larger droplets compared to the image of GGM in Figure 2a. In the case of GX and GX-pur (Figures 2b,d), the difference in the number and size of larger droplets was more evident and clearly reflects the droplet-size-distribution data.</p><!><p>Light microscope images of emulsions at the end of the 3-month storage period. The emulsifiers are in (a) starting GGM, (b) starting GX, (c) GGM-pur, and (d) GX-pur. The length of the scale bar is 10 μm in all images.</p><!><p>In order to observe the oxidative stability of emulsions, their peroxide values were measured periodically during the accelerated storage test. The oxidative stability of emulsions stabilized with GX was then investigated for the first time. Peroxide values indicate the formation of hydroperoxides, the initial oxidation products of rapeseed oil (Lehtonen et al., 2011, 2016). The results clearly show (Figure 3) that the phenolic fraction removed from the starting hemicelluloses GGM and GX was responsible for inhibiting lipid oxidation in emulsions.</p><!><p>Peroxide values for rapeseed oil in emulsions using starting GGM and GX and using purified GGM-pur and GX-pur as the emulsifiers.</p><!><p>The peroxide values for all emulsions were practically unchanged during the first 6 weeks of storage, which is compatible with the previously published results for concentrated PHWE GGM (Lehtonen et al., 2018). The peroxide values of all emulsions started to increase after 6 weeks, but the extent of oxidation was different at the end of the storage period. The starting hemicellulose GX was the most stabilizing of all the emulsifiers tested, because the increase of peroxide values during the 3-month storage period was fairly modest compared to other emulsifiers, although the stabilization of 6 weeks for GGM is also a notable result.</p><!><p>The carbohydrate composition of the starting hemicelluloses, GGM and GX, and their purified (pur) and phenolic (phe) fractions (Table 2) was analyzed in order to evaluate both the total amount of carbohydrates in each fraction and possible differences in carbohydrate composition. The total amount of carbohydrates was around 735 mg g−1 for GGM and 615 mg g−1 for GX, which is in agreement with results previously obtained for spray-dried PHWE GGM and GX (Mikkonen et al., 2019). The total carbohydrate contents for the purified fractions were 845 and 621 mg g−1, implying that the fractionation method increased the ratio of carbohydrates for GGM-pur but was very similar when starting GX and GX-pur were compared. For the fractions GGM-phe and GX-phe, the amount of carbohydrates was clearly lower: 251 and 181 mg g−1. This result indicates that 75–82% of these fractions are of an origin other than carbohydrates, which in the case of wood-based hot-water-extracted material is most likely composed of lignin or other phenolic residues.</p><!><p>Carbohydrate composition of starting GGM and GX, and the centrifugated fractions GGM-phe and GX-phe obtained from quantitative GC analysis of acid methanolyzed and silylated samples.</p><p>The results presented in normal font are expressed as mg/g of dry sample. For results presented in bold and italic font, the results have been normalized by setting a value of 100 for the main carbohydrate in the sample (i.e., for GGM, Man = 100, and for GX, Xyl = 100). Glucuronic acid (GlcA) was not detected.</p><!><p>The carbohydrate composition of the starting materials and purified fractions was more similar to that of the phenolic-rich fractions. Furthermore, certain carbohydrates seemed to be associated more closely with the phenolic fractions. In both GGM-phe and GX-phe, the presence of Araf, Rhap, Glcp, and MeGlcA was pronounced, even taking into account the high standard errors in the results. Different types of lignins are known to be associated with certain carbohydrates: glucomannan-lignin complexes have been isolated mainly from softwoods, whereas glucan-lignins have been found in hardwoods and xylan-lignins in both softwoods and hardwoods (Lawoko et al., 2005; Li et al., 2011; Du et al., 2013; del Río et al., 2016). Further separation and characterization of the different types of carbohydrate-lignins were beyond the scope of this work, and thus any clear conclusions about the identity of potentially different polysaccharides associated with lignin cannot be made at this point.</p><!><p>For more detailed chemical characterization, the non-acetylated starting hemicelluloses were analyzed by 2D HSQC NMR spectroscopy. The HSQC spectra for GGM and GX are presented in Figures 4, 5, respectively. The spectra of samples dissolved in d6-DMSO were tentatively identified based on existing data for GGM (Hannuksela and Hervé du Penhoat, 2004; Kim and Ralph, 2014; Berglund et al., 2019), for GX (Teleman et al., 2000, 2002; Rencoret et al., 2012; Kim and Ralph, 2014), for lignin (Liitiä et al., 2003), and for LCC γ-ester (Giummarella et al., 2019), combined with the HSQC-TOCSY NMR spectra (presented in Figures S1A,B). Furthermore, the centrifuged fractions GGM-phe and GX-phe were acetylated prior to analysis to improve solubility in d6-DMSO. The HSQC spectra of acetylated GGM-phe and GX-phe side-chain area are presented in Figure 6, for which the signals were identified according to previously published data (Ämmälahti et al., 1998; Qu et al., 2011; Wen et al., 2012; Du et al., 2013). The color codes and symbols used for the chemical structures identified are presented in Figure 7, and the list of peaks is presented in Table S2.</p><!><p>2D HSQC NMR spectrum of PHWE GGM starting material. Upper figure is the entire NMR spectrum, and lower figure is a magnification of the side-chain area. The nonacetylated sample was characterized dissolved in d6-DMSO. The symbols and chemical structures are presented and explained in Figure 7.</p><p>2D HSQC NMR spectrum of PHWE GX starting material. Upper figure is the entire NMR spectrum, and lower figure is a magnification of the side-chain area. The nonacetylated sample was dissolved in d6-DMSO for analysis. The symbols and chemical structures are presented and explained in Figure 7.</p><p>2D HSQC NMR spectrum of the fractions GGM-phe (upper figure) and GX-phe (lower figure). Only magnification of the side-chain areas is shown. The acetylated samples were dissolved in d6-DMSO for analysis. The symbols and chemical structures are presented and explained in Figure 7.</p><p>Chemical structures identified by NMR. The symbols and abbreviations used in NMR spectra are G (Glc), M (Man), MGA (MeGlcA), X (Xyl), A (β-O-4/β-aryl ether), B (β-β/resinol), C (β-5/phenyl coumaran), AEst (β-O-4 connected with MeGlcA by γ-ester bond), Est (acyl ester), GAr (guaiacyl), and SAr (syringyl). For acetylated carbohydrates, for example, the abbreviation X12OAc refers to xylose C-1 containing acetyl group in the C-2 position.</p><!><p>The NMR results support the analysis of carbohydrate composition, and the main carbohydrates shown in Table 2 were also present in the NMR spectra of the starting materials GGM and GX (Figures 4, 5). Thus, the most intensive signals in the spectra of GGM and GX were assigned to Manp and Xylp, respectively. Glcp could be assigned for both samples, whereas MeGlcA was found only in the spectrum of GX and Galp only in the spectrum of GGM. Interestingly, considering the signal of MeGlcA, for which the MGA4 (see the Figure 7 text for abbreviations used in NMR spectra) does not overlap with other signals, it seems that the threshold limit of this 2D NMR technique prevents the observation of MeGlcA in GGM, in which the relative amount of MeGlcA was lower compared to GX. Both starting hemicelluloses also contained acetates, which are naturally present in wood hemicelluloses, GGM and GX (Sjöström, 1993). The XG13OAc was identified based on a previous structural characterization of GX of birch, beech, and aspen, because the position of the cross-signal between X12,3OAc and X12OAc fits very well with the published data (Teleman et al., 2000, 2002). The abbreviation XG13OAc refers to the structural element (→ 4)[4O-Me-α-D-GlcpA-(1 → 2)][O-Ac-(1 → 3)]-β-D-Xylp-(1 → ). Because the previously published NMR data were obtained in a different solvent (D2O), and because there were now more overlapping signals, assignment of the other signals belonging to this xylopyranosyl ring was not possible.</p><p>The NMR spectra show also that both starting hemicelluloses contained lignin, for which the signals of β-aryl ether type were the most intensive (Figures 4, 5). Further, the results provide further evidence of the fact that lignin is the phenolic material improving emulsifying/functional properties of PHWE extracts. Because β-O-4 linkage is the most abundant type in native lignin (Sjöström, 1993), the intensity of the signals indicated that the structure of lignin was not extensively degraded but instead preserved during the PHWE process. The other lignin bonding types found in the spectra of the starting hemicelluloses were phenyl coumaran type (β-5), found for both hemicelluloses, and resinol type (β-β), found only for GX. According to the signals in the aromatic region at around 7 ppm, GGM contained only aromatic protons of the guaiacyl type, and GX contained mainly aromatic protons of the syringyl type and a small amount of guaiacyl type protons.</p><p>The starting hemicelluloses also contained -CH2- protons connected to ester functionality. For GX (Figure 5), the signal at 4.30/62.92 ppm was assigned to the γ-proton of the β-O-4-structure linked to MeGlcA through an ester bond, which is also known as the γ-ester type LCC bond (Li and Helm, 1995; Giummarella et al., 2019). In addition, the signal at 4.30/83.46 ppm was assigned to the β-proton belonging to the same LCC-bonding pattern type, confirming that γ-ester LCC bonds must be rather frequent in GX hemicelluloses produced by the PHWE process. According to model compound studies using smaller synthesized molecules, as well as to those using lignin dehydropolymer (DHP), urunosyl units can migrate to the γ-position (Li and Helm, 1995; Giummarella et al., 2019), and thus it is also possible that LCCs are formed during the PHWE process. Similarly for GGM (Figure 4), the signals at 4.04 and 4.27/63.25–63.4 ppm could be assigned to -CH2- protons connected to ester, but because no more of the signals present belonged to the β-O-4-type LCC-bonding pattern, these signals could not be unequivocally identified as originating from lignin. For example, the GGM of Aloe barbadensis contains acetyl groups at the C-6 position of Manp, which give signals at the same positions in the HSQC spectrum (Campestrini et al., 2013).</p><p>As already suggested by the small amount of carbohydrates, the centrifuged fractions GGM-phe and GX-phe (Figure 6) were composed mainly of lignin. The samples were also acetylated prior to analysis in order to improve their solubility in d6-DMSO. The typical bonding patterns for lignin were found, and the signals for the β-O-4 bond type were clearly the most intense, similarly to the starting hemicelluloses. For both GGM-phe and GX-phe, the other lignin bonding patterns, the β-5 and β-β structures, were also more clearly identified compared to the NMR spectra of the starting hemicelluloses. The LCC structures could not be clearly identified from these acetylated phenolic fractions, because the signals of γ-esters would in this case overlap with all the acetylated γ-signals in lignin.</p><!><p>The results from molar mass analysis by SEC for the starting hemicelluloses GGM and GX and for the purified and phenolic fractions are presented in Table 3. The molar masses of starting GGM, Mw of around 7,300 Da, and starting GX, Mw of around 3,100 Da, were in a similar range, but slightly lower compared to the previously reported values (Mikkonen et al., 2019). However, the previously obtained results, 8,200 Da for GGM and 4,000 Da for GX, were analyzed in water solutions compared to the DMSO used in this study, which could have affected the results slightly. The molar masses of both purified fractions were similar to those of the starting materials, 7,200 Da for pur-GGM and 3,400 Da for pur-GX.</p><!><p>Molar masses for starting hemicelluloses, GGM and GX, and the purified and phenolic fractions analyzed by size-exclusion chromatography (SEC) in DMSO.</p><!><p>For phenolic fractions containing mainly lignin, the estimated molar masses were lower than those of the starting materials, for GGM-phe significantly lower 2,800 Da and for GX-phe 2,500 Da. Although present knowledge indicates that analysis by SEC gives underestimated molar masses for lignins (Zinovyev et al., 2018), the results show that GGM-phe and GX-phe have similar molar masses but that their molar masses are different from those of the starting hemicelluloses and fractions of purified hemicelluloses. Furthermore, the polydispersity, Mw/Mn, for all GGM samples was in the range of 7.5–8.0, although for GX samples the value was higher for GX-phe (7.6) and lower for starting GX (4.9) and GX-pur (4.7). In this respect, the variation of molar masses was similar within the phenolic fractions (GGM-phe and GX-phe) as well as for the GGM starting material and purified fraction GGM-pur, whereas dispersity was slightly lower for starting GX and GX-pur.</p><!><p>The 2D DOSY results are shown in Table 4. The viscosity corrected value Dc(GGM) (0.17 × 10−10 m2s−1) is clearly lower than Dc(GGM-phe) (0.21 × 10−10 m2s−1), Dc(GX) (0.22 × 10−10 m2s−1), and Dc(GX-phe) (0.21 × 10−10 m2s−1), the latter three being practically identical. This is in line with the SEC results (Table 3), indicating approximately 7 kDa for GGM-pur and 3 kDa for the others. However, it must be pointed out that the absolute differences in these diffusion coefficients are not large. Because there is a spread in the DOSY-correlations (i.e., all the peaks of the molecule do not appear with the same Dc value), it is difficult to pick a representative average value. There are various reasons for the spread, such as possible overlap with other residual entities, success of DOSY fitting, non-optimized diffusion time/diffusion gradient area (in order to achieve sufficient decay), noise, etc. This, combined with the aforementioned small absolute differences, makes these DOSY results indicative at best, but still usable for qualitative purposes. Improvement could be achieved by optimizing the diffusion delays and/or diffusion gradient areas, increasing the number of diffusion steps in DOSY measurement, and increasing number of transients.</p><!><p>Diffusion coefficients (Dc) obtained from DOSY NMR of acetylated (or partially acetylated) starting GGM and GX and phenolic fractions GGM-phe and GX-phe.</p><p>Dc(MM) is the diffusion constant of macromolecule.</p><!><p>Furthermore, the lignin signals, which do not overlap with other residues, the β-O-4 (1H 5.95 ppm) and ArH (1H 6.65 ppm for GX and 6.97 ppm for GGM), have similar diffusion coefficients compared to carbohydrate signals. This means that lignin has a very similar diffusion coefficient compared to hemicelluloses, which provides further support for covalent association of carbohydrates and lignin.</p><!><p>The pyrolysis GC/MS technique (py-GC/MS) was used to evaluate the usefulness of this method for fast characterization of the phenolic content of the starting hemicelluloses and phenolic fractions. Py-GC/MS correlates with the lignin and carbohydrate composition, especially for pulp samples, and can be used fairly reliably for the determination of the S/G ratio, which is the ratio of syringyl and guaiacyl types of units in lignin (del Rio et al., 2002; Ohra-aho et al., 2013, 2018). Thus, a rough estimation of the lignin and carbohydrate content of the starting hemicelluloses GGM and GX and for the lignin-rich residues GGM-phe and GX-phe was made by grouping all the peaks from py-GC/MS and calculating areas of all groups, as presented in Table 5.</p><!><p>The products found in py-GC/MS of starting hemicelluloses GGM and GX and phenolic fractions GGM-phe and GX-phe.</p><p>The compounds originated from carbohydrates (Carb), p-hydroxyphenyl (H), guaiacyl (G), or syringyl-type (S) lignin units, other aromatic units (Ar), or rosin acids (RA).</p><p>May contain areas of two signals identified to the same compound by GC-MS. The significance to the total value is 1% or less.</p><p>The values are presented as percentages of the peak area compared to the total peak area.</p><!><p>The results were then compared to the acid methanolysis followed by GC analysis of total carbohydrates (in Table 2). Assuming that the starting hemicelluloses and phenolic fractions contained only hemicelluloses and lignin, the results of the methods should be compatible. However, the carbohydrate contents from determined py-GC/MS were much lower compared to the results obtained from acid methanolysis-GC method. On the other hand, the lignin content from py-GC/MS seemed fairly reasonable when compared to total carbohydrate content from acid methanolysis. When the total carbohydrate content from acid methanolysis and lignin content from py-GC/MS were summed, the total content (lignin + carbohydrates) was 94.34% for starting GGM, 100.01% for GGM-phe, 86.54% for starting GX, and 91.73% for GX-phe. According to previous results, the py-GC/MS-analysis of carbohydrate content is not necessarily reliable for comparing samples containing different carbohydrates (Ohra-aho et al., 2018), which probably affected the results presented here as well. However, for the rough evaluation of lignin content in samples of hemicelluloses, the method could be suitable, and it could provide estimations of the carbohydrate content in an indirect way.</p><p>The S/G ratio of the starting GX and GX-phe samples was very similar (4.70 and 4.89, respectively; Table 5). The reliability of analyzing the S/G ratio by py-GC/MS has been shown for eucalyptus samples (Ohra-aho et al., 2013), and the method is most likely valid for GX hemicelluloses. A recently published S/G ratio for birch wood from Sweden was 3.25 (Wang et al., 2018). Although the results are not necessarily comparable for samples from different wood materials, the S/G ratio obtained for lignin associated with GX seems fairly high, also taking into consideration the results obtained for other hardwood species. For example, in another study of eucalyptus samples, the S/G ratio was 1.9–3.1 (Ohra-aho et al., 2013).</p><!><p>A previous study showed that certain types of extractable small phenolic compounds of PHWE GGM concentrate were adsorbed on the oil droplets of rapeseed oil emulsions (Lehtonen et al., 2018). It was then assumed that LCC structures composed of phenolic and carbohydrate residues would improve the emulsification and stabilization ability of PHWE hemicelluloses. We now assume also that the extractable phenolic compounds would be associated with lignin present in the samples. The phenolic fractions GGM-phe and GX-phe were extracted and analyzed with UPLC, and the main peaks were identified with LC-MS and then quantified using corresponding standards.</p><p>The main small phenolic compounds identified according to LC-MS were vanillin (in both GGM and GX) and syringaldehyde (only in GX). The amounts found in GGM-phe and GX-phe are shown in Table 6. Both compounds were found mainly in the ethanol soluble fractions; GX was not even precipitated during extractions. The amounts of compounds dissolved in neutral solvent and additionally acid hydrolyzed were very similar. Clearly, the highest amount of these compounds was released by base hydrolysis.</p><!><p>The amounts obtained from UPLC analysis of main small extractable phenolic compounds, vanillin and syringaldehyde, which were identified as the main products extracted from the phenolic residues GGM-phe and GX-phe.</p><p>ND, not detected.</p><!><p>The total amount of vanillin and syringaldehyde extracted was <0.1 m-%, meaning that the amount was still much lower considering the starting hemicelluloses. However, the classification of vanillin and syringaldehyde would fit that of hydroxycinnamic acids (OHCs) in terms of the previously used classification (Lehtonen et al., 2018). On the other hand, ethanol soluble phenols belonging to OHCs were found solely adsorbed in the oil of the emulsion, which means that vanillin bound to GGM-phe and syringaldehyde bound to GX-phe also participate in the formation of emulsions. Furthermore, because these compounds are clearly mainly covalently bound to the phenolic fractions containing lignin, it is also likely that lignin is involved in the formation and stabilization of emulsions.</p><p>The S/G ratio of syringaldehyde and vanillin extracted and base-hydrolyzed from GX-phe was 4.79, which is very close to the value obtained from py-GC/MS for the whole lignin. Although the values could be similar by coincidence, it is more likely that the similar S/G ratio obtained reflects the presence of lignin adsorbed with hemicelluloses to the surface of emulsion droplets. Because we have not thus far been able to completely release hemicelluloses adsorbed on rapeseed oil droplets, this quantitation by UPLC is by far the best method for identifying the presence of lignin in emulsions stabilized with PHWE hemicelluloses, and it can be used to tag on lignin associated with hemicelluloses.</p><!><p>The results regarding the droplet-size distribution of emulsions (Figure 1, Table 1) can be explained by the presence of lignin. For fresh emulsions prepared using purified GGM-pur and GX-pur fractions, the D[3,2] values were smaller compared to starting GGM and GX. It is reasonable to assume that lignin's participation in the formation of oil droplets would increase their size.</p><p>Regarding the physical stability of emulsions, the droplet size increased faster during the storage of emulsions stabilized with GGM-pur and GX-pur compared to emulsions stabilized with the starting hemicelluloses. This indicates that the presence of lignin stabilizes the physical structure of emulsions. For PHWE GGM, it was recently demonstrated that the mixed mechanism involves Pickering stabilization with interfacial adsorption of GGM, which are probably associated with lignin (Valoppi et al., 2019a). In addition, the bimodal distribution of GX into smaller and larger droplets was less enhanced in the presence of lignin.</p><p>It is evident that lignin, as a natural antioxidant, is also responsible for the improved oxidative stability of emulsions. However, oxidation of phenolic compounds may also change their chemical structure, which could further induce structural changes and affect the physical stability of emulsions. The presence of LCC bonds was evident from the NMR spectrum of starting GX, and the γ-ester structures found could be at least partially responsible for the functional properties of PHWE hemicelluloses, allowing the lignin part anchor to the oil droplet surface, as hypothesized previously (Lehtonen et al., 2018). In this case, it is not necessary to debate whether the LCCs are derived from the starting wood material or produced during the PHWE process; the essential point is the excellent functional properties of hemicelluloses produced by the PHWE process.</p><!><p>We showed that phenolic structures, which were partially removed from both GGM- and GX-rich wood extracts by using centrifugal forces, played a key role in emulsion stability. The proportions, chemical compositions, and molar masses of the phenolic-rich fraction varied between GGM and GX hemicelluloses. Complementary chemical characterization of centrifuged materials showed that the phenolic-rich fraction contained mainly native lignin and a small amount of carbohydrates.</p><p>Using various approaches, the results confirmed that this phenolic-rich fraction improved both the physical and the oxidative stability of emulsions stabilized with PHWE extracts. The antioxidative properties of phenolic compounds coextracted with hemicelluloses may also be interlinked with the physical stability of emulsions. Furthermore, NMR analysis confirmed the presence of a high concentration of γ-ester type LCCs, which could explain the excellent emulsifying capacity of PHWE hemicelluloses. Both GGM and GX produced emulsions with high physical and oxidative stability, although the emulsions had slightly different types of characteristics depending on the source of hemicellulose. The results also showed that in order to achieve desired emulsifying properties, the total removal of lignin is not advisable; in fact, it introduces unnecessary complexities into the PHWE biorefining process.</p><!><p>All datasets generated for this study are included in the article/Supplementary Material.</p><!><p>KM planned and received funding for the project. ML mainly designed the experimental plan, with expertise in wood chemistry, with the help of KM (emulsions and hemicelluloses) and FV (emulsions, fractionation of hemicelluloses by centrifugal forces). ML performed and analyzed the 2D HSQC NMR of hemicelluloses, did part of the practical work during the preparation and characterization of emulsions, performed and analyzed the phenolic extraction procedures by UPLC and LC-MS, and assumed the main responsibility for writing the manuscript and interpreting the data. PK provided the materials for the study as well as knowledge about PHWE hemicelluloses and the process. VJ analyzed the carbohydrate content under the guidance of ML. VJ also contributed to the preparation and characterization of emulsions. SH designed and performed the DOSY NMR analysis, provided technical support during NMR analysis, and contributed to the writing of the experimental details of NMR for the manuscript. NM contributed to the calibration of SEC data and determination of molar masses. All authors read and commented on the manuscript.</p><!><p>The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.</p>

PubMed Open Access

Directed, Nickel-Catalyzed Umpolung 1,2-Carboamination of Alkenyl Carbonyl Compounds

We report a regioselective, nickel-catalyzed syn-1,2carboamination of non-conjugated alkenyl carbonyl compounds with O-benzoyl hydroxylamine (N-O) electrophiles and aryl/alkylzinc nucleophiles to afford β-and γ-amino acid derivatives. This method enables preparation of products containing structurally diverse tertiary amine motifs, including heterocycles, and can also be used to form quaternary carbon centers. The reaction takes advantage of a tethered 8-aminoquinoline directing group to control the regiochemical outcome and suppress two-component coupling between the N-O electrophile and organozinc nucleophile.

directed,_nickel-catalyzed_umpolung_1,2-carboamination_of_alkenyl_carbonyl_compounds

<p>Nitrogen-containing small-molecules comprise a significant portion of all known medicines. [1] Thus, novel methods for the formation of carbon-nitrogen (C-N) bonds have been actively pursued. [2] 1,2-Carboamination represents an appealing strategy for converting readily available alkene starting materials into valuable structurally complex amine products in an expedient manner (Scheme 1). This transformation can be carried out using different modes of reactivity, including a classical-polarity approach in which the nitrogen-based reactant functions as a nucleophile (i.e., R2NH) and an umpolung approach where the nitrogen-based reactant is an electrophile (i.e., R2NX, X = halide or pseudohalide), as depicted in Scheme 1a.</p><p>Catalytic intramolecular (two-component) alkene carboamination involving both polarity types has been extensively studied. [3] Intermolecular (three-component) variants, on the other hand, remain comparatively unexplored and have typically been limited to conjugated alkenes (e.g., styrenes or acrylates). [4] In terms of precedents involving non-conjugated, unstrained alkenes, [5] Liu and coworkers have reported palladium-catalyzed carbonylative 1,2-carboamination using 2oxazolidone or phthalimide nucleophiles to afford terminal βamino acids. [5a] Later, this group reported a similar net transformation involving an azide-containing hypervalent iodine reagent. [5b] These two reports rely on rapid migratory insertion of CO to outcompete side reactions, such as β-H elimination. Our group has reported a palladium-catalyzed directed 1,2carboamination of unactivated alkenes via a classical polarity approach (Scheme 1d). [6] In particular, we demonstrated regioselective anti-addition of imides, amides, sulfonamides, and various azaheterocycles with aryl iodides across alkenes. These contributions notwithstanding, 1,2-carboamination of nonconjugated alkenes employing aliphatic amines and alkyl carbon coupling partners remain unexplored. The goal of the present study was to address this knowledge gap through the development of a three-component umpolung carboamination of a non-conjugated alkene using a substrate directivity strategy.</p><p>Electrophilic aminating reagents have a rich history in enabling C-N bond formation. [7] During the past few years, examples of umpolung carboamination of alkenes and allenes have been described. [4a, 4c][8] For example, building on seminal reports by Narasaka, [8] Bower and coworkers described an intramolecular umpolung carboamination of 𝛾,δ-unsaturated oxime esters with arylboronic ester coupling partners. [9] The Zhu group later described analogous reactivity with 1,3,4-oxadiazole C-H nucleophiles (Scheme 1b). [10] Regarding intermolecular examples, in 2013 the Zhang group described an umpolung radical-based copper-catalyzed aminocyanation of styrenes employing N-F reagents (Scheme 1c). [4a] Last year Liu and coworkers published an enantioselective aminoarylation of styrenes also catalyzed by copper using N-F reagents as electrophiles. [4c] In contrast, analogous transformations involving the use of electrophilic aminating reagents in nickel-catalysis have been less extensively studied. [7g, 11] Two-component nickel-catalyzed C-N cross-couplings between organometallic nucleophiles and N-O electrophiles have been described by the Johnson, [11a] Jarvo, [11b] and Knochel groups. [11c] To the best of our knowledge, only a single example of nickel-catalyzed alkene carboamination has been reported to date, [8g] an intramolecular system developed by Selander and coworkers in 2017 (Scheme 1b). [7g, 11d] Realization of an intermolecular nickel-catalyzed carboamination</p><p>[N]-H = imides, amides, sulfonamides, hydroxamic acid derivatives, and azaheterocycles Table 1. Selected Optimization of Reaction Conditions.</p><p>[a]</p><p>[a] Reaction conditions: 1a (0.1 mmol), 2a (1.0-1.2 equiv), Me2Zn (1.0 M in heptane), 60 °C, 18-24 h. 1 H NMR yields reported with CH2Br2 as internal standard.</p><p>process would present the opportunity to rapidly generate medicinally motifs with dense functionality.</p><p>We recently described substrate-directed nickel-catalyzed three-component conjunctive cross-coupling reactions [12] that append differentiated alkyl/aryl fragments to β,𝛾-and 𝛾,δunsaturated carbonyl compounds using aryl/alkyl halides and aryl/alkyl zinc reagents. [13] The regioselectivity of these reactions is controlled by a tethered 8-aminoquinoline (8-AQ) directing group that stabilizes 5-or 6-membered metallacycles, thereby suppressing undesired side reactions, such as β-hydride elimination or two-component cross-coupling. Given these results we wondered if it would be possible to employ Obenzoylhydroxylamines as electrophiles in lieu of aryl/alkyl halides to synthesize β-and 𝛾-amino acid derivatives under nickel catalysis. We surmised that this approach would complement our previous palladium(II)-catalyzed method (Scheme 1d) in several respects. [6] Namely it would be synselective, proceed with the opposite sense regioselectivity, enable use of alkyl coupling partners, and potentially compatible with alkenes distal from the AQ group (Scheme 1e).</p><p>To test this idea, we elected to use alkene 1a as the pilot substrate given its unique effectiveness in earlier work [13][14] and 2a as the electrophilic nitrogen source based on its success in various other catalytic methods. These starting materials were combined with commercially available dimethylzinc solution in the presence of catalytic nickel. With 20 mol % Ni(cod)2 we observed formation of product 3a (Table 1) in 71% yield (Table 1, Entry 1). DMF, toluene, acetonitrile, and dioxane were also tested under conditions otherwise identical to those in entry 1. The reaction proceeded in toluene and dioxane, though yields were attenuated compared to in THF. Considering conditions from our previous work [13b] we attempted to drive the reaction to completion by using excess Me2Zn, but in this case we found significantly diminished yields when more than one equivalent was used. We also found that the reaction was higher yielding at lower concentrations, with the optimal concentration being 0.075 M 1a in THF. Lower catalyst loadings of 10-15 mol % gave comparable yields, though decreasing the catalyst loading further (5 mol %) led to slightly diminished yield. Increasing the amount of 2a to 1.2 equiv provided 3a in 84% 1 H NMR yield (79% isolated). We also found that the reaction performs comparably well using several bench-stable Ni(II) salts, enabling a glove-box free protocol. The product 3a was isolated in 77% yield using NiCl2 as the precatalyst.</p><p>Having optimized the reaction conditions, we proceeded to explore the O-benzoylhydroxylamine electrophile scope (Table 2). We found that several hetereocyclic motifs (3b-3g) frequently found in bioactive compounds were well-tolerated, including thiomorpholine, tert-butyoxycarbonyl-protected piperazine, 2-(piperazin-1-yl)pyrimidine, 4,5,6,7tetrahydrothieno[3,2-c]pyridine, piperidine, and pyrrolidine. An array of N-O reagents derived from acyclic amines (2h-2l), including N-methyl-N-benzylamine, diethylamine, dibenzylamine and diallylamine also reacted under optimized conditions. Sterically hindered and especially reactive N-O reagents could not be used as coupling partners in this reaction (see SI). The product 3l was obtained in a similar yield using NiCl2, highlighting its efficacy as a substitute for Ni(cod)2. The reaction was also compatible with a variety of diorganozinc and organozinc halide nucleophiles, though some reactions were found to proceed in diminished yields. In some cases this could be overcome by slow addition of the organozinc nucleophile, demonstrated in the synthesis of 3m, which was isolated in 87% yield. Several other primary alkylzinc nucleophiles were compatible, including propyl (3n), ethyl propionate (3o) and benzyl (3p), providing the corresponding products in moderate yields. Secondary alkylzinc nucleophiles such as cyclobutyl (3q) and cyclohexyl (3r) could also be employed and provided moderate yields. Tertiary carbon nucleophiles were unsuccessful under the optimized reaction conditions (see SI). We observed that monoalkylzinc halides were generally lower yielding that dialkylzinc reagents, likely due to their well-known attenuated nucleophilicity. We hypothesize that the need for excess alkylzinc halide (four times more than in the case of dialkylzinc reagents) is due to competitive reduction of the electrophile 2a before it is able to react in the desired pathway, leading to decreased yields. In the case of secondary nucleophiles, we believe competing β-hydride elimination pathways generates reducing species in solution that facilitate electrophile decomposition. We have also isolated aminoarylated product 3s in 27% yield.</p><p>We also explored the scope of alkene substrates (Table 3) and found that the reaction was compatible with a variety of substituted alkenes. The relative stereochemistry of 4a was determined by X-ray crystallography, establishing that the reaction proceeded in a syn-selective manner. [13] The stereochemistry of other products derived from internal 1,2disubstituted alkenes were assigned by analogy. 4a was also obtained using NiCl2 as the precatalyst and was obtained with similar yield. We also found that a phthalimide-protected amine could be tolerated under the reaction conditions to afford carboaminated product 4c in moderate yield. Given the success of setting quaternary carbon centers in our previously published dialkylation reaction, [13b] we wondered whether this carboamination reaction could also function in sterically congested environments. We were pleased to find tri-and 1,1disubstituted alkenes could be used to synthesize compounds 4d and 4e, respectively, in good yields, demonstrating the ability of this method to forge quaternary carbons centers at either the β and 𝛾 position. The reaction also proceeded in moderate yields with α-methyl substituted alkenyl carbonyl compounds (4f), though we found the benzyl-substituted analogue to proceed in significantly reduced yields (<10% isolated). We also found the reaction could be extended to 𝛾,δ-unsaturated substrates to afford products 4g-4i and 2-vinylbenzamide-derived product 4j.</p><p>We next performed the reaction on gram scale to demonstrate its synthetic utility. On 5-mmol scale, we were able to isolate 1.30 g of 3a in 83% yield (Scheme 2). We also validated two methods to remove the AQ directing group. Hydrolysis of 3a afforded β-amino acid 3a' in 76% yield. Using a method published by Ohshima and coworkers, [15] methanolysis of 3a afforded ester 3a'' in 79% yield. Furthermore, we found that a stereocenter at the carbon α to the carbonyl did not racemize under the reaction conditions (see SI for details).</p><p>Regarding the reaction mechanism, we surmised that two plausible redox manifolds could be operative, namely Ni(0)/Ni(II) or Ni(I)/Ni(III) catalysis (depicted in general form as Ni(n)/N(n+2)). Moreover, the reaction could proceed via two different orders of events (Fig. 3a). In Pathway A, substratebound nickel complex 6c would first undergo oxidative addition with 2 to form intermediate 6a. Transmetalation followed by insertion to the alkene would form 6b, which could reductively eliminate to form products 3-4 and regenerate the active catalyst. In the second potential mechanism, Pathway B, intermediate 6c would first react via transmetalation, after which migratory insertion would lead to intermediate 6d. This species could then oxidatively add to 2 to give nickel intermediate 6e.</p><p>Reductive elimination would form the key C(sp 3 )-N bond and regenerate the catalytically active low-valent nickel species. A third mechanistic scenario (see SI) in which C-N bond formation precedes transmetalation and C-C reductive elimination cannot be conclusively ruled out at this stage, though we consider it to be less likely because it would involve formation of larger nickelacycles in preference to smaller nickelacycles with both classes of substrates (6 versus 5 with products 3, and 7 versus 6 with products 4).</p><p>In an effort to disambiguate between these possibilities, we prepared radical clock electrophile 2m (Fig. 3b). Based on literature precedents, the corresponding aminyl radical-which would be formed if SET oxidative addition were operative [11d] - was expected to cyclize with a first-order rate constant of approximately 10 4 s -1 . [16] When this electrophile was subjected to standard reaction conditions, only non-cyclized product 3t was formed in 40% yield. No evidence of cyclization was observed by 1 H NMR of the crude reaction mixture. This result is consistent with a two-electron oxidation addition pathway or alternatively with an SET oxidation pathway involving a radical recombination step with a rate constant >10 4 s -1 .</p><p>The effect of radical inhibitors was next studied (Fig. 3c). The reaction was not inhibited by the addition of BHT (1 equiv). On the other hand, addition of TEMPO (1 equiv) dramatically suppressed product formation, leading to unreacted starting materials, as well as TEMPO-H and TEMPO-Me adducts, as monitored by 1 H NMR and LC-MS. This result suggests that a Ni(I)/(III) cycle involving a Ni(I)-Me intermediate and SET events may be operative; however, a more detailed mechanistic study is needed before firm conclusions can be drawn. In summary, we have developed an intramolecular umpolung carboamination of non-conjugated alkenes that affords a variety of β-and 𝛾-amino acid and ester derivatives. The reaction is enabled by a removable 8-aminoquinoline tethered directing group, which facilitates formation of stabilized 5-or 6-membered nickelacycles, suppresses β-hydride elimination and two-component coupling, and determines the regiochemical outcome. The reaction tolerates a range of alkenes with various substitution patterns and proceeds in the presence of several synthetically important functional groups.</p>

ChemRxiv

In Silico Studies of Polyaromatic Hydrocarbon Inhibitors of Cytochrome P450 Enzymes 1A1, 1A2, 2A6, and 2B1

A computational study was undertaken to understand the nature of binding and the structural features that play a significant role in the binding of arylacetylene molecules to cytochrome P450 enzymes 1A1, 1A2, 2A6 and 2B1. Nine polycyclic arylacetylenes determined to be mechanism-based P450 enzyme inhibitors were studied. The lack of polar substituents in these compounds causes them to be incapable of hydrogen bonding to the polar protein residues. The four P450 enzymes of interest all have phenylalanine residues in the binding pocket for potential \xcf\x80\xe2\x80\x93\xcf\x80 interactions with the aromatic rings of the inhibitors. The inhibition potency of these arylacetylenes toward P450s 1A1 and 2B1 showed a dependence on the proximity of the inhibitor\xe2\x80\x99s triple bond to the prosthetic heme Fe of the enzyme. In P450 enzyme 1A2, the inhibitor\xe2\x80\x99s potency showed more dependence on the \xcf\x80\xe2\x80\x93\xcf\x80 interactions of the inhibitor\xe2\x80\x99s ring systems with the phenylalanine residues of the protein; with proximity of inhibitor triple bond to the heme Fe weighing in as the second most important factor. The results suggest that maximizing the \xcf\x80\xe2\x80\x93\xcf\x80 interactions with phenylalanine residues in the binding pocket and optimum proximity of the acetylene moiety to the heme Fe will provide for a substantial increase in the potency of the polyaromatic hydrocarbon mechanism-based inhibitors. A fine balance of these two aspects of binding coupled with attention to supplementing hydrophobic interactions could address potency and selectivity issues for these inhibitors.

in_silico_studies_of_polyaromatic_hydrocarbon_inhibitors_of_cytochrome_p450_enzymes_1a1,_1a2,_2a6,_a

Introduction<!>Assays<!>7-Ethoxyresorufin O-deethylation (EROD), 7-methoxyresorufin O-demethylation (MROD), 7-pentoxyresorufin O-depentylation (PROD), and Coumarin 7-Hydroxylation Assays<!>Data Analysis (14)<!>Pre-incubation Assays in the Presence and Absence of NADPH<!>Protein Crystal Structures and Homology Modeling of P450s 1A1 and 2B1<!>Binding Modes by Docking Simulation<!>Results and Discussion<!>Conclusion<!>

<p>Cytochrome P450 enzymes constitute a large superfamily of heme proteins known to catalyze oxidation reactions of an expansive array of endogenous and exogenous compounds including drugs, sterols, fatty acids, carcinogens, vitamins and steroids (1–3). There are 57 known human P450 enzymes, including a few pseudogenes which are classified into families (sequence identity >40%) and subfamilies (sequence identity ≥55%), (see http://drnelson.utmem.edu/CytochromeP450.html). Some of these enzymes are polymorphically expressed, resulting in different metabolic activity that could influence the overall toxic effect, including carcinogenesis induced by environmental chemicals (4). The P450 enzymes 1A1, 1A2, 2A6, and 2B1 have been shown to oxidize polycyclic aromatic hydrocarbons to produce carcinogenic agents that can bind to DNA causing cancer (5–8). Targeting these enzymes for inhibition with small molecules offers new avenues in cancer prevention and therapy (9). Many acetylenic compounds are shown to irreversibly inactivate P450 enzymes by time-dependent destruction of heme via alkylation (6, 10–15). Ortiz de Montellano and coworkers have established the mechanism involving the formation of an intermediate by P450-dependent oxidation of acetylenic moiety followed by covalent binding of the unstable intermediate to a heme nitrogen (16, 17).</p><p>In our present study, we show that arylacetylenes, such as, 1-ethynylpyrene (1EP), 1-(1-propynyl)pyrene (1MEP), 4-(1-propynyl)pyrene (4-MEP), 1-(1-butynyl)pyrene (1-EEP), 2-ethynylphenanthrene (2-EPhen), 3-ethynylphenanthrene (3-EPhen), 9-ethynylphenanthrene (9-EPhen), 2-(1-propynyl)phenanthrene (2-MEPhen), and 9-(1-propynyl)phenanthrene (9-MEPhen) act as mechanism-based inhibitors of P450s 1A1 and 1A2, with many of them additionally inhibiting P450 2B1. None of these compounds showed significant inhibition of P450 2A6. The unique aspect of these compounds is the lack of any polar structural moiety capable of hydrogen bonding with polar protein residues. Our goal was to understand the structural basis of the interactions that these molecules have with the P450 enzymes 1A1, 1A2, and 2B1 through molecular modeling studies. We were also intrigued by the fact that none of these compounds inhibited P450 2A6. Reported X-ray crystal structures were utilized for the enzymes 1A2 and 2A6, while homology models were built for P450s 1A1 and 2B1. Docking studies of the arylacetylenes with the four P450 enzymes 1A1, 1A2, 2A6 and 2B1 were performed and the results were analyzed to obtain the consensus binding posture for each of the molecules of interest. Better understanding of the molecular nature of binding interactions will direct us in the design of more effective and selective inhibitors for these P450 enzymes.</p><!><p>Rat CYP2B1 supersomes (rat CYP2B1 + P450 reductase + cytochrome b5), and human CYP1A1, 1A2, and 2A6 supersomes (human CYP enzymes + P450 reductase) were purchased from B.D. Biosciences Corporation (Woburn, MA, USA). All other chemicals were purchased from Sigma Aldrich Company (Milwaukee, WI).</p><p>The P450 1A1, 1A2, and 2B1 dependent activities were assayed in 7-alkoxyresorufin dealkylation assays using ethoxyresorufin, methoxyresorufin, and pentoxyresorufin fluorescent substrates, respectively (18). P450 2A6 dependent 7-hydroxylation of coumarin was used in a similar assay with minor differences as described below for measuring 2A6 activity (19, 20).</p><!><p>Potassium phosphate buffer (1750 μL of a 0.1 M solution, pH 7.6) was placed in a 1.0 cm quartz cuvette, and 10 μL of a 1.0 M MgCl2 solution, 15 μL of a 1.0 mM corresponding resorufin or coumarin substrate solution in Me2SO (DMSO), 10 μL of the microsomal P450 protein, and 15 μL of an inhibitor in DMSO were added. For the controls, 15 μL of pure DMSO was added in place of the inhibitor solution. The reaction was initiated by the addition of 200 μL of a NADPH regenerating solution. The regenerating solution was prepared by combining 797 μL of a 0.10 M potassium phosphate buffer solution (pH 7.6), 67 μL of a 15 mM NADP+ solution in buffer, 67 μL of a 67.5 mM glucose 6-phosphate solution in buffer, and 67 μL of a 45 mM MgCl2 solution, and incubating the mixture for 5 minutes at 37°C before the addition of 3 units of glucose 6-phosphate dehydrogenase/mL and a final 5 minute incubation at 37°C. The final assay volume was 2.0 mL. The production of 7-hydroxyresorufin anion was monitored by a spectrofluorimeter (OLIS DM 45 Spectrofluorimetry System) at 530 nm excitation and 585 nm emission, with a slit width of 2 nm. The production of 7-hydroxycoumarin was monitored at 338 nm excitation and 458 nm emission, with a slit width of 2 nm. The reactions were performed at 37°C. For each inhibitor, a number of assay runs were performed using varying inhibitor concentrations.</p><!><p>The data obtained from these assays were analyzed by a computer analysis method of the reaction progress curve in the presence of various inhibitor concentrations and in the absence of the inhibitor as the control run. Results are tabulated in Table 1. A second order curve describing product formation with respect to reaction time in seconds was obtained for each inhibitor concentration and the control. The Microsoft Excel Program was used to fit the data (Fluorescence intensity vs. Time) obtaining the parameters of the best-fit second order curves (y=ax2+bx+c). Using the parameters (Coefficient b in the second order equation is enzymatic activity) obtained from the above, activities were calculated using first order derivatives. Dixon plots were used (by plotting the reciprocals of enzymatic activity (1/v) vs. inhibitor concentrations [I]) in order to determine IC50 values (x-intercepts) for the inhibitors.</p><!><p>To confirm mechanism-based inhibition, pre-incubation assays were performed as follows. All assay solution components had the same concentrations as in the above assays. For pre-incubation assays in the presence of NADPH, potassium phosphate buffer (1550 μL of a 0.1 M solution, pH 7.6) was placed in a 1.0 cm quartz cuvette followed by 10 μL of a 1.0 M MgCl2 solution, 10 μL of the microsomal P450 protein, 15 μL of an inhibitor in DMSO (for the control, 15 μL of pure DMSO was added in the place of the inhibitor solution), and 200 μL of a NADPH regenerating solution. The assay mixture was incubated for five minutes at 37°C, before reaction initiation by the addition of 200 μL of buffer and 15 μL of the corresponding substrate solution. The final assay volume was 2.0 mL. The production of 7-hydroxyresorufin anion was monitored at 530 nm excitation and 585 nm emission. The production of 7-hydroxycoumarin was monitored at 338 nm excitation and 458 nm emission. The reactions were performed at 37°C. For each inhibitor, a number of assay runs were performed using varying inhibitor concentrations. For the pre-incubation assays in the absence of NADPH, potassium phosphate buffer (1750 μL of a 0.1 M solution, pH 7.6) was placed in a 1.0 cm quartz cuvette followed by 10 μL of a 1.0 M MgCl2 solution, 10 μL of the microsomal P450 protein, and 15 μL of an inhibitor in DMSO (for the control, 15 μL of pure DMSO was added in the place of the inhibitor solution). The assay mixture was incubated for five minutes at 37°C, before reaction initiation by the addition of 200 μL of the NADPH regenerating solution and 15 μL of the corresponding substrate solution. The final assay volume was 2.0 mL. The production of P450-dependent reaction products were monitored as previously described (10). The reactions were performed at 37°C. For each inhibitor, a number of assay runs were performed using varying inhibitor concentrations.</p><!><p>All of our in silico studies were carried out using the Molecular Operating Environment (MOE) Program (Chemical Computing Group, Montreal, Canada). The coordinates of the template P450 enzymes 1A2 (PDB ID: 2HI4) and 2A6 (PDB ID: 1Z10) were taken from the Protein Data Bank (http://www.rcsb.org). Oxygen atoms representing water were removed and hydrogen atoms were added to the template proteins using Amber99 force field.</p><p>A homology model of P450 1A1 (Figure 1A) was built based on the protein crystal structure of P450 1A2 (2HI4) taken from the Protein Data Bank using the MOE-Homology program (Tripos SYBYL which was used for building 2B1 homology model lacked access to some modules at the time of 1A1 homology modeling and hence MOE was used). The target sequence of human P450 1A1 was obtained from the SwissProt database (21) (P04798). Target sequence was aligned with the sequence of 2HI4 using MOE's multiple sequence and structural alignment algorithm (22) with the structural alignment tool enabled and the blosum62 substitution matrix. The three-dimensional model building of P450 1A1 was performed using the MOE homology program based on a segment matching procedure and a best intermediate algorithm with the option to refine each individual structure enabled. Ten intermediate models were generated and the final model was taken as the cartesian average of all the intermediate models. The heme residue was positioned using the same coordinates as in the template and the complex model was energy minimized. The minimized homology model was validated with PROCHECK(23) and ProsaII (24).</p><p>A homology model of P450 2B1 (Figure 1B) was built based on the protein crystal structures of P450 2B4 (1SUO- mammalian and 2BDM- microsomal) using the molecular modeling package SYBYL (version 7.1, Tripos, St. Louis, MO). The sequence of rat P450 2B1 was used as the target which was obtained from the SwissProt database (21) (P00176). The sequence identity was 73%. The homology model of P450 2B1 was built using Composer in the Biopolymer module of SYBYL. Conformations of loop sequences were automatically generated by the program. The heme residue was positioned using the same coordinates as in the template. Hydrogen atoms were added to the structure and the initial model was refined by energy minimization by the steepest descent method followed by the conjugate gradient method until the root mean square gradient of the potential energy was <0.05 kcal.mol−1A−1. The minimized homology model was validated with PROCHECK and ProsaII.</p><!><p>The structures of the 9 molecules used in this study are given in Figure 2. The 3D structures of the molecules were built using the Tripos SYBYL 7.3 Program. Initial geometric optimizations of the ligands were carried out using the standard MMFF94 force field, with a 0.001 kcal/mol energy gradient convergence criterion and a distance-dependent dielectric constant employing Gasteiger and Marsili charges. Additional geometric optimizations were performed using the semi-empirical method molecular orbital package (MOPAC). The compounds were docked into the binding pockets of P450 enzymes using three programs, GOLD (Cambridge Crystallographic Data Centre), MOE and FlexX (Tripos). The consensus binding postures of the inhibitors were obtained by visual inspection of the docking simulations and their docking scores. The docked complexes were then minimized in three steps using MOE Energy Minimize application. Amber ff99 was used for standard residues of the protein and partial charges were calculated as required for this force field. The non-standard forcefield parameters for heme and the cysteine-iron bond were taken from the literature (25). MOE Energy minimization consists of finding a set of atomic coordinates that correspond to a local minimum of the molecular energy function (we used the potential energy model) by applying large scale non-linear optimization techniques to calculate a conformation (near to the starting geometry) for which the forces on the atoms are zero(26). MOE uses a success of three methods to effect an energy minimization: Steepest Descent (SD), Conjugate Gradient (SG) and Truncated Newton (TN) controlled by the root mean square (RMS) gradient of energy falling below 0.1 kcal A−1. In addition, atoms may be fixed during the calculation. First, hydrogen atom positions were relaxed by holding other atoms fixed. This was followed by allowing all side chain atoms to relax while holding the backbone atoms fixed. Finally, the entire structure was relaxed until root mean square (RMS) gradient of energy was less than 0.1 kcal A−1. Through all of the minimization steps, the planar structure of the heme residue was fixed. The distances between the external carbon of the inhibitor's triple bonds and the heme Fe were determined. The centroids of the phenyl rings of the phenylalanine residues and those of the inhibitor molecules were calculated using the Scientific Vector Language (SVL) module of MOE.</p><!><p>Cytochrome P450s 1A1, 1A2, 2A6 and 2B1 are known to metabolize xenobiotics and ubiquitous environmental carcinogens such as polycyclic aromatic hydrocarbons. Known X-ray crystal structures of P450s 1A2 (PDB ID: 2HI4) (27) and 2A6 (PDB ID: 1Z10) (28) were used in this study. P450s 1A1 (human) and 1A2 (human) show high sequence identity with 73% alignment score. P450s 2B4 (mammalian) and 2B1 (rat) also show high sequence identity with 70% alignment score. Based on the sequence identities of 1A1 with 1A2, and 2B1 with 2B4; homology models for P450s 1A1 and 2B1 were built using the X-ray structures 2HI4.pdb (human) (for 1A2) and 1SUO.pdb (mammalian) and 2BDM.pdb (microsomal) (for 2B1) respectively, as templates. The homology models of P450 1A1 and 2B1 are shown in Figure 1. P450 enzymes share a common overall fold and topology with twelve helices A-L and four β-sheets 1–4, despite less than 20% sequence identity across the gene superfamily. The binding site is defined by the helices E, I, J, K, and L as well as portions of β-sheet 1. The prosthetic heme group is confined between the distal I helix and proximal L helix and bound to the adjacent cysteine in the loop containing the P450 signature amino acid sequence FxxGx(H/R)xCxG. The absolutely conserved cysteine is the proximal or "fifth" ligand to the heme iron (29). The nine inhibitors (Figure 2) used in this study are unique in that they all have 3–4 fused benzene ring systems (polycyclic aromatic hydrocarbons- PAHs), incorporated with an acetylenic moiety in order to achieve mechanistic inhibition of P450 enzymes while lacking any polar group capable of hydrogen bonding with protein residues. The inhibition data for the nine inhibitors measured using human P450 enzymes 1A1, 1A2, and 2A6, and rat enzyme 2B1 are given in Table 1.</p><p>The binding sites of the four P450s of interest were studied to understand the nature of the residues defining the site (Table 2). Aromatic residues were of great interest to us as these residues would interact favorably with the PAH rings of our inhibitors through π–π interactions. Among the four P450 enzymes in our study, P450 1A2 was found to have the largest binding pocket, and P450 2B1 was found to have the smallest binding pocket. All of these enzymes have phenylalanine residues in their binding pockets, with many clustered together (highlighted in bold in Table 2 and shown in Figure 3), capable of making π–π interactions with aromatic rings of the ligand.</p><p>Docking studies were carried out using three different docking programs, GOLD, MOE and FlexX. We were greatly interested in the binding orientation of the molecules that exhibited significant inhibition (i.e., IC50 <10 μM) of the P450 enzymes 1A1, 1A2 and 2B1. Figure 4 depicts the binding of 1EP to P450s 1A1, 1A2 and 2B1 as an example. Two aspects of the ligand docking postures were considered. As the present inhibitors are mechanism-based inhibitors, the position of the triple bond with respect to the Fe atom of the prosthetic heme group of the P450 enzyme is crucial in the oxidation of the triple bond. The distance between the Fe atom of the heme residue and the external carbon of the triple bond was the first aspect for consideration in our docking studies. The second aspect was the effective π stacking of the inhibitors with the various aromatic rings of phenylalanine residues in the active site. The docking results revealed that all of the inhibitors in our study have π–π interactions with the phenylalanine residues in the binding pocket. The distances between the centroids of phenylalanine residue's aromatic rings and the nearest aromatic ring of the ligands were calculated. The distances between the docked ligands and the heme-Fe and phenylalanine aromatic rings for the P450 enzymes 1A1, 1A2, and 2B1 are given in Tables 3A, 3B and 3C respectively.</p><p>It has been shown that terminal acetylenes are metabolized by cytochrome P450 to form a reactive ketene intermediate by a 1,2-hydrogen shift. The ketene intermediate may react with the P450 heme moiety becoming covalently bound to heme nitrogen resulting in a time-dependent destruction of the heme chromophore leading to loss of P450 enzymatic activity (16, 30, 31). The putative ketene intermediate can also react with nucleophilic residues of the protein resulting in covalent binding to the protein with concomitant loss of P450 enzymatic activity (10, 32, 33). Alternatively, excess ketene generated in the active site of the enzyme may react with water to form a carboxylic acid metabolite (16, 30). These mechanistic studies indicate the relevance of proximity of the acetylenic moiety of the inhibitors to the heme Fe atom which is the first aspect of our study.</p><p>Interactions between aromatic moieties also known as π–π interactions are important forces in molecular recognition, and play an important role in controlling the conformations and substrate binding properties of nucleic acids and proteins. The geometry preferences and energetic forces of these interactions have been widely studied in recent years (34, 35). Hunter and coworkers have shown that edge-to-face type orientations and offset stacked orientations are electrostatically attractive geometries while face-to-face and edge-to-edge orientations are unfavorable (36–38). Observations in proteins support the presence of such favorable geometries (37).</p><p>The binding pocket of cytochrome P450 1A1 incorporates four phenylalanine residues at positions 123, 224, 258 and 384. Three of them (Phe123, Phe224, and Phe258) showed favorable interactions with the ligand molecules. The distance between aromatic rings of phenylalanine residues in the native enzyme was 6.0 Å (Phe123-Phe224), 7.7 Å (Phe224-Phe258) and 11.5 Å (Phe123-Phe258). The docked molecules exhibited edge to face interactions with Phe123 and offset stacked interactions with Phe224 and Phe258. The distance between the external carbon of the triple bond and the heme Fe (Table 3A) showed a direct linear correlation with inhibitory activity (Figure 5A). The molecule 1MEP with the closest triple bond to the heme Fe (4.6 Å) showed the most inhibition with an IC50 of 0.02 μM while the molecule 9MEPhen with the farthest triple bond from the heme Fe (11.9 Å) showed the least inhibition with an IC50 of 2.26 μM. The distance between the ligands and centroids of aromatic rings of phenylalanine residues ranged from 4.7 to 5.5 Å for Phe123, 3.7 to 4.9 Å for Phe224, and 3.1 to 7.3 Å for Phe258 (Table 3A). Even though the aromatic rings of phenylalanine residues had π–π interactions with the ligand molecules, they did not show any clear correlation to the activity profile. It is clear that the cytochrome P450 1A1 inhibition potency of these molecules depends on the distance of the triple bond from the heme Fe. The π stacking of the polycyclic aromatic core of the molecules with the aromatic rings of phenylalanine residues did not contribute to the differences in the potency of these inhibitors. The terminal methyl substituted aryl acetylenes (1MEP and 2MEPhen) showed higher inhibition potency due to many favorable interactions with the protein such as shorter distance of the triple bond to the heme Fe, effective offset π stacking distance with Phe224 and Phe258, and the additional fact that the methyl group fits perfectly in the hydrophobic region of the pocket defined by amino acid residues Val382 and Ile386 (Figure 6A).</p><p>The active site of cytochrome P450 1A2 depicts the presence of five phenylalanine residues Phe125, Phe226, Phe256, Phe260 and Phe319. The aromatic rings of Phe125, Phe226, Phe256 and Phe260 were clustered together in the binding pocket with the rings pointing inwards, whereas the Phe319 aromatic ring pointed outwards. Hence the contribution to π stacking with the inhibitors could be seen only from the first four phenylalanine residues. The distances between these aromatic rings in the native 1A2 enzyme is 5.9 Å for Phe125-Phe226, 6.5 Å for Phe226-Phe256, 5.8 Å for Phe256-260 and 6.9 Å for Phe226-Phe260. The distance between the external carbon of the ligand molecules and the heme Fe ranged between 4.3 Å and 14.6 Å (Table 3B). This distance did not show any clear relationship with the variation in inhibitory activity of the molecules. The distances between the centroids of the aromatic rings of the phenylalanine residues and the ligands were 5.3–7.9 Å for Phe125, 3.2–4.2 Å for Phe226, 5.7–8.7 Å for Phe256 and 4.2–5.6 Å for Phe260 (Table 3B). The variation in the ligand-phenylalanine aromatic ring distance was minimal for Phe125 and Phe226, implying that even though it can be a positive factor towards effective binding, it does not account for the differences in inhibition potency. Distances between the ligand and the phenylalanines Phe256 and Phe260 indeed show a linear relationship to their inhibition potency (Figure 5B). The data points that did not fit in the linear relationship for Phe256 and Phe260 had shorter distances to the heme Fe atom which factored into the increase in potency. These results show that a good π–π interaction of the ligand with all four of the phenylalanines increases the inhibition potency of these molecules towards P450 enzyme 1A2. This when combined with a shorter distance of the triple bond from the heme Fe should result in higher inhibition potency. 1MEP showed the highest potency among the ligands as its triple bond was one of the closest to the heme Fe, it had favorable π–π interactions with the phenylalanine residues lining the pocket, and the methyl group was positioned in the hydrophobic region of the pocket defined by protein residues Leu382, Ile386 and Leu497 (Figure 6B).</p><p>The binding pocket of P450 2B1 was the smallest of all the P450 enzymes studied. It had two phenylalanine residues Phe115 and Phe297. Phe297 made an edge-to-face interaction with the inhibitors whereas Phe115 made an edge-to-edge π interaction with the inhibitors. The distances between the centroid of the aromatic rings of phenylalanine residues and the nearest aromatic ring of the inhibitors ranged from 4.5–5.2 Å for Phe297, and 6.1–7.3 Å for Phe115 (Table 3C). The distance between the heme Fe and the ligand external carbon of the triple bond varied from 3.2 Å to 11 Å and showed a linear correlation to the inhibition potency (Figure 5C). The inhibitors did not show significant difference in their π stacking interactions with the phenylalanine residues. The molecule 1EP showed the highest inhibition which could be attributed to its triple bond being closest to heme Fe atom (3.2 Å).</p><p>The binding pocket of P450 enzyme 2A6 consists of many phenylalanine residues (Phe107, Phe11, Phe118, Phe209 and Phe480) that could potentially interact favorably with the inhibitors. However, the shape of the binding pocket and the orientation of many of the phenylalanine residues in the binding pocket cause distortion in the planar inhibitors when docked in the binding pockets. This is a highly unfavorable situation which is reflected by the absence of inhibitory activity of these ligands towards cytochrome P450 2A6.</p><!><p>The polyaromatic hydrocarbon inhibitors studied here do not have any polar substituents capable of hydrogen bonding with the polar residues of P450s 1A1, 1A2, 2A6 and 2B1. The binding pockets of all four target P450 enzymes contain aromatic residues that can form π–π interactions with the aromatic rings of the inhibitors. In P450s 1A1 and 2B1, the distance between the triple bond and the heme Fe is a crucial factor in determining the potency of the inhibitor. In P450 enzyme 1A2, the π–π interactions with the four phenylalanine residues play a more dominating role in determining the potency of the inhibitors. Further more favorable positioning of the hydrophobic groups of the inhibitor's side chains positioned in the hydrophobic region of the binding pocket further increases the potency of inhibition. Through these studies, we have gained valuable insight into the factors governing the potency of the inhibition of cytochrome P450s 1A1, 1A2, 2A6 and 2B1 by polyaromatic hydrocarbons. This knowledge will help us in designing inhibitors with increased potency and selectivity for these enzymes.</p><!><p>3D structures of P450s 1A1 (A) and 2B1 (B) obtained from homology modeling. The protein is shown as a ribbon model and the Heme residue is depicted as stick model.</p><p>Structures of arylacetylene inhibitors used for the study</p><p>Depicts binding pockets of P450 enzymes 1A1 (A), 1A2 (B), 2A6 (C) and 2B1 (D) with heme and Phenylalanine residues as stick models showing Gaussian contact molecular surfaces in green dots.</p><p>Illustrates the π–π interactions between the protein and the ligand and the proximity of the triple bond to the heme Fe atom for the representative molecule 1EP docked in the binding pockets of P45 enzymes 1A1 (A), 1A2 (B) and 2B1 (C). The residues and inhibitor are shown as stick models.</p><p>(A), (B), and (C) Relationship between the inhibitor concentration and Heme-C distance for P450 enzymes 1A1, 1A2 and 2B1; (D), (E), and (F) Relationship between the inhibitor concentration and distance between centroids of aromatic rings of Phenylalanine residues and inhibitor for P450 enzymes 1A1, 1A2 and 2B1.</p><p>(A) Represents binding modes of most potent inhibitors 1MEP and 2EPHEN for P450 enzyme 1A1; (B) represents binding modes of most potent inhibitor 1MEP for P450 enzyme 1A2. All interacting residues and the inhibitors are shown as stick models.</p><p>Inhibition data in supersomes for arylacetylenes measured using human P450 enzymes (1A1, 1A2, 2A6) and rat P450 enzyme (2B1).</p><p>Properties of the P450 enzyme binding pocket</p><p>Tripos SYBYL (SiteID Pockets) was used to identify the binding cavity by initiating grid-based algorithm for locating pockets. This works well for finding narrow and/or solvent inaccessible pockets. The volume of the binding cavity is also given by this algorithm.</p><p>Distances between Heme Fe, Phe residues and the ligand moieties for P450 1A1 (A), 1A2 (B) and 2B1 (C).</p><p>Heme-C distances represent the distance between the Heme Fe and the external carbon of the triple bond.</p><p>Distance between the centroids of the Phenyl ring of Phenylalanine residues and the nearest aromatic ring of the inhibitor.</p>

PubMed Author Manuscript

Synthesis and in vitro anti-proliferative activity of some novel isatins conjugated with quinazoline/phthalazine hydrazines against triple-negative breast cancer MDA-MB-231 cells as apoptosis-inducing agents

AbstractTreatment of patients with triple-negative breast cancer (TNBC) is challenging due to the absence of well- defined molecular targets and the heterogeneity of such disease. In our endeavor to develop potent isatin-based anti-proliferative agents, we utilized the hybrid-pharmacophore approach to synthesize three series of novel isatin-based hybrids 5a–h, 10a–h and 13a–c, with the prime goal of developing potent anti-proliferative agents toward TNBC MDA-MB-231 cell line. In particular, compounds 5e and 10g were the most active hybrids against MDA-MB-231 cells (IC50 = 12.35 ± 0.12 and 12.00 ± 0.13 μM), with 2.37- and 2.44-fold increased activity than 5-fluorouracil (5-FU) (IC50 = 29.38 ± 1.24 μM). Compounds 5e and 10g induced the intrinsic apoptotic mitochondrial pathway in MDA-MB-231; evidenced by the reduced expression of the anti-apoptotic protein Bcl-2, the enhanced expression of the pro-apoptotic protein Bax and the up-regulated active caspase-9 and caspase-3 levels. Furthermore, 10g showed significant increase in the percent of annexin V-FITC positive apoptotic cells from 3.88 to 31.21% (8.4 folds compared to control).

synthesis_and_in_vitro_anti-proliferative_activity_of_some_novel_isatins_conjugated_with_quinazoline

Introduction<!><!>Introduction<!><!>Introduction<!>Chemistry<!>General procedure for preparation of 4-chloroquinazolines 2a,b<!>General procedure for preparation of 4-hydrazinyl-2-substituted quinazoline 3a,b<!>General procedure for preparation of the targetisatin-quinazoline hybrids 5a–h<!>1-Chloro-4-phenylphthalazines 8a,b<!>General procedure for preparation of 1-hydrazinyl-4-phenylphthalazines 9a,b<!>General procedure for preparation of the target isatin-phthalazine hybrids (10a–h)<!>N-Benzylindoline-2,3-diones12a,b<!>General procedure for preparation of the target hybrids (13a-c)<!>In vitro anti-proliferative activity<!><!>ELISA immunoassay<!>Annexin V-FITC apoptosis<!>Statistical analysis<!>ADME profiling<!>Chemistry<!><!>Chemistry<!>In vitro anti-proliferative activity against TNBC MDA-MB-231 cell line<!>Structure activity relationship SAR<!>Effects on mitochondrial apoptosis pathway proteins Bax andBcl-2<!><!>Effects on the level of active caspase-9 and caspase-3<!><!>Effects on the level of active caspase-9 and caspase-3<!>Annexin V–FITC apoptosis assay<!><!>Annexin V–FITC apoptosis assay<!>In vitro anti-proliferative activity against four different cell lines<!><!>In vitro anti-proliferative activity against four different cell lines<!>ADME study<!><!>Conclusion

<p>Heading the list of the critical health problems for females, breast cancer continues to be the major life-threatening health issue. According to the recent records, an estimated 1.7 million women are expected to be diagnosed with breast cancer by 2020 with a make-up of 26% increase from current recorded levels1,2. Despite the huge and continuous efforts of the scientific community, no tangible improvements were recorded in the last decade. With the development of resistance toward the available anticancer drugs, the mission becomes even more challenging and the efficacious therapy remains questionable3,4.</p><p>Targeting breast cancer with tailored drugs depends on the fact of immunohistochemical expression of estrogen receptors (ERs), progesterone receptors (PRs) and human epidermal growth factor receptors-2 (HER-2). Being devoid of the three aforementioned targetable proteins, triple-negative breast cancer (TNBC) stands even as an unbeatable challenge5. Besides their well-known resistance to the current chemotherapy, TNBC is an aggressive phenotype of cancer with high recurrence risk, poor prognosis, reduced survival and distant metastasis that it may reach the lungs and even the central nervous system (CNS)5,6. Accordingly, patients largely receive systemic chemotherapy that puts them under higher risk of suffering the devastating side effects. This urgent necessity motivates medicinal chemists to search for alternative treatments targeting apoptotic machinery as a novel approach for TNBC therapy.</p><p>Targeting critical regulators of apoptosis with the goal of inducing apoptosis in cancer cells still stands as an attractive and successful strategy to drug discovery and development of new anti-cancer agents7–11. On April 11, 2016, the U.S. food and drug administration (FDA) approved Venetoclax (Venclexta®) for the treatment of patients with chronic lymphocytic leukemia (CLL) whose tumors have a specific genetic alteration. Venclexta is the first FDA-approved drug that targets the BcL-2 protein, an anti-apoptotic regulatory protein12.</p><p>Molecular hybridization is a contemporary concept in drug design and development gaining momentum worldwide. There are two main ways that can be explored to design and construct affordable and efficient hybrid molecules. The first method merges haptophoric moieties of different drugs and the second technique combines two or more entire drugs together using a linker chain13,14. This strategy affords novel molecule hybrids with improved affinity and efficacy when compared to the parent drugs, modified selectivity profile, different and/or dual modes of action and reduced undesired side effects. Thus, hybrid molecules emerged as magic bullets that trigger two or more cytocidal pharmacological mechanisms of action acting in synergy to inhibit cancer tumor growth15–17.</p><p>Pertaining to its wide presence endogenously in both human and other mammalian tissues and fluids besides its prevalence in many naturally occurring compounds, notably alkaloids, fungal metabolites and marine natural products, isatin (1H-indole-2,3-dione) as a privileged scaffold is endowed with excellent anticancer profile18. Thus, medicinal chemists embarked on exploring diverse isatin derivatives comprehending their potential to create novel bioactive compounds. By 2006, the FDA approval of Sunitinib (Sutent®) I (Figure 1) for the treatment of advanced renal carcinoma and gastrointestinal stromal tumors furnished the path for the creation of new isatin derivatives as drug targets19. Nintedanib (Ofev®) II (Figure 1), an orally available triple angiokinase inhibitor, received its first global approval in the US in October 2014 for the treatment of idiopathic pulmonary fibrosis (IPF)20. By 2015, Nintedanib II has been approved by the European medicines agency (EMA) in Europe, as a second-line treatment in combination with docetaxel of locally advanced, metastatic or locally recurrent non-small cell lung cancer of adenocarcinoma20.</p><!><p>Structures of some isatin-, quinazoline- and phthalazine-based I–VI approved anticancer drugs.</p><!><p>On the other hand, quinazolines constitute a leading class of heterocycles that displayed interesting biological activities, chiefly anticancer activity. Raltitrexed (Tomudex®) III, Afatinib (Gilotrif®) IV and Gefitinib (Iressa®) V (Figure 1) are examples for the clinically approved quinazoline-based anticancer drugs21. Besides, phthalazine nucleus has emerged as a promising and attractive one in the development of novel anticancer agents22–24. Olaparib (Lynparza®) VI (Figure 1) is an oral small molecule phthalazine-based poly ADP-ribose polymerase (PARP) inhibitor being developed for the treatment of solid tumors. In December 2014, Olaparib was approved in the EU and USA for the treatment of BRCA-mutated ovarian cancer25.</p><p>Surveying the literature revealed that different isatin26–30, 2-phenylquinazoline31–35 and 4-arylphthalazine23,36–38 derivatives were developed with significant activity against the TNBC MDA-MB-231 cell line. Figure 2 displays some of these derivatives with their IC50 values or growth inhibition (GI) percentage against MDA-MB-231 cells.</p><!><p>Structures of some reported isatins, quinazolines and phthalazines VII–XVII with anti-proliferative activity against the triple-negative breast cancer MDA-MB-231 cells, and structures of the target hybrids 5a–h, 10a–h and 13a–c.</p><!><p>In light of the above findings and as a part of our ongoing effort to develop potent isatin-based anti-proliferative agents39–45, we utilized the hybrid pharmacophore approach to design and synthesize three different series of novel isatin-based hybrids 5a–h, 10a–h and 13a–c (Figure 2), via merging the pharmacophoric elements of isatin and phthalazine or quinazoline in a single chemical framework through a hydrazine linker, with the prime goal of developing potent anti-proliferative agents against the TNBC MDA-MB-231 cell line. The three synthesized series were in vitro evaluated for their potential anti-proliferative activity against TNBC MDA-MB-231 cell line. Compounds 5e and 10g were further estimated for their apoptosis induction potential in MDA-MB-231 cells, to gain mechanistic insights into the anti-proliferative activity of the prepared hybrids. Furthermore, the activities against A549 alveolar carcinoma, Caco-2 colon cancer, LoVo human colorectal carcinoma and HepG2 hepatocellular carcinoma were in vitro examined.</p><!><p>Melting points were measured with a Stuart melting point apparatus (Bibby Scientific Limited, Staffordshire, UK) and were uncorrected. Infrared spectra were recorded as potassium bromide discs on Schimadzu FT-IR 8400S spectrophotometer (Shimadzu Scientific Instruments, Kyoto, Japan) and expressed in wave number (cm−1). The NMR spectra were recorded on Varian Gemini-300BB 300 MHz FT-NMR spectrometers (Varian Inc., Palo Alto, CA). 1H and 13C spectra were run at 300 and 75 MHz, respectively, in deuterated dimethyl sulfoxide (DMSO-d6). Chemical shifts (δH) are reported relative to tetramethylsilane (TMS) as internal standard. All coupling constant (J) values are given in hertz. Chemical shifts (δC) are reported relative to DMSO-d6 as internal standards. The abbreviations used are as follows: s, singlet; d, doublet; m, multiplet. Elemental analyses were carried out at the Regional Center for Microbiology and Biotechnology, Al-Azhar University, Cairo, Egypt. Reaction courses and product mixtures were routinely monitored by thin layer chromatography (TLC) on silica gel precoated F254 Merck plates (Merck Group, Darmstadt, Germany). Unless otherwise noted, all solvents and reagents were commercially available and used without further purification.</p><!><p>To a cooled stirred mixture of phosphorus oxychloride (30 mL) and N,N-dimethylaniline (1 mL), 2-Substituted-quinazolin-4(3H)-ones 1a,b (10 mmol) were added portion-wise. After refluxing for 6 h, the reaction mixture was cooled then poured onto ice-water and alkalinized with 2 N NaOH. The aqueous solution was extracted with methylene chloride. The combined organic layer was dried over anhydrous Na2SO4 then filtered, and the solvent was removed under reduced pressure. The obtained solid was crystallized from isopropanol to afford compounds 2a,b46.</p><!><p>To a magnetically stirred solution of 4-chloroquinazolines 2a,b (5 mmol) in ethanol (15 mL), hydrazine hydrate 99% (2.5 mL, 50 mmol) was added. The stirring was continued at refluxing temperature for 4 h. Upon cooling, the obtained precipitate was filtered, washed with methanol and water, dried and recrystallized from ethanol to furnish compounds 3a,b.</p><p>2–(2,6-Dichlorophenyl)-4-hydrazinylquinazoline (3b). White crystals (yield 86%), m.p. 237–239 °C; IR (KBr, ν cm−1): 3295 (NH, NH2) cm−1; 1H NMR (DSMO-d6) δ ppm: 4.77 (s, 2H, NH2, D2O exchangeable), 7.51–7.54 (m, 3H, Ar–H), 7.69–7.87 (m, 3H, Ar–H), 8.21 (d, 1H, Ar–H, J = 8.1 Hz), 9.77 (s, 1H, NH, D2O exchangeable); Anal. Calcd. for C14H10Cl2N4:C, 55.10; H, 3.30; N, 18.36; Found C, 54.89; H, 3.33; N, 18.41.</p><!><p>To a stirred solution of hydrazine derivatives 3a,b (1 mmol), in absolute ethyl alcohol, the appropriate isatin 4a–d (1 mmol) and catalytic amount of glacial acetic acid were added. The mixture was heated under reflux for 0.5 h, filtered while hot. The obtained solid was washed with ethanol collected and crystallized from an ethanol/DMF mixture to obtain hybrids 5a–h.</p><p>3–(2–(2–(4-Chlorophenyl)quinazolin-4-yl)hydrazono)indolin-2-one (5a). Orange powder (yield 75%), m.p. 291–292 °C; IR (KBr, ν cm−1): 3210 (NH) and 1675 (C = O); 1H NMR (300 MHz, DMSO-d6) δ ppm: 6.98 (d, 1H, Ar–H, J = 7.8 Hz), 7.14 (t, 1H, Ar–H, J = 7.5 Hz), 7.38 (t, 1H, Ar–H, J = 7.8 Hz), 7.61 (d, 2H, H-3 and H-5 of 4-Cl-C6H4, J = 8.4 Hz), 7.69–7.76 (m, 3H, Ar–H), 7.98–8.00 (m, 2H, Ar–H), 8.54 (d, 2H, H-2 and H-6 of 4-Cl-C6H4, J = 8.7 Hz), 11.38 (s, 1H, NH, D2O exchangeable), 13.89 (s, 1H, NH, D2O exchangeable); 13C NMR (75 MHz, DMSO-d6) δ: 111.23, 112.70, 120.00, 120.62, 122.63, 123.37, 127.46, 128.49, 128.66, 129.72, 131.19, 134.11, 135.48, 136.24, 141.83, 151.51, 156.15, 157.99, 163.29 (C = O); Anal. Calcd. for C22H14ClN5O: C, 66.09; H, 3.53; N, 17.52; Found C, 66.21; H, 3.49; N, 17.45.</p><p>3–(2–(2–(4-Chlorophenyl)quinazolin-4-yl)hydrazono)-5-fluoroindolin-2-one (5b). Orange powder (yield 77%), m.p. >300 °C; IR (KBr, ν cm−1): 3190 (NH) and 1690 (C = O); 1H NMR (300 MHz, DMSO-d6) δ ppm: 6.95 (dd, 1H, Ar–H, J = 4.2, 8.7 Hz), 7.19 (t, 1H, Ar–H, J = 8.7 Hz), 7.59–7.63 (m, 3H, Ar–H), 7.71–7.77 (m, 2H, Ar–H), 7.96 (d, 2H, Ar–H, J = 7.8 Hz), 8.53 (d, 2H, H-2 and H-6 of 4-Cl-C6H4, J = 8.4 Hz), 11.39 (s, 1H, NH, D2O exchangeable), 13.86 (s, 1H, NH, D2O exchangeable); 13C NMR (75 MHz, DMSO-d6) δ: 107.93, 112.72, 116.75, 121.22, 127.43, 128.54, 129.78, 133.67, 134.29, 135.54, 136.16, 137.95, 138.07, 151.61, 156.13, 157.95, 163.44, 167.34 (C = O); Anal. Calcd. for C22H13ClFN5O: C, 63.24; H, 3.14; N, 16.76; Found C, 63.39; H, 3.11; N, 16.68.</p><p>5-Chloro-3–(2–(2–(4-chlorophenyl)quinazolin-4-yl)hydrazono)indolin-2-one (5c). Orange powder (yield 70%), m.p. > 300 °C; IR (KBr, ν cm−1): 3285 (NH) and 1672 (C = O); 1H NMR (300 MHz, DMSO-d6) δ ppm: 6.98 (d, 1H, Ar–H, J = 8.4 Hz), 7.41 (d, 1H, Ar–H, J = 8.4 Hz), 7.61 (d, 2H, H-3 and H-5 of 4-Cl-C6H4, J = 8.4 Hz), 7.69–7.77 (m 3H, Ar–H), 7.97 (d, 2H, Ar, J = 7.8 Hz), 8.53 (d, 2H, H-2 and H-6 of 4-Cl-C6H4, J = 8.7 Hz), 11.49 (s, 1H, NH, D2O exchangeable), 13.86 (s, 1H, NH, D2O exchangeable); 13C NMR (75 MHz, DMSO-d6) δ: 112.68, 119.95, 120.10, 121.67, 123.47, 126.75, 127.50, 128.49, 128.68, 129.72, 134.29, 135.51, 136.11, 140.41, 151.58, 156.02, 157.90, 163.08 (C = O); Anal. Calcd. for C22H13Cl2N5O: C, 60.85; H, 3.02; N, 16.13; Found C, 60.97; H, 2.98; N, 16.05.</p><p>5-Bromo-3–(2–(2–(4-chlorophenyl)quinazolin-4-yl)hydrazono)indolin-2-one (5d). Orange powder (yield 77%), m.p. >300 °C; IR (KBr, ν cm−1): 3245 (NH) and 1679 (C = O); 1H NMR (300 MHz, DMSO-d6) δ ppm: 6.94 (d, 1H, Ar–H, J = 8.1 Hz), 7.54 (dd, 1H, Ar–H, J = 2.1, 8.1 Hz), 7.61 (d, 2H, H-3 and H-5 of 4-Cl-C6H4, J = 8.7 Hz), 7.69–7.79 (m 2H, Ar–H), 7.87 (s, 1H, Ar–H), 7.98 (d, 2H, Ar–H, J = 3.9 Hz), 8.56 (d, 2H, H-2 and H-6 of 4-Cl-C6H4, J = 8.7 Hz), 11.49 (s, 1H, NH, D2O exchangeable), 13.88 (s, 1H, NH, D2O exchangeable); Anal. Calcd. for C22H13BrClN5O: C, 55.20; H, 2.74; N, 14.63; Found C, 55.33; H, 2.71; N, 14.54.</p><p>3–(2–(2–(2,6-Dichlorophenyl)quinazolin-4-yl)hydrazono)indolin-2-one (5e). Orange powder (yield 65%), m.p. > 300 °C; IR (KBr, ν cm−1): 3170 (NH) and 1683 (C = O); 1H NMR (300 MHz, DMSO-d6) δ ppm: 6.88 (d, 1H, Ar–H, J = 7.5 Hz), 7.08 (t, 1H, Ar–H, J = 7.5 Hz), 7.31 (t, 1H, Ar–H, J = 7.5 Hz), 7.58–7.84 (m, 5H, Ar–H), 8.01 (s, 1H, Ar–H), 8.42 (d, 2H, Ar–H, J = 7.5 Hz), 10.62, 11.38 (s, 1H, NH, D2O exchangeable), 12.11, 13.93 (s, 1H, NH, D2O exchangeable); Anal. Calcd. for C22H13Cl2N5O: C, 60.85; H, 3.02; N, 16.13; Found 60.97; H, 2.99; N, 16.03.</p><p>3–(2–(2–(2,6-Dichlorophenyl)quinazolin-4-yl)hydrazono)-5-fluoroindolin-2-one (5f). Orange powder (yield 70%), m.p. > 300 °C; IR (KBr, ν cm−1): 3185 (NH) and 1675 (C = O); 1H NMR (300 MHz, DMSO-d6) δ ppm: 6.85 (dd, 1H, Ar–H, J = 4.5, 8.7 Hz), 7.20 (t, 1H, Ar–H, J = 8.7 Hz), 7.55–7.87 (m, 6H, Ar–H), 8.03 (s, 1H, Ar–H), 8.45 (d, 1H, Ar–H, J = 8.4 Hz), 10.65, 11.39 (s, 1H, NH, D2O exchangeable), 12.25, 13.91 (s, 1H, NH, D2O exchangeable); 13C NMR (75 MHz, DMSO-d6) δ: 112.37, 117.99, 121.07, 123.75, 127.34, 128.35, 129.42, 132.76, 134.26, 135.28, 136.92, 138.10, 139.76, 146.37, 151.24, 155.92, 156.80, 158.99, 159.42, 159.95, 163.40, 165.44 (C = O); Anal. Calcd. for C22H12Cl2FN5O: C, 58.43; H, 2.67; N, 15.49; Found C, 58.29; H, 2.71; N, 15.58.</p><p>5-Chloro-3–(2–(2–(2,6-dichlorophenyl)quinazolin-4-yl)hydrazono)indolin-2-one (5g). Orange powder (yield 75%), m.p. >300 °C; IR (KBr, ν cm−1): 3203 (NH) and 1688 (C = O); 1H NMR (300 MHz, DMSO-d6) δ ppm: 6.89 (d, 1H, Ar–H, J = 8.4 Hz), 7.37 (t, 1H, Ar–H, J = 8.7 Hz), 7.58 (d, 1H, Ar–H, J = 8.4 Hz), 7.72–7.88 (m, 5H, Ar–H), 8.01 (s, 1H, Ar–H), 8.39 (d, 1H, Ar–H, J = 8.4 Hz), 10.75, 11.48 (s, 1H, NH, D2O exchangeable), 12.37, 13.85 (s, 1H, NH, D2O exchangeable); Anal. Calcd. for C22H12Cl3N5O: C, 56.37; H, 2.58; N, 14.94; Found C, 56.51; H, 2.53; N, 14.82.</p><p>5-Bromo-3–(2–(2–(2,6-dichlorophenyl)quinazolin-4-yl)hydrazono)indolin-2-one (5h). Orange powder (yield 72%), m.p. >300 °C; IR (KBr, ν cm−1): 3305 (NH) and 1670 (C = O); 1H NMR (300 MHz, DMSO-d6) δ ppm: 6.84 (d, 1H, Ar–H, J = 8.7 Hz), 7.49–8.02 (m, 7H, Ar–H), 8.36 (d, 1H, Ar–H, J = 8.4 Hz), 8.53 (s, 1H, Ar–H), 10.76, 11.48 (s, 1H, NH, D2O exchangeable), 12.15, 13.85 (s, 1H, NH, D2O exchangeable); Anal. Calcd. for C22H12BrCl2N5O: C, 51.49; H, 2.36; N, 13.65; Found C, 51.30; H, 2.39; N, 13.71.</p><!><p>Compounds 8a,b were prepared according to the literature procedure47,48.</p><!><p>Compounds 9a,b were prepared according to the literature procedure49.</p><!><p>Hydrazino phthalazines 9a,b (1 mmol) and catalytic amount of glacial acetic acid were added to a stirred solution of indoline-2,3-dione 4a–d (1 mmol) in refluxing absolute ethyl alcohol, then the reaction mixture was refluxed for 1 h. The precipitated solid was filtered while hot, dried and recrystallized from ethanol/DMF mixture to afford the target hybrids 10a–h.</p><p>3–(2–(4-Phenylphthalazin-1-yl)hydrazono)indolin-2-one (10a). Orange powder (yield 65%), m.p. > 300 °C; IR (KBr, ν cm−1): 3183 (NH) and 1677 (C = O); 1H NMR (300 MHz, DMSO-d6) δ ppm: 6.87 (d, 1H, H-7 isatin, J = 7.8 Hz), 7.05 (t, 1H, H-5 isatin, J = 7.5 Hz), 7.29 (t, 1H, H-6 isatin, J = 7.5 Hz), 7.56–7.63 (m, 5H, H-2, H-3, H-4, H-5 & H-6 of C6H5), 7.68 (d, 1H, H-8 phthalazine, J = 8.1 Hz), 7.89 (t, 1H, H-6 phthalazine, J = 7.8 Hz), 7.96 (t, 1H, H-7 phthalazine, J = 7.8 Hz), 8.48 (d, 1H, H-4 isatin, J = 7.8 Hz), 8.68 (d, 1H, H-5 phthalazine, J = 7.8 Hz), 10.56 (s, 1H, NH, D2O exchangeable), 12.83 (s, 1H, NH, D2O exchangeable); Anal. Calcd. for C22H15N5O: C, 72.32; H, 4.14; N, 19.17; Found C, 72.44; H, 4.11; N, 19.12.</p><p>5-Fluoro-3–(2–(4-phenylphthalazin-1-yl)hydrazono)indolin-2-one (10b). Orange powder (yield 60%), m.p. >300 °C; IR (KBr, ν cm−1): 3234 (NH) and 1677 (C = O); 1H NMR (300 MHz, DMSO-d6) δ ppm: 6.86 (dd, 1H, H-6 isatin, J = 4.2, 8.7 Hz), 7.14 (t, 1H, H-7 isatin, J = 8.4 Hz), 7.57–7.64 (m, 5H, H-2, H-3, H-4, H-5 & H-6 of C6H5), 7.71 (d, 1H, H-8 phthalazine, J = 7.8 Hz), 7.91 (t, 1H, H-6 phthalazine, J = 7.8 Hz), 7.96 (t, 1H, H-7 phthalazine, J = 7.8 Hz), 8.20 (dd, 1H, H-4 isatin, J = 3.0, 8.4 Hz), 8.68 (d, 1H, H-5 phthalazine, J = 7.5 Hz), 10.58 (s, 1H, NH, D2O exchangeable), 13.04 (s, 1H, NH, D2O exchangeable); 13C NMR (75 MHz, DMSO-d6) δ: 109.52, 113.88, 118.19, 123.64, 126.08, 126.62, 127.00, 128.54, 129.37, 132.69, 133.78, 134.55, 139.06, 143.32, 149.31, 151.86, 155.86, 158.98, 165.87, 171.92; Anal. Calcd. for C22H14FN5O: C, 68.92; H, 3.68; N, 18.27; Found C, 68.77; H, 3.73; N, 18.33.</p><p>5-Chloro-3–(2–(4-phenylphthalazin-1-yl)hydrazono)indolin-2-one (10c). Orange powder (yield 72%), m.p. >300 °C; IR (KBr, ν cm−1): 3195 (NH) and 1679 (C = O); 1H NMR (300 MHz, DMSO-d6) δ ppm: 6.89 (d, 1H, H-6 isatin, J = 8.4 Hz), 7.34 (d, 1H, H-7 isatin, J = 8.4 Hz), 7.57–7.64 (m, 5H, H-2, H-3, H-4, H-5 & H-6 of C6H5), 7.72 (d, 1H, H-8 phthalazine, J = 8.1 Hz), 7.92 (t, 1H, H-6 phthalazine, J = 8.1 Hz), 8.02 (t, 1H, H-7 phthalazine, J = 8.1 Hz), 8.45 (s, 1H, H-4 isatin), 8.62 (d, 1H, H-5 phthalazine, J = 8.1 Hz), 10.68 (s, 1H, NH, D2O exchangeable), 13.03 (s, 1H, NH, D2O exchangeable); Anal. Calcd. for C22H14ClN5O: C, 66.09; H, 3.53; N, 17.52; Found C, 66.21; H, 3.49; N, 17.42.</p><p>5-Bromo-3–(2–(4-phenylphthalazin-1-yl)hydrazono)indolin-2-one (10d). Orange powder (yield 75%), m.p. >300 °C; IR (KBr, ν cm−1): 3315 (NH) and 1672 (C = O); 1H NMR (300 MHz, DMSO-d6) δ ppm: 6.85 (d, 1H, H-6 isatin, J = 8.4 Hz), 7.47 (d, 1H, H-7 isatin, J = 8.1 Hz), 7.57–7.64 (m, 5H, H-2, H-3, H-4, H-5 & H-6 of C6H5), 7.72 (d, 1H, H-8 phthalazine, J = 8.4 Hz), 7.92 (t, 1H, H-6 phthalazine, J = 8.1 Hz), 8.01 (t, 1H, H-7 phthalazine, J = 8.1 Hz), 8.59–8.62 (m, 2H, H-4 isatin and H-5 phthalazine), 10.70 (s, 1H, NH, D2O exchangeable), 13.06 (s, 1H, NH, D2O exchangeable); Anal. Calcd. for C22H14BrN5O: C, 59.47; H, 3.18; N, 15.76; Found C, 59.61; H, 3.14; N, 15.68.</p><p>3–(2–(4–(4-Chlorophenyl)phthalazin-1-yl)hydrazono)indolin-2-one (10e). Yellow powder (yield 70%), m.p. >300 °C; IR (KBr, ν cm−1): 3218 (NH) and 1676 (C = O); 1H NMR (300 MHz, DMSO-d6) δ ppm: 6.87 (d, 1H, H-7 isatin, J = 7.8 Hz), 7.05 (t, 1H, H-5 isatin, J = 7.5 Hz), 7.29 (t, 1H, H-6 isatin, J = 7.5 Hz), 7.60–7.65 (m, 4H, H-2, H-3, H-5 & H-6 of 4-Cl-C6H5), 7.67 (d, 1H, H-8 phthalazine, J = 7.8 Hz), 7.89 (t, 1H, H-6 phthalazine, J = 7.8 Hz), 7.96 (t, 1H, H-7 phthalazine, J = 7.8 Hz), 8.47 (d, 1H, H-4 isatin, J = 7.2 Hz), 8.68 (d, 1H, H-5 phthalazine, J = 7.5 Hz), 10.63 (s, 1H, NH, D2O exchangeable), 13.04 (s, 1H, NH, D2O exchangeable); Anal. Calcd. for C22H14ClN5O: C, 66.09; H, 3.53; N, 17.52; Found C, 65.89; H, 3.58; N, 17.61.</p><p>3–(2–(4–(4-Chlorophenyl)phthalazin-1-yl)hydrazono)-5-fluoroindolin-2-one (10f). Orange powder (yield 73%), m.p. >300 °C; IR (KBr, ν cm−1): 3247 (NH) and 1670 (C = O); 1H NMR (300 MHz, DMSO-d6) δ ppm: 6.86 (dd, 1H, H-6 isatin, J = 4.2, 8.1 Hz), 7.13 (t, 1H, H-7 isatin, J = 8.1 Hz), 7.64–7.66 (m, 4H, H-2, H-3, H-5 & H-6 of 4-Cl-C6H5), 7.70 (d, 1H, H-8 phthalazine, J = 8.1 Hz), 7.92 (t, 1H, H-6 phthalazine, J = 7.8 Hz), 8.01 (t, 1H, H-7 phthalazine, J = 7.8 Hz), 8.20 (dd, 1H, H-4 isatin, J = 3.0, 8.4 Hz), 8.66 (d, 1H, H-5 phthalazine, J = 7.8 Hz), 10.58 (s, 1H, NH, D2O exchangeable), 13.05 (s, 1H, NH, D2O exchangeable); Anal. Calcd. for C22H13ClFN5O: C, 63.24; H, 3.14; N, 16.76; Found C, 63.08; H, 3.09; N, 16.82.</p><p>5-Chloro-3–(2–(4–(4-chlorophenyl)phthalazin-1-yl)hydrazono)indolin-2-one (10g). Orange powder (yield 70%), m.p. >300 °C; IR (KBr, ν cm−1): 3195 (NH) and 1673 (C = O); 1H NMR (300 MHz, DMSO-d6) δ ppm: 6.90 (d, 1H, H-6 isatin, J = 8.1 Hz), 7.35 (d, 1H, H-7 isatin, J = 8.4 Hz), 7.66–7.69 (m, 4H, H-2, H-3, H-5 & H-6 of 4-Cl-C6H5), 7.70 (d, 1H, H-8 phthalazine, J = 8.1 Hz), 7.92 (t, 1H, H-6 phthalazine, J = 8.1 Hz), 8.01 (t, 1H, H-7 phthalazine, J = 7.8 Hz), 8.43 (s, 1H, H-4 isatin), 8.61 (d, 1H, H-5 phthalazine, J = 7.8 Hz), 10.69 (s, 1H, NH, D2O exchangeable), 13.12 (s, 1H, NH, D2O exchangeable); 13C NMR (75 MHz, DMSO-d6) δ: 111.44, 118.91, 125.29, 126.05, 126.35, 126.87, 128.75, 130.13, 131.32, 133.40, 133.84, 134.07, 141.45, 142.72, 148.30, 151.91, 165.62; Anal. Calcd. for C22H13Cl2N5O: C, 60.85; H, 3.02; N, 16.13; Found C, 61.03; H, 2.97; N, 16.06.</p><p>5-Bromo-3–(2–(4–(4-chlorophenyl)phthalazin-1-yl)hydrazono)indolin-2-one (10h). Yellow powder (yield 75%), m.p. >300 °C; IR (KBr, ν cm−1): 3210 (NH) and 1678 (C = O); 1H NMR (300 MHz, DMSO-d6) δ ppm: 6.85 (d, 1H, H-6 isatin, J = 8.1 Hz), 7.47 (d, 1H, H-7 isatin, J = 8.4 Hz), 7.66–7.69 (m, 4H, H-2, H-3, H-5 & H-6 of 4-Cl-C6H5), 7.71 (d, 1H, H-8 phthalazine, J = 7.8 Hz), 7.92 (t, 1H, H-6 phthalazine, J = 8.1 Hz), 8.01 (t, 1H, H-7 phthalazine, J = 7.8 Hz), 8.58–8.61 (m, 2H, H-4 isatin and H-5 phthalazine), 10.74 (s, 1H, NH, D2O exchangeable), 13.17 (s, 1H, NH, D2O exchangeable); 13C NMR (75 MHz, DMSO-d6) δ: 111.83, 113.01, 119.40, 123.42, 125.07, 126.06, 126.89, 128.66, 129.11, 131.32, 132.69, 133.39, 133.85, 134.07, 141.80, 142.60, 148.30, 151.89, 165.49; Anal. Calcd. for C22H13BrClN5O: C, 55.20; H, 2.74; N, 14.63; Found C, 55.36; H, 2.77; N, 14.55.</p><!><p>Compounds 12a,b were prepared according to the literature procedure45.</p><!><p>Following the same procedures described for preparation of the hybrids 5a–h, using N-benzylindoline-2,3-diones 7a,b instead of indoline-2,3-dione 4a–d.</p><p>1-Benzyl-3–(2–(2–(4-chlorophenyl)quinazolin-4-yl)hydrazono)indolin-2-one (13a). Orange powder (yield 65%), m.p. 291–293 °C; IR (KBr, ν cm−1): 1681 (C = O); 1H NMR (300 MHz, DMSO-d6) δ ppm: 5.00, 5.06 (s, 2H, –CH2–), 6.99 (d, 1H, Ar–H, J = 7.8 Hz), 7.16 (t, 1H, Ar–H, J = 7.5 Hz), 7.27–7.46 (m, 6H, Ar–H), 7.59–7.98 (m, 6H, Ar), 8.10 (d, 1H, Ar–H, J = 7.5 Hz), 8.47–8.57 (m, 2H, Ar–H), 11.42, 13.78 (s, 1H, NH, D2O exchangeable); Anal. Calcd. for C29H20ClN5O: C, 71.09; H, 4.11; N, 14.29; Found C, 70.91; H, 4.16; N, 14.21.</p><p>1-Benzyl-3–(2–(2–(2,6-dichlorophenyl)quinazolin-4-yl)hydrazono)indolin-2-one (13b). Orange powder (yield 70%), m.p. >300 °C; IR (KBr, ν cm−1): 1674 (C = O); 1H NMR (300 MHz, DMSO-d6) δ ppm: 4.98, 5.07 (s, 2H, –CH2–), 6.98 (d, 1H, Ar–H, J = 7.5 Hz), 7.10 (t, 1H, Ar–H, J = 7.8 Hz), 7.25–7.37 (m, 5H, Ar–H), 7.42 (t, 1H, Ar–H, J = 7.8 Hz), 7.61 (d, 1H, Ar–H, J = 7.8 Hz), 7.68–7.89 (m, 4H, Ar–H), 8.03 (d, 1H, Ar–H, J = 8.4 Hz), 8.51–8.53 (m, 2H, Ar–H), 12.28, 13.87 (s, 1H, NH, D2O exchangeable); Anal. Calcd. for C29H19Cl2N5O: C, 66.42; H, 3.65; N, 13.36; Found C, 66.57; H, 3.69; N, 13.28.</p><p>3–(2–(2–(4-Chlorophenyl)quinazolin-4-yl)hydrazono)-1–(4-fluorobenzyl) indolin-2-one (13c). Orange powder (yield 68%), m.p. > 300 °C; IR (KBr, ν cm−1): 1677 (C = O); 1H NMR (300 MHz, DMSO-d6) δ ppm: 4.99, 5.05 (s, 2H–CH2–), 7.02 (d, 1H, Ar–H, J = 7.2 Hz), 7.12–7.19 (m, 4H, Ar–H), 7.30 (t, 1H, Ar–H, J = 7.5 Hz), 7.42–7.81 (m, 6H, Ar–H), 7.98 (s, 1H, Ar–H), 8.18 (d, 1H, Ar–H, J = 8.1 Hz), 8.48–8.56 (m, 2H, Ar–H), 11.67, 13.79 (s, 1H, NH, D2O exchangeable); Anal. Calcd. for C29H19ClFN5O: C, 68.57; H, 3.77; N, 13.79; Found C, 68.72; H, 3.81; N, 13.73.</p><!><p>Human breast cancer (MDA-MB-231) cells, LoVo human colorectal carcinoma cells, human hepatoma (HepG2) cells and human non-small cell lung cancer (A549) cells were obtained from American Type Culture Collection (ATCC). Cells were maintained in Dulbecco's Modified Eagle's Medium (DMEM) (Sigma-Aldrich, St. Louis, MO), supplemented with 10% fetal bovine serum (Lonza Group, Basel, Switzerland), 100 IU/mL penicillin, 100 mg/mL streptomycin and 2 mmol/L L-glutamine (Sigma). Cells were seeded into 96-well plates at 0.4 × 104/well and incubated overnight. The medium was replaced with fresh one containing the desired concentrations of the test compounds.</p><p>Anti-proliferative activity of the prepared hybrids against MDA-MB-231 cells was evaluated by the cell GI assay using 5-Fluorouracil (5-FU) as a standard treatment. This assay was conducted by the use of WST-1 reagent for determination of IC50 for each compound and the results are given in Table 1. After 48 h of incubation with the synthesized agents in the treatment groups or the vehicle (1% DMSO) in the control group, 10 μL of the WST-1 reagent were added to each well and the plates were re-incubated for 4 h at 37 °C. The amount of formazan was quantified using ELISA reader at 450 nm. Moreover, cytotoxicity of the compounds and the reference drug, doxorubicin in LoVo, HepG2 and A549 cells was assessed against control group at the end of exposure using the sulforhodamine B (SRB) assay. Experimental conditions were tested using three replicates (three wells of the 96-well plate per experimental condition) and all experiments were performed in triplicates. IC50 was calculated according to the equation for Boltzman sigmoidal concentration–response curve using the nonlinear regression fitting models by Graph Pad, Prism version 5 (GraphPad Software Inc., La Jolla, CA).</p><!><p>In vitro anti-proliferative activity of the newly synthesized hybrids against MDA-MB-231 cell line.</p><p>IC50 values are the mean ± SD of three separate experiments.</p><!><p>Effects of treatment of MDA-MB-231 cells with the most promising compounds 5e and 10g on the cells' apoptotic machinery were further investigated. The levels of the apoptotic markers caspase-9, caspase-3 and Bax as well as the anti-apoptotic marker Bcl-2 were assessed using ELISA colorimetric kits (Neogen, Lexington, KY) as per the manufacturer's instructions.</p><p>MDA-MB-231 cells were cultured as a monolayer in T-25 flasks and were seeded to attain 30% confluency prior to treatment. Cells were then treated separately with compounds 5e and 10g at their IC50 concentrations (12.35 and 12 μM, respectively) for 48 h. At the end of treatment, cells were collected via trypsinization and centrifuged at 10,000 rpm. The pellet was then rinsed with phosphate buffered saline (PBS) and lysed in radio immunoprecipitation assay (RIPA) lysis buffer at 4 °C for 45 min, then centrifuged at 14,000 rpm for 20 min to remove the cellular debris. Lysates were then collected and stored at −80 °C for later protein determination using Pierce BCA Protein Assay Kit (Pierce Biotechnology, Rockford, IL) according to manufacturer's recommendations.</p><p>The cell lysate was diluted 10 times, and 100 μL (50 mg protein) was added to the wells of four separate microtiter plates for the four ELISA kits that were pre-coated with primary antibodies specific to caspase-9, caspase-3, Bax and Bcl-2 proteins, respectively. A secondary biotin-linked antibody specific to the protein captured by the primary antibody was further added to bind the captured protein, forming a "sandwich" of specific antibodies around the desired protein in the cell lysate. The streptavidin-horseradish peroxidase (HRP) complex was then used to bind the biotin-linked secondary antibody through its streptavidin portion. The HRP domain reacted with the added TMB substrate, forming a colored product that was measured at 450 nm by a plate reader (ChroMate-4300, Awareness Technology, Inc., Palm City, FL) after the reaction was terminated by the addition of stop solution.</p><!><p>Apoptotic cells were further analyzed by Annexin V-FITC/DAPI assay (Cayman Chemical, Ann Arbor, MI). Briefly, MDA-MB-231 cells were cultured to a monolayer then treated with compound 10g at the IC50 concentration (12 μM) as described earlier. Cells were then harvested via trypsinization, and rinsed twice in PBS followed by binding buffer. Moreover, cells were re-suspended in 100 μL of binding buffer with the addition of 1 μL of FITC-Annexin V (Becton Dickinson BD Pharmingen™, Heidelberg, Germany) followed by an incubation period of 30 min at 4 °C. Cells were then rinsed in binding buffer and re-suspended in 150 μL of binding buffer with the addition of 1 μL of DAPI (1 μg/μL in PBS) (Invitrogen, Life Technologies, Darmstadt, Germany). Cells were then analyzed using the flow cytometer BD FACS Canto II (BD Biosciences San Jose, CA) and the results were interpreted with FlowJo7.6.4 software (Tree Star, FlowJo LLC, Ashland, OR).</p><!><p>Data are presented as means ± SD. Individual groups were compared using the two-tailed independent student's t-test. Multiple group comparisons were carried out using one-way analysis of variance (ANOVA) followed by the Tukey–Kramer test for post-hoc analysis. Statistical significance was accepted at a level of p < 0.05. All statistical analyses were performed using GraphPad InStat software, version 3.05 (GraphPad Software, Inc., La Jolla, CA). Graphs were sketched using GraphPad Prism software, version 5.00 (GraphPad Software, Inc., La Jolla, CA).</p><!><p>Absorption, distribution, metabolism and excretion (ADME) profiling for the prepared hybrids was carried out using Discovery Studio 2.5 (Accelrys, San Diego, CA). All the tested hybrids were drawn as a small library then prepared using ligand protocol to find the suitable orientation in 3D. ADME profiling was predicted for the designed library using ADME descriptors protocol.</p><!><p>The synthetic strategies employed to prepare the new target hybrids are depicted in Schemes 1–3. In Scheme 1, 4-chloro-2-phenylquinazolines 2a,b were obtained (in 75 and 72% yield, respectively), by chlorination of 2-phenylquinazolin-4(3H)-ones 1a,b in excess refluxing phosphorous oxychloride. Next, 4-chloro-2-phenylquinazolines 2a,b were refluxed with hydrazine hydrate in ethanol to furnish the hydrazine derivatives 3a,b (in 81 and 86% yield, respectively), which were condensed with different isatins in refluxed ethanol in the presence of a catalytic amount of glacial acetic to afford the isatin-quinazoline hybrids 5a–h (in 65–77% yield).</p><!><p>Reagents and conditions: i, POCl3/N,N-dimethylaniline/reflux 6 h; ii, NH2NH2.H2O/EtOH/reflux 4 h; iii, EtOH/AcOH (catalytic)/reflux 0.5 h.</p><p>Reagents and conditions: i, NH2NH2.H2SO4/NaOH/reflux 1 h; ii, POCl3/N,N-dimethylaniline/reflux 6 h; iii, NH2NH2.H2O/EtOH/reflux 7 h. iv, EtOH/AcOH (catalytic)/reflux 1 h.</p><p>Reagents and conditions: i, DMF/K2CO3/reflux 3 h; ii, Compounds 4a,b/EtOH/AcOH (catalytic)/reflux 0.5 h.</p><!><p>IR spectra of the latter hybrids displayed absorption bands due to the NH groups in the region 3170–3305 cm−1, in addition to a carbonyl band in the region 1672–1690 cm−1. The 1H NMR spectra of compounds 5a–h confirmed the presence of two D2O-exchangeable singlet signals attributable to NH protons of the hydrazine function ( = N–NH–) and the isatin in the range δ 10.62–11.49 and 12.11–13.93 ppm, respectively.</p><p>For the synthesis of isatin-phthalazine hybrids 10a–h, 2-benzoylbenzoic acids 6a,b were cyclocondensed with hydrazine sulfate in the presence of sodium hydroxide to afford phthalazinone 7a,b. Next, chlorination of compounds 7a,b was carried out via refluxing with excess phosphorus oxychloride to furnish 1-chloro-4-phenylphthalazines 8a,b (in 65 and 72% yield, respectively). The chloro intermediates 8a,b were reacted with hydrazine hydrate in refluxing ethanol to provide the 1-hydrazinyl-4-phenylphthalazines 9a,b (in 84 and 80% yield, respectively), which upon reaction with different isatins afforded the target hybrids 10a–h (in 60–75% yield) (Scheme 2).</p><p>IR spectra of 10a–h revealed the presence of an NH stretching band about 3200 cm−1. Their 1H NMR spectra showed two D2O-exchangeable singlet signals from the hydrazine function ( = N–NH–) and the isatin protons in the range δ 10.56–10.74 and 12.83–13.17 ppm.</p><p>Finally, isatin was refluxed with benzyl bromide or 4-fuorobenzyl bromide in dry DMF and anhydrous K2CO3 to furnish N-substituted isatins 12a,b, respectively (in 88 and 82% yield, respectively). The reaction of derivatives 12a,b with the appropriate 4-hydrazinyl-2-phenylquinazoline derivative 3 in ethanol in the presence of a catalytic amount of glacial acetic acid afforded the target hybrids 13a–c (Scheme 3).</p><p>The structures of hybrids 13a–c were confirmed under the basis of spectral and elemental analyses which were in full agreement with the proposed structures.</p><!><p>Anti-proliferative activity of the newly prepared hybrids 5a–h, 10a-h and 13a–c was evaluated against MDA-MB-231 TNBC cell line using the WST-1 assay as described by Ngamwongsatit et al.50. Being a well-known broad-spectrum anticancer drug and a first line therapy for many tumors management, 5-FU was chosen as a positive control51,52. The anti-proliferative activity was expressed as the half maximal inhibitory concentration (IC50) values (Table 1).</p><p>From the obtained results, it is obvious that most of the tested hybrids have excellent to moderate growth inhibitory activity toward the tested MDA-MB-231 cancer cell line. In particular, compounds 5e and 10g were the most active members against MDA-MB-231 cells (IC50 = 12.35 ± 0.12 and 12.00 ± 0.13 μM, respectively), with 2.37- and 2.44-fold increased activity than the reference drug, 5-FU (IC50 = 29.38 ± 1.24 μM). Besides, compounds 5a–d, 5f–h, 10a–f, 10h and 13a displayed good anti-proliferative activity with IC50 values ranging from 12.86 ± 0.12 to 26.21 ± 2.07 μM. Whilst, hybrids 13a and 13b were moderately active with IC50 values of 48.72 ± 2.59 and 34.17 ± 2.42 μM, respectively.</p><!><p>By scrutinizing the aforementioned biological data, one can snugly reveal a clear structure activity relationship. First, we investigated the impact of the substitution at the 5-position of the isatin moiety. Regarding the isatin-quinazoline hybrids 5a–h, incorporation of unsubstituted isatin resulted in compounds 5a and 5e with moderate activity against the MDA-MB-231 cell line (IC50 = 13.18 ± 0.36 and 12.35 ± 0.12 μM, respectively). As it has an electronic and size properties similar to those of hydrogen, fluorine is introduced as a classical bioisostere to the hydrogen atom. Compounds 5b and 5f bearing fluorine substituent, displayed slight decrease in the anti-proliferative activity (IC50 = 15.15 ± 0.15 and 12.95 ± 0.31 μM, respectively) compared to the unsubstituted analogs, suggesting that halogens incorporation may not be advantageous for the activity. Furthermore, introduction of more bulky and lipophilic chlorine atom; compounds 5c and 5g, caused decrease of activity against MDA-MB-231 cells (IC50 = 17.9 ± 0.11 and 14.59 ± 0.23 μM, respectively). Also, incorporation of bromine atom (more bulky and lipophilic than chlorine) decreased the anti-proliferative activity (IC50 = 19.68 ± 0.26 and 17.31 ± 0.55 μM, respectively). Thus, the order of activity of the halogenated hybrids in such series, were smoothly decreased in the order of F > Cl > Br, hinting that increasing the lipophilicity and the size of the substituents of isatin moiety at the 5-position is not affirmative for the activity of the isatin-quinazoline hybrids regardless of the type of substitution of the 2-phenyl group of the quinazoline moiety.</p><p>Regarding the isatin-phthalazine hybrids 10a–h, the above-mentioned relationship could be, also, proposed for the hybrids 10a–d with unsubstituted 4-phenyl group in the phthalazine moiety. While, substitution of hybrids 10e–h (containing 4–(4-chlorophenyl) group) with F or Cl at 5-position of isatin moiety has no significant effect, introduction of Br led to a decreased activity toward the MDA-MB-231 cell line.</p><p>Further examination of the effect of the substitution pattern of the phenyl group on the prepared hybrids was then conducted. Regarding the isatin-quinazoline hybrids 5a–h, the increased IC50 values of members 5a–d with incorporated 4-chlorophenyl group (13.18 ± 0.36, 15.15 ± 0.15, 17.93 ± 0.11 and 19.68 ± 0.26 μM, respectively) than that of their corresponding analogs 5e–h with 2,6-dichlorophenyl group (12.35 ± 0.12, 12.95 ± 0.31, 14.59 ± 0.23 and 17.31 ± 0.55 μM, respectively) indicated that 2,6-dichloro substitution is more beneficial for activity rather than 4-chloro substitution, which could be attributed to the non-coplanarity or the increased lipophilicity. On the other hand, substitution of the 4-phenyl group in the isatin-phthalazine hybrids was found to be advantageous to the activity rather than unsubstitution, that is evidenced by the decreased IC50 values of compounds 10e–f (12.86 ± 0.12, 12.94 ± 0.23, 12.00 ± 0.13 and 15.21 ± 0.15 μM, respectively) than those of their corresponding analogs 10a–d (13.96 ± 0.10, 15.62 ± 0.14 17.63 ± 0.37 and 21.39 ± 0.13 μM, respectively).</p><p>Finally, investigation of the effect of N-benzylation of isatin moiety showed that this was not an advantageous approach for the anti-proliferative activity against MDA-MB-231 cell line, where the N-benzyl hybrids 13a–c displayed decreased potency (26.21 ± 2.07, 48.72 ± 2.59 and 34.17 ± 2.42 μM, respectively) than their unsubstituted counterparts 5a and 5e (IC50 = 13.18 ± 0.36 and 12.35 ± 0.12 μM, respectively).</p><p>In conclusion, we can deduce that unsubstitution of isatin moiety at C-5 or N-1 positions, disubstitution of the 2-phenyl group of quinazoline moiety and substitution of the 4-phenyl group of phthalazine moiety are crucial elements for the anti-proliferative activity against MDA-MB-231.</p><!><p>Induction of apoptosis in cancer cells is one of the successful strategies for the development of cancer therapy53–55. Therefore, we investigated the potential pro-apoptotic effects of our isatin-quinazoline and isatin-phthalazine hybrids attempting to explore the underlying mechanism for their anti-proliferative activity.</p><p>As indicated by the cytotoxicity results, compounds 5e and 10g were found to be the most active agents against MDA-MB-231 breast cancer cells. Thus, we investigated the potential pro-apoptotic effects of these promising agents in an attempt to define the principle mechanism for their anti-proliferative activity.</p><p>The Bcl-2 family of proteins is responsible for synchronizing the mitochondrial apoptotic pathway56. These proteins are classified into two groups: anti-apoptotic proteins such as Bcl-2 protein and the counteracting pro-apoptotic proteins including Bax protein57. It has been previously indicated that the overexpression of the anti-apoptotic Bcl-2 protein is a contributing factor to tumorigenesis in a multitude of cancers including breast cancer58. In this study, exposure of MDA-MB-231 breast cancer cells to compounds 5e and 10g for 48 h resulted in a significant increase in the expression of the pro-apoptotic protein Bax by ∼130 and 180%, respectively, compared to the control (Figure 3(A), Table 2). On the other hand, treatment of MDA-MB-231 breast cancer cells with compounds 5e and 10g significantly reduced the expression levels of the anti-apoptotic protein Bcl-2 by ∼29 and 52%, respectively, compared to the control (Figure 3(B), Table 2). A rather more important parameter is the ratio between Bax (apoptosis inducer) and Bcl-2 (apoptosis suppressor) that gave more real insight to the apoptotic activity of the compounds as it is considered a key indicator of therapeutic response to chemotherapy. Analyzing the results discloses that compound 5e increased the Bax/Bcl-2 ratio two folds compared to the control, while compound 10g even boosted it four folds in comparison to the control. The ability of the two compounds to down regulate Bcl-2 levels while boosting Bax levels further supports their effectiveness as apoptosis inducers.</p><!><p>Effect of compounds 5e and 10g on the protein levels of A) Bax; B) Bcl-2 in MDA-MB-231 cells treated with the compounds at their IC50 concentrations against control (1% DMSO). Data are mean ± SD (n = 3). The experiment was done in triplicates. *Significantly different from control at p < 0.05. **Significantly different from control at p < 0.01. ***Significantly different from control at p < 0.001.</p><p>Effect of compounds 5e and 10g on active caspase-9 and -3 levels, and the expression levels of Bcl-2 and Bax in MDA-MB-231 cancer cells treated with the compounds at their IC50 concentrations.</p><p>Data are mean ± SD of three separate experiments.</p><p>aSignificantly different from control (1% DMSO) at p < 0.05.</p><p>bSignificantly different from control (1% DMSO) at p < 0.01.</p><p>cSignificantly different from control (1% DMSO) at p < 0.001.</p><!><p>The down-regulation of the anti-apoptotic Bcl-2 results in increased levels of free pro-apoptotic Bax which then accumulates at the inner mitochondrial membrane forming channels thus altering membrane permeability. Apoptotic factors then leak into the cytoplasm resulting in the activation of caspases cascade59. Caspases are cysteine-containing aspartic acid-specific proteases, which exist as zymogens in the cytoplasm and play a pivotal role in the execution of cell death upon activation60. Caspase-3 is the key executioner protease that is activated by upstream initiator caspases such as caspase-960. Therefore, the elevated Bax/Bcl-2 ratios obtained with these treatments triggered the investigation of the protein expression levels of active caspases-9 and -3.</p><p>In this study, treatment of MDA-MB-231 cells with compounds 5e and 10g resulted in a significant elevation of active caspase-9 protein levels by around 2.3 and 2.7 folds, respectively, compared to control (Figure 4(A), Table 2). Moreover, MDA-MB-231 cells upon treatment with compounds 5e and 10g, exhibited a significant upregulation of active caspase-3 protein levels by ∼2.4 and 3.7 folds, respectively, compared to control (Figure 4(B), Table 2).</p><!><p>Effect of compounds 5e and 10g on the protein levels of A) active caspase-9; B) active caspase-3 in MDA-MB-231 cells treated with the compounds at their IC50 concentrations against control (1% DMSO). Data are mean ± SD (n = 3). The experiment was done in triplicates. ***Significantly different from control at p < 0.001.</p><!><p>It is worth noting that upon comparing the relative expression levels of caspases with the relative elevation patterns of Bax/Bcl-2 ratio observed in both treatments, it was found that the magnitude of the Bax/Bcl-2 ratio elevation is proportional to that of caspase-9 and that of caspase-3 up-regulation. This observation strengthens the correlation between the increased Bax/Bcl-2 ratio and the activation of proteolytic caspases that act as key players in the execution of apoptosis, which might be responsible for the anti-proliferative activity of the tested compounds. Moreover, compound 10g exhibited a more potent activity than compound 5e in terms of both elevating Bax/Bcl-2 ratio and further activating the caspases cascade at a relatively lower IC50. This finding directed further investigation of the apoptotic effect of this compound on MDA-MB-231 cells.</p><!><p>Externalization of the phospholipid phosphatidylserine (PS) at the cell membrane is one of the hallmarks of cells going into apoptosis61. Having a high Ca2+-dependent binding affinity to negatively charged phospholipid surfaces. Annexin A5 (AnxV A5), a 36-kDa human protein, is a suitable candidate for apoptosis imaging. Besides, Anx V-based flow cytometry analysis is a useful tool to comprehend whether cell death is due to physiological apoptosis or nonspecific necrosis61.</p><p>Further evaluation of the apoptotic effect of the most potent compound 10g was carried out using Anx V-FITC/DAPI dual staining assay (Figure 5). MDA-MB-231 cells treated with compound 10g showed a significant increase in the percent of Anx V-FITC positive apoptotic cells (apoptotic and late apoptotic) from 3.88 to 31.21% which comprises about 8.4-folds (p < 0.001) compared to control (Figure 5). Flow cytometric analysis of the differential binding of the cells to Anx V-FITC displayed a significant increase in the proportion of early apoptotic cells. This observation is in line with the simultaneous up-regulation of the upstream caspase-9 (an initiator caspase) together with the downstream caspase-3 (the hallmark key executor) belonging to the intrinsic apoptotic pathway. This strongly suggests as well that a cascade of proteinases' activation has been triggered as a consequence of an increase in Bax/Bcl-2 ratio, eventually leading to apoptosis. These findings clearly imply that compound 10g has potent pro-apoptotic activity that unravels the mechanism of its anti-proliferative activity.</p><!><p>Effect of compounds 10g on the percentage of Annexin V-FITC-positive staining in MDA-MB-231 cells versus control (1% DMSO). Data are mean ± SD (n = 3). The experiments were done in triplicates. The four quadrants identified as: Normal, viable; Apo, early apoptotic; LATE APO, late apoptotic; NEC, necrotic. ***Significantly different from control at p < 0.001 (student's t-test).</p><!><p>In conclusion, the reduced expression of the anti-apoptotic protein Bcl-2 in addition to the enhanced expression of the pro-apoptotic protein Bax as well as the up-regulated active caspase-9 and caspase-3 levels together with a harmonized increase in the Bax/Bcl-2 ratio, suggests that the cytotoxic effect of compounds 5e and 10g might be attributed, at least in part, to the induction of the intrinsic apoptotic mitochondrial pathway.</p><!><p>The in vitro anti-proliferative activity of the prepared hybrids 5a–h, 10a–h and 13a–c was further examined against a panel of cell lines, namely A549 alveolar carcinoma, Caco-2 colon cancer, LoVo human colorectal carcinoma and HepG2 hepatocellular carcinoma using SRB colorimetric assay as described by Skehan et al.62, to carry out further elaboration for their anti-proliferative activity. The results were expressed as IC50 values and listed in Table 3.</p><!><p>In vitro anti-proliferative activities of the newly synthesized hybrids against A549, Caco-2, LoVo and HepG2 cell lines.</p><p>IC50 values are the mean ± SD of three separate experiments.</p><p>NA: Compounds having IC50 value >100 μM.</p><!><p>The presented data revealed that some hybrids displayed moderate to fair anti-proliferative activity toward A549 cell line. In particular, compounds 5d and 10d showed the better activity against A549 with IC50 values of 28.73 ± 1.59 and 18.42 ± 1.52 μM, respectively. Besides, most compounds possessed good to moderate activity against LoVo cell line with IC50 values ranged from 13.09 ± 0.33 to 48.62 ± 2.21 μM. Regrettably, among the tested cancer cell lines, the colon Caco-2 and hepatocellular HepG2 were not susceptible cell lines to almost all hybrids influence.</p><!><p>The ADME of the prepared hybrids 5a–h, 10a–h and 13a–c was predicted through a theoretical kinetic study performed by Discovery Studio software (Table 4). To evaluate the lipophilicity and polar surface area, AlogP98 and PSA_2D descriptors were calculated. Besides, absorption, solubility and CYP2D inhibition levels were predicted. All compounds were expected to have very low-to-low aqueous solubility. Meanwhile, most of the examined derivatives seemed to possess good to poor absorption levels, and predicted to be CYP2D inhibitors except compounds 10a, 10e–h and 13b which are expected not to inhibit CYP2D.</p><!><p>Computer aided ADME study for the prepared hybrids.</p><p>Lipophilicity descriptor.</p><p>Polar surface area.</p><p>Solubility parameter. (0:−2 = optimal, −2:−4 = good, −4:−6 = low, −6:−8 = very low).</p><p>Solubility level (0 = extremely low, 1 = very low but possible, 2 = low, 3 = good, 4 = optimal).</p><p>Absorption level (0 = good, 1 = moderate, 2 = low, 3 = very low).</p><p>CYP2D inhibition (0 = non inhibitor, 1 = inhibitor).</p><!><p>In an effort to develop potent anti-proliferative agents, the molecular hybridization approach was adopted to design and synthesize new different series of isatin-quinazoline hybrids 5a–h, isatin-phthalazine hybrids 10a–h and N-benzylisatin-quinazoline hybrids 13a–c. The in vitro anti-proliferative activity of the newly synthesized hybrids was evaluated against the TNBC MDA-MB-231 breast cancer. Compounds 5e and 10g displayed the highest potency toward MDA-MB-231 with IC50 values of 12.35 ± 0.12 and 12.00 ± 0.13 μM, respectively. Subsequently, compounds 5e and 10g were further estimated for their apoptosis induction potential. Both compounds 5e and 10g proved to induce apoptosis, which was assured by the reduced expression of the anti-apoptotic protein Bcl-2 in addition to the enhanced expression of the pro-apoptotic protein Bax as well as the up-regulated active caspase-9 and caspase-3 levels together with a harmonized increase in the Bax/Bcl-2 ratio. Furthermore, compound 10g showed a significant increase in the percent of annexin V-FITC positive apoptotic cells from 3.88 to 31.21% which comprises about 8.4 folds compared to control. Finally the newly prepared hybrids were examined against a panel of cell lines, namely, LoVo human colorectal carcinoma, A549 alveolar carcinoma and Caco colon cancer cell lines to carry out further elaboration for their anti-proliferative activity. Based on the previous findings, we came into conclusion that the isatin- quinazoline/phthalazine hybrids are a good platform for further optimization as novel anticancer agents for TNBC.</p>

PubMed Open Access

Exosome Uptake Depends on ERK1/2-Heat Shock Protein 27 Signaling and Lipid Raft-mediated Endocytosis Negatively Regulated by Caveolin-1*

Background: Exosome vesicles can transfer molecular information previously shown to stimulate tumor development; however, the mechanism of exosome uptake is unknown.Results: Mammalian cells internalize exosomes through lipid raft-mediated endocytosis negatively regulated by caveolin-1.Conclusion: Our findings provide novel insights into cellular uptake of exosomes.Significance: Our data provide potential strategies for how the exosome uptake pathway may be targeted.

exosome_uptake_depends_on_erk1/2-heat_shock_protein_27_signaling_and_lipid_raft-mediated_endocytosis

Introduction<!>Cell Culture<!>Antibodies and Reagents<!>Exosome Purification and Characterization<!>Confocal Laser Scanning Microscopy and Live-cell Imaging of Internalized Exosomes<!>TIRF Microscopy of Cell Surface-bound Exosomes<!>Flow Cytometry Analysis<!>Electron Microscopy<!>Transfections and Transductions<!>Immunoblotting and Immunoprecipitation<!>Phosphokinase Array<!>Statistical Analysis<!>Exosomes Enter Cells by Endocytosis and Travel Along Endosomal Cytoskeletal Routes<!><!>Exosomes Enter Cells by Endocytosis and Travel Along Endosomal Cytoskeletal Routes<!>Role of Lipid Raft-mediated Endocytosis in Exosome Uptake<!><!>Role of Lipid Raft-mediated Endocytosis in Exosome Uptake<!>CAV1 Negatively Regulates Exosome Uptake<!><!>CAV1 Negatively Regulates Exosome Uptake<!>Role of ERK1/2-HSP27 Signaling Activation in Exosome Uptake<!><!>Role of ERK1/2-HSP27 Signaling Activation in Exosome Uptake<!>CAV1 Regulates Uptake of Exosomes through Suppression of ERK1/2 Signaling<!><!>CAV1 Regulates Uptake of Exosomes through Suppression of ERK1/2 Signaling<!>

<p>The process of endocytosis involves multiple mechanisms in mammalian cells differing in the type of cargo and fate of cargo upon internalization. Described mechanisms of classical endocytosis include clathrin-dependent endocytosis, macropinocytosis, clathrin-independent endocytosis pathways such as caveolae-mediated uptake associated with lipid rafts (or cholesterol-enriched membrane microdomains) in the plasma membrane and non-classical pathways involving non-clathrin, non-caveolae-mediated endocytosis (1–4). Secreted vesicles, here classified as exosomes (also referred to as microvesicles, shedded vesicles, ectosomes, microparticles, plasma membrane-derived vesicles) are complex biological vehicles enclosed with cytoplasmic components and genetic material (5–8). The putative function of exosomes in cancer are based on the recently described findings of the transfer of genetic material and signaling proteins, resulting in e.g. increased angiogenesis and metastasis (8–12). Exosomes are released from multivesicular bodies (MVBs)2 upon their fusion with the plasma membrane (13). Recent studies suggest that released vesicles may transfer functional RNA as well as transmembrane proteins contributing to the propagation of a transformed cell phenotype (9, 10, 14). Accordingly, exosomes can be regarded as multi-purpose delivery vehicles in analogy with an endogenous virus-like particle infecting cells in the surrounding environment. Importantly, recent reports have documented key processes involved in exosome release and have identified similarities in topology and mechanisms to retroviral budding (15, 16). It is still controversial whether vesicular uptake is cell type specific (17), and whether it involves membrane fusion or endocytosis (18, 19). Thus, studies elucidating the mechanisms involved in exosome uptake remain an important challenge. We have previously reported on microvesicle-induced pro-angiogenic signaling; a process highly relevant in the most aggressive brain tumor type GBM (20). Here we set out to elucidate the unknown mechanisms of the uptake of GBM cell-derived vesicles and the signaling events involved in the internalization process.</p><!><p>Human umbilical vein endothelial cells (HUVECs, Lonza) were cultured in endothelial basal medium supplemented with 10% heat-inactivated fetal bovine serum (FBS), 2 mm l-glutamine, 100 units/ml penicillin, 100 μg/ml streptomycin, 10 ng/ml hydrocortisone, and 20 μg/ml human recombinant EGF. Human cervix adenocarcinoma (HeLa, ATCC), HeLa cells previously generated to stably overexpress CAV1-YFP (21), human GBM cells (U87 MG, ATCC) and wild-type (WT) or CAV1 knock out (cav-1 (−/−)) mouse embryonic fibroblasts (MEF, ATCC) were cultured in DMEM supplemented with 10% FBS, 2 mm l-glutamine, 100 units/ml penicillin, 100 μg/ml streptomycin (growth medium). CHO-K1 and COS-7 cells were cultured in F12K supplemented with 10% FBS, 2 mm l-glutamine, 100 units/ml penicillin, 100 μg/ml streptomycin. U87 MG shRNA negative control (NC) and shRNA CAV1 were routinely cultured in growth medium supplemented with 1 μg/ml puromycin. U87 MG cells transfected with a CD63-mCherry plasmid were sorted with FACSAria by fluorescence intensity and clones were further selected by neomycin resistance. U87 MG-CD63-mCherry cells were routine cultured in growth medium supplemented with 1000 μg/ml G418. All cells were cultured in a humidified incubator with 5% CO2 at 37 °C.</p><!><p>Antibodies for CAV1 (ab2910), CD63 (ab8219), TSG101 (ab30871), clathrin heavy chain (ab21679), p-HSP27 (ab17937), HSP27 (ab5579), flotillin-1 (ab41927), calnexin (ab2798), β-actin (ab8227), and α-tubulin (ab7291) were from Abcam. TF antibody (10H10) and siRNA for clathrin (sc-35067) were from Santa Cruz Biotechnology. Secondary antibodies conjugated with 5 or 15 nm gold were from Electron Microscopy Sciences, Fort Washington, PA. Total ERK1/2 antibody (9102) and siRNA for HSP27 (#6526S) were from Cell Signaling. Normal mouse IgG was from Molecular Probes. siRNA for CD63 (#4392420, ID:s2701), Mission Lentiviral transduction particles (SHCLNV) (shRNA) for CAV1 (NM_001753 clone TRCN000008002) and negative control (NC) (SHCO16V, pLKO.1-puro), phosphorylated (Thr183/Tyr185) ERK1/2 antibody (M8159), PKH67 Green Fluorescent cell linker midi kit, PKH26 red fluorescent cell linker midi kit, CellVue labeling kit, cholera toxin subunit B FITC conjugate, FITC conjugated 10 kDa dextran, FITC-conjugated transferrin, Phalloidin-TRITC, amiloride, methyl-β-cyclodextrin (MβCD), filipin III, nocodazole, puromycin dihydrochloride, G418, Cytochalasin D, and Lantrunculin A were all from Sigma Aldrich. Dynabeads® Protein G and Hoechst 33342 were from Invitrogen. SEA Block Blocking buffer was from Thermo Scientific. DiI-labeled acetylated LDL was from Harbor Bioproducts. Simvastatin was from Calbiochem. Complete protease inhibitors and phosphatase inhibitors were from Roche. Human phosphokinase antibody array (#ARY003) was from R&D Systems. U0126 was purchased from Selleck Chemicals. Plasmid encoding CD63-mCherry was kindly provided by Dr Lippincott-Schwartz and CD63-GFP plasmids was a gift from Dr. Takahisa Takino (Kanazawa University). The plasmid encoding pIRESneo-EGFP-α-tubulin was obtained from Addgene (Patricia Wadsworth, plasmid #12298). Plasmid encoding CAV1-YFP (pEX_EF1_CAV1-YFP) was from ATCC.</p><!><p>Secreted vesicles were isolated from cell culture medium of U87 MG as previously described (20). Transmission electron microscopy was performed to validate the presence and purity of intact exosomes and analyzed for size using nanoparticle tracking analysis (NTA). Protein amounts were quantified with BCATM protein assay kit (Pierce) in each vesicle preparation. Isolated exosomes were in indicated experiments labeled with PKH67, PKH26 or cellvue midklaret fluorescent labeling kit according to the manufacturer's protocol (Sigma). Each vesicle preparation was stored at 4 °C and used within 5 days after isolation.</p><!><p>All experiments were performed in Zeiss LSM 710 confocal scanning equipment using excitation wavelengths of 405, 488, 546, 633 nm, and a Plan-Apochromat 20×/0.8M27 objective and a C-Apochromat 63X/1.20W korr M27 glycerol immersion objective. Image analysis was performed using the Zeiss LSM software. Live cell imaging of internalized exosomes was performed with the same microscope equipped with heat incubator set at 37 °C with 5% CO2. For live cell confocal laser scanning microscopy experiments, cells were grown in glass bottom chamber slides, and 10–20 μg/ml of labeled exosomes were added to subconfluent cells in phenol red-free and serum-free conditions and incubated as indicated in the figure legend. Surface-bound exosomes were removed by extensive washing with 1 m NaCl and serum-free medium, followed by live cell imaging of intracellular exosomes in phenol-red free medium. For colocalization studies, 10 μg/ml CtxB-FITC, 100 μg/ml Dx10-FITC, or 150 μg/ml Tfn-FITC was co-incubated with labeled exosomes and incubated for 30 min. Cells were washed as above and fixed using 2% (w/v) paraformaldehyde for 5 min at room temperature. Cells were immediately analyzed. The weighted colocalization coefficients were calculated in representative cells using the Zeiss LSM software, and represent the number of red pixels (exosomes) that colocalize with turquoise pixels (CtxB or Dx10) divided by the total number of turquoise pixels.</p><!><p>Total internal reflection fluorescence (TIRF) microscopy experiments were performed using an inverted microscope (Axio Observer Z1; Zeiss) equipped with a Zeiss TIRF module and a 100× 1.45 NA DIC M27 Zeiss TIRF oil immersion lens and acquisitions were made by using Slidebook 5.5 (3I). Subconfluent HeLa CAV1-YFP cells were incubated with labeled exosomes for 1 h in serum-free medium prior to live cell microscopy in phenol red-free medium.</p><!><p>Subconfluent cells were incubated with labeled exosomes for the indicated time periods. Cells were washed in PBS, detached with trypsin, and subsequently washed twice in PBS supplemented with 1% BSA (w/v) and analyzed by flow cytometry on a FACS-Calibur instrument integrated with Cell-Quest software (BD Biosciences). Graphs show mean fluorescent values (10,000 events/sample) of one of three representative experiments (n = 3) with similar results ± S.D. unless stated otherwise.</p><!><p>Isolated exosomes or cells incubated with or without exosomes were washed twice in Tris-buffered saline (TBS), pelleted, fixed for 1 h at 20 °C, and overnight at 4 °C in 2.5% glutaraldehyde in cacodylate buffer. For uptake of exosomes in HUVECs, cells were incubated with 10 μg/ml exosomes followed by trypsinization, washing in TBS and fixation in 2.5% glutaraldehyde in cacodylate buffer overnight at 4 °C. Some samples were incubated with the primary antibodies TF (10H10 titer 1:50) and tsg101 polyclonal (rabbit, titer 1:100) or flotillin-1 antibody (rabbit, 10 μg/ml) followed by detection with secondary antibodies conjugated with 5 nm gold, TF (1:10), or 15 nm gold (tsg101 and flotillin-1, titer 1:20) and analyzed using a JEOL JEM 1230 transmission electron microscope (JEOL, Peabody, MA) as previously described (20).</p><!><p>DNA plasmid transfections were performed in U87 MG cells (CD63-GFP), COS-7 (tubulin-GFP) or MEF, U87MG, CHO-K1, and HeLa (CAV1-YFP) seeded in chamber slides for confocal imaging or 24-well plates for flow cytometry analysis and grown in their respective medium w/o antibiotics. RNA interference was performed in HUVECs or U87 MG seeded in respective growth medium w/o antibiotics. All transfections were performed using Lipofectamine (Invitrogen) and 100 nm siRNA against CD63, HSP27, or clathrin following the recommendations of the manufacturer. U87 MG cells were transduced using Mission Lentiviral transduction particles (SHCLNV, Sigma Aldrich) for CAV1 (NM_001753 clone TRCN000008002) and negative control (negative control (SHCO16V, pLKO.1-puro)) at a MOI of 0.2 and selected by 1 μg/ml puromycin.</p><!><p>Cells or exosomes were washed with ice-cold PBS and lysed in a reducing RIPA buffer containing 20 mm Tris-HCl (pH 7.5), 150 mm NaCl, 1 mm Na2EDTA, 1 mm EGTA, 1% Nonidet P-40, 1% sodium deoxycholate, 2.5 mm sodium pyrophosphate, 1 mm β-glycerophosphate, 1 mm Na3VO4, 1 μg/ml leupeptin, and complete mini protease inhibitor mixture (Roche Diagnostics). For CD63 antibody requiring non-reduced conditions, samples were lysed in Triton-X buffer (20 mm Tris-HCl, pH 8.0, 137 mm NaCl, 1% (v/v) Triton X-100, 2 mm EDTA) supplemented with complete mini protease inhibitor mixture. Proteins were fractionated by electrophoresis, blotted, and developed using HRP substrate. The intensity of the bands was quantified using ImageJ software (NIH) using α-tubulin as loading control unless stated otherwise. For immunoprecipitation using magnetic Protein G® beads, cells were lysed in denaturing lysis buffer, and 500 μg of protein were immunoprecipitated using 1 μg of CAV1 antibody and incubated with rotation at 4 °C overnight. Resuspended Dynabeads were added (1.5 mg) and incubated with rotation at 4 °C for 3 h. Dynabeads®-Ab-antigen complexes were washed four times in PBS with Ca2+/Mg2+ before elution and further Western blotting analysis using 20% SEA Block in TBS supplemented with 1% (v/v) Tween20 as blocking buffer and for incubation with primary and secondary antibodies. Proteins were fractionated by electrophoresis, blotted, and developed using HRP substrate.</p><!><p>HUVECs were starved for 16 h in serum free medium and were either untreated or stimulated with exosomes (20 μg/ml) for 10 or 30 min for phosphokinase antibody array (#ARY003, R&D) before lysate collection. Levels of phosphorylated proteins (200 μg per sample) were analyzed in cell lysates according to the protocol provided by the manufacturer. Corresponding protein amounts were used as a control of phosphoproteins residing in exosomes. The arrays were quantified using ImageJ software (NIH). Values are expressed as the mean intensity relative to reference (Ref) of the respective blot.</p><!><p>All data are presented as the mean of triplicates ± S.D. Statistical significance was evaluated with Student's two-tailed unpaired t test using Microsoft Excel. In some cases, the error bars were smaller than the symbols.</p><!><p>The GBM cell line U87 MG secretes exosome-like microvesicles or exosomes previously characterized by us and others for features and function (9–11, 20). Nanotracking analysis and electron microscopy showed 50–400 nm-sized vesicles (Fig. 1, A and B) that were further characterized for the presence of the exosomal markers CD63, Tissue Factor (TF), and flotillin-1 and the absence of the ER marker calnexin, demonstrating purity of vesicles (Fig. 1C). The internalization of exosomes was visualized by confocal fluorescence microscopy in human umbilical vein endothelial cells (HUVECs) (Fig. 1D, left panel) and U87 MG cells (supplemental video S1). The specificity of the uptake of fluorescent exosomes was supported by efficient attenuation by an excess of unlabeled exosomes (Fig. 1D, right panel).</p><!><p>Endocytosis of GBM cell-derived exosome-like extracellular vesicles. A, characterization of U87 MG-derived vesicles by nanoparticle tracking analysis. B, electron microscopy validates intact vesicles. Scale bar, 100 nm. C, immunoblot analysis of cells and exosome-like vesicles for the exosomal markers CD63, TF, and flotillin-1, and the ER marker calnexin. D, confocal microscopy analysis of exosome uptake in the absence or presence of an excess (×4) unlabeled exosomes. Scale bars, 15 μm. E and F, time (E), and concentration (F)-dependent uptake of exosomes using flow cytometry analysis. G, insignificant passive uptake of exosomes at 4 °C. H, exosome uptake in human cervix adenocarcinoma cells (HeLa), chinese hamster ovary cells (CHO K1), mouse embryonic fibroblasts (MEF), U87 MG and HUVEC cells (n = 3). Values are normalized to HUVEC ( = 1). Control (Ctl) represents cells without exosomes. Flow cytometry graphs represent mean fluorescent values of one of two representative experiments generating similar results, error bars are ± S.D.; a.u., arbitrary units. I, exosomes move along microtubule tracks. COS-7 cells were transfected with pIRESneo-EGFP-α-tubulin 24 h prior to the addition of PKH26 labeled exosomes for an additional 16 h. Movie sequences display (4 boxed individual images, 0 s, 12 s, 25 s, 54 s) of exosome transport (yellow) along microtubule (white). Arrowheads depict intracellular, motile exosomes. Images were captured using a C-Apochromat 20X/0.8 M27 objective, 4.0 zoom, pinhole setting of 31 μm and laser gains of 5.5% (561 nm) and 5% (488 nm). Image size was x:512, y:512, and images were captured during 2 min and 11 s. Scale bar, 20 μm. For full-length movie, see supplemental video S2. J, reduced exosome transport in HUVECs treated for 10 min with 10 μg/ml nocodazole (lower panels) compared with Control (no treatment, upper panels). Pictures shown (Exo movement) represent color-coded data from time series of exosome movement in an overlay in which every time point (during 40 s) corresponds to a color. Images were captured using a C-Apochromat 63X/1.20W korr M27 objective (zoom 3.6). Scale bars, 50 μm. For full-length movies, see supplemental videos S3 and S4. K, cell viability is intact as measured by trypan blue exclusion after 10 min of nocodazole (10 μg/ml) treatment. L, electron microscopy images of compartments with internalized tissue factor (TF)-bearing exosomes over time. Low magnification overviews (upper panels, scale bars, 100 μm) confirm the intracellular localization of exosomes, and cropped pictures (lower panels, scale bars, 100 nm) demonstrate intraluminal vesicles (red arrowheads) positive for α-TF 5 nm gold particles. Note that endogenous vesicular structures in HUVECs are negative for TF (lower left panel, white arrowhead). M, electron microscopy colocalization studies in HUVECs of GBM cell-derived exosomes detected by α-TF 5 nm gold particles and MVBs distinguished by anti-TSG101 15 nm gold particles (scale bars, 100 μm). Ctl: no addition of exosomes. N–O, exosomes reside in a CD63-positive compartment after long-term incubation (24 h) but do not colocalize with CD63 at the cell surface. U87MG cells stably expressing CD63-mCherry (N) or transiently transfected with CD63-GFP (O) were incubated with PKH-labeled exosomes for the indicated time periods. Cells were washed in 1 m NaCl and PBS to remove nonspecifically bound exosomes before fixation and confocal microscopy analysis. CD63 (turquoise) and internalized exosomes (red) were captured at the indicated time points. Scale bars, 15 μm.</p><!><p>The uptake mechanism of exosomes has been a matter of debate, i.e. whether exosomes enter cells through direct fusion with the plasma membrane of recipient cells or through endocytosis (18, 19). Here, exosomes displayed time and concentration-dependent uptake kinetics (Fig. 1, E and F), and incubation at 4 °C efficiently attenuated uptake, suggesting an energy-dependent process rather than passive membrane passage (Fig. 1G). Further, several normal and transformed cell-lines were able to internalize exosomes at significant levels (Fig. 1H), which argues against that exosome transfer is restricted to specific cell-types as has previously been suggested (17). Live confocal imaging with tubulin-GFP-expressing cells revealed that internalized exosomes travel along microtubules following their uptake (Fig. 1I and supplemental video S2). The mobility of internalized exosomes was substantially reduced by interference with microtubule polymerization using nocodazole (Fig. 1J and supplemental videos S3 and S4). Importantly, under these conditions nocodazole had no unspecific effect on cell viability (Fig. 1K). These data are consistent with an endocytic process rather than membrane fusion as exosomes followed a time, concentration, and temperature-dependent pathway. The fact that disruption of microtubule attenuated intracellular exosome transport further indicates that exosomes are not fused with the plasma membrane, but rather are dependent on the regular endosomal transportation machinery for further intracellular sorting. To corroborate these data, tissue factor (TF), previously shown to reside in GBM cell-derived exosomes while absent in HUVECs (20), was used to discriminate between endogenous vesicles and internalized exosomes in electron microscopy analysis. Immunolabeling for TF demonstrated that internalized exosomes are enclosed in double membrane structures and do not merge with the plasma membrane of recipient HUVECs (Fig. 1L).</p><p>Exosomes were shown to be sorted to larger compartments positive for the MVB marker TSG101 (tumor susceptibility gene 101 protein) (22) (Fig. 1M). In further support of the sorting of endocytosed exosomes to MVBs, a substantial fraction of internalized vesicles were located in a CD63 (LAMP-3, Tspn30)-positive compartment (Fig. 1N). Experiments with exosomes derived from CD63-GFP transfected cells pre-incubated with PKH-labeled exosomes, suggested that mixing of exosomal constituents can occur during maturation in MVBs, resulting in the generation of compound vesicles (supplemental Fig. S1, A and B). However, siRNA-mediated knockdown of CD63 or antibody-mediated cell-surface blocking of CD63 did not significantly affect exosome uptake (supplemental Fig. S1, C–E). These results, and the fact that exosomes did not colocalize with CD63 at short incubation times (Fig. 1O), suggest that although exosomes are sorted to CD63-positive vesicles in recipient cells, CD63 has no direct role in the endocytic uptake of exosomes.</p><!><p>We next sought to define the distinct, cellular pathways associated with the endocytic uptake of exosomes. We found no colocalization of exosomes either with transferrin (Tfn) or acetylated LDL (Fig. 2A), i.e. conventional ligands of classical, clathrin-dependent endocytosis (2). Accordingly, 80% knockdown of clathrin (heavy chain) (Fig. 2, B and C) had no effect on exosome uptake (Fig. 2D). Interestingly, when evaluating by confocal microscopy the colocalization coefficient of exosomes and the lipid raft marker cholera toxin subunit B (CtxB) at short-term incubation, we found ∼20% colocalization (Fig. 2, E and G). CtxB binds strongly to GM1 (monosialotetrahexosylganglioside), an established constituent of membrane lipid rafts, the integrity of which depends on the high abundance of cholesterol. We found some colocalization of exosomes and the macropinocytosis/fluid-phase marker 10 kDa dextran (Dx10); however, significantly less than between exosomes and CtxB (Fig. 2, F and G). Amiloride, i.e. an inhibitor of Na+/H+ exchange important for macropinocytotic uptake expectedly inhibited Dx10 uptake (Fig. 2H); however, amiloride had no significant effect on exosome uptake at any concentration tested (Fig. 2I). In support of a role of membrane rafts, exosome uptake was highly sensitive to membrane cholesterol depletion by MβCD. Exosome internalization was inhibited by MβCD in a dose-dependent manner, and at the highest concentration used ∼60% reduction of uptake was shown in HUVECs (Fig. 2J). Importantly, this effect was not restricted to HUVECs, as MβCD similarly inhibited exosome uptake in U87 MG cells (Fig. 2K). The uptake of fluorescently labeled Tfn was intact while CtxB uptake was expectedly inhibited (Fig. 2L), confirming that MβCD preferentially inhibits non-clathrin dependent endocytosis under these conditions. MβCD could potentially also disrupt cholesterol-rich domains of the exosomal membrane. We therefore performed experiments with a statin (simvastatin), which acts through inhibition of the rate-limiting enzyme of cholesterol biosynthesis, 3-hydroxyl-3-methylglutaryl coenzyme A (HMG-CoA) reductase, and that is widely used as a cholesterol lowering drug in humans. Notably, simvastatin was previously shown to reduce intracellular cholesterol levels in HUVECs (23). We show that simvastatin can dose-dependently inhibit exosome internalization in these cells (Fig. 2M).</p><!><p>Endocytic uptake of exosomes requires intact lipid membrane rafts. A, confocal microscopy analysis shows no colocalization of Tfn (turquoise, upper panel) or AcLDL (red, lower panel) with exosomes (red/turquoise) at 30 min. Scale bars, 15 μm. B, knockdown validation of clathrin heavy chain using siRNA against clathrin (si Clath) as compared with negative control sequence (si NT) and normalized to α-tubulin (C). D, flow cytometry analysis of exosome uptake shows no difference in si Clath as compared with si NT-transfected cells. E, confocal microscopy analysis shows colocalization of CtxB (turquoise) and exosomes (red) at 30 min of uptake in HUVECs (upper panel) and U87 MG cells (lower panel). Scale bars, 15 μm. F, confocal microscopy analysis shows limited colocalization of Dx10 (turquoise) and exosomes (red) at 30 min of uptake in HUVECs. Scale bars, 15 μm. G, weighted colocalization coefficients display 20% (mean value) colocalization of CtxB and exosomes, and ∼15% for exosomes and 10 kDa dextran (Dx10). Colocalization coefficients were calculated (Dx10/exosomes, n = 23 cells; CtxB/exosomes, n = 22 cells) using Zeiss Zen software. *, p = 0.0182. All images were captured using a C-Apochromat 63X/1.20W korr M27 objective using laser gain of 6.0% in both lasers. H, macropinocytosis inhibitor amiloride (100 μm) decreases Dx10uptake (*, p = 0.01) while Tfn uptake is less affected. I, amiloride has no significant effect on exosome uptake at a wide range of concentrations. J and K, cholesterol-depleting drug MβCD dose-dependently inhibits exosome uptake. J, HUVECs (p values, *, 0.005, **, 0.0023, ***, 0.0008); K, U87 MG cells (*, p = 0.068). L, uptake of Tfn is not affected by MβCD (2.5 mm), while CtxB and exosome uptake are reduced, suggesting specific inhibition of lipid raft-dependent uptake (p values, *, 0.0006, **, 0.02). M, simvastatin dose-dependently inhibits exosome uptake (p values, *, 0.014, **, 0.011). N, sequestration of lipid rafts by filipin III substantially inhibits exosome uptake (left panel) while Dx10 (middle panel) and Tfn (right panel) uptake are less affected (*, p = 0.001). Presented graphs show mean fluorescent values (10,000 events/sample) of one of three representative experiments (n = 3) with similar results ± S.D.</p><!><p>Non-classical endocytosis pathways (also defined as lipid raft associated pathways) can be either dependent or independent of CAV1 (2, 4). Caveolae-dependent endocytosis is a well studied clathrin-independent pathway and shares many features with membrane lipid rafts (24, 25). Importantly, filipin III, an inhibitor of lipid raft-dependent and caveolar endocytosis, was shown to inhibit exosome uptake by ∼50% (Fig. 2N, left panel). As controls of filipin specificity, the uptake of Dx10 and Tfn was almost unaffected (Fig. 2N, middle and right panels).</p><!><p>The above data prompted further studies on the role of CAV1 in endocytic uptake of exosomes. Intriguingly, CAV1 knock out cells (MEF cav 1 (−/−)) (Fig. 3A) displayed increased levels of exosome uptake as compared with wild type cells (MEF cav 1 (+/+), here denoted as MEF WT) as demonstrated by confocal microscopy (Fig. 3B) and flow cytometry analyses (Fig. 3C). Somewhat unexpectedly, these results suggested a negative regulatory role of CAV1 in exosome uptake. This was also true for U87 MG cells, as stable knockdown of CAV1 (Fig. 3D) resulted in significantly increased uptake of exosomes (Fig. 3E, left panel). As important controls, knockdown of CAV1 did not significantly alter the uptake of clathrin-mediated endocytosis, and macropinocytosis was slightly decreased, as evaluated by Tfn and Dx10 uptake, respectively (Fig. 3E, middle and right panels). Moreover, rescue experiments in which CAV1-YFP was ectopically expressed in MEF cav 1 (−/−) (Fig. 3F), showed reduced exosome uptake by ∼50% as compared with control MEF cav 1 (−/−) cells (Fig. 3, G–I). In line with these results, HeLa cells stably transfected to highly overexpress CAV1-YFP exhibited significantly reduced uptake of exosomes (Fig. 3, J and K). We could show that these effects were not restricted to MEF and HeLa cells, as U87 MG and CHO-K1 cells transfected with CAV1-YFP plasmid displayed significantly reduced exosome uptake as compared with control plasmid transfected cells (Fig. 3L).</p><!><p>Exosome internalization is negatively regulated by CAV1. A, mouse embryonic fibroblasts from wild-type (MEF WT) and cav-1 knock-out mice (MEF cav-1(−/−)) were analyzed for CAV1 protein. B, confocal images show elevated uptake of exosomes (red) in MEF cav-1 (−/−) as compared with MEF WT cells. Scale bars, 15 μm. C, graph shows quantitative measurement of exosome uptake in MEF WT and MEF cav-1 (−/−) cells by flow cytometry, and represents mean fluorescent values (20,000 events/sample) of three independent experiments (n = 9) ± S.D. (*, p = 0.000003). D, stable knockdown (by ∼80%) of CAV1 by lentiviral shRNA transduction in U87 MG cells. E, flow cytometry analysis demonstrates increased exosome uptake (left panel), no significant difference in Tfn uptake (middle panel), and decreased uptake of Dx10 (right panel) in CAV1 shRNA cells, as compared with cells transfected with control shRNA (shRNA NC). Graph represents mean fluorescent values (20,000 events/sample) of one of three representative experiments (n = 4) ± S.D. (*, p = 0.0019 in left panel). F, flow cytometry analysis of CAV1-YFP transfection efficiency; MEF WT (gray area), MEF cav-1 (−/−) (black line), and MEF cav-1 (−/−) cells transfected with pEX_EF1_CAV1-YFP plasmid (gray line). G, introduction of CAV1-YFP in MEF cav-1 (−/−) cells (YFP, turquoise) reduces uptake of exosomes (red). Note that the high CAV1-YFP-expressing cell (arrowhead) displays reduced exosome uptake as compared with the low CAV1-YFP expressing cell. Scale bars, 15 μm. H, dot plot analysis of exosome uptake versus CAV1-YFP expression in MEF cells, as indicated. I, quantification of the uptake in H. Graph represents mean fluorescent values (20,000 events/sample) of a representative experiment (n = 4) ± S.D. (*, p = 0.0009). J, immunoblotting for CAV1 in HeLa cells stably transfected with CAV1-YFP. K, reduced uptake of exosomes in CAV1 overexpressing HeLa cells as compared with HeLa WT cells. L, reduced uptake of exosomes in transiently CAV1 overexpressing U87 MG cells (left panel) and CHO-K1 cells (right panel) as compared with control cells transiently transfected with eGFP. Graph represents mean fluorescent values (20,000 events/sample) of one out of two independent experiments (n = 3) ± S.D. (*, p = 0.018). M, no colocalization between CAV1-YFP (turquoise) and exosomes (red) using confocal microscopy analysis. Scale bars, 15 μm. N, no colocalization between CAV1-YFP (green) and exosomes (red) using TIRF microscopy analysis. Scale bars, 10 μm. O, electron microscopy colocalization studies in HUVECs of exosomes detected by α-TF 30 nm gold particles and lipid rafts distinguished by anti-flotillin-1 10 nm gold particles. Scale bars, 100 μm. Left panel: Ctl, no addition of exosomes.</p><!><p>It has previously been shown that endocytic uptake of virus particles is negatively regulated by CAV1 by stabilization of specific plasma membrane lipid raft domains (1, 26, 27). We next employed confocal microscopy co-localization studies of CAV1-YFP and PKH-labeled exosomes. As shown in Fig. 3M, we could not detect any colocalization at 30 min of incubation, suggesting that exosomes are not associated with CAV1-positive endosomal structures. Confocal microscopy cannot fully discriminate between newly internalized and cell-surface associated exosomes, i.e. the above data do not exclude the possibility of exosomal association with CAV1-containing lipid rafts at earlier time points than 30 min. We therefore applied TIRF microscopy to visualize exosomes present in the membrane region close to the cell surface, and during early phases of internalization. These analyses confirmed that exosomes do not appear to colocalize with CAV1 upon internalization (Fig. 3N). Together, these findings indicate that the negative regulatory function of CAV1 does not occur through stabilization of lipid raft domains directly involved in exosome uptake. To corroborate confocal microscopy data, showing that internalized exosomes are associated with lipid rafts (Fig. 2E), we next performed immunoelectron microscopy colocalization studies of exosomes (by tissue factor antibody staining to discriminate from endogenous endosomes) and the lipid raft marker flotillin-1. We found a substantial colocalization at 30 min of internalization, signifying that exosomes indeed are taken up through lipid rafts (Fig. 3O).</p><!><p>Our findings that CAV1 negatively regulates membrane raft-dependent exosome uptake without apparent co-localization with exosomes, prompted further mechanistic studies on the role of CAV1. Apart from its role as a structural component of lipid rafts, CAV1 has been shown to modify the activity of several signaling proteins, such as Src family members, epidermal growth factor receptor, and integrins (28–30). Consequently, we set out to explore potential signaling mechanisms involved in exosome uptake. In initial experiments, using phospho-kinase arrays, we found that short-term incubation with exosomes resulted in 2–4.5-fold induction of several lipid raft associated proteins; p-FAK, the heat-shock protein p-HSP27, and p-ERK1/2 and its downstream target p-MSK1/2 (Fig. 4, A and B) (for a complete list of relative phospho-kinase levels, see supplemental Fig. S2). Exosomal induction of p-ERK1/2 and p-HSP27 were validated by Western blotting (Fig. 4, C and F). The mitogen-activated protein kinase (MAPK) pathway may be initiated at the cell surface and continue during endosomal sorting, while more recent studies suggest that MAPK signaling is a required element of endocytosis (31). Interestingly, pharmacological targeting of ERK1/2 signaling using the specific inhibitor.</p><!><p>Exosome internalization depends on ERK1/2 and HSP27 signaling activation and an intact cytoskeleton. A, levels of phosphorylated kinases in HUVECs with no treatment (Ctl), or incubated with exosomes for 10 min (10 min Exo) or 30 min (30 min Exo). As a comparison, same protein amount of exosomes was used to visualize phosphoproteins residing in exosomes (Exo content). B, quantification of the mean value (n = 2 in each blot) of p-ERK1/2, p-MSK1/2, p-HSP27, and p-FAK relative unstimulated cells (Ctl). C, relative protein levels of p-ERK1/2 with or without exosome stimulation in the absence or presence of U0126. D, reduced exosome uptake in HUVECs treated with U0126. Graphs represents mean fluorescent values (20,000 events/sample) of one of three independent experiments (n = 3) (*, p = 0.01, **, p = 0.002) ± S.D. E, reduced exosome uptake in HeLa cells treated with U0126, expressed as mean fluorescent values (20,000 events/sample) of one of two independent experiments (n = 4) (*, p = 0.000009) ± S.D. F, induction of p-HSP27 protein by exosomes is counteracted by ERK1/2 inhibition using U0126. G, representative blot of relative HSP27 protein expression after siRNA knockdown. H, reduced exosome (red) uptake in HSP27 knockdown cells. White, f-actin; blue, nuclei. Scale bars, 15 μm. I, flow cytometry analysis of cells in H from a representative experiment (n = 4) (*, p = 0.002) ± S.D. J, actin cytoskeleton phalloidin stainings in HSP27 siRNA and siRNA NT-transfected cells. Note the abnormal cytoskeleton in siHSP27-transfected cells (right panel, red arrowheads). Scale bars, 100 μm. K, quantification of the average number of cells with abnormal cytoskeleton relative the total number of cells counted in seven independent microscopic fields; siNT cells (n = 116) and siHSP27 knockdown cells (n = 209). Data are mean values ± S.D. (*, p = 0.00018). L, disruption of the actin cytoskeleton (f-actin, white) using 0.5 μm Cytochalasin D or Lantrunculin A. Scale bars, 100 μm. M and N, exosome uptake in cells treated with varying concentrations of Cytochalasin D (M) or Lantrunculin A (N). Graphs are mean fluorescent values (10,000 events/sample) of two independent experiments with similar results ± S.D. All values (*) were significantly different from control (0 μm) with a p value of < 0.0001.</p><!><p>U0126 (Fig. 4C) dose-dependently decreased exosome uptake in HUVECs (Fig. 4D) as well as in HeLa cells (Fig. 4E). We next turned our interest to exosome-dependent induction of p-HSP27 (Fig. 4, A and B) as HSPs besides from being localized in the cytosol, have been associated with the cellular membrane and lipid rafts (32). Accordingly, previous reports have shown a role of p-HSP27 in macromolecular internalization by regulation of actin cytoskeleton dynamics (33, 34). We found that inhibition of ERK1/2 signaling dose-dependently reduced p-HSP27 levels (Fig. 4F), suggesting that, at least in the context of exosome uptake, HSP27 is activated downstream of ERK1/2. A direct role of HSP27 as an effector molecule in exosome internalization was supported by significantly reduced exosome uptake upon siRNA-mediated knock-down of HSP27 (Fig. 4, H and I). In agreement with the previously understood role of HSP27 in cytoskeleton rearrangement, cells transfected with siRNA for HSP27 as compared with non-target siRNA exhibited an abnormal actin cytoskeleton as assessed by phalloidin staining (Fig. 4, J and K). Pharmacological disruption of the actin cytoskeleton (Fig. 4L) using Cytochalasin D or Lantrunculin A inhibited uptake of exosomes under similar conditions (Fig. 4, M and N), reinforcing the notion that an intact actin cytoskeleton is required for efficient exosome uptake. We conclude that exosomes may trigger lipid raft mediated endocytosis through signaling activation of ERK1/2-dependent pathways that include a specific role of HSP27.</p><!><p>From the above findings, we next explored the hypothesis that CAV1 negatively regulates exosome uptake through interference with ERK1/2 signaling. We could appreciate that the induction of p-ERK1/2 during exosome uptake was enhanced in MEF cav-1 (−/−) as compared with MEF WT cells at all time points tested (Fig. 5A). Previous studies suggested CAV1-mediated down-regulation of MAPK signaling in rodents but not in human fibroblasts (35); however, in the context of exosome uptake we observed CAV1-mediated suppression of p-ERK1/2 induction in both mouse (MEF) and human (HeLa)-derived cells (Fig. 5, A and D). Consistent with the findings in HUVECs (Fig. 4D) and HeLa cells (Fig. 4E), reduced levels of p-ERK1/2 by U0126 treatment in MEF cells (Fig. 5B) resulted in decreased internalization of exosomes (Fig. 5C). More importantly, enhanced exosome uptake by CAV1-defiency could be directly linked to ERK1/2 signaling, as U0126 treatment efficiently counteracted exosome uptake in MEF cav-1 (−/−) cells (Fig. 5C). Accordingly, in HeLa-CAV1-YFP overexpressing as compared with wild-type HeLa cells, we found reduced capability of exosomes to induce p-ERK1/2 signaling and subsequent p-HSP27 induction at multiple time points of exosome stimulation (Fig. 5D). We conclude that exosome uptake commences through lipid raft-mediated endocytosis associated with and dependent on signaling activation of ERK1/2 and HSP27. Further, our data indicate that CAV1 localized in the plasma membrane negatively regulates endocytic uptake of exosomes at least partly through suppression of ERK1/2 signaling activation (for a schematic overview, see Fig. 5E).</p><!><p>CAV1 negatively regulates ERK1/2-dependent endocytosis of exosomes. A, shown is a representative blot for p-ERK1/2 and total ERK1/2 (t-ERK1/2) in MEF WT and MEF cav-1 (−/−) cells stimulated with exosomes for the indicated times (left panel), and quantification of relative protein levels (right panel). B, reversal of exosome-mediated induction of p-ERK1/2 by U0126 in MEF cells. Graph shows quantification of Western blot analysis. C, up-regulation of exosome uptake in CAV1-deficient cells is counteracted by ERK1/2 inhibition. Graph represents mean fluorescent values (20,000 events/sample) of one of three independent experiments with similar results ± S.D. *, p = 0.02, **, p = 0.00001. D, overexpression of CAV1-YFP suppresses exosome-mediated induction of p-ERK1/2 and p-HSP27 in HeLa cells. Shown are representative blots for p-ERK1/2, t-ERK1/2, p-HSP27, total HSP27, and tubulin (upper panel), and quantification of relative protein levels (lower panel) in HeLa WT and CAV1-YFP cells stimulated with exosomes for the indicated times. E, schematic figure of the major findings of the present work. Exosomes are internalized by lipid raft-associated endocytosis, which is under negative control by CAV1. Additional signaling proteins involved are ERK1/2 and HSP27, and probably additional ERK1/2 downstream targets. HSP27 is known to be involved in rearrangement of the actin cytoskeleton important for the invagination of the plasma membrane during endocytosis.</p><!><p>Here, we present several significant findings that advance our understanding of exosome uptake at the level of membrane uptake pathways and signaling regulation. First, we establish that exosome uptake mainly occurs through non-clathrin dependent, lipid raft-mediated endocytosis. To circumvent limitations caused by the high sensitivity of labeled exosomes to detergents and fixation solutions, we have in our studies applied live microscopy using fluorescent fusion protein constructs, fluorophore-labeled ligands and/or mild fixation procedures in combination with electron microscopy studies. Data from these studies strongly suggest that exosomes are internalized by endocytosis rather than by membrane fusion at the plasma membrane. More importantly, we identify specific signaling mechanisms implicated in the uptake process, and find an unexpected and significant role of CAV1 in regulating exosome uptake. Our data suggest that negative regulation of endocytic uptake of exosomes occurs through stabilization of lipid rafts by CAV1. However, these results do not exclude multilevel signaling through the described or other pathways regulated by or independent of CAV1. Further studies investigating other related pathways are of high interest.</p><p>Although we provide convincing evidence that ERK1/2 is activated by exosomes especially in the context of CAV1 deficiency, and that ERK1/2 activity is required for efficient exosome uptake, our studies do not fully elucidate the protein interactions involved. A reciprocal regulation of ERK1/2 and CAV1 has previously been described in other systems, where the activation of ERK1/2 was suppressed by CAV1, and the up-regulation of ERK1/2 in turn was shown to suppress CAV1 mRNA levels (36). Moreover, ERK1/2 was shown to localize to caveolae in the plasma membrane, and the caveolar localization of ERK1/2 negatively regulated further signal transduction to the nucleus (36, 37). In the context of exosome internalization, however, we failed to demonstrate a direct protein interaction of CAV1 and ERK1/2, as determined by confocal microscopy analyses and anti-CAV1 antibody pulldown experiments (supplemental Fig. S3, A and B). The protein interactions responsible for CAV1-mediated negative regulation of ERK1/2 signaling during exosome uptake remains has to be determined in future studies. Based on our findings of a role for CAV1 and ERK1/2 in exosome internalization, it may be speculated that uptake of extracellular vesicles is governed by the signaling status of recipient cells. In the context of the tumor microenvironment, this may be determined by specific oncogenetic events in malignant cells and the availability of e.g. growth factors, cytokines, and their respective receptors in the stromal compartment. Thus, future studies should explore the possibility of differential transfer of vesicles in the context of CAV1 and ERK1/2 expression, as these proteins are frequently deregulated at various stages of tumor development (38, 39). Moreover, as exosome-based delivery of mRNA and miRNA has been proposed as a feasible, therapeutic approach in cancer and other pathological conditions (40), further comprehensive understanding of vesicular uptake mechanisms should offer a more rational design of exosomes as drug delivery vehicles.</p><!><p>This work was supported by grants from the Swedish Cancer Fund; the Swedish Research Council; the Swedish Society of Medicine; the Physiographic Society, Lund; the Crafoordska, Gunnar Nilsson, Lundbergs, and Kamprad Foundations; the Skåne University Hospital donation funds; and the Governmental Funding of Clinical Research within the National Health Services (to A. L. F.).</p><p>This article contains supplemental videos S1–S4 and Figs. S1–S3.</p><p>multivesicular body</p><p>glioblastoma</p><p>caveolin</p><p>heat shock protein</p><p>nanoparticle tracking analysis</p><p>total internal reflection fluorescence</p><p>tissue factor</p><p>transferrin</p><p>cholera toxin</p><p>methyl-β-cyclodextrin</p><p>3-hydroxyl-3-methylglutaryl coenzyme A.</p>

PubMed Open Access

Structural Basis for Multiple Sugar Recognition of Jacalin-related Human ZG16p Lectin*

Background: ZG16p is a soluble mammalian lectin with a Jacalin-related β-prism-fold.Results: ZG16p binds to short α-mannose-related glycans and glycosaminoglycans via the canonical shallow mannose-binding pocket and an adjacent basic surface area, respectively.Conclusion: ZG16p possesses a unique feature of multiple-ligand binding.Significance: Structural insights of ZG16p ligand binding are crucial for understanding the biological functions of the protein.

structural_basis_for_multiple_sugar_recognition_of_jacalin-related_human_zg16p_lectin*

Introduction<!>Materials<!>Preparation of the Heparin and Hyaluronic Acid Oligosaccharide Fractions<!>Synthesis of Man-α-O-Ser/Thr<!>Synthesis of Man-α-O-heptapeptide<!>Protein Expression and Purification<!>Glycan Microarray Analyses<!>NMR Titration Experiments<!>Crystallization, X-ray Data Collection, and Structure Determination<!><!>Ligand Preference of ZG16p<!><!>NMR Titration Experiments<!><!>ZG16p Binding to α-O-Methylmannose<!><!>Recognition of Mannose-α-O-Serine/Threonine<!><!>Recognition of Mannose-α-O-Serine/Threonine<!>Two Binding Modes for Disaccharides<!><!>Two Binding Modes for Disaccharides<!>Heparin Binding Is Independent from Mannose Binding<!><!>Heparin Binding Is Independent from Mannose Binding<!><!>Heparin Binding Is Independent from Mannose Binding<!><!>Heparin Binding Is Independent from Mannose Binding<!>

<p>Zymogen granule protein 16 (ZG16p)5 is a 16-kDa soluble protein, originally identified by screening a cDNA expression library of rat pancreas with a polyclonal antibody to purified rat zymogen granule membranes (1). Early on, rat ZG16p was postulated to be a carbohydrate-binding protein on account of its amino acid similarity to the plant lectin Jacalin (1, 2). To gain clues to the function of the ZG16p tissue expression profile, localization and putative carbohydrate binding were investigated. By Northern blot analysis expression of rat ZG16p mRNA was detected in the pancreas, duodenum, and colon (1), and by quantitative PCR analysis human ZG16p mRNA was detected in the liver, colon, small intestine, and pancreas (3, 4). By immunochemical staining, rat ZG16p was found in acinar cells of pancreas and goblet cells of duodenum and colon (1). Human ZG16p was similarly investigated, and detected in mucus-secreting cells of the digestive system also including acinar cells of the pancreas and goblet cells of the intestine, as well as serosanguineous acinar cells of the parotid gland (4). From the expression pattern and localization, it was deduced that the main function of ZG16p is in the digestive system.</p><p>Expression level of ZG16p mRNA was reported to be regulated by various factors. Dexamethasone, which is known to induce high rates of synthesis of secretory enzymes and secretion via zymogen granules (5, 6), increased remarkably the ZG16p gene expression in the rat pancreatic tumor cell line AR4-2J (1). Cerulein, a peptide hormone that up-regulates expression and transport of zymogens, led to short term down-regulation of ZG16p mRNA in the mouse pancreas (7). It was shown that the mRNA level of ZG16p was down-regulated in human hepatocellular carcinoma (3). The varying expression levels of ZG16p raises the possibility that the protein is able to perform different functions.</p><p>ZG16p lacks a transmembrane region. The protein is associated with the luminal surface of pancreatic zymogen granule membrane (ZGM) in cholesterol-glycosphingolipid-enriched microdomains together with pancreatic secretory granule membrane major glycoprotein GP-2, syncollin, and sulfated proteoglycans (8, 9). ZG16p was identified in the ZGM fraction by proteomic analyses of rat pancreatic zymogen granules (10, 11). There is evidence for the functional importance of ZG16p in the process of selective packing and sorting of pancreatic enzymes to the ZGM in pancreatic acinar cells (referred to as condensation-sorting). Pretreatment of ZGM with anti-ZG16p antibody resulted in a 50–60% inhibition of condensation-sorting (2). The condensation-sorting and the association of ZG16p with ZGM were also inhibited by pretreatment of ZGM with chondroitinase ABC or heparinase (2). ZG16p was therefore suggested to mediate condensation-sorting as part of a proteoglycan/glycoprotein scaffold at the luminal side of the ZGM through its binding to glycosaminoglycans (GAGs) (2, 8). Recent structural analysis of the purified rat pancreatic zymogen granule proteoglycans revealed the presence of highly sulfated heparan sulfate chains (12). Binding assays showed that rat ZG16p interacts with heparan sulfate proteoglycans via their GAG chains (12). In addition, it was shown that ZG16p can bind to heparin (12), which is a binding partner described for several digestive enzymes in pancreatic granules (13, 14). ZG16p is a highly basic protein, and the isoelectric point of rat ZG16p is approximately pH 9.0 (2, 11, 15). It has been postulated that the positively charged lysine and arginine residues on the ZG16p surface are involved in the interactions with sulfated GAGs (2). In support of this, a site-directed mutagenesis study has indicated that the basic amino acid residues Lys33, Lys36, Arg37, Arg55, and Arg58 on rat ZG16p are involved in heparin binding (12).</p><p>We described earlier the crystal structure of human ZG16p, and the three-dimensional structure revealed a β-prism I-fold, the first to be described in mammals (16). The β-prism-fold consists of three β-sheets and each is made up of three to four β-strands. The human ZG16p structure is similar to those of the other Jacalin-related mannose-binding lectins: Banlec (17), Heltuba (18), Atrocarpin (19), MornigaM (20), Parkia lectin (21), and Griffithsin (22). These lectins do not require Ca2+ ions for mannose binding and their carbohydrate-binding sites consist of three exposed loops, GG loop, recognition loop, and binding loop at the top of the β-prism I-fold (23). The GXXXD motif is a common carbohydrate binding motif in the Jacalin-related mannose-binding lectins (23). In our previous structural study it was observed that the putative carbohydrate-binding site of ZG16p was occupied by a glycerol molecule, which mimics part of the mannose residue bound by Jacalin-related lectins in complex with mannosyl ligands (16). Human ZG16p was shown to bind to an α-linked mannose-polyacrylamide-biotin probe, and there was binding also to the β-linked anomer (4). Mutation of the evolutionarily conserved amino acid residue Asp151 in human ZG16p, which is involved in mannose binding in Jacalin-related mannose-binding lectins, was shown to abolish the binding to mannose (4). Furthermore, ZG16p was shown to bind to pathogenic fungi, Candida, and Malassezia species, and the binding was inhibited by Candida albicans mannan (4). Considering the mannose-binding ability, it was speculated that secreted ZG16p in the intestinal mucus layer might protect against invading pathogens (4).</p><p>The recent studies on ZG16p suggest roles of ZG16p in both the luminal side of ZGM and in the mucus layer of the digestive system (1, 4). However, the relationship of functions of ZG16p to its carbohydrate-binding mechanism is still not fully understood. Also no structural information is available to date on ZG16p in complex with its ligands. Here, based on the binding specificities of human ZG16p revealed in glycan microarray analyses, we performed crystallographic and NMR studies of ZG16p in complex with carbohydrate ligands.</p><!><p>Glcβ1–3Glc was purchased from Seikagaku Corp., and Manα1–3Man was from Sigma and Santa Cruz Biotechnology. α-O-Methyl-mannose and β-O-methyl-galactose were from Sigma, and heparin tetrasaccharide from Iduron.</p><!><p>The oligosaccharide fractions (with an estimated size of 20–25-mer) of heparin and hyaluronic acid were prepared by controlled digestion with heparinase I and hyaluronate lyase followed by fractionation by gel filtration chromatography as described (24, 25).</p><!><p>Chemical synthesis of Man-α-O-Ser and Man-α-O-Thr was performed as described previously (26).</p><!><p>Fmoc-O-(2,3,4,6-O-tertaacetyl-α-d-mannopyranosyl)-l-threonine was synthesized as described (26). The desired peptide sequence Val-Glu-Pro-(Man-α-O-Thr)-Ala-Val-Ala, which is a fragment of O-mannosylated dystroglycan (27), was synthesized with solid phase peptide synthesis using an Applied Biosystems peptide synthesizer (ABI 433A). Conventional Fmoc strategy was employed for the elongation of the peptide sequence and coupling with the acetylated Man-α-O-Thr building block. After cleavage from the resin under mild acidic conditions, the glycopeptide with acetylated Man was obtained after HPLC purification. The acetyl groups were removed with 0.05 m NaOMe in MeOH (28). Upon stirring for 1.5 h at room temperature, the reaction mixture was directly applied for size exclusion chromatography (Sephadex G-15, GE Healthcare, H2O) to give the desired Man-α-O-heptapeptide; m/z (MALDI-TOF) Found: [M + Na]+, 911.39; C38H64N8O16Na requires [M + Na]+, 911.43.</p><!><p>Human ZG16p(21–159) for initial glycan microarray analyses and crystallization was expressed in Escherichia coli as a (His)6-MBP-fused form as described previously (16). The fused protein was purified using a Ni-Sepharose column (GE Healthcare). For structural studies the protein was treated with tobacco etch virus protease and the (His)6-MBP tag was removed by the Ni-Sepharose column. The untagged ZG16p proteins were further purified with size exclusion chromatography (HiLoad 16/60 Superdex 75 pg, GE Healthcare). Uniformly 15N-labeled ZG16p(21–167) for NMR study was prepared by culturing E. coli in 15N-labeled Spectra 9 medium (Cambridge Isotope Laboratories, Inc.) using pCold-PDI vector (29). The purification procedure of 15N-labeled ZG16p(21–159) was essentially the same as described above. For additional glycan microarray analyses, human ZG16p(21–167) was expressed in E. coli using pCold-GST vector (30) and the GST-fused proteins and GST alone were purified using the Glutathione-Sepharose 4B column (GE Healthcare) according to the manufacturer's instructions. The purified GST-ZG16p proteins were dialyzed against PBS (8.1 mm Na2HPO4, 1.5 mm KH2PO4, 137 mm NaCl, 2.7 mm KCl, pH 7.4) including 0.02% (w/v) sodium azide, 1 mm EDTA, protease inhibitor mixture (Complete, EDTA-free, Roche Applied Science), 2 mm DTT, 2% (v/v) glycerol.</p><!><p>Microarray analyses were performed using the neoglycolipid-based system in which the glycan probes are lipid-linked and include neoglycolipids and glycolipids (31). These were robotically printed on nitrocellulose-coated glass slides at 2 and 5 fmol/spot using a non-contact arrayer, and analyses were performed as described (32, 33). For the analysis of (His)6-MBP-fused ZG16p, the results of 492 oligosaccharide probes at 5 fmol/spot are shown in the supplemental Microarray Data. For analysis of GST-ZG16p, a different version of the microarray was used; binding results of the 62 oligosaccharide probes at 5 fmol/spot are shown in supplemental Table S1.</p><p>For binding analyses, microarray slides were blocked at ambient temperature for 60 min with 3% (w/v) BSA in PBS. For the (His)6-MBP-tagged ZG16p, the protein was analyzed at 200 μg/ml, and the binding was detected with anti-His and biotinylated anti-mouse IgG (both from Sigma and used at 10 μg/ml). The GST-tagged proteins, GST-ZG16p and GST-ZG16p (Y104F), were overlaid at 200 μg/ml followed by rabbit anti-GST antibody Z-5 (Santa Cruz), 1:200, and then by biotinylated anti-rabbit IgG (Sigma), 1:200. Biotinylated BanLec (EY lab) was tested at 5 μg/ml. Binding was detected with Alexa Fluor 647-labeled streptavidin (Molecular Probes) at 1 μg/ml. After quantification, microarray data analyses and presentation were performed using dedicated software (34). On-array inhibition experiments were performed with 100 μg/ml of GST-ZG16p in the presence of Man-α-OMe or Gal-β-OMe (1 and 10 mg/ml), and at 200 μg/ml of GST-ZG16p in the presence of heparin or hyaluronic acid oligosaccharide fractions (1 and 3 mg/ml).</p><!><p>NMR experiments were performed at 298 K using a cryoprobe-equipped 500 MHz spectrometer (BrukerBiospin). The backbone amide signals of 13C/15N-labeled ZG16p(21–167) were assigned sequentially via the analysis of the two-dimensional 1H-15N HSQC, and three-dimensional HNCA, HN(CO)CA, HNCACB, and CBCA(CO)NH spectra.6 NMR titration experiments were carried out by acquiring 1H-15N HSQC spectra on samples of 0.1–0.15 mm 15N-labeled ZG16p with the addition of increasing amounts of carbohydrate ligands. Chemical shift differences were calculated as Δδ = [(ΔδH)2 + (0.2 × ΔδN)2]1/2, where ΔδH and ΔδN are the observed chemical shift changes for 1H and 15N, respectively. For determination of the dissociation constant (Kd), Δδ was plotted as a function of the molar ratio (ligand:protein) and the titration curve was fitted using the maximum chemical shift change and dissociation constant as variable parameters. 15N-Labeled ZG16p and ligands were dissolved in PBS, pH 6.5, including 10% D2O. Data were collected with 1024 (F2) × 256 (F1) data matrix points with either 4 or 8 scans. 1H NMR chemical shifts indicated with parts per million (ppm) were calibrated based on an outer standard of a chemical shift of 4,4-dimethyl-4-silapentane-1-sulfonic acid, given a singlet at 0 ppm. 15N chemical shifts (ppm) are calibrated using indirect reference based on the IUPAC-IUB recommended 15N/1H resonance ratio of 0.10132911 (35).</p><!><p>Crystals of human ZG16p(21–159) were obtained by sitting drop vapor diffusion under a previously reported condition (16). All ZG16p·ligand complex crystals were obtained by using soaking ligand-free ZG16p crystals with carbohydrate ligand dissolved in reservoir solution (0.09 m MES, pH 6.5, 0.09 m NaH2PO4, 0.09 m KH2PO4, and 1.8 m NaCl) at 0.1 mg/μl concentration. Data sets were collected from synchrotron radiation (1.0000-Å wavelength) at BL5A, NE3A, and NW12A beamlines of the Photon Factory, High Energy Accelerator Research Organization (KEK) (Tsukuba, Japan). The crystals were cryo-protected with a reservoir solution containing 25% (v/v) ethylene glycol. The diffraction data were processed using HKL2000 (36). The structures of the ZG16p·ligand complexes were solved by the molecular replacement method using the program Molrep (37) with ligand-free ZG16p structure (PDB code 3APA) (16). Further model building was manually performed using the program COOT (38). Refinement was carried out using programs CNS1.1 (39) and REFMAC5 (40). The stereochemical quality of the final models was assessed by PROCHECK (41). Crystallographic parameters and refinement statistics are summarized in Table 1.</p><!><p>Statistics of crystallographic data</p><p>Values in parentheses are for the highest resolution shell.</p><!><p>Previous reports have revealed two different types of ligands for ZG16p. One is sulfated GAGs, in particular heparan sulfate and heparin (12), and the other is mannose (4). This dual binding specificity seems to be a unique feature of ZG16p. To examine in greater detail the carbohydrate-binding specificity of ZG16p, we performed binding analyses using a neoglycolipid-based glycan microarray system (33), initially with a (His)6-MBP-fused ZG16p construct and using an array comprising 492 lipid-linked oligosaccharide probes (supplemental Microarray Data). These encompassed a variety of mammalian type sequences, representative of N-glycans (pauci- and high mannose-type and complex type), peripheral regions of O-glycans; blood group antigen-related sequences on linear or branched backbones and their sialylated and/or sulfated analogs; linear and branched poly-N-acetyllactosamine sequences, gangliosides, oligosaccharide fragments of GAG and polysialic acid. The arrays also included size-defined fragments (homo-oligomers) of microbial and plant-derived glucan polysaccharides. Interestingly, the results showed very strong binding to O-mannosyl-related probes, Ser/Thr-linked O-mannose (numbers 335–338). In contrast, no detectable binding was observed for high-mannose N-glycans Man8GN2 and Man9GN2 and only weak binding was observed to pauci-mannose N-glycans, such as Man2GN1 and Man3GN2 (supplemental Microarray Data). Binding signals of various intensities were also observed to glucosyl disaccharides with differing linkages and to glucan oligosaccharide sequences, in particular the malto-oligosaccharide series (with α1–4-glucosyl linkages). The (His)6-MBP tag itself only showed marginally detectable binding signals with glucan oligosaccharides of the malto-series but not to other probes (supplementary Microarray Data). The lack of binding to GAG-related sequences could be due to the relatively low affinity of the protein to GAG oligosaccharides presented on the array surface also due to the protein being analyzed in monomeric form.</p><p>To address the concern that binding observed for glucose-related probes with the (His)6-MBP tagged construct may be nonspecific, and to investigate the GAG binding property of ZG16p, GST-tagged ZG16p was used for additional microarray analyses. The dimeric nature of the GST construct (42, 43) is postulated to be advantageous for the detection of potentially weak ligand binding compared with the monomeric (His)6-MBP construct. These additional analyses were carried out with a focused microarray composed of 62 glycan probes including those of diverse N-glycan sequences and GAG oligosaccharides, as well as those of the O-linked glycans (monosaccharides linked to Ser/Thr) and glucan oligosaccharides (supplemental Table S1). GST alone did not show significant binding to any of the glycans tested (data not shown). In contrast, positive signals were observed with GST-tagged ZG16p. The probes bound can be categorized into 5 groups: (i) short α-manno-oligosaccharides (Manα1–3Man, Manα1–3[Manα1–6]Man, Manα1-3Manβ1–4GlcNAc), (ii) oligomannose with a 6-phosphate group, (iii) α-mannose attached to Ser/Thr, (iv) β-glucan hexasaccharide (Glcβ1–3Glcβ1–3Glcβ1–3Glcβ1–3Glcβ1–3Glc), and (v) sulfated GAG oligosaccharides (the 16-mer of heparin and chondroitin sulfate B) (Fig. 1A). The short α-mannose glycans bound have the Manα1–3Man unit in common, whereas the pauci and high mannose N-glycans having terminal Manα1–3Man unit did not give strong signals. For comparison, we performed the same microarray analysis using the well characterized mannose-binding β-prism lectin, Banlec (17). In contrast to ZG16p, Banlec bound to most of the pauci and high mannose N-glycans (Fig. 1D). The differences in binding specificities toward pauci- and high mannose N-glycans by ZG16p and Banlec is possibly attributed to the different valency. Banlec monomer has two potential mannose-binding sites (first and second sites), and exists as a homodimer. Thus four ligand-binding sites are potentially available for Banlec, whereas ZG16p is envisaged to have only the first mannose-binding site homolog, because the second site homolog does not have the conserved GXXXD motif (Fig. 2). Microarray analyses revealed "new" binding partners for ZG16p, phosphorylated high mannose glycans, Man-O-Ser/Thr, and β-glucan oligomer. The relatively strong binding to Man5GlcNAc2 and Man6GlcNAc2 with 6-phosphate groups may be attained, we propose, by α-mannose binding to the ligand binding pocket coupled with the interaction of phosphate group with a positively charged patch composed of Lys102, Lys106, and Lys122 (16). Among the GAG sequences tested, oligosaccharides of heparin and chondroitin sulfate B (CSB) were significantly bound by ZG16p. Both heparin and CSB are rich in iduronic acid residues, the conformational flexibility (44) of which (compared with glucuronic acid found in hyaluronic acid, chondroitin sulfates A and C) may be one of the determinants in the interaction of heparin and CSB with ZG16p via anionic groups matching. Heparan sulfate, which was recently reported to be a possible endogenous partner for ZG16p, also has iduronic acid, in particular the highly 6-O-sulfated domains. The detailed survey on the heparan sulfate motifs bound by ZG16p is beyond the scope of this study. Banlec did not bind to any GAG sequences tested, likely due to its negatively charged surface potential (Fig. 3).</p><!><p>Glycan microarray analyses. GST-fused human ZG16p (A), mutant GST-ZG16p (Y104F) (B), mutant GST-ZG16p (D151N) (C), and biotinylated BanLec (D) were overlaid. Numerical scores of the binding signals are means of duplicate spots at 5 fmol/spot (with error bars). The results are representative of multiple analyses. The complete list of probes and their sequences are provided in supplemental Table S1.</p><p>Sequence alignment of human ZG16p, rat ZG16p, and Jacalin-related mannose binding-type lectins. Pink boxes indicate residues involved in mannose binding at the first sugar-binding site. Proposed heparin-binding residues (Lys33, Lys36, Arg37, Arg55, and Arg58 in rat ZG16p) and conserved residues in human ZG16p (Lys36, Arg37, and Arg55) are shown in purple. Second binding site in Banlec and Griffithsin is shown in light green. Asp151 in ZG16p and corresponding aspartic acid residues in other lectins are colored in cyan. Secondary structure of human ZG16p is shown above the amino acid sequence. Sequences are aligned with MATRAS (52) and CLUSTAL W (53).</p><p>Electrostatic surface potential of ZG16p and Banlec. The surface models of human ZG16p (PDB code 3VZF, upper) and Banlec (PDB code 1X1V, lower) are colored according to the electrostatic surface potential (blue, positive; red, negative; scale from −8 to 8 kT/e). Bound ligands are shown as a ball-and-stick model.</p><!><p>To confirm the binding capability of these glycans for ZG16p, we performed NMR titration experiments to determine the dissociation constant (Kd) for each ligand. Due to the low affinities of many carbohydrate-protein interactions, we monitored the binding by using 1H-15N HSQC titration experiments. Five ligands, Manα1–3Man, Glcβ1–3Glc, Manα-O-Thr, and Man-α-O-heptapeptide (Val-Glu-Pro-(Man-α-O-Thr)-Ala-Val-Ala), and heparin tetrasaccharide, were used in the NMR. Upon addition of each ligand, a particular set of signals were specifically perturbed (supplemental Fig. S1, Fig. 9A). The dissociation constants were determined using the chemical shift changes (Fig. 4 and Fig. 9B) and summarized in Table 2. All five ligands were bound weakly by ZG16p: the dissociation constants were in the 2–20 mm range. The presence of flanking peptides at both N-terminal and C-terminal sides of Man-α-O-Thr did not significantly affect the binding to ZG16p. The relatively weak binding of ZG16p toward these ligands is in part due to the monovalent nature of the ZG16p-carbohydrate interaction. Dissociation constants of mannose-immobilized agarose and yeast mannan to ZG16p were reported to be 1.3 and 1.7 μm, respectively (4). Relatively strong binding to mannose-immobilized agarose and yeast mannan is attributed to the polyvalent nature of mannose residues possibly by reducing the dissociation rate constant.</p><!><p>Titration curves for the selected peaks in two-dimensional NMR spectra of 15N-labeled ZG16p. (A) Man-O-Thr, (B) Man-O-peptide, (C) Manα1–3Man, and (D) Glcβ1–3Glc are shown. The corresponding 1H-15N HSQC spectra are shown in supplemental Fig. S1.</p><p>Kd values of ZG16p from NMR titration experiments (n = 3)</p><p>a Val-Glu-Pro-(Man-α-O-Thr)-Ala-Val-Ala.</p><!><p>To gain insight into the mode of interaction of ZG16p with the various ligands, we conducted x-ray crystallographic studies in the presence of ligands. Initial soaking experiments using various sugars were unsuccessful, because of the competition of the high concentration of glycerol present (20–25%, v/v). In previous experiments, glycerol was observed to bind to the mannose-binding site, mimicking the sugar ligand (16). After optimization of cryoprotectant and the concentration, we found that the cryoprotectant ethylene glycol (25% v/v) was a good alternative to glycerol that would not compete with the ligand.</p><p>The structure of human ZG16p in complex with α-O-methylmannose was successfully solved at 2.8-Å resolution by the molecular replacement method using the structure of ligand-free ZG16p (Fig. 5A). A structural comparison of the ZG16p·O-methylmannose complex and ligand-free ZG16p revealed that the two structures were almost identical, with a root mean square deviation value of 0.27 Å for corresponding Cα atoms (22–159, Fig. 5B). As predicted from the glycerol-bound ZG16p structure (16), Gly35 in the GG loop, Gly147, Ser148, and Leu149 in the binding loop and Asp151 on the β12 strand contribute to the mannose binding (Fig. 5, C and D). Eight conserved hydrogen bonds are identified with the α-O-methylmannose that contributes to the intermolecular interaction: Man O3-Gly35 N, Man O4-Gly35 N, Man O4-Asp151 O (side chain), Man O5-Ser148 O (side chain), Man O5-Ser148 N, Man O6-Asp151 O (side chain), Man O6-Leu149 O, and Man O6-Leu149 N. The Man O1, O2, and methyl groups are exposed to the solvent and not directly involved in the ZG16p interaction. In the crystal structures of mannose binding-type Jacalin-related plant lectins in complex with mannose ligands, the same hydrogen bond networks were observed (16). Taking the data together, we conclude that ZG16p and the mannose binding-type Jacalin-related plant lectins have the same mannose-binding mode.</p><!><p>Crystal structures of the ZG16p·methyl-O-mannose complex. Overall structure (A), superimposition with ligand-free ZG16p (B), and close-up view (D) of the structure is shown. C, superposition of glycerol-bound ZG16p and O-methylmannose-bound ZG16p structures. GG and GXXXD motifs of ZG16p are shown as stick models, and bound glycerol and O-methylmannose are shown as ball-and-stick and stick models, respectively. Oxygen and nitrogen atoms are colored with red and blue, respectively. In A and D, the secondary structures are highlighted in yellow (β1, β2, β11, and β12), cyan (β3-β6), green (β7-β10), and orange (α-helix). Bound O-methylmannose is shown as ball and stick model. In A, the Fo − Fc electron density map of the ligand is shown in gray mesh contoured at 1σ level. In B, the main chains of the ligand-free form (gray) and ligand complex (green) are superimposed in a wire model. In D, residues binding to O-methylmannose are shown as stick models in green and potential hydrogen bonds are indicated as dotted lines.</p><!><p>An interesting finding from the glycan microarray analysis is that Man-O-Ser and Man-O-Thr give strong binding signals with ZG16p. The structures of human ZG16p in complex with Man-O-Ser and Man-O-Thr were determined at 2.1- and 2.7-Å resolutions, respectively (Fig. 6, supplemental Fig. S2, a and b). Little structural change was observed for ZG16p upon binding to Man-O-Ser or Man-O-Thr. The mannose binding mode in two complex structures was almost identical to that in the ZG16p·O-methylmannose complex. A striking feature was the presence of water-mediated hydrogen bonds formed between Tyr104 Oη (binding loop), Man O4, and Ser N (ligand) in ZG16p·Man-O-Ser complex (Fig. 6A). The observed hydrogen bonds seem to stabilize the conformation of the Tyr104 side chain and Ser main chain. In the ZG16p·Man-O-Thr complex, the water molecule was also found at the corresponding site and formed hydrogen bonds with Tyr104 Oη and Man O4 (Fig. 6B). Thr N (ligand) does not participate in hydrogen bonding. Instead, the Thr Cγ methyl group is located on the hydrophobic surface of the Man residue, thus stabilizing the Man ring-Thr Cγ hydrophobic interaction. Thus ZG16p can recognize both Man-O-Ser and Man-O-Thr efficiently by tuning the interaction mode.</p><!><p>Crystal structures of ZG16p·Man-O-Ser and ZG16p·Man-O-Thr complexes. Close-up views of the ligand-binding site for ZG16p·Man-O-Ser (A) and ZG16p·Man-O-Thr (B) complexes. The secondary structures are highlighted in yellow (β1, β2, β11, and β12), cyan (β3-β6), green (β7-β10), and orange (α-helix). Bound ligands are shown as ball and stick models. Potential ligand-protein hydrogen bonds are indicated as dotted lines. Oxygen and nitrogen atoms are colored with red and blue, respectively.</p><!><p>O-Mannosylation is conserved from bacteria to humans (45), and the O-mannose structures are particularly abundant in fungi (46) and Mycobacterium tuberculosis (47). Protein O-mannosylation in fungi is thought to be important in the formation and maintenance of a stable cell wall (45). It has been demonstrated that O-mannosylation of surface mannoproteins of fungal pathogens contributes significantly to virulence (48, 49). Invading pathogens with O-mannose structures can potentially become targets of ZG16p in the digestive system. ZG16p has been proposed to contribute to the immune system by capturing pathogenic fungi (4).</p><p>The intestinal epithelium is the major interface with the indigenous microbiota and a primary portal of entry for bacterial pathogens. Several RegIII proteins (mouse RegIIIγ and human hepatointestinal pancreatic/pancreatitis-associated protein HIP/PAP), which are highly expressed in the small intestine, are secreted C-type lectins that kill Gram-positive bacteria by recognizing the peptidoglycan carbohydrate backbone (50, 51). Analogously to the function of RegIII proteins, ZG16p may function as one of the proteins that protect against enteric infections and limit opportunistic invasion by symbiotic bacteria.</p><!><p>Manα1–2/3Man unit and β-glucans containing the Glcβ1–3Glc unit are commonly found in fungal cell walls. Elucidation of the ZG16p-binding mode provides clues to a protective role against pathogens. We have determined the structures of ZG16p in complex with Manα1–3Man and Glcβ1–3Glc at 1.9- and 2.0-Å resolution, respectively (Fig. 7, A and B, supplemental Fig. S2, c and d). Little structural changes were observed for ZG16p upon binding to these ligands. In the ZG16p·Manα1–3Man complex, the non-reducing mannose residue is accommodated in the mannose binding pocket, with a similar binding mode to the other ZG16p complex structures. O1 of the reducing mannose residue makes a hydrogen bond with the Ser148 O, thus stabilizing the reducing mannose outside the binding pocket. On the other hand, in the ZG16p·Glcβ1–3Glc complex, the reducing glucose residue is accommodated in the binding pocket, with the same hydrogen bonding pattern. As O1 of the reducing end glucose is exposed to the solvent, we infer that this could be an internal glucose residue of a trisaccharide or longer glucan oligomer. The O3 atom of non-reducing glucose makes a weak hydrogen bond with Lys36 Nϵ. In addition, O2 atoms of the two glucose residues are within hydrogen bonding distance (3.4 Å). These inter- and intra-molecular hydrogen bonds may stabilize the non-reducing glucose residue. The two binding modes observed for Manα1–3Man and Glcβ1–3Glc to ZG16p are similar to those of Banlec (Fig. 7, C and D). In the crystal structures of Banlec in complex with pentamannose and Glcβ1–3Glc, non-reducing mannose and reducing glucose are embedded in the binding pocket, respectively.</p><!><p>Two different disaccharide-binding modes of ZG16p. Close-up views of ZG16p·Manα1–3Man complex (A), ZG16p·Glcβ1–3Glc complex (B), Banlec·pentamannose complex (17) (PDB code 3MIU) (C), and Banlec·Glcβ1–3Glc complex (23) (PDB code 2BN0) (D). In C, the electron density of only a Manα1–3Man unit was observed and shown. Bound ligands are shown as ball and stick models, and potential ligand-protein hydrogen bonds are indicated as dotted lines. Residues binding to ligand are shown as stick models and potential hydrogen bonds are indicated as dotted lines. Oxygen, nitrogen, and carbon atoms are colored with red, blue, and green, respectively.</p><!><p>Interestingly, in the structures of ZG16p in complex with both of the disaccharides Manα1–3Man and Glcβ1–3Glc, Tyr104 Oη is involved in water-mediated hydrogen bond with Man O4 and Glc O4, respectively. Similar hydrogen bonding patterns are observed in ZG16p/Man-O-Ser and ZG16p/Man-O-Thr structures, indicating the importance of Tyr104 for binding to these ligands. This Tyr residue in human ZG16p is well conserved among the other mammalian ZG16p homologs. Indeed when we performed microarray analysis of the Y104F mutant of ZG16p the binding to α-mannose, Man-O-Ser/Thr, β-glucan oligomer, and high mannose glycans with 6-phosphate was diminished, but not the binding to heparin and CSB (Fig. 1B). This observation confirms the characteristic involvement of Tyr104 in Man/Glc-type ligand binding. Unexpectedly with the Y104F mutant, higher binding intensity was observed to the heparin probe in the array. Tyr104 is solvent exposed and has a potential to form a hydrogen bond with a water molecule as observed in the crystal structures of sugar-ZG16p complexes. When Tyr104 is mutated to Phe, such hydrogen bonds cannot be formed and a local structural change could occur that might affect the heparin binding. It is interesting that in a separate study we observed that β-glucan polysaccharide curdlan was not bound by ZG16p.6 Further studies are required to understand the biological significance of glucose binding by ZG16p.</p><!><p>Previous mutagenesis studies on rat ZG16p suggest that Lys33, Lys36, Arg37, Arg55, and Arg58 are involved in heparin binding (12). Among the residues, Lys36, Arg37, and Arg55 are conserved in human ZG16p (Fig. 2) and the location of the three residues is shown in Fig. 8. The three residues are spatially close to the mannose-binding pocket; therefore we investigated whether the two ligands, mannose and heparin, compete with each other. For this purpose, we prepared the D151N mutant, whose mutation is reported to abolish mannose binding (4). We first performed the glycan microarray on the D151N mutant and wild-type ZG16p (Fig. 1C). Consistent with the previous report, the D151N mutant showed diminished binding to the mannose-related probes. In contrast, binding to heparin was rather enhanced, indicating the mannose binding pocket is not involved in heparin binding and the negative charge of Asp151 might be unfavorable for the interaction with negatively charged heparin.</p><!><p>Residues of human ZG16p important for heparin binding. Crystal structure of human ZG16p with O-methylmannose highlighting the mannose-binding site (Asp151 in red and bound O-methylmannose in ball and stick model) and putative heparin-binding site (Lys36, Arg37, and Arg55 in blue).</p><!><p>To determine the binding site of heparin on ZG16p, we conducted solution NMR analysis using 15N-labeled ZG16p. Upon addition of heparin tetrasaccharide, a set of NMR signals were perturbed (Fig. 9A). The residues showing significant chemical shift changes (Δδ > 0.04) were mapped on the ZG16p structure (Fig. 9, C and D). The perturbed residues are located on one side of the ZG16p surface, indicating that ZG16p binds to heparin using the area.</p><!><p>ZG16p binds heparin using its large surface area. A, 1H-15N HSQC spectra of 15N-labeled ZG16p in the absence and presence of heparin tetrasaccharide. B, titration curves for the selected peaks in 1H-15N HSQC spectra of 15N-labeled ZG16p upon addition of heparin tetrasaccharide. C, chemical shift changes of amide signals upon binding to heparin tetrasaccharide. D, mapping of perturbed residues on the ZG16p structure. The residues showing chemical shift change are colored in blue (Δδ > 0.08 ppm) and sky blue (0.04 < Δδ < 0.08 ppm).</p><!><p>We further conducted on-array inhibition assays. ZG16p binding to α-mannose-related and glucan-derived oligosaccharide probes was suppressed in the presence of α-O-methyl mannose but not β-O-methyl galactose (Fig. 10A). Importantly, ZG16p binding to heparin or CSB was not inhibited by the presence of Man or Gal. Long chain (>20-mer) oligosaccharides of heparin, but not hyaluronic acid, were found to have an inhibitory effect for ZG16p binding to α-mannose-related probes (Fig. 10B). Little or no inhibition was found to the Man-O-Thr/Ser probes or to immobilized heparin 16-mer when soluble heparin oligosaccharides were used as inhibitor. It is difficult to design assay conditions to have a mutual effect of strongly and very weakly bound probes on the array. Nonetheless, these data indicate that mannose binding does not compete with the heparin binding site on ZG16p, whereas heparin binding may affect to some extent the binding to mannose-related ligands, in particular to those that have relatively low affinity. The independent binding sites on ZG16p but with some cross-effects after binding may be important for the association of monomeric ZG16p to various partners under physiological conditions.</p><!><p>Binding of GST-fused human ZG16p to selected ligands in microarrays in on-array inhibition assays. A, Man-α-OMe and Gal-β-OMe were used as inhibitors. B, oligosaccharide fractions (>20-mer) of heparin (Hep) and hyaluronic acid (HA) were used as inhibitors. Numerical scores of the binding signals are means of duplicate spots at 5 fmol/spot with error bars.</p><!><p>In conclusion, ZG16p possesses a unique multiple-ligand binding property toward short chain α-manno-oligosaccharides, and sulfated GAG oligosaccharides. The differing carbohydrate-binding modes are attained by a shallow mannose binding pocket and a basic amino acid patch on the protein surface. Furthermore, two disaccharide-binding modes were observed, enabling the non-reducing or reducing sugar to be accommodated in the binding pocket. Such differing ligand-binding modes may confer multiple functions to ZG16p at different locations: glycosaminoglycan binding in zymogen granules and α-mannose binding at mucosal surfaces. Further analyses of the glycosaminoglycan-binding mode are under way to better understand the functions of ZG16p.</p><!><p>This work was supported by Grant-in-aid for Scientific Research (C) 25460054 (to Y. Y.) and by United Kingdom Research Council Basic Technology Initiative Glycoarrays Grant GRS/79268 and Translational Grant EP/G037604/1 and Wellcome Trust Grants WT093378MA and WT099197MA (to T. F.).</p><p>This article contains supplemental Microarray Data, Figs. S1 and S2, and Table S1.</p><p>The atomic coordinates and structure factors (codes 3VZF, 3VY7, 3VZG, 3VZE, and 3VY6) have been deposited in the Protein Data Bank (http://wwpdb.org/).</p><p>S. Hanashima, S. Götze, Y. Liu, A. Ikeda, K. Kojima-Aikawa, N. Taniguchi, D. Varón Silva, T. Feizi, P. H. Seeberger, and Y. Yamaguchi, unpublished data.</p><p>zymogen granule protein 16</p><p>zymogen granule membrane</p><p>N-(9-fluorenyl)methoxycarbonyl</p><p>glycosaminoglycans</p><p>chondroitin sulfate B</p><p>maltose-binding protein</p><p>Protein Data Bank.</p>

PubMed Open Access

The competing effects of microbially derived polymeric and low molecular-weight substances on the dispersibility of CeO2 nanoparticles

To understand the competing effects of the components in extracellular substances (ES), polymeric substances (PS) and low-molecular-weight small substances (SS) <1 kDa derived from microorganisms, on the colloidal stability of cerium dioxide nanoparticles (CeNPs), we investigated their adsorption to sparingly soluble CeNPs at room temperature at pH 6.0. The ES was extracted from the fungus S. cerevisiae. The polypeptides and phosphates in all components preferentially adsorbed onto the CeNPs. The zeta potentials of ES + CeNPs, PS + CeNPs, and SS + CeNPs overlapped on the plot of PS itself, indicating the surface charge of the polymeric substances controls the zeta potentials. The sizes of the CeNP aggregates, 100-1300 nm, were constrained by the zeta potentials. The steric barrier derived from the polymers, even in SS, enhanced the CeNP dispersibility at pH 1.5-10. Consequently, the PS and SS had similar effects on modifying the CeNP surfaces. The adsorption of ES, which contains PS + SS, can suppress the aggregation of CeNPs over a wider pH range than that for PS only. The present study addresses the non-negligible effects of small-sized molecules derived from microbial activity on the migration of CeNP in aquatic environments, especially where bacterial consortia prevail.The cerium dioxide (CeO 2 , F m m3 ) nanoparticle (CeNP) is a nanomaterial that is finding a wide variety of applications to a vast number of products involving fuel additives 1 , fuel cell components 2 , biomedical applications 3,4 , combustion accelerators and abrasives 5,6 , and specialized polishing agents 7 . With all of these applications, it is inevitable that CeNPs will be found in the environment. Unfortunately, in vitro and in vivo experiments with CeNPs have shown that this material can cause chronic toxicity to aquatic organisms 8 , cell death to E. coli 9 , increase of reactive oxygen species levels relevant to human lung cells 10 , and decrease of glutathione levels in cultured human lung epithelial cells 11 . Due to their small size, ~10 nm, the inhaled CeNPs can penetrate into the deep respiratory system 12 and potentially cause adverse health effects despite the existing study reported that CeNPs have low human toxicity 13 . Thus, the distribution and migration behavior of CeNP, as well as other engineered nanoparticles in the ambient environment, is a central issue that requires careful monitoring and modeling 14 . The mobility of CeNP follows the general rules for colloid transport in surface and subsurface environments [15][16][17][18][19] . Colloid transport can be controlled by several processes: sedimentation, filtering effects, hydrodynamic chromatographic effects, and capillary effects. All of these processes are largely dependent on the aggregation processes of CeNPs in natural aquifers 17,20 .The aggregation of colloids is mainly constrained by several factors, including solution pH, electrolyte concentrations [21][22][23] , adsorbed ions 23,24 , and adsorbed organic matter 23,25 . On the other hand, natural and engineered

the_competing_effects_of_microbially_derived_polymeric_and_low_molecular-weight_substances_on_the_di

<!>Materials and Methods<!>Preparation of the extracellular substances (ES), the extracellular polymeric substances (PS) and the extracellular small substances (SS).<!>Results and Discussion<!>Morphology of polymeric substances within the ES. The topological and phase contrast AFM images<!>The effects of ES, PS, and SS on the surface electric potential (ζ potential).<!>Conclusion

<p>nanoparticles, including CeNPs, can encounter microbial consortia in the subsurface environment 26 due to the ubiquitous occurrence of microorganisms 27,28 . During the interaction between microorganisms and nanoparticles, the extracellular substances (ES) that are released by the microorganisms 29,30 , as essential constituents to form biofilms 31 , adsorb onto the nanoparticle's surface and occasionally lead to particle dissolution 32 , promoting electron transfer 33 , and changing the dispersibility of the nanoparticles in solution 29 . The polymer substances (PS) included in the ES category are generally composed of 40-95% polysaccharides, <1-60% protein, <1-10% nucleic acids, and <1-40% lipids 34,35 . Adsorption of the PS onto the nanoparticles changes the zeta (ζ) potential of aggregates and promotes the dispersibility of particles, increasing the critical aggregation concentration 29,[36][37][38][39] . Adeleye et al. 36 reported that extracellular PS adsorbed onto Cu and CuO nanoparticles can change the ζ potential from positive to negative at pH 4 and narrowed the ζ potential range. Miao et al. 39 also reported the enhanced stability of CuO nanoparticles after adsorption of extracellular PS and polysaccharides due to the electrostatic repulsion and formation of a steric barrier. Adeleye and Keller 37 carried out adsorption experiments of extracellular PS onto TiO 2 nanoparticles. This resulted in a reversal of the surface charge and enhanced particle stabilization. Lin et al. 38 performed adsorption experiments for extracellular PS onto TiO 2 nanoparticles and found that both electric repulsion and steric hindrance were mechanisms of stabilization with increasing mass of the adsorbed extracellular PS. Our previous study 29 revealed enhanced stabilization of CeNPs through steric hindrance and the critical aggregation concentration of NaCl increased from 10 mM to 250 mM when ES was adsorbed onto CeNPs. Despite the fact that effects of extracellular PS have been explored, the previous studies have focused on the polymers only, excluding the effect of small molecules in the ES. Thus, there is limited knowledge on the total competing effects of ES components including small substances. The aim of the present study is to understand the properties of the PS in microbially derived ES, PS, and SS. Secondarily, we aim to evaluate their competing effects on the adsorption processes onto CeNPs, changes in the CeNP surface properties, and the aggregation and sedimentation of CeNPs at various pHs.</p><!><p>CeO 2 nanoparticles (CeNPs). Synthetic CeNPs were commercial products purchased from Strem Chemicals, Inc., Newburyport, MA, USA, (part# 58-1400, ~7 nm). The CeNPs had spherical shape and the average diameter was ~7 nm. The surface area was determined to be 70.2 m 2 g −1 using a BET single point analysis, which was smaller than that calculated assuming fully dispersed spherical nanoparticles having <10 nm in size. This indicates that the CeNPs already aggregated prior to use in the present experiments. For a detailed description of these CeNPs, see references 29,30 .</p><!><p>In the present study, Saccharomyces cerevisiae (X-2180) was used as a representative microorganism. First, S. cerevisiae was harvested in 200 mL of sterilized YPD medium, which was composed of 10 g L −1 yeast extract, 20 g L −1 peptone, and 20 g L −1 dextrose. The yeast was incubated for 20 h on a rotary shaker at 120 revolutions per minute (rpm) at 25 ± 1 °C. The suspension of the yeast cells were centrifuged for 10 min at 3000 rpm to be separated. The separated cells were washed three times with 1 mmol L −1 NaCl solution. The yeast cells were put in a polypropylene tube filled with 50 mL of 1 mmol L −1 NaCl. In all solutions the cell density was adjusted to 2.0 ± 0.1 dry g L −1 . The pH of the solutions was initially adjusted to 3.0 ± 0.1 with 1.0 mol L −1 HNO 3 solution. In our previous study 29 , a high concentration of organic matter was extracted at this pH and the composition was similar to the organic matter extracted at higher pHs. A pH meter (TOA tpx-999i; PCE108CW-SR) equipped with an electrode was used to measure pH.</p><p>After extracting ES for 72 h, the suspension was filtered through a polytetrafluoroethylene (PTFE) membrane filter (Advantec) with 0.20 μm pore size to remove the yeast cells. The filtrate was named as the ES solution. This ES solution contained both polymers and low-molecular-weight species. A portion of ES was dialyzed for 72 hours at 4 °C using a 1000 MWCO Spectra/Por ® 7 (Spectrum) cellulose dialytic membrane. The volume ratio of the ES to ultrapure water was set to 1:3. The water outside the dialytic membrane was exchanged with ultrapure water every 24 hours. This outside solution after the first 24 hours of dialysis was labeled the "extracellular small substances (SS)" solution. The conductivity after 72 hours was measured to be ~0.0 μS. The solution that remained in the membrane tube was labeled PS. The ES, PS, and SS solutions were preserved at 4 °C in a refrigerator and the solutions were adjusted back to room temperature prior to use in experiments.</p><p>The morphology of the ES was observed by scanning probe microscopy (SPM, DimensionIcon, Bruker AXS, Billerica, USA). The observations were performed under ambient atmospheres using a ScanAsyst probe (ScanAsyst-Air). The specimen for SPM was prepared by dropping the ES solution onto the cleaved pristine surface of biotite and air-dried. Then, the specimen was rinsed with ultrapure water three times and air-dried again.</p><p>The ES contained various kinds of polymers, organic matter, and inorganic ions, such as H 3 PO 4 . The total phosphate concentrations in the ES, PS, and SS were determined using inductively coupled plasma atomic emission spectrometry (ICP-AES; Agilent 7500c). The detection limit of P was 15 ppb. The concentrations of dissolved organic carbon (DOC) were determined by using a total organic carbon analyzer (TOC; TOC-VE, Shimadzu). The detection limit was 50 μg L −1 and the error was <2%. To further characterize the ES, the dried ES, PS, and SS were analyzed using an attenuated total reflectance Fourier transform infrared spectrometer (ATR-FTIR; Jasco, FT/IR-620) equipped with a deuterated L-alanine triglycine sulfate (DLATGS) detector, a single bounce attenuated total reflectance attachment, and a ZnSe crystal. Thirty-two spectra were obtained with a spectral resolution of 4 cm −1 and averaged. To prepare the dried samples, the pH of the ES solutions was first adjusted to 6.0 ± 0.1 with 1.0 mol L −1 NaOH solution. The ES, PS, and SS solutions were lyophilized and preserved at −10 °C until the measurement. In addition to the FTIR analysis, elemental analysis was completed on the lyophilized ES, PS, and SS to determine the concentrations of C, N, and H.</p><p>Adsorption of ES, PS, and SS onto the CeNPs. The 5000 mg L −1 CeNPs stock suspension was prepared and ultra-sonicated for 10 min. Five different solutions were prepared in the present experiment: (i) 1 mM NaCl (control); (ii) 1 mM NaCl + 160 μM H 3 PO 4 (160 μM P), of which the P concentration was adjusted to that of the ES solution in the previous study 29 ; (iii) ES solution containing1 mM NaCl solution (conditions during the extraction procedure); (iv) PS + 1 mM NaCl, to adjust the ionic strength to be similar to the other solutions; and (v) SS + 1 mM NaCl. The pH of these suspensions was adjusted to 6.0 ± 0.1 with NaOH. Each of these five solutions were mixed with an aliquot of CeNPs stock solution, in which the concentration of CeNPs was set to 100 mg L −1 so that multiple analytical techniques could be employed. In this study, we did not adjust the C content prior to the adsorption experiments, because the C content does not reflect the actual concentration of specific organic molecules. All ES, PS, and SS contain organic matter with various molecular weights. Thus, it is difficult to quantify the actual concentrations of the non-specified molecules. Rather, the CeNP surfaces were saturated with the organic matters that have concentrations as prepared in the experiments.</p><p>High-Angle Annular Dark-Field Scanning TEM (HAADF-STEM) and energy-dispersive x-ray spectroscopy (EDX) were completed using a scanning transmission electron microscope (STEM, JEOL, JEM-ARM200CF and JEM-ARM200F, Akishima, Japan). The TEM specimens were prepared by desalinating three times with ultrapure water and dropping the suspension sample on a 300 mesh Cu with Ge or holey carbon supporting membrane followed by air-drying. For the ATR-FTIR analyses, ES, PS, or SS were adsorbed onto the CeNPs at pH 6.0. These suspensions were statically reacted for 24 h. The duration of 24 hours is enough to achieve the apparent equilibrium in this experiment according to our previous study 29 . After the adsorption, the CeNPs associated with organic matter were separated using a 0.025 μm nitrocellulose membrane filter and lyophilized. The analytical procedure of ATR-FTIR for ES, PS, or SS adsorbed to CeNPs was the same as the one described in the previous section.</p><p>A Zeta Sizer Nano ZEN (Malvern Instruments Inc) was used to measure the ζ potential and average hydrodynamic diameter for the CeNPs suspensions in 1 mM NaCl solution with a capillary cell. The starting pH was set to 6.0 and the pH was shifted to the targeted value using NaOH or HNO 3 .</p><!><p>Characterization of ES, PS, and SS. The composition of the ES extracted in 1.0 mmol L −1 NaCl after 72 hours of incubation is summarized in Table 1. The ES contains ~172 mg L −1 organic carbon, ~0.44 mmol L −1 K + , and ~0.59 mmol L −1 total P. The compositions of PS and SS are also given in Table 1. The ES used in the present study contained 2-4 times higher concentrations of organic and inorganic species than the ES characterized in our previous study 29 . The amounts of dissolved organic species in the solutions before and after the adsorption experiments are given in Table S1.</p><p>Figure 1 shows ATR-FTIR spectra (900 to 1800 cm −1 ) of the lyophilized samples before the adsorption experiments: ES (a), SS (b), and PS (c). The peak assignments, with references, are summarized in Table 2. In line (a), the absorption band at ~1397 cm −1 is assigned to symmetric stretching of COO − groups (ν s C-O) that are included in proteins and polypeptides, and carboxylated polysaccharides [40][41][42] . The band at ~1604 cm −1 is assigned to the stretching vibration of C=O groups derived from amide I bonds, which represent amides associated with proteins and polypeptides. The band at ~1518 cm −1 is assigned to the stretching vibration of C-N groups and deformation vibration of N-H groups included in amide II bonds, which correspond to -CO-NH-of proteins and polypeptides [41][42][43] . Thus, the two bands at 1604 and 1513 cm −1 suggest that the ES contains proteins and polypeptides. The band at ~1119 cm −1 can be assigned to ring vibrations of C-O-C bonds included in polysaccharides and the stretching vibrations of P=O bonds in proton-dissociated orthophosphate. The band at ~1343 cm −1 is assigned to the carbon backbone coupled with C-O and P-O stretching 44 . Further, the band at ~1050 cm −1 is assigned to symmetric stretching vibrations of P=O derived from phosphoryl groups 24,40,45 . It is difficult to separate the phosphate bands and the polysaccharide vibration bands due to their overlaps; however, the ES released from S. cerevisiae typically contains both polysaccharides and phosphoryl species, which are also included in the ES released from Bacillus subtilis 40,46 and Pseudomonas aeruginosa 45,47 .</p><p>The FTIR spectrum of SS (b), molecules smaller than 1 kDa, is similar to that of the ES before dialysis. Although the peak position of the PS spectrum is similar to that of the ES spectrum, there is a slight difference in the relative intensity between the peaks. The relative intensity of the peaks derived from phosphate (1044 cm −1 ) and carboxyl (1399 cm −1 ) groups was weaker than that of proteins (1521 cm −1 and 1635 cm −1 ) in the PS fraction. These results indicate that SS contains almost the same compounds as ES, such as inorganic phosphate, amino acids, polysaccharides, and polypeptide, whereas PS contains mainly polysaccharides, proteins and polypeptides of larger molecular sizes >1 kDa.</p><!><p>were obtained for the freshly cleaved biotite (Fig. 2a), the freshly cleaved biotite without the ES after desalination (Fig. 2b), and the ES adsorbed onto the freshly cleaved biotite after desalination (Fig. 2c), which showed the presence of nanoparticles 20-30 nm in diameter and ~3 nm in height (Fig. 2c). The same mode images without ES after desalination did not show any nanoparticles on the surface of cleaved biotite (Fig. 2b). Thus, the particles detected in the former images (Fig. 2c) are polymeric substances of the ES. The shallow height of the nanoparticles indicates a flattened shape after substrate adhesion; thus, the true particle size in solution is likely smaller than 20 nm.</p><p>STEM of the ES + CeNPs, PS + CeNPs, and SS + CeNPs. Figure 3 shows that HAADF-STEM image and the EDX elemental maps of the samples. ES + CeNPs, PS + CeNPs, and SS + CeNPs exhibit CeNP aggregation ranging from 100 to 500 nm, on which C, N, and P are distributed uniformly, indicating that the polymeric substances of the ES adsorbed onto the CeNP surface. Note that the C map contains interference from the holey carbon supporting mesh. In the PS + CeNPs specimen, the EDX spectrum reveals a clear peak of the S K-line in an aggregate, derived from thiol groups, which is likely attributed to the presence of amino acids such as cysteine and methionine. The P/Ce molar ratio on the aggregates in ES + CeNPs, PS + CeNPs, and SS + CeNPs varies between 0.01 to 0.08, indicating that the adsorption of ES, PS, and SS to CeNP is not homogeneous (Fig. 4).</p><p>Representative EDX spectra for ES + CeNPs, PS + CeNPs, and SS + CeNPs can be found in Fig. S1.</p><p>FTIR of ES, PS, and SS adsorbed to CeNPs. ATR-FTIR difference spectra of the experimental samples are shown in Fig. 1: ES + CeNPs (d), SS + CeNPs (e), PS + CeNPs (f), PS + 160 μM P + CeNPs (g), and 160 μM P + CeNPs (h). The CeNPs spectrum was subtracted from the raw spectra to display only the spectra of the adsorbed species. All of the spectra after the adsorption treatment appear similar. After adsorption, the bands for phosphate were broadened, indicating the formation of inner-sphere complexes on the CeNPs, as previously 44 .</p><p>reported 29 , and proteins and polypeptides adsorbed preferentially onto the CeNP surfaces. Interactions between the extracellular PS and metal oxides generally occur via amide, hydroxyl, and carboxylic groups on the PS amino acids in addition to the phosphate groups from phospholipids or nucleic acids 37 . The amide I peak, derived from proteins and polypeptides, shifted toward higher frequencies and the peak derived from the carboxyl group became minimized after adsorption to the CeNPs, regardless of dialysis treatment. The predominant adsorption of proteins from the bacterial extracellular PS to metal oxides has also been reported in several previous studies 40,48 by forming inner-sphere complexes 45 . Proteins can adsorb onto hydrophilic surfaces and the shift in the peak position typically occurs due to protein conformational changes after adsorption [49][50][51][52] .</p><p>Although the chemical compositions between ES, PS, and SS were different, the FTIR spectra after adsorption to CeNPs appear identical, strongly suggesting that the molecules adsorbing onto the CeNP surfaces possess similar functional groups, despite the molecular size differences in the ES, PS, and SS constituents.</p><!><p>Figure 5a shows the pH dependence of ζ potential for: CeNPs (control), ES + CeNPs, 160 μM P + CeNPs, and PS only. The point of zero charge (PZC) of CeNPs (control) was determined to be 6.9, which was within the range of the reported PZC values from several literature sources; 6.5-8.0 21,53 . The ζ potentials of the CeNPs in 160 μM P were −40 to −50 mV at pH > 5.0, and the isoelectric point (iep) was determined to be ~1.6. In the pH 6 solution, the ζ potential decreased as H 2 PO 4 − and HPO 4 2− adsorbed to the CeNPs. In the ES + CeNPs, the ζ potentials were plotted between the control and 160 μM P values. The P concentration (592 μM) in the present ES was determined to be higher than that measured in our previous study 29 , but within the same order of magnitude. As reported in 29 , phosphate in the ES also adsorbed to CeNPs, although the ζ potential was not affected by the adsorbed phosphate. Furthermore, the ζ potentials of PS were plotted almost identical to the plots of ES + CeNPs. The ζ potentials of ES (Fig. S2) were plotted deviated from PS and ES + CeNPs, indicating that the ES compounds adsorbed onto the CeNP is similar to PS rather than the total ES compounds. It was impossible to measure the ζ potential of SS due to their small sizes. The electrophoretic mobility distribution for the ES + CeNPs was similar to that for PS at pH 2.9-10.0 (Fig. S3). In the diagram for PS at pH 9.99 (Fig. S3), the single peak split into multiple peaks at high pH, most likely because there were several aggregates with different functional groups on the CeNP surface. In the case of ES + CeNPs, the peaks were located at the same mobility value as the case of PS, which may indicate that the large molecules with similar specific functional groups preferentially adsorbed onto the CeNP surfaces. The effect of phosphate adsorption did not appear in the ζ potential in the ES solution because the preferential adsorption of macromolecules, such as proteins and polypeptides, on the outermost surface hinders the effects of orthophosphate.</p><p>The ζ potentials of three additional systems (PS + CeNPs, PS + 160 μM P + CeNPs, and SS + CeNPs) were also plotted in Fig. 6a, confirming that the presence of inorganic phosphate in the ES did not influence the ζ potential of CeNP in any system where PS was present; that is, the ζ potential of CeNPs reacted with ES was governed by the polymers in the ES rather than the small charged molecules, such as phosphate. In addition, the ζ potential of SS + CeNPs, which contained inorganic phosphate, also exhibit the same trend as that of ES + CeNPs. There are two factors that caused the similarity. One factor is the steric barrier created by the organic matter, even by molecules size smaller than 1 kDa, because the FTIR spectra of ES, PS, and SS revealed that the functional groups of the compounds adsorbed on the CeNPs were nearly identical (Fig. 1). The other factor is the decreased concentration of SS due to dialysis. The SS solution was diluted to approximately one quarter, implying the possibility of decreased amounts of adsorbed inorganic phosphate in the ES. Indeed, the adsorption experiment of inorganic phosphate using various concentrations of phosphate revealed that the ζ potential increases gradually as the P concentration decreases, and the pH dependence at the P concentration of 1.6 µM, which is two orders of magnitude less than the P concentration in ES, became identical to that of the control (Fig. 7a). In addition, the average hydrodynamic diameter also changed concurrently; when P concentration decreased from 160 μM to 16 μM, the ζ potential shifted to a positive value, the iep shifted from ~1.6 to ~4.3, and the pH, at which the average hydrodynamic diameter becomes the maximum, shifted from ~1.9 to ~5.0. The pH dependence of the average hydrodynamic dimeter of the SS + CeNPs appeared similar to that of P + CeNPs (16 µM) (Fig. 7b). When the CeNPs were exposed to 107 µM P, the same P concentration in SS, the iep shifted from ~1.6 to ~2.0, though the pH dependence of the ζ potential was almost identical to the 160 µM P case. Thus, the pH dependence of the ζ potentials for SS + CeNPs and ES + CeNPs can be ascribed to the similar characteristics of the polymeric substances, even when of different molecular sizes.</p><p>The effects of ES, PS, and SS on the size of the aggregates. Figure 5b shows the average hydrodynamic diameter of CeNPs in three conditions: (i) 1.0 mM NaCl; (ii) 1.0 mM NaCl + 160 μM H 3 PO 4 ; (iii) ES + 1 mM NaCl, over pH 1-11. The size increases at near the iep under all conditions. This indicates that the electrostatic repulsive force resulted from the outermost charge of the particle constrains the aggregation behavior of CeNPs.</p><p>When the pH was less than 3, the ζ potentials of the ES + CeNPs and the 160 μM P + CeNPs cases were both close to zero, meaning that the electrostatic repulsive force was not effective under low pH conditions (Fig. 5a). However, the size of ES + CeNPs was less than half of that of 160 μM P + CeNPs, indicating that the steric barrier formed by the polymeric substances effectively suppressed the aggregation of CeNPs. Indeed, as described above, the FTIR results indicated a preferential adsorption of proteins onto the CeNP surfaces, causing steric hindrance 39 . On the other hand, between pH 3 and 4, the average particle size of ES + CeNPs was larger than that of the 160 μM P + CeNPs and control cases. Under this pH condition, the ζ potential of ES + CeNPs was close to zero, while that of 160 μM P + CeNPs was as low as −30 mV. This indicates that the aggregation behavior of CeNPs was constrained by both electrostatic and steric repulsion, and the effects of electrostatic repulsion were greater than that of the steric barrier.</p><p>Figure 6b shows the pH dependence of the average hydrodynamic diameter of ES+ CeNPs, SS + CeNPs, PS + CeNPs, and PS + 160 μM P + CeNPs. The average hydrodynamic diameter for SS + CeNPs exhibited a different trend from that of 160 μM P + CeNPs or ES + CeNPs, most likely because the electrostatic repulsive forces in SS + CeNPs were less than that of the 160 μM P + CeNPs case, due to the lower P concentration in SS, and the steric barrier derived from the molecules <1 kDa in SS was weaker than that in the ES + CeNPs case. In the PS + CeNPs case, the average hydrodynamic diameter increased around the iep of PS itself, whereas that in the PS + 160 μM P + CeNPs case showed only a slight increase around pH ~3. The difference might be attributed to the presence of P adsorbed onto CeNPs rather than PS adsorption on to the CeNPs. CeNP aggregation should Figure 3. HAADF-STEM image with the elemental maps of the CeNP specimen after contact with ES, SS, or PS at pH 6.0 in 1.0 mM NaCl for 24 hours followed by desalination washing with ultrapure water thrice. The suspension of ES + CeNPs was dispersed on the Ge-mesh without using C, while the samples of PS + CeNPs and SS + CeNPs were prepared on holey carbon mesh with a Cu supporting grid.</p><p>have been promoted at between pH 3 to 5, near the iep of PS (~3.5), because the ζ potential is determined by the largest molecules, regardless of the presence of phosphate. However, aggregation appeared to be suppressed by the inorganic phosphate adsorbed onto CeNPs, likely because the repulsive forces derived from the adsorbed inorganic phosphate became predominant when the distance between CeNP surfaces become sufficiently small. At pH <3, which is close to iep (pH ~1.6) of 160 μM P + CeNPs, both electrostatic repulsive forces and steric barriers from the adsorbed PS reduced aggregation. Although the same trend can be seen in ES + CeNPs, aggregation in PS + 160 μM P + CeNPs was suppressed more profoundly than that in ES + CeNPs. The ES contains free cations and small organic molecules, including amino acids with carboxyl groups. Thus, in the ES + CeNPs system, the electric repulsive forces of inorganic phosphate can be neutralized 54 or other organic matter having hydrophobic adsorption to the CeNP surfaces may lead to a higher affinity between the particles. Indeed, Mosley and Hunter 55 reported that adsorption of organic matter suppresses aggregation through the formation of steric barriers, whereas it can also prevent dissociation of colloid aggregate once they formed aggregates. Because of such effects owing to the presence of other small organic matters in ES, the ES + CeNPs case was found to be slightly less aggregated than that of the PS + 160 μM P + CeNPs case. Figure S4 shows the results of additional experiments measuring the sedimentation rates of CeNPs at pHs of 2.0, 3.0, 3.5, 6.0, 7.0, and 10.0 under three conditions: CeNPs, 160 μM P + CeNPs, and ES + CeNPs. These rates were calculated based on the turbidity time course monitored using UV-Vis spectroscopy. The sedimentation rate of CsNPs was the fastest when the solution pH is close to the iep: 160 μM P + CeNPs at pH 2, ES + CeNPs at pH 3.5, and CeNP at pH 7.0. The sedimentation rate of CeNPs greatly depends on the solution pH, which is consistent with the DLS analysis.</p><p>Comparison with the other macromolecules. Figure 8 summarizes the surface properties and sizes of CeNPs aggregates under the present experimental conditions. Because the ES is a part of natural organic matter (NOM), and vice versa, it is useful to compare the present results with the well-known effects of NOM such as humic substances (HS) on the aggregation behavior of various engineered nanoparticles 19,23,25,[56][57][58][59][60][61] . The presence of NOM typically stabilizes the colloids and reduces aggregation by coating the nanoparticle surfaces: forming a steric barrier and modifying the surface charges 62 . Our results on the role of microbial ES appear to be similar to that of NOMs previously reported 63 . The similarity between the effects of exudate and NOM on the aggregation behavior of TiO 2 nanoparticles has already been recognized 64 . However, comparing the elemental analysis and FTIR data of ES (Table 1) with those of fulvic acid (FA) and humic acid (HA) 65 , the ES contains higher concentrations of nitrogen, reflecting a higher protein and polypeptides content, whereas HS consist of carbohydrates, alginate, amino acids, lignin, and pectin. The clear difference between ES and NOM is the preferential adsorption of protein and polypeptides to the CeNP surfaces.</p><p>In general, the molecular weights of the differently sized components in HS, FA and HA, range from a few hundred to thousands and from thousands to several millions of Da, respectively 65 . Several studies have reported that large molecular size of HA can reduce aggregation more efficiently than FA 66,67 by forming thick polymer layers. In the case of NOM, it has been reported that adsorbed layer thickness, aromaticity, and molecular weight, are all correlated with aggregation behavior 26,63 . It is also noted that typical surfactants with low-molecular weights can desorb or be displaced by larger molecules, such as NOMs 63 ; however, such phenomena were not observed in the present experiments. This can be ascribed to the fact that the functional groups of the organic substances that adsorbed onto the CeNP surfaces were similar in SS and PS. This is reasonable because proteins and polypeptides, which preferentially adsorbed to the CeNP surfaces in the present experiment, generally adsorb by forming chemical bonds 37,45 . Furthermore, in case of NOMs, large fibrillary polymers can form bridges between nanoparticles, and this occasionally promotes aggregation 68 . Such an enhanced aggregation mechanism did not occur in our PS case because the conformation of the proteins and polypeptides present, as revealed in the AFM images (Fig. 2), were approximately spherical. Rather, the PS further reduced aggregation compared to the SS case (Fig. 8).</p><!><p>Adsorption experiments using ES (PS + SS), PS (>1 kDa), and SS (<1 kDa) to CeNPs were conducted to understand the effect of each fractionated component on the modification of the surface and dispersibility of the CeNPs at pHs ranging from 1.5 to 10. Microscopic and spectroscopic characterization of these components, with and without CeNPs, revealed that the three fractions were composed of organic matter that contained similar functional groups, despite the size difference of the molecules within each fraction because polypeptides and amino acids were present in the SS fraction. The preferential adsorption of proteins and/or polypeptides, and inorganic phosphates were observed in all fractions. The polymeric substances in the ES case formed aggregates as large as a few tens of nm and the chemical composition showed heterogeneous distributions on the CeNP surfaces from the adsorption of multiple components. The ζ potential of CeNP with ES, PS, and SS exhibited the same pH dependence, suggesting that the polymeric substances, even smaller-sized molecules, modified the surface charge of the CeNPs with apparent similarity to the PS case. Although the ζ potentials were governed by the polymeric substances in the ES, other small polymers with different ieps can also adsorb to the CeNP surfaces and reduce nanoparticle aggregation under conditions where the ζ potential is nearly zero. Thus, the suppression effects on the aggregation by ES adsorption onto CeNPs can be expressed under wider pH conditions than those derived from PS adsorption only.</p><p>There is a wide range of amounts and chemical variations in small and large polymeric substances, as well as geochemical parameters, in natural surface and subsurface environments. Hence, the results of the present study highlight the non-negligible impact of microbially derived polymeric substances, of various molecular sizes, on the migration of CeNP in the environment. The dynamics of adsorption, aggregation, and transformation of various organic matter-nanoparticle couples in realistic environments still remains to be explored as previously pointed out 69 . The quantitative data obtained in the present study can be useful for understanding the role of small-sized molecules derived from microbial activity on the migration of CeNPs in aquatic environments, especially where bacterial consortia prevail.</p>

Scientific Reports - Nature

Asymmetric olefin isomerization of butenolides via proton transfer catalysis by an organic molecule

An unprecedented enantioselective and general olefin isomerization was realized via biomimetic proton transfer catalysis with a new chiral organic catalyst. A broad range of mono- and di-substituted \xce\xb2,\xce\xb3-unsaturated butenolides were transformed into the corresponding chiral \xce\xb1,\xce\xb2-unsaturated butenolides in high enantioselectivity and yield in the presence of as low as 0.5 mol% catalyst. Mechanistic studies have revealed the protonation as the rate-determining step.

asymmetric_olefin_isomerization_of_butenolides_via_proton_transfer_catalysis_by_an_organic_molecule

<p>As exemplified by the Δ5-3-ketosteroid isomerase (KSI)-catalyzed conversion of β,γ- to α,β-unsaturated steroidal ketones (Scheme 1), enzyme-mediated olefin isomerization via a proton transfer from one carbon atom to another in the same substrate molecule constitutes a common and important class of chemical reactions in biology.1 In contrast, only metal-mediated hydride transfer catalysis has been employed in small molecule-catalyzed asymmetric olefin isomerizations, which include enantioselective olefin isomerizations of allylic amines by a chiral Rh+/BINAP complex2 and isomerization of allylic alcohols with a Rh+/planar-chiral phosphaferrocene complex.3 Although substrate-directed diastereoselective olefin isomerizations with either achiral acids or bases have been applied in natural product synthesis,4 only a single example of olefin isomerization by enantioselective proton transfer catalysis, mediated by a bimetallic gadolinium complex, was reported in the literature.5 Herein, we wish to report the realization of a general and highly enantioselective olefin isomerization with a chiral organic catalyst.</p><p>Although the KSI-catalyzed olefin isomerization does not generate a stereocenter, the mechanism underlying the enzymatic proton transfer catalysis is illuminating (Scheme 1). It involves the deprotonation of the C4-β-proton of steroidal ketone by Asp38, while Tyr14 and Asp99 serve as acidic catalytic residues by providing hydrogen bonds to the ketone group. With the oxyanion engaging in hydrogen bonding interactions with Tyr14 and Asp99, the γ-carbon of the dienolate intermediate undergoes a protonation by Asp38 from the β-face. As illustrated in Scheme 2, a 6'-OH cinchona alkaloid in a gauche-open conformation also has a syn-arrangement of acidic and basic active sites, which is similar to how Tyr14 and Asp38 are arranged in the active site of KSI. This observation in combination with the ability of 6'-OH cinchona alkaloids to serve as an acid-base bifunctional catalyst and as a chiral proton donor for asymmetric protonation6,7,8,9 led us to envision the possibility of 6'-OH cinchona alkaloids as a catalyst for an enantioselective isomerization via proton transfer catalysis (Scheme 2).10</p><p>The γ-substituted α,β-unsaturated butenolides constitute a structural motif shared by many biologically interesting natural and synthetic products. Furthermore, they are versatile chiral building blocks for asymmetric synthesis. Thus, the preparation of optically active γ-alkyl α,β-unsaturated butenolides have been a topic of interest. Accordingly, several approaches utilizing easily accessible intermediates from the chiral pool11 or synthetic chiral precursors have been developed.12 Among these only two were catalytic approaches, which involved the Os-mediated asymmetric dihydroxylation of β,γ-unsaturated esters12f,12h, and the Cu-catalyzed heteroallylic asymmetric alkylation, respectively, as the asymmetric induction step.12j,12k Thus, an efficient and general catalytic enantioselective isomerization of γ-substituted β,γ-butenolides to the corresponding α,β-unsaturated butenolides could provide a complementary method that may also expand the scope of chiral butenolides accessible by catalytic asymmetric synthesis (Scheme 2).</p><p>Guided by these considerations we investigated a series of existing 6'-OH cinchona alkaloids and other structurally distinct cinchona alkaloids reported in the literature for their ability to promote the enantioselective isomerization of the commercially available γ-methyl β,γ-butenolide 8a to the corresponding γ-methyl α,β-unsaturated butenolide 9a in dichloromethane at room temperature (Table 1). To our surprise, the catalytic activity of these cinchona alkaloids for this olefin isomerization was found to be critically dependent on not only whether a hydrogen bond donor is presented, but also where it is located in the cinchona alkaloid skeleton. Specifically, while 6'-OH cinchona alkaloids 1a afforded promising activity, cinchona alkaloid 2a, which bears no hydrogen bond donor, was found to be inactive as a catalyst. Although hydroquinine (2b), a 9-OH cinchona alkaloid was active, its activity was significantly lower than that of 1a (entry 3 vs. 1, Table 1). Moreover, the 9-thiourea cinchona alkaloid 4 failed to catalyze this reaction. Even in the presence of both cinchona alkaloid 2a and 6'-OH quinoline 5, no isomerization occurred. These data indicated that the specific spatial relationship between the hydrogen bond donor and acceptor is largely responsible for the outstanding activity of 1a toward this isomerization.</p><p>We next focused our attention to the optimization of enantioselectivity by screening various 6'-OH cinchona alkaloids (1 and 3) already reported in the literature (Table 2). As the optical purity of chiral α,β-butenolide 9a was found to gradually decrease at room temperature in the presence of catalyst 1a (table 1, footnote c), the enantioselectivity afforded by different catalysts was evaluated by ee of 9a obtained at 10 min after the initiation of the isomerization. To our disappointment, 6'-OH cinchona alkaloids bearing 9-aryl and alkyl groups of various bulk (1a, 1b, 3a, 3b) afforded similarly modest selectivity (entries 1–5, Table 2). Interestingly, the enantioselectivity was improved with cuperidine 1c (entry 3, Table 2), albeit not to a synthetically useful level. We also investigated a 6'-thiourea cinchona alkaloid 6. In contrast to the completely inactive 9-thiourea cinchona alkaloid 4, 6 proved to be remarkably active but furnished only a modest enantioselectivity. Following the hypothesis that a perturbation of the acidity of the 6'-OH might improve the catalytic activity and enantioselectivity of the 6'-OH cinchona alkaloids,13 we prepared a novel cinchona alkaloid derivative Q-7 in which the quinoline ring was oxidized into the corresponding N-oxide analogue as 6-hydroxyquinoline-N-Oxide was determined to be more acidic than 6-hydroxyquinoline.14 To our delight, it was shown to be a more active as well as selective catalyst in comparison to 6'-OH cinchona alkaloids 1 and 3 (entry 7 vs. entries 1–5, Table 2). Importantly, QD-7 afforded even higher activity and selectivity while providing the expected opposite sense of asymmetric induction (entry 8, Table 2). By decreasing the reaction temperature to −20 °C, the QD-7-mediated isomerization of 8a afforded 9a in 92% ee (entry 9, Table 2). At this temperature, the racemization of 9a was found to be minimal, if not completely suppressed (entry 10 vs. 9, Table 2).</p><p>With 7 as the catalyst we explored the scope of the asymmetric olefin isomerization (table 3). With either QD- or Q-7 the isomerization of 8a produced 9a in 90% ee. The conversion of this isomerization reached the maximum at near 80% due to a reverse isomerization. Importantly, the separation of 8a and 9a could be achieved via standard chromatographic procedure, thereby allowing the isolation of optically active 9a in moderate yields (entry 1, Table 3). The enantioselectivity of the reactions promoted by 7 was insensitive to the alteration of the steric bulk of the alkyl substituent (entries 2–4). Due to minimal racemization of 9d, the isomerization of γ-isopropyl β,γ-butenolide 8d could be carried out at room temperature at which the reaction proceeded rapidly (entry 4). Moreover, the isomerization of butenolide 8e, which bears a free hydroxyl group on the alkyl chain, was also tolerated by QD-7 (entry 5, Table 3).</p><p>Interestingly, the isomerizations with α,γ-disubstituted β,γ-unsaturated butenolides (8f–k) proceeded readily to completion in a highly enantioselective fashion without detectable racemizations of the corresponding α,β-unsaturated butenolide (9f–k) even at room temperature. Consequently, rapid asymmetric isomerizations of high yield and enantioselectivity could be achieved. For example, with 10 mol% QD-7 the isomerization of 8f was complete within 1.0 h to afford 9f in 95% yield and 90% ee at room temperature (entry 6, Table 3). The catalyst loading could be reduced to as low as 0.5 mol% without sacrificing either the enantioselectivity or the yield (entry 7, Table 3), though a longer reaction time of 12h was required. As exemplified with the isomerization of 8f, a complete and high yielding reaction could still be attained even with 0.1 mol% catalyst, although the enantioselectivity was slightly decreased from 90% to 86% ee (entry 8, Table 3). Importantly, such catalytic efficiency could be extended to a series of substrates bearing various α- and γ-alkyl groups (entries 9–13, Table 3). The β,γ-disubstituted β,γ-unsaturated butenolides (8l–m) turned out to be a more challenging class of substrates. The isomerization of these butenolides proceeded in lower yet still useful level of enantioselectivity. In light of the lack of an efficient approach toward optically active β,γ-disubstituted α,β-butenolides,12f,12i,15 the current isomerization represents a valuable access toward these chiral building blocks.</p><p>To gain insights into the reaction mechanism, we carried out a preliminary kinetic study of the QD-7-mediated isomerization of 8f. The reaction was found to show first-order dependence on the QD-7 and 8f, respectively.16 In addition we measured the carbon isotope effect on carbon 2–6 of 8f employing Singleton's NMR technique at natural abundance to discern the rate-limiting step of the olefin isomerization.16,17 A pronounced carbon isotope effect was observed on the γ carbon when the 13C ratio of recovered 8f at 71% conversion was compared to that of the virgin sample (13C(recovered)/13C(virgin) at Cγ = 1.023, average of three runs) (eq. 1). This kinetic isotope effect indicates the γ-protonation step is the rate-limiting step of the isomerization reaction.</p><p> (1)</p><p>Based these results from our mechanistic studies and the drastic difference in catalytic activity between 6'-OH cinchona alkaloid 1a and the corresponding 6'-OMe congener 2a (entry 1 vs. 2, Table 1), we propose a catalytic cycle for this enantioselective isomerization reaction (Scheme 3). In this mechanism, the deprotonation of β,γ-butenolide 8 occurs after the hydrogen bond-based complexion of 8 with catalyst 7, which is followed by the rate-determining γ-protonation. It should be noted that in the protonation step, either the protonated quinuclidine or the 6'-OH could serve as the proton donor.</p><p>In summary, we have realized an unprecedented enantioselective olefin isomerization via biomimetic proton transfer catalysis with a new chiral organic catalyst. This asymmetric transformation is applicable to a broad range of β,γ-butenolides bearing one or more substituents. With a low catalyst loading and a simple experimental protocol, this reaction should provide a valuable method for the asymmetric synthesis of chiral α,β-butenolides. Mechanistic studies have revealed that the protonation step is the rate-determining step of this organocatalytic olefin isomerization. Interestingly, this stands in contrast to the enzyme-catalyzed olefin isomerizations, which feature a rate-determining deprotonation.1a</p><p> ASSOCIATED CONTENT </p><p>Supporting Information. Experimental procedures and characterization of the products. This material is available free of charge via the Internet at http://pubs.acs.org.</p>

PubMed Author Manuscript

Development of an in vivo active, dual EP1 and EP3 selective antagonist based on a novel acyl sulfonamide bioisostere

Recent preclinical studies demonstrate a role for the prostaglandin E2 (PGE2) subtype 1 (EP1) receptor in mediating, at least in part, the pathophysiology of hypertension and diabetes mellitus. A series of amide and N-acylsulfonamide analogs of a previously described picolinic acid-based human EP1 receptor antagonist (7) were prepared. Each analog had improved selectivity at the mouse EP1 receptor over the mouse thromboxane receptor (TP). A subset of analogs gained affinity for the mouse PGE2 subtype 3 (EP3) receptor, another potential therapeutic target. One analog (17) possessed equal selectivity for EP1 and EP3, displayed a sufficient in vivo residence time in mice, and lacked the potential for acyl glucuronide formation common to compound 7. Treatment of mice with 17 significantly attenuated the vasopressor activity resulting from an acute infusion of EP1 and EP3 receptor agonists. Compound 17 represents a potentially novel therapeutic in the treatment of hypertension and diabetes mellitus.

development_of_an_in_vivo_active,_dual_ep1_and_ep3_selective_antagonist_based_on_a_novel_acyl_sulfon

<p>Prostanoids are a family of oxidative metabolites of arachidonic acid that act in a autocrine and paracrine fashion. The cyclooxygenase activity of COX-1 and COX-2 converts arachidonic acid to prostaglandin G2 (PGG2) and the peroxidase activity of the same enzymes reduces PGG2 to prostaglandin H2 (PGH2). PGH2 is then isomerized to the five principal prostanoids by their respective synthases. Prostanoids bind to and activate a family of cell surface G-protein coupled receptors.1 Prostaglandin E2 (PGE2) is formed from PGH2 by prostaglandin E synthases (cPGES, mPGES-1, mPGES-2) and is a major prostanoid produced by the kidney and the vasculature.2 The bioactivity of PGE2 is mediated through four subtypes of E-Prostanoid (EP) receptors, designated EP1-EP4.3 EP2 and EP4 couple to stimulatory G-proteins, which increase intracellular cAMP when activated. EP3 canonically couples to inhibitory G-proteins, suppressing cAMP accumulation. Both EP1 and EP3 are known to induce calcium flux into the cell.4, 5 The tissue localization of each of these EP receptors produces diverse and sometimes opposing biological activities of PGE2 in vivo.3</p><p>Hypertension and diabetes are the primary causes of 62 % of patients with End-Stage Renal Disease (ESRD) and 72 % of patients that develop ESRD each year6, which requires life-long dialysis or kidney transplantation for survival. Elimination of PGE2 production with COX inhibitors,7, 8 like NSAIDs,9 is not a viable option as highlighted in a number of clinical trials. Recent studies in rodents and humans have suggested a role for the EP1 receptor in mediating at least part of the pathophysiology of diabetes mellitus10-12 and hypertension.13-16 EP1 has been prosecuted as a potential therapeutic target for chronic pain.17-21 As such, small molecule, drug-like antagonists of EP1 have been developed. Human prostanoid receptor-targeting molecules are often nonselective,22 owing to the evolution of the EP family of GPCRs to recognize the same endogenous ligand, PGE2. The molecular pharmacology of these compounds at mouse prostanoid receptors is less well known, often poorly selective, and not always comparable to human pharmacology.23 In order to study these molecular targets more precisely, we developed EP1 antagonists selective for the mouse receptor to use in mouse models of hypertension and diabetes mellitus.</p><p>To develop antagonists selective for the mouse EP1 receptor, we started with compound 7 (Figure 1), synthesized as previously described (Scheme 1)24. Diethyl dipicolinic acid (1) was reduced with NaBH4 to 2. Parikh-Doering oxidation of 2 with sulfur trioxide-pyridine complex and DMSO produced the unstable aldehyde 3. 4-chlorophenoxide was then reacted with 3, followed by neutralization with HCl to form 4. Reduction of the secondary alcohol of 4 under H2 and Pd/C with the addition of H2SO4 and ZnBr2 gave 5. Alkylation of 5 with 2-fluoro-4-chlorobenzyl bromide and cleavage of the ester by refluxing with NaOH produced the sodium salt (6) of the lead (7) which was formed by protonation of 6. The lead compound was shown to have good affinity for the human EP1 receptor and was stable in microsomes and S9 fractions of several species. However, 7 was previously reported to have a high-affinity interaction with the human thromboxane (TP) receptor.24 We evaluated the molecular pharmacology of 7 at the mouse EP receptors as well as the mouse TP receptor.</p><p>Compound 7 was confirmed to be a functional antagonist of mEP1 in vitro and to have submicromolar affinity for the mouse EP1 receptor by Schild Analysis (Figure 2). 7 had no detectable affinity for mouse EP3 or EP4 receptors by radioligand binding assays. 7 had poor, but detectable affinity at mouse EP2, and suppressed signaling through mouse TP receptor at concentrations 100-fold higher than at the human receptor (Table 1), confirming the off-target activity of 7 at mouse TP.</p><p>Results from in vivo pharmacokinetics experiments (Table 2) revealed compound 7 to possess a moderate systemic plasma clearance (CLp) and volume of distribution predicted at steady-state (Vss), subsequently displaying a short half-life (t1/2, > 60 min) in mice receiving a parenteral administration of the EP1 receptor antagonist. We observed a bioavailability (%F) of approximately 14 % following the oral administration (10 mg/kg) of 7 to mice.</p><p>Recently, Ostenfeld et al. have shown that in rats 7 is cleared primarily by glucuronidation and sequestration into the bile.25 With the goal of inhibiting glucuronidation while improving molecular pharmacology of 7, a series of carboxylic acid bioisosteres of 7 were pursued. N-acylsulfonamides are common carboxylic acid bioisosteres that have been successfully implemented in antagonists of angiotensin II AT1 receptors26 as well as EP3 receptors.27 A series of analogs (8 - 21) resulting from the amidation of 7 was prepared (Table 3). Tertiary amide analogues of 7 (9 and 14) were included to evaluate a structure-activity role of an acidic proton in ligand-receptor interactions. Each was synthesized by coupling 7 to a series of primary and secondary amines (8 - 16) and sulfonamides (17 – 21) employing common activators such as EDC·HCl, HOBt, and DIPEA in DMF (Scheme 1).</p><p>The molecular pharmacology observed for 8 - 21 was determined at mEP1 - mEP4 and mTP (Table 3). Generally, N-acylsulfonamides retained mEP1 affinity similar to 7 (representative data for 17, Figure 3) while the amide series had reduced affinity for mEP1. Each analog displayed reduced affinity for mEP2 and mTP. Interestingly, four analogs (11, 17, 18, and 21) displayed enhanced affinity for mEP3, a potential therapeutic target for hypertension- and diabetes mellitus-related ESRD. EP3 is of particular interest as it shares signaling pathways and endogenous ligands with EP1 and may represent a compensatory signaling pathway in the event of EP1 blockade.3, 5, 16, 28, 29 These dual-selectivity compounds were confirmed to be functional antagonists of mEP3 by Schild Analysis (data not shown).</p><p>We subsequently determined the intrinsic clearance (Clint) of several potent amide and N-acylsulfonamide analogs (Table 4). Results indicated an exceptional instability to metabolism in vitro, displaying estimated predicted hepatic clearance (CLHEP) values that approached the hepatic blood flow in mice (QH, 90 mL/min/kg).</p><p>Results from metabolite identification studies in hepatic subcellular fractions indicated extensive biotransformation of the amide 11 and the N-acylsulfonamide 17, including NADPH-independent hydrolysis (i.e., esterases) and NADPH-dependent oxidation (i.e., P450) of these analogs. Figure 4 depicts the metabolism of 17, including the hydrolysis of the sulfonamide (M1), and P450-mediated oxidation of the methylene linker (M2) and benzylic oxidation (M3). The extent of plasma protein binding (fraction unbound, Fu) in mouse was determined to be extensive for three compounds assessed (Fu: 7 = 0.005, 11 = 0.010, 17 = 0.004).</p><p>Given the molecular pharmacology and in vitro metabolism data, we proceeded to evaluate the in vivo pharmacokinetics of 17. Mice (n = 3) were subsequently administered a subcutaneous dose (5 mg/kg) with intermittent plasma collections to measure systemic levels of 17 (Figure 5). Compound 17 achieved a maximum plasma concentration (Cmax) of 504 nM (± 167) 2 hours (tmax) following subcutaneous administration and displayed an area-under-the-curve (AUC) of 7508 nM*h.</p><p>To evaluate 17 as an antagonist of EP1 and EP3 in vivo, we measured blockade of mEP1 and mEP3 acute vasopressor activity in mice. Left common carotid arteries and right jugular veins of anesthetized mice were cannulated. Direct arterial pressure was measured via carotid catheter. Vasoactive substances were administered via jugular catheter. 17PTPGE2 was used to acutely raise mean arterial pressure (MAP) via mEP1 and sulprostone was used for mEP3 (Figure 6). Agonists were administered IV through the jugular catheter 2 h after subcutaneous administration of 17. Pretreatment of mice with 5 mg/kg 17 administered subcutaneously significantly attenuated the pressor activity of an IV bolus of 20 μg/kg 17PTPGE2 (ΔMAP 50.3 ± 5.5 mmHg vs. 27.0 ± 3.6 mmHg). Pretreatment with 17 also significantly suppressed pressor activity of an IV bolus of 10 μg/kg sulprostone (ΔMAP 53.3 ± 2.3 mmHg vs. 32.0 ± 3.5 mmHg). To ensure the observed effect was selective for EP-mediated vasoconstriction, phenylephrine (10 μg/kg) was shown to be unaffected by pretreatment with 17 (data not shown).</p><p>In conclusion, we have identified a novel, dual-selectivity antagonist (17) of the mouse EP1 and mouse EP3 receptors possessing an acylsulfonamide bioisostere for the prototypical carboxylic acid moiety of EP ligands. 17 was found to have indistinguishable affinity for mEP1 as for mEP3 (mEP1 pKD vs. mEP3 pKI, P = 0.40, Student's two-tailed t test). 17 had improved selectively over mEP2 and mTP. 17 was less stable in mouse hepatic microsomes than 7, due in part to hydrolysis of 17 to 7, a problem effectively circumvented by subcutaneous administration of 17. Finally, we confirmed 17 is a functional antagonist of mEP1 and mEP3 in vivo by blocking mEP1/mEP3-mediated acute vasopressor activity in anesthetized mice. While the attenuation of pressor activity appears to be incomplete, these results recapitulate experiments performed in mice with genetic disruptions of EP1.13 Dual specificity EP1/EP3 antagonists represent a novel class of potential ESRD therapeutics we hypothesize will be more beneficial than blocking either receptor alone.</p>

PubMed Author Manuscript

Evaluation of in vitro anti-proliferative and immunomodulatory activities of compounds isolated from Curcuma longa

The rhizome of Curcuma longa (CL) has been commonly used in Asia as a potential candidate for the treatment of different diseases, including inflammatory disorders and cancers. The present study evaluated the anti-proliferative activities of the isolated compounds (3 curcuminoids and 2 turmerones) from CL, using human cancer cell lines HepG2, MCF-7 and MDA-MB-231. The immunomodulatory activities of turmerones (\xce\xb1 and aromatic) isolated from CL were also examined using human peripheral blood mononuclear cells (PBMC). Our results showed that the curcuminoids (curcumin, demethoxycurcumin and bisdemethoxycurcumin) and \xce\xb1-turmerone significantly inhibited proliferation of cancer cells in dose-dependent manner. The IC50 values of these compounds in cancer cells ranged from 11.0\xe2\x80\x9341.8 \xce\xbcg/ml. Alpha-turmerone induced MDA-MB-231 cells to undergo apoptosis, which was confirmed by annexin-V & propidium iodide staining, and DNA fragmentation assay. The caspase cascade was activated as shown by a significant decrease of procaspases-3, -8 and -9 in \xce\xb1-turmerone treated cells. Both \xce\xb1-turmerone and aromatic-turmerone showed stimulatory effects on PBMC proliferation and cytokine production. The anti-proliferative effect of \xce\xb1-turmerone and immunomodulatory activities of ar-turmerone were shown for the first time. The findings revealed the potential use of CL crude extract (containing curcuminoids and volatile oil including turmerones) as chemopreventive agent.

evaluation_of_in_vitro_anti-proliferative_and_immunomodulatory_activities_of_compounds_isolated_from

1. Introduction<!>2.1 Materials<!>2.2 Preparation of crude extract and sub-fractions<!>2.3 Isolation of \xce\xb1-turmerone and ar-turmerone from lipophilic fraction (F1)<!>2.4 Isolation of curcumin, demethoxycurcumin and bisdemethoxycurcumin from lipophilic fractions (F2 and F3)<!>Cell line culture<!>Preparation of human peripheral blood mononuclear cells (PBMC)<!>2.6 Cell viability assay<!>2.7 Cell cycle analysis<!>2.8 Annexin V-FITC/PI staining assay<!>2.9 DNA fragmentation assay<!>2.10 Western blot analysis<!>2.11 Detection of mitochondrial transmembrane potential (\xce\x94\xcf\x88m)<!>2.12 Cell proliferation assay of PBMC and collection of culture supernatant<!>2.13 Determination of human IL-2, IFN-\xce\xb3 and TNF-\xce\xb1 production by PBMC<!>2.14 Statistical analysis<!>3.1 Isolation and characterization of \xce\xb1-turmerone, ar-turmerone, curcumin, demethoxycurcumin and bisdemethoxycurcumin from lipophilic fractions<!>3.2 Comparison of inhibitory activities on cell proliferation of HepG2, MCF-7 and MDA-MB-231 cells among different curcuminoids and turmerones<!>3.3 Effects of curcuminoids and turmerones on cell viability of normal skin cells<!>3.4 Presence of apoptotic cell fraction in \xce\xb1-turmerone-treated MDA-MB-231 cells<!>3.5 \xce\xb1-turmerone treatment induced phosphatidylserine externalization in MDA-MB-231 cells<!>3.6 \xce\xb1-turmerone induced apoptosis via activation of caspase-dependent pathway<!>3.7 \xce\xb1-turmerone caused mitochondrial transmembrane depolarization in MDA-MB-231 cells and altered expression of Bcl-2 protein<!>3.8 Effects of \xce\xb1-turmerone and ar-turmerone on cell proliferation and cytokine production of PBMC<!>4. Discussion<!>

<p>Various species of the perennial plant Curcuma (family Zingiberaceae) are found as common ingredients in many health supplements in Asia. Curcuma longa Linn. (CL) is a spice commonly used in Indian and Chinese cuisine and traditional medicine. In Ayurvedic medicine, the rhizome of CL was suggested to be used as a stimulant, tonic, stomachic and depurative. It was also used in combinations for sprains and bruises. According to the Ayurvedic Pharmacopoeia of India, essential oil from rhizome of CL was used as a carminative, stomachic and tonic (Ministry of Health & Family Welfare, Government of India, 2001). In ancient Chinese medicine, the use of the plant Curcuma firstly appeared in "Yao Xing Lun" during Tang dynasty (A.D.618–907) and was formally recorded in the Compendium of Materia Medica (Ben Cao Gang Mu) of Ming dynasty (A.D.1590). According to the Chinese Pharmacopoeia, CL was suggested to have the functions of eliminating blood stasis, promoting the flow of "qi", stimulating menstrual discharge and relieving pain (Chinese Pharmacopoeia Commission, 2005).</p><p>In modern pharmacological studies, CL constituents, in particular curcumin had previously been shown to possess anti-inflammatory, antioxidant, and chemopreventive properties. Curcumin was known to suppress NF-κB activation (Singh and Aggarwal, 1995) and inhibited COX-2 through the arachidonic acid metabolism pathway which accounted for its anti-inflammatory activities (Plummer et al., 1999). Curcumin also suppressed tumor growth by blocking signal transduction pathways in the target cells (Mori et al., 2001). Recent reports suggested that curcumin down-regulated expression of cell proliferation and anti-apoptotic and metastatic gene products (Aggarwal et al., 2006) and potentiated antitumor activity of gemcitabine in pancreatic cancer model (Kunnumakkara et al., 2007). Furthermore, curcumin has been shown to inhibit proliferation and induce apoptosis in human leukemic cell lines (Anuchapreeda et al., 2008), human breast (Ramachandran et al., 2002, 2005; Simon et al., 1998), hepatoma (Cao et al., 2006, 2007; Chan et al., 2005), pancreatic (Kunnumakkara et al., 2007; Li et al., 2004) and colon (Cheng et al., 2007) cancer cells. On the other hand, curcumin has been suggested to modulate the proliferation and cellular response of various immune cell types, such as T cells, B cells, macrophages, neutrophils, NK cells and dendritic cells (Jagetia and Aggarwal, 2007; Bhaumik et al., 2000).</p><p>In addition to curcumin, other constituents of CL such as sesquiterpenoids, also possessed various biological activities. Turmeric oil was shown to offer protection against benzo(a)pyrene induced increase in micronuclei in human lymphocytes. It also decreased the number of micronucleated cells in oral submucous fibrosis patients (Hastak et al., 1997). Volatile oil isolated from CL had been reported to have antibacterial (Norajit et al., 2007) and antifungal activities (Roth et al., 1998), anti-proliferation in human leukemia (Shi et al., 2003) and lung cancer cells (Wang et al., 2005). Previous studies have shown that aromatic turmerone isolated from CL could induce apoptosis in human leukemia cells (Aratanechemuge et al., 2002) and murine leukemia cells (Ji et al., 2004). The aromatic turmerone also possessed antioxidant (Jayaprakasha et al., 2002) and antiplatelet activities (Lee, 2006).</p><p>There are only a few pharmacological studies regarding the extracts of CL and the individual curcuminoids of CL (other than curcumin) (Anuchapreeda et al., 2008; Funk et al., 2006; Deters et al., 2008). The present study aimed to isolate and characterize the bioactive components from CL using bioassay-guided fractionation and to evaluate the biological activities of such components using in vitro models. The anti-proliferative effects in human hepatoma cells and breast cancer cells, as well as the immunomodulatory activities in human peripheral blood mononuclear cells of the isolated compounds from CL prepared in our laboratory were evaluated.</p><!><p>Dried rhizome of Curcuma longa L. (CL) was purchased from a herbal supplier in Hong Kong with the source of origin in India. Organoleptic, microscopic and chemical authentication were accomplished in accordance with the Chinese Pharmacopoeia (State Pharmacopoeia Commission of People's Republic of China, 2005). Authenticated voucher specimen (no. HK 40400) was deposited in the Hong Kong Herbarium of the Agriculture, Fisheries and Conservation Department of Hong Kong Special Administration Region, China.</p><p>The organic solvents, dichloromethane, 95% ethanol, ethyl-acetate, n-hexane and methanol were purchased from Lab-Scan (Thailand) and were either HPLC-grade or AR grade.</p><!><p>One kilogram of CL rhizome was cut into small pieces and soaked in 1.0 L of 95% (v/v) ethanol for 60 minutes. It was then extracted twice with 1.5 L of 95% (v/v) ethanol under reflux for 90 minutes. Following filtration, the ethanolic extracts were combined and centrifuged at 5250×g for 10 minutes. About 2.0 L of supernatant was collected and evaporated under reduced pressure at 60°C to dryness (with a yield of 7.7% w/w).</p><p>The dried ethanolic extract was allowed to re-dissolve in 1.0 L of 10% (v/v) methanol in water, followed by partitioning twice with 400 ml of n-hexane. The hexane layers were combined and evaporated to dryness to give fraction 1 (F1, 34.8% w/w). The remaining aqueous layer was subjected to partition with firstly dichloromethane and then ethyl acetate successively by following the same procedure to give dried fraction F2 (37.8% w/w) and F3 (15.1% w/w), respectively. The remaining aqueous layer was collected and concentrated under reduced pressure at 50°C. The concentrated aqueous solution was subjected to lyophilization to give dried fraction 4 (F4, 11.8 % w/w). The crude ethanolic extract and its fractions (F1-F4) have been tested for their inhibitory and stimulatory activities in cancer cells and peripheral blood mononuclear cells (PBMC), respectively. Fraction F1 showed the most potent stimulatory effect on PBMC proliferation while fractions F2 and F3 exhibited potent inhibitory activities in cancer cells (data not shown). Therefore, fractions F1, F2 and F3 were subjected to further fractionation.</p><!><p>The fraction F1 was subjected to silica gel column chromatography eluting with dichloromethane and methanol (98:2). The eluents were monitored by thin layer chromatography (TLC) and similar fractions were combined to afford 5 fractions (F1-A, F1-B, F1-C, F1-D and F1-E). The inhibitory activities in cancer cells of the fractions had been tested and the fraction F1-B was shown to be most potent (data not shown). The fraction F1-B was further fractionated by silica column chromatography eluting with hexane and ethyl acetate (96:4) to afford 5 subfractions (F1-B1, F1-B2, F1-B3, F1-B4 and F1-B5). Subfractions F1-B3 and F1-B5 were characterized and identified as α-turmerone (0.03% w/w) and ar-turmerone (0.027% w/w), respectively, using 1H and 13C NMR spectral analysis and mass spectrometry.</p><!><p>The fractions F2 and F3 were combined as similar constituents were present in these fractions. The combined fraction was subjected to silica gel column chromatography eluting with dichloromethane and methanol (95:5). The eluents were monitored by thin layer chromatography (TLC) and similar fractions were combined to afford 6 fractions (F2&3-A, B, C, D, E and F). The solvents from elutes were concentrated under vacuum and recrystallized. Curcumin (0.35% w/w), demethoxycurcumin (0.08% w/w) and bisdemethoxycurcumin (0.04% w/w) were recrystallized with ethanol from F2&3-A, F2&3-C and F2&3-E, respectively. The purified compounds were characterized using 1H and 13C NMR spectral analysis and mass spectrometry.</p><!><p>The human hepatoma cell line HepG2, human breast cancer cell lines MCF-7 and MDA-MB-231, and human skin fibroblast cell line Hs-68 were obtained from American Type Culture Collection (MD, USA). The cancer cell lines were maintained in RPMI-1640 medium, while Hs-68 cells were maintained in Dulbecco's modified Eagle's medium (DMEM), containing 10% heat-inactivated fetal bovine serum (FBS), 100 units/ml penicillin, and 100 μg/ml streptomycin. Media and supplements were obtained from Invitrogen GIBCO (NY, USA). The cells were incubated at 37°C in a humidified atmosphere of 5% CO2 and those cells in the exponential growth phase were used for experiments.</p><!><p>Fresh human buffy coat obtained from the Hong Kong Red Cross Blood Transfusion Service was diluted with phosphate-buffered saline (PBS) at a ratio of 1:1. The diluted buffy coat sample was layered on equal volume of Ficoll-Paque™ Plus solution (GE healthcare, UK) and then centrifuged at 800×g for 20 minutes at 18°C. The thin white middle PBMC layer was collected and the PBMC were washed with PBS twice and centrifuged at 100×g for 10 minutes at 18°C. The supernatant was discarded and the PBMC were resuspended in RPMI-1640 medium plus 10% FBS, 100 units/ml penicillin, and 100 μg/ml streptomycin. The cell number was counted with a hematocytometer and the viability of the cells was checked by trypan blue exclusion assay. Only the isolated cells with 95% or above viability were resuspended to the target density, which was 4 × 106 cells/ml, for experiments. The PBMC were incubated at 37°C in a humidified atmosphere of 5% CO2.</p><!><p>The HepG2, MCF-7, MDA-MB-231 and Hs-68 cells (1 × 104/well) or PBMC (4× 105/well) were seeded in 96-well flat-bottom culture plates (Iwaki, Japan) and incubated with various concentrations (3.125–100 μg/ml) of isolated compounds from CL prepared in our laboratory (as mentioned in sections 2.3 and 2.4) for 48 hours. Plain medium was added to the control wells. The turmerones were diluted in absolute ethanol at 100 mg/ml and stored at −20°C. The lyophilized powder of curcuminoids was dissolved in DMSO at 100 mg/ml and stored at −20°C. For experiments, the final concentrations of the tested compounds were prepared by diluting the stock with culture medium. Control wells and drug-treated wells received the vehicle solvent (1% v/v ethanol or 0.5% v/v DMSO).</p><p>Following the incubation of cells with the extracts, the medium was discarded and 30 μl of 5 mg/ml 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl-tetrazolium bromide (MTT; Sigma, USA) in PBS were added to each well and the plates were further incubated for 3 hours at 37°C. For the PBMC, the plates were centrifuged at 300×g for 10 minutes. The supernatant was then removed and 100 μl of DMSO was added to each well to dissolve the purple formazan crystals. The absorbance at a wavelength of 540 nm was measured spectrophotometrically with a BMG FLUOstarOptima microplate reader (BMG LABTECH GmbH, Germany). Results were expressed as the percentage of MTT absorbance with respect to vehicle-treated control cells.</p><!><p>Human breast cancer cells MDA-MB-231 (3 × 105/well) were seeded at 6-well culture plates and incubated overnight. The cells were treated with α-turmerone at various concentration for 24 hours. Cells were harvested after incubation, washed and fixed in 70% v/v ethanol at 4°C overnight. Before analysis, the cells were resuspended in PBS containing propidium iodide (PI, 20 μg/ml) and RNase A (10 μg/ml) at 37°C in the dark for 30 minutes. Cell cycle distribution was analyzed by flow cytometry (Becton Dickinson FACSCanto II, USA). Data from 10,000 cells per sample were collected and Modfit LT (Verity Software House, ME, USA) was used for cell cycle and apoptotic peak modeling.</p><!><p>Human breast cancer cells MDA-MB-231 (3 × 105/well) were treated with α-turmerone at various concentration for 24 hours. Then, the cells were collected, washed and stained with Annexin-V FITC kit according to the manufacturer's protocol (Trevigen, MD, USA). Briefly, cells were washed twice with 1X binding buffer and incubated in 100 μl labeling solution containing 1 μl annexin-V FITC conjugate and 10 μl PI in the dark for 15 min at room temperature. The fluorescence of the samples was detected by flow cytometry (Becton Dickinson FACSCanto II, USA).</p><!><p>Human breast cancer cells MDA-MB-231 (1 × 106) were seeded in 100 mm culture dish and incubated for 24 hours to allow attachment. Various concentrations (40–80 μg/ml) of α-turmerone or 0.4% v/v ethanol (control) were added to the dishes and incubated for 48 hours. After treatment, attached and floating cells were harvested and washed twice with cold PBS. The cell pellets were then lysed with 200 μl DNA lysis buffer (200 mM Tris-HCl, 100mM EDTA, 1% SDS, pH 8.3) for 15 minutes at 37°C. The samples were centrifuged at 6000×g for 5 minutes to collect the DNA-containing supernatants. The supernatant was added with 50 μl of 5% SDS and 10 μl RNase A (0.4 mg/ml) at 56°C for 90 minutes. After incubation, 20 μl of proteinase K (1.5 mg/ml) was added into the mixture and incubated at 56°C for further 90 minutes. The DNA was precipitated with 30 μl of 3M sodium acetate and 750 μl of absolute ethanol. The resulting DNA was collected by centrifugation, rinsed with cold 75% v/v ethanol and then absolute ethanol. The DNA pellet was left to air dry and the dried pellet was dissolved in 30 μl TE buffer at 37°C for 30 minutes. The extracted DNA was subjected to electrophoresis through a 1.5% agarose gel containing ethidium bromide. The DNA bands were visualized under ultraviolet illumination.</p><!><p>Human breast cancer cells MDA-MB-231 (1 × 106) were seeded in 100 mm culture dish and incubated for 24 hours to allow attachment. Various concentrations (40–80 μg/ml) of α-turmerone or 0.4% v/v ethanol (control) were added to the dishes and incubated for 24 or 48 hours. After treatment, attached and floating cells were harvested and washed twice with PBS. The cell pellets were then lysed with whole cell extraction buffer (2% SDS, 10% glycerol, 625 mM Tris-HCl, pH6.8) for 15 minutes on ice. The samples were heated at 95°C for 10 minutes and then centrifuged at 14000×g for 20 minutes at 4°C. The supernatant proteins were resolved by 12% SDS-polyacrylamide gel and transferred to 0.45μm polyvinylidene fluoride (PVDF) membrane (Immobilon, Millipore, USA). The membrane was blocked with 5% non-fat milk in Tris buffered saline containing Tween-20 (20mM Tris-HCl, pH7.6, 150 mM NaCl, 0.1% Tween-20). The blots were incubated overnight with primary antibodies against human β-actin (1:2000, Sigma, USA), Bcl-2, procaspases-3, and -8 (1:100, Santa Cruz, USA) procaspase-9 and p21 (1:1000, Cell Signalling, USA) and cyclin D (1:200, BD Pharmingen, USA). After incubation with the secondary horseradish peroxidase-conjugated antibodies (1:2000, Invitrogen, USA) for 1 hour, detection was performed using enhanced chemiluminescence assay kit (GE Healthcare, USA).</p><!><p>The Δψm was analyzed by JC-1, a cationic carbocyanine fluorescent dye accumulated in mitochondria. JC-1 forms monomers and emits green fluorescence when Δψm is relatively low. Cells with high Δψm were indicated by red fluorescence as JC-1 aggregated. Human breast cancer cells MDA-MB-231 (3 × 105/well) were treated with α-turmerone at various concentrations for 24 hours. Then, the cells were collected, washed and incubated with 10 μM JC-1 for 30 min at 37°C in darkness. Then, changes in JC-1 signals were analyzed by flow cytometry (Becton Dickinson FACSCanto II, USA).</p><!><p>The isolated PBMC were seeded in a 96-well flat bottom microplate and incubated with various concentrations (3.125–100 μg/ml) of turmeric extracts or isolated compounds for 24, 48 and 72 hours. To ensure the effects of Turmeric extracts on the PBMC were not due to the presence of LPS, polymyxin B (10 μg/ml), an antibiotic that can bind LPS (Duff and Atkins, 1982), were added to the extracts. The mitogen, phytohaemagglutinin (PHA), was added to target wells when appropriate at a final concentration of 10 μg/ml. Polymyxin B and PHA were purchased from Sigma (USA). Proliferation rates were estimated by (methyl-3H)-thymidine incorporation. After the incubation period, the microplates were centrifuged at 300×g for 10 minutes to obtain cell-free supernatant. The supernatant was collected and stored at −80°C until cytokine ELISA experiments.</p><p>The effects of Turmeric extracts on the proliferation of resting and mitogen (PHA)-activated PBMC were determined by (methyl-3H)-thymidine incorporation. Following incubation of cells with the extracts for 72 hours, tritiated thymidine (0.5 μCi/well; GE healthcare, UK) was added into each well and further incubated for 6 hours. Then the cells were harvested on glass fiber filters by cell harvester. Radioactivity in the filters was measured by Packard TopCount NXT™ Microplate Scintillation and Luminescence Counter (PerkinElmer Inc., USA).</p><p>For in vitro cell proliferation assay, the turmerones were dissolved in absolute ethanol. During experiment, all the drug-treated or control wells were added same concentrations of vehicle (0.5% DMSO or 1% ethanol) so that the only variable was the concentrations of test compounds. In fact, control wells without vehicle have also been set in each experiment and our data showed that the presence of vehicle (DMSO or ethanol) did not significantly change in proliferation or cytokine production of PBMC cultures (n=8, data not shown).</p><!><p>The PBMC culture supernatants were subjected to test for the concentrations of cytokines such as IL-2 and IFN-γ by enzyme-linked immunosorbent assay (ELISA; BD Pharmingen, USA). The assay was carried out according to the procedures recommended in the ELISA kit manual.</p><!><p>Data were expressed as the mean ± standard deviation (S.D.). Statistical analysis and significance, as measured by the Student's t-test for paired samples or one way analysis of variance (ANOVA) followed by Bonferroni post-test, as appropriate, were performed using GraphPad PRISM software version 3.0 (GraphPad Software, USA). In all comparisons, p < 0.05 was considered statistically significant.</p><!><p>Compounds F1-B3 and F1-B5 were characterized and identified as α-turmerone and ar-turmerone, respectively, using 1H and 13C NMR spectral analysis and mass spectrometry (Table 1a). Chemical shifts of the turmerones were in accordance with the reported values (Golding and Pombovillar, 1992; Dhingra et al., 2007). The compounds F2&3-A, F2&3-C and F2&3-E were characterized and identified as curcumin, demethoxycurcumin and bisdemethoxycurcumin, respectively, using 1H NMR spectral analysis and mass spectrometry (Table 1b). Chemical shifts of the curcuminoids were in accordance with the reported values (Jayaprakasha et al., 2002; Park et al., 2005; Uehara et al., 1992). Their chemical structures are shown in Fig. 1.</p><!><p>Assays for proliferative response in human hepatoma cells (HepG2) and human breast cancer cells (MCF-7 and MDA-MB-231) were performed to evaluate the anti-proliferative effects of 3 curcuminoids isolated from CL ethanolic extracts (Fig. 2). The cells were treated with a series of concentrations of curcuminoids from 1.6–50.0 μg/ml for 48 hours. As shown in Fig. 2, all three curcuminoids showed significant anti-proliferative activities in HepG2, MCF-7 cells and MDA-MB-231 cells in a concentration-dependent manner (Fig. 2A-2C). The concentrations producing 50% growth inhibition (IC50) of the curcuminoids on the three cell lines were listed in Table 2. Among the species tested, curcumin exhibited the most potent inhibitory effect on MDA-MB-231 cells proliferation (IC50 = 11.0 μg/ml). Demethoxycurcumin and bisdemethoxycurcumin could achieve 50% proliferation inhibition with a concentration of 11.4 and 12.1 μg/ml, respectively (Table 2).</p><p>To evaluate the effects of turmerones isolated from CL on cell proliferation of cancer cells, cells were treated with a series of concentrations of turmerones from 3.125–100 μg/ml for 48 hours. In Fig. 2, α-turmerone at concentration of 50 μg/ml caused nearly 70% inhibition of cell proliferation compared to the control in all three cell lines. The inhibitory activities of α-turmerone (12.5–100 μg/ml) were significantly lowered than those of ar-turmerone. To inhibit cell proliferation of HepG2 and MDA-MB-231 cells at 50%, the concentrations of α-turmerone were lowered than that required for MCF-7 cells (Table 2).</p><!><p>To determine whether the inhibition of cell proliferation on these cancer cell lines was selective, human normal fibroblast (Hs-68) cells were used as a negative control. Hs-68 cells were incubated with the curcuminoids and turmerones under the same conditions as the cell viability assay of the cancer cells. The tested curcuminoids (3.125–50 μg/ml) induced no significant suppression on the proliferation of Hs-68 (Table 2). The maximum percentage of inhibition of cell proliferation of each tested curcuminoids (50 μg/ml) was ranged from 23.5% to 63.4% (n = 4). Besides, the tested turmerones (3.125–100 μg/ml) did not show significant suppression on the proliferation of Hs-68 (Table 2). When compared with the cancer cells, Hs-68 cells were much less susceptible to the anti-proliferative effects of the curcuminoids and turmerones at the concentration of 50 μg/ml, respectively.</p><!><p>Since the IC50 value of α-turmerone was found to be the smallest in MDA-MB-231 cells, further experiments were performed in MDA-MB-231 cells to elucidate the mode of cell death induced by α-turmerone. Cell cycle phases were detected by propidium iodide (PI) staining. As shown in Fig. 3A & 3B, the population of sub-G1 phase increased from 2.80% to 38.86% after 48 hours of α-turmerone exposure (30–45 μg/ml). Apoptotic cells, as judged from the appearance of a sub-G1 peak, were observed in α-turmerone-treated cells. However, there was no arrest at any phase of the cell cycle after treatment. In addition, DNA agarose gel electrophoresis experiment showed that DNA fragmentation, a marker of apoptosis, was detected in α-turmerone-treated MDA-MB-231 cells after 24 and 48 hours (Fig. 3C). However, the pattern was not detected in the solvent control (0 μg/ml).</p><!><p>To further confirm whether α-turmerone caused apoptosis in MDA-MB-231 cells, annexin V-FITC/PI staining experiment was performed. The occurrence of phosphatidylserine (PS) externalization onto the cell surface is a characteristic of apoptotic cells. The results showed that the proportion of early apoptosis cells (Annexin V-FITC positive and PI negative) increased with the concentrations of α-turmerone applied (Fig. 4).</p><!><p>For elucidating the molecular mechanisms of action of α-turmerone in MDA-MB-231 cells, the apoptosis signal transduction was investigated. Western blotting (Fig. 5) showed that α-turmerone significantly decreased the expression of cyclin D1, which is involved in cancer cell proliferaton. Caspases play a crucial role in initiation and execution of apoptosis (Nunez et al., 1998). Caspases-8 and -9 are known as initiator caspases, whereas caspase-3 is considered as executioner caspase (Budihardjo et al., 1999). Two distinct and well-known initiator caspases, caspase-8 for the death receptor-mediated and caspase-9 for the mitochondria-mediated pathways, have been shown to initiate apoptosis. Therefore, the expression levels of caspases-3, -8 and -9 as well as PARP, the substrate of caspase-3, were detected by Western blot analysis. As shown in Fig. 5, α-turmerone treatment significantly decreased the expression levels of pro-caspases-3 and -8. Proteolytic cleavage of mitochondria-mediated apoptotic cell death initiator pro-caspase-9 was observed in α-turmerone-treated MDA-MB-231 cells and the smaller fragments p35 and p37 were shown. Cleaved PARP was also observed in α-turmerone-treated MDA-MB-231 cells.</p><!><p>The mitochondria membrane potential change (Δψm) was measured by JC-1 staining. Mitochondria with normal Δψm concentrated JC-1 into aggregates (red fluorescence), while in depolarized mitochondria, JC-1 formed monomer (green fluorescene). The Δψm in cells treated with α-turmerone (25–45μg/ml) for 24hrs was decreased in a concentration-dependent manner. The percentage of depolarization cells was increased from 22.6% to 92.2% (Fig. 6). On the other hand, the Bcl-2 protein expression level in α-turmerone-treated cells was investigated. As shown in Fig. 5, treatment of MDA-MB-231 cells with α-turmerone slightly down-regulated Bcl-2 expression.</p><!><p>Different concentrations of α-turmerone (AL) and ar-turmerone (AR) were tested in cultured PBMC to examine their effects on the proliferation of PBMC. The turmerones were added to PBMC with polymyxin B in order to rule out the possible contamination by the endotoxin. As shown in Fig. 7, α-turmerone (AL, 10 or 15 μg/ml) and ar-turmerone (AR, 5 -15 μg/ml) were found to significantly stimulate the proliferation of PHA-activated PBMC after 72 hours treatment. However, turmerones added at 5 -15 μg/ml did not alter the proliferative response of resting PBMC (without mitogen) (data not shown). When ar-turmerone was added to PHA-activated PBMC at 10 μg/ml, the resulting proliferation responses were slightly higher than those stimulated by α-turmerone (Fig. 7A). After incubating the PBMC with α-turmerone (AL) and ar-turmerone (AR), the culture supernatants were collected. The concentrations of TNF-α, IL-2 and IFN-γ were determined by ELISA. After 24 hours stimulation in PHA-activated PBMC, the TNF-α production was significantly increased by α-turmerone (AL) and ar-turmerone (AR) at 10 and 15 μg/ml (Fig. 7C). The productions of IL-2 and IFN-γ in PHA-activated PBMC were higher than those in resting PBMC (IL-2: 5867.2 ± 3185.4 vs 103.4 ± 81.2 pg/ml; IFN-γ: 6158.2 ± 1954.1 vs 325.7 ±135.1 pg/ml). However, neither α-turmerone (AL) nor ar-turmerone (AR) further altered the productions of IL-2 and IFN-γ in PHA-activated PBMC (data not shown). On the other hand, in resting PBMC, ar-turmerone (AR) significantly increased TNF-α, IL-2 and IFN-γ productions (Fig. 7B, 7D and 7E).</p><!><p>The use of traditional medicine has expanded and health supplement consist of different types of herbal medicines have become very popular in Asia in recent years. Curcuma longa (CL) and its extracts, curcumin, are widely consumed as food additive and medicine, which are believed to possess anti-inflammatory, anti-oxidant, anti-cancer and immunomodulatory properties. However, scientific investigations to compare their pharmacological effects are seldom reported. In the present study, the anti-proliferative activities on human cancer cells and immunomodulating activities in human PBMC of the CL extracts were evaluated in vitro. All the extract fractions and isolated compounds were prepared by bioassay-guided fractionation process.</p><p>The present study demonstrated that the curcuminoids (curcumin, demethoxycurcumin and bisdemethoxycurcumin) and α-turmerone isolated form CL lipophilic fraction significantly inhibited proliferation of cancer cells in a dose-dependent manner, with curcumin being the most potent. Dose-response curves were obtained and demonstrated IC50 values of the curcuminoids ranged from 11.0 to 28.2 μg/ml in three cancer cell lines. On the contrary, the tested compounds (demethoxycurcumin, bisdemethoxycurcumin, α-turmerone and ar-turmerone, 3.125–25 μg/ml) did not induce any significant inhibition of cell proliferation in human normal skin fibroblasts, illustrating their selective cytotoxic effects on cancer cell lines (Table 2).</p><p>The mechanisms of action of the cytotoxic effects of curcumin in hepatoma cells (Cao et al., 2006, 2007) and breast cancer cells (Ramachandran et al., 2002, 2005; Mehta et al., 1997; Shao et al, 2002) have been demonstrated. They suggested that curcumin induced mitochrondrial and nuclear DNA damage in HepG2 cells and apoptosis in MCF-7 and MDA-MB-231 cells. These results were consistent with our present findings that all three tested curcuminoids significantly inhibited proliferation of HepG2, MCF-7 and MDA-MB-231 cells in a dose-dependent manner (Fig. 2). The anti-inflammatory activities (Funk et al., 2006; Lantz et al., 2005) and inhibitory activities on WT1 gene expression in leukemic cells (Anuchapreeda et al., 2008) of demethoxycurcumin and bisdemethoxycurcumin have been reported and compared with the activities of curcumin. Their results suggested that curcumin showed higher efficacy. However, a recent study demonstrated that demethoxycurcumin and bisdemethoxycurcumin showed higher antimetastasis potency than curcumin by the differentially down-regulation of extracellular matrix degradation enzymes (Yodkeeree et al., 2009). Our results also demonstrated that the inhibitory activities on cancer cells proliferation among curcuminoids have no significant difference. Moreover, the effects of α-turmerone and ar-turmerone on cancer cell proliferation were compared with the curcuminoids in the present study; the activities of the turmerones on cancer cells could be manifested.</p><p>The cell line MCF-7 is prototypes for estrogen-sensitive breast cancer cells, representative of relative early stages in breast cancer development. Contrary to MCF-7 cells, MDA-MB-231 cells are prototypes for estrogen-insensitive, highly-invasive breast cancer cells, representative of relative late stages in breast cancer development (Toillon et al., 2002). The anti-invasive effects of curcumin in estrogen receptor-negative MDA-MB-231 cells have been demonstrated by Shao et al. (Shao et al., 2002), nevertheless, the suppressive activities of other constituents of CL, such as α-turmerone, on this cell type are worth attention. Besides, apoptosis induced by ar-turmerone in leukemia cells has been reported (Aratanechemuge et al., 2002; Ji et al., 2004). To date, the molecular mechanisms of action of α-turmerone and its effects on breast cancer cells are not fully understood. Therefore, mechanistic study of α-turmerone on breast cancer cells were performed in the present study. Cell cycle analysis was performed to evaluate the effect of α-turmerone. The results showed the accumulation of sub-G1 cell population in a dose-dependent manner (Fig. 3A). DNA fragmentation and apoptotic cell population were shown in DNA laddering assay and Annexin V/PI staining assay, respectively. These implied that apoptosis occurred in the α-turmerone-treated MDA-MB-231 cells. The role of mitochondria in the intrinsic pathway of apoptosis was examined by detecting changes of fluorescent probe JC-1. Disruption of mitochondria membrane potential was observed in α-turmerone-treated cells (Fig. 6). This further activated the procaspase-9 cleavage (Fig. 5), which then activated the procaspase-3 and finally triggered apoptosis.</p><p>The inhibition of cell proliferation in both breast cancer cell lines by curcuminoids and α-turmerone was demonstrated in this study, implying that the potential use of turmeric oil and curcuminoids as health supplement for breast cancer treatment at different stages could be justified. Meanwhile, Lantz et al. suggested that the fraction containing curcuminoids and turmeric oils was more effective than the fraction of curcuminoids alone at inhibiting PGE2 production (Lantz et al., 2005).</p><p>Our results also showed that both α-turmerone and ar-turmerone exhibited stimulating effects on PBMC proliferation and cytokine production (Fig. 7). Since the activities of the turmerones were not affected by the addition of polymyxin B (10 μg/ml) (data not shown), the observed mitogenic effect was unlikely to be mediated by the endotoxin (lipopolysaccharide), which is a potent activator of B cells. Besides, α-turmerone and ar-turmerone were isolated from the lipophilic fraction (F1) of Curcuma longa crude ethanolic extract. The effects of crude ethanolic extract and the fractions (F1-F4, section 2.2) have been tested in PBMC and fraction F1 exhibited the most potent stimulatory effect on PBMC proliferation (data not shown) and this fraction was subjected to further column chromatography. Among the subfractions of F1, only the isolated compounds α-turmerone and ar-turmerone showed stimulating activities in PBMC. Therefore, the response of PBMC may be specific to such turmerones instead of other contaminants. On the other hand, the compounds curcumin, demethoxycurcumin and bisdemethoxycurcumin, which were isolated from F2 and F3, have not been tested in PBMC because fraction F2 and F3 showed neither stimulatory nor inhibitory effect on PBMC proliferation (data not shown). Moreover, a previous study has demonstrated that in fact curcumin inhibited PHA-induced proliferation and IL-2 production in PBMC (Yadav et al., 2005).</p><p>Along with these results, α-turmerone and ar-turmerone were shown to have immunostimulatory effects. Researches on turmeric oil have focused on the anti-inflammatory effects using macrophages. Nevertheless, little information is available about the immunomodulatory activities in other cell types of the immune system. Alpha-turmerone and ar-turmerone were shown for the first time to exert modulatory activities in human PBMC. Last but not least, the multiple targets abilities of CL extracts (including curcuminoids and turmerones) could be illustrated in the present study.</p><p>In conclusion, bioassay-guided fractionation of the rhizome of Curcuma longa led to the isolation of the curcuminoids (curcumin, demethoxycurcumin and bisdemethoxycurcumin), α-turmerone and ar-turmerone. Curcuminoids and α-turmerone significantly inhibited proliferation of cancer cells in a dose-dependent manner. Both α-turmerone and ar-turmerone stimulated PBMC proliferation and cytokine production in vitro. This is possibly the first report on the anti-proliferative effect of α-turmerone and immunomodulatory activity exerted by ar-turmerone. These findings revealed the potential use of Curcuma longa extracts (including curcuminoids and volatile oil) as chemopreventive/antitumor agent.</p><!><p>Chemical structures of α-turmerone, ar-turmerone, curcumin, demethoxycurcumin and bisdemethoxycurcumin.</p><p>Effects of curcuminoids and turmerones on cell viabilities of (A&D) HepG2, (B&E) MCF-7 and (C&F) MDA-MB-231 cells. Cells were treated with increasing concentrations of curcumin (open square), demethoxycurcumin (closed triangle), bisdemethoxycurcumin (open circle), α-turmerone (closed circle) and ar-turmerone (x) for 48 hours, and cell viability was determined by the MTT assay. Results are expressed as percentages of MTT absorbance with respect to the untreated vehicle control wells (mean ± S.D. of 5–6 independent experiments with 6 wells each).</p><p>Effects of α-turmerone on apoptosis induction on MDA-MB-231 cells. (A) Cell cycle analysis of MDA-MB-231 cells treated with vehicle or α-turmerone for 48 hours. After treatment, cells were analyzed by flow cytometry (n=3). (B) The fractions of cells in each cell cycle phase were summarized in the table. (C) Agarose gel of electrophoresis of genomic DNA of MDA-MB-231 cells treated with different concentrations of α-turmerone for 24 or 48 hours (n=3). M represents 100 bp DNA marker.</p><p>α-turmerone-induced phosphatidylserine externalization on MDA-MB-231 cells. Cells were stained with Annexin-V FITC and PI and detected by flow cytometry. The lower right quadrant (Annexin-V+ PI−) represents early apoptosis, whereas upper right quadrant (Annexin-V+ PI+) represents late apoptosis. All results are representative of three independent experiments.</p><p>Effects of α-turmerone on cyclin D1, Bcl-2, PARP, procaspase-3, -8 and -9 expression levels in MDA-MB-231 cells. Total proteins of α-turmerone-treated (40 or 60μg/ml) cells was isolated for testing the expression by Western blotting. All results are representative of three independent experiments.</p><p>α-turmerone caused mitochondrial transmembrane depolarization in MDA-MB-231 cells. Cells were stained with JC-1 fluoresence dye and changes in Δψm were determined by flow cytometry using green (FITC) and red (PI) channels. All results are representative of three independent experiments.</p><p>Proliferative response and cytokine productions of α-turmerone (AL) and ar-turmerone (AR) treated PBMC. (A) PHA-activated PBMC were seeded in 96-well plates and incubated with 5, 10 or 15 μg/ml of α-turmerone (AL) or ar-turmerone (AR) in culture medium for 72h, and expressed as the mean % ratio of count per minute in treated and untreated control cells of 8 individual blood donors. The resting PBMC (B, D, E) and PHA-activated PBMC (C) were seeded in 96-well plates and incubated with 5, 10 or 15 μg/ml of α-turmerone (AL) or ar-turmerone (AR) for 24 hours. Culture supernatants were collected and the cytokine concentrations were specifically determined by ELISA. Results were expressed as mean concentration + SD of 8 blood samples with four wells each. Differences between the treated and untreated control group were determined by Student's unpaired t-test. * P < 0.05, **P < 0.01 as compared to the control group.</p><p>1H NMR and 13C NMR spectral data of α-turmerone and ar-turmerone</p><p>1H NMR and 13C NMR spectral data of curcumin, demethoxycurcumin and bisdemethoxycurcumin</p><p>Concentrations producing 50% growth inhibition (IC50) of curcuminoids and turmerones isolated from Curcuma longa on cancer cell lines and normal skin fibroblasts.</p>

PubMed Author Manuscript

Synthetic Inositol Phosphate Analogs Reveal that PPIP5K2 Has a Surface-Mounted Substrate Capture Site that Is a Target for Drug Discovery

SummaryDiphosphoinositol pentakisphosphate kinase 2 (PPIP5K2) is one of the mammalian PPIP5K isoforms responsible for synthesis of diphosphoinositol polyphosphates (inositol pyrophosphates; PP-InsPs), regulatory molecules that function at the interface of cell signaling and organismic homeostasis. The development of drugs that inhibit PPIP5K2 could have both experimental and therapeutic applications. Here, we describe a synthetic strategy for producing naturally occurring 5-PP-InsP4, as well as several inositol polyphosphate analogs, and we study their interactions with PPIP5K2 using biochemical and structural approaches. These experiments uncover an additional ligand-binding site on the surface of PPIP5K2, adjacent to the catalytic pocket. This site facilitates substrate capture from the bulk phase, prior to transfer into the catalytic pocket. In addition to demonstrating a “catch-and-pass” reaction mechanism in a small molecule kinase, we demonstrate that binding of our analogs to the substrate capture site inhibits PPIP5K2. This work suggests that the substrate-binding site offers new opportunities for targeted drug design.

synthetic_inositol_phosphate_analogs_reveal_that_ppip5k2_has_a_surface-mounted_substrate_capture_sit

Introduction<!>Stimulation of the ATPase Activity of PPIP5K2KD by 5-PA-InsP5 and 2-O-Bn-5-PA-InsP4<!>Design and Synthesis of Inositol Phosphate and Diphosphoinositol Phosphate Analogs<!>Stimulation of the ATPase Activity of PPIP5K2KD by Synthetic Inositol Phosphate Analogs<!>Inhibition of PPIP5K2KD Catalytic Activity by Inositol Phosphate Analogs<!>Structural Analysis Reveals that PPIP5K2KD Has Two Adjacent Ligand-Binding Sites<!>Site-Directed Mutagenesis of the Substrate Capture Site<!>Significance<!>Protein Expression, Purification, Crystallization, and Structure Determination<!>Enzyme Assays<!>Statistical Analysis<!>Author Contributions<!>Accession Numbers<!>Supplemental Information

<p>The process of signal transduction that governs many cellular activities frequently relies upon evolutionarily conserved families of small, regulatory molecules. Among them are the diphosphoinositol polyphosphates (inositol pyrophosphates: 5-PP-InsP4, 1-PP-InsP5 [1-InsP7], 5-PP-InsP5 [5-InsP7], and 1,5-[PP]2-InsP4 [InsP8]; Figure 1), in which six to eight phosphate groups are crammed around the six-carbon inositol ring. These high-energy molecules are synthesized by two distinct classes of kinases, IP6Ks and PPIP5Ks. The IP6Ks add the 5-diphosphate group (Draskovic et al., 2008); mammals express three IP6K isoforms (Thomas and Potter, 2014). The PPIP5Ks synthesize the 1-diphosphate (Wang et al., 2012); there are two isoforms in mammals (Thomas and Potter, 2014). Interest in this field has recently been heightened by demonstrations that diphosphoinositol polyphosphates operate at the interface of cell signaling and organismic homeostasis (Choi et al., 2005; Szijgyarto et al., 2011; Shears, 2009; Illies et al., 2007; Chakraborty et al., 2010; Pulloor et al., 2014). Here, a dynamic balance between the activities of IP6Ks and PPIP5Ks is of particular significance. For example, the synthesis of 5-PP-InsP5 by IP6Ks inhibits the PtdIns(3,4,5)P3/PDK1/AKT/mechanistic target of rapamycin (mTOR) cascade (Chakraborty et al., 2010) that controls cell growth and metabolism in response to changes in levels of nutrients, growth factors, and bioenergetic status (Benjamin et al., 2011). This inhibitory action of 5-PP-InsP5 is reversed through its further phosphorylation by the PPIP5Ks (Gokhale et al., 2013). There may be therapeutic value in inhibiting PPIP5K activity to elevate 5-PP-InsP5 levels and attenuate the mTOR pathway, which is hyperactivated in 70% of human tumors, contributing to the derangement of cell growth and metabolism that accompanies cancer development and progression (Benjamin et al., 2011). We recently published proof-of-principle of the latter idea by demonstrating that AKT phosphorylation in myoblasts is inhibited when PPIP5K1 expression is "knocked-down" (Gokhale et al., 2013). It is just such therapeutic motives that frequently drive the development of drugs that can specifically target kinases such as PPIP5Ks. Candidate molecules may be rationally designed when information on protein structure is available. To this end, we recently solved the structure of the N-terminal kinase domain of PPIP5K2 (PPIP5K2KD) in complex with natural substrate within the catalytic site (Wang et al., 2012). However, the architecture of the active site exhibits substantial geometric and electrostatic constraints that raise challenges for the design of an effective yet specific inhibitor.</p><p>In the current study, we set out to prepare substrate analogs that might modify PPIP5K2 activity. The synthesis of analogs of diphosphoinositol polyphosphates presents particular technical challenges due to the reactive nature of the diphosphate group and the protected diphosphate intermediates (Best et al., 2010). The high negative charge density of these materials also presents purification problems (Capolicchio et al., 2013). Although several of the naturally occurring diphosphoinositol polyphosphates have been synthesized (Albert et al., 1997; Best et al., 2010; Wu et al., 2013; Capolicchio et al., 2013), the preparation of useful analogs has only recently been accomplished (Riley et al., 2012; Wu et al., 2013). In the latter studies, analogs of 5-PP-InsP4 and 5-PP-InsP5 were synthesized in which the diphosphate groups were replaced with metabolically stabilized phosphonoacetate (PA) or methylenebisphosphonate (PCP) groups. In the current study, we describe the synthesis of a series of diphosphoinositol polyphosphate analogs. We demonstrate how we used these reagents to gain insight into a previously described (Weaver et al., 2013) substrate-stimulated ATPase activity of PPIP5K2KD. These experiments also led us to uncover a second ligand-binding site in PPIP5K2KD that performs an important aspect of the catalytic cycle by enhancing capture of substrate from the bulk phase.</p><!><p>We recently reported that PPIP5K2KD exhibits an unusual, non-productive, substrate-stimulated ATPase activity (e.g., we observed a 2- to 3-fold activation by 25 μM of either Ins(1,3,4,5,6)P5 or InsP6; Figure 2A; Weaver et al., 2013). We now report that 25 μM of either of two previously described analogs of diphosphoinositol polyphosphates (Riley et al., 2012) also stimulate ATP hydrolysis 5-fold by 5-O-α-phosphonoacetyl-myo-inositol 1,2,3,4,6-pentakisphosphate (5-PA-InsP5 [1]), and 9-fold by 2-O-benzyl-5-O-α-phosphonoacetyl-myo-inositol 1,3,4,6-tetrakisphosphate (2-O-Bn-5-PA-InsP4 [2]; Figures 2A and 2B). In view of the precise geometric and electrostatic specificity constraints within the active site (Wang et al., 2012), we did not anticipate that it could accommodate 2-O-Bn-5-PA-InsP4, which sports a bulky hydrophobic group. It was therefore unexpected that 2-O-Bn-5-PA-InsP4 should be a more effective activator of ATPase activity than the two natural substrates, InsP6 (Figure 2A) and PP-InsP5 (Weaver et al., 2013). We sought further information on this phenomenon.</p><!><p>To increase insight into the origin of ligand-stimulated ATPase activity of PPIP5K2KD, we explored the contributions of the benzyl (Bn) and α-phosphonoacetyl (PA) groups in 2-O-Bn-5-PA-InsP4 (2). Our approach was to synthesize a series of analogs (Figure 2B) that exchanged Bn for either OH, monophosphate, or alternative hydrophobes, whereas the PA group was replaced by either a monophosphate or a diphosphate. For analogs containing acyl groups at O-2, the appropriate synthetic precursors are orthoesters of myo-inositol, because acid hydrolysis of these compounds gives rise to 2-O-acyl esters with very high regioselectivity (Godage et al., 2006, 2013). Thus, 2-O-benzoyl-InsP5 (4) and 2-O-butanoyl-InsP5 (5; Figures 2B and 3A) were synthesized from myo-inositol orthobenzoate (11) and myo-inositol orthobutanoate (12), respectively. Acid hydrolysis of 11 and 12 (Godage et al., 2013) gave pentaols 13 and 14, respectively, and subsequent phosphorylation and deprotection provided 2-O-benzoyl-InsP5 (4) and 2-O-butanoyl-InsP5 (5).</p><p>For analogs containing alkyl ethers or hydroxyl groups at C-2, synthetic routes proceeding from the symmetrical butane-2,3-diacetal (BDA) protected diol 17 are expedient, as in our recent synthesis of 2-O-(2-aminoethyl)-InsP5 (6; Riley et al., 2014). The remaining compounds 7–10 were also synthesized from diol 17 using a unified approach (Figures 2B and 3B). Thus, regioselective benzylation of the axial hydroxyl group of diol 17 gave 2-O-benzyl ether 18, and removal of BDA protecting groups gave pentaol 19. Phosphorylation of 19 afforded the symmetrical pentakisphosphate 20 and cleavage of phosphate protection gave 2-O-Bn-InsP5 (7) in 59% overall yield from 17.</p><p>For the synthesis of 2-O-Bn-5-PP-InsP4 (8) and 5-PP-InsP4 (9), the C-5 hydroxyl group of 18 was first phosphitylated with bis(cyanoethyl)(N,N-diisopropylamino)phosphine, followed by oxidation to provide the protected monophosphate 21. Removal of BDA groups followed by phosphorylation afforded the fully protected pentakisphosphate 23 in high yield (Figure 3B). We constructed the diphosphate group by modifying a recently described methodology (Capolicchio et al., 2013); compound 23 was treated with DBU (1,8-diazabicyclo[5.4.0]undec-7-ene) and BSTFA (N,O-bis(trimethylsilyl)trifluoroacetamide) to remove both cyanoethyl protecting groups on the 5-phosphate, followed by methanolysis of the temporary trimethylsilyl protection to give the phosphate monoester at the C-5 position. Phosphitylation then afforded a mixed P(III)-P(V) intermediate, which was oxidized to produce 24 (Figure 3B). We next prepared 2-O-Bn-5-PP-InsP4 (8) from crude 24 (which contained DBU), by hydrogenolysis over palladium hydroxide, which removed the benzyl esters from the phosphates; amines (either in DBU or in triethylamine; Riley et al., 2012) inhibit hydrogenolytic cleavage of the 2-O-benzyl ether. Note that 2-O-Bn-5-PP-InsP4 (8) is a synthetic analog of an inositol pyrophosphate that retains the actual diphosphate group.</p><p>Hydrogenolysis of purified 24 from which DBU was removed yielded 5-PP-InsP4 (9), in 51% overall yield from 23 (Figure 3B). This molecule is a naturally occurring diphosphoinositol polyphosphate that features an unphosphorylated 2-OH group (Figure 1); 5-PP-InsP4 has been implicated in telomere maintenance (York et al., 2005; Saiardi et al., 2005). Finally, for the synthesis of 2,5-di-O-Bn-InsP4 (10), diol 17 was benzylated followed by cleavage of BDA groups. Subsequent phosphorylation and deprotection of cyanoethyl phosphate esters gave (10) in 63% overall yield from 17 (Figure 3B).</p><!><p>As discussed above (see also Figure 2A), 2-O-Bn-5-PA-InsP4 (2) strongly activated nonproductive ATPase activity of PPIP5K2KD. The degree of stimulation was dramatically reduced when the latter's benzyl group was replaced with a hydroxyl group to give 5-PA-InsP4 (3, the PA analog of 5-PP-InsP4; Riley et al., 2012; Figure 2). Furthermore, 5-PP-InsP4 (9) stimulated ATPase activity at about half the rate shown by 2-O-Bn-5-PP-InsP4 (8). Taken together, these data (Figure 2) lead to an unexpected conclusion that the ligand's benzyl group makes an important contribution to its association with PPIP5K2KD.</p><p>In 2-O-Bn-5-PP-InsP4 (8), the natural and more negatively charged diphosphate (PP) group replaces the PA group at O-5 in 2-O-Bn-5-PA-InsP4 (2). We found that 2-O-Bn-5-PP-InsP4 (8) was slightly (20%) more effective at stimulating ATPase activity than was 2-O-Bn-5-PA-IP4 (2; Figure 2). Nevertheless, the PA group still contributes to efficacy more than a monophosphate group because 2-O-Bn-InsP5 (7) was less effective than was 2-O-Bn-5-PA-InsP4 (2; Figure 2).</p><p>Even in 2-O-Bn-InsP5 (7), the 2-O-benzyl ether has an enhancing effect, roughly doubling the rate of ATP hydrolysis that is induced by InsP6, which has a phosphate group at C-2 (Figure 2). To investigate the effect of hydrophobes other than benzyl at O-2, we compared 2-O-Bn-InsP5 (7) with 2-O-Bz-InsP5 (4) and 2-O-But-InsP5 (5). The change from the 2-O-benzyl ether to the structurally related 2-O-benzoyl ester led to only a minimal reduction in ATPase-stimulating activity, while substitution with a 2-O-butanoyl ester gave a sharp reduction in activity (Figure 2). Furthermore, the analog with a positively charged aminoethyl group at O-2 in (2-O-AminoEt-InsP5 [6]) did not stimulate ATPase activity (Figure 2).</p><p>We next tested one further analog in which the diphosphate group in 2-O-Bn-5-PP-InsP4 (8) was replaced with a second benzyl group (i.e., 2,5-di-O-Bn-InsP4; 10). This compound elicited the highest degree of ATPase activity among all of the analogs that we have synthesized (Figure 2). This was initially a puzzling conclusion, as structural considerations indicate that 2,5-di-O-Bn-InsP4 (10) would at best be a poor ligand for the active site.</p><p>Several of the compounds described above were selected for detailed dose-response curves; we used the six compounds that yielded the greatest stimulation of ATPase activity, with the exception of 2-O-Bz-InsP5 (4), which in any case showed similar efficacy to 2-O-Bn-InsP5 (7; Figure 4A). The most potent of this series of compounds was 2,5-di-O-Bn-InsP4 (10, half-maximal effective concentration [EC50] = 340 nM; Figure 4A). In general, the rank order of the EC50 values for these compounds approximated the rank order of their maximal effects, with the notable exception of 2-O-Bn-5-PP-InsP4 (8; Figure 4A).</p><!><p>The six compounds that were selected for the dose-response curves in the assays of ATP hydrolysis (Figure 4A) were now investigated for their effects upon diphosphoinositol polyphosphate metabolism by PPIP5K2KD (Figure 4B). We used a high-throughput reverse-kinase assay that records ATP generated from ADP during the dephosphorylation of 100 nM 1,5-[PP]2-InsP4 (Riley et al., 2012; Weaver et al., 2013). Each of the tested analogs inhibited [PP]2-InsP4 metabolism by PPIP5K2KD (Figure 4B). The two most potent of these analogs were 5-PA-InsP5 (1) and 2,5-di-O-Bn-InsP4 (10) (Figure 4B). The inhibitory effect of 5-PA-InsP5 (1) was not in itself surprising becasue we have already demonstrated it to be a PPIP5K substrate (Riley et al., 2012). However, the inhibition by 2,5-di-O-Bn-InsP4 (10) was more unexpected because due to its bulk, and its less negative charge compared to physiological substrates, we predict it to be a poor ligand for the active site of the highly specific PPIP5K2KD (Wang et al., 2012). Nevertheless, its ability to inhibit reverse-kinase activity was directly confirmed by high-performance liquid chromatography (HPLC) analysis of [PP]2-[3H]InsP4 dephosphorylation (Figures S1A–S1C available online). We confirmed that 2,5-di-O-Bn-InsP4 (10) also inhibited PPIP5K2KD in the "forward," kinase direction using InsP6, a physiologically relevant substrate (Figures S1D and S1E). A Dixon plot demonstrated that inhibition by 2,5-di-O-Bn-InsP4 (10) was competitive in nature (Ki = 240 nM; Figure 4C), arguing that its actions are not allosteric in nature.</p><p>The rank order of potencies (as half-maximal inhibitory concentration [IC50] values) with which this group of molecules inhibited [PP]2-InsP4 dephosphorylation by PPIP5K2KD (Figure 4B) is different in two key respects from the rank-order of efficacy (EC50 values) for their separate stimulation of ATPase activity (Figures 4A and 4B). First, 2-O-Bn-InsP5 (7) is 4-fold less potent than 2-O-Bn-5-PA-InsP4 (2) at stimulating ATPase activity (Figure 4A), whereas the two analogs are equally efficient at inhibiting [PP]2-InsP4 metabolism (Figure 4B). Second, 5-PA-InsP5 (1), the weakest activator of ATPase (Figure 4A), is the most potent inhibitor of [PP]2-InsP4 dephosphorylation (Figure 4B). These are observations that suggest there is some uncoupling of inositol phosphate turnover from the ATPase activity.</p><!><p>We have previously published an atomic-level description of both 5-PP-InsP5 and InsP6 substrates bound into the catalytic pocket of PPIP5K2KD (Wang et al., 2012). In a separate study, we (Riley et al., 2012) soaked 2 mM 5-PA-InsP5 (1) into crystals of PPIP5K2KD, and reported that this ligand occupies the active site in the same orientation as the natural substrate. At the time of that earlier study, we contoured the simulated annealing omit map at 2 σ, and observed some additional but uninterpretable electron density at the entrance of the active site (data not shown). For the current study, we increased the soaking concentration of 5-PA-InsP5 (1) to 10 mM, and have now unequivocally detected an additional ligand-binding site, at 1.7 Å resolution, that is located near the surface of PPIP5K2KD, at the entrance to the catalytic center (Figures 5A and 5D; Table S1). This surface-mounted ligand-binding site is too remote from ATP to permit ligand phosphorylation. It should be noted that both this site and the catalytic pocket cannot be occupied simultaneously due to steric clashing (Figures 5A and S2A). That is, these particular structural data arise from a mixture of crystal complexes in which 5-PA-InsP5 (1) is separately bound to either of the two sites.</p><p>The stimulation of ATPase activity of PPIP5K2KD by either 2-O-Bn-5-PA-InsP4 (2) or 2,5-di-O-Bn-InsP4 (10) indicates that these compounds also interact with PPIP5K2KD, so we soaked 5–10 mM of either 2-O-Bn-5-PA-InsP4 (2) or 2,5-di-O-Bn-InsP4 (10) into the PPIP5K2KD crystals. X-ray analysis revealed that both of these analogs were exclusively associated with the second ligand-binding site (Figures 5B, 5C, 5E, and 5F; Table S1). Neither of these analogs was found to occupy the catalytic site. These data suggest that the stimulation of ATPase activity by either natural substrates or their analogs is associated with their occupation of the second binding site, not the catalytic site. That explanation is in turn consistent with other data (Figures 4A and 4B, and see above) that suggest ligand-stimulated ATPase activity is uncoupled from the inositol phosphate kinase activity.</p><p>The architecture of this second ligand-binding site is represented by a deep cleft, which is walled on one side by K53, K54, and K103. The opposite face is formed from R213, and a loop created from residues E192 to H194 (Figure 5). In this binding site, the shared groups in 5-PA-InsP5 (1) and 2-O-Bn-5-PA-InsP4 (2) exhibit almost identical conformations (Figure 6A). As for 2,5-di-O-Bn-InsP4 (10) this molecule is rotated ∼20° clockwise (Figure 6B). For both 2-O-Bn-5-PA-InsP4 (2) and 2,5-di-O-Bn-InsP4 (10), their phosphate groups at C-3 and C-4 would clash with the positions that the C-4 and C-3 phosphates of InsP6/5-PP-InsP5 normally occupy in the active site (Figures S2A–S2C). This phenomenon presumably contributes to the potency of both 2-O-Bn-5-PA-InsP4 (2) and 2,5-di-O-Bn-InsP4 (10) as inhibitors of inositol phosphate turnover by PPIP5K2KD (Figures 4B and 4C). These structural considerations explain how both compounds can inhibit inositol phosphate kinase activity without occupying the catalytic site.</p><p>The association of 2-O-Bn-5-PA-InsP4 (2) and 2,5-di-O-Bn-InsP4 (10) with the second binding site is facilitated by their phosphate groups at C-3 and C-4 making multiple electrostatic interactions with K53, K54, and R213 (Figures 5, S2E, and S2F). The terminal phosphonate of 2-O-Bn-5-PA-InsP4 (2) also forms polar contacts with the backbone of K103 and the H194 side chains (Figures 5 and S2E). Additionally, the 5-O-benzyl group in 2,5-di-O-Bn-InsP4 (10) has van der Waals interactions with the side chains of H101 and E192 (Figures 5 and S2F). The 2-O-benzyl group in this analog is disordered, indicative of its mobility. Nevertheless, our catalytic data (Figure 4) indicate that this benzyl group makes a significant contribution to ligand potency.</p><p>The data that we have obtained allow important conclusions to be drawn concerning the likely significance of the second binding site. For example, we can exclude the possibility that, as is the case with certain enzymes (Reed et al., 2010), the role of the second, noncatalytic substrate-binding site is to regulate enzyme activity by substrate inhibition; PPIP5K2KD does not exhibit that property (Weaver et al., 2013). Other enzymes use noncatalytic ligand-binding sites to promote enzyme activation allosterically (Grant, 2012). That also seems unlikely to be the case for PPIP5K2KD, because our inositol phosphate analogs do not enhance kinase activity, but inhibit it instead (Figure 4B; Riley et al., 2012). The nature of the inhibition was competitive (Figure 4C), further arguing against an allosteric effect. Instead, the proximity to the active site of the second ligand-binding site (Figure 4) suggests that the latter plays a role in catalysis of natural substrates, by facilitating their capture from the bulk phase. A precedent for such a phenomenon has been reported for microbial anthranilate phosphoribosyltransferases (Castell et al., 2013; Marino et al., 2006); anthranilate substrate is first captured at a surface-mounted binding-site, before delivery to a proximal catalytic site (Castell et al., 2013). A similar substrate transfer is feasible in PPIP5K2KD, and our structural data provide some atomic-level insight into such a phenomenon. For example, we presume that delivery of natural substrate is efficient, so that substrate only occupies the capture site transiently; this may explain why we could not detect substrate in the capture site in our crystal structures. We found that such substrate transfer within PPIP5K2KD requires that the inositol ring be flipped and rotated by ∼100° (Figures 6D and 6E). This motion is presumably facilitated by the flexible amino-acid side chains that comprise the active site (Wang et al., 2012). Such conformational dynamics can reinforce catalytic specificity (Herschlag, 1988), which is indeed a notable feature of PPIP5K2KD (Wang et al., 2012).</p><!><p>Mutagenesis offers a valuable means of pursuing conclusions drawn from structural analysis. The loop that is formed from residues A191 to H194 screens the catalytic site from ligands that are associated with the second binding site (Figures 5, 7A, and S3). The influence of the carbonyl oxygen of A191 may be indirect, through a hydrogen bond with a water molecule that in turn coordinates with an Mg2+ ion that interacts with the nucleotide's β-phosphate moiety (Figure 7A). The role of H194 seems more direct; it can form a hydrogen bond with the oxygen atom of the nucleotide's β-phosphate (Figures 7A and S3). This could stabilize the transition state, or have some other catalytic role. These interactions could be relevant to both kinase and ATPase activities of PPIP5K2KD. Consistent with these ideas, an H194A mutant did not exhibit any detectable ATPase activity, either in the presence or absence of 2-O-Bn-5-PA-InsP4 (2) or 2,5-di-O-Bn-InsP4 (10), and its inositol phosphate kinase activity was reduced 80-fold compared to that of the wild-type enzyme (data not shown).</p><p>As discussed above, our structural and biochemical data led us to hypothesize that the stimulation of ATPase activity by either natural substrates or their analogs is associated with their occupation of the substrate capture site, not the catalytic site. We attempted to consolidate this idea by selecting for mutation residues that might separate ATPase activity from inositol phosphate kinase activity. For example, our X-ray data (Figure 5) indicate that K54 and R213 interact with ligand at the substrate capture site. However, previous mutagenic work has shown that both K54 and R213 also contribute directly to inositol phosphate kinase activity (Wang et al., 2012). Nevertheless, we did observe that K54A and R213A mutants exhibited a substantially impaired degree of stimulation of ATPase activity by either 2-O-Bn-5-PA-InsP4 (2) or 2,5-di-O-Bn-InsP4 (10) (Figures 7B and 7C). These mutagenic data are consistent with the idea that K54 and R213 participate in substrate capture.</p><p>The side chain of K103 is also suggested by our structural data to interact with ligand that is bound to the capture site (Figures 5 and S2). Nevertheless, a K103A mutant showed only a slight reduction in 2-O-Bn-5-PA-InsP4 (2)-activated and 2,5-di-O-Bn-InsP4 (10)-activated ATPase activity, compared to wild-type enzyme (Figures 7B and 7C). Likewise, the InsP6 kinase activity of the K103A mutant (71 ± 4 nmol/mg protein/min, n = 3) was similar to that of wild-type enzyme (61 ± 4 nmol/mg protein/min).</p><p>We also mutated E192. This residue is present in the loop between the two ligand-binding sites (Figure 7A). It is too distant from ATP and the catalytic site to influence either directly, and electrostatic repulsion would prevent it from interacting with negatively charged groups on substrates located in the second binding site. Nevertheless, E192 is evolutionarily conserved in PPIP5Ks from mammals to yeast (not shown), suggestive of functional significance. We investigated the significance of E192 by preparing E192G and E192Q mutations that we posited would eliminate electrostatic repulsion between the amino acid side chain and phosphorylated ligands, enhancing ligand binding to the substrate capture site. We further hypothesized that such an effect might reduce the efficiency of transfer of substrate to the catalytic site, which in turn would lead to a reduction in inositol phosphate kinase activity. Indeed, we found that these E192G and E192Q mutations reduced the rate of InsP6 phosphorylation by 12- and 18-fold respectively, compared to wild-type PIP5K2KD (Figure 7D). It was of further note that neither of these particular mutations altered the rate of nonproductive, ligand-stimulated ATPase activity elicited by either 2-O-Bn-5-PA-InsP4 (2), 2-O-Bn-5-PP-InsP4 (8), or 2,5-di-O-Bn-InsP4 (10; Figure 7E), confirming the uncoupling of this aspect of PPIP5K2KD activity from its kinase activity.</p><p>We have shown that 2,5-di-O-Bn-InsP4 (10) binds to a substrate-capture site on the surface of PPIP5K2KD (Figure 5), thereby inhibiting the enzyme's catalytic activity (Figure 4). This particular ligand-binding site may be exploitable as a pharmacological target. We therefore examined the specificity of 2,5-di-O-Bn-InsP4 (10) by investigating if it interacted with IP6K2, a member of a different class of kinases that also phosphorylate InsP6 and synthesize diphosphoinositol polyphosphates (see Figure 1). Because the affinity of IP6K2 for InsP6 (430 nM; Saiardi et al., 2000) is very similar to that for InsP6 phosphorylation by PPIP5K2KD (390 nM; Weaver et al., 2013), an identical substrate concentration ([InsP6] = 500 nM) was used to assay both enzymes. As shown in Figure 4C, 2,5-di-O-Bn-InsP4 (10) inhibits PPIP5K2KD with an IC50 value of 1 μM (Figure 4C). This analog was found to be a much weaker inhibitor of IP6K2 activity (IC50 = 63 μM; data not shown). Thus, 2,5-di-O-Bn-InsP4 (10) may be a useful lead molecule for future development of a drug that can specifically inhibit PPIP5Ks and not IP6Ks, even though both groups of enzymes phosphorylate InsP6.</p><p>In conclusion, our studies have uncovered a "catch-and-pass" aspect to the catalytic cycle of PPIP5K2KD. Substrate (either InsP6 or 5-PP-InsP5) first associates with a ligand-binding site on the surface of the kinase. We propose that this phenomenon enhances kinase activity by improving substrate capture from the bulk phase; substrate is then delivered into the catalytic pocket. We are not aware of any other examples of a substrate-capture mechanism in a small-molecule kinase. Indeed, the only example that we have found in the literature for a dedicated substrate capture site on any enzyme is that observed for certain microbial anthranilate phosphoribosyl-transferases (Castell et al., 2013; Marino et al., 2006). Our structural, biochemical, and mutagenic data have also led us to conclude that the stimulation of ATPase activity of PPIP5K by inositol phosphate analogs is associated with their occupation of the substrate capture site, not the catalytic site. Moreover, the previously puzzling observation (Weaver et al., 2013) that a degree of nonproductive ATP hydrolysis is also stimulated by natural substrate, can now be rationalized as a consequence of its interaction with the capture site. Furthermore, the fact that InsP6-stimulated ATPase activity is abolished by an R213A mutation (Weaver et al., 2013) can now be viewed as resulting from an impairment to substrate occupation of the capture site.</p><p>Our atomic-level description of the substrate-capture site on PPIP5K2KD indicates its structural determinants of specificity differ substantially from those of the catalytic site. We have further shown that synthesis of diphosphoinositol polyphosphates by PPIP5K2KD is inhibited by ligands that bind to the capture site but not the catalytic site. These findings in turn raise the possibility that there may be cellular constituents that might inhibit catalytic activity by binding to this site. Finally, the substrate capture site offers a selective target for the purposes of rational drug design, including screening in silico (Kitchen et al., 2004), that is free from many of the usual specificity constraints within the catalytic site that typically complicate pharmacological targeting of small-molecule kinases. It may also be possible to design a ligand that occupies both sites.</p><!><p>PPIP5K1 and PPIP5K2 are small-molecule kinases that synthesize diphosphoinositol polyphosphates, which function at the interface of cell signaling and organismic homeostasis. Synthetic chemical modulators of cell-signaling enzymes such as the PPIP5Ks can provide valuable insight into catalytic mechanisms, they decipher the biological roles of the enzymes in situ, and they generate leads for therapeutic drug development. Substrate analogs offer one approach for the preparation of such chemical reagents. However, the architecture of the active site of PPIP5Ks, as revealed by recent X-ray analysis, has identified substantial geometric and electrostatic constraints that limit the options for designing a substrate analog that might be an effective modulator. In the current study, we describe the chemical synthesis of both the naturally occurring 5-PP-InsP4 and a family of inositol polyphosphate analogs that include molecules with hydrophobes at the 2- and/or 5-positions. Importantly, these molecules have led us to characterize a second, less constrained substrate-binding site on the surface of PPIP5K2, adjacent to the catalytic pocket. We provide an atomic-level description of this surface-mounted site and describe its role in the catalytic cycle in capturing substrate from the bulk phase. With the assistance of site-directed mutagenesis of this site, we show that its occupation is associated with an unusual, ligand-activated ATPase activity. This considerable amount of information had not been accessible with our experiments with natural substrates and represents a success for our chemical biology approach. In addition to adding to the repertoire of catalytic specializations of the PPIP5Ks, our structural and functional characterization of this ligand-binding site offers a promising target for drug development; to this end, 2,5-di-O-Bn-InsP4 is a significant lead compound.</p><!><p>The kinase domain of human PPIP5K2 (PPIP5K2KD; residues 1–366) and the N-terminally truncated domain used for the crystallography studies (residues 41–366) were subcloned, expressed, and purified as before. The latter was crystallized by hanging drop vapor diffusion against a well buffer of 12% (w/v) PEG 3350, 20 mM MgCl2, 0.1 M HEPES, pH 7.0, 1 mM AMP-PNP, and 2 mM CdCl2 at 4°C. The crystals were then soaked with 5–10 mM compounds in a stabilizing buffer containing 22% (w/v) PEG 3350, 10 mM MgCl2, 0.1 M sodium acetate, pH 5.2 or 7.0, at 4°C for 3 days. Cryosolvent was prepared by adding 33% ethylene glycol into the soaking buffer. Diffraction data were collected using APS beamlines 22-BM and 22-ID. All data were processed with the program HKL2000. The structure was determined using rigid body and direct Fourier synthesis, and refined with the equivalent and expanded test sets. The structure was further manually rebuilt with COOT and refined with PHENIX and REFMAC from the CCP4 package. Ligand topology and parameter files were prepared using the PRODRG server. The molecular graphics representations were prepared with the program PyMol (Schrödinger). The 2D ligand-protein interaction diagrams were generated by LigPlot+.</p><!><p>HPLC was used to assay [3H]InsP6 phosphorylation by human PPIP5K2KD or human IP6K2 in 100 μl buffer containing 50 mM KCl, 20 mM HEPES pH 7.0, 7 mM MgSO4, 5 mM ATP, and 1 mM EDTA (for IP6K2, [MgSO4] was 12 mM and [ATP] was 10 mM). The initial InsP6 concentration was either 500 nM, 70 nM, or 40 nM. Various concentrations of 2,5-di-O-Bn-InsP4 were present as indicated. Radioactivity was assessed using an in-line Flo1 detector. The ADP-driven dephosphorylation of 100 nM 1,5-[PP]2-InsP4 by PPIP5K2KD was usually determined by a luminescence-based assay of ATP accumulation (Riley et al., 2012; Weaver et al., 2013). The IC50 values were determined using GraphPad Prism v6.02 (n ≥ 3). In some experiments, 1,5-[PP]2-[3H]InsP4 dephosphorylation by PPIP5K2KD was assayed in 50 μl buffer containing 50 mM KCl, 20 mM HEPES pH 7.0, 7 mM MgSO4, 5 mM ADP, 1 mM EDTA. Reactions were then quenched, neutralized, and analyzed with HPLC using a Partisphere SAX column (Weaver et al., 2013); 1 ml fractions were collected for liquid scintillation spectrometry. The ATPase activity of PPIP5K2KD (10–270 μg/ml) was assayed from Pi release following incubation at 37°C for 120–180 min in 20 μl reaction mixtures containing 20 mM Tris/HCl, pH 7.5, 10 mM ATP, 100 mM KCl, 0.1 mM EDTA, and 13 mM MgCl2.</p><!><p>In the figures, error bars represent SEs from three experiments.</p><!><p>H.W., H.Y.G., A.M.R., J.D.W., and S.B.S. carried out the experiments; S.B.S. and B.V.L.P. designed and coordinated the research; and all authors provided input during writing of the paper.</p><!><p>The Protein Data Bank accession numbers for the protein/ligand structures reported in this paper are 4NZM, 4NZN, and 4NZO.</p><!><p>Document S1. Supplemental Experimental Procedures, Figures S1–S3, and Table S1Document S2. Article plus Supplemental Information</p>

PubMed Open Access

Spectral Tuning and Photoisomerization Efficiency in Push–Pull Azobenzenes: Designing Principles

This work demonstrates how push–pull substitution can induce spectral tuning toward the visible range and improve the photoisomerization efficiency of azobenzene-based photoswitches, making them good candidates for technological and biological applications. The red-shifted bright ππ* state (S2) behaves like the lower and more productive dark nπ* (S1) state because less potential energy along the planar bending mode is available to reach higher energy unproductive nπ*/S0 crossing regions, which are responsible for the lower quantum yield of the parent compound. The stabilization of the bright ππ* state and the consequent increase in isomerization efficiency may be regulated via the strength of push–pull substituents. Finally, the torsional mechanism is recognized here as the unique productive route because structures with bending values attributable to the inversion mechanism were never detected, out of the 280 ππ* time-dependent density functional theory (RASPT2-validated) dynamics simulations.

spectral_tuning_and_photoisomerization_efficiency_in_push–pull_azobenzenes:_designing_principles

<!>Introduction<!><!>Results and Discussion<!><!>Results and Discussion<!>trans-AB Systems<!><!>trans-AB Systems<!><!>trans-AB Systems<!><!>trans-AB Systems<!>cis-AB Systems<!><!>cis-AB Systems<!>Conclusions<!><!>Author Contributions<!>

<p>This article published November 10, 2020 with an error in Scheme 2. The corrected file published November 19, 2020.</p><!><p>Azobenzene (AB) is a prototypical photoresponsive molecule undergoing a reversible photoinduced isomerization between its cis and trans configurations, which is strongly attractive for a widespread range of applications. The trans ↔ cis interconversion mechanism has been debated for a long time:1−10 it could take place through rotation around the central double bond (torsion) or through an in-plane bending motion (Scheme 1). Eventually, hybrid torsion-bending processes were recently proposed.11−14 Interestingly, the well-separated absorption wavelengths of the two isomers make this molecule suitable for optical switches in technological15−17 or biological18−20 devices and in the development of light-powered molecular machines.2,3,21−27 Both isomers show two absorption bands in the UV–vis window: the more intense one is associated to a π → π* transition, peaking in the UV region (301/265 nm in the gas phase, trans/cis, respectively28), while the much weaker band in the visible range (440/425 nm28) is associated to a symmetry-forbidden n → π* transition. These ππ*/nπ* bands are separated enough to allow their selective irradiation: interestingly, excitation in the UV (ππ*) and in the visible region (nπ*) shows significantly different quantum yields (QYs), about 11% and 25%, respectively, in the trans case and 27% and 56% in the cis case in n-hexane.29 The QY wavelength dependence, which is in contrast with Kasha's rule, suggests that different reaction mechanisms may take place starting from the ππ* or nπ* excited states (ESs),12 an issue that is still under discussion in experimental8,30 and theoretical8,11,31−33 studies. Because of the reversibility of the isomerization, its speed, and the simplicity of incorporating azobenzene in complex structures, many studies are focused on red-shifting the intense ππ* bands, whose UV absorption is limiting technological and biological applications. For this purpose, push–pull substituents have demonstrated to be good candidates:37−40 the simultaneous destabilization of the last π orbital (electron-donating substituent) and stabilization of the π* LUMO (electron-withdrawing substituent), results in red shift of the ππ* absorption,8,18,33,41−46 which influences the ππ–nπ* energy gap and leads to a change in the photoisomerization properties. The aim of this work is to evaluate how push–pull substituents could control the capability of AB-based photoswitches, tuning the linear absorption energy and the isomerization efficiency, depending on the mechanism behind. For this purpose, we compare the behavior of the parent AB with two different push–pull-substituted systems with increasing electron-donating/withdrawing strength: 4-methoxy-4′-cyanoazobenzene (NC–AB–OMe) and 4-(4-nitrophenylazo)aniline (O2N–AB–NH2, also known as Disperse Orange 3 or DO3); see Figure 1. The comparison is made by means of time-dependent density functional theory (TD-DFT) semiclassical dynamics simulations (RASPT2-validated at crossing points, see Table S11) accounting for multireference dynamically correlated energies. The results allow us to identify the control knobs of productive (i.e., photoisomerization) versus nonproductive (i.e., aborted photoisomerization) radiationless decays, thus paving the way to a rational design of AB derivatives with tuneable spectral properties and increased photoisomerization efficiency.</p><!><p>Selected AB-systems (bottom) considering an ensemble of eight push–pull derivatives: correlation between the strength of push–pull substituents and the lowest nπ*/ππ* vertical excitation energies (yellow/blue lines, respectively).</p><!><p>Comparison between AB and the two push–pull-substituted systems NC–AB–OMe and O2N–AB–NH2 (Figure 1) was done by conducting mixed quantum–classical dynamics simulations at the TD-DFT/CAM-B3LYP/6-31G* level, following 40 gas-phase trajectories on both the cis and trans isomer of the three systems (240 trajectories in all), initiated on the bright ππ* state. Nonadiabatic events were treated with a simplified hopping scheme, relying on the energy gap as a criterion for changing the electronic state, fixed lower than 3 kcal/mol. Back hopping was always allowed between ESs, while it was not permitted once the trajectory decayed on the ground state (GS). We ran 40 additional trajectories on the trans-AB nπ* state that were employed as a reference for the photobehavior of the more productive dark state in the parent compound. TD-DFT/CAM-B3LYP/6-31G* was validated by benchmarking against RASPT2 static calculations at the S1/S0 crossing points, using an accurate setup that was previously tested for the parent system (MS-RASPT2/RASSCF/ANO-L-VDZP),12,47 where the active space (including the valence π-orbitals and the nitrogen lone pairs) was appropriately enlarged for the push–pull-substituted systems (Figures S1–S3 in the Supporting Information). However, because TD-DFT fails to produce correctly shaped potential energy surfaces (PESs) in the region surrounding intersections with S0, we limit our analysis to the ES dynamics until the S1/S0 gap is lower than 3 kcal/mol (S1/S0 crossing seam). The O2N–AB–NH2 and NC–AB–OMe derivatives were selected after a preliminary study (at the CAM-B3LYP/6-31G* level) of eight systems with increasing push–pull strength: Figure 1 clearly shows how the substituents red-shift the ππ* state, leaving the dark nπ* roughly unchanged. Increasing the push–pull strength reduces the ππ*/nπ* energy gap, until inversion of the ππ*/nπ* energy order (Figure 1 and Table S1). Because of their small size, the selected systems are good candidates to make accurate predictions about these promising push–pull derivatives. The quantitative accuracy of the employed method is supported by the good agreement between the experimental and computed vertical energies for trans/cis-AB, trans-NC–AB–OMe, and trans-O2N–AB–NH2 (see Table 1). This also validates the prediction for the absorption values (ππ* and nπ*), which are not available in the literature, in particular for the push–pullcis-conformers, which are thermally unstable and therefore difficult to isolate and characterize.40 Besides stabilizing the ππ* state, the growing strength of the push–pull substituents also affects the charge distribution on the two phenyl rings and, consequently, the molecular dipole moment (see Figure S4 and Table S1 in the Supporting Information). The charge separation is proportional to the electron-donating/withdrawing strength, as shown in Figure 1: −NH2 is a better push group than −OMe because of the lower electronegativity of the nitrogen atom; at the same time, the −NO2 substituent "pulls" more than the −CN. The charge excess on the two halves is notably larger on the bright ππ* ES, with a consequent increase in the dipole moment value: 0.162/0.247 D and 0.241/0.327 D on ππ* for trans/cis NC–AB–OMe and the O2N–AB–NH2, respectively, compared with 0.074/0.071 D and 0.115/0.106 D of the GS. The larger dipole moment of the cis conformer could be referred to the nonplanar geometry that hinders the orbital delocalization, leading to a larger charge separation between the two halves.</p><!><p>Optimized GS bending and torsional parameters are shown at the top of Figures 2 and 4 (Cartesian coordinates for the trans and cis conformers are given in the Supporting Information). Details on the S2 and S1 average lifetimes, calculated on all trajectories or separately on the torsional or bending paths, are documented in Tables S4–S6, respectively.</p><p>In ethanol at room temperature.</p><p>In 2-methyltetrahydrofuran (MTHF) at 77 K.</p><!><p>We show how the push–pull derivatives behave dynamically different, compared to the parent system, when they are excited to the bright ππ* state in the following. Because photoexcited trans- and cis-isomers lead to quite different paths11 (as demonstrated by the experimental lifetimes in Table 1), we will discuss them separately.</p><!><p>Looking at the S2 dynamics leading to the initial S2 → S1 decay, we notice that the CNNC dihedral angle stays close to 180° in all the systems, while the CNN bending angles close and then oscillate around a value that is a bit smaller than that in the FC geometry, in agreement with recent studies on the AB photoisomerization.12Figure 2 shows the normalized distribution of the CNNC torsion (top) and CNN bending (bottom) trajectories along the trans-dynamics of the three systems, including the nπ* trans-dynamics of AB. The left panels refer to the dynamics on S2 [panels (a–c) and (h–j)], while the right plots refer to the dynamics on S1 after decay from S2 [panels (e–g) and (l–n)] or after direct excitation [for trans-AB, panels (d,k)]. The most significant effect of push–pull substitution is a drastically shorter ππ* lifetime with respect to the parent compound, where S2 is living two times longer than in the substituted trans-systems (168 fs for AB against 70 and 86 fs for NC–AB–OMe and O2N–AB–NH2, respectively; see vertical dashed lines in Figure 2, left part).</p><!><p>Normalized distribution of the CNNC torsional value (top panels) and widest CNN bending value (bottom panels) over time for the trans-system dynamics (40 for each panel) on S2 (left) and on S1 (right) until decay to the GS. The color scale refers to the normalized density of trajectories. The panels (d,k) refer to the 40 dynamics initiated in the nπ* for trans-AB. Vertical dashed lines: ES lifetimes averaged over all trajectories (black) and over torsional (red) or bending paths (green). Horizontal dotted lines: FC value of the relative coordinate. Top left structures: CNNC torsion and CNN bending values in the S0 minimum trans-systems (DFT/B3LYP/6-31G* optimized).</p><!><p>On the other hand, in the subsequent dynamics on the nπ* state (S1), bending oscillations accompany the torsional motion (Figure 2, right part), leading to a S1 → S0 crossing region spanning from planar to fully rotated CNNC values (Tables S4–S6 in the Supporting Information). This is due to an extended S0/S1 crossing seam, that has been extensively documented in previous studies,11,47,48 covering both bending and torsional modes, where the fully (∼90°) rotated structures are the lowest in energy, but also, higher energy, less-rotated structures could be accessible through the bending mode, provided that enough kinetic energy is available in the dynamics. Based on the characteristics of the S1 → S0 seam, we have grouped the trajectories in two different sets, labelled torsional and bending paths, based on the CNNC torsional value at the S1 → S0 decay: the former includes trajectories decaying on S0 at CNNC < 135° (half between 180 and 90°), the latter includes trajectories which, to a great extent, preserve the planarity of azobenzene until decaying to the GS (CNNC > 135°). Most trajectories for all the three trans-systems follow the bending path (82.5/65/65% for AB/NC–AB–OMe/O2N–AB–NH2, respectively; see Table 2), but none of them reach bending values that could justify a possible inversion-driven isomerization process (i.e., close to 180°; see Scheme 1, bottom part of Figure 2, and Tables S4–S6). This explains the smaller QY of trans-AB from ππ* (11% vs 25% from nπ*29): the most populated bending paths are reaching S0/S1 CIs with neither bending nor torsion values large enough to allow the trans–cis isomerization. Moreover, the bending motions are mainly symmetric (see values in Tables S4–S6), and even a hypothetical concerted bending mechanism would lead back to the reagent. Consequently, on the basis of the large number of dynamics on the three trans-systems (120, 40 for each system), we conclude that the only productive process follows the torsion mechanism. However, because our analysis is limited to the ES dynamics until decay to the GS, we can only have an upper bound estimate of the ππ* QY, which is given by the number of torsional paths populated for each system: we obtained 17.5%, 35%, and 35% for trans-AB, NC–AB–OMe, and O2N–AB–NH2, respectively, envisaging a larger QY in the push–pull systems than that in the parent AB (see Table 2). To prove that the isomerization QY correlates well with the population of the torsion mode, we ran 40 dynamics for trans-AB (using the same initial conditions as for the ππ* state) starting from the more productive nπ* state (experimental QY = 25%29): in this case, 32.4% of the trajectories belong to the torsional path (Table 2), a value that is close to the observed QY. Previous semiclassical dynamics by Granucci and Persico49 employing a semiempirical Hamiltonian reported values for the QYs of 15% and 33% starting from the ππ* and nπ*state, respectively, which is perfectly in line with the amount of torsional trajectories obtained in each case from our simulations. Remarkably, the ratio between the torsional paths populated when initiating the dynamics either in the ππ* or nπ* state (Table 2) matches well with the experimental ππ* and nπ* QY ratio (theoretical estimate: 17.5/32.4 = 0.54, experimental QY ratio in n-hexane:29 11/25 = 0.44). This strengthens the hypothesis that CN=NC torsion is the productive mechanism, which explains the larger QY in the transpush–pull systems.</p><!><p>Trans-system dynamics: torsional path = CNNC < 135° at the S1/S0 decay and bending path = 135° < CNNC < 180° at the S1/S0 decay. The geometrical parameters are averaged over all the set of trajectories belonging to each torsional/bending group. Cis-system dynamics: all trajectories are ascribable to the torsional path (>99%), for which CNNC > 45° at the S1/S0 decay.</p><!><p>To further support and rationalize that the push–pull systems could be more productive than the parent trans-AB because of the larger population of torsional paths, Figure 3 shows the S2 → S1 (red) and the S1 → S0 (dark blue) hopping point distribution, along the bending/torsional coordinates, for the ππ* trajectories of the three different systems. Interestingly, for the push–pull systems, the S1 → S0 hopping point distribution obtained starting from the bright ππ* state is matching with the trans-AB distribution obtained starting from the more efficient nπ* state [light-blue points in Figure 3 panel (a) versus blue points in panels (b,c)], envisaging that the push–pull derivatives excited to ππ* behave exactly as AB excited to nπ*, for which a larger isomerization productivity is experimentally documented. Instead, the S1 → S0 decay points for trans-AB when excited to S2 show a clearly different distribution, largely concentrated in the bending region. Additionally, the average bending values at the S1 → S0 hopping points are a bit smaller for the torsional trajectories (and in the nπ* dynamics) than for the bending ones (see Table 2 and Figure 3), which is perfectly in line with the shape of the S1/S0 crossing region depicted in our earlier studies,11 showing that fully rotated CIs (∼90°) display smaller bending values than less-rotated (and therefore less productive) ones.11,47 It is thus apparent that by calibrating the strength of push–pull substituents, one could red-shift the absorption maximum of the bright ππ* state, bringing it closer to that of the productive nπ* and concurrently increase the photoisomerization efficiency, two main achievements in the design of photoactive AB-based systems.</p><!><p>Projection of all the decay geometries in the torsion/bending space for the trans (left part) and cis (right part) dynamics. Red points = S2 → S1, blue points = S1 → S0 hopping point distribution populated along the ππ* (S2) dynamics of the three systems. Light-blue points in panel (a) correspond to S1 → S0 hopping points populated by the 40 trajectories starting from the trans-AB nπ* (S1) state. The vertical line in each panel defines the torsional and the bending regions (i.e., half way between 180 and 90° for the trans-isomers and between 0 and 90° for the cis ones).</p><!><p>Concerning the lifetimes, we see a nice agreement between experiments and theory: time-resolved photoelectron spectroscopy experiments5 show two decay time constants for trans-AB: the shorter (170 fs) is perfectly matching our trans-AB S2 → S1 average decay time value of 168 fs (black dashed line in Figure 2a,h and in Table 1); the longer one (420 fs) is close to the computed S2 + S1 average decay time of 323 fs of the slower torsional paths (red dashed line in Figure 2e,l; see also Table 1 and details on average lifetimes in Table S4). Even though the original work5 attributed the longer experimental lifetime of 420 fs to two higher lying ππ* states (S3–S4), the low oscillator strength reported for them5,11 suggests that the population of S2 is by far more probable and that the 420 fs time constant could instead be associated to the S2 + S1 deactivation following the CNNC torsional motion toward the twisted S1/S0 crossing region. This hypothesis was already proposed by Granucci et al.,49 and it is also supported by the following theoretical50−52 and experimental34 studies reporting a S1 lifetime of about 0.4 ps.</p><p>An insight into the behavior of the dynamics following S2 → S1 decay clearly shows an average nπ* S1 lifetime that is almost doubled in the push–pull derivatives than in AB (155 fs, 141 fs, and 63 fs for NC–AB–OMe, O2N–AB–NH2, and AB, respectively, black dashed lines in Figure 2e–g). Interestingly, the S1 average lifetimes of the push–pull systems resemble those of the more productive dark nπ* state of the parent AB when it is directly excited, (130 fs, see black dashed line in Figure 2d,k), once again showing that the dynamics of the push–pull systems excited to the ππ* resembles that of the nπ* state of AB. Eventually, we observe that the S1torsional path average lifetime in the transpush–pull systems (red dashed line, Figure 2f,g) is about three times longer than that in the bending paths (green dashed lines Figure 2f,g), which is again similar to the dynamics of trans-AB from the nπ* state (62 vs 270 fs, Figure 2d,k). The longer lifetime of the torsional versus bending path could be simply referred to the time needed for internal vibrational energy redistribution from the bending to torsional mode, which is necessary to populate the MEP leading to the nπ* decay process to the GS.47 This is in line with the recently published AB ππ* CASPT2 dynamics,12 indicating that the productive CN=NC torsional mechanism is slower than the unproductive route characterized by symmetric bending modes.</p><p>To explain the opposite trend in the S2 and S1 lifetimes observed in push–pull AB as compared to the parent compound, we propose a simple model, which rationalizes entirely the differences documented in both ESs for the three systems. Because the push–pull substituents stabilize only the bright state, while keeping the nπ* energy unaffected, we imagine a simple shift of the ππ* PES, as shown in Scheme 2. By lowering the ππ* state, the crossing with nπ* becomes more accessible (i.e., lower activation energy), thus leading to a shorter S2 lifetime for the push–pull derivatives (Figure 2). Additionally, less energy becomes available along the initially populated bending modes on S1 to eventually access the higher energy S1/S0 crossing region at roughly planar structures (torsional angle CNNC around 180°). Eventually, vibrational energy redistribution takes place, triggering population of the nπ* (torsional) minimum energy path and populating the slower, but more productive, torsional paths leading to rotated S1/S0 CI structures.</p><!><p>Cis-isomers behave similarly to the trans ones: the push–pull substituents red-shift the ππ* intense band according to their electron-donating/withdrawing strength, leaving the nπ* state energy roughly unchanged (Table 1). The main difference with respect to the trans-conformers is that except for few outliers, more than 99% of the 120 cis-dynamics reach S1 → S0 regions, which is always attributed to the CNNC torsional decay mechanism (CNNC > 45°; see Tables 2 and S7–S9 in the Supporting Information), as clearly shown in Figure 3. This is in line with the larger experimental QY observed in cis-AB (Φ = 0.27 vs 0.11 of the trans(29)). Moreover, torsion is activated already on S2 (reaching torsion values up to 50°; see Figure 3) and becomes notably larger on the S1, as shown by the torsion panels of Figure 4, because of the nonplanar FC starting structure. The earlier activation of the torsional motion, compared to the trans analogues, impedes the early decay to the nπ* state through the bending funnel, resulting in longer S2 lifetimes of the cis-isomers, compared to the trans ones, in agreement with previous dynamics simulations of AB from the ππ* state.52 The bending motions are more asymmetric than those in the trans-systems (Table 2), and to be more specific, the larger bendings are mainly attributed to the fragment bearing the electron donor group (−OMe or −NH2; see Tables S8 and S9 in the Supporting Information). Anyways, none of the cis-dynamics reach bending angles close to 180° (see Tables S7–S9), suggesting that the inversion path is not populated, as already noted for trans-systems. The S2 lifetime is shortening with the increasing push–pull strength (Figure 4a–c and g–i), supporting the previously explained hypothesis that the ππ* red shift speeds up the access to the ππ*/nπ* crossing seam (Scheme 2). Instead, the nπ* lifetime in the cis-isomers is not affected by the push–pull substituents (Figure 4d–f and j–l) because the steeper gradient along the torsional coordinate drives the system straight to the rotated nπ*/S0 peaked CIs, as documented previously.11,53 These differences in the S1 PES shape (compared to the flat trans-nπ* surface) correlate with a larger amount of kinetic energy along the torsional mode, inevitably leading to an increased photoisomerization QY with respect to the trans analogues.</p><!><p>Normalized distribution of the CNNC torsional value (top panels) and widest CNN bending value (bottom panels) over time for the cis-system dynamics (40 for each panel) on S2 (left) and on S1 (right) until decay to the GS. The color scale refers to the normalized density of trajectories. Vertical dashed lines: ES lifetimes averaged over all trajectories. Horizontal dotted lines: FC value of the relative coordinate. Top left structures: CNNC torsion and CNN bending values in the S0 minimum cis-systems (DFT/B3LYP/6-31G*-optimized).</p><!><p>That said, looking at the S2/S1 and S1/S0 CI distribution along the torsion/bending coordinates in Figure 3, we see that the parent and push–pull-derivatives behave similarly, populating the same photoisomerization processes and thus suggesting similar photoisomerization QYs (which is expected to remain higher than that in the trans analogues).</p><!><p>Kasha's rule violation in AB systems was often attributed to two different decay channels that are populated when exciting directly the ππ* (S2) or nπ* (S1) state. The present work supports and extends this hypothesis by proposing a unified mechanistic model, which can be applied to both azobenzene and its push–pull derivatives, foreseeing a higher QY for the latter, with respect to the parent compound. By analyzing a large number of TD-DFT (RASPT2 validated) ππ* molecular dynamics on AB and two push–pull AB derivatives, we see that S2 trajectories in the parent compound are mainly characterized by CNN/NNC bending motions preserving the system planarity and eventually leading to S2/S1 and, subsequently, to S1/S0 crossing regions, which are unproductive and drive the systems back to reactant repopulation. Indeed, only a small number of trajectories redistributes the vibrational energy along the torsional mode that could drive the system to the fully rotated S1/S0 CI (∼90°), triggering the isomerization. We demonstrate that push–pull substituents mitigate this situation, leading to a behavior from the ππ* state (bright) that is similar to that of the productive nπ* state (dark). Indeed, the substituents induce a ππ* red shift, bringing the bright state closer to the dark nπ* and therefore leading to the population of the same (and more productive) torsional pathways (Scheme 2). This demonstration, based on a significant number of trajectories, endorses the push–pull derivatives as flexible candidates for more efficient and visible light-activated switches, which are attractive for technological and biological applications. Moreover, the large number of trajectories is a strong statistical support to finally assign the photoisomerization process exclusively to the torsion mechanism, even if it is assisted by large CNN/NNC bending motions.11,12,14,48,53 Indeed, structures distorted enough to support a photoisomerization driven by the inversion route are never reached (Scheme 1). Therefore, only the torsion is the productive path, while the pure bending is an unproductive reaction coordinate, justifying the lower QY observed in AB when exciting the ππ* state (bending-dominated) as compared to direct nπ* excitation (torsion-driven). Because of the importance of the embedding on the excited-state dynamics,54−56 QM/MM studies are currently undergoing to disclose effects of solvent polarity and viscosity on the photoactivity of these systems.</p><!><p>Computational details, vertical excitation energies and dipole moments of AB push−pull derivatives, nature of TD-DFT and RASPT2 ESs, charge distribution in the FC geometry, details on TD-DFT dynamics, RASPT2 energies at S1/S0 crossings (0 K dynamics), and Cartesian coordinates of the B3LYP/DFT/6-31G*-optimized GS minima (PDF)</p><p>jp0c08672_si_001.pdf</p><!><p>∥ F.A. and A.N. contributed equally.</p><!><p>Financial support from the PHANTOMS project, PRIN: PROGETTI DI RICERCA DI RILEVANTE INTERESSE NAZIONALE – Bando 2017, Prot. 2017A4XRCA, is acknowledged. M.G., I.C., and A.N. acknowledge support from the H2020-NMBP-TO-IND-2018-2020/DT-NMBP-09-2018 project SIMDOME, grant agreement no. 814492.</p><p>The authors declare no competing financial interest.</p>

PubMed Open Access

A 9-Connected Zirconium-Based Metal-Organic Framework for Ammonia Capture

Construction of multifunctional metal-organic frameworks (MOFs) with asymmetric connectivity have the potential to expand the scope of their utilization. Herein, we report a robust 9-connected microporous Zr-based MOF, NU-300, assembled from asymmetric tri-carboxylate ligands and Zr6 nodes. As indicated by single-crystal X-ray diffraction analysis, there exist uncoordinated carboxylate groups in the structure of NU-300 that can participate in ammonia (NH3) sorption through acid-base interactions which yield high uptake of NH3 at low pressure regions (<0.01 bar). In situ infrared (IR) spectroscopy shows the interactions between Brønsted acidic sites and NH3, which 2 suggests that NU-300 can be used as a sorbent for NH3 capture at low pressures.

a_9-connected_zirconium-based_metal-organic_framework_for_ammonia_capture

Introduction<!>3<!>Result and Discussion<!>6<!>Conclusions

<p>Metal-organic frameworks (MOFs) are a class of porous crystalline materials assembled by metal nodes and organic ligands. 1,2 Because of their high porosity, 3 versatile pore structures, 4 and tunable chemical functionalities, [5][6][7] MOFs can be precisely designed at the molecular level for targeted applications including, but not limited to, gas storage [8][9][10] and separation, [11][12][13][14][15][16] catalysis [17][18][19] , chemical sensing, 20,21 and more. 22,23 Particularly, zirconium-based MOFs (Zr-MOFs) have attracted extensive attention in recent years due to their high thermal and chemical robustness, as well as topological diversity. 24,25 Many reported Zr-MOFs contain 12-, 10-, 8-, 6-or 4connected Zr6 nodes with di-, tri-, and tetra-carboxylate ligands, showcasing different network topologies. 26 In designing stable zirconium MOFs, in general, highly symmetric di-, tri-, and tetra-carboxylate ligands are used which results in high symmetry MOFs when combined with symmetric Zr6 nodes. 24,25 In the case of asymmetric ligands, although copper and rare-earth (RE) based MOFs have been constructed with non-planar tri-carboxylate ligands, 27,28 however, Zr-based MOFs have rarely been explored. 29 Asymmetric ligands can introduce new types of coordination environments and potentially unlock new topological structures for novel MOF materials. 27 Thus, the construction of Zr-MOFs with asymmetric ligands remains under explored for understanding the relationship between the different connected Zr6 nodes and ligands within their unique structures, as well as exploiting the potential chemical properties related to the asymmetric ligands.</p><!><p>The exceptional stability of Zr-MOFs renders them as promising candidates for the capture of ammonia (NH3). [30][31][32] Considering the associated corrosiveness and toxicity, the capture of NH3 at extremely low concentrations using porous materials under ambient conditions in industrial settings is of great importance to comply with limits of short-term exposure (35 ppm) and long-term exposure (25 ppm) set by the Occupational Safety and Health Administration (OSHA). 31 Several MOFs including HKUST-1, 33 MOF-74, 34 M(isonicotinic acid)2 (M = Zn, Co, Cu, Cd), 35 MFM-300(Al), 36 M2Cl2BBTA (M = Co, Mn), 37 M2Cl2(BTDD) (M = Mn, Co, Ni and Cu), 38 and UiO-66 series 30,32 have been tested for NH3 uptake. However, the majority of MOFs studied for NH3 uptake showed structural degradation upon exposure or significant loss of uptake after consecutive cycles. So far, only a limited number of MOFs are reported to exhibit good reversible NH3 sorption over multiple cycles, e.g. MFM-300(Al), Co2Cl2BBTA and M2Cl2(BTDD). [36][37][38] The development of robust MOFs with reversible NH3 sorption that can withstand multiple cycles remains challenging.</p><p>With the aforementioned challenges in mind, herein, we report a robust and functional 9-connected microporous Zr-MOF (NU-300) assembled from an asymmetric tri-carboxylate ligand and novel Zr6 node with an unusual linker connectivity. Notably, NU-300 has a high ammonia uptake at low pressures by exploiting Brønsted acidic sites on both the ligand and the node, which can be used an adsorbent for NH3 capture at low concentration.</p><!><p>The solvothermal reactions of ZrCl4 and 3,5-di(4'-carboxylphenyl)benzoic acid 4 (H3L) in N,N-dimethylformamide (DMF) with formic acid as a modulator, yielded colorless rhombic-shaped crystals of NU-300 (NU stands for Northwestern University).</p><p>Single crystal X-ray diffraction revealed that NU-300 crystallizes in the orthorhombic Imma space group. The asymmetric unit of NU-300 contains four Zr 4+ atoms, each uniquely eight-coordinated. As shown in Figure 1a, Zr1 is coordinated by four distinct oxygen atoms from different carboxylates of four H3L ligands and four μ3-O entities.</p><p>Zr2 is coordinated by two distinct oxygen atoms from carboxylates of two H3L ligands, one oxygen atom from formic acid, one oxygen atoms from DMF and four μ3-O entities.</p><p>Zr3 is coordinated by three oxygen atoms from various carboxylates of three H3L ligands, one oxygen from DMF and four μ3-O entities. Finally, Zr4 is coordinated by three oxygen atoms from various carboxylates of three H3L ligands, one oxygen from terminal OH/H2O and four μ3-O entities. Two Zr1, two Zr2, one Zr3 and one Zr4 atoms are connected together by eight μ3-O atoms to form the Zr6O8 cluster (Figure S2). This cluster differs from previously reported Zr6 nodes that contain only one or two crystallographically independent Zr 4+ atoms. 26,29,[39][40][41][42] Moreover, the H3L ligand adopts two types of coordination modes. In mode I, two carboxylate groups of H3L adopt a bridging bis-monodentate mode while one is monodentate (Figure 1b). In mode II, two carboxylate groups of H3L adopt monodentate and bridging bis-monodentate modes, respectively, while one carboxylate group remains uncoordinated (Figure 1c) and points to the channel along the a-axis in the 3D structure of NU-300 (Figure 1d). The topological analysis indicates that NU-5 tritopic linkers, since monodentate carboxylic acids are not included in the connectivity counting. Thus, the 3D framework of NU-300 can be simplified as a (3, 3, 3, 9)connected network with a point symbol of (4.6 2 ) (4 2 .6)2 (4 8 .6 20 .8 8 ) (Figure 1e), which is a new topology. The 9-connected Zr6 nodes of NU-300 are different from a previously reported 9-connect node, wherein the carboxylate ligands bridge adjacent Zr atoms in the node.</p><!><p>The phase purity of bulk NU-300 is confirmed by comparison of simulated and experimental PXRD patterns (Figure 2a). Thermogravimetric analysis (TGA) reveals that the framework of NU-300 starts to decompose at around 400 °C in air (Figure S3), demonstrating the high thermal stability of NU-300. The permanent porosity of NU-300 is confirmed by N2 adsorption measurements at 77 K (Figure 2b). NU-300 exhibits a type I isotherm, indicative of the microporous character of the material. The Brunauer-Emmett-Teller (BET) area and total pore volume for NU-300 are calculated to be 1470 m 2 /g and 0.58 cm 3 /g, respectively, while the pore size distribution based on DFT modeling indicates micropores of ~11 Å (Figure 2b). The chemical stability of NU-300 is then investigated by soaking NU-300 in 100 °C H2O, 0.01 M aqueous HCl (pH=2) and 0.001 M aqueous NaOH (pH=11) solutions for 24 h. As illustrated by PXRD patterns (Figure 3a), the crystallinity of the NU-300 is retained after these treatments. To further confirm the chemical stability of NU-300, N2 sorption measurements are also conducted after these treatments (Figure 3b). The N2 isotherms of NU-300 in hot water and acidic conditions are almost identical to that of 8 pristine NU-300, confirming its structural integrity and permanent porosity after exposure to boiling water and dilute acid. However, a decrease is observed in surface area and pore volume after base treatment. In light of the presence of free carboxylate groups (-COOH) within the framework, NH3 sorption tests were conducted on NU-300 to investigate potential guest-host 9 interaction. The first run adsorption-desorption isotherm shows an adsorbed NH3 amount of 8.28 mmol/g at 298 K and 1.0 bar (Figure 4a). At pressures less than 0.01 bar (Figure 4b), NH3 molecules preferentially adsorbed to the Zr6 nodes and Brønsted acidic sites of the free -COOH groups in NU-300, exhibiting the steep NH3 isotherm.</p><p>After regeneration at room temperature under vacuum, an NH3 uptake of 3.30 mmol/g remained in NU-300, likely due to the chemisorption process by the formation of strong interactions between the uncoordinated -COOH groups and NH3 molecules, apart from the acidic -OH groups on Zr6 nodes (Figure S5). This indicates that Brønsted acid sites, particularly the free -COOH groups, aid NU-300 in NH3 uptake at low pressures, allowing NU-300 to reach approximately 4 mmol/g uptake by 0.10 bar and 1.5 mmol/g by 0.01 bar, and the latter can be recycled for at least three times (Figure 4b). There was a loss in NH3 uptake capacity at 1 bar between the first and second cycles while the third cycle of NH3 sorption was nearly identical (5.71 and 5.41 mmol/g at 1.0 bar), suggesting that the loss in capacity occurs primarily in the initial sorption cycle. We then turned to IR spectroscopy to further assess how the carboxylic acid sites in NU-300 interact with NH3 molecules during the adsorption and desorption processes.</p><p>As observed in Figure 5, with NH3 exposure for 60 min on NU-300, two characteristic NH3 bands are observed, indicating NH3 interactions with NU-300: the degenerate and symmetric deformation of NH3 at 1625 and 1360 cm -1 , respectively. 43 The location of C=O stretching vibration of the free -COOH groups at 1730 cm -1 (Figure S7) decreases 11 to 1710 cm -1 upon NH3 exposure, possibly due to deprotonation and subsequent resonance that weaken the C=O bond strength. 44 The appearance of overlapping bands between 3300 and 3700 cm -1 supports the deprotonation of -COOH by NH3. 45 However, none of these bands between 3300 and 3700 cm -1 disappear upon Ar purge. NH3 exposure also results in a new band at 1480 cm -1 , which could be assigned to the vibration of N-H in NH4 + . 43 These observations indicate that NH3 molecules were protonated in acid-base reaction with Brønsted acidic sites. IR spectra showed that even after Ar purge residual adsorbed ammonia bands were still present which rationalized the loss in uptake between the first and second cycles.</p><!><p>In summary, we have designed a robust 9-connected Zr-based MOF, NU-300 using an asymmetric tri-carboxylate ligand. The presence of free -COOH groups on the ligands provides NU-300 with relevant functional properties for the chemisorption of NH3 molecules at low partial pressures. For the first cycle adsorption, the adsorbed uptake of NH3 was 8.28 mmol/g at 298 K and 1.0 bar. On the other hand, the second and third</p>

ChemRxiv

Predicting Fixation Tendencies of the H3N2 Influenza Virus by Free Energy Calculation

Influenza virus evolves to escape from immune system antibodies that bind to it. We used free energy calculations with Einstein crystals as reference states to calculate the difference of antibody binding free energy (\xce\x94\xce\x94G) induced by amino acid substitution at each position in epitope B of the H3N2 influenza hemagglutinin, the key target for antibody. A substitution with positive \xce\x94\xce\x94G value decreases the antibody binding constant. On average an uncharged to charged amino acid substitution generates the highest \xce\x94\xce\x94G values. Also on average, substitutions between small amino acids generate \xce\x94\xce\x94G values near to zero. The 21 sites in epitope B have varying expected free energy differences for a random substitution. Historical amino acid substitutions in epitope B for the A/Aichi/2/1968 strain of influenza A show that most fixed and temporarily circulating substitutions generate positive \xce\x94\xce\x94G values. We propose that the observed pattern of H3N2 virus evolution is affected by the free energy landscape, the mapping from the free energy landscape to virus fitness landscape, and random genetic drift of the virus. Monte Carlo simulations of virus evolution are presented to support this view.

predicting_fixation_tendencies_of_the_h3n2_influenza_virus_by_free_energy_calculation

1 Introduction<!>2.1 Scheme of the Free Energy Calculation<!>2.2 Einstein Crystal<!>2.3 Modified Hydrogen Atoms<!>2.4 Expressions of Free Energies<!>2.5 Implementation of Free Energy Calculation Algorithm<!>3.1 Free Energy Landscape<!>3.2 Historical Substitutions in Epitope B<!>4.1 Fitness of the Virus Strains<!>4.2 Selection in the Epitope<!>4.3 A Picture of the H3N2 Virus Evolution<!>4.4 Multiple Substitutions<!>4.5 Prediction of Future Virus Evolution<!>5 Conclusion

<p>Influenza A virus causes annual global epidemics resulting in 5–15% of the population being infected, 3–5 million severe cases, and 250,000–500,000 fatalities.1 The subtype of influenza A is determined by two surface glycoproteins—hemagglutinin (H) and neuraminidase (N). The H3N2 virus has been one of the dominant circulating subtypes since its emergence in 1968. The antibodies IgG and IgA are the major components of the immune system that control influenza infection, binding to the influenza hemagglutinin.2 There are five epitopes at the antibody binding sites on the top of H3 hemagglutinin, namely epitopes A–E. The epitope bound most prolifically by antibody is defined as the dominant epitope, and it is central to the process of virus neutralization by antibody and virus escape substitution.3 The cellular immune system, on the other hand, plays a relatively less recognized role in handling the invasive influenza virus.2 The cellular system along with the innate immune system exerts a somewhat more homogeneous immune reaction against genetically distinct influenza strains.2,4</p><p>Vaccination is currently the primary method to prevent and control an influenza epidemic in the human population.1 Influenza vaccination raises the level of antibody specific for hemagglutinin and significantly enhances the binding affinity between antibody and hemagglutinin. Vaccine effectiveness depends on the antigenic distance between the hemagglutinin of the administered vaccine strain and that of the dominant circulating strain in the same season.3,5 Memory immune response from virus in previous seasons as well as vaccination in the current and previous seasons impose selective pressure on the current circulating virus to force it to evolve away from the virus strains recognized by memory antibodies that selectively bind to hemagglutinin.</p><p>As a result of the immune pressure and the escape evolution of the influenza virus, which is largely substitution in the dominant epitope of hemagglutinin, the influenza vaccine must be redesigned and administered each year, and the vaccine effectiveness has been suboptimal in some flu seasons.3,6 The escape evolution in the dominant epitope is at a higher rate than that in the amino acid sites outside the dominant epitope.7 Sites in the dominant epitope also show higher Shannon entropy of the 20 amino acids than do those outside the dominant epitope.8 High substitution rate and Shannon entropy in the dominant epitope of hemagglutinin suggest that the dominant epitope is under the strongest positive selection by human antibodies. The immune pressure against each genotype of the dominant epitope can be at least partially quantified by the binding constant between antibody and hemagglutinin.</p><p>The H3N2 virus and human immune system in this work are simplified to be a system consisting of the H3 hemagglutinin and the corresponding human antibody. Exposure by infection or vaccination produces an affinity-matured antibody with the binding constant to the corresponding hemagglutinin equal to 106–107 M−1, while the binding constant of an antibody uncorrelated to the hemagglutinin is below 102 M−1.2 Escape substitutions may decrease the binding constant by changing the antibody binding free energy ΔG. Some substitutions decrease the antibody binding constant more than others and have higher probabilities to be fixed, because decrease in the antibody binding constant is favorable to the virus. Here we define the difference of antibody binding free energy as ΔΔG =ΔG42 − ΔG31 in which ΔG31 and ΔG42 are antibody-wildtype hemagglutinin binding free energy and antibody-evolved hemagglutinin binding free energy, respectively, as shown in Figure 1. The fixation tendency of each substitution is a function of the difference of the antibody binding free energy9 of the escape substitution.</p><p>Epitope A or B of the H3N2 virus was dominant in most influenza seasons.3 Epitope B of the H3N2 virus was the dominant epitope presenting more substitutions than any other epitope in the recent years. Epitope B was also dominant in 1968 when H3N2 virus emerged. Thus during these periods of time, the substitutions in epitope B directly affect the antibody binding constant and reflect the direction of the virus escape substitution. To attain a global view of the effects of substitutions in epitope B, it is necessary to compute a matrix containing the differences of antibody binding free energy caused by each possible single substitution in epitope B. There are 21 amino acid sites in epitope B, and each residue in the wild type strain may substitute to any of the 19 different types to amino acid residues, hence we need to calculate a 19 × 21 matrix with 399 elements. Such a matrix is a free energy landscape quantifying the immune selection over each evolved influenza strain. In this free energy landscape, the virus tends to evolve to a position with low binding affinity of antibody to evade antibodies and reduce the immune pressure. Calculation of this landscape will enable us to study the mechanism of immune escape from a quantitative viewpoint, providing a criterion to describe and foresee the evolution of influenza virus.</p><p>This paper is organized as follows: In Materials and Methods section, we describe the protocol for the free energy calculation and the system of hemagglutinin and antibody. In Results section, we present and analyze the calculated free energy landscape. The substitutions observed in history are also compared with the results of the calculation. In the Discussion section, a general picture of H3N2 virus evolution under the selection pressure of the immune system is discussed and simulation results are discussed. Finally, our work is summarized in the Conclusion section.</p><!><p>The expression of the binding constant K depends on the antibody binding free energy ΔG, K = exp(−ΔG/RT). The Boltzmann constant R = 1.987 × 10−3 kcal/mol/K. The temperature is fixed to the normal human body temperature T = 310 K. Shown in Figure 1, one substitution in hemag-glutinin changes the antibody binding free energy from ΔG31 to ΔG42. The first and second subscripts define the end state and the starting state of the binding process, respectively. The ratio of the antibody binding constant after and before substitution is written as (1)K1K0=exp(−ΔΔG/RT) where K1 and K0 are the antibody binding constant to substituted and wildtype hemagglutinin, respectively.</p><p>The difference of the antibody binding free energy ΔΔG =ΔG42 − ΔG31 =ΔG43 − ΔG21 is calculated by applying the Hess' Law to the thermodynamic cycle defined by State 1–4 in Figure 1. The processes corresponding to ΔG43 and ΔG21 are unphysical but more convenient to simulate. We calculated AG21 and AG43 for each amino acid substitution in the unbound hemagglutinin and hemagglutinin bound by antibody, respectively On the surface of the virus particle, hemagglutinin exists in the form of a trimer in which three monomers are encoded by the same virus gene. Thus we simultaneously substituted the amino acids in three hemagglutinin monomers in the trimer. The antibody has a Y-shaped structure with two heavy chains and two light chains. In the resolved structure (PDB code: 1KEN), the hemagglutinin trimer is bound by two Fab fragments. Thus, we incorporated the Fab dimer into the system for MD simulation.</p><p>Using the software CHARMM,10 we calculated each of AG21 and AG43 using thermodynamic integration.11 We used molecular dynamics (MD) simulation to obtain the ensemble averages of the integrand from which each of ΔG21 and ΔG43 is calculated. The potential energy for the MD algorithm to sample the conformation space of the system is (2)U(r,λ)=(1−λ)Ureac(r)+λUprod(r) in which r is the coordinates of all the atoms, λ is the variable of integration, Ureac is the potential energy of the system corresponding to wildtype hemagglutinin, and Uprod is the potential energy of the system corresponding to substituted hemagglutinin. The value of ΔG21 or ΔG43 is (3)ΔG=∫01〈∂U(r,λ)∂λ〉λdλ=∫01〈Uprod(r)−Ureac(r)〉λdλ. The integrand 〈Uprod (r) − Ureac (r)〉λ is the ensemble average with fixed λ of potential energy difference between the system after and before substitution. The interval of integration λ ∈ (0,1) was equally divided into four subintervals in each of which a 16-point Gauss-Legendre quadrature was applied to numerically integrate the ensemble averages. The ensemble averages with 64 distinct λ ∈ (0,1) were calculated by MD simulation with the potential energy defined in Eq. (2).</p><!><p>We introduce the Einstein crystals to calculate the free energy of the reference state in the dual topology at both endpoints of the thermodynamic integration. To illustrate the function of the Einstein crystals, we analyze the free energy of the dual topology without Einstein crystals when λ = 0 as an example. We denoted by n1, n2, and n0 numbers of the reactant atoms, product atoms, and all the remaining atoms in the system, respectively. We denoted by r, rproduct, and x the coordinates of the reactant atoms, product atoms, and all the remaining atoms in the system. The momenta of reactant atoms, product atoms, and the remaining atoms are denoted by pr,i, pp,i, and px,i. The masses are similarly denoted by mr,i, mp,i, and mx,i. The Hamiltonian of the system with λ = 0 is (4)H=∑i=1n0px,i22mx,i+∑i=1n1pr,i22mr,i+∑i=1n2pp,i22mp,i+Un0(x)+(1−λ)Un0+n1(x,r)+λUn0+n2(x,rproduct). The partition function is (5)Q=∏i=1n0(2πmx,ih2β)3/2∏i=1n1(2πmr,ih2β)3/2∏i=1n2(2πmp,ih2β)3/2 ∫dxdrexp[−βUn0(x)−βUn0+n1(x,r)]×∫drproduct1=Qreal×∏i=1n2[(2πmp,ih2β)3/2V]=Qreal×Qproduct. When λ = 0, this partition function is the product of Qreal, the partition function of the real system without product atoms, and Qproduct, the partition function of the product atoms when λ = 0.</p><p>The free energy is given by −1/β times the logarithm of the above partition function. The free energy is (6)G=Greal−1β∑i=1n232log  (2πmp,ih2β)−1βn2logV. As shown in the above equation, the effect on the translational entropy from the product atoms is proportional to the logarithm of system size V. It diverges in the thermodynamic limit. This divergence exists, no matter what λ scaling is performed. Note that we do not use the Einstein crystals to handle the translational entropy a ligand loses or gains when binding a flexible biomolecular receptor, which is taken into account by the thermodynamic cycle in Figure 1. The translational entropy, proportional to logV in Eq. (6), is that of the dummy product atoms, not that of the bound or unbound complex.</p><p>The value of G depends on the identity of the product atoms. Thus, the contribution to the thermodynamic integration is different at the two endpoints, i.e. −kT log Qreactant ≠ −kT log Qproduct, in which Qreactant is the partition function of the reactant atoms when λ = 1. Note also that the expression of the partition function contains the factor Qproduct for the product atoms. Relating the conventional expression for thermodynamic integration, Eq. (3), to ΔΔG of Eq. (1) requires one to account for this term. This term arises from the use of a dual topology in CHARMM, and this term is typically ignored. While the contribution from the decoupled atoms is not constant, it can be exactly calculated if the restricted partition function over the decoupled atoms can be calculated. This calculation is what the Einstein crystal performs, using an Einstein crystal for the reference state rather the ideal gas in Eq. (4).</p><p>In four 16-window thermodynamic integrations, the smallest variable of integration is λ = 1.32 × 10−3. Since λ is close to zero, product atoms in the system have potential energy near zero and behave as ideal gas atoms, with translational entropy proportional to the logarithm of system size, see Eq. (6). Exact calculation of the translational entropy terms of product atoms at λ = 0 by explicit dynamics seems difficult, because the translational entropy of the product atoms grows as the logarithm of the system size. These relatively free product atoms destabilize the system. This entropy divergence is a fundamental feature of the statistical mechanics, not a numerical artifact. Unrestrained product atoms induce large fluctuation of the Hamiltonian in the MD algorithm. These fluctuations increase the standard error of the quantity Uprod (r)−Ureac (r), which is defined in Eq. (3) and is computed from the trajectory of the MD simulation. These fluctuations often cause the numerical integration algorithm in the MD simulation to be unstable.12 In this case, the energy of the simulated system increases rapidly. This phenomenon causes CHARMM to terminate abnormally. The translational entropy introduced by the free atoms at λ = 0 and 1 affects the result. Reactant atoms cause the same problem near λ = 1.</p><p>We noticed that the non-linear scaling, i.e. using a high power of λ such as the fourth power of λ, in Eq. (2)13,14 did not work. The high power of the smallest λ is extremely close to zero and the product atoms are almost free, which cause the MD simulation to terminate abnormally at several windows with small λ. Additionally, the issue of translational entropy of reactant and product atoms needs to be addressed. Even when the MD algorithm with the non-linear scaling of λ13,14 terminates and appears to have generated a converged simulation trajectory, this does not necessarily imply that the translational entropy of reactant or product atoms has been properly controlled. In fact, the λ scaling approach may hide the entropy divergence at λ = 0 or λ = 1 by letting the algorithm terminate due to numerical roundoff error, rather than building statistical mechanical reference states for each of λ = 0 and λ = 1 to account for or control the effect of translational entropy.</p><p>An alternative to λ scaling introduces the soft-core potential as a way to turn off the potential.15,16 The soft-core approach, like the lambda-scaling approach, does not address the translation entropy of the atoms at λ = 0 or λ = 1. Previous studies with non-constrained atoms at both end-points have been performed.17–23 Besides the classical molecular dynamics with a non-ideal-gas reference state introduced into the dual topology, quantum molecular dynamics via metadynamics has been used to analyze a deamidation process.24 Other applications of quantum molecular dynamics based free energy calculation include chorismate conversion to prephenate,25 isomerization of glycine,26 and histone lysine methylation.27 As illustrated in Eq. (6), the translational entropy of the uncoupled atoms causes error in the final free energy results if it is not accounted for.</p><p>One way to calculate the free energy change exactly is to use a non-ideal-gas reference state. This is quite natural, since the protein is not composed of ideal gas atoms. Deng and Roux introduced restraint potentials to confine the translational and rotational motion of a bound ligand to accelerate convergence of the simulation.28 We use this idea to exactly include the contribution from the restrained states and built two Einstein crystals as the reference states for reactant and product atoms, respectively. Our calculation allows a theoretically exact determination of the free energy due to amino acid substitution.</p><p>To handle these two difficulties at both endpoints of the integration in a theoretically exact way, we use two Einstein crystals as the reference states for reactant and product atoms, respectively. The Einstein crystal has been used as a reference state for free energy calculations. Frenkel and Ladd computed free energy of solids by building a path connecting the real solid and the reference Einstein crystal.29 Noya et al. showed that a restrained Einstein crystal is a suitable reference in the free energy calculation of biomolecules.30 The Einstein crystal, a solid state model, is consistent with the nature of antibody binding process in liquid phase. First, although the importance of biomolecular flexibility in protein-protein binding process is well-accepted, and is fully and exactly included in our calculation, we simply need to localize the product atoms when λ = 0 and the reactant atoms when λ = 1. Moreover, we need to calculate the contribution to the free energy of these localized atoms.</p><p>The choice of Einstein crystals as the reference states removes the singularity in thermodynamic integration in Eq. (3). As an example, an Einstein crystal was used as the reference state for the free energy calculation of hard-sphere fluid in order to remove the singularity in Eq. (3) at the end point λ = 0.31 In this example, the reference Einstein crystal was achieved by harmonically coupling the particles to their equilibrium positions and removing all interactions between particles.32</p><p>We here use Einstein crystals as the reference states to calculate the binding free energy change due to amino acid substitution. The Einstein crystal is a model for localized atoms. The free energy of the Einstein crystal can be exactly calculated. One Einstein crystal contains distinguishable and non-interacting atoms under harmonic constraints around reference positions fixed in space. In the Einstein crystal, the atom i with coordinates ri has potential energy (7)Ui(ri)=Ki2∥ri−ri0∥2 in which ri and ri0 are the actual and reference position of the atom, respectively, and Ki is the force constant of the harmonic constraint. We denote by mi the mass of atom i. The canonical partition function of an Einstein crystal is (8)QE(N,V,T)=1h3N∫exp(∑i=1N−βpi22mi)exp(∑i=1N−βKi∥ri−ri0∥22)dpdr=(2πhβ)3N∏i=1N(miKi)3/2. The spatial fluctuation of atom i in the Einstein crystal is (9)〈(δri)2〉=3βKi.</p><p>In our system, we let the potential energy for MD simulation defined by Eq. (2) become (10)U(r,λ)=(1−λ)Ureac(r)+λUprod(r)+λUein,reac(r)+(1−λ)Uein,prod(r). Therefore reactant and product atoms are localized at both λ = 0 and λ = 1. The reference positions of atoms in Einstein crystals are the equilibrium positions of corresponding reactant and product atoms. To minimize the numerical error during the thermodynamic integration calculation, we minimized the fluctuation of the integrand of thermodynamic integration 〈∂U (r, λ) /∂λ〉λ = 〈Uein,reac (r) − Ureac (r)〉λ+ 〈Uprod (r) − Uein, prod (r) 〉λ. Minimization of the terms on the right hand size is approximately achieved by letting the average spatial fluctuation of each atom in Einstein crystals equal to that of the corresponding reactant or product atom, i.e. (11)〈(δri)2〉reac=〈(δri)2〉ein,reac=3βKireac (12)〈(δri)2〉prod=〈(δri)2〉ein,prod=3βKiprod For each atom in the Einstein crystal, the force constant of harmonic constraint, Kireac or Kiprod, was calculated from the monitored fluctuations of the corresponding reactant or product atom with Eq. (11) or Eq. (12). In the scheme in Figure 1, the states with Einstein crystals are states 1b, 2b, 3b, and 4b.</p><!><p>The frequency of atom vibration depends on its mass. Hydrogen atoms generally have the highest vibration frequencies in the system. Such high frequencies require short time step in MD simulation and increase computational load. To limit vibration frequencies and allow a longer time step, one can apply the SHAKE algorithm to fix the length of any bond involving hydrogen atoms.33 The SHAKE algorithm decreases the degrees of freedom in the system by introducing additional constraints between atoms. Instead, we artificially changed the mass of hydrogen atoms from 1.008 to 16.000 amu in order to preserve degree of freedom in the system following the suggestion by Bennett.34 A larger mass of hydrogen atoms allows a longer time step in the MD algorithm. Pomes and McCammon showed that changing the hydrogen mass to 10 amu allow using a 0.01 ps time step to simulate a system which consists of 215 TIP3P water molecules, smaller than our system.35 Feenstra et al. change the mass of hydrogen atoms to 4 amu to increase the simulation stability of a system which contains protein and water molecules and resembles our system.36 We set the time step as 0.001 ps, a value widely used in simulations with physical masses for all atoms, to gain higher stability in the simulation of our large system with a hemagglutinin trimer, a Fab dimer, and water molecules. As with the Einstein crystals, we exactly calculated and subtracted off the contribution of the change to the hydrogen mass to ΔΔG. Note that the modification of hydrogen mass is independent to the reference states in the simulation, which is selected to be Einstein crystals in this project. In fact, most of the hydrogen atoms in the system are neither reactant nor product atoms. In Figure 1, the states with Einstein crystals and modified hydrogen atoms are states 1a, 2a, 3a, 4a, 1b, 2b, 3b, and 4b.</p><!><p>Introducing two Einstein crystals and heavier hydrogen atoms changes the potential energy in the system, as well as the canonical partition functions. After modification of hydrogen atoms, the mass of atoms changed from mr,i to mr,i′, from mp,i to mp,i′, or from mx,i to mx,i′. Canonical partition functions of the states in Figure 1 are: (13)Q3(n0+n1,V,T)=1h3(n0+n1)∏i=1n0(2πmx,iβ)3/2∏i=1n1(2πmr,iβ)3/2×Z3(n0+n1,V,T) (14)Q3a(n0+n1,V,T)=1h3(n0+n1)∏i=1n0(2πmx,i′β)3/2∏i=1n1(2πmr,i′β)3/2×Z3(n0+n1,V,T) (15)Q3b(n0+n1+n2,V,T)=1h3(n0+n1)∏i=1n0(2πmx,i′β)3/2∏i=1n1(2πmr,i′β)3/2×Z3(n0+n1,V,T)(2πhβ)3n2∏i=1n2(mp,i′Kiprod)3/2 (16)Q4(n0+n2,V,T)=1h3(n0+n2)∏i=1n0(2πmx,iβ)3/2∏i=1n2(2πmp,iβ)3/2×Z4(n0+n2,V,T) (17)Q4a(n0+n2,V,T)=1h3(n0+n2)∏i=1n0(2πmx,i′β)3/2∏i=1n2(2πmp,i′β)3/2×Z4(n0+n2,V,T) (18)Q4b(n0+n1+n2,V,T)=1h3(n0+n2)∏i=1n0(2πmx,i′β)3/2∏i=1n2(2πmp,i′β)3/2×Z4(n0+n2,V,T)(2πhβ)3n1∏i=1n1(mr,i′Kireac)3/2 in which the states are denoted by the subscripts. Contribution of the potential energy part of the Hamiltonian to the partition function is (19)Z3(n0+n1,V,T)=∫exp(−βUn0+n1(r))dr (20)Z4(n0+n2,V,T)=∫exp(−βUn0+n2(r))dr From the partition functions, free energies defined in Figure 1 are calculated: (21)ΔG3a=−32β∑i=1n0ln(mx,i′mx,i)−32β∑i=1n1ln(mr,i′mr,i) (22)ΔG4a=32β∑i=1n0ln(mx,i′mx,i)+32β∑i=1n2ln(mp,i′mp,i) (23)ΔG3b=−3n2βln(2πhβ)−32β∑i=1n2ln(mp,i′Kiprod) (24)ΔG4b=3n1βln(2πhβ)+32β∑i=1n1ln(mr,i′Kireac) (25)ΔG43b=−1βln[∏i=1n1(mr,i′/Kireac)3/2Z4(n0+n2,V,T)∏i=1n2(mp,i′/Kiprod)3/2Z3(n0+n1,V,T)]. The free energy between state 3 and 4 is (26)ΔG43=ΔG43b−1βln(2π/hβ)3n2∑i=1n2(mp,i/Kiprod)3/2(2π/hβ)3n1∑i=1n1(mr,i/Kireac)3/2=ΔG43b−1βlnQE2(n2,V,T)QE1(n1,V,T) in which QE1 and QE2 are the partition functions of the Einstein crystals for product atoms and reactant atoms, respectively. The free energy ΔG43b was calculated by thermodynamic integration while ΔG43 was used to calculate the free energy difference of one substitution. Note that the correction term between ΔG43b and ΔG43 is independent of the masses of atoms. Canonical partition functions as well as free energies of the state 1, 1a, 1b, 2, 2a, and 2b are calculated in a similar way.</p><!><p>The above discussion is the theoretical basis for the implementation of our free energy calculation algorithm. The free energy calculation protocol consists of four steps. First, we built the dual topology with reactant and product atoms in the amino acid substitution site in separated antibody and hemagglutinin or antibody-hemagglutinin complex. We then solvated the protein system and modified the mass of hydrogen atoms. Second, two Einstein crystals were introduced as the reference states for the reactant and product atoms, respectively Third, the MD simulation was run at 64 windows. The thermodynamic integration algorithm obtained the free energy values ΔG21 for separated antibody and hemagglutinin or ΔG43 for antibody-hemagglutinin complex, as in Figure 1. This step gave the ΔΔG value. Fourth, we calculated the error bar of the ΔΔG value obtained in the last step. The technical details of these four steps are illustrated in the text below. Also described are the verification of the free energy calculation protocol, the software and hardware information, and the CPU hours consumed by the protocol.</p><p>The hemagglutinin trimer of H3N2 virus strain A/Aichi/2/1968 with bound dimer antibody HC63 (PDB code: 1KEN) was used in our calculation. For each amino acid substitution, we built the dual topology with side chains of both amino acids prior to the simulation. Reactant and product atoms were defined as the side chains in the original and substituting amino acid, respectively. All the covalent and non-bonded interactions between reactant and product atoms were removed. The protein was in an explicit water box with periodic boundary condition. The mass of hydrogen atoms was changed from 1.008 to 16.000 amu.</p><p>All the simulations were performed by CHARMM c33b2 with CHARMM22 force field.10 We first fixed the positions of hemagglutinin trimer except reactant atoms and minimized the system with 200 steps of steepest descent (SD) algorithm and 5000 steps of adopted basis Newton-Raphson (ABNR) algorithm. We ran a 5 ps MD simulation of the system, the trajectory of which gave the spatial fluctuation 〈(δri)2〉 of each reactant atom. Then we fixed reactant atoms, released product atoms, and ran a 5 ps MD simulation to obtain the spatial fluctuation of each product atom. Final positions of both reactant and product were adopted as the reference positions of the corresponding Einstein crystal. The force constant Ki of each atom in Einstein crystals was obtained from 〈(δri)2〉 by Eq. (11) and Eq. (12). With modified hydrogen atoms and two Einstein crystals as the reference states of reactant and product atoms, state 1b, 2b, 3b, and 4b in Figure 1 were generated for thermodynamic integration.</p><p>In thermodynamic integration, MD simulations were run at 64 windows with distinct λ. In each window, pressure of the system was first calibrated with a 10 ps MD simulation in an isothermal-isobaric (NPT) ensemble. The duration of 10 ps is appropriate because it is long enough to equilibrate the pressure and short enough to prevent the protein from drifting away from the original location. We fixed coordinates of the residues and water molecules except for those within 15 Å from the three alpha carbons. Then we removed amino acid residues and water molecules other than those within 27.5 Å from the three alpha carbons of substituted residues in the hemagglutinin trimer to reduce the system size, because the fixed atoms are not included in the topology of movable atoms and the cutoff of the non-bonded forces is 12 Å. The Ewald sum was used to calculate charge interactions. Note that this substantial reduction of the system relies on the assumption that the free energy change due to the amino acid substitution is mostly affected by atoms near the binding site after the system reaches equilibrium. This assumption is based on two facts: the conformations of hemagglutinin and antibody are stable once the system reaches equilibrium, and all the removed or fixed atoms have invariant interactions with the substituting amino acid residues. The stable protein conformation means amino acid residues far away from the substituting residue do not move during the amino acid substitution process. In the CHARMM22 force field used in this project, the cutoff of non-bonded force is 12 Å and less than the 15 Å threshold for system reduction. The system reduction does not directly affect the force on the substituted residue because of absence of the long-range non-bonded force between the substituted residue and atoms removed from the system. This system reduction method was also applied to compute binding free energy of subtilisin,37 of tripsin,21 and of Src SH2 domain.22 Robust results were obtained in all of these applications. Generally, this system reduction strategy can produce reliable result if the reduced system contains the residues and molecules critical to the binding process.21 We note that the system reduction method could be a limitation of the free energy calculation model. The fixing of amino acid residues and water molecules described in section 2.5 substantially reduced the CPU time needed, but is an approximation to the real system containing the whole proteins. This limitation reflects the tradeoff between model accuracy and required computational resource. In the canonical ensemble, the new system was equilibrated for 200 ps and simulated for another 900 psas the data production phase. The integrand of thermodynamic integration is the ensemble average of the sampled trajectory 〈∂U (r, λ) /∂λ〉λ = 〈Uein, reac (r) − Ureac (r) − Uein,prod (r) + Uprod (r)〉λ. The free energy ΔG21 and ΔG43 between the real states was calculated by adding a correction term of the Einstein crystals in Eq. (26). Finally, the difference of antibody binding free energy is ΔΔG = ΔG43 − ΔG21.</p><p>Error bars of ΔΔG are also given. The convergence behavior of the simulation was analyzed using the block average method developed by Flyvbjerg and Petersen.38 As mentioned above, the MD simulation for either the unbound hemagglutinin or the hemagglutinin-antibody complex contains 64 windows with distinct λ. The 900 ps data production phase contains 9 × 105 simulation steps. The values A = Uprod (r) − Ureac (r), as in equation Eq. (3), computed in consecutive simulation steps were grouped into bins, and consecutive bins were merged progressively. The quantity σ2 (A) / (n − 1), in which σ2 (A) is the variance of the average of each bin A1,A2, …,An and n is the number of bins, increases with the bin size and reaches a plateau when the bin size is 1 × 104 steps. We fixed the bin size to 1 × 104 steps and estimate the variance of ensemble average 〈A〉 as σ2 (A) / (n − 1), following Flyvbjerg and Petersen's method.38</p><p>This protocol, without the Einstein crystal contribution, was verified by recalculating published free energy differences of amino acid substitution T131I.9 Without the Einstein crystal contribution, our protocol gave the ΔΔG = 5.69 ± 0.07 kcal/mol, compared to the ΔΔG = 5.20 ± 0.94 kcal/mol in the published work.9 Theoretically exact results presented here include the Einstein crystal contribution. We note that the theoretically exact ΔΔG for T131I, including the Einstein crystal contribution, is 3.71 ± 0.07 kcal/mol.</p><p>The simulation was performed using CHARMM22 force field at three clusters: tg-steele.purdue.teragrid.org (Intel Xeon E5410, 2.33 GHz), sugar.rice.edu (Intel Xeon E5440, 2.83 GHz), and biou.rice.edu (IBM POWER7, 3.55 GHz), as well as at the condor pool tg-condor.rcac.purdue.edu at Purdue University. Simulation of each substitution took approximately 7.5 thousand CPU hours on average, and so this work consumed about three million CPU hours.</p><!><p>For each of the 21 amino acid sites in epitope B, we substituted from alanine to each one of the 19 other amino acids, in which we used the neutral histidine (CHARMM code: Hse) as the model of histidine. The free energy difference and standard error of each substitution were calculated by the MD simulation (see Materials and Methods). The wildtype amino acid in each site of epitope B was extracted from the hemagglutinin sequence of the H3N2 strain A/Aichi/2/1968. The free energy difference and standard error of the substitution from the wildtype amino acid in each site were then calculated from the values for the change from the wildtype amino acid to alanine and from alanine to the new amino acid. The values are listed in Table 1.</p><p>As described in Eq. (26), each ΔΔG value listed in Table 1 contains the contribution of two Einstein crystals. The contribution of Einstein crystals to the final ΔΔG values was calculated for each of the 399 amino acid substitutions in epitope B. The average fraction of the contribution of Einstein crystals in the calculated ΔΔG values is 44%. The contribution of Einstein crystals is far greater than that of the statistical error of our free energy calculation in Table 1, which is 4.5% on average. Thus, the Einstein crystal contribution is both theoretically exact and practically important. In 371 of the 399 substitutions, the absolute values of the contribution of Einstein crystals is greater than 1.96 standard errors of the final ΔΔG values. That is, the contribution of Einstein crystals is significant with p < 0.05 in 93.0% of all the amino acid substitutions. Consequently, it is essential to incorporate Einstein crystals in the free energy calculation to eliminate the error caused by the methods that neglect the unknown effect of the translational entropy of the free atoms in thermodynamic integration. The contribution of the translational entropy of ideal gas-like atoms (λ = 0 or λ = 1) needs to be either calculated or removed by a theoretically exact method to perform an exact free energy calculation.</p><p>The obtained ΔΔG values allow us to analyze the character of each of the 20 amino acids. We first averaged over all the 21 amino acid sites in epitope B the ΔΔG value caused by the single substitutions from alanine to the other amino acids. The averaged ΔΔG values are listed in Table 2. The largest ΔΔG are caused by the negatively charged amino acids (Glu, Asp) and the positively charged amino acids (Arg, Lys), indicating that introduction of charged amino acids in the dominant epitope decreases the binding affinity between antibody and hemagglutinin. Note that amino acid substitutions that change the charge of hemagglutinin significantly affect the calculated free energy values.39–41 The issue of how to best calculate free energy differences when charge changes has been debated over the years. In the present paper, we are using the standard Ewald approach with explicit solvent. We note that the evolutionary history of H3 hemagglutinin since 1968 shows an increasing trend of the number of charged amino acids in epitope B,42 which agrees with the results that introduction of charged amino facilitates virus evasion from antibody, as illustrated in Table 2. The result that introduction of charged amino acid on average increases ΔΔG is not an artifact, is supported by data from the influenza evolution, and is expected on the basis that charge is hydrophilic. In addition to the charge, the rank of free energy differences also largely correlated to the size of amino acid. By the definition used by RasMol,43 the 16 uncharged amino acids are tagged as hydrophobic (Ala, Gly, Ile, Leu, Met, Phe, Pro, Trp, Tyr, Val), large (Gln, Hse, Ile, Leu, Met, Phe, Trp, Tyr), medium (Asn, Cys, Pro, Thr, Val), and small (Ala, Gly, Ser), as shown in Table 2. The ranks of small amino acids are lower than those of medium amino acids (p = 0.036, Wilcoxon rank-sum test) and those of large amino acids (p = 0.085, Wilcoxon rank-sum test). In contrast, the hydrophobicity of the uncharged amino acids is largely uncorrelated to their ranks by ΔΔG. As a result, charged amino acids in the dominant epitope are essential to the immune evasion while the virus escape substitution among small amino acids have minimal effect.</p><p>Epitope B comprises 21 amino acid sites in the top of the hemagglutinin trimer. Taking the probability for one substituting amino acid to exist at each site to be proportional to the relative frequency of this amino acid in H3 hemagglutinin, the weighted average free energy difference in each of the 21 sites was calculated. The relative frequencies of 20 amino acids were obtained from 6896 H3 hemagglutinin sequences deposited between 1968 and 2009 in the NCBI database44 and listed in Table 2. Also using the ΔΔG values in Table 1, we calculated and tabulated in Table 3 for each site i the value of 〈ΔΔG〉i, which is the average ΔΔG weighted by the probability for each different amino acid to be introduced, where probability is proportional to the relative frequencies of 20 amino acids counted from the H3 sequences in NCBI database from 1968 to 2009.</p><p>As shown in Table 3, there is obvious variation among the expected free energy differences 〈ΔΔG〉i caused by single substitutions at amino acid site i of epitope B. This variation is partly due to the wildtype amino acids in the sites. For instance, the wildtype amino acid in site 190 is Glu that has the highest rank in Table 2. As shown in Table 3, any amino acid substitution in site 190 tends to have a negative ΔΔG. Another cause of variation in 〈ΔΔG〉i is that distinct sites affect differently the antibody binding process. Epitope B of the wildtype A/Aichi/2/1968 hemagglutinin sequence contains five sites with threonine: 128, 155, 160, 187, and 192. The mathematical expectancies 〈ΔΔG〉i in these five sites are −7.746, 4.471, 4.956, 1.182, and −1.737 kcal/mol, respectively. Therefore, each site in epitope B has a specific effect on the virus escape substitution. A random substitution in epitope B affects the antibody binding free energy differently depending on the site and the substituting amino acids.</p><p>The variation of 〈ΔΔG〉i is also reflected by the tertiary structure of the epitope B bound by the antibody By looking into the structure of epitope B shown in Figure 2. Epitope B resides in two protruding loops from amino acid site 128 to 129, and from site 155 to 165, respectively, and in a α-helix from site 186 to 198. Site 128 has a negative average free energy difference 〈ΔΔG〉128 = −7.746 ± 0.098 kcal/mol. All the other sites in these two loops show a positive 〈ΔΔG〉i value of a random substitution, with the minimum 〈ΔΔG〉157 = 3.944 ± 0.090 kcal/mol in site 157. The α-helix is located between hemagglutinin and antibody. In the α-helix, the sites facing towards the antibody usually present large positive 〈ΔΔG〉i values such as site 193 and 196, while the sites facing towards the hemagglutinin show lower 〈ΔΔG〉i such as site 189, 192, and 197. Thus in the one dimensional sequence from site 186 to 198, the 〈ΔΔG〉i values oscillate with peaks and valleys corresponding to the sites in the α-helix facing alternatingly to the antibody and hemagglutinin. Consequently, the variation of the expected free energy changes in distinct sites depends on the structure of the hemagglutinin-antibody complex.</p><!><p>The simulation results are supported in two aspects by amino acid sequence data of H3 hemagglutinin collected since 1968. These historical sequences are downloaded from the NCBI Influenza Virus Resource45 and aligned. First, Pan et al. analyzed the number of charged amino acid in epitope B of H3 hemagglutinin in each year since 1968, and found an increasing trend of charged amino acids.42 This finding supports the results that amino acid substitution introducing charged residues on average facilitates virus escape from antibody, as illustrated in Table 2. Second, amino acid substitutions in epitope B between 1968 and 1975 also verified the free energy calculation, as shown below.</p><p>With the knowledge of the free energy landscape of the single substitutions, we are able to recognize favorable single substitutions in epitope B. Substitutions with large positive ΔΔG values enable the virus to evade the immune pressure and increase the virus fitness. Favorable substitutions grow in the virus population. Selection for substitutions with large ΔΔG is part of the evolutionary strategy of the virus. The results of free energy calculation can also explain the substituted virus strains collected in history.</p><p>We analyzed the hemagglutinin sequence information of H3N2 strains evolving from the A/Aichi/2/1968 strains. H3 hemagglutinin circulating from 1968 to 1971 was mainly in the HK68 antigenic cluster while those circulating from 1972 to 1975 were mainly in the EN72 antigenic cluster.46 Table 4 shows that in the dominant epitope B, there were 17 substitutions occurred in 12 sites collected between 1968 to 1975,47 which contributed to the immune evasion and corresponding virus evolution from the HK68 cluster to the EN72 cluster. Also listed in Table 4 are the free energy differences of these historical substitutions. The 17 substituting amino acids have significantly higher ranks compared to the corresponding wildtype amino acids (p = 0.0044, Wilcoxon signed-rank test). This significant difference is expected because 15 of 17 substituting amino acids have ranks between 1 and 10, while 10 of 12 wildtype amino acids in the substituted site have ranks between 11 and 20. In all the 21 sites in epitope B, 15 of 21 wildtype amino acids have ranks between 11 and 20. Additionally, the ΔΔG values of these 17 substitutions listed in Table 4 are greater than the expected free energy differences 〈ΔΔG〉i in Table 3 of random substitutions in the 12 substituted sites (p = 0.013, Wilcoxon signed-rank test).</p><p>We also looked into the historical escape substitutions in epitope B evading the immune pressure of the vaccine strains. For each influenza season, the amino acids in the administered vaccine strain were defined as the wildtype ones and those in the dominant circulating strain as the substituting amino acids. In each of the 19 seasons in which H3N2 virus was the dominant subtype from 1971 to 2004, the substitutions in epitope B were located3 and their ΔΔG values were obtained from Table 1. As shown in Table 5, escape substitutions in epitope B as of 1973 mostly had positive ΔΔG and generated substituting amino acids with increased rank (p = 0.047, Wilcoxon signed-rank test). Such tendency to introduce amino acids with higher ranks was not observed after 1973: the ranks of wildtype and substituting amino acids after 1973 present little significant difference (p = 0.28, Wilcoxon signed-rank test). The hemagglutinin of A/Aichi/2/1968 used in the free energy calculating is in the HK68 antigenic cluster. Perhaps after the virus evolved into the next EN72 cluster, change in the virus antigenic character stimulates the immune system to produce new types of antibody other than the HC63 antibody used in the calculation. A different binding antibody changes the free energy landscape of the substitutions in epitope B. Thus the application of the present free energy landscape should be limited within the HK68 and EN72 clusters. Free energy differences of substitutions in the EN72 cluster would need to be calculated using the updated antibody crystal structure.</p><!><p>The free energy landscape shown in Table 1 gives the change of the antibody binding affinity, K1/K0 = exp(−ΔΔG/RT), induced by each possible substitution in epitope B of the wildtype hemagglutinin. The majority of the substitutions lead to positive ΔΔG, and yield a reduced binding affinity K1 that is smaller than the binding affinity of the original mature antibody K0. Decreased antibody binding constant grants the virus a higher chance of evading the immune pressure and infecting host cells. We propose that virus fitness is positively correlated to the free energy difference ΔΔG. The other factor affecting virus fitness is the capability of the hemagglutinin to maintain the normal biochemical functions, such as virus entry. Most sites in epitope B changed amino acid identities during 1968 to 2005 as the H3N2 virus kept circulating.47 We therefore postulate that the substitutions in epitope B do not greatly interfere with the biochemical function of hemagglutinin, and virus fitness is dominantly determined by the free energy difference resulted from substitutions in epitope B.</p><p>The binding constant between hemagglutinin and antibody after the first round of maturation is about 106 M−1, and the binding constant of an uncorrelated antibody is below 102 M−1.2 On average, four substitutions in epitope B change the substituted hemagglutinin sufficiently so that the immune response of the original antibody binding to epitope B is abrogated.3 Since this is a reduction of the binding constant from roughly 106 M−1 to 102 M−1, one amino acid substitution that contributes to immune escape causes on average a 10-fold decrease in antibody binding constant, or equivalently ΔΔGcrit = 1.42 kcal/mol at 310 K. Assuming the effect of immune evasion can be broken into the sum of individual amino acid substitutions in the dominant epitope,3 we define the virus fitness w as the sum of the contribution in each site of epitope B (27)w=A0+∑epitopeBδwi. We denote by ΔΔGiαγ the free energy difference to substitute amino acid α to amino acid γ at site i. We investigated two versions of the virus fitness landscape. The first is to define δwi as a linear function of the free energy difference of the substitution (28)δwi=A1ΔΔGiαγΔΔGcrit. The second is to define δwi as a step function (29)δwi=A2H(ΔΔGiαγ−ΔΔGcrit) in which H is the Heaviside step function. Illustrated in the simulation below, either definition of fitness is sufficient to explain the observed immune evasion of the H3N2 virus.</p><!><p>Evolution of the H3N2 virus is driven jointly by neutral evolution and selection.48 Neutral evolution may be ongoing in sites outside the epitopes. The high substitution rate in epitope B suggests that selection is the major factor shaping the pattern of evolution in that epitope.47 Shown in Table 4 and Table 5 are the historical substitutions. The significantly increased ranks of free energy differences suggests the existence of selection by the immune pressure for substitutions that have increased the free energy difference ΔΔG and decreased the antibody binding constant. The immune selection is directional: certain types of amino acids such as charged ones were initially more likely to be added into the epitope B42 because they maximally decreased the antibody binding constant as indicated in Table 2. The heterogeneity of the expected free energy difference of a random substitution in Table 3 shows that each site in epitope B has a specific weight with regard to immune escape.</p><p>Table 4 also illustrates that the immune selection did not necessarily pick the amino acid with the highest rank of ΔΔG as the substituting amino acid. Amino acids with moderate rank were introduced into epitope B even for the fixed substitution T155Y. Therefore the historical evolution did not simply substitute amino acids by maximizing the free energy differences in Table 1. This phenomenon is possibly due to two causes. First, the virus fitness may be insensitive to the ΔΔG values, e.g. A1 in Eq. (28) may be small, or amino acid substitutions with large ΔΔG values may contribute equivalently to the fitness, as in Eq. (29). Second, only a small fraction of virus in one host is shed by the host and infects the next host, so the population size of propagated virus from one host is smaller by several orders of magnitude than the total virus population size in the same host. Additionally, a seasonal bottleneck exists in the influenza virus circulation.49 Both random mutation and small population sizes lead to dramatic randomness in the evolution. Consequently, the evolution of H3 hemagglutinin is not solely determined by maximizing the free energy differences in Table 1 and minimizing the antibody binding constant, even if the virus is under immune selection. Instead, randomness plays a key role in the H3N2 virus evolution.</p><!><p>Selection depends on the fitness of each virus genotype that is quantified as a non-decreasing function of the free energy difference ΔΔG. Moderate selection in epitope B requires that fitness improvement is limited when ΔΔG is large. One possibility is that the ratio A1/A0 in Eq. (28) is small. Another is that the fitness takes the form of Eq. (29) in which all substitutions with ΔΔG > ΔΔGcrit have equal fitness.</p><p>The virus evolution is also affected by the genetic drift. Genetic drift is a term which captures the random component of evolution due to the large size of the phase space of possible substitutions relative to the single set of substitutions that lead to the highest viral fitness. The effect of genetic drift is quantitatively reflected in the fixation process of a new strain, as shown in the simulation below A narrow bottleneck of virus propagation allows only a small fraction of the progeny to survive, imposing a notable probability that a favorable substitution is lost in the next generation. The effect of genetic drift is to increase the randomness in the virus evolution so that observed substitutions are based on chance in addition to the fitness of these substitutions.</p><p>To model the H3N2 evolution discussed above, we ran two Monte Carlo simulations of the influenza evolution model. A population of N sequences of epitope B with 21 sites were created and initialized as the wildtype A/Aichi/2/1968 sequence. Here N = 103 to account for a narrow genetic bottleneck of hemagglutinin and for tractability of the simulation. We iterated the simulation program for 5,000 generations or about five years to recreate a pattern of evolution similar to that in history and shown in Table 4. The random substitution rate of H3 hemagglutinin is roughly 4.5 × 10−6 amino acid substitution/site/generation.50 We let the number of substitutions follow a Poisson distribution with mean λ = 21 × 4.5 × 10−6 N = 9.5 × 10−5N and randomly assigned the substitution sites. The substituting amino acid at each substitution site was randomly picked from the remaining 19 amino acids proportional to the historical frequencies observed in hemagglutinin. The fitness w in the first simulation was calculated for each sequence using Eq.(28) with A0 = 100 and A1 = 3 and that in the second simulation was calculated for each sequence using Eq. (29) with A0 = 100, A2 = 9, and ΔΔGcrit = 1.42 kcal/mol. Note that by choosing A1 = 3 for the first simulation, a random substitution causes the expected fitness to change from 100 to 104.9, and by choosing A2 = 9 for the second simulation, a random substitution changes the expected fitness from 100 to 105.0. The size of the progeny of each sequences equals the fitness w of the sequence if w > 0, and equals 0 if w ≤ 0. The next generation of sequences was initialized by randomly sampling N sequences from the progeny sequences.</p><p>The results of both simulations showed remarkable similarity to the observed substitutions in Table 4 with the bottleneck N equal to 103. See Figure 3 and Figure 4. Amino acid substitutions generated in the simulation are usually distinct with those in Table 4 observed in history. The ΔΔG values of each substitution emerging in the simulation are nevertheless similar to those of the historical substitutions listed in Table 4. As was observed in history in Table 4, most of the substituted strains in the simulations with relative frequency greater than 1% have positive ΔΔG values with the ranks of the substituting amino acids ranging from 1 to 10. The fixation of a newly emerged substitution takes about 1,000 generations or one year on average. Fixed substitutions mostly introduce amino acids with positive ΔΔG values in Table 1 and higher ranks in Table 2, and several of these fixed substitutions in simulation, such as E190D and N188D, have the highest ΔΔG values in the current site. However, fixed substitutions in the simulation are not always the substitutions with the highest ΔΔG values in Table 1. These observations suggest that the Monte Carlo simulation considering the effect of substitution, selection, and genetic drift is able to reproduce the pattern of evolution observed in history. This simulation also shows that besides the free energy difference of each substitution, the mapping from the free energy landscape to the fitness landscape as well as the random genetic drift are dominant factors of the evolution in virus epitopes.</p><p>Shown in Figure 3 and Figure 4 for both simulations are the trajectories of relative frequencies of substituting amino acids. The trajectories are similar to historical observations of human H3N2 virus data.47 For influenza, 1000 generations roughly equal one year. The two substitutions T155Y and N188D were fixed in epitope B in 1968–1973. As indicated by Figure 3 and Figure 4, substitution T155Y emerged between generation 3000 and 4000, or equivalently between 1971 and 1972 from the emergence of the H3N2 virus in 1968.47 Substitution T155Y was fixed between generation 4000 and 5000. Similarly, substitution N188D emerged between generation 2000 and 3000 and was fixed between generation 4000 and 5000. The first simulation in which virus fitness is calculated using Eq. (28) generated two fixed substitution, G129A that emerged at generation 4000 and was fixed by generation 5000, and E190D that emerged at generation 3600 and was fixed by generation 3900. The second simulation using Eq. (29) generated one fixed substitutions, V196D emerging at generation 2900 and fixed by generation 5000, and one substitution that nearly fixed, N188D emerging at generation 4100 and acquiring the relative frequency 0.84 at generation 5000. The trajectories in both simulations resemble those of substitutions T155Y and N188D observed in history. From these results, the two Monte Carlo simulations appear to capture the main factors of immune selection and genetic drift in evolution of the H3N2 virus.</p><!><p>In this work, we calculated the free energy difference for each possible substitution in epitope B. The free energy calculation for multiple substitutions is intractable using the current technology due to the combinatorially increasing calculation load for multiple substitutions. The issue of multiple substitutions is here addressed by assuming that the effect of immune evasion is well represented by the sum of the contribution in each substituted site in epitope B. Data indicate the independence of the immune evasion effect of the sites in epitope B.3 We may, thus, assume that the free energy difference of the multiple substitution is the sum of the individual ΔΔG values available in Table 1 plus a minor correction term.</p><!><p>The result of this work quantifies the reduction of the binding constant of antibody to virus for substitutions in epitope B with larger ΔΔG values and higher ranks of substituting amino acids. A newly emerging virus strain with larger antibody binding free energy difference has a greater probability to become the dominant strain in the next flu season. Note that due to random fluctuations in the large phase space of possible substitutions, actual trajectories deviate from the trajectory determined by choosing sites and substituting amino acids with greatest free energy differences. With a three dimensional structure of hemagglutinin of the current circulating virus and binding antibody, one is able to calculate the free energy landscape for all the possible single substitutions in the dominant epitope and estimate the a priori escape probabilities in the next season. The dominant circulating influenza strain usually possesses amino acid substitutions from the vaccine strain against which memory antibodies are generated. Usually these substitutions disrupt the antibody binding process by decreasing the binding constant, as shown in Table 5. Thus one can predict vaccine effectiveness by evaluating the antibody binding constant against the dominant circulating strain, which is acquired by calculating free energy difference of the amino acid substitutions between the vaccine strain and the dominant circulating strain.3 More accurate predictions of evolutionary pattern of virus as well as epidemiological data such as vaccine effectiveness may be obtained by optimally mapping the free energy landscape to the fitness landscape and taking into account random factors such as genetic drift in the evolution process.</p><!><p>We introduced the Einstein crystal as a technology to improve the results of free energy calculation. By calculating the free energy difference of each amino acid substitution, we obtained the free energy landscape for substitutions in epitope B of hemagglutinin. There is notable variation between the values of free energy differences of different substitutions at different sites, because the identities of original and substituting amino acids, as well as the locations of amino acid substitutions, affect to differing degrees the antibody binding process. In this free energy landscape, we suggest that virus tends to evolve to higher ΔΔG values to escape binding of antibody. Counterbalancing this selection is random drift. Historical amino acid substitutions in epitope B and Monte Carlo simulations of the virus evolution using the free energy based virus fitness, in which random genetic drift of the virus adds statistical noise into the virus evolution process, showed that selected substitutions are biased to those with positive ΔΔG values.</p>

PubMed Author Manuscript

Overexpression of MET is a new predictive marker for anti-EGFR therapy in metastatic colorectal cancer with wild-type KRAS

PurposeSince the KRAS mutation is not responsible for all metastatic colorectal cancer (mCRC) patients with resistance to anti-epidermal growth factor receptor (EGFR) monoclonal antibody (MoAb) therapy, new predictive and prognostic factors are actively being sought. MethodsWe retrospectively evaluated the efficacy of anti-EGFR MoAb-based therapies in 91 patients with mCRC according to KRAS, BRAF, and PIK3CA mutational status as well as PTEN and MET expression.ResultsIn the patient group with wild-type KRAS, the presence of BRAF mutation or PIK3CA mutations was associated with lower disease control rate (DCR), shorter progression-free survival (PFS), and shorter overall survival. Patients with MET overexpression also showed lower DCR and shorter PFS when compared with patients with normal MET expression. In a separate analysis, 44 patients harboring wild-type KRAS tumors were sorted into subgroups of 25 patients without abnormality in three molecules (BRAF, PIK3CA and MET) and 19 patients with abnormality in at least one of these three molecules. The former group showed significantly higher DCR and longer PFS following anti-EGFR therapy than the latter group.ConclusionsOur data point to the usefulness of MET overexpression, in addition to BRAF and PIK3CA mutations, as a new predictive marker for responsiveness to anti-EGFR MoAbs in mCRC patients with wild-type KRAS. This study also suggests that application of multiple biomarkers is more effective than the use of a single marker in selecting patients who might benefit from anti-EGFR therapy.Electronic supplementary materialThe online version of this article (doi:10.1007/s00280-014-2401-4) contains supplementary material, which is available to authorized users.

overexpression_of_met_is_a_new_predictive_marker_for_anti-egfr_therapy_in_metastatic_colorectal_canc

Introduction<!>Patients<!>Mutational analysis of KRAS, BRAF, and PIK3CA by direct sequencing<!>Immunohistochemistry of PTEN and MET<!><!>Statistical analysis<!>Patient characteristics<!><!>PIK3CA mutational analysis<!>PTEN immunohistochemical evaluation<!>MET immunohistochemical evaluation<!><!>Discussion<!>

<p>Cetuximab and panitumumab are monoclonal antibodies (MoAbs) that inhibit the activation of the epidermal growth factor receptor (EGFR) and its downstream pathways, namely the RAS/RAF/MAPK and the PI3K/PTEN/Akt axes [1, 2]. As the response rate (RR) to anti-EGFR MoAbs remains as low as 10–20 % in patients with metastatic colorectal cancer (mCRC) [2], several studies have been performed to identify markers predicting the efficacy of these agents. Tumors carrying oncogenic KRAS mutations typically do not respond to anti-EGFR MoAbs therapy [3]. This finding led the European Medicines Agency and, subsequently, the US Food and Drug Administration to limit the use of cetuximab and panitumumab only to patients with wild-type KRAS tumors [4]. However, since only 40–60 % of patients with wild-type KRAS tumors respond to anti-EGFR MoAb therapy, new predictive and prognostic factors are actively being sought [5, 6]. In this regard, the presence of oncogenic deregulation of EGFR and other members of its downstream signaling pathways, such as BRAF, PIK3CA, and PTEN, has been shown to influence the responsiveness to cetuximab and panitumumab and could, therefore, help to identify nonresponder patients [4, 6–10]. While many studies have demonstrated the BRAF mutation, PIK3CA mutation, and PTEN overexpression as markers for resistance to anti-EGFR MoAb therapy, some failed to show such association [4, 7, 8, 10–13]. Therefore, analysis of these genetic markers in different patient populations, in particular in different ethnic groups, will help determine their clinical significance.</p><p>Furthermore, recent studies also have suggested that activation of MET, a tyrosine kinase that acts as a receptor for hepatocyte growth factor (HGF) and can activate the RAS/RAF/MAPK and PTEN/PI3K/Akt pathways, may be a novel mechanism of cetuximab resistance in CRC [13–18]. However, it remains unclear whether MET activation can serve as a predictive marker for the response to the anti-EGFR therapy in patients with wild-type KRAS.</p><p>Therefore, we investigated the status of MET expression together with PTEN expression and mutations of BRAF and PIK3CA in tumors of Japanese mCRC patients with wild-type KRAS. The main purpose of this study was to examine these genetic profiles for potential correlations with clinical response to anti-EGFR MoAb therapy.</p><!><p>Clinical outcomes of anti-EGFR MoAb therapy were retrospectively analyzed for possible associations with the molecular features of tumors in mCRC patients. The study enrolled 91 patients who were treated at the Department of Gastroenterological Surgery and Medical Oncology, Kyorin University Hospital, between November 2008 and December 2012. All patients had presented with histologically confirmed mCRC and had been treated with salvage chemotherapy incorporating anti-EGFR MoAbs. Clinical features of the patients and pathological profiles of the tumors were obtained from patient medical records.</p><p>All patients received cetuximab- or panitumumab-based therapy for mCRC (11 as first-line, 29 as second-line, 39 as third-line, and 12 as fourth-line or greater). Cetuximab, as monotherapy or in combination with irinotecan, was administered intravenously (i.v.) at a loading dose of 400 mg/m2 over 2 h, followed by weekly doses administered at 250 mg/m2 over 1 h. Panitumumab was administered i.v. every 2 weeks at a dose of 6 mg/kg. Treatment was continued until disease progression (PD) or toxicity occurred. Clinical evaluation and tumor response was analyzed according to Response Evaluation Criteria in Solid Tumors (RECIST) [19]. This study was approved by the Research Ethics Committee, Hospital of Kyorin University School of Medicine.</p><!><p>Paraffin-embedded tissues (primary or metastatic) were sectioned at 10 μm thicknesses and mounted as three separate slides per tissue. The resulting slides were treated three times with xylene and then washed with ethanol. To minimize contamination by normal DNA, areas in which at least 70 % of the cells exhibited disease-specific pathology were dissected under a binocular microscope, from which DNA was extracted using the QIAamp FFPE Tissue Kit (QIAGEN). Segments of the KRAS, BRAF, and PIK3CA genes were amplified using gene-specific primers and subjected to direct DNA sequencing as previously described [4, 13, 20]. KRAS point mutations were screened for codons 12 and 13 within exon 2, two hot spots that cumulatively include >95 % of mutations in this gene [21]. BRAF mutations were screened for V600E within exon 15, in which >95 % of point mutations occur [7, 9]. PIK3CA mutations were screened within exons 9 and 20, in which >80 % of point mutations occur [4, 10, 12].</p><!><p>PTEN and MET expression levels were evaluated by immunohistochemistry performed on 4-μm tissue sections of paraffin-embedded specimens. PTEN was assessed using the 17.A mouse MoAb (1:25 dilution; Neomarkers, Thermo Fisher Scientific Inc., Fremont, CA); MET was assessed using the SP44 rabbit MoAb (Spring Biosciences, Pleasanton, CA) [22, 23]. Negative controls were incubated with nonimmune solution instead of primary antibody. Endothelial cells and hepatocellular carcinoma cells were used as positive controls for PTEN and MET expression, respectively. The PTEN and MET staining intensities were evaluated by a pathologist (Y.O.) who was blinded to the diagnosis of individual patients.</p><p>To our knowledge, there currently are no validated scoring systems for interpretation of PTEN or MET staining intensity. Both PTEN and MET are localized primarily in the cytoplasm [11, 24, 25]; we therefore adopted a scoring system that has been used for other cytoplasmic proteins and is based on the intensity of immunoreactivity and percentage of stained cells [26, 27]. Specifically, intensity was scored according to a four-tier system: 0, no staining; 1, weak; 2, moderate; and 3, strong. An additional one, two, or three points were assigned if the percentage of positive cells was <25, 25–50 %, or >50 %, respectively [4, 11].</p><!><p>Representative examples of immunohistochemical staining in colorectal cancer. PTEN, normal expression (a) and loss of expression (b); MET, low expression (c) and overexpression (d)</p><!><p>Comparison of categorical variables was performed with the χ 2 test or the Fisher's exact test. The progression-free survival (PFS) and overall survival (OS) were calculated using the Kaplan–Meier method. Comparisons between different groups were performed using log-rank tests. To identify independent biomarkers, multivariate analyses were performed using a logistic regression model for response and a Cox regression model for PFS and OS. Two-tailed P values of <0.05 were considered significant. All analyses were performed using SPSS software (SPSS for Windows Version 15.0; SPSS Inc., Chicago, IL).</p><!><p>All study patients were Japanese; they were 66 men and 25 women with a mean age of 67 years (range 38–85 years). At a median follow-up of 13.3 months (range 1.3–24.4 months), 78 patients (86 %) had progressed, and 41 patients (45 %) had died. Response to anti-EGFR therapy was evaluable in all patients. We observed no patients with complete response (CR), 27 with partial response (PR), 24 with stable disease (SD), and 40 with PD. Therefore, the overall RR was 29.7 %, and the disease control rate (DCR) was 56.0 %. In the whole group, PFS and OS were 3.9 and 13.3 months, respectively.</p><!><p>Characteristics of patients with wild-type KRAS (n = 67)</p><p>FOLFIRI folinic acid, fluorouracil, and irinotecan, FOLFOX folinic acid, fluorouracil, and oxaliplatin</p><p>Effect of biomarkers on RR and DCR of patients with wild-type KRAS: univariate analysis</p><p>PR partial response, SD stable disease, PD disease progression, RR response rate, DCR disease control rate</p><p>Effect of biomarkers on PFS and OS in patients with wild-type KRAS: univariate analysis</p><p>PFS progression-free survival, OS overall survival, HR hazard ratio, CI confidence interval</p><p>a Progression-free survival (PFS) and b overall survival (OS) in wild-type KRAS patients classified by BRAF mutational status. c Progression-free survival (PFS) and d overall survival (OS) in wild-type KRAS patients classified by PIK3CA mutational status. e Progression-free survival (PFS) and f overall survival (OS) in wild-type KRAS patients classified by PTEN expression status. g Progression-free survival (PFS) and h overall survival (OS) in wild-type KRAS patients classified by MET expression status</p><!><p>The mutational status of PIK3CA was determined in 84 patients. Mutations were detected in three (13 %) of 23 patients with KRAS mutations and three (5.2 %) of 58 patients with wild-type KRAS (P = 0.339; Supplementary Table 2). None of the PIK3CA-mutant patients exhibited a response to MoAb therapy. When analysis was limited to patients with wild-type KRAS, DCR was significantly associated with the PIK3CA mutational status (P = 0.027; Table 2). PIK3CA mutations also were significantly associated with shorter PFS (1.8 vs. 5.4 months; HR 2.22; 95 % CI 1.07–3.86; P = 0.005) and shorter OS (5.1 vs. 15.4 months; HR 2.16; 95 % CI 0.84–4.29; P = 0.031) (Table 3, Fig. 2c, d).</p><!><p>Of 91 patients, 75 patients were evaluable for PTEN. Twenty-four patients (32 %) showed loss of PTEN expression in the cytoplasmic compartment of the tumor cells. No significant correlation was found between PTEN expression and KRAS mutational status (Supplementary Table 2). No significant association between PTEN expression and RR, DCR, PFS, or OS was detected in patients with wild-type KRAS, although patients with loss of PTEN tended to show lower RR and DCR than those with normal PTEN expression (Tables 2, 3, Fig. 2e, f).</p><!><p>Of 91 patients, 75 patients were evaluable for MET, with overexpression of the protein detected in 36 samples (48 %) (Supplementary Table 1). As with PTEN, there was no correlation between MET expression and KRAS mutational status (Supplementary Table 2). In 54 wild-type KRAS patients evaluable for MET, MET overexpression was associated with lower DCR (53.9 % vs. 82.1 %, P = 0.040; Table 2). Furthermore, MET overexpression was associated with shorter PFS (4.7 vs. 6.8 months; HR 1.46; 95 % CI 1.06–2.02; P = 0.018; Table 3), but exhibited no correlation with OS (12.8 vs. 15.4 months; HR 1.16; 95 % CI 0.73–1.82; P = 0.524) in this patient subgroup (Table 3, Fig. 2g, h).</p><!><p>Effect of biomarkers on PFS in patients with wild-type KRAS: multivariate analysis</p><p>CI confidence interval</p><p>Effect of biomarkers on OS in patients with wild-type KRAS: multivariate analysis</p><p>CI confidence interval</p><!><p>In the present study, the MET expression, PTEN expression, and mutations of BRAF and PIK3CA in mCRC patients with wild-type KRAS were investigated in association with clinical response to anti-EGFR MoAb therapy. The most striking finding in this study was that MET overexpression was associated with lower DCR and shorter PFS in patients with wild-type KRAS. One previous study reported an association of MET overexpression with the response to anti-EGFR therapy in mCRC [13], although those researchers did not report the KRAS status of their study subjects. To the best of our knowledge, the present study is the first to clarify an association of MET overexpression with inferior clinical response to anti-EGFR MoAbs in mCRC patients with wild-type KRAS. The rate of MET overexpression in mCRC in the present study was 48 %, similar to those examined in the previous studies (17–79 %) [13, 28, 29].</p><p>MET is involved in many mechanisms of cancer proliferation and metastasis. MET contains a tyrosine kinase domain that initiates a range of signals to regulate various cellular functions [30]. MET can activate the RAS/RAF/MAPK and PTEN/PI3K/Akt pathways by itself or via EGFR transphosphorylation [15–18]. In fact, MET overexpression or genetic alteration has been shown to play a role in the pathogenesis of several tumor types. In CRC, overexpression of MET has been suggested to be associated with tumor progression [28, 31]. In addition, MET also contributes to cancer resistance against EGFR inhibitors through bypass signaling. In nonsmall cell lung cancer, amplification of MET is associated with resistance to gefitinib, the reversible EGFR tyrosine kinase inhibitor, via ErbB3 activation [17, 18, 32]. Resistance in that example is mediated by MET-ErbB3 transactivation, leading to restored signaling via the PI3K/AKT pathway [14]. Our present data revealed that MET overexpression is associated with shorter PFS, but not with altered OS, in mCRC patients with wild-type KRAS who received anti-EGFR MoAbs, suggesting that MET contributes to resistance against these therapies. If confirmed, these results attest to the feasibility of the recent development of MET-targeted agents against malignant diseases, a therapeutic approach that has already been reported in several phase I and II trials [33]. MET-targeted agents, alone or in combination with EGFR inhibitors, may offer the potential for improving patients' outcome in mCRC.</p><p>This study also adds to the growing evidence that BRAF mutational status predicts efficacy of anti-EGFR therapy in mCRC patients with wild-type KRAS. Therefore, assessment of BRAF mutations before initiation of anti-EGFR therapy appears to be justified in this patient group. However, the clinical impact of BRAF gene testing depends on the prevalence of BRAF mutations. In this study, the frequency of BRAF mutations was 5 %, a value lower than that previously reported (7.9–16.6 %), possibly reflecting the fact that BRAF mutation is a negative prognostic marker that affects OS [34, 35]. In the present study, OS was shorter in BRAF-mutated patients compared with patients with wild-type BRAF, an observation that is consistent with the results of previous studies [4, 34, 35]. Therefore, some patients with BRAF mutations may not have survived long enough to be recruited into this study. The frequency of BRAF mutations might have been higher in a prospective study, which is expected to enroll all CRC patients. In addition, the prevalence of BRAF mutations was reported to be lower in Asian people than in Western people [34]. Taken together, these data suggest that the clinical relevance of analyzing BRAF mutation in Asian mCRC patients should be assessed by prospective studies in the future.</p><p>The frequency of PIK3CA mutations in the present study (8 %) was comparable to those in previous reports (7–18 %) [4, 10, 25]. Previous studies employing wild-type KRAS patients generally reported shorter median PFS or OS in PIK3CA-mutant patients than in patients with wild-type PIK3CA [4, 7, 10]. In concordance with these results, our patients with PIK3CA mutation showed significantly shorter PFS and OS and lower DCR than those without mutation. The present results confirmed that mutation of PIK3CA is also a predictive marker for response to anti-EGFR MoAb.</p><p>Low PTEN expression has been associated with shorter PFS in CRCs treated by anti-EGFR MoAbs in several reports [4, 11, 25], while no correlation was demonstrated in another report [8, 13]. We did not detect any association between PTEN expression status and clinical response to anti-EGFR MoAb therapy. This discrepancy may reflect differences in patient characteristics or study design, and notably, the distinct IHC scoring algorithms used in the present study. The use of a standardized methodology for assessment of PTEN expression would be crucial in the future studies.</p><p>This study has some limitations. Our study was performed retrospectively in a relatively small and heterogeneous population. The majority of our population (90 %) was treated with two or more chemotherapy regimens before anti-EGFR MoAb therapy. In addition, the anti-EGFR treatment protocols were heterogeneous. The discrepancy observed in the results between univariate and multivariate analyses might reflect these factors. Our findings therefore should be validated in subsequent prospective studies before they are applied in the clinical practice.</p><p>In conclusion, our data point out the usefulness of MET overexpression and mutations of BRAF, as a new predictive marker for response to anti-EGFR MoAbs in mCRC patients with wild-type KRAS. Using these two genes may be more useful for predicting the response to anti-EGFR MoAbs. These results support the emerging view that a comprehensive assessment of genetic alterations in EGFR signaling pathways will enable an accurate identification of patients who will benefit from anti-EGFR treatment and other molecular-targeting therapies, including MET inhibitors.</p><!><p>Supplementary material 1 (PDF 138 kb)</p><p>(A) PFS and (B) OS of 91 patients classified by KRAS mutational status (PDF 640 kb)</p>

PubMed Open Access

Understanding the General Packing Rearrangements Required for Successful Template Based Modeling of Protein Structure from a CASP Experiment

As an alternative to the common template based protein structure prediction methods based on main-chain position, a novel side-chain centric approach has been developed. Together with a Bayesian loop modeling procedure and a combination scoring function, the Stone Soup algorithm was applied to the CASP9 set of template based modeling targets. Although the method did not generate as large of perturbations to the template structures as necessary, the analysis of the results gives unique insights into the differences in packing between the target structures and their templates. Considerable variation in packing is found between target and template structures even when the structures are close, and this variation is found due to 2 and 3 body packing interactions. Outside the inherent restrictions in packing representation of the PDB, the first steps in correctly defining those regions of variable packing have been mapped primarily to local interactions, as the packing at the secondary and tertiary structure are largely conserved. Of the scoring functions used, a loop scoring function based on water structure exhibited some promise for discrimination. These results present a clear structural path for further development of a side-chain centered approach to template based modeling.

understanding_the_general_packing_rearrangements_required_for_successful_template_based_modeling_of_

Introduction<!>Stone Soup Template Based Structure Prediction Procedure<!>3SP: Side-chain Driven Backbone Refinement<!>CorTorgles: Correlated Torsion Angle Loop Modeling<!>hrMD: high resolution MD Scoring Function<!>Watloop: Water Path Estimation of Solvation Score<!>Packing Analysis of Results<!>Stone Soup Performance<!>3SP Core Packing<!>CorTorgles Loop Modeling<!>Scoring functions<!>Packing Rearrangements<!>Conclusion<!>

<p>Template based protein structure prediction methods (TBM) have the potential to rapidly 'solve' the structures of many gene sequences of unknown structure. The explicit aim of the structural genomics initiative is to solve the structures of sequences without obvious homology to known structures, increasing the number of templates available for TBM. Despite these efforts, the Protein Data Bank1 has seen virtually no growth in the number of protein folds in the last several years, suggesting that most soluble, globular protein folds have been discovered. As the Critical Assessment of Protein Structure Prediction (CASP) experiments have shown, the ability of predictors to significantly improve these templates' similarity to the target structure remains unimproved2–7. In this work, we characterize the packing rearrangements that need to be modeled to move a template closer to the native structure.</p><p>Current template based methods rely on variations of backbone based chain assembly methods8–12. In the constraint based approach8,9, template structures are used to define short and long range distances that act as constraints on atom positions. A simulated annealing procedure is then used to generate backbone models, which can be scored or clustered and scored, before rebuilding and packing the side-chains. In the fragment-based approaches10–12, a threading procedure is used to identify template structure, and these templates are then used to identify short peptide fragments. These fragments are sampled and reassembled to search the backbone conformational space. What is common between these methods and the majority of protein structure prediction approaches is the sampling of conformational space based solely on the protein backbone and without direct influence of the amino acid side-chains3,13,14. Information about the location of the side-chain centers of mass is retained indirectly in the backbone fragments, but the move space is fundamentally defined by the backbone fragments. This approach seems at odds with the basic idea behind template based structure modeling: because the proteins share the same fold, it is the changes in sequence and therefore the contributions from amino acid side-chains that determines the differences between template and native protein structures. Sampling protein conformational space of packing using backbone move sets may account for the current lack of progress in template-based modeling.</p><p>In contrast to these backbone based-methods, our group has been actively seeking approaches to capture the side-chain influences on the protein backbone15–19. In particular, we have shown that cliques in the contact graphs of proteins (i.e. sets of residues where all residues contact all other residues) can be geometrically clustered within and between protein families19. These clusters represent tertiary packing motifs that are common to all protein structures. For the set of structures from the 9th CASP9 experiment20, we combined this side-chain approach with other methods such as loop modeling17 and conformational scoring21 developed in the group to make template based structures predictions. Our approach seeks to make moves directly in the packing of the templates, and allow the perturbations in the side-chain packing to define the position of the backbone. The tertiary motifs also allow us to precisely analyze the results of our prediction method and exactly characterize the side-chain perturbations necessary to move a template towards the native protein structure.</p><!><p>The Stone Soup template based structure prediction procedure is an agglomeration of a number of novel methods that approach different parts of the template based structure prediction problem (Figure 1). The workflow is described generally in the following paragraph and more detail to each individual method is given below. At the start, the sequence provided by the CASP 9 organizers was compared to all known protein structures using PSI-Blast22. To identify templates, a cutoff of 20% sequence identity was used with a coverage cutoff of 90%. If multiple templates were identified, they were aligned to one another using MUSTANG23. This alignment was then used in a profile-profile alignment to the target using MUSCLE24. The aligned structures were averaged to produce a starting template structure, which was then analyzed by our novel tertiary structure prediction (3SP) method that defined regions for core refinement and the remaining for loop modeling: essentially regions with few or no 3SP constraints. For the core of the starting averaged template structure, 3SP was used to identify cliques in the templates, find similar cliques from known protein structures, and statistically model those set of cliques. Side-chain driven backbone samples were drawn from the 3SP distribution of residue cliques and substituted into an average template structure. Since this perturbation of the templates broke the backbone connectivity, all-atom models were generated using Pulchra25. Models were scored with the high resolution molecular dynamics (hrMD) derived volume (see below) and torsion angle score for every second structure, and the top scoring structures were identified to combine with modeled loops. Ranging from 56 to 368, the number of top scoring structures selected depended primarily on available computational resources since the number and size of targets as well as their loops varied over the course of prediction season. For the loops, the new approach of CorTorgles26 uses template data to model backbone ϕ,ψ distributions in unstructured regions of the proteins. Samples from these distributions were converted to loops in Cartesian coordinates using the SNerf algorithm27. Loops were filtered out if their C-terminal stem α-carbons were greater than 2 Å of the template C-terminal stem when the N-terminal stem α-carbons were aligned (non-closure) or if any loop α-carbon were within 3.76 Å of any protein α-carbons (backbone overlap, class score). The remaining loops were built onto the best scoring structures from the core refinement. The loops were further filtered according to the bridging water score WatLoop, a new water path based scoring function described below. Because completion of the WatLoop score did not occur until halfway through the prediction season, this scoring step was only applied to the latter half of the targets. Combining the best core and loop structures, complete all-atom models were again generated using Pulchra25 and scored using hrMD. Complete structures were built by selecting the best scoring set of loops for each 3SP structure identified in step 3. All these loops were then combined on each 3SP structure. Each 3SP structure was considered independently, so different base structures could have different sets of best scoring loops. All-atom models of the complete structures were once again generated using Pulchra25, then steepest descent minimized for 1000 steps using the OPLS force field28 in Gromacs29 and subjected to a final scoring using hrMD.</p><!><p>The underlying concept of 3SP is to drive backbone perturbations based on the interactions of side-chains. This is accomplished by creating a move-set library that relates side-chain packing variations in Cartesian space to the ϕ,ψ torsion angle space of the backbone mainchain. This library is generated by clustering the maximal contact cliques30 computed from the 95% sequence unique ASTRAL31 set of known protein structures (hereafter referred to as move set cliques) based on the relative positions of their Cα atoms and side-chain centers of mass (centroids)17. These move-set cliques represent the maximally self-interacting clusters of residues (all residues in the set are in contact with all other residues in the set). For these clustered packing cliques, the distributions of Cα and centroids at each residue position are modeled using a kernel density estimation approach17. The distribution of a given centroid position for a packing clique is a mixture of trivariate normal distributions centered on the centroid locations of known cliques. The model also permits straightforward conditional sampling, allowing perturbations at a single clique position to be propagated to other positions. To properly model the residue cliques, this statistical modeling is applied in 2 steps: first the side-chain centroid positions are modeled with respect to each other and then individual side-chain centroid positions are modeled to their respective residue's Cα position as well as backbone ϕ,ψ torsion angles.</p><p>To select a specific set of 3SP moves for a particular target, the residues for core refinement need to be identified. From the averaged template structure, maximal contact cliques30 are first computed. These template cliques are compared to pre-calculated library of clustered move-set cliques. The move-set cliques that are within 1.2 Å RMSD of the template cliques are pulled for modeling and are further filtered according to the distances between the Cα atoms and centroid for individual residues to ensure that the modeled positions are consistent with the target sequence. For each selected move-set clique, the modeled 3SP distribution of 1000 side-chain positions to the backbone Cα position and torsion angles constitutes the sampling of core repacking. The set of these distributions represents the overall set of moves from which draws are taken during the template-based modeling. A 3SP move consists of making draws first from the centroid distributions, and then obtaining the respective Cα position conditioned on the centroid draws. In this way, the selected side-chain positions inform changes in the backbone structure. These positions were used to build up the model structure as described above.</p><p>During the template based modeling, a model's clique is selected at random and the positions of its α-carbon and centroid atoms are changed to those of a randomly selected draw from the 3SP distribution for that clique. The move is accepted if it results in Cα-Cα distances less than 4.4 Å for consecutive residues (the maximum distance observed in the PDB) and if there are no overlaps between centroid atoms (as determined by the minimum observed distance between pairs of centroids of the 20 amino acid types in the PDB). For the next move, a new clique that shares at least two residues with the previously moved clique is selected and moved as described. If no overlapping cliques that were not moved in the previous two steps are found, a new starting clique is chosen at random and moved as described. A single run consisted of 5000 steps.</p><!><p>Contiguous segments of two or more residues with no modeled cliques were modeled with our loop modeling algorithm CorTorgles26, which applies a statistical estimation of continuous backbone ϕ,ψ distributions18. The ϕ,ψ angles for the loop region to be modeled plus two flanking residues on each side are calculated from all the templates (total residues = n+4). As described in detail18,26, these are used to fit the parameters of a Dirichlet process mixture of bivariate von Mises distributions centered on a hidden Markov model that describes a continuous distribution in ϕ,ψ space. This unique centering distribution allowed us to develop informative template based conformation distributions even at alignment positions with little or no observed data, which allowed us to cope with sparse data and effectively extend a homology modeling approach to loop regions. Samples from this distribution are converted into Cartesian coordinates by building from the template backbone N, Cα, and carboxyl C atom positions of the first residue in the modeled segment (N-terminal stem)27. The resulting positions of the Cα atoms of the last two residues in the segment (C-terminal stem) are then compared to their positions in the template. If the average of these distances is greater than 2 Å, loop closure is not satisfied and the loop is rejected. Accepted loops are built onto low scoring 3SP structures by aligning the four stem residues. Loops with backbone clashes are eliminated by requiring at least 3.76 Å between non-sequential Cα atoms.</p><!><p>The hrMD scoring functions is based on an extensive set of molecular dynamics (MD) simulations or dynameome of conformations around the native state ensemble21 and compares main-chain torsion angles, side-chain volume, and side-chain torsion angles of individual residues to values observed in the aforementioned molecular dynamics simulations. Volumes were calculated using Voronoi polyhedra32. In order to obtain the volumes of surface residues, the protein was inserted in an equilibrated water box in 10 randomly selected orientations and the resulting residue volumes were averaged. The hrMD score is a unified probabilistic scoring function incorporating a) distribution of exposed polar groups (eSA), b) backbone dependent residue volumes (v) and c) backbone dependent χ1 angles (abbreviated as below). These were calculated from the native state dynameome described above. The unified score Su is proportional to how plausible each candidate structure (denoted by ) given the sequence information. With the assumptions of independence between individual amino acids (denoted by aai), and the nature of research question that we are comparing the candidate structures for the same sequence, the scoring function is reduced to the following form: (3)∝∏i=1nP(ϕi,ψi,eSAi,χi,vi,aai)=∏i=1nP(vi|ϕi,ψi,aai)P(χi|ϕi,ψi,ai)P(eSAi|aai)</p><p>The proposed unified score Su is in the log scale of this estimated probability due to the sparsity of the knowledge space, (4)Su=∑i=1n[logP(vi|ϕi,ψi,a)+logP(χi|ϕi,ψi,ai)+logP(eSAi|ai)]</p><p>One issue that does bear consideration is the computation time involved in calculating the volume portion of the hrMD score. The calculation of Voronoi polyhedra is a much faster algorithm than other volume calculation algorithms, but the overall calculation is still quite slow, especially since a small solvation shell is added to the protein. The computational expense of this calculation could be minimized by introducing a clustering scheme and only scoring cluster centers or by only using the hrMD score in the later stages of prediction.</p><!><p>The population density of the distances between polar groups connected through hydrogen bonded water network on the loop region obtained from the dynameome data was used as reference data21. The 0.1Å bin was used for distances ranging from 1.7 to 12.0Å to give total 104 bins. For each distance bin, the path was counted as one if it connected from loop to loop, and as a half if it connected loop to helix or sheet. The frequencies were normalized, transformed so the highest peak has 1.0 and lowest valley has -1.0 as score. The candidates are put in the water box obtained from an MD simulation, and then the waters within 1.4Å from protein surface are removed. Hydrogen bonded water paths between two polar groups in the loop region are searched and counted. Water paths connected through up to six waters are considered to be consistent with the reference data. This polar group distribution was compared with the dynameome data and scored. Since the water structure around the protein is not equilibrated, we rotated the candidate structures three times along each x, y, and z axes and searched the water paths for each rotation. The ten scores are averaged and this averaged score was used to select the good loop structures on the protein after adding the loop to the protein core structures.</p><!><p>Analysis of differences in packing between templates and the native target structure was performed using the contact order defined packing cliques as described previously17. For each template structure, a sequence/structural alignment was performed using MUSTANG23. At the simplest level, the number of residues in packing cliques were compared for equivalent positions. Then, the packing clique class based on the contact order classification system was compared between equivalent sites for the template and native target structure. For example, a template packing clique of 3+1 (3 local residues packed against a non-local residue) would not be the same as a native packing clique of 2+2 (2 local residues packed against 2 non-local residues), and this would be considered a change. Lastly, cliques were compared based on position of residues in space in a similar fashion to what was done with 3SP to define the move set using a 4.4 Å RMSD cutoff between Cα atoms.</p><!><p>The Stone Soup template based structure prediction algorithm was used on 59 CASP9 targets, of which native structures were released for 45. A breakdown of the Stone Soup results is shown in Table 1. Predicted targets ranged in amino acid length from 79 to 611 residues. These targets had from 1 to 67 templates in the PDB. These templates had between 61% to 100% coverage by our packing cliques. With 100% coverage, there were targets with no loops, but there were also targets with up to 21 loops. Template CαRMSD ranged from 1.74 to 19.32 Å, while final CαRMSD values from the closest of the 5 Stone Soup predictions is from 2.13 to 19.32 Å. As shown by the open diamonds in Figure 2, the averaged template structure in general moved the starting template structure away from the native structure. While this was a major source of error and affected overall performance, it did not significantly impact the sampling and selection capabilities of our approach. Therefore, the discussion will focus on the particular results from the 3SP and CorTorgles components.</p><!><p>For each target 56 to 368 minimized models with all loops were generated, depending on the number of processors available (Table 1). Yet, the diversity of structures generated in 3SP was generally low. The short vertical bars around the unity line in Figure 2 indicate that 3SP sampled conformational space only around the starting template structure. The skew in the distributions above the unity line shows the sampling was more away from the native structure than towards it. This is partly attributable to our sampling scheme, which was essentially a random walk, with no scoring function to allow us to keep "good" moves and reject "bad" moves. However, the more fundamental issue is that our method is too conservative in its move-set since it relies heavily on information from the template structure. Because the 3SP repacks the conserved residues found in the template core, the approach does not contain those new clique conformations that make the difference between that template and native structure. Even if we used the closest template, our sampling of packing space remains close to the starting structure. As shown in Figure 2, some targets began with that closest native template, but this did not improve our sampling. To test that limitation of move-set library, model structures for each target were built up by selecting the cliques from our library that were closest in CαRMSD to the template cliques. The dotted line in Figure 2 shows the average improvement of about 0.2 Å CαRMSD to native, which lies just below the unity line. The best improvement found was just under 0.5 Å CαRMSD and our worst was an increase in CαRMSD from native of 1.2 Å. This increase was due to small lever arm effects in a region where the native had strained backbone torsion angles nearer to disallowed regions. This result shows that the limit of this 3SP approach is not a significant improvement over the starting structure.</p><!><p>For all but two targets, there were regions of the protein that could not be modeled by 3SP. This could be due to lack of coverage in the templates, unpacked residues, or rare packing arrangements that were not well represented in the PDB. We consider all of these cases as candidates for loop modeling and modeled the ϕ,ψ distributions for these residues using the DPM-HMM method detailed in Lennox et al, 201018. Once the ϕ,ψ distributions are calculated, making draws and building putative structures is very fast, allowing us to generate 1,000,000 models for each loop region. However, the stem filter, which enforces loop closure, removes the vast majority of these loops and only a small fraction could be grafted back onto the template structure (Figure 3A). The median number of remaining loops after applying the stem filter is 572. These loops are then built onto the best scoring 3SP models and a backbone clash filter is applied, leading to a further reduction in the number of structures that must be scored. As Figure 3B shows, it was difficult to improve on loops of shorter length from 5 to 7 residues. In this regime, the starting loops usually began very close to native leaving little room for improvement. At the other end of the spectrum, longer loops of 16 and up were not sampled well by CorTorgles, and increases in CαRMSD to native is seen. Cortorgles exhibits its best performance from 8 to 15 residues, where improvements to the overall template by the loops pushed 0.3 Å. As the template put constraints on the starting and end points of the loop as well as the path it takes to connect those points, this improvement is a significant contribution to building better models for loops in the 8 to 13 residue length range.</p><!><p>Addressing the effectiveness of the scoring function used in this study is difficult as we generally did not move the templates significantly closer to or further from the target (Figure 2). Using just the hrMD scoring function to select from the distribution of model structures on average selected structures on average that were just slightly worse than the starting template structure. The long dashed line in Figure 2 shows this average and indicates that hrMD performs consistently regardless of how close the template structure starts to the native. The hope was that the using the molecular dynamics data would be able to discriminate structures that were closer to native, which is not the trend shown by the line. As scoring function based on physical principles of a protein structure, these results are consistent and suggest the limitation of a score like hrMD. Since the hrMD scores structures on their physical reasonableness, deviations that unfold or perturb a fold would not be allowed. For structures far from native that require large rearrangement, hrMD would score movement away from the template structure poorly. In a similar manner, the hrMD only selects close structures with templates that are nearer to native. Therefore, the hrMD scoring function is good at keeping the structure stable, but inappropriate for sampling across conformational space.</p><p>Adding the water path distance filter (WatLoop) displayed a small improvement over hrMD alone, so the improvement of the template that WatLoop provided was investigated for the loops. Figure 2B shows the fitted line to the average improvement of loops selected by WatLoop. Consistent with the ability of the Cortorgles to make loops, the WatLoop was able to generally find the better candidates. In 39% of the 213 loops modeled, WatLoop selected loops that were farther from native. In the remaining 61%, WatLoop was able to select loops that moved the structure towards the native structure. In 3 instances, WatLoop was able to select the best loop made by Cortorgles. In each of these, the starting structure was below 3Å CαRMSD to the native. Since the WatLoop scoring function relies on the network of waters around the loop and its respective structure, the WatLoop scoring function is promising for refinement in template based structure prediction when the template structure is close to the native structure. Alternatively, these results suggest that WatLoop should be used in the final steps of model sampling when the model structure is hopefully closer to the native structure to allow the WatLoop discrimination.</p><!><p>Overall, the templates possessed a certain amount of variation in their packing from the samples and representation in the PDB as explored by Figure 4. Even for templates that are geometrically similar to the target (Cα / side-chain center of mass CαRMSD <4 Å), the fraction of cliques that are identical in target and templates is always less than 60% and may be as low as 30% (Figure 4A). As expected, clique conservation is even lower for poorer templates. In contrast, the structural similarity between cliques that are conserved is high regardless of the similarity between target and template (Figure 4B). Thus, if regions of low clique conservation could be predicted, we could be confident that the remaining regions provide a good template for packing in the template. Another issue that may affect a packing centered template based modeling approach is the completeness of the PDB in describing different packing arrangements. This issue is investigated by considering the number of representatives in the filtered PDB set for each target clique in all CASP9 targets. The 3 Cα / side-chain center of mass RMSD cutoffs were considered for defining representatives. At the shortest cutoff, 0.5 Å, the PDB set appears to be quite incomplete. Three in four target cliques have fewer than 10 representatives in the PDB that are <0.5 Å RMSD. The PDB appears to be much more complete when a 1.0 Å RMSD cutoff is used, with less than 5% of cliques having fewer than 10 representatives. At a 1.5 Å RMSD cutoff, the PDB is essentially complete. While the incompleteness of the PDB at the 0.5 Å RMSD cutoff suggests a lower limit on the resolution of packing based approaches to TBM, the clique conservation issues discussed above represent a much larger practical challenge.</p><p>A simple analysis of the types of cliques that are conserved provides some insight into what regions of the protein are more likely to be different in the target and template (Figure 5). Cliques formed by four residues are by far the most common in the PDB. These also appear to be the most likely to be conserved between target and template. Two and three residue cliques are poorly conserved in the CASP 9 targets (Figure 5A) by a significant margin. Even when they are conserved, three residue cliques are less likely than larger cliques to be geometrically similar to the template clique (Figure 5B). Taken together, these observations at first suggest that regions of the protein with many two or three residue cliques are poorly packed and more sensitive to changes in sequence. Based on a new analysis34, isolated changes in two and three residue cliques occurs in less than 9% of the cases for two and three body cliques. The 91% majority of changes in two and three body cliques results from repacking and rearrangements of three and four body clique packing. Cliques can also be classified according to the backbone connectivity of the residues in the clique.</p><p>In Figure 4C and 4D, we classify four residue cliques as local if all residues are near each other in primary sequence, secondary if they all come from the same element of hydrogen bonded secondary structure (i.e. consecutive turns of a helix or neighboring strands in a sheet), and tertiary if one or more residues is neither local nor hydrogen bonded to the other members of the clique. We find that 4 residue local cliques are the least conserved and the least structurally similar. At 82%, most of these local cliques are found in regions classified as loops. The remaining 18% are those local cliques that start in defined secondary structure and extend into loop regions. Secondary cliques defined by relatively rigid secondary structural elements are more likely to be conserved than tertiary cliques, which are more sensitive to changes in the detailed orientation of different secondary structure elements.</p><!><p>Stone Soup, a novel side-chain based packing algorithm coupled with a new loop modeling protocol, was tested against the CASP9 set of template based homology modeling targets4,33. An analysis of Stone Soup's performance indicated that the scoring functions are limited and the approach's move set is overly conservative. As a physical scoring function, hrMD restricts a model structure from sampling into physically unreasonable regions of conformational space. The WatLoop function shows promise, but requires the core structure to be close to the native (< 3Å) to perform well. For the move set, larger perturbations from the template structure need to be included, because template and target side-chain packing contacts tend to differ significantly even if the sequences are close homologs. Therefore, improvement requires the prediction of regions packing rearrangements between template and native structure, which has been a core problem in template based structure prediction.</p><p>To this end, a recent development in our group has the characterization of a basic principle underlying protein packing of knobs into sockets that allows us to identify the amino acid code for protein structure34. The knob-socket construct allows us to identify the exact changes in packing between the template and native structure that need to be modeled. In effect, sockets identify regions of protein structure that will form or not form interactions with other parts of the protein. If there is an interaction, the socket packs with a socket. So, perturbations in structure that either create or remove tertiary packing can be identified and then move sets can be sampled using 3SP. Inclusion of the knob-sockets and the corresponding amino acid code has great potential and should improve our predictions.</p><!><p>This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.</p>

PubMed Author Manuscript

Multiple Proton Confinement in the M2 Channel from the Influenza A Virus

The tetrameric M2 protein bundle of the influenza A virus is the proton channel responsible for the acidification of the viral interior, a key step in the infection cycle. Selective proton transport is achieved by successive protonation of the conserved histidine amino acids at position 37. A recent X-ray structure of the tetrameric transmembrane (TM) domain of the protein (residues 22\xe2\x80\x9346) resolved several water clusters in the channel lumen, which suggest possible proton pathways to the His37 residues. To explore this hypothesis, we have carried out molecular dynamics (MD) simulations of a proton traveling towards the His37 side chains using MD with classical and quantum force fields. Diffusion through the first half of the channel to the \xe2\x80\x9centry\xe2\x80\x9d water cluster near His37 may be hampered by significant kinetic barriers due to electrostatic repulsion. However, once in the entry cluster, a proton can move to one of the acceptor His37 in a nearly barrierless fashion, as evidenced both by MD simulations and a scan of the potential energy surface (PES). Water molecules of the entry cluster, although confined in the M2 pore and restricted in their motions, can conduct protons with a rate very similar to that of bulk water.

multiple_proton_confinement_in_the_m2_channel_from_the_influenza_a_virus

Introduction<!>Methods<!>Classical MD Simulations<!>QM/MM MD Simulations<!>Identification of Excess Protons<!>Potential Energy Scans<!>Free Energy of Translocation of a Cation through the Pore<!>Dynamics of a Proton in the Entry Cluster<!>Energetics of Proton Diffusion through the Entry Cluster<!>Conclusions

<p>Among cations in solution, H+ features by far the highest mobility. The dominant mechanism of diffusion through liquid water is Grotthuss hopping,1 by which a local excess proton jumps from one water molecule to the next, rather than relying on the much slower diffusion of the H3O+ molecule. Proton channels and transporters utilize the Grotthuss mechanism by propagating an excess proton through an ordered water file.2–4 An intriguing example is the M2 proton channel of the influenza A virus, which performs selective proton transport across the viral envelope.5–7 When the virus is surrounded by an environment of pH less than 68—e.g., the endosome of a host cell during endocytosis—protons flow through M2 and acidify the viral interior. This in turn triggers the uncoating of the virus and release of its genetic material, thereby beginning the infection cycle.9</p><p>M2 is both a pharmaceutical target and a rich source of information on the fundamental processes involved in proton transport through biological membranes. Therefore, significant work has been performed to study its mechanism of conduction.10–17 M2 is composed of four identical peptide chains.18–20 As the pH surrounding the virus decreases, the four histidine residues at position 37 become sequentially protonated, until the channel is "activated" at around the 3+ state,10,12 thereby achieving pH-controlled, proton-selective conduction. The transmembrane (TM) sections of the M2 chains,5,21 spanning residues 25 to 46, also assemble into a functional tetrameric channel, 22 identified in the following as M2TM. High resolution structures of the TM region of the M2 channel14,15,23 have provided detailed insights into how the protein's conformational changes may affect conduction. However, it is still unclear whether the charged His37 side chains simply repel each other and open up space for conductive water wires in a pH-controlled fashion ("shutter" mechanism) or instead actively transport protons by relaying them from one side of the His37 gate to the other ("shuttle" mechanism).11,20 Experimental and computational techniques have been employed to try to elucidate which mechanism is employed by M2.12,13 But a conclusive answer has yet to be provided.</p><p>In this paper, we investigate the initial step of the proton translocation pathway, namely the introduction of a proton into the channel in its highest-charged nonconducting state, 2+. As a basis for our MD simulations we used the recent X-ray structure of a functional mutant of the TM peptide (Figure 1),23 which was recorded at a pH 6.5, slightly above the pKa for protonation of a third histidine residue (6.3±0.312), which is believed to activate the channel. This structure shows alternating layers of structured water clusters and protein side chains in the region surrounding the key His37 and Trp41 residues, which are respectively responsible for proton selectivity10,24 and asymmetry in the proton conduction with respect to the direction of the proton gradient.25 On the extraviral side of His37 is the entry cluster, comprising six H-bonded water molecules, four of which are H-bonded to the four His37 residues, arranged as a box (His-box). Understanding the mechanism of proton translocation through this cluster towards the His-box should offer insights into the overall mechanism of proton transport.</p><p>To address the above issues we adopted a pair of computational approaches chosen to allow us to investigate the H-bonding structure surrounding a positive ion in the pore and the structural diffusion of a excess proton through the entry cluster. We first apply classical molecular dynamics to sample the response of the pore water to the presence of a positive cation in the region of the pore ranging from the Val27 to the Trp41 side chains. We choose the methylammonium cation molecule (NH3CH3+), because of its similarities to the transient H3O+ ions involved in the structural diffusion process. We then present a study which directly investigates the structural diffusion of an incoming proton through the entry cluster to the four histidines by ab initio and hybrid quantum mechanical/molecular mechanical (QM/MM) methods. Importantly, the presence of crystallographic water molecules in the channel pore23 allowed us to start with a reasonable molecular picture of possible protonation pathways. Finally, we summarize our results and draw conclusions.</p><!><p>All simulations in this work were performed on the M2TM' peptide which is defined as the M2TM peptide with Gly34 mutated to alanine. This mutation was chosen because it is known to be active (though 40% less active than the wild type),26 and the water clusters observed in the high-resolution structure of G34A described above23 provide us with initial information about possible proton transport pathways. Though there are some quantitative differences in the pH dependence of the activity for the G34A mutant, experimental evidence suggests that it conducts by the same multistep mechanism as wild type M2.26</p><!><p>We have applied classical molecular dynamics simulations to sample various conformations of a methylammonium cation as it passes through the upper portion of the channel. In doing so we hope to investigate the structure of the pore waters surrounding an incoming positive charge. Several key approximations made for the purposes of this study are discussed below.</p><p>The TM region of M2 features a broad configurational ensemble due to a conformational exchange among distinct substates occurring on several different timescales. Each of these substates is preferentially populated at a specific pH condition.14,15,23 This conformational heterogeneity is believed to be relevant to the proton conduction mechanism.14,15,23 Despite the fact that large amounts of experimental and simulation data have provided insights into the nature of these structural transitions, a completely satisfactory picture of the conformational equilibrium of M2TM is still missing. The lack of detailed structural information about this highly dynamical helix bundle poses a challenge to computational modeling as a satisfactory sampling of the relevant conformers would require very long MD trajectories and an apt choice of lipid composition and boundary conditions on the lipid bilayer. Therefore, rather than attempting a complete description of the system, in this work we addressed the less ambitious goal of characterizing the response of water molecules lying in the pore of M2 to an incoming positively charged moiety for a specific bundle conformation, namely the structure at intermediate pH solved by X-ray crystallography.23 We, thus, applied harmonic position restraints with a force constant k = 20 kcal/mol/Å2 to the heavy atoms of the protein backbone to restrict the accessible portion of the configurational space. Due to this approach we do not expect our calculations to provide quantitative predictions concerning the energetics of methylammonium translocation. However, no relevant constriction regions are found within the outer section of the pore (between residues 27 and 34), and molecules as large as the drug amantadine have been detected experimentally in this region without significant alteration of the overall bundle structure.27 Thus, structural information about metastable states of the cation in this region of the pore of this particular conformation are likely to be well represented.</p><p>At the pH 6.5 conditions of crystallization,23 M2 features either 2 or 3 charged histidines, with all of the Nε atoms protonated.12 These protonation states are also reasonable for the presently studied G34A mutant, which features a high similarity in the current-pH profile.26 We have previously verified that M2TM' is stable around the X-ray structure with 2 charged histidines, but evolves towards the low-pH structure15 when 3 histidines are charged.23 Therefore, we initialized the protein with 2 charged histidines, and modeled the entry of the third charge by the methylammonium cation.</p><p>The M2TM' bundle and the methylammonium were solvated in a 54 × 54 × 64 Å3 water box containing three Cl− ions to neutralize the total charge of the system. It is worth stressing that, due to the lack of the lipid bilayer and the presence of only three counterions, this setup does not provide a faithful representation of the full electrostatic environment of the channel. As with the position restraints, this simplification is expected to impact on the energetics of methylammonium translocation, and therefore the potentials of mean force calculated for this model system are not expected to be quantitatively accurate.</p><p>The methylammonium molecule and the four M2TM' protein chains were modeled using the CHARMM27 force field,28 and water molecules using the TIP3P force field.29 All classical MD simulations were performed with NAMD.30</p><p>Periodic boundary conditions were applied, and the electrostatic potential was solved by the particle mesh Ewald (PME) method31 with an accuracy threshold of 10−6, a real space spherical cutoff of 12 Å, and a fast Fourier transform (FFT) grid spacing of 0.8 Å. Lennard-Jones interactions were cut off at 12 Å, with a switching function starting from 10 Å. The equations of motion were solved with the velocity Verlet integrator using a timestep of 1.5 fs. The lengths of all bonds involving hydrogen atoms were kept constrained with the SHAKE method.32 The system was run at 310 K and 1 atm using Langevin temperature33 and Langevin piston pressure34,35 coupling schemes. Decay times for the thermostat and barostat were chosen to be 1 ps and 0.1 ps, respectively.</p><p>We applied external biasing forces to enhance the sampling rate of configurations relevant in the context of a His protonation pathway. This biasing potential was constructed via the metadynamics algorithm,36 using the position of the cation's nitrogen atom along the membrane normal as a collective variable. We made use of Gaussian hills of 0.25 Å width and a height of 0.001 kcal/mol. Hills were added every 0.2 ps until approximately 50 ns of trajectory data was collected. During this time, the methylammonium cation spanned all positions between the Val27 and the Trp41 side chains (Figure 1).</p><!><p>Hybrid quantum mechanical/molecular mechanical (QM/MM) calculations of the M2TM' peptide embedded in a lipid bilayer were performed with the CP2K software package.37 The QM region was defined to include all His37 sidechains (up to Cα) and 7 water molecules, 6 of which correspond to the entry cluster23 plus an additional water on the extraviral side of this cluster as observed in classical MD simulations. All of the Nε atoms of His37 were protonated, along with 2 of the Nδ atoms. A third excess proton was placed in the entry cluster as specified below. Diffusion of QM atoms outside the QM box was prevented with a quadratic confining potential around a 16 Å-edge cubic box with a spring constant k = 4.48 kcal/mol/Å2. However, this force was not necessary in any of the simulations performed.</p><p>The protein was embedded in a membrane of palmitoyl oleoyl phosphatidyl choline (POPC) molecules, and periodic boundary conditions were applied with a 80 Å × 80 Å × 77 Å periodic box. A 15 Å water layer on both sides of the membrane was added. This created a buffer of 30 Å of water between two periodic images of the membrane or of the protein. The total system (QM+MM) contains ~54,600 atoms.</p><p>The QM region was treated at the DFT level of theory in the Gaussian-plane wave (GPW) approximation.38 The BLYP functional,39,40 molecular optimized triple-zeta basis sets with two polarization functions,41 and Goedecker-Teter-Hutter pseudopotentials42 were used. An energy cutoff of 360 Ry was employed in the plane wave representation of the density. A wavelet based Poisson solver43 was used to remove the spurious interactions of the QM region with its periodic images. The QM system box size was chosen such that no atom is within 5 Å of the box edge.</p><p>The His37 side chains were capped with hydrogens, the forces on which were treated via the IMOMM scheme,44 with a scaling factor of 1.5 applied to relate the QM C-H bond distance to the MM C-C distance. The CHARMM forcefield28 was used for the protein and lipid atoms with the TIP3P model for water outside the QM box.29 The electrostatic interactions of the total QM/MM system were treated with the particle mesh Ewald method31 with an accuracy threshold of 10−6 and a spherical cutoff of 8 Å for the real space part. A 12 Å spherical cutoff was used for the Lennard-Jones interactions.</p><p>As in the classical MD runs, two His37 residues were doubly protonated while the remaining two were protonated only on Nε. The membrane, water, and protein atoms were then equilibrated via classical MD for 4 ns starting from the high resolution crystal structure, which we previously observed to be stable on a time scale of tens of ns for this charge state.23 (See Supporting Figure S1 for more details.)</p><p>To model the motion of an incoming proton, we generated 20 different initial configurations of the system generated by sampling configurations from the equilibration MD trajectory, adding a proton to various locations in the entry cluster, and optimizing the structure of the QM region and surrounding residues at the QM/MM level of theory. Each configuration was run with a time step of 0.25 fs for approximately 3 ps (about 60 ps in total). In each simulation, we maintained a temperature of 300 K using the Nosé-Hoover chains thermostat45 with a time constant of 1 ps.</p><!><p>In order to explore the proton motion in the channel it is necessary to identify specific protons as the excess protons. This identification was performed according to a quantitative, two-step scheme previously used in a study of an excess proton in liquid water.46 First a predetermined number of acceptor atoms bound to excess protons are identified. Then, in the event that the acceptor atom is bound to multiple protons, an excess proton is chosen among them.</p><p>In the case of the QM/MM MD simulations described above, three acceptor atoms of interest are identified, corresponding to the three excess charges in the QM subsystem. This is done by assigning a hydrogen coordination number to each potential acceptor atom (7 water oxygen atom and 4 imidazole Nδ in the system). The hydrogen coordination is simply the integer number of hydrogen atoms that are closer to a given acceptor atom than to any other acceptor. The three acceptor atoms with a coordination number corresponding to the value expected for a charged species (3 coordinated hydrogens for a hydronium oxygen, 1 for a charged imidazole Nδ) are identified as the acceptor atoms of interest.</p><p>Then we proceed to identify the excess hydrogen itself. If the acceptor atom of interest is an imidazole Nδ, there is only a single hydrogen associated to it. In the case of hydronium, we associate an asymmetric bond length with each proton, δ = |RO−H − RH−O|, where RO−H and RH−O are the distances from the hydrogen in question to the nearest two acceptor atoms, respectively. The proton with the smallest asymmetric bond length, δ, is identified as the excess proton for purposes of our analysis.</p><!><p>Potential energy scans were performed along a coordinate describing proton transport through the entry cluster to the His-box to investigate the energetics of this process. Two levels of theory were employed. First, the highly accurate MP2 method was employed with the cc-pvdz basis set47 to study the proton transfer in a gas phase cluster constructed to mimic the entry cluster/His-box region of the protein. The cluster includes the entire QM region from the QM/MM calculations above (His37 side chains, 6 entry water molecules, plus 1 additional water) with the addition of the four Ala34-Ile35 peptide groups and the two waters of the bridging cluster. With the inclusion of these groups, all excess protons are provided with a hydration shell. The core (1s) electrons were left uncorrelated in all MP2 calculations to reduce the computational cost. The MP2 calculations were performed with NWChem version 5.1.48 In addition to the MP2 scans, potential energy scans were performed with the QM/MM scheme described above to investigate the effects of the environment.</p><p>The reaction pathway employed in our scans was defined by first locating several stable points in the PES by geometry optimizations at the QM/MM level of theory. MM atoms farther than one covalent bond or hydration shells from the QM region were constrained to their initial positions. Three local minima in the PES were located, which differ in the location of the three excess protons. Intermediate structures between these minima were generated by linear interpolation of the system's Cartesian coordinates. The coordinates of the gas phase cluster used for the MP2 PES were taken directly from the QM/MM optimized geometries. Thus, the gas phase cluster geometries reflect the effect of the protein and lipid environment. Dangling bonds were capped with hydrogens according to the IMOMM procedure described above.</p><!><p>To gain insight into the mechanism of charge translocation across the channel pore we computed the potential of mean force (PMF) for displacing a cation along the direction of the channel axis of symmetry, from the plane containing the four Val27 residues (a constriction region acting as a valve at the N-term side of the channel pore) to the Trp41 side chains, which approximately delimits the entry region on the C-term side (Figure 1). Note that the absence of bond rearrangement effects in classical models precludes the possibility of describing the structural diffusion of a proton through the channel, but this approach allows us to achieve better sampling of the conformational space of the pore waters than would be possible with more costly methods that allow bond breaking. We choose to model the passage of a methylammonium in (NH3CH3+) through the pore because the pattern of H-bonds with its first solvation shell is very similar to that of hydronium.</p><p>In the PMF shown in Figure 2A, the reaction coordinate is the position of the nitrogen atom of the amine group relative to the Cα atoms of the His37 residues. As expected the positive charge localized on the His-box results in high free energy values as the methylammonium approaches the histidines. The energy barriers involved in the methylammonium translocation are much higher than in previous computational studies of conduction of other cations, such as Na+ and NH4+,13,49 possibly because these calculations were based on different structures of M2TM, featuring a much wider pore49,50 compared to the recent high resolution structures.14,15,23 However, it is important to stress that in our calculations, due to the presence of positional restraints on the backbone atoms, the structure of the protein is not allowed to relax as the cation approaches the His-box; therefore we expect our calculations to overestimate the free energy barriers for the passage of the cation through the His-box.</p><p>The most stable states are found within the 10–14 Å interval, corresponding to the region close to the N-term end of the channel pore. A close inspection of the PMF reveals two distinct local minima with a negligible free-energy difference, located approximately at 13 and 10 Å and separated by a barrier of ~2 kcal/mol. The first minimum (around 13 Å) corresponds to configurations in which the nitrogen atom is localized within the Val constriction region and in which, therefore, the amine group is surrounded by the hydrophobic moieties of the Val residues (Figure 2B). These configurations can be regarded as the first stepping stone along the conduction pathway: after the excess proton has approached the channel mouth, the formation of a metastable hydronium moiety within the Val sidechains may establish a connection between the bulk and the pore waters and allow diffusion of the proton via Grotthuss hopping. The second minimum (around 10 Å) represents configurations in which the amine group is located right below the Val sidechains and is completely solvated by the water molecules filling the upper part of the channel pore. This region corresponds to a region of diffuse density detected in the X-ray crystal structure23 (Figure 2C).</p><p>Although very similar in energy, the two minima feature significantly different structural properties. Indeed, analysis of the populations in these two minima reveals difference in the orientation of the C-N bond; configurations where the reaction coordinate falls in the 12–14 Å interval show a consistent orientation of the C-N bond, while those in the 9–11 Å interval show a nearly isotropic distribution (Figure 2E). The almost constant tilt of the C-N bond with respect to the channel axis (~60°) observed in the case of the first minimum likely results from a geometric constraint. In order to exchange H-bonds with water molecules located on both sides of the Val-gate, the amine group needs to project the N-H bonds along both directions of the channel axis, thus the C-N bond is locked in a nearly horizontal orientation. An important implication of the fact that the amine group is stably H-bonded to water molecules on both the sides is that, arguably, a hydronium moiety does not need to be desolvated in order to cross the Val-gate, and therefore this step is not expected to be rate-limiting in the conduction mechanism.</p><p>Another interesting feature of the PMF profile is the presence of a clearly defined local minimum around 4 Å. Although these configurations show a much higher free-energy compared to the previous ones (~20 kcal/mol), the presence of a metastable state is nevertheless remarkable given the close proximity of the methylammonium to the doubly-charged His-box. For these configurations we again observed a well-defined orientation of the C-N bond, which indicates that the amine group is restrained by a structured H-bonding pattern (Figure 2D). Indeed four of the water molecules constituting the entry cluster and another one displaced toward the N-term side of the pore are constantly oriented in such a way that all the dipole moments point toward the positive charge. Such an electrostatic stabilization mechanism may also be at work for other molecules besides methylammonium. All of the molecules known so far to bind the M2 channel-pore are alkyl-amines and, at least for one of them (amantadine), the amine group has been shown experimentally to be located in the same region.27</p><p>As the His-box is further approached the free-energy increases to a value of ~40 kcal/mol, decreasing again only after the His-box region is crossed. However, particularly in this region, we expect our calculations to strongly deviate from the energetics of an excess proton in a realistic system. Indeed, in the context of a structured water cluster and in presence of a dipolar field created by structured water molecules, it is possible that, besides the Eigen-like ones, the hydronium cation can significantly populate other configurations for which methylammonium may not be an adequate model. Therefore, in the next sections we will reconsider the structure and the stability of an excess proton in the entry cluster by adopting a QM/MM approach allowing bond breaking and formation.</p><!><p>A total of sixty picoseconds of QM/MM molecular dynamics simulations were run to identify proton transport pathways from the region near Ala34 to His37 and investigate the ease with which the proton passes through the entry cluster. Because the water cluster and His37 sidechains are treated at the DFT level of theory it is possible to observe structural diffusion of the proton through the cluster. All simulations involved a total charge of 3+ in the His-box/entry cluster region. In all simulations at least two excess protons were assigned to the His-box, and a third excess proton placed at different positions between the entry cluster and the His-box. The free evolution of this free proton was observed.</p><p>The left panel of Figure 3 shows the averaged spatial density of the three excess protons taken from these simulations. Significant proton density is observed in interstitial positions between all neighboring pairs of water oxygen atoms, and adjacent to all four imidazole Nδ atoms.</p><p>An example proton transport event, drawn from a representative trajectory, is depicted in the right panels of Figure 3. The proton moves easily between water molecules via Grotthuss hopping. In the process of this reaction two different forms of the excess proton are observed: Eigen (characterized by an H3O+ cation stabilized by three surrounding hydrogen bond acceptors) and Zundel (H5O2+, characterized by the near-equal sharing of a proton between two neutral water molecules). Protons starting in the water cluster were observed to move to a previously uncharged His37 residue via Grotthuss hopping in sixteen out of twenty simulations. In the remaining four replicas the proton remains in the entry cluster for the duration of the simulation. No protons were observed passing from a charged His37 to the entry cluster, suggesting that the state with three charged His37 residues is relatively stable.</p><p>In a previous computational study, Chen et al. directly simulated Grotthuss hopping through the channel using the multi-state empirical valence bond (MS-EVB) method.13 The MS-EVB method allows for bond-breaking and forming in the pore water molecules, so, like the QM/MM results presented here, the calculations are capable of reproducing structural diffusion. However, unlike the present QM/MM results, bond breaking/forming in the histidine groups was not allowed in the EVB study. Thus, the possibility of shuttling a proton via the His37 residue during translocation was not considered.</p><p>A histogram describing the configuration of the atoms surrounding an excess proton in the water cluster is given in Figure 4. The structure is described by two coordinates: the asymmetry of the hydrogen bonds and the water oxygen-oxygen distance. The maxima of the histogram is at asymmetric configurations which can be characterized as Eigen-like, but about one third of the population is in the region of the plot corresponding to nearly symmetric, Zundel-like configurations around the proton. An excess proton in bulk water was simulated by similar methods and reported by Marx et al.46 A histogram as a function of the same coordinates was reported, and the similarity between that distribution and Figure 4 is striking; nearly the same ratio between Eigen-and Zundel-like configurations is observed in the water cluster constrained in the pore of M2 as was seen in bulk water.</p><p>This similarity suggests that the mobility of the proton in the entry cluster may be comparable to that in bulk water. Figure 5 shows the root mean squared deviation (RMSD) of the acceptor atoms currently bound to the excess proton as a function of time for each of the 20 individual replica simulations. The acceptor atoms move in discrete "hops" of approximately 2, 4, or 6 Å, corresponding to the concerted motion of the excess proton through 1, 2, or 3 acceptor atoms (Figure 5). Rapid "rattling" with an amplitude of approximately 2 Å is observed in some trajectories. This corresponds to Zundel-like configurations where the location of the excess proton cannot be well described by the choice of a single acceptor atom.</p><p>A fit of the data by a curve of the form RMSD=Ct demonstrates that the proton moves through the entry cluster at a rate comparable to the rate of proton diffusion in bulk water. The fitted constant, C = 1:97 Å2/ps, is of the same magnitude of the diffusion constant of a proton in bulk water (0.931 Å2/ps at 25 °C51). The preference of the system for the histidine-bound state indicates that the process observed in our simulations is not unbiased diffusion, and thus it would be incorrect to interpret our fitted value as the proton diffusion constant in the pore. However, it does indicate that despite the ordered nature of the water cluster, the proton moves with a rate comparable to that of diffusion in bulk water.</p><!><p>Based on the pKas of imidazolium and hydronium, the excess proton is expected to be localized preferentially on the imidazole moieties rather than bind to the water molecules in the entry cluster. However, given the striking resemblance of the entry cluster to onehttp://www.google.com/ of the preferred conformations of a gas phase H+(H2O)6 protonated water cluster (in which the excess proton is shared by the two central water molecules in a barrierless fashion52–54) it is worth investigating the relative energies of these states.</p><p>To this end and to quantify the height of the energy barriers in proton diffusion, we have calculated a potential energy surface along a proton transfer reaction path following the Grotthuss hopping of a proton through the entry cluster—from the apical water dimer, through the lower (four-water) plane in the cluster, and finally to one of the singly protonated His37 residues. The three QM/MM-optimized minima along this path are illustrated in Figure 6A; two of these minima correspond to states where a hydronium (Eigen) cation is present in the entry cluster while the third stable configuration has three charged His37 residues and an uncharged entry cluster. The process thus encompasses two proton transfers, the first between two water molecules, the second from a water molecule to a histidine. The PES along this path is shown in Figure 6B. The resolution of our calculations allows us to see not only the energies of stable states along the reaction path, but also to estimate the heights of the barriers to hopping between them.</p><p>The PES calculated at the MP2/cc-pvdz level of theory spans an energy range of 3 kcal/mol, with a maximum barrier of 3 kcal/mol (Figure 6B), consistent with the fast diffusion of protons across the water molecules of the cluster observed in our MD trajectories. Particularly noteworthy is the unexpectedly low energy of the state in which a hydronium is located adjacent to the neutral histidine, which is attributed to the local electrostatic environment. The dipoles stabilizing the cation (the His37 side chain and the peptide group of Ala34-Ile35, pictured in Figure 6C) are even larger than those provided by H-bonded water molecules, which themselves are known to stabilize a hydronium state by tens of kcal/mol.55 Another source of stabilization is that the positive charge keeps a distance of ~5.5 Å from the histidine bound protons when bound to this water. This is farther than the His-His distance in a purely His-bound 3+ state (4.5 Å) but still shorter than the 6.5 Å diagonal distance between charged His residues in the 2+ state.</p><p>The inclusion of the remainder of the protein-lipid-water environment alters the stability of the hydronium states relative to the His-bound ones. A QM/MM model based on a DFT-BLYP description of the system predicts a total energy drop of 2 kcal/mol over the course of the path (Figure 6B). However, the barriers are not greatly affected; they are never higher than 4 kcal/mol.</p><p>In both the QM and QM/MM calculations the overall energy of transfer is small, indicating that the reactant complex—with the third proton residing on the water cluster—is nearly as stable as the product complex with a neutral entry cluster and a triply protonated His-box. The entry cluster thus enables permeant protons to diffuse to the charged His-box without encountering a large energetic barrier or falling into any deep energy wells. The nearly flat energy landscape suggests that the geometry of the entry cluster and of the local protein environment plays an important role in stabilizing a charged 2+ and 3+ state in a highly restricted nano-environment embedded in a bilayer; through the entire Grotthuss hopping pathway the excess positive charge is lined with properly oriented dipoles providing a strong electrostatic stabilization. Moreover the presence of many distinct states arguably results in an entropic stabilization of the excess protons. It is worth mentioning that proton delocalization via nuclear tunneling (not included in our description) might also contribute additional stabilization.</p><p>The results of the above calculations are relevant to the discussion of the overall proton transport mechanism of M2: "shuttle" vs. "shutter." The existence of a small driving force pushing the excess proton towards His37, even in the presence of a +2 charge, suggests a shuttle-type mechanism. Such a mechanism requires the transfer of a proton from a His37 Nδ to its own Nε (tautomerization) or that of a neighboring His residue at one point in the transport cycle. Though our simulations do not directly address this reaction, the relative ease by which protons can move around the His-box suggests that such a reaction may be accessible on a relatively short time scale.</p><!><p>A recent X-ray structure of the TM region of the M2 proton channel unveiled an unexpectedly well-defined configuration of pore water molecules.23 Rather than showing a homogeneous water accessibility, the channel pore hosts several stacked layers of water molecules held in position by an extended network of tight water-water H-bonds. The confinement within the sub-nanometer enclosure of the channel and the small number of H-bonds exchanged with the pore-lining residues result in the formation of cage-like assemblies of waters bearing striking resemblance to some of the most energetically favorable structures of positively charged gas-phase water clusters.52–54 This analogy prompts the appealing hypothesis that the excess protons, transported between the water molecules via Grotthuss hopping during the conduction cycle, may be transiently stored in one of these water clusters filling the channel pore. To address this issue we employed a variety of computational schemes to investigate the equilibrium configurations and the dynamics of excess positive charges in the channel pore.</p><p>We first considered the PMF for displacing a classical hydronium-like moiety, namely methylammonium, from the channel mouth to the His-box along the channel axis. We found a relatively smooth free-energy landscape featuring a broad global minimum located immediately below the channel mouth in correspondence to the Ser31 residues. In this region the cation diffuses freely without showing any tight binding conformation. It is of note that this minimum corresponds to a region (below the Val27 sidechains) of diffused density detected in the X-ray structure.23 Although the overall positive charge on the protein results in a strongly repulsive PMF as the His-box is approached, the free-energy landscape shows a local minimum in the region of the pore near Ala34, which is immediately above the entry cluster. The presence of a metastable state at this location is attributed to the orientation of the dipole moments of the waters and of carbonyl groups of the peptide backbones, which are able to electrostatically stabilize a positive charge in this position.</p><p>Therefore we focused on the entry cluster by performing QM and QM/MM calculations. By allowing bond breaking/formation, we were able to sample all the other accessible configurations of the water-bound excess proton, besides the H3O+ hydronium-like one. In particular, frequent exchanges between Eigen-like(H9O4+)and Zundel-like(H5O2+) configurations were observed during the timespan of our MD (QM/MM) simulations. Despite the confining environment, the equilibrium between Zundel and Eigen species in the entry cluster is indistinguishable from that seen for bulk water,46 therefore very similar kinetics are expected for protonation/deprotonation events of specific water molecules. Indeed, the motion of the simulated excess proton in the entry cluster is comparable to that determined for proton diffusion in bulk water.51</p><p>The nearly-free diffusion of the excess proton across the continuous wire of H-bonded water molecules of the entry cluster is also supported by the potential energy surfaces calculated at the MP2/cc-pvdz and DFT-BLYP levels of theory. Indeed, the potential energies along a putative proton transport pathway, in which an excess proton is displaced from one of the two apical waters of the entry cluster to one of the neutral histidines, reveals the presence of several local minima separated by small barriers (compared to kBT).</p><p>In summary, the water clusters observed in the lumen of the M2 channel of the influenza A virus are able to sustain a highly diffusive behavior of the excess proton. Therefore proton conduction from the channel mouth to the crucial amino acid His37 is likely to occur in a very similar fashion to proton transport in bulk water.</p>

PubMed Author Manuscript

A Modular System for the Synthesis of Multiplexed Magnetic Resonance Probes

We have developed a modular architecture for preparing high-relaxivity multiplexed probes utilizing click chemistry. Our system incorporates azide bearing Gd(III) chelates and a trialkyne scaffold with a functional group for subsequent modification. In optimizing the relaxivity of this new complex we undertook a study of the linker length between a chelate and the scaffold to determine its effect on relaxivity. The results show a strong dependence on flexibility between the individual chelates and the scaffold with decreasing linker length leading to significant increases in relaxivity. Nuclear magnetic resonance dispersion (NMRD) spectra were obtained to confirm a tenfold increase in the rotational correlation time from 0.049 ns to 0.60 ns at 310 K. We have additionally obtained a crystal structure demonstrating that modification with an azide does not impact the coordination of the lanthanide. The resulting multinuclear center has a 500% increase in per Gd (or ionic) relaxivity at 1.41 T versus small molecule contrast agents and a 170% increase in relaxivity at 9.4 T.

a_modular_system_for_the_synthesis_of_multiplexed_magnetic_resonance_probes

Introduction<!>Results and Discussion<!>General Synthetic Methods<!>(7) Tri-tert-butyl 2,2\xe2\x80\xb2,2\xe2\x80\xb3-(10-(3-azido-2-hydroxypropyl)-1,4,7,10-tetraazacyclododecane-1,4,7-triyl)triacetate<!>(1) 1-(3-azido-2-hydroxypropyl)-4,7,10-tris(carboxymethyl)-1,4,7,10-tetraazacyclododecyl-gadolinium(III)<!>(8) Ethyl 5-(2,4,6-tribromophenoxy)pentanoate<!>(9) Ethyl 5-(2,4,6-tris((trimethylsilyl)ethynyl)phenoxy)pentanoate<!>(10) Ethyl 5-(2,4,6-triethynylphenoxy)pentanoate<!>(2) 5-(2,4,6-triethynylphenoxy)pentanoic acid<!>General procedure for click chemistry<!>(3) 5-(2,4,6-tris(1-(2-hydroxy-3-(1H-1,2,3-triazol-1-yl)propyl)-4,7,10-tris(carboxymethyl)-1,4,7,10-tetraazacyclododecylgadolinium(III))phenoxy)pentanoic acid<!>(11) Ethyl 5-(2,4,6-tris(1-(2-((2-(1H-1,2,3-triazol-1-yl)ethyl)amino)-2-oxoethyl)-4,7,10-tris(carboxymethyl)-1,4,7,10-tetraazacyclododecylgadolinium(III))phenoxy)pentanoate<!>(12) Ethyl 5-(2,4,6-tris(1-(2-((2-(1H-1,2,3-triazol-1-yl)propyl)amino)-2-oxoethyl)-4,7,10-tris(carboxymethyl)-1,4,7,10-tetraazacyclododecylgadolinium(III))phenoxy)pentanoate

<p>Magnetic resonance imaging (MRI) is a widely used modality in experimental research and clinical diagnostics.2 The intrinsic contrast of MR images can be augmented by the use of MR contrast agents (CAs) that are derived from two primary classes: superparamagnetic particles3–5 and paramagnetic chelates.6 The relaxivity (mM−1s−1) of a paramagnetic contrast agent reflects its ability to shorten the T1 relaxation time of water which results in a brighter MR image. The higher the relaxivity, the more sensitive the agent.6,7</p><p>During the last decade there has been a surge in the development of new gadolinium-based MR contrast agents. These efforts have focused on signal amplification, targeting, and bioactivated or responsive probes.7–22 Recently, multimodal agents have emerged that enhance MR contrast and simultaneously provide the ability to image the target with other techniques (fluorescence microscopy or positron emission tomography)12,23 for in vivo targeting and fate mapping of cells.8,22</p><p>The relaxivity of MR contrast agents depends heavily on magnetic field strength.24 At low magnetic fields (1.5 to 3 T) the relaxivity of small molecule contrast agents are limited by their short rotational correlation time (τR).7 As shown in Equation 1 an optimal relaxivity occurs when the correlation time (τc1) of the Gd(III) CA is equal to the inverse of the Larmor frequency (1/ωI) of the proton. This correlation time contains contributions from three process, τR the rotational correlation time, τM the mean water residence lifetime, and T1e the electronic relaxation time. By attaching a small molecule CA to a macromolecule (e.g., protein, polymer, nanoparticle, viral capsid) the molecular tumbling decreases resulting in a longer τR and a subsequent increases in relaxivity.7,25–32 Although macromolecular systems have shown significant promise as high relaxivity CAs at low field, they are difficult to characterize and are frequently polydisperse.31,33 For macromolecular CAs, an increase in relaxivity is observed between 0.5 and 1.5 T due to the field dependence of T1e. The increase in relaxivity is due to the T1e contribution to τc1 resulting from the long τR of the macromolecular CAs. However, this increase is followed by a rapid decrease in relaxivity as the field strength increases beyond 1.5 T. This effect results from the long τc1 value which shifts the onset of the relaxation dispersion to smaller proton Larmor frequencies.34 The need for increased sensitivity for molecular and cellular imaging in research laboratories has driven the development of high field MR systems.24,35 Caravan and coworkers demonstrated that as field strength increases the optimal value for τR decreases, from ≈ 20 ns for low field to ≈ 0.5 ns for high field magnets.24</p><p>To address the need for contrast agents that can operate effectively in the high field MR instruments and simultaneously provide the option to install multiple modalities, the development of a synthetically flexible contrast agent core is required. Toward this goal we have designed a new class of contrast agent conjugates that are modular, monodisperse, exhibit high relaxivity, and can be subsequently modified (Scheme 1).</p><!><p>Several examples of multimeric conjugates with intermediate τR values between (0.5 and 4 ns) exist in the literature.22,36 For example, the work of Martin and coworkers is an early example of a modular agent that has been modified with a fatty acid to bind to BSA.37 Ranganathan and coworkers have found a 25% increase in relaxivity when the number of freely rotating atoms in their linker was decreased.38 Additionally a 40% increase in relaxivity was seen by Henig and coworkers in their silsesquioxane contrast agents by changing from an ethylene to a benzene linker.39 The multimeric complexes of Livramento and Kotková demonstrate very high relaxivities at 20 and 60 MHz due to increased size and hydration number.40–43 These last two examples have shown promise as high field agents but do not possess further functionality for targeted or multimodal imaging projects. The properties of these agents are compared in Table 1. In related work with macromolecular agents Zhang and Nair use peptide tetramers designed to bind to human serum albumin (HAS) or fibrin and demonstrate that multilocus binding results in an increase in relaxivity due to increased rigidity.44–46 Avedano and Datta show that increasing the rigidity of their linkers (alkyl vs aromatic) to HSA and viral capsids respectively increased relaxivity by more than 50%.29,47 The works of Rudovsky and coworkers demonstrate the importance of internal motion in dendritic systems by utilizing ion paring between the PAMAM dendrimer and a polyarginine to reduce internal motion and increase relaxivity.48,49</p><p>We have previously demonstrated that the triazole linkages formed by click chemistry are effective in generating multimeric CAs with relatively high relaxivities as a result of increased linker rigidity.22 Here, we describe three new azide-functionalized chelates and a new phenol-based scaffold with three alkyne groups to take advantage of facile click chemistry. The alcohol functional group on the phenol scaffold is orthogonal to click chemistry and can be used in subsequent modification.</p><p>The new macrocyclic chelates 1, 4, and 5 were designed to investigate the effect of linker length on relaxivity (Figure 1). Preparation of begins with a ring opening of epichlorohydrin with sodium azide to provide 1-azido-3-chloropropan-2-ol 6. Addition of 6 to the tri-substituted macrocycle (tris-t-butyl-DO3A) provides the protected ligand 7. Deprotection of the t-butyl esters with formic acid at 50 °C followed by metalation provides contrast agent 1 after HPLC purification (Scheme 2). Chelates 4 and 5 were prepared from tris-t-butyl-DO3A-monoacetic acid and N3(CH2)nNH2 through peptide coupling where n = 2, 3 see (Scheme 1-SI). These chelates were metalated using Gd(OAc)3 and purified using reverse phase HPLC.</p><p>Complex 1 was characterized by X-ray crystallographic analysis (v:v:v, 1:1:1, mixture of water, acetonitrile, and acetone) and revealed no significant differences between the chelate structure of 1 and [Gd(HP-DO3A)(H2O)]. (Figure 2).50 The O8-Gd bond length is consistent with that reported for [Gd(HP-DO3A)(H2O)] (ProHance®) indicating that the overall structure is not perturbed by the addition of the azide at C16 and the coordination geometry of 1 is similar to [Gd(HP-DO3A)(H2O)].</p><p>Preparation of scaffold 2 is shown in Scheme 3. The synthesis begins with the alkylation of tribromophenol with ethyl 5-bromovalerate in DMF with K2CO3 overnight at 50 °C to form 8 in high yield. A Sonogashira reaction using 8 and trimethylsilyl (TMS) acetylene provides the protected trialkyne 9. The TMS groups are removed using KF in ethanol and the ethyl ester is saponified in dioxane/water with NaOH to form the desired scaffold 2 (Scheme 3). This scaffold includes a carboxylic acid that can be functionalized before or after the click chemistry to form the final conjugate.</p><p>A series of agents (3, 11, 12) were synthesized with 1, 3, or 4 methylene groups between the scaffold and the chelate. Click chemistry was used to conjugate 1, 4, and 5 to scaffold 2 using standard conditions resulting in an observable correlation between relaxivity and linker length. The longest linker (12) has the lowest relaxivity (7.3 mM−1s−1 at 1.41 T) while the shortest linker (3) exhibits the highest relaxivity (15.4 mM−1s−1 at 1.41 T) (Figure 3). This result is not surprising considering that increasing the rigidity of the linker between the scaffold and the individual chelates limits local rotation. This effect has been shown to increase relaxivity in other systems.22,26,38 It should be noted that τm could be an additional limitation to the relaxivity of (11) and (12) due to the monoamide ligand structure of these chelates.</p><p>[Eu(HPN3DO3A)(H2O)] was synthesized to confirm that the addition of the azide to the structure did not perturb the hydration number, q in solution. Fluorescence analysis of [Eu(HPN3DO3A)(H2O)] showed a of 0.90±0.1. Following attachment of [Eu(HPN3DO3A)(H2O)] to scaffold 2 the hydration number was maintained at 0.87±0.1. These results confirm that the addition of the azide and the subsequent click chemistry do not affect the hydration number. The relaxivities of 1 and 3 were measured at 1.41 T and 1 is similar to [Gd(HP-DO3A)(H2O) ](Table 2). At this low field strength, the per Gd relaxivity of 3 is five-fold higher than that of 1.</p><p>To observe the effect of field strength on the relaxivities of 1 and 3, these values were measured at 9.4 T. At this field strength, the relaxivity of 1 is only 10% lower than at 1.41 T, while 3 shows a 70% decrease. This effect is typical of small molecule chelates like 1 with short τR values. In these cases, the relaxation rates are relatively unaffected by field strength. The pronounced decrease in relaxivity of 3 at higher field strength suggests the complex has a long τR. The per Gd relaxivity of 3 at 9.4 T is 170% that of 1. By controlling the size of the CA τR can be regulated to achieve a significant increase in relaxivity.</p><p>To obtain a better estimate of the increase in the τR between compounds 1 and 3, water proton relaxivity was measured from 0.01 to 40 MHz proton Larmor frequency (corresponding to fields of 0.0002 to 0.95 T) for 1 and 3 at 298 and 310 K (Figure 4). The nuclear magnetic resonance dispersion (NMRD) profiles35 of 1 exhibit a single dispersion. This result is ascribed to one water molecule coordinated to the paramagnetic Gd(III) ion, the protons of which are dipolarly coupled to the electron spins with a correlation time corresponding to the tumbling time of a small molecule, and the additional contribution of diffusing water molecules. The relaxivity of 3 is i) much larger than that of 1 and ii) exhibits a dispersion occurring at smaller frequencies. Both features indicate that the correlation time τc is sizably increased.51 A peak in relaxivity appears in the high frequency region, indicating that the correlation time is field dependent, and must be affected by the electron relaxation time T1e. In turn, this result means that for 3, τR is longer than T1e at least at low fields, so that at such fields T1e dominates the correlation time τc1. T1e increases with increasing magnetic field strength and therefore the correlation time τc1 is determined by both τR and T1e in the high field region of Figure 4. The temperature dependence observed in the relaxivity profiles of both complexes indicates that water molecules are in the fast exchange regime. Therefore, their τM is not strongly limiting the relaxivity.</p><p>The profiles for 1 have been fit (solid lines in Figure 4) using the standard Solomon-Bloembergen-Morgan (SMB) model,52,53 i.e. neglecting the presence of static zero-field splitting (ZFS), and the Freed equation.54,55 These models calculate the inner-sphere and the outer-sphere contributions, respectively. The SBM model describes the relaxation profiles from the number of coordinated water molecules and their distance from the unpaired electrons, and from several dynamic parameters, that are the molecular reorientation time, the electron relaxation time and the lifetime of coordinated water protons. The Freed equation describes the contribution to relaxation due to water molecules freely diffusing around the paramagnetic complex through the diffusion constant, the distance of closest approach and the electron relaxation time. We assume that the two protons of the coordinated water molecule are located at 3.1 Å from the metal ion, with lifetimes τM of 350 and 300 ns at 298 and 310 K respectively, as previously found for [Gd(HP-DO3A)(H2O)] at 298 K.56 Diffusion water molecules were considered with standard values for the diffusion constants of 2.5·10−5 and 3.5·10−5 cm2s−1 at 298 and 310 K,51 respectively, and a distance of closest approach of 3.6 Å. The best fit parameters were the field dependent electron relaxation time, T1e, described by the transient ZFS Δt and the electron correlation time τv, and the tumbling time τR. The calculated best fit values were 0.027±0.003 cm−1 for Δt, 28±2 and 26±3 ps for τv at 298 and 310 K, respectively, and 0.067±0.004 and 0.049±0.004 ns for τR at 298 and 310 K, respectively. These values correspond well to those previously obtained for similar small Gd(III) complexes by analysing the relaxation data using the same model.6</p><p>The profiles for 3 could not be satisfactorily fit using the SBM model presumably due to the effect of a static ZFS in the presence of a slower molecular tumbling. The profiles were fit according to the "modified Florence" approach53,57–59 to obtain estimates of the parameters on which the relaxivity depends. In such an approach the effect of static ZFS on the splitting of the energy levels of all spin states is considered to be best calculated using nuclear spectral densities, provided that τR is much larger than T1e. The solid lines in Figure 4 are the best-fit curves obtained for the two protons of the coordinated water molecule at 3.1 Å from the metal ion with a τR of 0.74±0.03 ns and of 0.60±0.02 ns at 298 K and 310 K, respectively. All other parameters have the same values used in the analysis performed with the SBM and Freed models, except for the transient ZFS Δt of 0.018±0.001 cm−1, τv of 23±2 ps, and the static ZFS of 0.024±0.005 cm−1. Figure 4 shows that higher field data are in reasonably good agreement with the best-fit profiles obtained from relaxation data measured up to 40 MHz.</p><p>The overall agreement between the calculated curves and the experimental data shows that the main features of all profiles can be reproduced as the result of an increase in τR upon conjugation of 1 to 3. (Table 3) The change in the values of electron relaxation parameters between 1 to 3 (similar to that observed for other gadolinium complexes when bound to macromolecules)27,60,61 is likely determined by the simultaneous presence of both static and transient ZFS, which is not fully accounted for in fast rotating systems by available fitting programs.53, 62 Both profiles of 1 and 3 were fit according to the "modified Florence" approach, and the main features of all profiles could be reproduced as the result of an increase in τR on passing from 1 to 3 (see Supporting Information, Figure S1). Independent from the electronic parameters, the analysis clearly shows that the NMRD profiles of 3 can only be reproduced for tumbling times that are approximately one order of magnitude larger than those of 1.</p><p>Conjugation of 1 to 3 results in an increase in τR from 0.049 ns to 0.60 ns at 310 K resulting in a 2 fold increase in relaxivity at low field and a 5 fold increase at the peak value of 40 MHz. The analysis of the data is consistent with an essentially unaltered hydration of the Gd(III) complex upon conjugation to the scaffold and shows that the increase of relaxivity is due to the increase in τR. The tenfold increase in τR upon the conjugation of 1 to 3 places the molecule in the intermediate τR range between (0.5 and 4 ns). This range has been predicted to be optimal for imaging at higher magnet field strengths.24</p><p>MR images were obtained of complexes 1 and 3 in glass capillaries (1 mm) at 7 T. (Figure 5) Molecular concentrations ranged from 60–7.5 μM. The image was obtained using a RARE pulse sequence with a TE and TR of 11 and 200 ms. Not surprisingly, 3 is clearly brighter than 1 at the same molecular concentration. This data demonstrates the improved performance of 3 vs 1 at high field strength which results in significant contrast enhancement.</p><p>In conclusion we have developed a multiplexed and modular MR contrast agent scaffold with a 500% increase in per Gd relaxivity at 1.41 T and a 170% increase at 9.4 T versus small molecule MR agents. Using this approach we are preparing scaffolds that support multiple CAs while simultaneously providing the ability to incorporate functional groups for subsequent modification with targeting ligands, fluorophores, and nanoparticles.</p><!><p>Unless noted, materials and solvents were purchased from Sigma-Aldrich Chemical Co. (St. Louis, MO) and used without further purification. GdCl3·6H2O and 1,4,7,10-tetraazacyclododecane (cyclen) were purchased from Strem Chemicals (Newburyport, MA) and used without further purification. Unless noted, all reactions were performed under a nitrogen atmosphere. THF, acetonitrile, and dichloromethane were purified using a Glass Contour Solvent system. Deionized water was obtained from a Millipore Q-Guard System equipped with a quantum Ex cartridge (Billerica, MA). Thin-layer chromatography (TLC) was performed on EMD 60F 254 silca gel plates. Visualization of compounds was accomplished using either an iodoplatinate or UV light. Standard grade 60 Å 230–400 mesh silca gel (Sorbent Technologies) was used for flash column chromatography.</p><p>1H and 13C NMR spectra were obtained on a Bruker 500 MHz Avance III NMR Spectrometer or a Varian Inova 400 MHz NMR Spectrometer with deuterated solvent as noted. Electrospray ionization mass spectrometry (ESI-MS) spectra were taken on a Varian 1200 L single-quadrupole mass spectrometer. High resolution mass spectrometry data was aquired on an Agilent 6210 LC-TOF (ESI, APCI, APPI). Analytical reverse-phase HPLC-MS was performed on a Varian Prostar 500 system with a Waters 4.6 × 250 mm 5 μM Atlantis C18 column. This system is equipped with a Varian 380 LC ELSD system, a Varian 363 fluorescence detector, and a Varian 335 UV-Vis detector. Preparative runs were performed on a Water 19 × 250 mm Atlantis C18 Column. The mobile phases consisted of Millipore water (A) and HPLC-grade acetonitrile (B). HPLC method 1: 0–5 min 100% A, 5–24:08 min 57.5% A, 24:08–30 min 0% A, 30–35 min 0% A, 35–40 min 100% A.</p><p>Determination of r1 was accomplished using a Bruker minispec 60 MHz (1.41 T) magnet and a Varian Inova 400 MHz (9.4 T) NMR Spectrometer. At 1.41 T the T1 relaxation times were determined using an inversion recovery method while at 9.4 T a saturation recovery method was used. The saturation recovery method utilized a 2 second presaturation pulse centered on the water frequency. All measurements were done at 37 °C at an approximate 1 mM concentration of CA in 10 mM DPBS purchased from Invitrogen.</p><p>NMRD measurements were performed with a Stelar Spinmaster FFC-2000-1T fast field cycling relaxometer in the 0.01–40 MHz proton Larmor frequency range at 298 and 310 K. Standard field cycling protocol was used. Longitudinal water proton relaxation rates were obtained with an error smaller than 1%. Proton nuclear magnetic relaxation dispersion (NMRD) profiles were obtained by plotting proton relaxation rates as a function of applied magnetic field after subtraction of the diamagnetic contribution of buffer alone and normalization to 1 mM Gd(III) concentration.</p><p>(6) 1-azido-3-chloropropan-2-ol was synthesized as described by Ingham et al with the following modifications.63 Diethyl ether was used in the place of dichloromethane in the procedure. The final product was not distilled as in the literature procedure but simply extracted into ether and evaporated. The product was used directly in subsequent reactions. Caution: Safe handling procedures for perchlorates and small molecule azides should be reviewed before performing this reaction, as there is a danger of explosion if heat, friction, or shock is applied.</p><p>Tri-tert-butyl 2,2′,2″-(1,4,7,10-tetraazacyclododecane-1,4,7-triyl)triacetate hydrobromide (trist-butyl-DO3A HBr) was synthesized according to the procedure of Oskar with the following modifications.64 To a 500 mL RB flask was added 10.279 grams (59.8 mmoles) of cyclen to this was added 14.8405 grams (179.2 mmoles) of NaOAc. The solids were dissolved in 180 mL of dimethylacetamide (DMA). The reaction was cooled to 0 °C with ice and 26.5 mL (179.3 mmol) of tert-butyl bromoacetate dissolved in 70 mL of DMA was added drop wise over 40 minutes at 0 °C. The reaction was allowed to warm to RT and stirred for two days and was poured into a solution of 16.6 grams of KBr in 1000 mL of H2O. The solution was brought to a basic pH with 17.7g (3.5 eq) of NaHCO3. (Caution: A large amount of gas is produced.) Add 10 mL of ether to initiate precipitation of the HBr salt of tris-t-butyl-DO3A. The final white to off white powder was filtered and dried under vacuum to give a yield of 21.4976 grams (60% yield). 1H NMR (500 MHz, CDCl3) δ 3.56 – 2.57 (m, 21H), 1.46 (d, J = 3.6 Hz, 27H). 13C NMR (126 MHz, CDCl3) δ 170.53, 169.63, 81.86, 81.71, 77.29, 77.04, 76.78, 58.22, 51.31, 51.12, 49.14, 47.53, 28.23, 28.19, 28.03, 0.00.</p><!><p>To a 250 mL round bottom was added 4.9963 grams (8.397 mmol) of tris-t-butyl-DO3A and 2.8957 grams (20.98 mmol) of K2CO3. The flask was charged with 80 mL of anhydrous acetonitrile. The flask was sealed and placed under a nitrogen atmosphere. 1.6957 grams (12.50 mmol) of (6) were dissolved in 5 mL of acetonitrile and added to the solution of tris-t-butyl-DO3A. The reaction was heated to 50 °C and stirred overnight. The reaction was monitored by mass spectrometry to follow the disappearance of the starting materials. 0.399 g (2.94 mmol) of (6) was added and allowed to stir for another 24 hours. The reaction was checked by MS to determine the disappearance of tris-t-butyl-DO3A. Once all of the tris-t-butyl-DO3A had disappeared the reaction was filtered and evaporated to provide a yellow oil. The oil was dissolved in minimal methanol and 100 mL of diethyl ether was added and the flask was placed at −20 °C overnight. 3.968 grams (77% yield) of clear to yellow crystals of (7) formed overnight in the freezer and were filtered and washed with cold ether. M/Z observed: 614.5, calculated: 614.4 [M+H]+.</p><!><p>1.0201 grams (1.661 mmol) (7) was dissolved in 100 mL of formic acid and heated overnight. MS was used to observe the removal of the t-butyl protecting groups, once complete the formic acid was evaporated on a rotary evaporator. The resulting oil was re-dissolved in water 3 × 10 mL and evaporated to remove most of the formic acid. The resulting glassy solid was dissolved in 30 mL of water and Gd(OAc)3·6H2O was added. The pH was adjusted to ~6.5 with 1M NaOH and the reaction was heated to 50 °C. The pH was adjusted back to 6.5 every 6 to 10 hours until no further change occurred (typically 1 to 2 days). The reaction was evaporated and purified by reverse phase HPLC according to method 1, retention time of 15.5 minutes, and 99.9% purity, followed by lyophilization to yield 412 mg of (1) as a white powder in 41% yield based on the starting mass of tris-t-butyl-DO3A. M/Z observed: 597.13151, calculated: 597.13367 [M+H]+.</p><!><p>To a 100 mL round bottom flask was added 970.4 mg (2.933 mmol) of tribromophenol, 483.1 mg (3.500 mmol) of K2CO3, and 0.5 mL (3.120 mmol) of ethyl 5-bromovalerate. The flask was charged with 30 mL of dry DMF and nitrogen gas was bubbled through the reaction for 5 minutes. The reaction was left under nitrogen and brought to 50 °C and stirred for 12 hours. Once the reaction was shown to be complete by TLC (9:1 Hex:EtOAc) the reaction was diluted with 50 mL of H2O and extracted three times with 50 mL of diethyl ether. Afterwards, the combined organic layers were dried over MgSO4, filtered, and concentrated using a rotary evaporator. The residue was purified by column chromatography (9:1 hexane: ethyl acetate) to give 1181.8 mg of a yellow to clear oil (88% yield). 1H NMR (500 MHz, CDCl3) δ 7.62 (s, 2H), 4.12 (q, J = 7.1 Hz, 2H), 3.97 (t, J = 5.8 Hz, 2H), 2.40 (t, J = 4.6 Hz, 2H), 1.88 (dt, J = 6.5, 3.2 Hz, 4H), 1.24 (t, J = 7.1 Hz, 3H). 13C NMR (126 MHz, CDCl3) δ 173.70, 153.04, 135.21, 119.27, 117.49, 73.13, 60.54, 34.17, 29.55, 21.66, 14.48. M/Z observed: 456.8653, calculated: 456.8644 [M+H]+.</p><!><p>To a flame dried 50 mL round bottom flask was added 0.5793 grams (1.265 mmol) of S-8, 29.3 mg (0.154 mmol) of copper(I) iodide, and 107.2 mg (0.4092 mmol) of triphenylphosphine. The flask was charged with 12 mL of dry triethylamine. Nitrogen was bubbled through the solution for 5 minutes followed by the addition of 1.8 mL (13 mmoles) of TMS-acetylene and 90.4 mg (0.128 mmoles) of bis(triphenylphosphine) palladium (II) chloride were added. The flask was heated to 75° C and left to stir overnight under nitrogen. The reaction was concentrated using a rotary evaporator. 30 mL of hexane was added to the flask and the remaining solids were filtered off. The hexane solution was concentrated using a rotary evaporator and the final residue was purified by silica gel chromatography (40:1 hexane : ethyl acetate). 606.7 mg of clear oil was obtained (93.8% yield). 1H NMR (500 MHz, CDCl3) δ 7.46 (s, 2H), 4.21 (t, J = 5.9 Hz, 2H), 4.10 (d, J = 7.1 Hz, 2H), 2.37 (t, J = 7.2 Hz, 2H), 1.93 – 1.73 (m, 4H), 1.23 (t, J = 7.1 Hz, 3H), 0.27 – 0.13 (m, 27H). 13C NMR (126 MHz, CDCl3) δ 173.66, 161.79, 137.59, 118.55, 117.80, 103.10, 100.05, 99.86, 94.69, 73.70, 60.46, 34.27, 29.96, 21.87, 14.46, 0.08, 0.01. M/Z observed: 511.2505, calculated: 511.2515 [M+H]+.</p><!><p>To a 100 mL RB flask was added 214 mg (0.418 mmol) of S-9 which was dissolved in 20 mL of ethanol. To this was added 366 mg (6.29 mmol) of KF and stirred for 4 h or until the reaction is complete by TLC (19:1 hexanes:ethyl acetate). Once complete the reaction was evaporated and the solids washed with hexanes. The crude product was purified by column chromatography using 19:1 hexanes:ethyl acetate to provide 116 mg of a colorless to yellow oil in 94% yield. 1H NMR (499 MHz, CDCl3) δ 7.53 (s, 2H), 4.23 (t, J = 5.9 Hz, 2H), 4.11 (q, J = 7.1 Hz, 2H), 3.26 (s, 2H), 3.01 (s, 1H), 2.37 (t, J = 7.2 Hz, 2H), 1.96 – 1.73 (m, 4H), 1.23 (t, J = 7.1 Hz, 3H). 13C NMR (126 MHz, CDCl3) δ 173.74, 162.57, 138.31, 117.76, 117.21, 82.72, 81.55, 78.72, 77.95, 74.05, 60.46, 34.18, 29.75, 21.69, 14.47, 14.40. M/Z observed: 295.1322, calculated: 295.1329 [M+H]+.</p><!><p>105 mg (0.375 mmol) of S-10 was dissolved in 5 mL of 1,4 dioxane. To this was added 1 mL of 1M NaOH and stirred for 4 hours or until completion 9:1 hexanes:ethyl acetate was indicated by TLC. 5 mL of water were added to the reaction and the dioxane was evaporated on a rotary evaporator. The remaining water was further diluted by 5 mL of water and acidified with 3M HCl. Upon reaching an acidic pH the product precipitated as yellow to orange crystals 86 mg, 91% yield. 1H NMR (499 MHz, CDCl3) δ 7.56 (s, 2H), 4.26 (t, J = 5.9 Hz, 2H), 3.29 (s, 1H), 3.04 (s, 1H), 2.48 (t, J = 7.3 Hz, 2H), 2.01 – 1.80 (m, 4H). 13C NMR (126 MHz, CDCl3) δ 178.25, 162.27, 138.12, 117.61, 116.99, 82.55, 81.30, 78.47, 77.78, 73.72, 33.37, 29.41, 21.24. M/Z observed: 265.0865, calculated: 265.087 [M-H]−.</p><!><p>All click chemistry reactions were done in a 2:1 mixture of t-butanol and water. Compounds 1, 4, and 5 (3.3 eq) were dissolved in the water and compound 2 (1 eq) was dissolved in t-butanol. The solution of the Gd(III) complex in water was added to the solution of 2 in t-butanol. This solution was bubbled with nitrogen to remove any adventitious oxygen followed by the addition of [Cu(MeCN)4]PF6 (0.02 eq) and tris[(1-benzyl-1H-1,2,3-triazol-4-yl)methyl]amine (TBTA) (0.02 eq). The reaction was left under a nitrogen atmosphere and heated to 50 °C overnight. The reaction was checked by MALDI or HPLC to determine completeness of the reaction. The solvent was removed by lyophilization and the products 3, 6, and 7 were purified by reverse phase HPLC utilizing method 1.</p><!><p>Purified by HPLC according to method 1, retention time of 17.1 minutes, and 99.1% purity, M/Z found: 1045.76034, calculated: 1045.76040 [M+2H]2+</p><!><p>Purified by HPLC according to method 1, retention time of 18.6 minutes, and 99.2% purity, M/Z found: 1085.2790, calculated: 1085.2787 [M+2H]2+</p><!><p>Purified by HPLC according to method 1, retention time of 19.1 minutes, and 98.3% purity, M/Z found: 1105.79946, calculated: 1105.80178 [M+2H]2+</p>

PubMed Author Manuscript

Recent advances in the asymmetric phosphoric acid-catalyzed synthesis of axially chiral compounds

In recent years, the synthesis of axially chiral compounds has received considerable attention due to their extensive application as biologically active compounds in medicinal chemistry and as chiral ligands in asymmetric catalysis. Chiral phosphoric acids are recognized as efficient organocatalysts for a variety of enantioselective transformations. In this review, we summarize the recent development of chiral phosphoric acid-catalyzed synthesis of a wide range of axially chiral biaryls, heterobiaryls, vinylarenes, N-arylamines, spiranes, and allenes with high efficiency and excellent stereoselectivity.

recent_advances_in_the_asymmetric_phosphoric_acid-catalyzed_synthesis_of_axially_chiral_compounds

<!>Introduction<!><!>Introduction<!><!>Introduction<!><!>Introduction<!>Enantioselective synthesis of atropisomeric biaryls<!><!>Enantioselective synthesis of atropisomeric biaryls<!><!>Enantioselective synthesis of atropisomeric biaryls<!><!>Enantioselective synthesis of atropisomeric biaryls<!><!>Enantioselective synthesis of atropisomeric biaryls<!><!>Enantioselective synthesis of atropisomeric biaryls<!><!>Enantioselective synthesis of atropisomeric biaryls<!><!>Synthesis of atropisomeric arylindoles<!><!>Synthesis of atropisomeric arylindoles<!><!>Synthesis of atropisomeric arylindoles<!><!>Synthesis of atropisomeric arylindoles<!><!>Synthesis of atropisomeric arylindoles<!><!>Synthesis of atropisomeric arylindoles<!><!>Synthesis of atropisomeric arylindoles<!><!>Synthesis of atropisomeric arylindoles<!><!>Synthesis of atropisomeric arylindoles<!><!>Synthesis of miscellaneous atropisomeric heterobiaryls<!><!>Synthesis of miscellaneous atropisomeric heterobiaryls<!><!>Synthesis of miscellaneous atropisomeric heterobiaryls<!><!>Synthesis of miscellaneous atropisomeric heterobiaryls<!><!>Synthesis of miscellaneous atropisomeric heterobiaryls<!><!>Synthesis of miscellaneous atropisomeric heterobiaryls<!><!>Synthesis of miscellaneous atropisomeric heterobiaryls<!><!>Synthesis of miscellaneous atropisomeric heterobiaryls<!><!>Enantioselective synthesis of axially chiral arylalkene and N-arylamines<!><!>Enantioselective synthesis of axially chiral arylalkene and N-arylamines<!><!>Enantioselective synthesis of axially chiral arylalkene and N-arylamines<!><!>Enantioselective synthesis of axially chiral arylalkene and N-arylamines<!><!>Enantioselective synthesis of axially chiral allenes and spiranes<!><!>Enantioselective synthesis of axially chiral allenes and spiranes<!><!>Enantioselective synthesis of axially chiral allenes and spiranes<!><!>Enantioselective synthesis of axially chiral allenes and spiranes<!><!>Enantioselective synthesis of axially chiral allenes and spiranes<!><!>Enantioselective synthesis of axially chiral allenes and spiranes<!><!>Conclusion

<p>All costs related to the publication of this open access article were entirely funded by the Beilstein-Institut.</p><!><p>Axial chirality is one of the most important properties of nature, resulting from the nonplanar arrangement of four groups in pairs about a chirality axis. These include atropisomerism [1], chiral allenes, spiranes, spiroindanes, and so on [2–3]. Recently, there emerged an enormous demand for enantiopure compounds, not only for pharmaceutical and fine chemical industries, but also for fragrance, flavor, agrochemical, and food industries [4]. Consequently, the importance of axially chiral compounds has been widely recognized in both academic and industrial chemical societies [5]. Therefore, asymmetric synthesis of axial chiral compounds has been paid much attention [6], and great progress has been made in recent years. For example, many remarkable activities have been undertaken to develop strategies such as dynamic kinetic resolution, atroposelective coupling, cycloaddition, and chirality conversion for the construction of axial chirality [7–14].</p><p>Axially chiral biaryl and heterobiaryl units are widely used as basic building blocks for chiral ligands [14], chiral catalysts [14–17] (Figure 1), various natural products, drugs and bioactive molecules [18–19], pharmaceutical agents [20–21] (Figure 2), and chiral building blocks in modern organic synthesis [5]. During the past decades, both C2 and non-C2 symmetric axially chiral biaryl compounds such as BINAP, BINAM, NOBIN and their derivatives BINOL have played a crucial role as ligands in the development of transition-metal-catalyzed enantioselective transformations [9,14,22–23].</p><!><p>Representative examples of axially chiral biaryls, heterobiaryls, spiranes and allenes as ligands and catalysts.</p><p>Selected examples of axially chiral drugs and bioactive molecules.</p><!><p>Axial chirality is also found in chiral stationary phases for enantioselective separation, dopants in liquid-crystalline materials, chiroptical molecular switches, microporous soluble polymers, and interlocked nanotubes (Figure 3) [24]. In addition, axially chiral allenes and spiranes [25] are well-known scaffolds widely used in natural products, ligands, organocatalysts, and functional materials as well as versatile chiral building blocks in organic synthesis [14,26–27].</p><!><p>Axially chiral functional materials and supramolecules.</p><!><p>Chiral phosphoric acids represent an important and widely used class of catalysts for a variety of enantioselective transformations, especially for carbon–carbon and carbon–heteroatom bond-forming reactions [15,28]. They are important for the development of axially chiral compounds, which are involved in the design of chiral catalysts and ligands. Currently, chiral phosphoric acids are widely used in stereoselective oxidative/cross-coupling of two aryl counterparts, asymmetric control of aromatic ring formation, atroposelective functionalization of biaryl compounds, and so on [17,29–30]. In this context, Akiyama (2004) described that chiral phosphoric acids (CPAs) have great potential for catalysis of a wide range of reactions to achieve good to perfect enantioselectivities. This is due to their ability to act as synergistic bifunctional catalysts bearing both Brønsted acidic and Lewis basic sites, with the 3,3′-substituents playing a crucial role in achieving excellent enantioselectivity [31]. The widespread use of phosphoric acids and phosphates as chiral acids, chiral anions, and ligands is one of the most important achievements of modern enantioselective catalysis. The atropochiral BINOL, H8-BINOL and SPINOL [32] derived phosphoric acids (Figure 4) [33–34] play a crucial role in asymmetric catalysis, the construction of numerous axially chiral biaryls/heterobiaryls [21,35–38], and other useful asymmetric transformations [5].</p><!><p>Important chiral phosphoric acid scaffolds used in this review.</p><!><p>Although chiral phosphoric acids have shown promising properties in asymmetric catalysis and play a significant catalytic role in the development of axially chiral compounds with biaryls, heterobiaryls, atropisomeric arylalkenes, allenes, and spiranes, there are only few comprehensive reviews in this area. Therefore, the aim of this review is to provide readers with an overview of recent advances in the asymmetric phosphoric acid-catalyzed synthesis of axially chiral compounds and to suggest that much more attention should be paid to these catalysts in order to promote asymmetric synthesis.</p><!><p>Direct C–H functionalization strategies for the atroposelective aryl–aryl cross-coupling using various transition-metal catalysts has rarely been successful for the enantioselective construction of hindered biaryls [39] due to the discord between the temperature tolerance of the rotational axis and the high temperature required for C–H activation and suffered from poor chemoselectivity [40]. However, the realization of this redox-neutral aryl–aryl cross-coupling is a formidable challenge. Therefore, the discovery of efficient catalysts and ligands to achieve high stereoselectivity is a fundamental issue in catalytic asymmetric synthesis [14]. In this section, we will present pioneering examples of chiral phosphoric acid-catalyzed asymmetric syntheses of axially chiral biaryls. In 2013, Kürti and co-workers reported a chiral phosphoric acid-catalyzed aryl–aryl-bond formation process for the regio- and atroposelective synthesis of 2,2′-diamino-1,1′-binaphthalenes (BINAMs) from achiral N,N′-binaphthylhydrazines (Scheme 1). In the presence of chiral phosphoric acids (CPA 1), the reaction undergoes a simple [3,3]-sigmatropic rearrangement, giving the corresponding products 2 in good yield (up to 88%) and enantioselectivity (up to 93:7 er). The density functional calculations showed that the chiral phosphoric acid proton forms an H-bond with nitrogen atoms of 1 and the phosphate acts as a chiral counterion, resulting in a [3,3]-sigmatropic rearrangement with controlled stereoselectivity [14,41].</p><!><p>Atroposelective aryl–aryl-bond formation by employing a facile [3,3]-sigmatropic rearrangement.</p><!><p>In 2017, Tan and co-workers developed an organocatalytic atroposelective synthesis of axially chiral biarylamino alcohols from the reaction of quinone ester (3) and various 2-naphthylamines 4. In the presence of CPA 2, the axially chiral biarylamino alcohols 5 were obtained in moderate to good yields (up to 85%) and high to excellent enantioselectivities (up to 99% ee) (Scheme 2). The electronic properties of the substituents on the 2-naphthylamine showed remarkable effects on the chemical yield but had negligible impact on the enantioselectivity [42].</p><!><p>Atroposelective synthesis of axially chiral biaryl amino alcohols 5.</p><!><p>The direct arylation reaction of quinones 6 and 2-naphthols 7 was described by Tan and co-workers in 2015. The corresponding axially chiral biaryl diols 8 were prepared in good yields (60–90%) and excellent enantioselectivities (90–99% ee) in the presence of CPA 2 (Scheme 3). The stereoselectivity of the reaction is only moderately affected by the position and electronic properties of the substituents on the aromatic ring [36]. Experiments showed that chiral phosphoric acid CPA 2 acted as a bifunctional organocatalyst, that activates 2-naphthols and quinone derivatives via multiple H-bonds and promotes the first step of the enantioselective conjugative addition to generate intermediate I-1 and transfers the its central chirality information to the axial chirality to give the chiral biaryldiols (Scheme 3) [14].</p><!><p>The enantioselective reaction of quinone and 2-naphthol derivatives.</p><!><p>In 2013, Akiyama and co-workers described the enantioselective preparation of multisubstituted biaryls by the desymmetrization strategy, which was further enhanced by the subsequent asymmetric reaction (kinetic resolution) in the presence of chiral phosphoric acid CPA 3. In this work, various EWG- and EDG-containing substrates were incorporated, and chiral biaryls 10 were obtained with good to excellent selectivities of 81–93% ee by desymmetrization and 63–96% ee by kinetic resolution. The subsequent asymmetric bromination reaction afforded monobrominated biaryls with excellent enantioselectivities up to 99% ee (Scheme 4). The experimental and computational studies showed that the highly organized hydrogen-bonding network between a substrate, a chiral phosphoric acid catalyst (CPA 3), and a brominating reagent (N-bromophthalimide) plays a crucial role in achieving the excellent selectivities [43].</p><!><p>Enantioselective synthesis of multisubstituted biaryls.</p><!><p>The axially chiral biaryl-2-amines, N-heteroarenes [44] and derivatives are ubiquitous structural motifs and have profound applications in biochemistry and asymmetric catalysis [45]. These chiral biaryls can be prepared by asymmetric C–H activation. The C–H activation or functionalization can be achieved by a metal-catalyzed chiral phosphoric acid ligand-assisted method, which offers distinct possibilities to provide various chiral biaryl compounds by changing different directing groups (DGs) [46]. Since the important reports by Akiyama and Terada [31,47], chiral phosphoric acids have received much attention in the efficient construction of chiral molecules not only by using them as organocatalysts, but also by applying them as ligands in transition-metal catalysis due to their multifunctionality (Brønsted acidity/basicity, hydrogen-bonding units, and counter-anions toward metals) [48–49]. In 2019, Shi, Lin, and co-workers achieved an enantioselective synthesis of axially chiral quinoline-derived biaryl atropisomers via Pd-catalyzed C–H olefination of 8-phenylquinoline (11) using a novel chiral spirophosphoric acid (CPA 4). A wide range of quinoline-derived biaryls 13 with various substituents was synthesized in good to excellent yields (up to 99%) with excellent enantioselectivities (up to 98% ee) (Scheme 5). A working model for the origin of enantioselectivity in C–H olefination was provided by density functional theory calculations [46].</p><!><p>Enantioselective synthesis of axially chiral quinoline-derived biaryl atropisomers mediated by chiral spirophosphoric acid catalyst CPA 4.</p><!><p>Recently, Shi, Lin and co-workers reported the atroposelective synthesis of axially chiral biaryls by Pd(II)-catalyzed free amine-directed atroposelective C–H olefination using chiral spirophosphoric acid CPA 5 as an efficient ligand and Ag2CO3 as the oxidant. The C–H olefination of various biaryl-2-amines with acrylates and styrenes was carried out and afforded the desired products 15 in good yields (up to 96% yield) with high enantioselectivities (up to 96% ee). The electronic properties of the side chains of the biaryl compounds have a significant effect on the reactivity, but little on the enantiocontrol. In general, biaryl-2-amines with electron-donating groups showed higher reactivity than those with electron-withdrawing groups (Scheme 6). CPA 5 activates the reaction by creating a rigid and narrow chiral pocket, thus leading to better noncovalent interactions with the substrates. Moreover, the SPA ligand loading of 1 mol % is effective for this reaction under mild conditions [50].</p><!><p>Pd-Catalyzed atroposelective C–H olefination of biarylamines.</p><!><p>Moreover, Shi, Lin and co-workers reported the first Pd(II)-catalyzed directed atroposelective C–H allylation of 2-(naphthalen-1-yl)aniline derivatives 14 with methacrylates 16 via β-H elimination using (S)-CPA 4 as a ligand (Scheme 7) [51]. The directing group (NH2) facilitated the C–H activation on the other aromatic ring and ensured its coordination with Pd to form palladacycle I-2 to restrict bond rotations and promote exclusive allylic selectivity via β-H1 elimination as opposed to styrenyl selectivity. In addition, the methacrylates were used as allyl surrogates to overcome the reactivity problem caused by steric hindrance of 1,1,-disubstituted alkenes, and the electron-withdrawing esters enhance the migratory insertion to form palladacycle I-2. The CPA generates the chiral environment and played a crucial role in favoring of atropisomerism. A series of axially chiral biaryl-2-amines with 1,1-disubstituted alkenes 17 was efficiently synthesized in moderate to excellent yields (42–96%) with excellent enantioselectivities (up to >99% ee).</p><!><p>Palladium-catalyzed directed atroposelective C–H allylation.</p><!><p>The indole scaffold is found in many natural products, drugs, and bioactive molecules. Moreover, the introduction of axial chirality into the indole scaffold is receiving attention due to its widespread use [45,52]. In 2019, Li and co-workers developed a synthetic strategy for the atroposelective construction of phenylindole 20 by the chiral phosphoric acid-catalyzed cross-coupling of quinones 18 and indoles 19. In this reaction, the chiral phosphoric acid (R)-CPA 6 acts as a bifunctional catalyst to activate indoles and quinones through a dual H-bond activation mode to form the intermediate I-3, and aromatization with central-to-axial chirality transfer occurs to afford the axially chiral phenylindoles in good yield (76–92%) with good to excellent enantioselectivity (88–96% ee, Scheme 8a). The position and electronic properties of the substituents on the aromatic rings of the indole have limited influence on the reactivity and remarkable effects on the enantioselectivity. For example, in the case of indole 20b, the enantioselectivity was drastically reduced due to steric hindrance present at the 2-position of the indole, which plays a crucial role in the stability of axial chirality.</p><!><p>Enantioselective synthesis of axially chiral (a) aryl indoles and (b) biaryldiols.</p><!><p>In 2019, Li and co-workers further investigated the feasibility of 5-hydroxyindoles 22 with iminoquinones 23 under optimized reaction conditions. In the presence of 5 mol % (R)-CPA 7, the desired biaryldiol products 23 were obtained in moderate to good yields (58–95%) with excellent enantioselectivity (90–95% ee, Scheme 8b) [53]. The synthesized axially chiral phenylindoles have the potential to be used as chiral ligands in asymmetric catalysis [54] and are ubiquitous in natural products with considerable bioactivity [55–56].</p><p>Organocatalytic aryl C–H activation via a nonradical process represents an enormous challenge in organic synthesis, although the nucleophilic aromatic substitution with cleavage of the electrophilic aryl C–H bond has only recently been developed by transition-metal-catalyzed aryl C–H activation [57]. In the presence of a chiral phosphoric acid, the azo group has recently been revealed to be a useful moiety that may efficiently activate an aromatic ring for formal nucleophilic aromatic substitution, resulting in the cleavage of the aryl C–H bond and direct arylation of the nucleophile [58]. In 2018, Tan and co-workers showed that azo groups enable the organocatalytic asymmetric arylation of indoles. The nucleophilic aromatic substitution between the azobenzene derivative 24 and indoles 25 was carried out in the presence of 2.5 mol % chiral phosphoric acid (CPA 8, Scheme 9a), leading to the intermediate I-4, which was subsequently aromatized to give the intermediate I-5 (Scheme 10). The axially chiral arylindole 26 was synthesized from intermediate I-5 via chirality transfer with high isolated yield (up to 98%) and excellent enantioselectivity (up to >99% ee, Scheme 9a) [59]. The authors also reported that the reaction of 2-substituted indoles with azobenzenes proceeded smoothly in the presence of the chiral phosphoric acid catalyst (R)-CPA 9. At a catalyst loading of 0.2 mol %, the intermediate I-5 was simultaneously cyclized to I-6, followed by β-H elimination and chirality transfer to afford another type of axially chiral indoles 27 (Scheme 10) bearing aniline groups by direct arylation and rearrangement in moderate to good yields with excellent enantioselectivities of mostly 99% ee (Scheme 9b) [59]. The electronic properties and position of the substituents on the azobenzene ring did not affect the yields, whereas the electronic properties of the substituents on the indole ring did.</p><!><p>Asymmetric arylation of indoles enabled by azo groups.</p><p>Proposed mechanism for the asymmetric arylation of indoles.</p><!><p>In 2019, Lin and co-workers reported the asymmetric three-component cascade reaction of 2,3-diketo esters 28, aromatic amines 29, and 1,3-cyclohexanediones 30 to prepare axially chiral arylindoles 31 in a highly enantioselective manner. A wide range of substrates was subjected to the reaction in the presence of the chiral phosphoric acid CPA 9 to afford axially chiral N-arylindoles 31 in good yields (up to 93%) with good to exceptional enantioselectivities (up to 99% ee, Scheme 11). The chiral spirocyclic phosphoric acid catalyst developed by our group is critical for increasing the enantioselectivity in this cascade reaction [38]. This catalyst can facilitate the aldol reaction to generate a stereocenter (I-7), which can then be converted to axial chirality (I-8 to I-10) and finally aromatized to give 31 (Scheme 11) [60].</p><!><p>Enantioselective synthesis of axially chiral N-arylindoles [38].</p><!><p>Zhou and co-workers published an excellent paper in 2019 on the conversion of central to axial chirality in an enantioselective [3 + 2] annulation of 1-styrylnaphthols 32 with azonaphthalenes 33. Under defined conditions, the cycloaddition product 34 was prepared in high yield (99%) with exclusive diastereoselectivity and 99% ee in the presence of the chiral phosphoric acid CPA 2. Subsequently, using the chiral phosphoric acid-catalyzed [3 + 2] formal cycloaddition and a moderate DDQ oxidation method over 34, enantiomerically enriched 2,3-diarylbenzoindoles 35 were successfully prepared by performing a central-axial chirality conversion by oxidative aromatization. With excellent diastereoselectivities (all >20:1) and enantioselectivities (87–99% ee), the target products 35 containing different groups were obtained in high yield (up to 94%) (Scheme 12). In this transformation, electron-donating and withdrawing groups were tolerated [7].</p><!><p>Enantioselective [3 + 2] formal cycloaddition and central-to-axial chirality conversion.</p><!><p>In 2020, Tan and co-workers disclosed the phosphoric acid-catalyzed atroposelective arene functionalization of nitrosonaphthalene with indoles to form atropisomeric indole-naphthalenes 39 and indole-anilines 40 by a nucleophilic aromatic substitution reaction [61] (Scheme 13). Various 2-nitrosonaphthalenes 36 and indoles 37 with different substitutions were subjected to nucleophilic aromatic substitution reaction catalyzed by CPA 11 to form an intermediate I-11, which was then re-aromatized to give I-12, and then oxidized to give 39. In addition, the intermediate I-12 was cyclized to form the intermediate I-13 followed by β-H elimination to give another axially chiral arylindole framework 40 (Scheme 14). Both products 39 and 40 were obtained in moderate to excellent yields (up to 99%) and good to excellent enantioselectivity (83–99% ee, Scheme 13). Moreover, the reactions provide easy access to the privileged NOBIN (2-amino-2'-hydroxy-1,1'-binaphthyl) structures 41. The additional insights into the origins of enantiocontrol were described by DFT calculations.</p><!><p>Organocatalytic atroposelective arene functionalization of nitrosonaphthalene with indoles.</p><p>Proposed reaction mechanism for the atroposelective arene functionalization of nitrosonaphthalenes.</p><!><p>Owing to the presence of axially chiral indole-based biaryl scaffolds such as naphthylindoles and phenylindoles in bioactive molecules and chiral catalysts [62–64], the construction of this scaffold has recently become valuable and attracted the attention of chemists [2,64]. In this context, Shi and co-workers reported a new strategy for the enantioselective synthesis of axially chiral naphthylindoles via the asymmetric addition reaction of racemic naphthylindole 42 with azodicarboxylate 43 under chiral phosphoric acid catalysis. In the presence of CPA 2, 42 and 43 reacted and underwent dynamic kinetic resolution to afford naphthylindoles 44 with axial chirality in moderate to good yields (50–98%) and high enantioselectivity (91:9 to 98:2 er, Scheme 15) [65].</p><!><p>Asymmetric construction of axially chiral naphthylindoles [65].</p><!><p>In addition, the authors also succeeded in preparing naphthylindoles 46, which exhibit both axial and central chirality, through the addition reaction of racemic naphthylindoles 42 and o-hydroxybenzyl alcohols 45 using chiral phosphoric acid CPA 13. This reaction afforded products 46 in moderate to good yields (50–97%), excellent diastereoselectivities (>95:5 dr), and high enantioselectivities (91:9 to 98:2 er, Scheme 15). The control experiments showed that the N–H group in naphthylindoles, the OH group in phenol, and the carboxylate group in azodicarboxylate play a crucial role in both the reactivity and enantioselectivity of the reaction due to the formation of hydrogen bonds with CPA. In addition, the chiral phosphoric acid catalysts generate the chiral environment and showed the pleasing role in favoring atropisomerism [65].</p><p>In 2019, Shi and co-workers also achieved the first catalytic asymmetric construction of axially chiral 3,3'-bisindole scaffolds 49 bearing both axial and central chirality by employing the CPA-14-catalyzed asymmetric addition reaction of 2-substituted 3,3'-bisindoles 47 to isatin-derived 3-indolylmethanols 48. The isatin-derived 3-indolylmethanols and 2-substituted 3,3'-bisindole substrates bearing different substituents afforded the axially chiral 3,3'-bisindole scaffolds 49 in moderate to high yields (up to 98%) and with excellent stereoselectivities (all >95:5 dr, >99% ee, Scheme 16). Moreover, the introduction of a bulky group into the ortho-position of prochiral 3,3'-bisindoles allowed the synthesis of new members of axially chiral biaryls (3,3'-bisindole derivatives) bearing both axial and central chirality in high yields and excellent stereoselectivities. Moreover, this methodology represents a new strategy for the catalytic enantioselective synthesis of axially chiral 3,3'-bisindole backbones from prochiral substrates [66].</p><!><p>Enantioselective synthesis of axially chiral 3,3'-bisindoles [66].</p><!><p>The atroposelective synthesis of a new class of 3,3'-bisindoles 51 with axial and central chirality by dynamic kinetic resolution of 2-substituted 3,3'-bisindoles via asymmetric nucleophilic addition reactions using isatin-derived imines 50 as electrophiles was reported by Tan, Shi and co-workers in 2020. Enantioenriched 3,3'-bisindoles with a biologically essential chiral 3-aminooxindole unit can be readily prepared by these methods. CPA 9 utilizes hydrogen bonding to activate both reactants, resulting in enantioenriched 3,3'-bisindole derivatives with multiple chirality. In the presence of the chiral phosphoric acid catalyst CPA 9, various groups of 2-substituted 3,3'-bisindoles and isatin-derived imines were tolerated and gave good to high yields (up to 80%) and moderate to excellent stereoselectivities (up to 98:2 er, >95:5 dr, Scheme 17) [67].</p><!><p>Atroposelective synthesis of 3,3'-bisiindoles bearing axial and central chirality.</p><!><p>In 2020, Shi and co-workers reported the chiral phosphoric acid (CPA)-catalyzed enantioselective addition reaction of 3,3'-bisindoles 47 and ninhydrin-derived 3-indolylmethanols 52 for the synthesis of 3,3'-bisindoles 53 with single axial chirality. In the presence of CPA 15, the axially chiral 3,3′-bisindoles 53 were synthesized via a dynamic kinetic resolution (DKR) process in overall moderate yields (up to 79%) and good enantioselectivities (up to 94:6 er, Scheme 18) [68]. The authors investigated the high rotation barrier and stability of the axially chiral products and found that 3,3'-bisindole scaffolds with a single axial chirality are quite stable because of the bulkiness of the ninhydrin group. They also investigated the rotational barrier of substrate 47 and confirmed that it could be rapidly racemized during the reaction process, suggesting that dynamic kinetic resolution could be used to accomplish enantioselective synthesis of axially chiral 3,3'-bisindoles. The mechanistic studies showed that the N–H group of 3,3′-bisindoles 47 played a crucial role in carrying out the addition reaction with substrates 52 possibly by forming a hydrogen bond with CPA 15. Moreover, the N–H group of ninhydrin-derived 3-indolylmethanols 52 was crucial in increasing the reactivity and enantioselectivity, but it is not important for ninhydrin-derived substrates to carry out the addition reaction. This suggests that this substrate could produce an ion-pair interaction with CPA 15 in addition to the hydrogen-bonding interaction.</p><!><p>Enantioselective synthesis of axially chiral 3,3'-bisindoles bearing single axial chirality.</p><!><p>Axially chiral pyrazole scaffolds are commonly found in natural products and drugs and are often used as valuable building blocks in organic synthesis [69]. Herein, Li and co-workers developed a new asymmetric synthesis of atropisomeric pyrazole derivatives 56 via an enantioselective reaction of azonaphthalene 54 with pyrazolone 55 using 5 mol % chiral phosphoric acid CPA 16. Substrates bearing electron-donating and electron-withdrawing groups gave the desired axially chiral pyrazole derivatives 56 in good yields (up to 99%) with excellent enantioselectivities (up to 98% ee, Scheme 19). Density functional theory calculations revealed that the chiral phosphoric acid acts as a proton-transfer shuttle to assist proton transfer from the OH group of pyrazolone-enol intermediates to the N–N double bonds of azonaphthalenes [70].</p><!><p>Enantioselective reaction of azonaphthalenes with various pyrazolones.</p><!><p>On the other hand, N-arylcarbazole frameworks are also abundant in pharmaceuticals, agrochemicals, natural products, and functional OLED materials [71–72]. Therefore, the construction of axially chiral N-arylcarbazoles is desirable. In this sense, Tan and co-workers (2020) developed the first chiral phosphoric acid-catalyzed atroposelective C–H amination of arenes for the synthesis of N-arylcarbazole structures. The atroposelective N-arylcarbazoles 58 were prepared by C–H amination of azonaphthalene derivatives 54 and carbazole derivatives 57 via asymmetric addition and chirality transfer. In the presence of CPA 9, axially chiral N-arylcarbazoles were obtained in moderate to good yields (51–97%) with good to excellent enatioselectivity (87–96% ee, Scheme 20a) [73].</p><!><p>Enantioselective and atroposelective synthesis of axially chiral N-arylcarbazoles [73].</p><!><p>More importantly, the same group reported the synthesis of axially chiral N-arylindole atropisomers from azonaphthalene 54 and indole substrates 59 in the presence of chiral phosphoric acid CPA 17. The atropisomeric adducts were obtained in moderate to good yields (up to 93%) with good to excellent enantioselectivity (up to >99% ee, Scheme 20b). Therefore, this synthetic method (nucleophilic aromatic substitution reaction) is crucial for the development of alternative C–N bond-forming reactions to conventional metal-involved cross-couplings, providing axially chiral N-arylcarbazoles 60 in good yields with remarkable enantiocontrol through a rearomatization-enabled central to axial chirality transfer pathway [73].</p><p>Citing the crucial role of di-carbazole-substituted arenes in OLED materials [74], Tan and co-workers have recently developed structural motifs with two chiral N-aryl axes from 2,6-diazonaphthalene 61 and carbazoles 62. For this reaction, a high loading of 20 mol % CPA 18 with a 1-pyrenyl group at the 6,6'-position of the spiro backbone was preferred as the best catalyst for this transformation. Encouragingly, the double atroposelective C–H amination reaction took place and afforded the desired 1,5-dicarbazole naphthalene derivative 63 in moderate yield with >90% enantioselectivity (Scheme 20c) and tolerable diastereoselectivity (4:1) [73].</p><p>The optically enriched benzimidazoles are N-heterocycles which are of great interest as drug-like molecules [75], and exhibit biological activities such as anticancer, antiviral, antifungal, and antibacterial effects [76]. In this context, Miller and co-workers reported the performance of C2-symmetric chiral phosphoric acids (C2-type) and phosphothreonine-embedded, peptidic phosphoric acids (pThr-type CPAs) to catalyze a wide range of atroposelective cyclodehydration reactions for the synthesis of benzimidazoles. In the presence of CPA 7, the corresponding products 65 were obtained with the highest selectivity (up to 96% ee) at full conversion while catalyst CAT 1 (Scheme 21) afforded 65 with up to 94% ee at 96% conversion. The highly substituted axially chiral benzimidazoles 65 were all formed with excellent enantioselectivity when either CAT 1 (93–97% ee) or CPA 7 (93–96% ee) was used as catalyst (Scheme 21). The authors described that, BINOL-derived CPAs and pThr-type CPA scaffolds were found to be effective for atroposelective cyclodehydrations. Both the DFT and catalyst correlation studies showed that the steric repulsion between the large 3,3′-substituent of the C2-type CPAs catalyst and the bottom aromatic ring of the substrate seems to determine the enantioselectivity [77].</p><!><p>Atroposelective cyclodehydration reaction.</p><!><p>In 2020, Fu and co-workers described the chiral phosphoric acid-catalyzed atroposelective construction of axially chiral N-arylbenzimidazoles involving a carbon–carbon bond cleavage under optimal reaction conditions. In the presence of CPA 2, N1-(aryl)benzene-1,2-diamines 66 were used in the reaction with multicarbonyl compounds 67 and 68 and afforded the corresponding products, axially chiral N-arylbenzimidazoles 69 and 70 in high yields (up to 89%) with excellent enantioselectivity (up to 98% ee, Scheme 22) [78]. The primary amino group in the N1-(aryl)benzene-1,2-diamines reacts with the carbonyl group in 67 to give imine intermediate I-15 mediated by the chiral phosphoric acid. Meanwhile, isomerization of I-15 leads to enamine I-16. In the presence of the CPA catalyst, an intramolecular Michael addition of the amino group to the enamine in I-16 leads to I-17. Subsequently, the imines I-15 and I-17 are converted to the intermediate I-18, which through elimination by cleavage of the C–C bond gives the target product 69 (Scheme 23). This catalytic approach encompasses a wide range of substrates and functional groups to construct atropisomeric N-arylbenzimidazoles that are widely used in natural products, biologically active molecules, ligands, and catalysts.</p><!><p>Atroposelective construction of axially chiral N-arylbenzimidazoles [78].</p><p>Proposed reaction mechanism for the atroposelective synthesis of axially chiral N-arylbenzimidazoles.</p><!><p>Axially chiral arylpyrroles are the core scaffold for a variety of natural products, bioactive compounds and pharmacological agents [79] as well as for a variety of chiral phosphine ligands [80]. As a result, the preparation of axially chiral arylpyrroles has been one of the most important areas of investigation in synthetic chemistry. In the last decade, the optical activity of arylpyrroles has been explored by optical resolution of racemates using chiral resolving agents or chiral column chromatography [81]. Although Zhang et al. developed the first catalytic asymmetric Paal–Knorr reaction for accessing highly enantioenriched axially chiral arylpyrroles in 2017 [82], a complicated catalytic system and narrow substrate range limited its application. Therefore, the use of chiral phosphoric acids to modulate axial chirality enables the preparation of highly enantioenriched axially chiral arylpyrroles. In this context, Tan and co-workers discovered in 2019 a highly effective approach using organocatalytic atroposelective desymmetrization and kinetic resolution to obtain enantioenriched axially chiral arylpyrroles. The axially chiral arylpyrroles 73 were prepared in high yields (up to 99%) and with excellent enantioselectivities (up to 98% ee) from the reaction of 71 and diethyl ketomalonate (72) under remote control of CPA 19 as chiral phosphoric acid catalyst (Scheme 24). The hydrogen bonding between the ketomalonate and CPA 19 proved to be the most important interaction in the formation of the chiral pocket for the induction of chirality and could considerably improve the stereocontrol of the reaction [21].</p><!><p>Atroposelective synthesis of axially chiral arylpyrroles [21].</p><!><p>The axially chiral arylquinazolinones form the backbones of a large number of natural products and biologically active compounds as well as chiral ligands [83]. Nevertheless, a simple chiral phosphoric acid-catalyzed enantioselective approach to access optically pure arylquinazolinones has never been developed in the last decades. However, in 2017, Tan and co-workers reported the enantioselective synthesis of axially chiral arylquinazolinones 76 from the reaction of N-arylanthranilamides 74 and benzaldehyde (75) by accelerating imine formation (I-19), and under the catalysis of a chiral phosphoric acid, intramolecular nucleophilic addition occurs to form I-20, followed by oxidative dehydrogenation with 2,3-dichloro-5,6-dicyano-1,4-benzoquinone (DDQ). In the presence of 10 mol % chiral phosphoric acid CPA 7, the axially chiral arylquinazolinones 76 were obtained in high yield (up to 99%) with high enantioselectivity (83–97% ee, Scheme 25). Both the position and the electronic properties of the substituents on the aromatic ring had a minor influence on the reaction efficiency and enantioselectivity of this transformation [35].</p><!><p>Synthesis of axially chiral arylquinazolinones and its reaction pathway [35].</p><!><p>Quinolines are widely used in natural and synthetic products and exhibit remarkable pharmacological properties [84]. To synthesize this valuable scaffold, Cheng and co-workers performed for the first time in 2019 an atroposelective Friedländer heteroannulation reaction of 2-aminoaryl ketones 77 with α-methylene carbonyl derivatives 78 catalyzed by a chiral phosphoric acid. In the presence of (R)-CPA 9, the Friedländer heteroannulation reaction between aromatic amines 77 and the carbonyl derivative 78 was carried out to form imine I-21, which tautomerized to generate intermediate I-22. Under CPA 9 catalysis, the intermediate I-22 is converted to I-23, which after dehydration forms the desired enantioenriched products, polysubstituted 4-arylquinolines 79 in high yield (up to 94%) with good to excellent enantioselectivities (up to 97% ee, Scheme 26). The chiral phosphoric acid catalyst played an important role in the asymmetric induction by establishing a favorable chiral environment in the cyclization step through supporting hydrogen bonds [85].</p><!><p>Synthesis of axially chiral aryquinoline by Friedländer heteroannulation reaction and its proposed reaction mechanism [85].</p><!><p>In 2019, the group of Bertuzzi and Corti developed chiral phosphoric acid-catalyzed Povarov reactions of N-arylimines 80 with 3-alkenylindoles 81 to give enantioenriched, highly substituted, 1,2,3,4-tetrahydroquinolines, which can be oxidized to axially chiral 4-(indol-3-yl)quinolones 82, in a central-to-axial chirality conversion approach with up to 99% yield and high enantioselectivity (up to 99% ee, Scheme 27). In presence of 5 mol % (R)-CPA 2, the benzyl-substituted 3-alkenylindole 81 (dienophile) was subjected to a Povarov cycloaddition with the commonly used imine 80, giving the 2,3,4-tetrahydroquinoline as a single diastereomer in high yield and enantioselectivity and finely tolerating the high steric requirements necessary to provide stable atropisomeric quinolines after oxidation by DDQ [86].</p><!><p>Povarov cycloaddition–oxidative chirality conversion process.</p><!><p>Although many elegant strategies have been developed to enable the atroposelective construction of axially chiral biaryls and heterobiaryls [87–89], the synthesis of axially chiral styrenes or vinylarenes has rarely been developed [90–91], due to their low rotational barrier and weak configurational stability. Moreover, the application of an organocatalytic kinetic resolution strategy to access axially chiral styrenes has rarely been reported. In this regard, Shi and co-workers reported the first atroposelective access to oxindole-based axial chiral styrenes or vinylarenes by kinetic resolution of racemic oxindole-based styrenes 83 and azalactones 84. In the presence of CPA 20, racemic oxindole-based styrenes reacted with azalactones via hydrogen bonding, and the azalactones underwent asymmetric ring opening to form axially chiral oxindole-based styrenes 85 in up to 53% yield with good diastereoselectivities (up to 94:6 dr) and enatioselectivities (up to 95% ee). Moreover, the racemic oxindole-based styrenes underwent kinetic resolution to afford other axially chiral styrenes 86 in moderate yields (up to 54%) with excellent enantioselectivities (up to 96% ee) and high selectivity factors (SF up to 106) [92]. This strategy is critical for accessing axially chiral the oxindole-based styrenes and provides a robust method for the synthesis of bisamide derivatives that are both axially and centrally chiral (Scheme 28).</p><!><p>Atroposelective synthesis of oxindole-based axially chiral styrenes via kinetic resolution.</p><!><p>Very recently, the same group reported the first CPA-catalyzed asymmetric assembly of axially chiral arylalkene-indole scaffolds by organocatalytic (Z/E)-selective and enantioselective (4 + 3)-cyclization using (3-alkynylindol-2-yl)methanols 87 as electrophile and 2-naphthols 88 or phenols 90 as nucleophile (Scheme 29) [45]. The (3-alkynylindol-2-yl)methanol 87 is expected to convert to the allene-iminium intermediate I-24 by accepting a proton from CPA 14. Then, the CPA anion activates the nucleophilic addition between 2-naphthol (88) and allene-iminium intermediate I-24 to form axially chiral I-25, followed by rearomatization of the naphthol ring of I-25 and isomerization to I-26. Thereafter, CPA forms two hydrogen bonds with the two OH groups of I-26 to generate a carbocation and facilitates an intramolecular nucleophilic addition to afford axially chiral aryl-alkene-indole frameworks (89) in up to 98% yield, >95:5 E/Z, and up to 97% enantioselectivity (Scheme 30).</p><!><p>Synthesis of axially chiral alkene-indole frame works [45].</p><p>Proposed reaction mechanism for axially chiral alkene-indoles.</p><!><p>The nonbiaryl N–C atropisomer is an important structural scaffold, which is present in natural products, medicines. and chiral ligands due to the restricted rotation around an N–C single bond. There are few strategies for the catalytic atroposelective construction of N–C nonbiaryl atropisomers, which mainly consist of enantioselective cyclization [35], N-functionalization [93], and desymmetrization [94] of the existing achiral N–C bond. In this context, Bai and co-workers (2019) developed the direct intermolecular enantioselective C–H amination of N-aryl-2-naphthylamines 92 with azodicarboxylates 93 to prepare N–C atropisomers of nonbiaryl naphthalene-1,2-diamine 94. In the presence of chiral phosphoric acids (CPA 15), the desired product 94 was obtained in moderate to high isolated yield (up to 95%) and enantioselectivity (up to 94% ee, Scheme 31). However, the reactivity and stereoselectivity decreased the closer the substituents were to the reaction site, probably due to steric effects. On the other hand, the substrates with electron-withdrawing groups would decrease the electron cloud density, limiting the reactivity and thus reducing the yield. In general, the electronic properties of the substituents did not affect the stereoselectivity of the reaction. The mechanistic study revealed that, CPA 15 simultaneously activates N-phenyl-2-naphthylamine and azodicarboxylate through a dual hydrogen bond activation mode and a π–π interaction strategy (Scheme 31) [2].</p><!><p>Atroposelective C–H aminations of N-aryl-2-naphthylamines with azodicarboxylates.</p><!><p>Diarylamines and related scaffolds are among the most common potentially atropisomeric chemotypes in medicinal chemistry. For example, the drugs binimetinib and bosutinib contain diarylamines that exist as rapidly interconverting atropisomers, and a VEGFR inhibitor from Wyeth contains a potentially atropisomeric N-arylquinoid [95–96]. In 2020, Gustafson and co-workers reported the first chiral phosphoric acid-catalyzed atroposelective electrophilic halogenation of N-arylquinoids 95, a class of compounds similar to diarylamines. In this reaction, CPA 21 was an effective catalyst, providing stereochemically stable brominated product N-arylquinoids 96 in high yield (up to 95%) and atroposelectivity (up to 98:2 er, Scheme 32). Remarkably, these products existed as stereochemically stable class 3 atropisomers under protic and aprotic conditions due to a strong intramolecular N–H–O hydrogen bond that locks one of the axes into a planar conformation and proposed that the quinoid-nitrogen axis exists in a planar exo conformation [97].</p><!><p>Synthesis of brominated atropisomeric N-arylquinoids.</p><!><p>The enantioselective construction of spirocyclic centers is an exciting synthetic challenge that plays a prominent role in the discovery of new catalysts and complex molecules [98]. Spirocyclic frameworks are present, for example, in biologically active natural products (e.g., fredericamycin A and acutumine [99]), SPINOL-based ligands and catalysts [100], and organometallic complexes with important applications. The use of chiral catalysts for spirocyclization reactions brings asymmetry at the site of the spirocyclic center while tolerating different electronic and steric functional groups [98].</p><p>In this context, Tan and co-workers succeeded in the enantioselective synthesis of axially chiral SPINOLs 98 from ketals 97 by an intramolecular fashion in the presence of 1 mol % chiral phosphoric acid CPA 22 to afford the axially chiral product (R)-98 in high yield (62–95%) with good to excellent enantioselectivity (90–96% ee, Scheme 33) [17]. The electronic properties and steric bulk of the substituents on the catalysts as well as the axial chiral backbone have a very strong influence on the reactivity and enantioselectivity as shown in Scheme 33. The synthesis of the products (R)-98 at a gram scale was carried out without loss of chemical yields and stereoselectivities under optimized conditions. The authors also found that the excellent stereocontrol was due to the simultaneous interaction between the bifunctional phosphoric acid and the intermediate formed in the reaction via hydrogen bonds.</p><!><p>The enantioselective syntheses of axially chiral SPINOL derivatives.</p><!><p>Axially chiral allenes and derivatives occur in overwhelming numbers in natural products, pharmaceuticals, functional materials, and useful intermediates in organic synthesis [101]. Therefore, great efforts have been made to synthesize these axially chiral scaffolds [102]. In this regard, Wang and co-workers reported the organocatalytic asymmetric synthesis of tetrasubstituted α-amino allenoates by a dearomative γ-addition reaction of 2,3-disubstituted indoles 99 to β,γ-alkynyl-α-imino esters 100 [103]. In the presence of chiral phosphoric acid CPA 13, a series of tetrasubstituted α-amino allenoates 101 was prepared in moderate to excellent yields (69–99%), dr (9:1 to >20:1), and excellent enantioselectivity (91–99% ee, Scheme 34). The mechanistic studies showed that the substituents at the second and third positions of the indole play a crucial role in the chemo- and stereoselectivity. Moreover, the chiral phosphoric acid group attains a bifunctional role by activating both partners via hydrogen bonds, and the chiral backbone of the catalyst controls the stereoselectivity through steric effects and π–π interactions.</p><!><p>γ-Addition reaction of various 2,3-disubstituted indoles to β,γ-alkynyl-α-imino esters.</p><!><p>Moreover, Li and co-workers recently developed CPA-catalyzed regio- and stereoselective γ-addition reaction of isoxazol-5(4H)-ones 103 to β,γ-alkynyl-α-imino esters 102 for the synthesis of axially chiral tetrasubstituted α-amino allenoates 104 containing an adjacent quaternary carbon stereocenter and an axially chiral tetrasubstituted allene scaffold [104]. Although isoxazol-5(4H)-ones have different nucleophilic sites, the authors succeeded in the C-allenylation of isoxazol-5(4H)-ones with high efficiency (up to 91% yield) and high regioselectivities and stereoselectivities (up to 94% ee and >20:1 dr, Scheme 35).</p><!><p>Regio- and stereoselective γ-addition reactions of isoxazol-5(4H)-ones to β,γ-alkynyl-α-imino esters.</p><!><p>Because of the importance of chiral tetrasubstituted allenes with aryl substituents, the asymmetric synthesis of these scaffolds has received much attention. In 2020, Lu and co-workers carried out the enantioselective dehydrative γ-arylation of α-indolyl-α-trifluoromethylpropargyl alcohol 105 and 1-naphthol (106) or 2-naphthol (107) catalyzed by chiral phosphoric acids CPA 13 and CPA 23, to produce a wide range of chiral tetrasubstituted allenes 108 and naphthopyrans 109 in high yield (≤98% yield) with excellent regio- and enantioselectivities (>99:1 er, Scheme 36) [105]. However, when substituents are present at the C-4 or C-7 position of the indole ring of the propargyl alcohol, the yield decreases, which could be due to steric effects. Since the chiral phosphoric acid catalysts can interact with these groups via double hydrogen bonds, control studies have shown that the free OH on naphthol/phenol and the NH groups on the α-indolyl-α-trifluoromethylpropargyl alcohol are very important for the reaction.</p><!><p>Synthesis of chiral tetrasubstituted allenes and naphthopyrans.</p><!><p>Li and co-workers developed a chiral phosphoric acid-catalyzed asymmetric remote 1,8-conjugate addition of thiazolones 111 and azlactones 112 to propargyl alcohols 110 for the synthesis of the chiral allenes 113 and 114, respectively. In the presence of 1 mol % CPA 24, 5H-thiazol-4-ones 111 and p-quinone methides generated in situ from propargyl alcohols 110 were incorporated and afforded the axially chiral tetrasubstituted allenes 113 with a chiral thiazolone moiety in high yield (65–97%), high enantioselectivity (76 to >99% ee) and diastereoselectivity (10:1 to >20:1 dr). In addition, the enantioselective 1,8-conjugate addition of azlactones 112 to para-quinone methides generated in situ from propargyl alcohols 110 were carried out in the presence of 1 mol % chiral phosphoric acid CPA 7 and afforded the chiral allenes 114 in high yields (65–97%) with good to excellent enantioselectivities (76–97% ee) and good diastereoselectivities (up to 20:1 dr, Scheme 37) [106]. The electronic nature and position of the substituents on the aromatic ring of the thiazolones or propargylic alcohols did not significantly affect the stereoselectivity of the reaction. Moreover, both electron-withdrawing and donating groups on the aromatic rings of propargyl alcohols or azlactones smoothly participated in the asymmetric 1,8-conjugate addition and afforded the corresponding chiral allenes in good yields with high enantioselectivity. The control experiments showed that the propargyl alcohol 110 was firstly converted to para-quinone methide (p-QM) in the presence of CPA and then led to the successful formation of the corresponding products through hydrogen-bonding interaction of para-quinone methide and thiazolones/azlactones with CPA.</p><!><p>Asymmetric remote 1,8-conjugate additions of thiazolones and azlactones to propargyl alcohols.</p><!><p>Recently, using chiral phosphoric acid catalysis, we developed a highly regio-, diastereo-, and enantioselective dearomatization reaction of 1-substituted 2-naphthols 115 and β,γ-alkynyl-α-imino esters 100. The highly functionalized naphthalenone derivatives 116 with an allene moiety, exhibiting both a quaternary stereocenter and axial chirality, were obtained in good yields (up to 82%) with high diastereoselectivities (up to >99:1 dr), and enantioselectivities (up to 96% ee, Scheme 38). Control experiments showed that the high to excellent stereoselectivity is the result of a dual hydrogen-bonding interaction between the substrates and the chiral phosphoric acid, with the substituent at the 1-position of the 2-naphthol playing a key role in controlling the regioselectivity [107].</p><!><p>Synthesis of chiral allenes from 1-substituted 2-naphthols [107].</p><!><p>In recent decades, chiral phosphoric acids have been recognized as privileged chiral catalysts and ligands and have therefore become an important tool in asymmetric organic synthesis. This review summarizes the state of the art in the chiral phosphoric acid-catalyzed asymmetric synthesis of axially chiral biaryls, heterobiaryls, arylamines, arylalkenes or styrenes, spiranes, and allenes. These axially chiral compounds have attracted considerable attention in recent years due to their wide application in the total synthesis of axially chiral natural products, functional materials, bioactive compounds, privileged chiral ligands, and have further potential applications in asymmetric catalysis and drug discovery. Accordingly, considerable efforts have been made to find new efficient routes for the enantioselective construction of various atropisomeric aryl–aryl or aryl–heteroaryl, enantioselective spiranes and allenes. Despite the advances mentioned above, much of this area of research is still relatively unexplored. Compared to the use of chiral phosphoric acids in the preparation of centrally chiral compounds, their application in the synthesis of axially chiral biaryls and heterobiaryls, axially chiral allene, atropisomeric aryl alkenes, and spirane moieties is still quite limited. Given the increasing importance of axially chiral compounds, novel chiral phosphoric acid-catalyzed asymmetric methods to address these challenges are highly desirable. Therefore, we believe that the future development of asymmetric syntheses using chiral phosphoric acid catalysts will play a crucial role in the preparation of other complex organic molecules with axial chirality, which will be widely used in science and industry in the near future.</p>

PubMed Open Access

High-performance photoacoustic probe for biopsy-free assessment of copper status in murine models of Wilson's disease and liver metastasis

The development of high-performance photoacoustic (PA) probes that can monitor disease biomarkers in deep-tissue has the potential to replace invasive medical procedures such as a biopsy. However, such probes must be highly optimized for in vivo performance and exhibit an exceptional safety profile. In this study, we have developed PACu-1, the first PA probe designed for biopsy-free assessment (BFA) of hepatic Cu via photoacoustic imaging. PACu-1 features a Cu(I)-responsive trigger appended to an aza-BODIPY dye platform that has been optimized for ratiometric sensing. Owing to its excellent performance, we were able to detect basal levels of Cu in healthy wildtype mice, as well as elevated Cu in a Wilson's disease model and in a liver metastasis model. To showcase the potential impact of PACu-1 for BFA, we conducted a blind study where we were able to successfully identify a Wilson's disease animal from a group of healthy control mice with greater than 99.7% confidence. Significance StatementThe ability to non-invasively detect and track disease biomarkers via photoacoustic imaging can potentially serve as a substitute for invasive medical procedures such as a liver biopsy. While achieving this goal can have a profound impact on disease management, it is an immense challenge that requires novel chemical tools that are sensitive, selective, and safe. Here we report an acoustogenic probe designed for Cu(I), which becomes dysregulated in many disease states. In addition to demonstrating in vivo efficacy in multiple models, we designed a blind study to assess its utility for biopsy-free assessment of hepatic copper levels in Wilson's disease. This work sets the stage for future studies to evaluate the performance of acoustogenic probe designs for biomedical applications.

high-performance_photoacoustic_probe_for_biopsy-free_assessment_of_copper_status_in_murine_models_of

Figures 1 to 6<!>Design and Characterization of PACu-1<!>Metabolic Stability, Biodistribution, and Safety Profile of PACu-1<!>Imaging Exogenous Cu(I) in BALB/c Mice<!>PA Imaging of Hepatic Cu(I) in Wilson's Disease<!>BFA of Hepatic Cu(I) in Wilson's Disease via a Blind Study<!>PA Imaging of Cu(I) in a Liver Metastasis Model<!>Discussion<!>Materials and Methods<!>Determination of the diagnostic threshold.

<p>Introduction Photoacoustic (PA) imaging is a light-in, sound-out technique that has emerged as a promising biomedical approach for the non-invasive assessment of various ailments in humans, ranging from arthritis to cancer. 1,2 Excitation of an endogenous pigment such as hemoglobin in blood or melanin in tissue can provide contrast since relaxation via non-radiative decay can trigger thermoelastic expansion of the surrounding tissue. Repeatedly irradiating a region of interest with a pulsed laser can result in pressure waves that can be readily detected by ultrasound transducers. Since ultrasound at clinically relevant frequencies can travel through the body with minimal perturbation, it is possible to accurately pinpoint the source of the signal to afford high resolution images at centimeter imaging depths. 3 Beyond label-free applications, the utility of PA imaging for disease detection has been augmented by the recent development of acoustogenic probes (activatable PA probes) that give an off-on signal enhancement or ratiometric readout. 4,5 Notable examples include those that can visualize dysregulated enzymatic activities, [6][7][8][9] properties of the disease tissue microenvironment, [10][11][12] as well as small molecule-and metal ion-based disease biomarkers. [13][14][15][16][17][18] However, replacing an invasive medical procedure such as a liver biopsy with an acoustogenic probe is an immense challenge since the in vivo performance and safety profile of such a chemical tool must be exceptional. Thus, in spite of undesirable shortcomings such as the potential to develop severe infections, false negatives due to collection of nondiseased tissue, and the inability to directly monitor disease progression in real-time, 19 liver biopsies are still commonly employed to assess biomarkers in conditions such as Wilson's disease (WD) 20 and cancer. 21 It is noteworthy that elevated levels of hepatic copper (Cu) is a common biomarker shared by these conditions. In WD, Cu accumulates in the liver due to a genetic mutation in the Cu-exporter, ATP7B, and this can lead to chronic liver damage which can become fatal if not treated. 22,23 In the context of cancer, Cu is elevated in many solid tumors including breast 24,25 and lung 26,27 cancers which generally metastasize to the liver. Since Cu can promote angiogenesis and drive tumor progression, BFA of Cu in metastatic lesions is critical. While several probes have been developed for in vivo imaging of Cu, these early examples were designed for fluorescent 28 and bioluminescent 29 methods which are more suitable for shallow imaging depths (mm range) owing to scattering and attenuation of light. More recently, our group 13,30 and others 31 have developed Cu probes for PA imaging to achieve greater tissue penetration and higher resolution. However, these probes are designed to target Cu(II), whereas intracellular Cu exists predominantly in the +1-form owing to a highly reducing environment of the cell. 32 To overcome this challenge, we present the development of PACu-1, the first acoustogenic probe for Cu(I) and its application in BFA of hepatic Cu in a WD model and a liver metastasis model. Moreover, we designed an unbiased BFA blind study to identify a Wilson's disease mouse from a group of healthy wildtype controls using PACu-1.</p><!><p>To target the +1-oxidation state of Cu, we installed a Cu(I)-responsive tris[(2-pyridyl)methyl]amine (TPA) 33 trigger onto an optimized aza-BODIPY dye platform to yield PACu-1 which features ratiometric imaging capabilities. Specifically, we hypothesized capping of the 2,6-dichlorophenol moiety will result in a blue-shift of the wavelength of maximum absorbance (λmax) relative to the uncapped probe. However, the binding of Cu(I) to TPA will induce an oxidative cleavage event of the pendant ether linkage to release the latent dye (Figure 1a). Subsequently, selective irradiation of each form (probe and product) at their corresponding λmax will yield two signals, from which a ratio can be determined. This probe design feature is important for BFA, especially in the liver since we anticipate there to be significant background interference from blood. In addition, we selected the aza-BODIPY platform to develop PACu-1 due to their large extinction coefficients (10 4 to 10 5 M -1 cm -1 ) in the near infrared range and low fluorescence quantum yields since both of properties translate to a stronger PA signal. 30 Lastly, we have determined empirically that many of the aza-BODIPY-based probes we have developed intrinsically localizes to the liver owing to its relatively high hydrophobic properties.</p><p>After synthesizing PACu-1 (Scheme S1), we evaluated its in vitro response to 20 equivalents of Cu(I) (introduced as [Cu(CH₃CN)₄]PF₆). After 1 h incubation at room temperature, we observed a large spectral shift of 91 nm from 678 nm (probe, Ɛ = 5.3 × 10 4 M -1 cm -1 ) to 767 nm (product, Ɛ = 3.7 × 10 4 M -1 cm -1 ) (Figure 1b). Given that the extinction coefficient of a molecule is a reliable proxy for its PA output, we estimate the ratiometric turn-on will be ~10.4-fold (defined as (770/680Final)/(770/680Initial)). Importantly, irradiation at 680 nm will predominately excite PACu-1, whereas light at 770 nm will only generate a signal that corresponds to the turned over product (Figure 1c-d). Moreover, we were able to observe a dose-dependent response to Cu (LOD = 0.2 µM) (Figure 1e). We were also able to show that PACu-1 can function in the presence of glutathione (GSH), an abundant biological thiol, that can compete with PACu-1 to bind Cu(I) (Figure 1f). Indeed, GSH is present at high levels in the liver and most solid tumors (up to 10 mM). 34,35 Lastly, PACu-1 was shown to exhibit excellent selectivity for Cu(I) against a panel of monovalent and divalent metal ions (Figure 1g). This finding is significant because in addition to Cu(I), the TPA trigger has been tuned to sense other metal ions and thus, may exhibit off-target reactivity. 36,37</p><!><p>Because our objective is to employ PACu-1 for BFA of hepatic Cu via PA imaging, it is critical to 1) demonstrate that it is not metabolized in the liver to give false positives and 2) show that it is biocompatible with an excellent safety profile. To this end, we treated PACu-1 with rat liver microsomes rich in metabolic enzymes (e.g., CYP450s). After an incubation period of 1 h, we did not observe any change in the absorbance spectra indicating there would be minimal off-target activation of PACu-1 that can lead to false positive results (Figure S1). We corroborated these results with mass spectroscopy analysis that showed the latent aza-BODIPY was not being released. Next, we performed MTT assays to assess the cytotoxicity of the probe in mammalian cell lines. For instance, HEK293 cells incubated with up to 25 µM of PACu-1 for 24 h were shown to have no significant loss of viability (Figure 2a). Next, we sought to determine the biodistribution of PACu-1 after systemic administration in BALB/c mice via ex vivo PA imaging analysis of the vital organs. Our data indicates that PACu-1 predominantly localizes to the liver and does not accumulate in the heart, kidneys, or spleen (Figure S2). Before PACu-1 could be considered further as a chemical tool for BFA applications, we examined its in vivo safety profile. First, we performed H&E staining on liver samples obtained from mice treated with either a vehicle control or PACu-1. Our results show that the nuclear staining patterns were identical, suggesting that PACu-1 is non-toxic (Figure 2b). Second, we conducted a comprehensive liver function test to measure the levels of albumin, alkaline phosphatase (ALP), alanine aminotransferase (ALT), aspartate transaminase (AST), bilirubin, blood urea nitrogen (BUN), cholesterol, and glucose in serum. We did not observe any statistical difference between vehicle-and PACu-1-treated animals which further demonstrates that PACu-1 is safe and thus, is ideal for BFA applications (Figure 2c).</p><!><p>To determine whether PACu-1 can be employed to detect elevated hepatic Cu(I) in live animals, we treated BALB/c mice with CuCl2 via intraperitoneal administration 2 h prior to the introduction of PACu-1. Of note, Cu(II) is rapidly reduced to Cu(I) upon uptake into cells. PA imaging revealed that the PA770/680 ratio was 1.48 ± 0.23 for the Cu-treated animals, whereas the corresponding ratio for the vehicle control was 0.94 ± 0.13 (Figure 3, red and blue, respectively). To confirm that these results were due to the detection of Cu(I), we administered ammonium tetrathiomolybdate (TM), a high affinity FDA-approved Cu chelator drug (Kd = ~10 -20 ), 38 prior to treatment with PACu-1. As anticipated, we did not observe any activation (0.94 ± 0.13) when TM was present since it can outcompete PACu-1 for binding to Cu(I) (Figure 3, yellow). To further validate this finding, we administered Ctrl-PACu-1, a non-responsive control probe that features an attenuated Cu(I) binding trigger (Scheme S2), to a fourth group of animals. Interestingly, PA imaging demonstrated that the ratio was also lower than both the vehicle group and the TM group (0.82 ± 0.12) (Figure 3, green). The lower ratio suggests that PACu-1 can detect basal levels of Cu that are present in the liver. Finally, we performed ICP-MS analysis on liver samples obtained from mice treated with CuCl2. Compared to animals that received a vehicle control, the concentration of hepatic Cu was twice as high (Figure 3c).</p><!><p>Cu accumulation in the liver is a pathological hallmark of WD which is typically assessed clinically via liver biopsies. 39 Using an established ATP7B genetic knockout model of WD developed by Lutsenko and co-workers (JAX stock #032624), we measured the levels of hepatic Cu in wildtype mice and WD mice using ICP-MS analysis after obtaining biopsied tissue. On average, we found that the Cu levels in WD mice were 17.5-fold greater than wildtype mice (Figure 4a). Likewise, when we employed PACu-1 and PA imaging for BFA of Cu, we found that the PA770/680 ratio was significantly higher in WD mice (1.24 ± 0.16) relative to wildtype mice (0.80 ± 0.11) ( Figure 4b-d).</p><p>It is critical to note that while ICP-MS analysis reports on total Cu levels, PACu-1 can only access the labile pool which is defined as Cu weakly associated with intracellular chelators such as GSH. To confirm the in vivo imaging results, we harvested the heart, kidneys, liver, and spleen from WD and wildtype mice treated with PACu-1 to perform ex vivo PA imaging. This experiment was performed to demonstrate that the PA signal intensity is higher in the liver of WD mice owing to activation of PACu-1 (Figure S3).</p><!><p>To evaluate the potential efficacy of PACu-1 for BFA of hepatic Cu(I) in WD, it is critical to perform a rigorous study that is free of potential bias. To this end, we designed a blind experiment where one investigator randomly selected mice belonging to either the WD or wildtype groups (eight total) for the study (Figure 5a). Each of the animals were then tagged, and their identities were concealed until the completion of the study. A second investigator then administered PACu-1 and employed PA imaging to identify the WD mice. There was no physical indicator that would allow us to distinguish the mice based on appearance. Prior to BFA, a reliable diagnostic threshold was determined in wildtype mice, which is defined as the PA770/680 ratio (0.82 ± 0.10) (vide infra). With this in mind, we identified seven animals with a PA770/680 ratio (0.63, 0.72, 0.82, 0.86, 0.87, 0.93, 0.94) within two standard deviations of the diagnostic threshold which were assigned to Group 1 (wildtype mice) (Figure 5b-d). In contrast, only one of the animals had a PA770/680 ratio (1.16) greater than three standard deviations of the diagnostic threshold and was correspondingly assigned to Group 2 (WD mouse). When the identity of the eight animals were revealed at the end of the study, we were able to correctly identify the WD mouse with greater than 99.7% confidence.</p><!><p>Finally, we turned our attention to a second model to further showcase the potential clinical impact of PACu-1. Elevated Cu in cancer of the bone, breast, gastrointestinal tract, and lungs has been associated with aggressive phenotypes and poorer prognosis. 25 There are ongoing efforts to employ Cu chelation therapy to reduce the copper status in primary tumors, as well as in metastatic lesions to treat cancer. 40,41 Since the liver is one of the most common sites of metastasis in the body, BFA of Cu(I) levels would facilitate real-time monitoring during tumor progression and treatment with a chelator. Nu/J mice were either implanted with A549 cells in the liver or received sham surgeries. After four weeks, PACu-1 was administered for PA imaging. We elected to use a PA instrument (MSOT inVision, iThera Medical) capable of whole-body crosssectional imaging for this study because a built-in feature would allow us to readily perform spectral unmixing to distinguish the signal from PACu-1 and blood. Compared to the animals that received sham surgeries (1.06 ± 0.28), the PA fold turn-on (defined as PAFinal/PAInital) of tumorbearing mice was 2.31 ± 0.78 (Figure 6a-c). This indicates that in addition to being able to sense hepatic Cu(I) in WD, PACu-1 can also detect elevated Cu in a lung cancer liver metastases model.</p><!><p>One of the major goals of molecular imaging research is to develop high-performance chemical tools that can non-invasively detect and monitor disease biomarkers in a deep-tissue context. PA imaging is ideal for this application because it involves the conversion of safe near infrared light to non-toxic ultrasound waves. Since sound at clinically relevant frequencies can readily pass through the body, it is possible to obtain high resolution images beyond 10 cm in depth. 42 Despite the emergence of various acoustogenic probes for analyte sensing, none have been explored to date for BFA of disease biomarkers of the liver. Thus, our goal is to develop PA probes that can potentially replace or complement invasive biopsies currently in use to provide real-time monitoring capabilities.</p><p>In this study, we chose to target Cu because while it is an essential metal ion required by all living organisms, aberrant levels are linked to genetic disorders such as WD, as well as most solid cancer types. Our group has previously developed several PA probes for Cu(II), 13,30 however we found that they were not stable when incubated with RLMs. Likewise, after synthesizing RPS1, a PA probe designed to image Cu(II) in a murine Alzheimer's disease model, 31 we discovered that it could not detect exogenous copper in the liver (Figure S4). These results are not surprising since each of these examples were designed to respond to Cu in its +2-oxidation state. PACu-1 on the other hand, is highly selective for Cu(I), affords a robust PA signal enhancement when irradiated at 770 nm, is compatible with ratiometric sensing, intrinsically targets the liver, and most importantly, exhibits an exceptional safety profile.</p><p>It is worth noting that one of the major differences between BFA using PACu-1 and traditional biopsies is that our probe is designed to detect the labile Cu pool (Cu associated with GSH), whereas the latter technique reports on the total Cu content in the sample. Despite this difference, we can still reliably distinguish WD mice from wildtype controls as shown in our blind study. In addition to detecting Cu in WD, we also demonstrate PACu-1 can be used to detect elevated Cu in a liver metastasis model. We envision PACu-1 can be used to aid in the development of new Cu chelators or in conjunction with existing Cu binding drugs to monitor changes in real-time. As previously mentioned, we employed two different PA instruments for the WD and cancer studies. This indicates that PACu-1 will be compatible with a range of imaging systems including new hand-held scanners, 43,44 wearable devices, 45,46 and endoscopic setups. 47,48 Lastly, we envision this work will inspire the development of other PA probes for BFA applications.</p><!><p>In vitro selectivity assay. The initial absorbance (400-800 nm) of PACu-1 (5 µM, 1:1 DMF:HEPES, pH 7.4) was measured before the addition of a panel of metal ions (100 µM). These initial measurements were used to determine the initial ratio770/680 via UV-vis spectroscopy. After addition, the cuvette was sealed and incubated for 1 h. Final measurements were recorded, and the ratiometric fold turn-on was calculated by dividing the final ratio with the initial ratio. All metal solutions were prepared in water from their chloride salt, except for Ag2CO3. Cs2CO3, and tetrakis(acetonitrile)copper(I) hexafluorophosphate.</p><p>Biopsy assessment of hepatic Cu via ICP-MS. BALB/c mice were anesthetized using isoflurane (1.5 -2.0%). The mice were then intraperitoneally injected with a solution of CuCl 2 (5 mg/kg) or vehicle (sterilized saline). After 2 hours, the mice were euthanized, then the liver was excised and weighed for ICP-MS analysis. The 2-hour incubation time was used to reduce Cu in vivo. To determine the Cu concentration in WT (B6129SF2/J) and WD (B6;129S1-Atp7b tm1Tcg /LtsnkJ) mice, the livers were similarly prepared as the BALB/c mice for ICP-MS analysis, except no intraperitoneal injections were performed.</p><p>Ex vivo biodistribution of PACu-1 via PA imaging. BALB/c mice were anesthetized using isoflurane (1.5 -2.0%) and retro-orbitally injected with either a solution of PACu-1 (50 μM) or vehicle (10% DMSO in sterilized saline, 50 µL)). After 1 hour, the mice were euthanized, and the liver, spleen, heart, and kidneys were excised. Photoacoustic imaging of the organs was performed at 680 and 770 nm using continuous mode with a 6 second rotation time (Nexus 128+, Endra Life Sciences) The ratio of the PA signals in PACu-1 treated mice obtained upon excitation at 680 nm and 770 nm were normalized to the ratio of the PA signals in vehicle treated mice.</p><!><p>A group of 10 wildtype mice (B6129SF2/J), which are direct controls of the WD mice (B6;129S1-Atp7b tm1Tcg /LtsnkJ), were used to determine the diagnostic threshold for hepatic Cu in Wilson's disease via PA imaging. After the mice were anesthetized using isoflurane (1.5 -2.0%), their abdomens were shaved, and they were positioned in the PA tomographer to facilitate direct imaging of the abdomen. After an image was acquired, an ROI was drawn around the liver to determine the signal intensity. The ratio of the PA signals obtained upon excitation at 680 nm and 770 nm in the ROI provided the initial PA770/680 ratio. The mice were then treated with a 50 μM solution of PACu-1 in saline containing 10% DMSO (50 μL) via retro-orbital injection. The mice were returned to their cages for 60 minutes while PACu-1 was allowed to react with the hepatic Cu. The mice were anesthetized and their livers were imaged as described previously to obtain the final PA770/680 ratio. The diagnostic threshold value (mean ± 2×SD) was determined by dividing the final PA770/680 ratio with the initial PA770/680 ratio.</p><p>Identification of WD via PA imaging in a blind study. A group of eight mice consisting of one WD animal (B6;129S1-Atp7b tm1Tcg /LtsnkJ) and seven wildtype animals (B6129SF2/J mice) was tagged and randomized by the first researcher. Their identity and the total number of WD mice present was concealed until the end of the study. Importantly, these mice had no distinguishing physical features that would allow us to identify them based on appearance. PA imaging of hepatic Cu using PACu-1 was then performed by a second researcher to determine the PA ratiometric fold turn-on for each animal. Mice with a PA770/680 ratio value greater than 1.02 was assigned to Group 1 (WD) and mice with a PA770/680 ratio value between 0.62 to 1.02 was assigned to Group 2 (wildtype). After PA imaging was performed on all animals, the assignment and identity were revealed to and validated by the corresponding author.</p><p>Statistical analyses. Statistical analyses were performed in Microsoft Excel. Sample sizes in all experiments were sufficiently powered to detect at least a p value < 0.05, which was significant. All data are expressed as mean ± SD. Multiple group analysis was performed using the Kruskal-Wallis Test. All other in vivo imaging data was analyzed by performing the Student's t-test (α = 0.05). *p > 0.05; **p > 0.01.</p>

ChemRxiv

Strand Invasion of DNA Quadruplexes by PNA: Comparison of Homologous and Complementary Hybridization

Molecular recognition of DNA quadruplex structures is envisioned as a strategy for regulating gene expression at the transcriptional level and for in situ analysis of telomere structure and function. The recognition of DNA quadruplexes by peptide nucleic acid (PNA) oligomers is presented here, with a focus on comparing complementary, heteroduplex-forming and homologous, heteroquadruplex-forming PNAs. Surface plasmon resonance and optical spectroscopy experiments demonstrated that the efficacy of a recognition mode depended strongly on the target. For a quadruplex derived from the promoter regulatory region found upstream of the MYC proto-oncogene, the homologous PNA readily invades the DNA target to form a heteroquadruplex at high potassium concentration mimicking the intracellular environment, whereas the complementary PNA exhibits virtually no hybridization. In contrast, the complementary PNA is superior to the homologous PNA in hybridizing to a quadruplex modeled on the human telomere sequence. The results are discussed in terms of the different structural morphologies of the quadruplex targets and the implications for in vivo recognition of quadruplexes by PNAs.

strand_invasion_of_dna_quadruplexes_by_pna:_comparison_of_homologous_and_complementary_hybridization

Introduction<!>Results<!>SPR Analysis of PNA Hybridization to the Myc19 Quadruplex<!>CD Analysis<!>Cyanine Dye Binding to PNA-DNA Heteroduplexes<!>PNA Hybridization to an Alternative Quadruplex Target<!>DISCUSSION<!>Myc19 Quadruplex<!>hTelo22 Quadruplex<!>Materials<!>Equipment<!>Surface plasmon resonance (SPR)<!>Circular Dichroism Spectropolarimetry<!>Absorption spectroscopy<!>Emission spectroscopy

<p>The ability of DNA to adopt nonduplex secondary and tertiary structures has long been of fundamental and biological interest, but the discovery of quadruplex DNA is driving considerable new activity in this field.[1] Intramolecular quadruplex DNA forms from guanine-rich sequences that can fold back on themselves to allow hydrogen-bonded G-tetrad formation and pi-stacking mediated stabilization.[2] Quadruplex DNA is most stable in physiologically relevant concentrations of potassium, due to favorable coordination of the guanine O-6 atoms by potassium cations.[3] Bioinformatics studies showing the prevalence of quadruplex-forming sequence (QFS) motifs in the genome, particularly in promoter regions,[4] transcriptional reporter assays[5] and endogenous gene expression profiling[6] all point toward functional significance of quadruplex DNA.[7]</p><p>There are numerous reports of endogenous quadruplex-binding proteins.[8] These cofactors likely play important roles in stabilizing quadruplex secondary structures and recruiting other proteins involved in regulating gene expression. Artificial quadruplex-binding proteins have also been obtained through screening of diverse libraries and the resulting proteins have been used in a variety of experiments to identify quadruplex sites.[6, 9] A recent report by Balasubramanian and coworkers illustrates the use of a quadruplex-binding protein as the primary recognition component in an immunofluorescence method for identifying chromosomal quadruplex motifs.[10]</p><p>A diverse collection of synthetic quadruplex-binding molecules has been reported over the past 20 years. These include small molecules obtained through traditional medicinal chemistry approaches, which are expected to recognize quadruplexes through binding to specific three-dimensional structural motifs.[11] The most notable of these molecules feature sub-micromolar affinities and demonstrate significant intracellular activity against suspected quadruplex targets.</p><p>An alternative approach to recognizing quadruplexes relies on specific hydrogen bonding to the nucleobases by oligomeric compounds. The most straightforward design involves Watson-Crick recognition, i.e. a cytosine-rich oligomer that can bind in a complementary fashion to a G-rich target. The majority of reports describing this approach have involved C-rich peptide nucleic acids (PNAs).[12] Typically, binding of the complementary PNA competes with quadruplex formation,[13] although a recent report from Mayol and coworkers describes the binding of short PNAs to accessible loops without disruption of the underlying quadruplex secondary structure.[14]</p><p>Quadruplexes can also be recognized by G-rich PNAs, in which case the resulting hybrids are heteroquadruplexes, rather than heteroduplexes.[15] A variety of quadruplex structures and stoichiometries can be obtained in this manner. For example, targeting a folded DNA quadruplex with a PNA having two G3 tracts results in formation of a PNA2-DNA complex consisting of two PNA-DNA heteroquadruplexes linked via a short DNA tether that was originally a loop in the DNA homoquadruplex target.[16] Alternatively, a PNA having a single G3 tract can hybridize to a DNA having three G3 tracts to form a 1:1 heteroquadruplex.[17]</p><p>While targeting quadruplexes with either C-rich complementary or G-rich homologous PNA appears to involve sequence-based recognition as opposed to the structure-based recognition commonly associated with small molecules, we recently discovered that a quadruplex-forming PNA exhibited more than 10-fold difference in association kinetics within a group of morphologically diverse DNA quadruplexes, indicating that the structure of the quadruplex can play an important role in PNA heteroquadruplex formation, at least from a kinetic perspective.[16a] This motivated our current studies to determine (a) if similar structural factors would be observed in the hybridization of complementary PNA to DNA quadruplexes and (b) if there are significant differences in the kinetics of hybridization between complementary and homologous PNAs for two DNA quadruplex targets.</p><!><p>The overall goal of this work was to compare binding of homologous and complementary PNAs to DNA G-quadruplex targets. We started with a parallel quadruplex formed by a sequence modeled from the promoter region of the MYC proto-oncogene.[18] We previously demonstrated effective heteroquadruplex formation with this target by several G-rich PNAs with KD values in the low nanomolar range.[16] One of those PNAs, which we previously referred to as Pmyc, we now call PmycH to distinguish it from the complementary PNA, PmycC (Table 1). The homologous PNAs form 2:1 heteroquadruplexes with the DNA targets, so the complementary PNAs were designed to form 2:1 heteroduplexes. Also, note that PNA-DNA heteroduplex formation is known to favor alignment of the PNA N-terminus with the DNA 3′-terminus,[19] whereas the opposite orientational preference was found previously for a PNA-DNA heteroquadruplex.[15] Figure 1 illustrates our previously proposed heteroquadruplex-binding model along with the expected 2:1 heteroduplex structure.</p><!><p>Surface plasmon resonance (SPR) experiments were previously helpful in studying heteroquadruplex formation at high ionic strength, where UV melting temperatures are too high to determine, allowing study of these nanomolar binding interactions from both kinetic and equilibrium perspectives.[16c] Therefore, we began this study by using SPR to compare hybridization of PmycH and PmycC to an immobilized Myc19 target. Figure 1A shows binding of the homologous PNA to the DNA target in the presence of 100 mM KCl, where the DNA should be folded into a stable quadruplex. We observe a concentration dependent increase in the binding of the PNA at relatively low nanomolar concentrations. As we observed previously, dissociation of the bound PNA is quite slow, reflecting formation of a stable heteroquadruplex. In contrast, when we performed the identical experiment with the complementary PNA PmycC, negligible hybridization was observed (Figure 1B). This indicates that, under these conditions, formation of a PNA-DNA heteroduplex based on Watson-Crick base pairing is kinetically unfavorable, compared with PNA-DNA heteroquadruplex formation.</p><p>One possible explanation for the much slower hybridization of the complementary PNA is the stability of the Myc19 quadruplex structure in high potassium concentration. It is well known that replacing potassium by sodium reduces the stability of both DNA homoquadruplexes[3] and PNA-DNA heteroquadruplexes.[15, 16c] Therefore, we repeated the SPR experiments in sodium containing buffer (Figures 1C and 1D). There is relatively little change in the binding of PmycH to the target in NaCl versus KCl but PmycC now binds comparably to the homologous PNA. Changing the buffer to lithium further destabilizes the Myc19 quadruplex as well as the PNA-DNA heteroquadruplex but not the heteroduplex as shown by the results in Figures 1E and 1F, where PmycH shows much lower binding response whereas PmycC binding is further accelerated.</p><p>To more easily compare hybridization as a function of ion, Figure 2 shows SPR sensorgrams for a single PNA concentration (20 nM) in KCl, NaCl and LiCl. PmycH binds similarly in Na+ versus K+, indicating that the destabilizing effect of the change in ion is greater for the DNA homoquadruplex than for the PNA2-DNA heteroquadruplex, allowing more PNA to hybridize during the association phase. In the presence of LiCl, the large drop in PNA binding during the association phase and the significantly faster dissociation are consistent with the overall quadruplex destabilizing effect of lithium. However, the opposite trend is observed for PmycC; the complementary PNA clearly requires destabilization of the Myc19 quadruplex target in order to hybridize.</p><p>There is an additional interesting result evident from comparing the data obtained in different salts, particularly in Figures 1D and 1F: the binding signal is nearing saturation at ca. 150 response units in LiCl, whereas the signal in NaCl exceeds this value at the highest PNA concentration. We believe that weaker, higher-order complexes are forming at the higher PNA concentrations, based on the significantly faster dissociation rates evident in these sensorgrams compared with the lower PNA concentrations.</p><!><p>Circular dichroism measurements are useful in the characterization of nucleic acid secondary structure and in the study of hybridization events. CD spectra were recorded for Myc19 DNA alone and in a 1:2 mixture with PmycC in 100 mM KCl buffer (Figure 3A). As, observed previously, Myc19 alone exhibits a negative peak at 240 nm and a positive peak at 265 nm, indicative of a parallel quadruplex.[20] Annealing PmycC with Myc19 yields a CD spectrum that is similar to that of Myc19 alone. This is consistent with the SPR data indicating minimal binding of the complementary PNA to the DNA quadruplex under these conditions. However, PNA-DNA heteroduplexes also exhibit CD maxima at 265 nm and minima at 240 nm,[19] hence, it is difficult at this point to infer whether the CD spectrum of the mixture is due to a heteroduplex or a homoquadruplex. In contrast, destabilizing the Myc19 quadruplex in NaCl or LiCl leads to more significant changes in the CD spectra upon addition of PmycC (Figures 3B and 3C). These results indicate that heteroduplex formation is favorable in sodium and lithium, consistent with the SPR results.</p><!><p>Further examination of heteroduplex formation was done using two cyanine dyes, DiSC2(5) and DiSC1(3). We previously demonstrated binding of DiSC2(5) to PNA-DNA heteroduplexes.[21] The dye binds by assembling into a helical aggregate using the PNA-DNA duplex as a template and is readily detected as a blue-purple color change as well as distinct visible absorption and induced CD bands. Figure 4 shows induced CD spectra for the dye in the presence of 1:2 Myc19:PmycC. A very weak CD band is observed in the presence of KCl, with much higher intensity observed in NaCl and LiCl. Combined with the SPR and CD data described above, these results confirm that the complementary PNA binds very weakly to the Myc19 target in KCl solution. Moreover, the samples that were analyzed by CD in Figures 3 and 4 were annealed by heating to 90 °C followed by slow cooling to room temperature, giving the PNA ample opportunity to hybridize. Therefore, the poor binding evident in Figure 2B cannot be ascribed solely to a kinetic barrier to hybridization.</p><p>Based on the sequence of PmycC, a 2:1 heteroduplex should form (Scheme 1). From the SPR, CD and UV results, we do not observe the formation of Myc19-PmycC duplex in 100 mM KCl so the binding stoichiometry was determined in 100 mM NaCl. We recently demonstrated the utility of another cyanine dye, DiSC1(3), in determining PNA-DNA binding stoichiometries, based on differences in fluorescence intensity for binding to DNA homoquadruplexes and PNA-containing hybrids.[22]</p><p>DiSC1(3) exhibits distinctive UV-vis and fluorescence spectra in the presence of the Myc19 homoquadruplex and Myc19-PmycC heteroduplex (Figure S1). These spectral differences are sufficient to allow a continuous variations experiment to be used to determine the binding stoichiometry. In this experiment, the dye concentration was held constant at 1 μM. The total DNA+PNA concentration was also held constant, but the ratio of the two oligomers was varied. Plotting the fluorescence intensity versus mole fraction of PNA results in an inflection at 0.67, i.e. a 1:2 DNA:PNA ratio (Figure 5). Thus, the complementary PNA is capable of binding to the Myc19 homoquadruplex in 1:2 stoichiometry, analogous to our previous findings for the homologous PNA, although only PmycH has sufficient affinity to do so in high potassium concentration.</p><!><p>Overall, the results described above demonstrate both kinetic and thermodynamic advantages for targeting the Myc19 DNA quadruplex with heteroquadruplex-forming PNA over a comparable heteroduplex-forming PNA. However, we recently reported that the kinetics of PNA-DNA heteroquadruplex formation can vary over a wide range depending on the structure of the DNA quadruplex target.[16a] Therefore, we extended these studies to a DNA quadruplex that was more resistant kinetically to heteroquadruplex formation.</p><p>We used a DNA oligonucleotide, hTelo22 (Table 1), derived from the human telomeric repeat sequence (GGGTTA)n. DNA oligonucleotide models based on this sequence have been shown to fold (in potassium solution) into a hybrid structure that features both parallel and antiparallel strand orientations.[23] In our previous work, homologous PNAs such as PmycH and PteloH exhibited more than 10-fold slower hybridization to hTelo22 than to Myc19. Thus, we were curious to see if the advantages described above for homologous PNA binding to Myc19 would be preserved for hTelo22.</p><p>Figure 6 provides SPR data for binding of PteloH and PteloC to the immobilized hTelo22 DNA target. In KCl, binding of PteloH to hTelo22 is much slower than the binding of PmycH to Myc19 (compare Figures 6A and 1A), consistent with our earlier findings. A similar experiment with the complementary PNA PteloC revealed significant hybridization even in the presence of KCl, in contrast to complementary hybridization to Myc19 (compare Figures 6B and 1B). Thus, in KCl solution, complementary PNA is better suited to binding to the telomeric quadruplex, whereas homologous PNA performs better with the Myc19 quadruplex.</p><p>We next repeated the SPR experiments in quadruplex-destabilizing sodium- or lithium-containing buffer. As shown in Figures 6C and 6E, PteloH binding to hTelo22 is only modestly improved in sodium and completely abolished in lithium, indicating that, while destabilization of the DNA quadruplex should accelerate heteroquadruplex formation, corresponding destabilization of the heteroquadruplex is also sufficiently high to minimize PNA hybridization. In contrast, hybridization of PteloC to hTelo22 increases significantly in the order KCl < NaCl < LiCl. Thus, complementary PNA hybridization to the two DNA quadruplexes follows a qualitatively similar cation dependence, although heteroduplex formation is noticeably better for hTelo22. In contrast, homologous PNA hybridization to the two quadruplexes is markedly better for Myc19, consistent with our previous findings.</p><p>Figure 7 overlays SPR sensorgrams for hybridization of the two PNAs to the hTelo22 quadruplex in the different salts. The significant advantage of the complementary PNA over the homologous PNA is evident for all three salts.</p><p>CD spectra were recorded for hTelo22 DNA alone and in a 1:2 mixture with PteloC in 100 mM KCl, NaCl or LiCl buffer (Figure 8). The CD spectrum for the quadruplex alone in KCl exhibits features associated with a hybrid structure, specifically a positive band at 295 nm and a shoulder at 260 nm. In the presence of the complementary PNA, the spectrum shifts to that expected for a PNA-DNA heteroduplex. Similar results upon addition of PteloC are observed in the other salts, consistent with the SPR results.</p><p>CD spectra and cyanine dye binding experiments confirmed the formation of a PteloC:hTelo22 heteroduplex. As shown in Figure 9A, the induced CD signal from DiSC2(5) increased in the order KCl < NaCl < LiCl, consistent with increasing heteroduplex formation as the hTelo22 quadruplex is destabilized. The stronger CD signals compared with the Myc19 system (Figure 4) could reflect sequence preferences for DiSC2(5) binding to PNA-DNA duplexes, but this requires additional study to verify. In LiCl, the PteloC:hTelo22 heteroduplex is formed in a 2:1 stoichiometry based on Job plot analysis using the fluorogenic cyanine DiSC1(3) (Figure 9B; absorbance and emission spectra shown in Figure S2).</p><!><p>Previous findings from our labs and others have shown that complementary and homologous PNA probes bind intramolecular RNA/DNA quadruplexes and form stable PNA-DNA heteroduplex or heteroquadruplex structures, respectively. In earlier studies, we compared the binding of homologous and complementary PNAs to an RNA aptamer, which had been selected for binding to the Fragile X mental retardation protein and adopts a G-quadruplex structure.[24] There we found that that a 1:1 hybrid duplex and a 2:1 hybrid quadruplex were formed by complementary and homologous PNA probes respectively, but the complementary PNA was designed to target a central 7 nucleotide segment of the RNA, precluding 2:1 binding. In the current study, we specifically designed the complementary PNA to form a 2:1 complex to facilitate comparisons with the homologous PNA.</p><!><p>The Myc19 DNA quadruplex adopts a parallel morphology, with one-, two- and one-base loops formed as the strand folds back on itself in order to begin forming the next quadrant of the structure. The SPR results in Figure 1A and 1B illustrate the significant advantage possessed by the homologous PNA in invading the quadruplex structure in order to hybridize under high KCl conditions. The failure of the complementary PNA to bind under these conditions was also reflected in the CD results shown in Figures 3 and 4, indicating that even thermal annealing is insufficient to promote heteroduplex formation. (The consistency of these results also suggests that the parallel morphology of the Myc19 quadruplex is maintained whether in solution or immobilized on the SPR chip surface.) Thus, heteroquadruplex formation by the homologous PNA is both kinetically and thermodynamically favored over heteroduplex formation by the complementary PNA for this particular target. Only weakening of the DNA quadruplex by switching the cation to sodium or lithium allows the complementary PNA to bind to the Myc19 target. In contrast, Amato and coworkers reported that a short PNA hexamer (ACCCCA) was unable to invade a four-tetrad DNA quadruplex but could form stable heteroduplexes if thermally annealed with the target.[13c] The fluorogenic dye experiment shown in Figure 5 verifies that the complementary PNA forms a 2:1 heteroduplex, as designed.</p><!><p>In contrast to the strong preference for heteroquadruplex versus heteroduplex recognition of Myc19 by PNA, SPR results shown in Figures 6 and 7 indicate that heteroduplex formation is favored for PNA recognition of hTelo22 under all ionic conditions and Figure 9B is consistent with a 2:1 stoichiometry for the heteroduplex.</p><p>It will be interesting to compare the results for telomeric DNA with telomeric RNA (TERRA), which (i) is transcribed using the C-rich strand as a template, (ii) is known to fold in vitro into a parallel quadruplex, and (iii) localizes to telomeric regions.[25] The results described here suggest that homologous recognition of TERRA by quadruplex-forming PNA should be much more efficient than recognition of the corresponding DNA due to the parallel structure of TERRA. However, the corresponding SPR experiments are difficult to perform with immobilized RNA due to the harsh (i.e. high pH) conditions needed to regenerate the chip after PNA hybridization. The reverse experiment, namely hybridization of DNA or RNA to immobilized PNA could be illuminating, although such experiments would preclude 2:1 PNA-DNA/RNA hybridization. Nevertheless, other kinetic methods such as monitoring fluorescence enhancement of a PNA-appended dye such as thiazole orange might be useful for comparing hybridization rates with different targets in solution.</p><p>The results described here illustrate significant differences between quadruplex-invading PNAs that form heteroduplexes versus heteroquadruplexes and how kinetic preferences depend strongly on the structure of the quadruplex target. Although there have been numerous reports describing the intracellular effects of quadruplex-binding small molecules, biological effects of quadruplex-targeting PNAs have yet to be investigated. Significantly, a recent report from Xu and coworkers describes the ability of quadruplex-forming short RNAs to down-regulate expression of an EGFP reporter gene in live cells.[26] The high affinity of PNA and recent advances in delivery of PNA into cells are expected to lead to potent effects on gene expression.[27] It will be interesting to see if the differences reported here for complementary versus homologous PNA are reproduced in the context of targeting an intracellular reporter gene or endogenous gene.</p><!><p>DNA oligonucleotides were purchased from Integrated DNA Technologies (www.idtdna.com) and used without further purification. t-Boc protected peptide nucleic acid monomers were purchased from Applied Biosystems and used for standard solid phase synthesis of the PNA oligomers.[28] (PNA monomers are no longer sold by this company. Presently they can be purchased from ASM Research Chemicals; Hannover, Germany, or synthesized in-house). The PNA oligomers were purified by reverse phase HPLC and verified by MALDI-TOF mass spectrometry (Applied Biosystems, Voyager DE sSTR) using sinapinic acid as the matrix PmycH: expected m/z, 2460.2; found, 2462.52. PteloH: expected 2701.47; found, 2703.28; PmycC: expected m/z, 2181.3; found, 2179.1 PteloC: expected 2471; found, 2470.7.</p><p>All DNA and PNA concentrations were determined by measuring the absorbance at 260 nm at 85°C on a Cary 3 Bio spectrophotometer. At high temperatures the bases are assumed to be unstacked and the extinction coefficient of the oligomer is estimated as the sum of the individual bases. For the DNA oligomers the extinction coefficients were used as reported in literature.[29] The PNA extinction coefficients at 260nm were obtained from Applied Biosystems. (A: 13700 M−1cm−1; C: 6600 M−1cm−1; G: 11700 M−1cm−1 and T: 8600 M−1cm−1). The cyanine dye DiSC2(5) was purchased from Molecular Probes, Inc. (Eugene, OR) and used without further purification. Stock solutions were prepared in methanol, and concentrations were determined using the manufacturer's extinction coefficient (ε651 = 260,000 M−1cm−1). DiSC1(3) was synthesized by Dr. Gloria Silva. Product was characterized using ESI mass spectrometry (Thermo-Fischer LCQ ESI/APCI Ion Trap). (DiSC1(3), expected m/z for M+, 337.4; found, 337.3). UV-vis spectra for the compound matched the literature.[30]</p><!><p>UV-vis measurements were performed on a Varian Cary 3 spectrophotometer equipped with a thermoelectrically controlled multicell holder. Circular dichroism (CD) spectra were recorded on a Jasco J-715 spectropolarimeter equipped with a thermoelectrically controlled single cell holder.</p><!><p>SPR measurements were performed by using a BIACore 2000 system with streptavidin-coated sensor chips (SA) for all experiments. DNA was immobilized on the surface by noncovalent capture to streptavidin. To prepare sensor chips for use, they were conditioned with five consecutive 1-min injections of 1 M NaCl in 50 mM NaOH followed by extensive washing with HEPES buffer, pH 7.4 [0.01 M HEPES, 0.15 M LiCl, 3 mM EDTA, and 50 μl/liter Surfactant P20]. 5′-Biotinylated oligonucleotide (25nM) in coupling buffer (10 mM HEPES, pH 7.4, 3 mM EDTA, and 150 mM LiCl) was heated at 95 °C for 5 min and cooled slowly to room temperature, and then injected at a flow rate of 2 μL/min to achieve long contact times with the surface and to control the amount of the DNA immobilized. Approximately 420 response units (RUs) of either Myc19 or hTelo22 were immobilized on separate flow cells. This is approximately 4-fold higher surface density than we used previously for SPR analysis of PNA-DNA hybridization and provided improved signal-to-noise ratios at lower PNA concentrations without altering the observed kinetics (Figure S3). Direct binding experiments involved flowing homologous or complementary PNA over the chip surface as described previously.[16c]</p><!><p>CD measurements were performed on a Jasco J-715 CD spectropolarimeter equipped with water circulating temperature controller. Samples were prepared by mixing 1 μM DNA and 2 μM PNA together in 10 mM Tris-HCl (pH 7), 0.1mM EDTA and 100 mM Salt. Samples were annealed by heating to 95°C for five minutes and then slowly cooling to room temperature. All spectra were collected at 37°C by equilibrating the solutions at this temperature for 10min prior to recording. Each spectrum represents an average of 6 scans, collected at a rate of 100 nm min−1. The spectra were baseline corrected. For recording the induced CD spectra of DiSC2(5), 10 μM of DiSC2(5) dye was added into pre-annealed mixture of 2 μM DNA and 4 μM PNA. The spectra were recorded at 25°C by equilibrating the solutions at this temperature for 10 min prior to recording. Each spectrum represents an average of 2 scans, collected at a rate of 200 nm min−1. The spectra were baseline corrected.</p><!><p>Mixtures of DNA and 2 equivalents of PNA were annealed by heating to 95°C for 5 min and then slowly cooling to room temperature. The dye (DiSC2(5) or DiSC1(3)) was added to the samples and allowed to equilibrate at room temperature for 5 minutes after which the absorption spectra were collected using Varian Cary 300 Bio UV-Visible Spectrophotometer. Corresponding baseline corrections were made prior to absorbance measurement.</p><!><p>Emission (fluorescence) spectra of pre-annealed samples were collected using a Varian Cary Eclipse Fluorescence Spectrophotometer. Samples containing DNA, PNA and DiSC1(3) dye were prepared in a buffer solution containing 10 mM Tris-HCl (pH 7), 0.1 mM EDTA and 100 mM salt.</p><p>For determining PNA-DNA stoichiometries, continuous variation experiments were performed in which PNA and DNA were mixed at varying ratios but constant total concentration of 1 μM. Dye was also present at 1 μM concentration. Fluorescence spectra were recorded with excitation at 520 nm and the fluorescence intensity at 580 nm was plotted versus the PNA mole fraction to determine the empirical stoichiometry of PNA-DNA complex.</p>

PubMed Author Manuscript

Dynamic profiles and predictive values of some biochemical and haematological quantities in COVID-19 inpatients

IntroductionSevere acute respiratory syndrome coronavirus-2 (SARS-CoV-2) infection in some hospitalized patients has shown some important alterations in laboratory tests. The aim of this study was to establish the most relevant quantities associated with the worst prognosis related to COVID-19.Materials and methodsThis was a descriptive, longitudinal, observational and retrospective study, in a cohort of 845 adult inpatients from Bellvitge University Hospital (L’Hospitalet de Llobregat, Barcelona, Spain). A multivariate regression analysis was carried out in demographic, clinical and laboratory data, comparing survivors (SURV) and non-survivors (no-SURV). A receiver operating characteristic analysis was also carried out to establish the cut-off point for poor prognostic with better specificity and sensibility. Dynamic changes in clinical laboratory measurements were tracked from day 1 to day 28 after the onset of symptoms.ResultsDuring their hospital stay, 18% of the patients died. Age, kidney disease, creatinine (CREA), lactate-dehydrogenase (LD), C-reactive-protein (CRP) and lymphocyte (LYM) concentration showed the strongest independent associations with the risk of death in the multivariate regression analysis. Established cut-off values for poor prognosis for CREA, LD, CRP and LYM concentrations were 75.0 μmol /L, 320 U/L, 80.9 mg/L and 0.69 x109/L. Dynamic profile of laboratory findings, were in agreement with the consequences of organ damage and tissue destruction.ConclusionsAge, kidney disease, CREA, LD, CRP and LYM concentrations in COVID-19 patients from the southern region of Catalonia provide important information for their prognosis. Measurement of LD has demonstrated to be very good indicator of poor prognosis at initial evaluation because of its stability over time.

dynamic_profiles_and_predictive_values_of_some_biochemical_and_haematological_quantities_in_covid-19

Introduction<!>Subjects/Materials<!>Methods<!>Statistical analysis<!>Results<!><!>Results<!><!>Discussion<!>

<p>At the end of 2019, a cluster of pneumonia cases of unknown etiology was reported in Wuhan (Hubei Province, China). On January 30th, the World Health Organization (WHO) declared Coronavirus Disease 2019 (COVID-19) as a public health emergency of international concern (1, 2). The next day, the first case in Spain was declared. The virus responsible is a species of SARS-related coronavirus, a novel enveloped RNA virus, which was named severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) (1).</p><p>SARS-CoV-2 is transmitted person-to-person through droplets or direct contact, although it began as a zoonotic disease associated with a seafood market. Based on the evidence of rapidly increasing incidence of infections and the possibility of transmission by asymptomatic carriers, SARS-CoV-2 can be transmitted effectively among humans (3). In addition, the advancement and convenience of global travel may further enhance its worldwide spread (4). Primarily, COVID-19 patients have fever, myalgia or fatigue, and a dry cough at the time of admission. Although most patients have a mild clinical course with broad range symptoms, COVID-19 causes a complicated clinical situation for people with underlying conditions, including diabetes, hypertension, and cardiac, kidney and respiratory diseases, which result in rising rates of hospitalisation and mortality (5, 6). Notably, the highest number of comorbidities has been seen in infected patients admitted to the intensive care unit (ICU), suggesting that chronic diseases are likely to be risk factors for adverse clinical outcomes (4).</p><p>Since the outbreak of the pandemic, COVID-19 has been an important challenge for hospital care, both due to the significant increase in healthcare pressure and the need to optimise the clinical management of the new disease and its associated pathology, which was unknown at the time.</p><p>Beyond the recognised risk factors, several laboratory markers (biochemical, haematological and coagulation quantities) have been identified that indicate the course of COVID-19. The most studied biological quantities in COVID-19 disease are related to the main organs damaged in the infection by this virus: lung, kidney, liver, muscular system, heart. Determining the role of these dynamic blood biochemical changes in the course of the infection is of great importance, and plays an essential role in estimating patients' diagnosis and prognosis (2, 7). Moreover, identification of laboratory quantities capable of discriminating between cases with a low and high risk of mortality will allow for improved clinical situational awareness.</p><p>Most of the studies to date have been conducted with a limited number of cases and only a few biomarkers have been investigated. Therefore, in this retrospective study, we studied changes in laboratory tests in a large group of COVID-19 patients, including more laboratory quantities (7).</p><p>The aim of this study was to establish the most relevant laboratory quantities associated with the worst prognosis in order to identify early on patients at risk of severe disease progression, and to monitor the dynamic laboratory quantities during hospitalisation.</p><!><p>This was a descriptive, longitudinal, observational and retrospective study that includes a cohort of 845 adult inpatients from Bellvitge University Hospital (L'Hospitalet de Llobregat, Barcelona, Spain). The recruitment period was from March 16th 2020 to April 5th, 2020. The follow-up censoring data was June 8th 2020. This period of time includes the majority of patients diagnosed with COVID-19 in the first wave of the COVID-19 pandemic in our hospital.</p><p>Our study represents one of the few longitudinal studies on laboratory characteristics in COVID-19 patients with poor and good outcomes. To the best of our knowledge, it is the first study representing the Spanish population.</p><p>All consecutive adult patients discharged or deceased after hospital admission with SARS-CoV-2 infection were eligible for inclusion in the study. COVID-19 was diagnosed using a positive result of real-time reverse transcriptase-polymerase chain reaction (RT-PCR) testing of oro/nasopharyngeal swab sample according to the WHO interim guidance.</p><p>The inclusion criteria were: a) patients > 18 years, b) confirmed COVID-19 diagnosis, c) hospital discharge or death in hospital. The exclusion criterion was subsequent admission of the same patient.</p><p>Personal data relating to patients obtained through the present study were processed in accordance with the Regulation (UE) 2016/679 of the European Parliament on Data Protection. In addition, the study complied with the ethical principles for medical research involving human subjects adopted in the Declaration of Helsinki of the World Medical Association. The Research Ethics Committee of the Bellvitge University Hospital approved this paper for its publication.</p><!><p>Demographic, clinical and laboratory data were extracted retrospectively from electronic medical records and the laboratory information. Demographic data collected included age and gender, and the clinical data recorded were ICU admission, diabetes, hypertension, and cardiac, kidney and respiratory diseases. Data on the following laboratory quantities were collected: alanine transaminase (ALT), albumin (Alb), alkaline phosphatase (ALP), aspartate aminotransferase (AST), bilirubin (TBIL), calcium (Ca), creatine kinase (CK), creatinine (CREA), C-reactive protein (CRP), D-dimer (DD), ferritin (FER), gamma-glutamyltransferase (GGT), haemoglobin (Hb), interleukin-6 (IL-6), lactate dehydrogenase (LD), leukocytes (WBC), lymphocytes (LYM), neutrophils (NEU), platelets (Plt), potassium ion (K), procalcitonin (PCT) and troponin T (TnT). Laboratory data chosen depicted the second test result occurring within the 48h from presentation to the Emergency department, coinciding with the routine analytics, which presents a greater number of requested laboratory quantities. Due to the fact that it is a retrospective study, there are no results for all the laboratory parameters in all patients.</p><p>To determine the major clinical features that appeared during hospitalization, the dynamic changes in 22 clinical laboratory measurements, including haematological and biochemical measurements, were tracked from day 1 to day 28 after the onset of symptoms, at 3-day intervals. It was decided to pool the results of three days to obtain a greater number of data, since blood tests were not performed on all patients every day.</p><p>Venous blood was collected by routine phlebotomy using the BD Vacutainer Serum Separator II Advance Tubes (SST II) (ref. 366468) and the VACUETTE Plasma Lithium Heparin Blood Collection Tube (ref. 454029) for biochemical biomarkers, the VACUETTE TUBE 3.5 ml 9NC Coagulation sodium citrate 3.2% (ref. 474327) for DD, and the BD Vacutainer Plastic K3EDTA Tube 3ml with Lavender Hemogard Closure (ref. 368857) for haematological parameters.</p><p>Biochemical biomarkers were measured by spectrophotometric assays using a Cobas e502/e702 (Roche Diagnostics, Basel, Switzerland) and by electrochemiluminiscence assays, using a Cobas e602/e801 (Roche Diagnostics, Basel, Switzerland).</p><p>D-dimer was analysed by latex turbidimetric principle using an ACL-TOP 550/750 (Werfen Solutions, Barcelona, Spain).</p><p>Cell blood count was analysed by spectrophotometric, impedance and flux citometry using Sysmex XN-10/20 (Sysmex Corporation, Kobe, Japan).</p><p>All measurement systems have been correctly checked. Therefore, internal quality control materials have been processed daily, in accordance with the requirements of our laboratory. Quality specifications were met during the period studied.</p><p>The clinical laboratory is accredited according to the Internal Standard ISO 15189:2012.</p><!><p>Continuous variables were tested using the Shapiro-Wilk test to check the normality of distribution. Demographic, clinical and laboratory data were compared between survivors and non-survivors using the Chi-squared test for categorical data and the Mann-Whitney U test for continuous data.</p><p>To define the independent predictors associated with in-hospital death, a univariable logistic regression model was used. Then, a multivariable regression analysis was fitted. We excluded variables from the multivariable analysis if their univariable analysis was not significant. Some variables were excluded after carrying out a correlation matrix evaluation, which allows to assess the correlation between pairs of variables (Pearson's correlation coefficient). Those magnitudes with a correlation coefficient > 0.5 were discarded. The receiver operating characteristic (ROC) analysis was carried out to pick an optimum cut-off point in biomarker values that presented significant differences after multivariable analysis (throughout the entire hospital stay).</p><p>Results were considered to be statistically significant when P < 0.05. Statistical analyses were performed using the Stata software version 12.0 (College Station, TX: StataCorp LP).</p><!><p>Altogether 150 (18%) patients died during their hospital stay.</p><p>Demographic and clinical data of survivors (SURV) and non-survivors (noSURV) are shown in Table 1.</p><p>Laboratory data from the second set of tests after hospital admission and their comparison between groups are shown in Table 2. Test for normality using the Shapiro-Wilk test showed that data do not follow a normal distribution. The results of the univariate analysis are shown in Table 3.</p><p>In the final multivariate model age, kidney disease and CREA, and LD, CRP and LYM concentrations showed the strongest independent associations with the risk of death (Table 3).</p><p>Figure 1 and 2 show time series results of biochemical and haematological laboratory measurements.</p><!><p>Dynamic changes in laboratory biomarkers during hospitalisation. Dynamic changes at 4-day intervals in alanine transaminase, albumin, alkaline phosphatase, aspartate aminotransferase, bilirubin, calcium, creatine kinase, creatinine, C-reactive protein, D-dimer, ferritin, and gamma glutamyltransferase are shown.</p><p>Temporal changes in laboratory biomarkers during hospitalisation. Dynamic changes at 4-day intervals in haemoglobin, lactate dehydrogenase, leukocytes, lymphocytes, neutrophils, platelets, potassium, and troponin T are shown.</p><!><p>Results for ROC curves are shown in Figure 3. Areas under the curve (AUC) of CREA, LYM, CRP and LD concentrations were 0.67 (95% CI 0.62-0.72), 0.782 (95% CI 0.74-0.82), 0.76 (95% CI 0.71-0.80) and 0.80 (95% CI 0.75-0.84), respectively, and cut-off values were 75.0 μmol /L, 0.69 x109/L, 80.9 mg/L and 320 U/L.</p><!><p>Receiver operating characteristic (ROC) curves and cut-off point in laboratory biomarkers that have presented significant differences after multivariable analysis where P < 0.05 were considered statistically significant. AUC - area under the curve. 95% CI - 95% confidence interval.</p><!><p>The in-hospital mortality in our study was 18%, whereas Berenguer et al. reported a mortality rate in other multicenter studies in Spain of up to 28% (8). This variability can be attributed to the different criteria for patient admission and different resources available in each centre during the study period. The mortality was higher in men than in women and increased progressively with age as reported by Zhou et al., Zhang et al. and Berenguer et al. (3, 4, 8).</p><p>Procalcitonin, IL-6 and TnT did not show significance in the univariate regression despite the fact that their association with a poor prognosis has been widely described in the literature (9-11). Limited data available for IL-6 and PCT could affect the obtained results and it would be necessary to include more data in order to determine whether the expected results are obtained or not for this measurement.</p><p>Troponin T has been associated with poor prognosis in Spanish population, as stated by García de Guadiana-Romualdo et al., and also foreign populations, as stated by Rath et al. and Qin et al. (9, 12, 13). Regarding our data, significant differences were observed between SURV and noSURV groups, but the result of the univariate regression was not significant.</p><p>Ferritin did show statistically significant differences in the univariate model. However, it has not been included in the multivariate model due to its correlation with LD, which has been found as a predictor of progression toward severe forms.</p><p>Considering the dynamic profile of laboratory findings, most were in agreement with the consequences of organ damage and tissue destruction.</p><p>When serial results at 3-day intervals between SURV and noSURV were compared, similar results to laboratory findings from the second set of tests after hospital admission were obtained. The following quantities were significantly increased in noSURV patients: ALP, AST, TBIL, CRP, CK, CREA, FER, LD, K, PCT, TnT, DD, WBC and NEU. However, Alb, Ca, Hb, LYM and Plt concentrations were significantly lower in noSURV patients.</p><p>In noSURV group, the measurements which remained stable throughout the studied period were ALB, CK, PCT, K and LD concentrations. The quantities that trended towards an increase were TBIL, ALP, FER, AST, DD, WBC, NEU and Plt concentrations. The quantities that trended toward a decrease over time were LYM, Hb, TnT, Ca and CREA concentrations.</p><p>Concerning CREA concentration, despite being significantly higher in noSURV group, it showed a decreasing trend until death occurred. Some studies reported a progressive increase in CREA concentration in noSURV group (5, 7, 14, 15). Our data may suggest that a long hospitalisation period and the severity of illness are also associated with the loss of muscle mass, which is directly related to CREA concentration.</p><p>Mertoglu et al. reported a dynamic profile of higher CK concentration in SURV group than in noSURV (7). Considering the ability of the virus to cause tissue destruction or organ damage to the heart and muscle, our results may be consistent from a pathophysiological point of view.</p><p>Our study showed Plt concentration with a similar trend and a slight increase over time in both study groups. These results were consistent with Mertoglu et al., who obtained similar results in both groups (7). Results of published studies of dynamics of Plt concentration are variable. Ferrari et al. and Ding et al. showed a marked and increased trend in SURV group over time, and they highlighted this quantity to enhance the assessment of prognosis in hospitalised patients with COVID-19 (14, 15).</p><p>According to the measures determined to be significant by the multivariate analysis, a decreasing trend was found in CREA and LYM concentrations in noSURV group. The concentration of CRP, despite having a maintained tendency from day 4 to day 16, showed a decrease up to day 25, with a final rebound in its concentration at day 28. An established cut-off value for each quantity at initial analysis will provide an indicator of prognosis with a certain sensitivity and specificity, but they will vary throughout the period of hospitalisation. In contrast, LD showed a maintained trend in noSURV group throughout the studied period. These results were consistent with Ferrari et al., who showed a similar trend in the poor prognosis group from the 5th hospitalization day (14). The higher values of LD in noSURV group were in agreement with the extent injuries caused by the virus to lung and kidney. Establishing a cut-off value will be essential for initial analysis and a good indicator of prognosis with sensitivity and specificity values maintained over time.</p><p>There were several limitations of this study. First, despite having a great number of cases, this was a single-centre study. Second, the viremia of these patients was unknown. The viral load is a potentially useful marker associated with disease severity and it should be determined in all cases. Concerning laboratory data, as there was no established protocol to request analytics from patients with COVID-19 at the time the study was carried out, there are insufficient data for some measurements (PCT and IL-6) and they may be subject to bias.</p><p>This study shows that age, kidney disease, CREA, LD, CRP and LYM concentrations in COVID-19 patients from the southern region of Catalonia provide important information for their prognosis. Some measurements, such as LD, have demonstrated to be very good indicators of poor prognosis at initial evaluation because of their stability over time.</p><!><p>Potential conflict of interest</p><p>None declared.</p>

PubMed Open Access

Specificity and promiscuity in human Glutaminase Interacting Protein (GIP) recognition: Insight from the binding of internal and C-terminal motif

A large number of cellular processes are mediated by protein-protein interactions, often specified by particular protein binding modules. PDZ domains are an important class of protein-protein interaction modules that typically bind to the C-terminus of target proteins. These domains act as a scaffold where signaling molecules are linked to a multiprotein complex. Human Glutaminase Interacting Protein (GIP), also known as Tax Interacting Protein, is unique among PDZ domain containing proteins since it is composed almost exclusively of a single PDZ domain rather than one of many domains as part of a larger protein. GIP plays pivotal roles in cellular signaling, protein scaffolding and cancer pathways via its interaction with the C-terminus of a growing list of partner proteins. We have identified novel internal motifs that are recognized by GIP through combinatorial phage library screening. Leu and Asp residues in the consensus sequence were identified to be critical for binding to GIP through site-directed mutagenesis studies. Structure-based models of GIP bound to two different surrogate peptides determined from NMR constraints revealed that the binding pocket is flexible enough to accommodate either the smaller carboxylate (COO\xe2\x88\x92) group of a C-terminal recognition motif or the bulkier aspartate side chain (CH2 COO\xe2\x88\x92) of an internal motif. The non-canonical ILGF loop in GIP moves in for the C-terminal motif but moves out for the internal recognition motifs, allowing binding to different partner proteins. One of the peptides co-localizes with GIP within human glioma cells indicating that GIP might be a potential target for anti-cancer therapeutics.

specificity_and_promiscuity_in_human_glutaminase_interacting_protein_(gip)_recognition:_insight_from

<!>Protein expression and purification<!>Screening of the phage displayed peptide library<!>Phage binding assay<!>GIP-peptide titration by NMR<!>GIP-peptide models<!>Fluorescence spectroscopy<!>Immunocytochemical localization of GIP in cancer cells<!>Peptide internalization and co-localization with GIP in cancer cells<!>MTT assay<!>Identification of GIP-binding peptides by phage display<!>Binding Affinities Determination by Fluorescence Assay<!>GIP binding to internal motif peptides monitored by NMR spectroscopy<!>Chemical shift perturbation of GIP upon binding to internal motif peptide ligands<!>Role of the residues at P0 and P+1 of the peptide in GIP-peptide binding<!>Structural characterization of internal motif recognition by GIP<!>Co-localization of GIP and internal motif peptide in human glioma cells<!>Internal motif recognition by GIP<!>Mechanisms of internal motif recognition and comparison with canonical C-terminal recognition by GIP<!>Comparison of the binding of the ESSVDLLDG and GSGTDLDAS peptides to GIP<!>Evidence of internal motif recognition by GIP<!>Co-localization of GIP and internal motif peptide<!>New potential partner proteins of GIP<!>CONCLUSIONS

<p>PDZ domains, which are named after founding members Post synaptic density 95, Discs Large, and Zonula Occludens-1, are one of the most important protein-protein interaction modules found in living systems. These domains act as a scaffold where signaling molecules are linked to a multiprotein complex. PDZ domains mediate this organization of signaling complexes by recognizing the C-terminal amino acid sequence motifs of the interacting protein (1, 2). The most important functions of PDZ domains appear to be localization and clustering of ion channels (3), G-protein coupled receptors (4) and downstream effectors (5) at epithelial cell tight junctions, neuromuscular junctions and postsynaptic densities of neurons (6). These clustering and localization functions play significant roles in signal transduction pathways (7).</p><p>Glutaminase Interacting Protein (GIP) (8) also known as Tax Interacting Protein (TIP-1) (9) is a small globular protein (124 amino acid residues) uniquely composed of one PDZ domain that is flanked by flexible N- and C-termini. PDZ domains are small (80-100 residues) protein-protein interaction modules that typically bind the C-terminal motifs of the interacting partner proteins (10), but on rare occasions may interact with internal motifs that mimic a C-terminus (11, 12). To date, GIP has been shown to interact only with the C-termini of a growing list of partner proteins including Glutaminase L (8), HTLV-1 Tax (9), HPV E6 (13), β-catenin (14), Rhotekin (15), FAS (16) and Kir 2.3 (17). These GIP partner proteins play important roles in cell signaling, ion transport, transcription and/or cancer. GIP has also been shown to act as a scaffold in both astrocytes and neurons (18).</p><p>Discerning the protein interaction networks in and between different cell types forms the foundation for the design of new anti-cancer drugs. Thus, development of drugs targeting a specific protein is achievable when its network is fully characterized to minimize unwanted side-effects. To further explore the GIP interaction network, we used an f8-type phage displayed peptide library to screen for new GIP-binding peptides that may lead to new partner proteins. Such peptides may serve as leads for the development of novel anti-cancer therapeutics that specifically target GIP.</p><p>Here, we report the identification of 18 new GIP-binding peptides with novel internal motifs that map to a number of candidate human GIP partner proteins. All of these proteins are involved in various cancer pathways and/or other important cellular functions such as cellular adhesion, transcription, recombination and cell death. Alanine replacement studies confirmed that the identified internal-binding sequence motif is necessary for direct binding to GIP. Here, we report the structure-based models of internal motif binding to a PDZ domain obtained from docking of the peptide to the protein using NMR distance constraints obtained from intermolecular NOEs. Finally, we demonstrate that one of the peptides co-localizes with GIP inside human glioma cells and decreases their metabolic rate in culture.</p><!><p>Expression and purification of GIP was carried out as described previously (16). Briefly, the recombinant pET-3c/GIP plasmid was expressed in Escherichia coli (E. coli) BL21 DE3pLys cells, in M9 minimal media containing 13C-labeled glucose and/or 15N-labeled ammonium chloride. The overexpressed recombinant GIP was purified in a single-step by size exclusion chromatography on a Sephacryl S-100 column (GE Healthcare). Pooled fractions of the pure protein were exchanged to NMR buffer containing 50 mM sodium phosphate at pH 6.5, 1 mM EDTA and 0.01% (w/v) NaN3.</p><!><p>For identification of GIP-binding peptides, a pVIII 9-mer phage displaylibrary was used (19). The library contains 2×109 different phage clones with multivalently displayed foreign peptides, providing incredible diversity for finding target proteins in non-stringent conditions (20). Prior to the library selection, GIP was purified as described above and dialyzed against 0.1 M phosphate buffer at pH 8.0. Two wells of a Medisorp 96-well plate were coated with the purified protein at a 100 μg/ml concentration overnight at 4 °C. The protein-coated well was blocked with 1% Bovine Serum Albumin (BSA) in Tris Buffered Saline (TBS) for 1 h at room temperature. To select for the target-binding phage, an aliquot of 109 colony forming units (cfu) of the library (depleted on an unrelated target) was added to the well for additional 1 h incubation at room temperature. After incubation, unbound phages were discarded and the wells were washed 10 times with TBS containing 0.1% Tween 20 (TBST). The bound phage was eluted with 0.2 M glycine, pH 2.2 for 10 minutes and immediately neutralized using 1 M Tris-HCl, pH 9.1. The eluted phages were amplified in E. coli K91BluKan bacteria, purified and titered for the next round of selection. In rounds two and three, 1010 cfu aliquots were used in the selection procedures. After the third round, phage DNA in the area of the gpVIII gene was PCR amplified from 33 random phage-infected bacterial colonies, purified and sequenced. Sequences of GIP-binding peptides were deduced from phage DNA sequences using Chromas software.</p><!><p>Medisorp 96-well plates were coated with GIP at a 70 μg/ml concentration at 4 °C overnight and blocked with 1% BSA in TBS for 1 h at room temperature. An additional set of uncoated wells was also blocked for the negative control. The wells were washed with TBST washing buffer, pH 7.0 two times. Each selected phage clone was amplified individually and added at 5×106 cfu/well to the GIP-coated wells for 1 h incubation at room temperature. After incubation, the wells were washed 10 times with TBST washing buffer. Bound phage were recovered by adding 25 Hl of lysis buffer (2.5% CHAPS, 0.1% BSA in TBS buffer, pH 7.0) to the wells for 10 minutes at room temperature. After that, freshly prepared E. coli starved cells (125 Hl/well) were added to the wells for 15 minutes to allow phage infection. Next, 180 Hl of NZY broth (pH 7.5) containing 0.4 μg/ml tetracycline was added to each well and the plates were placed in a 37 °C incubator for 45 minutes. Finally, the content of each well was plated on NZY plates containing 20 μg/ml of tetracycline for overnight incubation at 37 °C. To quantify the phage, bacterial colonies were counted by a colony counter next morning.</p><!><p>Interaction studies were carried out by titration of 100 HM GIP with peptides containing several different internal sequences: ESSVDLLDG, ASSSVDDMA, GTNLDGLDG, GSSLDVTDN, GSGTDLDAS, and GSSAAVTDN. The target peptides were obtained with > 95 % purity from Chi Scientific (MA). The 10 mM stock solutions of the above peptides were prepared in 10 mM phosphate buffer at pH 6.5. The amide chemical shift perturbations (Δδ) were calculated as Δδ = ∣ Δδ15N∣/f + ∣ Δδ1H∣ (16). The introduction of the f factor and its value were justified by the difference in the spectral widths of the backbone 15N resonances and the 1H signals (15N range, 131.5-100.8 ppm = 30.7 ppm; 1H range, 10.1-6.6 ppm = 3.5; correction factor f=30.7/3.5 = 8.7). Thus, the correction factor f = 8.7 was used in order to give roughly equal weighting for each of the 1H and 15N chemical shift changes. For ligand titration experiments, uniformly 15N-labeled GIP was titrated with increasing concentrations of peptide to a GIP:peptide ratio of 1:10, and the corresponding two-dimensional {1H,15N} HSQC spectra were recorded. Beyond the ratio of 1:10, solid peptide was added in increasing amounts to an excess that approached saturation with protein to peptide ratios ranging from 1:40 to 1:140 for certain individual peptides.</p><!><p>To model the structure of GIP in complex with ESSVDLLDG and GSGTDLDAS peptides, we performed the following experiments: 2D TOCSY (21) and 2D ROESY (22) on each peptide, 2D 15N/13C F1, selectively filtered NOESY (23), 3D 13C-edited/filtered HSQC-NOESY and 3D 15N-edited/filtered HSQC-NOESY (24) on each peptide/protein complex. The sample contained ~400 UM uniformly 15N/13C labeled GIP, unlabeled 8 mM ESSVDLLDG or 16 mM GSGTDLDAS, 50 mM phosphate buffer containing 5 % D2O (pH 6.5), 1 mM EDTA and 0.01% (w/v) NaN3. Peptide-peptide and peptide-protein NOEs were added to the set of previously determined protein NOEs from free GIP for structure calculation using ARIA (25). Previous studies on the binding of GIP to various peptides by both X-ray crystallographic and NMR methods demonstrated that the core structure of GIP is not significantly affected by the ligand binding (26-28). In our previous study (28), we solved the NMR structure of GIP in the free-state and also in the bound-state with a known ligand from the protein Glutaminase using a whole new set of NOEs obtained from the NOESY data collected on the complex. The overall three-dimensional structure of GIP both in free and bound forms were the same except for minor conformational changes in the ligand binding regions of the protein in the bound form. The above NMR observation (28) was consistent with the results of X-ray structures of GIP bound with β-catenin (27) and KIR 2.3 (26). Thus, both NMR and X-ray studies showed that the overall structure of GIP remains unaffected except for minor conformational changes at the binding site to accommodate the ligand (26-28). Additionally, the chemical shift perturbations of GIP titration with the above 3 peptides were reported to be significant only at or near the binding regions (16, 28). Interestingly, in the present study, the chemical shift perturbations observed for GIP when titrating with the different internal motif peptides were very similar to what was observed previously for Glutaminase, β-catenin and KIR 2.3 (16, 28) indicating that the overall structure of GIP remains unaffected except for the ligand binding regions. To build the NMR models of GIP complexed with two different ligands (two different internal motif peptides) accurately, we did not follow the usual procedure of simply docking the ligand to the three-dimensional structure of a protein using the intermolecular NOE constraints between the protein and ligand. Instead, we removed the intra-NOE distance constraints of the regions of the protein (the α2 helix and the β2 strand) that form the binding site. This approach provided flexibility to the regions of the protein involved in ligand binding allowing them to adopt the conformational change induced upon ligand binding. Next, the experimentally derived intermolecular NOE constraints between GIP to each peptide (Table S1) were added to the structure calculation process carried out with the program ARIA. We used 37 and 32 intermolecular NOE distance restraints that were experimentally identified between the ligand binding regions of the protein to each peptide in the two GIP-peptide complexes. The rest of the structure calculation process to determine the structure-based model of each complex was followed as described previously (28). The above experimental intermolecular NOE distance constraints were critical for the determination of the NMR models of the ligand-bound proteins that showed the ligand-induced conformation changes. The NMR experiments for free GIP and each GIP-peptide complex were conducted under identical conditions such as pH, temperature, buffer, protein concentration etc. On an iterative basis, the structures were evaluated and refinements made to the ARIA inputs using VMD (29) to visualize the structures. For the final ensemble of structures, out of total of 200 starting structures, 25 structures with lowest energy were chosen for water refinement. Of those, 20 structures with the lowest energy were selected for analysis with Procheck (30).</p><!><p>All fluorescence spectra were recorded on a PerkinElmer LS 55 Luminescence spectrofluorometer at 25 °C. Titration experiments were done as described previously (16). The dissociation constant KD was determined using the OriginPro 6.1 software. The equation corresponding to single binding site was used to fit the data as described previously (31) .</p><!><p>D54 MG human glioma cells were plated onto 4-chamber slides (Nunc, Naperville, IL) at the density of 3×104 cells/chamber in Dulbecco's modified Eagle's medium (DMEM F-12) supplemented with 10% fetal bovine serum (FBS) and grown under 5% CO2 at 37 °C for 24 h. For immunocytochemical localization of GIP, the cells were fixed by 1% paraformaldehyde in PBS for 30 min, and then permeabilized with 0.5% Triton X-100 in PBS for 25 min at room temperature. The cover slips were blocked with MACS buffer (0.5% BSA, 2 mM EDTA in PBS, pH 7.2) for 1 h. The cells were incubated with primary anti-GIP mouse monoclonal IgG (Novus Biologicals, Littleton, CO) diluted 1:15 in PBS with 1% BSA, overnight at 4 °C. The primary antibody was removed and the slides were rinsed with PBS. Secondary goat anti-mouse Alexa 488-conjugated antibody (Novus Biologicals, Littleton, CO) diluted 1:40 in PBS/BSA was added and incubated for 1 h at room temperature. Unbound secondary antibody was removed by washing with PBS. Slides were mounted with cover slips using Vectashield DAPI mounting medium (Vector Laboratories, Inc., Burlingame, CA). Fluorescence images were acquired with an Olympus BH-2 fluorescence microscope equipped with Nikon Digital Sight DS-L1 camera.</p><!><p>To demonstrate the ability of the peptide to be internalized by human glioma D54 MG cells, the cells were plated on chamber slides and cultured overnight as described in the previous section. The cells were treated with TAMRA-labeled ESSVDLLDGGG(R)7 peptide at 1 HM for 25 min. After incubation, the cells were washed three times with PBS and fixed with 1% paraformaldehyde for 15 min. Fixed cells were mounted with cover slips using Vectashield DAPI mounting medium. The slides were evaluated by fluorescence microscopy.</p><p>For GIP-peptide co-localization studies, cells plated on chamber slides as above were treated with TAMRA-labeled ESSVDLLDGGG(R)7 peptide at 1 HM for 25 min and fixed with 1% paraformaldehyde for 15 min, followed by three PBS rinses. Fixed cells were permeabilized with 0.5% Triton X-100 in PBS for 25 min, rinsed three times with PBS, and blocked with MACS buffer for 1 h at room temperature. The cells were then incubated with primary anti-GIP mouse monoclonal IgG antibody overnight at 4 °C and washed three times with PBS as in the previous section. Fluorescein Alexa 488 anti-mouse secondary antibody was then added and incubated for 1 h at room temperature. After washing the cells three times with PBS, the cells were mounted and evaluated by fluorescence microscopy.</p><!><p>The effect of the peptide on D54 MG cells was examined by an assay that utilizes MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) (Sigma-Aldrich, St. Louis, MO) salt. This assay measures cellular oxidative metabolism. The dye is cleaved to a colored product by the activity of NAD(P)H-dependent dehydrogenase, and this indicates the level of energy metabolism in cells. The color development (yellow to blue) is proportional to the number of metabolically active cells. For these experiments, D54 MG cells were plated on 96-well culture plates at a density of 3× 103 cells/well and cultured overnight in DMEM F-12 medium (Mediatech Inc, Manassas, VA) containing 10% fetal bovine serum at 37 °C. Next day, the peptide was added to the cells at 10, 20, 40, 50, 75, 100, 200 HM concentrations. The cells were incubated at 37 °C until the total treatment time reached 16 h. After that, 10% volume of MTT stock solution (5 mg/ml) was added to the cell cultures for four hours for color development. The converted dye was then solubilized, and the absorbance was measured at 550 nm. Each data point was normalized against the control cells.</p><!><p>GIP-binding peptides were selected from a f8-type 9-mer phage displayed peptide library (32) that displays 4000 copies of the foreign nonamers in the N-terminal part of the major coat protein pVIII of phage fd–tet (landscape library). The library was constructed by replacement of amino acids 2–5 of pVIII with random nonamers. The landscape library allows selection of highly homologous families of peptides in non-stringent conditions due to its multivalency and avidity effect (20) with easily recognized binding motifs (33). To reveal GIP-binding motifs, the gene gpVIII DNAs were amplified by PCR from 33 phage clones, sequenced, and translated into 18 unique peptide sequences. Based on sequence alignment, they were placed into two groups (Table 1). Group 1 contained peptides with S/T-S-V/L-Da as a common motif. Interestingly, this motif was identified in different positions within the nine amino acid peptide sequences, including 2-5, 3-6, and 4-7. Group 2 contained a three residue N-L-D motif, which occupied positions from 2-4 and 3-5 within the peptides. An additional sequence, GSGTDLDAS, was also identified. Comparative analysis of all sequences revealed S/T-X-V/L-D to be the consensus motif.</p><p>The specificity of the selected phage clones to GIP was confirmed through a phage-binding assay by comparing relative binding of individual phage clones to the target protein in comparison with the controls, BSA or empty wells of the plastic plates used for phage selection. As an additional control for binding specificity, the above assay was repeated with phage f8-5, the vector that does not display any fusion foreign peptides (19). Equal numbers of individual phage clones were added to the wells containing either GIP or the above controls followed by incubation and quantification of the bound phage by titering in the host E. coli cells K91BK. It was observed that GIP-selected clones do not bind either to BSA or to the plastic. The vector phage alone did not bind to GIP (not shown).</p><!><p>Fluorescence assays involving titration of the protein to the peptides were studied by monitoring the decrease in the protein fluorescence by the addition of increasing amounts of various peptides. The KD values were calculated from the fluorescence intensity of GIP by plotting (F0-FC)/(F0-Fmin) versus [C] where F0 and FC, are the fluorescence intensities of the free protein and of the protein at a peptide concentration [C], and Fmin, the fluorescence intensity upon saturation of all ligand binding sites of the protein was obtained.</p><p>A plot of (F0-FC)/(F0-Fmin) versus [C] was established using an equation that defines a single binding site. The data were fitted to this plot to obtain the KD values using the OriginPro 6.1 software. The KD values of the internal motif peptides were within the range of 0.2-0.8 HM suggesting a moderate affinity of GIP for these peptides.</p><!><p>Several GIP-specific peptides revealed in the selection experiments were synthesized to assess their interactions with GIP using NMR spectroscopy. Five peptides representing motifs with either S/T or V/L amino acids in positions P−2 or P0 according to standard PDZ nomenclature (28), were selected for the NMR studies.</p><p>Chemical shift mapping is a powerful method frequently employed to investigate possible protein ligand interactions by NMR. The 2D 1H-15N HSQC spectrum provides the fingerprint region of a protein. This NMR experiment is a sensitive technique to study protein-ligand interactions in solution (16, 28, 34). Any perturbation in the chemical shift resonances from their original positions in this region indicates a change in the local environment of the affected residues within a protein (16). Based on the local chemical shift changes, we know that the overall fold and shape of the protein remains unchanged upon ligand binding as similar changes were observed in structures solved for GIP-peptide complexes with C-terminal peptides from Glutaminase-L (28), KIR2.3 (26) and β-catenin (27). To elucidate a molecular mechanism of GIP-ligand binding, we studied the interaction of GIP with selected peptides by 2D HSQC titration experiments. The amide proton and nitrogen resonances in the HSQC spectra were followed for each titration point. Resonances from most of the residues of GIP followed fast exchange kinetics on the NMR time scale as observed by gradual and systematic changes in their chemical shift positions (Fig. 1A). A few specific residues such as Leu29 and Gly30 followed intermediate exchange kinetics as seen by the disappearance of these peaks (Fig. 1A). The decrease in peak intensity of these residues is due to the exchange between amide resonances of free and bound GIP. Residues Leu29 and Gly30 are a part of the ILGF binding loop that makes specific hydrogen bonds to the negatively charged terminal carboxylate group of the partner protein with a C-terminal recognition motif during binding (28). This causes large chemical shift perturbations in these residues (35) despite very small structural changes (26-28). For our titration experiments, the magnitude of changes in the chemical shifts of residues in GIP can be correlated to the relative proximity to the peptide in the complex.</p><!><p>The chemical shift perturbation for each residue was calculated from the chemical shift changes of both 1H and 15N nuclei. When internal motif peptides were added, systematic changes of the amide resonances occurred in the titration spectra (Supplementary Information, Fig. S1). The significant chemical shift perturbations were grouped into three categories: medium shifted (>0.1 ppm), large shifted (>0.2 ppm), and intermediate exchange (Table 2). The intermediate exchange for certain residues within or very near the ILGF loop indicates that this loop is highly flexible as it has dramatically different kinetic properties compared to the rest of the protein. Unfortunately, because of the intermediate exchange that greatly broadens the resonances, the exact kinetic parameters of this region could not be studied. The magnitudes of the amide chemical shift changes upon binding to different internal motif peptides are mapped on to the ribbon diagram of GIP as indicated by different colors (Fig. 2).</p><p>Chemical shift perturbation analysis shows that the ILGF loop, β2 strand, and α2 helix are the regions of GIP that are most affected. It also shows that residues in the region Ile18-Gln23, Ile55-Glu62, and Glu67 which belong to the β1, βa, and β3 strands as well as the α1 helix are also affected, but to a lesser extent. This observation suggests that the peptides with internal binding motifs bind to the same binding site nestled between the β2 strand and α2 helix of GIP as the canonical C-terminal motif. This binding is allosterically driven, reminiscent of the way GIP binds to C-terminal motifs (16, 26-28).</p><!><p>To analyze the role of specific residues in the internal motif recognition by GIP, we created a double alanine substitution for LD in the GSSLDVTDN peptide. NMR titrations were performed to determine the effect of these substitutions on GIP binding. GIP was titrated with various concentrations of the alanine substituted peptide GSSAAVTDN. Compared with the wild-type GSSLDVTDN peptide (Fig. 1A), the chemical shift perturbation is negligible for the AA substitution (Fig. 1B). This indicates that any interaction between the peptide and GIP was completely eliminated. Interestingly, the observation that GSSAAVTDN peptide has virtually no binding to GIP suggests that both L and D are important for optimal interactions. Titrations with each of the identified peptides show that Leu29 and Gly30 are always in intermediate exchange for both residues (Table 2). Since LD or VD is present in each peptide and Leu29 and Gly30 are in intermediate exchange for the titration of each peptide, this supports our hypothesis that LD interacts directly with Leu29 and Gly30 of the ILGF carboxylate binding loop as a mimic of a hydrophobic C-terminal residue from a canonical C-terminal motif.</p><!><p>Structure-based models of GIP bound to each of the two internal motif peptides were obtained through docking studies using intermolecular NOEs measured by NMR. These docking studies used experimentally derived NOE distance restraints that provided the details of the interactions between each internal motif peptide and the GIP protein. We also used the intrapeptide NOEs from the peptide while it was bound to GIP to determine the internal structure of each peptide in the complex. The chemical shift perturbations of GIP binding to the internal motif peptides, ESSVDLLDG and GSGTDLDAS, were separately mapped onto the same region as that of the C-terminal peptides reported earlier in our laboratory (16, 28). The chemical shift perturbation studies detailing which regions of the GIP protein were most affected upon binding to the internal motif peptides showed similar patterns as those for previously solved complexes with GIP and C-terminal binding peptides (16, 28). This similarity in binding patterns allowed us to use our previously solved structure of the protein as a starting point in our structure-based model using the experimentally derived intermolecular NOEs between the GIP protein and each of the internal motif peptides. The experimentally derived NOEs demonstrated that each peptide bound to the protein in an extended strand conformation analogous to the previously determined C-terminal binding peptides (Supplementary Information, Table S1). There are four critical points of contact between GIP and both internal motif peptides. First, it binds by β-strand addition to form an antiparallel β-sheet with the β2 strand from GIP. Both peptides bind to GIP as antiparallel β-stands through this process. Second, the hydrophobic residue at P0 buries itself into the hydrophobic pocket created by Leu29, Phe31, Leu97 and Ile33. For ESSVDLLDG and GSGTDLDAS the role of P0 is taken by V4 and L6 respectively. Both side chains bury themselves into the hydrophobic pocket in the same way and with the same relative orientation. Third, either S or T at the P−2 position forms a hydrogen bond with His90 at the α2:1 position in GIP. Both ESSVDLLDG (Fig. 3A,B) and GSGTDLDAS (Fig. 3C,D) have more than one S or T in their respective sequences but it is S2 and T4 which are at the P−2 position from V4 and L6 at the P0 position within each peptide. The fourth and perhaps most important key point of contact is between the negatively charged carboxylate group of the ligand with the backbone amides from Leu29 and Gly30 within ILGF loop of GIP. For canonical binding this takes place with the C-terminal carboxylate of the interacting partner (28). For the internal recognition motifs, the side chain of aspartate acts as a substitute for the C-terminus. This role is taken by D5 and D7 in ESSVDLLDG and GSGTDLDAS respectively. In order to bind to the side-chain carboxylate in an internal recognition motif, we found that the flexible loop between the non-canonical βb and β2, which includes residues Leu29 and Gly30 in the ILGF loop adjust slightly so that the amide protons orient themselves toward the side-chain carboxylate to form a set of hydrogen bonds similar to the set of hydrogen bonds formed to the C-terminus during canonical binding.</p><!><p>GIP has been reported to be involved in many cancer pathways and represents a promising drug target (14, 36, 37). Our searches of protein databases (UniProt) indicated multiple cancer-related proteins containing the novel internal motif identified through our phage library screen (Supplementary Information, Table S2). We studied intracellular distribution of one of the peptides in human D54 MG glioma cells. The cells were treated with a synthetic ESSVDLLDG peptide fused to a cell-penetrating peptide G2R7 labeled with TAMRA. By fluorescence microscopy, the labeled peptide was shown to be uniformly distributed in the cytoplasm of the glioma cells (Fig. 4A). Next, cultures of D54 MG cells were treated with the TAMRA-labeled peptide followed by GIP staining with an anti-GIP antibody detected with a secondary antibody conjugated to Alexa 488 (Fig. 4). Both, the peptide (red) as well as GIP (green), were found to be co-localized in the cytoplasm of the cells. To investigate whether the above peptide will have any effect on the glioma cells, the cells were treated with the peptide at concentrations ranging from 10 to 200 μM for 16 h and their metabolism was measured by the MTT assay. The cell metabolism was suppressed in a dose dependent manner with increasing peptide concentrations (Fig. 5). The peptide concentration required to suppress 50% of the cell metabolism (IC50) was found to be equal to 69±10 μM.</p><!><p>In this study, a phage landscape library f8/9 with multivalently displayed foreign nonamers was used to identify new binding motifs for GIP, a single PDZ domain containing protein. The library used here was diverse, composed of two billion different phage clones. A randomized DNA segment was inserted into the N-terminus of the gene gpVIII that encodes the major phage coat protein (32). PDZ-binding phage clones were isolated from the library in three successive rounds of biopanning. In the GIP-phage binding assay, all of the selected phage clones were confirmed to be specific to GIP. Analysis of the peptide sequences led to the identification of a consensus internal-binding motif S/T-X-L/V-D. In the majority of previously reported phage display studies on PDZ-binding motifs, the identified peptide ligands were C-terminal recognition motifs (6, 38). To our knowledge, this is the first report of GIP recognition of internal binding motifs. In the selected sequences, S or T, which were followed by variable amino acids in position P−1, always occupied the P−2 position. Position P0 was always occupied by V or L, but not by I. This might indicate that steric factors are involved in the binding, thus, only the symmetric V or L side chains fit into the hydrophobic cavity, but not the asymmetric I side chain. Aspartate in the P+1 position was absolutely required.</p><!><p>Here, we also report structure-based models of PDZ domain recognition by two distinct internal motif peptides (Fig. 3). The binding of ESSVDLLDG and GSGTDLDAS to GIP shows key similarities to and differences from the canonical PDZ C-terminal binding by GIP with its interacting partner proteins. The similarities include: the β-strand addition mechanism, S or T at P−2 forms a hydrogen bond with His90, and V or L at P0 binds within the hydrophobic pocket created by Leu29, Phe31, Ile33 and Leu97. This explains the similar pattern of chemical shift perturbations within GIP for the binding of different internal motif peptides. The key difference is that in an internal motif P0 is not the C-terminus with a free carboxylate group. Instead a hydrophobic residue at P0 and D at P+1 serve as a structural mimic of a C-terminus with the side-chain carboxylate group of D forming the same set of hydrogen bonds to the backbone amides from Leu29 and Gly30 within the ILGF binding loop. Aspartate at P+1 has a different geometry than a C-terminal carboxylate group and needs to accommodate four additional heavy atoms. As a result, the backbone atoms of V/L at P0 and D at P+1 of the peptide loop around so that the side chain carboxylate group of D at P+1 points back toward the binding pocket. Analysis of the identified phage sequences shows that D is absolutely conserved among all the internal binding motifs. Each synthetic peptide derived from a phage clone did bind to GIP as monitored by NMR titrations. Thus, while E is also negatively charged, it appears that its side-chain is simply too bulky for the geometry to accommodate the binding pocket in an energetically favorable way.</p><p>Furthermore, while both Leu29 and Gly30 make the same set of hydrogen bonds to either a canonical C-terminal carboxylate group or carboxylate group from the side-chain of D at P+1 for an internal motif, the ILGF loop appears to be somewhat flexible and accommodating. It moves in to bind to a terminal carboxylate group of a C-terminal motif or moves out to bind to a carboxylate group from D at P+1 of an internal motif. The flexibility of this loop may be due to the non-canonical βa-βb hairpin loop of GIP. In most PDZ domains, the GLGF motif, also known as the carboxylate-binding loop comes directly between β1 and β2. However, in GIP, the non-canonical βa-βb hairpin loop uniquely positions the ILGF carboxylate-binding loop at a pivot point between βb and β2, thus allowing it to accommodate both sets of geometries for a terminal carboxylate group of C-terminal motif or a side-chain carboxylate group of D at P+1 for an internal motif.</p><p>Previously, X-ray crystal structures of a PDZ domain with internal motifs were solved (11, 12, 39). X-ray structures show that GLGF motif plays an important role for the interaction process. Interestingly, our GIP-peptide model suggests that the ILGF motif of GIP moves out to accommodate the internal motif. This flexible nature of the GLGF/ILGF motif helps to recognize both C-terminal and internal motif ligands.</p><!><p>Both ESSVDLLDG and GSGTDLDAS bind as part of an antiparallel β-sheet to the β2 strand. However, after P+1, the C-terminal segments of the peptides diverge in different directions. The direction of divergence appears to be controlled by whether P0 is L or V. For ESSVDLLDG, the alignment of V at P0 in the hydrophobic pocket of GIP followed by the alignment of D at P+1 allows the rest of the peptide to continue roughly antiparallel to βb. The hydrophobic L at P+3 makes a hydrophobic contact with Leu27 that further contributes to the binding stability of this particular peptide to GIP. In the case of GSGTDLDAS, the positioning of the larger hydrophobic residue L at P0 into the hydrophobic pocket causes D at P+1 to be positioned such that the remaining A and S at positions P+2 and P+3 point away orthogonal to both βb and β2. Also in contrast to ESSVDLLDG, GSGTDLDAS appears to form a slightly more extended antiparallel β-sheet with β2. Overall, it appears that binding to GIP the following conditions must be met: the ability to form a β-strand and the sequence S/T-X-L/V-D. ESSVDLLDG has both VD and LD pairs in its sequence, but only VD binds to the ILGF loop because it contains S at the relative position P−2. However, if the LD pair was bound to the ILGF loop, D would be at P−2 instead of S which is not energetically favorable since His90 is present at position α2:1. The His90 at α2:1 is responsible for the selectivity of S/T at P−2 of the interacting partner.</p><!><p>Very interestingly, endonuclein, a cell cycle regulated WD-repeat protein, was recently reported to interact with GIP (40). Endonuclein does not contain a canonical C-terminal PDZ binding motif, but contains the sequence EISGLDL (387-393) within its five WD repeats. WD repeats are β-sheet domains that contain multiple β-hairpin turns. It is possible that endonuclein interacts with GIP through this region that serves as an internal motif. If confirmed, this would be the first independent example of an interaction of GIP with a non-canonical internal motif.</p><!><p>GIP was shown to have the same subcellular localization (Fig. 4) as the synthetic peptide, ESSVDLLDG. The peptide was found to inhibit metabolism of the glioma cells in a preliminary test.</p><!><p>Using protein database searches, we have identified several proteins with the S/T-X-V/L-D internal motif that were previously shown to be involved in various cancer pathways and tumorigenesis (Supplementary Information, Table S2). For example, reduced expression of the mediator complex subunit 1 (MED 1) protein containing the above motif was associated with a more pronounced tumorigenic phenotype in human melanoma cells (41). The CYLD gene that encodes the cylindromatosis 1 protein also has this motif and was found to be down-regulated in human hepatocellular carcinoma cells and involved in their apoptotic resistance (42). Growing evidence indicates that CYLD deficiency may promote the development of various cancers (43). Another S/T-X-V/L-D internal motif-containing protein, MYO18B, was suggested to act as a tumor suppressor in the development of lung cancer (44). The MYO18B protein was also shown to play an essential role in ovarian cancer (45).</p><!><p>Our studies reveal new internal recognition motif for GIP. GIP recognizes target proteins containing S/T-X-V/L-D internal motif. This is the first report of GIP recognition of an internal motif. We have identified 18 new target proteins containing the above internal motif expanding the GIP interaction network. Structure-based models of GIP-peptide complexes reveal that the binding pocket of GIP is flexible and can accommodate either C-terminal or internal recognition motifs. The involvement of GIP in many cancer pathways suggests that this protein might be a potential target for drug design.</p>

PubMed Author Manuscript

Fractionation and evaluation of radical-scavenging peptides from in vitro digests of buckwheat protein \xe2\x8b\x86

Buckwheat protein (BWP) isolate was subjected to a two-stage in vitro digestion (1 h pepsin followed by 2 h pancreatin at 37 \xc2\xb0C). The antioxidant potential of the BWP digests was compared by assessing their capacity to scavenge 2,2\xe2\x80\xb2-azinobis (3-ethylbenzothiszoline-6-sulphonic acid) (ABTS+\xe2\x80\xa2) and hydroxyl (\xe2\x80\xa2OH) radicals. The 2-h pancreatin digest, which demonstrated the strongest activity against both radicals, was subjected to Sephadex G-25 gel filtration. Of the six fractions collected, fractions IV (456 Da) and VI (362 Da) showed the highest ABTS+\xe2\x80\xa2 scavenging activity and were 23\xe2\x80\x9327% superior to mixed BWP digest (P < 0.05). Fraction VI was most effective in neutralizing \xe2\x80\xa2OH and was 86 and 24% more efficient (P < 0.05) than mixed BWP digest and fraction IV, respectively. LC-MS/MS identified Trp-Pro-Leu, Val-Pro-Trp, and Val-Phe-Pro-Trp (IV), Pro-Trp (V) and tryptophan (VI) to be the prominant peptides/amino acid in these fractions.

fractionation_and_evaluation_of_radical-scavenging_peptides_from_in_vitro_digests_of_buckwheat_prote

1. Introduction<!>2.1. Extraction of buckwheat protein (BWP)<!>2.2. Preparation of protein digests<!>2.3. Gel filtration<!>2.4. Radical-scavenging activity (RSA)<!>2.4.1. \xe2\x80\xa2OH scavenging<!>2.4.2. ABTS+\xe2\x80\xa2 scavenging<!>2.4.3. LC-MS-MS<!>2.5. Statistical analysis<!>3.1. RSA of BWP in vitro digests<!>3.2. Peptide fractionation<!>3.3. RSA of peptide fractions<!>3.4. Peptide sequence<!>

<p>The gastrointestinal (GI) tract is one of the most vulnerable tissues inside the human body to oxidative attack by reactive oxygen species (ROS). Oxidative stress is believed to be one important cause of GI inflammation, ulcer, and colitis (Blau, Rubinstein, Bass, Singaram, & Kohen, 1999). The upper GI mucosa, itself a natural defense layer, is constantly exposed to ROS derived from endogenous as well as exogenous sources, i.e., foods, which can contain high amounts of unsaturated lipids, prooxidant transition metal ions and even directly, free radicals. For example, a diet containing iron and ascorbic acid in the presence of unsaturated fatty acids predisposed the GI lining to hydroxyl radical (•OH) mediated injury, which can lead to colitis (Carrier, Aghdassi, Platt, Cullen, & Allard, 2001). Moreover, it has been demonstrated that •OH can form in the gastric juice, and the radical generation is implicated in GI mucosa damage and ensuing ulcer (Nalini, Ramakrishna, Mohanty, & Balasubramanian, 1992). Hence, identifying potential antioxidants that may help neutralize radicals, particularly •OH, thereby protecting the GI system, is of great importance.</p><p>There has been growing interest in recent years to produce bioactive peptides that can exert radical scavenging activity. Carnosine, a naturally occurring dipeptide rich in muscle foods, is a classical example of peptides that can act as a strong radical scavenger and inhibit ROS-initiated lipid oxidation (Boldyrev & Johnson, 2002). Most reported antioxidants are derived from common food protein sources using commercial enzymes. For example, canola protein hydrolysate prepared using Flavourzyme was shown to be antioxidative and can enhance water-holding capacity in cooked pork meat (Cumby, Zhong, Naczk, & Shahidi, 2008). Hydrolyzed animal proteins, e.g., gelatin hydrolysate from Alaska polluck skin (Kim, Kim, Byun, Nam, Joo, & Shahidi, 2001), also show antioxidant activity in food model systems.</p><p>Proteins in raw and processed foods can possess antioxidant peptide sequences and structural domains; the active fragments are released during the GI digestion process. Reported high-efficiency radical scavenging peptides released through in vitro pepsin and pancreatin digestion include those from casein (Hernandez-Ledesma, Amigo, Ramos, & Recio, 2004), maize zein (Zhu, Chen, Tang, & Xiong, 2008), oyster protein (Crassostrea gigas) (Qian, Jung, Byun, & Kim, 2008), and mussel protein (Mytilus coruscus) (Jung et al., 2007).</p><p>Buckwheat, a traditional grain widely considered as a functional food source, has gained its fame due to published studies that linked its proteins to various health benefits, e.g., cholesterol reduction (Kayashita, Shimaoka, Nakajoh, Yamazaki, & Norihisa, 1997), tumor inhibition (Liu et al., 2001), and hypotension regulation (Ma, Bae, Lee, & Yang, 2006). Because many of the health promoting functions are inherently related to the radical scavenging activity of peptides from the protein digests, it is hypothesized that hydrolysis of buckwheat protein can release the peptide fragments capable of stabilizing ROS and inhibiting lipid oxidation. A preliminary study supported this hypothesis (Ma & Xiong, 2009). However, the specific peptides or peptide fractions responsible for the antioxidant functions have not been elucidated.</p><p>In the present study, the ability of mixed as well as individual fractions of in vitro pepsin-pancreatin sequential digests of buckwheat protein to stabilize •OH and ABTS+• radicals was investigated. The objective was to identify the most effective antioxidant peptide fraction(s) from buckwheat in vitro digests. Initially, the digest with the highest radical scavenging capacity was fractionated by means of gel filtration. The ability to stabilize hydroxyl radical by each post-column fraction was subsequently examined, and the prominent peptides in active fractions were sequenced by liquid chromatography-tandem mass spectrometry (LC-MS/MS).</p><!><p>Low-fat buckwheat flour was purchased from Bulkfoods.com (Toledo, OH, USA). The product specification sheet from the supplier indicated 3.6% fat, 71.4% total carbohydrate and 25% protein. Before protein extraction, the flour was stirred with hexane (1:1 w/v, four changes) for 48 h to remove residual fat. After vacuum evaporation of residual hexane, the dried defatted flour powder was subjected to the process of protein extraction according to the method of Tomotake, Shimaoka, Kayashita, Nakajoh, and Kato (2002) with some modifications. Defatted buckwheat flour (1 kg) was manually dispersed into 10 L of deionized water, and the pH was adjusted to 8.0 using 1 M NaOH. After stirring with a propeller (~50 rpm) at 4 °C for 2 h, the suspension was centrifuged at 5000 g for 20 min. The supernatant (protein extract) was decanted and adjusted to pH 4.5 using 1 M HCl to isoelectrically precipitate protein. The protein precipitate was washed with deionized water two times and then neutralized with 0.1 M NaOH before lyophilization. Freeze dried BWP powder was stored at −20 °C before use.</p><!><p>BWP in vitro digests were prepared according to the method of Lo and Li-Chan (2005). The suspension of BWP (5%, w/v) in nanopure deionized water was adjusted pH 2.0 with 1 M HCl, followed by the addition of pepsin (4%, w/w, protein basis). The mixture was incubated 1 h in a shaking water bath set at 37 °C to allow pepsin digestion. Subsequently, the pH was adjusted to 5.3 using 0.9 M NaHCO3. After the addition of pancreatin (4% w/w, protein basis), the pH was adjusted to 7.5 with 1 M NaOH. The digestion was restarted and continued in the 37 °C shaking water bath for another 2 h. Aliquots of hydrolysates were removed at 0, 30, 60, 90, 120, and 180 min during the pepsin → pancreatin sequential digestion, adjusted to neutrality (pH 7.0) with 1 M NaOH/HCl, and heated at 96 °C for 5 min to inactivate the enzymes. Each aliquot was freeze dried and kept at −20 °C before use.</p><!><p>Preliminary results showed that the two-stage in vitro digestion yielded a high radical scavenging activity in the final BWP digest (i.e., 180 min total digestion time). Therefore, this digest, referred to as "D180min", was subjected to peptide fractionation using a low-pressure size exclusion chromatography with a 2.6 cm (dia.) × 70 cm (length) Sephadex G-25 fine column (Pharmacia XK 26/70, Piscataway, NJ, USA).</p><p>A 2 mg/mL of D180min solution, prepared from lyophilized powder by dissolving in the elution buffer (0.02 M phosphate, pH 7.4), was clarified and sterilized through a 25-mm syringe filter with a 0.22 μM membrane (Fisher Scientific, Pittsburgh, PA). The purified solution (10 mL) was loaded to the Sephadex column and eluted in a 4 °C cold room with the elution buffer at a 0.9 mL/min flow rate. Peptide fractions were collected using an automated fraction collector, and the absorbance (215 nm) of the eluents was measured. In order to collect enough peptides for antioxidant assays, a total of 23 chromatographic runs were conducted. The corresponding peptide fractions from the 23 replicates were pooled and lyophilized. Freeze dried fractions were stored at −20 °C for further analysis.</p><p>Molecular weight (MW) distribution of the individual peptide fractions was estimated from a MW calibration curve generated from the elution volume of the following standards (Sigma Chemical Co., St. Louis, MO) that were chromtographed separately in the Sephadex G-25 column under the same condition as described above: cytochrome C (12327 Da), aprotinin (6512 Da), bacitracin (1423 Da), and tetrapeptide GGYR (452 Da). The evolution volume (mL) of blue dextran was used to establish the void volume of the column. Data were fitted in the exponential decay model (modified single with 3 parameters) of the SigmaPlot Ver. 9 software (Systat Software, Inc., Chicago, IL, USA), which yielded the following equation: LogMW=2.3429e(34.8528Vol−79.1592)</p><!><p>The RSA of BWP in vitro digests and the peptide fractions of the final digest (D180min) was evaluated using •OH and ABTS+• systems. The •OH assay involved the inhibition of radical formation rather than scavenging radicals that are already produced (i.e., pre-existed), while the ABTS+• scavenging assay was carried out by using pre-generated cationic radicals. Furthermore, ABTS+•, a synthetic radical species, is much larger in size than •OH, which is known to be most reactive of all the reactive oxygen species in food systems. Therefore, the analysis of RSA of BWP digests in the two different radical generation systems may lead to a better understanding of RSA of BWB peptides than would the individual assay systems.</p><!><p>The •OH scavenging activity measurement was carried out according to the method of Moore, Yin, and Yu (2006). Briefly, 30 μL samples were each mixed with 170 μL of 9.28 × 10−8 M fluorescein in a 96-well polystyrene plate (Fisher Scientific, Pittsburgh, PA, USA), followed by the addition of 40 μL of 0.1999 M H2O2 and 60 μL FeCl3. The mixed solution was immediately transferred to a Cary Eclipse fluorescence spectrophotometer equipped with a microplate reader (Varian, Victoria, Australia). The measurement with 0.1 s reading time per well and 1 min per plate was conducted with a 485 nm excitation wavelength and a 535 nm emission wavelength for 3 h to obtain the fluorescein decay curve. The •OH scavenging capacity was expressed as trolox equivalent (μM), which was determined from the regression equation built on a series of trolox standards (20, 40, 80, and 100 μM). The concentration of the standards was set as the x-axis and the net area under the decay curve was set as the y-axis. The calculation of the area under curve (AUC) is shown below, where f represents the fluorescence value at a particular time during the decay: AUC=0.5+f1/f0+f2/f0+f3/f0+…+fi−1/f0+0.5(fi/f0)</p><!><p>The ABTS+• scavenging ability was determined by the decolorization assay (Re, Pellegrini, Proteggente, Pannala, Yang, & Rice-Evans, 1999). Briefly, ABTS+• was generated by a mixed solution of 7 mM ABTS and 2.45 mM potassium persulphate. After 12–16 h reaction, a dense green-blue colored solution with excessive accumulation of ABTS+• radical was diluted with 0.2 M phosphate buffer (pH 7.4) to the absorbance level of 0.7 ± 0.02 at 734 nm. The RSA was then determined by mixing 10 μL samples (2 mg/mL protein) and 990 μL diluted ABTS+• solution for digestion aliquots (0, 30, 60, 90, 120 and 180 min) assay, and 100 μL samples (0.189 mg/mL protein) into 900 μL of diluted ABTS+• solution for post-column fractions and final digest mixture (D180min) for comparison. The absorbance of was recorded at 1, 2, 5 and 10 min during the reaction. The extent of decolorization represented the magnitude of scavenging ability and was calculated from a standard curve generated with 50, 100, 250, 500, and 1000 μM of trolox. Trolox equivalent antioxidant capacity (TEAC) was used to express RSA.</p><!><p>Gel filtration fractions (5 μL each) that exhibited strong radical scavenging activity were subjected to LC-MS/MS analysis using a QSTAR XL quadruple time-of-flight mass spectrometer (Applied Biosystems, Foster City, CA, USA) coupled with a nano-flow HPLC system (Eksigent Technologies, Dublin, CA, USA) through a nano-electrospray ionization source (Protana, Toronto, Canada) (Lu & Zhu, 2005). The samples were injected by an autosampler, desalted on a trap column (300 μm i.d. × 5 mm length, LC Packings, Sunnyvale, CA, USA), and subsequently separated by reverse phase C18 column (75 μm i.d. × 150 mm length, Vydac, Columbia, MD, USA) at a flow rate of 200 nL/min. The HPLC gradient was linear from 5% to 80% mobile phase B in 55 min using mobile phase A (H2O, 0.1% formic acid) and mobile phase B (80% acetonitrile, 0.1% formic acid).</p><p>Peptides eluted out of the reverse phase column were analyzed online by mass spectrometry (MS) and selected peptides were subjected to tandem mass spectrometry (MS/MS) sequencing. The automated data acquisition using information-dependent mode was performed on QSTAR XL under control by Analyst QS software (Applied Biosystems, Foster City, CA, USA). Each cycle typically consisted of one 1-sec MS survey scan from 150 to 1200 (m/z) and two 2-sec MS/MS scans of singly, doubly and triply charged species with mass range of 100 to 1200 (m/z). The spectra were interpreted using the de novo peptide sequencing module of the Analyst QS software.</p><!><p>The study employed a randomized complete block design with replication as the block. There were a minimum of three replications. Each analysis was done in duplicate. Data were subjected to analysis of variance using the general linear model's procedure of the Statistix software 9.0 (Analytical Software, Tallahassee, FL). When a treatment effect was found significant, Tukey HSD all-pairwise multiple comparisons were performed to identify significant differences between individual means.</p><!><p>In the •OH scavenging test, the potential of an antioxidant to inhibit •OH formation or to stabilize the radical was indicated by the rate of fluorescence decay of fluorescein (Moore et al., 2006). All BWP in vitro digest samples showed a slower fluorescence decay than non-hydrolyzed BWP (data not shown), indicating that hydrolysis improved •OH scavenging activity of BWP. Despite some variations, there was an overall trend that the •OH scavenging activity increased with digestion (Fig. 1). In the Fenton reaction (Fe2+ + H2O2 → Fe3+ + •OH + OH−), the impact of •OH on susceptible compounds could be restrained by either Fe2+ chelation or •OH stabilization or both. Hence, the strong •OH elimination activity of BWP digests, notably that of D180min, can be attributed to the removal of free Fe2+ prooxidant and stabilization of radicals through hydrogen or electron donation. The first mechanism was a postulation on the basis of the •OH assay, which can be supported by the strong iron chelation ability of 2 h pancreatin digests (i.e., D180min) (Ma and Xiong, 2009).</p><p>The ABTS+• radical scavenging activity assay also demonstrated a positive and more consistent rise in the scavenging capacity with digestion time, culminating at 2 h of pancreatin treatment (i.e., end of the total 180 min in vitro digestion) (Fig. 2). In the ABTS+• method, the antioxidant activity was measured exclusively by the ability of an antioxidant to act as a hydrogen or electron donor to neutralize preformed ABTS+• radicals (Re et al., 1999). However, in the •OH method, a test antioxidant is placed in a radical generation system where the antioxidant capacity was expressed both as inhibition of the radical initiation and elimination of formed radicals (Moore et al., 2006). The remarkable similarity in the results of •OH and ABTS+• assays on different digests suggested that inhibition of radical initiation was probably not a critical factor determining the efficacy of BWP digests as antioxidants.</p><!><p>Six peptide fractions were obtained by the Sephadex G-25 size exclusion gel filtration (Fig. 3). Based on the regression equation between elution volume and MW of each standard, the estimated mean MWs of these fractions were 3611 (I), 960 (II), 529 (III), 456 (IV), 365 (V), and 362 Da (IV). Assuming an average MW of 135 Da for amino acids, fraction I would be a peptide mixture consisting predominantly of those with 26 amino acid residues, and fractions II, III, IV and V or VI would have a preponderance of heptamerci, tetrameric, trimeric, or dimeric peptides, respectively. Fraction VI would also contain free amino acids. Based on the nitrogen analysis of the freeze dried eluent powders, these fractions accounted for 23.7% (I), 60.2% (II), 10.5% (III), 3.2% (IV), 1.4% (V), and 1.0% (VI) of the total eluent protein.</p><p>The incomplete separation of fractions I–IV suggested that they each contained peptides with various sizes, some of which were overlapping in neighboring fractions. A relatively broad distribution of molecular weight masses in hydrolyzed proteins is commonly associated with gel filtration (Adler-Nissen, 1986). The tailing parts of the eluents (fractions V and VI) would consist of mixed short peptides and free amino acids (Zhu et al., 2008). Thus, a more robust separation system (e.g., a high performance liquid chromatography) that enables a precise MW distribution measurement is desirable for better elucidating the relationship of peptide MW distribution with antioxidant activity.</p><!><p>The •OH and ABTS+• scavenging activities of various gel filtration fractions of the final BWP in vitro digest (D180min) are shown in Fig. 4 (•OH scavenging capacity vs. fractions) and Fig. 5 (ABTS+• scavenging capacity vs. fractions). Fraction VI exhibited the strongest •OH scavenging activity, which was 86% greater (P < 0.05) than the mixed BWP final digest (D180min). Fractions IV and VI showed the highest potential in scavenging ABTS+•, and the activity was enhanced by an average of 24.5% (P < 0.05) compared to mixed final digest.</p><p>Although the ABTS+• scavenging activity of fractions IV and VI indicated no significant differences, fraction VI showed a particularly strong •OH elimination activity and was 24% more effective than fraction IV. These results suggested that fractions IV and VI were primary contributors to the radical scavenging capacity of the final BWP in vitro digest. The weaker antioxidant activity of the final mixed digest compared to the individual fractions IV and V can be explained by the dilution effect of relatively ineffective larger peptides. Fractions I (3611 Da), II (960 Da) and III (529 Da), which were present in the mixed final digest (D180min), all displayed extremely low activity against ABTS+• and •OH. Yet, they made up the bulk of the total proteins/peptides in the mixed final digest.</p><!><p>Gel filtration fractions IV, V and VI, which exhibited relatively strong radical scavenging activity as indicated above, were subjected to individual peptide separation and sequence identification. The extracted ion chromatograms (XIC) and the tandem MS/MS spectra of the prominent peptides in these fractions are shown in Fig. 6. There were three prominent peptides in fraction IV, their XICs with the m/z values of 415.22, 401.2 and 548.27 are displayed in panels (A), (C) and (E), respectively. Their MS/MS spectra are shown in panels (B), (D) and (F). De novo sequencing determined the sequences of the peptides as Trp-Pro-Leu (B), Val-Pro-Trp (D) and Val-Phe-Pro-Trp (F), respectively. Fraction V produced relatively lower total signals in the LC-MS/MS analysis. A peptide with the m/z value of 302.14 was evident (panel G) and the MS/MS spectrum (panel H) supported the sequence of dipeptide Pro-Trp. Fraction VI showed a dominant XIC peak with elution time from 10 to 25 min (m/z at 205.12, panel I). The MS/MS spectrum displayed a characteristic pattern of fragments (m/z values of 118, 146 and 188) of the amino acid tryptophan as previously reported (Yamada, Miyazaki, Shibata, Hara, & Tsuchiya, 2008). Tryptophan was also detected in fractions V and IV with lower abundance (data not shown). The decreasing sizes of the prominent peptides in fractions IV, V and VI were consistent with the estimated molecular weights based on the gel filtration fractionation (Fig. 3).</p><p>These results were in concert with the general finding that short peptides with 2–10 amino acids exert greater antioxidant potential and other bioactive properties than their parent native proteins or large polypeptides (Kitts & Weiler, 2003). Through peptide bond cleavage, hydrolysis allows the release of active peptides capable of sequestering oxygen radicals, chelating prooxidant metal ions, and inhibiting lipid peroxidation in food systems (Elias, Kellerby, & Decker, 2008). For example, an oligopeptide with the sequence of His-Gly-Pro-Leu-Gly-Pro-Leu (797 Da), which was purified from fish skin hydrolysate, showed strong activity against ROS as well as linoleic acid peroxidation (Mendis, Rajapakse, & Kim, 2005). Two antioxidant short peptides, with the sequences of His-Val-Thr-Glu-Glu and Pro-Val-Pro-Ala-Glu-Gly-Val, were identified from chicken essence (Wu, Pan, Chang, & Shiau, 2005). Proline, as well as leucine, were found to play an important role in the antioxidant activity of peptides derived from soy protein, e.g., Leu-Leu-Pro-His-His (Chen, Muramoto, Yamauchi, & Nokihara, 1996). In our study, proline was present in all four prominent peptides identified in BWP digest fractions that showed strong antioxidant activity. In addition, valine was also present in two of four peptides. Zhu et al. (2008) reported that the peptide fractions rich in di-, tri-, and tetrapeptides from the zein in vitro digest (1–8 mg/mL protein) had comparable or stronger antioxidant activity than that of 0.1 mg/mL ascorbic acid or BHA. These peptides were superior to nonhydrolyzed zein.</p><p>The dominant existence of tryptophan in fraction VI suggested that tryptophan was a potent antioxidant in BWP digests, consistent with previous findings (Christen, Peterhans, & Stocker, 1990; Elias, McClements, & Decker, 2005). It is noteworthy that tryptophan was also present in all four peptides with significant abundances in the three strong radical scavenging fractions (IV, V, VI), further supporting its antioxidant property. Furthermore, because short peptides are smaller molecules than intact proteins, at an equal weight concentration basis, the peptides' molar concentration would be significantly greater than that of intact proteins. This difference would also contribute to the higher antioxidant activity of BWP peptides.</p><p>In conclusion, free radical scavenging activity of BWP was accentuated by in vitro digestion, especially after 2 h pancreatin digestion following the 1 h pepsin treatment. On an equal weight concentration basis, fractions enriched with di-, tri- and tetrameric peptides containing tryptophan and proline exhibited the strongest radical scavenging activity. These short peptides are implicated in the protection of the upper digestive tract of humans from oxidative stresses and may partially explain why dietary BWP promotes the health of the GI system.</p><!><p>Hydroxyl radical scavenging capacity of in vitro sequential digests of buckwheat protein. Means (n = 3) without a common letter differ significantly (P < 0.05). Sample solutions: 0.10 mg/mL protein.</p><p>ABTS+• scavenging capacity of in vitro sequential digests of buckwheat protein. Means (n = 3) without a common letter differ significantly (P < 0.05). Sample solutions: 0.159 mg/mL protein.</p><p>Sephadex G-25 gel filtration of the final in vitro digest (180 min total digestion time) of buckwheat protein (n = 23), and regression plot of log MW vs. elution volume (n = 3).</p><p>Hydroxyl radical scavenging capacity of gel filtration fractions of final in vitro digest (180 min total digestion time) of buckwheat protein. Means (n = 3) without a common letter differ significantly (P < 0.05). Sample solutions: 0.10 mg/mL protein.</p><p>ABTS+• scavenging capacity of gel filtration fractions of the final in vitro digest (180 min total digestion time) of buckwheat protein. Means (n = 3) without a common letter differ significantly (P < 0.05). Sample solutions: 0.159 mg/mL protein.</p><p>Extracted ion chromatograms (A, C, E, G, I) and tandem MS/MS spectra (B, D, F, H, J) of the prominent peptides present in gel filtration fractions IV (A–F), V (G; H), and VI (I; J). The peptides/amino acid were identified as Trp-Pro-Leu (B), Val-Pro-Trp (D), Val-Phe-Pro-Trp (F), Pro-Trp (H) and Trp (J).</p>

PubMed Author Manuscript

PvdF of pyoverdin biosynthesis is a structurally unique N10-formyltetrahydrofolate-dependent formyltransferase

The hydroxyornithine transformylase from Pseudomonas aeruginosa is known by the gene name pvdF, and has been hypothesized to use N10-formyltetrahydrofolate (N10-fTHF) as a co-substrate formyl donor to convert N5-hydroxyornithine (OHOrn) to N5-formyl- N5-hydroxyornithine (fOHOrn). PvdF is in the biosynthetic pathway for pyoverdin biosynthesis, a siderophore generated under iron-limiting conditions that has been linked to virulence, quorum sensing and biofilm formation. The structure of PvdF was determined by X-ray crystallography to 2.3 \xc3\x85, revealing a formyltransferase fold consistent with N10-formyltetrahydrofolate dependent enzymes, such as the glycinamide ribonucleotide transformylases, N-sugar transformylases and methionyl-tRNA transformylases. Whereas the core structure, including the catalytic triad, is conserved, PvdF has three insertions of 18 or more amino acids, which we hypothesize are key to binding the OHOrn substrate. Steady state kinetics revealed a non-hyperbolic rate curve, promoting the hypothesis that PvdF uses a random-sequential mechanism, and favors folate binding over OHOrn.

pvdf_of_pyoverdin_biosynthesis_is_a_structurally_unique_n10-formyltetrahydrofolate-dependent_formylt

INTRODUCTION<!>Preparation of PvdF Overexpression Plasmid.<!>Preparation of K72A,K74A-PvdF expression plasmid.<!>Wildtype PvdF expression and purification.<!>Expression and purification of K72A,K74A-PvdF.<!>Selenomethionine substituted PvdF expression and purification.<!>PvdA protein expression and purification.<!>Preparation of hydroxyornithine (OHOrn).<!>Preparation of 10-formyl-5,8 dideazafolate (fDDF) and 5,8 dideazafolate (DDF).<!>Steady state activity assays.<!>Coupled steady state activity assay.<!>Wildtype and K72A,K74A-PvdF progress curves.<!>Mass Spectrometry.<!>PvdF Crystallization.<!>Single crystal X-ray diffraction data collection and processing.<!>PvdF crystallographic model.<!>Preparation of PvdF.<!>Monomer architecture.<!>Structural homologues.<!>Folate binding pocket.<!>OHOrn binding.<!>Citrate.<!>Catalytic Triad.<!>PvdF steady state kinetics.<!>Product detection.<!>The observed DDF binding mode is a crystallization artifact.<!>DISCUSSION

<p>Iron is required for major metabolic processes such as cellular respiration and nucleotide biosynthesis. Due to insolubility and toxicity, iron is sequestered and highly regulated in human cells and is thus unavailable to bacterial pathogens, a phenomenon that has been called nutritional immunity1. Pathogens have developed elaborate mechanisms to overcome the paucity of available iron in the human host, including producing high affinity chelators called siderophores. Once secreted, siderophores bind iron, and are taken up in the iron-loaded form to provide the pathogen with the required iron2, 3.</p><p>The focus of this study is the second step in the biosynthesis of the siderophore pyoverdin, which is linked to virulence, quorum sensing, and biofilm development2 in the ESKAPE pathogen4, 5 Pseudomonas aeruginosa. Pyoverdins, whose structure and composition is dependent on bacterial strain, are composed of a dihydroxyquinoline chromophore core with an α-ketoacid sidechain attached to a 6–14 amino acid peptide that is assembled by nonribosomal peptide synthethase (NRPS) enzymes (Figure 1)3. Along with the NRPS enzymes, there are also enzymes in pyoverdin biosynthesis that are required for the production of precursors, maturation and tailoring of the peptide, and chromophore synthesis2, 3, 6, 7 . All pyoverdins from P. aeruginosa strains include N5-formyl-N5-hydroxyornithine, a nonproteinogenic amino acid derived from ornithine that has been hydroxylated and formylated on the sidechain amine resulting in a hydroxamate moiety of the siderophore. The biosynthetic operons for production of pyoverdin include proteins for conversion of L-ornithine (Orn) to N5-hydroxyornithine (OHOrn) by the ornithine hydroxylase PvdA, and for subsequent formation of N5-formyl-N5-hydroxyornithine (fOHOrn) by the hydroxyornithine transformylase PvdF (Figure 1)2. PvdA has been structurally and biochemically characterized8–10, but little is known about PvdF. Deletion strains of PvdF do not produce pyoverdin and are avirulent, and cell extracts from those strains showed formation of ornithine hydroxylamine without conversion to the hydroxamate form11.</p><p>N5-formyl-N5-hydroxyornithine is a component of other siderophores, including rhodochelin (Rhodococcuss jostii RHA1)12, coelichelin (Streptomyces coelicolor)13 and amychelin (Amycolatopsis sp.AA4)14. These chelators are similarly constructed by NRPS assembly lines that are dependent on accessory enzymes to generate fOHOrn. The enzyme characterizations for these pathways are at the initial stages, with activities confirmed and some steady state analyses performed12–14. Rhodochelin formyltransferase (Rft) has been definitively shown to perform a N10-formyltetrahydrofolate (N10-fTHF) dependent formylation reaction to convert OHOrn to fOHOrn12.</p><p>Here we report structural characterization of a hydroxyornithine transformylase, the PvdF enzyme from Pseudomonas aeruginosa. The structure reveals a core fold common among N10-fTHF dependent transformylases, including the glycinamide ribonucleotide transformylases (GART)15–19, the methionyl-tRNA transformylase (MTF)20, 21, and N-sugar transformylases of O-antigen formation22–26. However, the structure reveals large, unique insertions that we propose are important for binding the substrate OHOrn, and that place PvdF as the first documented member of a new structural subclass. This work includes a steady state kinetic analysis that indicates a partially ordered, formally random-sequential bireactant system that favors folate binding.</p><!><p>The pvdF gene was cloned from Pseudomonas aeruginosa (PAO1) genomic DNA, using polymerase chain reaction (PCR) with Herculase polymerase (Stratagene). The reaction was supplemented with 8% (v/v) DMSO as per manufacturer instructions due to high G-C content (61%). The forward primer (5'-AAT TAT ATA CAT ATG ACG AAA AGG AAA CTG GCC TA −3') contains an NdeI restriction site (underlined), and the reverse primer (5'-AAT ATA ATA CAG ATC TGG GAG CTT CTC GGC GAG CAG C-3') contains an BglII restriction site (underlined). The amplified DNA fragment was ligated into the correspondingly digested pET29b vector (Novagen) with T4 DNA ligase (New England BioLabs). The overexpression plasmid generates the PvdF protein with a C-terminal thrombin cleavage site followed by a histidine tag. This construct was further modified by site-directed mutagenesis to incorporate two stop codons at the C-terminus of the PvdF sequence so that the native PvdF protein, without purification tags, could be expressed. The Quik-Change® site-directed mutagenesis kit (Stratagene) was used with the forward primer (5'-CTG CTG GCC GAG AAG CTC TGA TGA CTG GGT ACC CTG GTG-3') and reverse complement primer (stop codons underlined).</p><!><p>The PvdF K72A,K74A expression plasmid was prepared by the Genscript plasmid preparation and mutagenesis services. The pvdF gene was cloned into the pET29b expression plasmid at the HindIII restriction site on the 3' end and the NdeI restriction site on the 5' end following the stop codon. The gene was synthesized such that the codon for K72 (AAA) was changed to encode alanine (GCA) and the codon for K74 (AAG) to encode alanine (GCG).</p><!><p>The PvdF plasmid was transformed into BL21(DE3) E.coli (New England BioLabs) for expression. Baffled flasks containing 1 L of LB Miller media containing 50 μg/ml of kanamycin were inoculated with 10 mL of overnight culture and grown at 37 °C in a shaker incubator (225rpm). When the OD600 reached 0.5, the temperature was lowered to 25 °C and allowed to equilibrate for 15 min. Expression was induced with isopropyl-β-D-1-thiogalactopyranoside (IPTG) to a final concentration of 0.2 mM with shaking incubation for 16 hours. The cells were harvested by centrifugation (6000 × g, 5 min, 4 °C). The cell pellets were resuspended in 20 mL of 50 mM Tris-HCl pH 8.5 and lysed by three passes through a French press apparatus (35,000 psi). The lysate was centrifuged (12000 × g, 30 min, 4 °C) and the supernatant was injected onto a Source 30Q affinity column (GE Healthcare) pre-equilibrated with 50 mM Tris-HCl pH 8.5. The protein was eluted with a linear gradient of increasing NaCl to 500 mM. Protein fractions containing PvdF were confirmed by 15% SDS-PAGE and pooled. The salt concentration was adjusted to 1 M final concentration by slow addition of solid NaCl with gentle mixing. The protein was injected onto a Source Phenyl Sepharose (GE Healthcare) column pre-equilibrated in 50 mM Tris-HCl pH 8.5, 1 M NaCl. The PvdF protein was eluted from the column using a gradient to a buffer with no NaCl (50 mM Tris-HCl pH 8.5). Fractions containing PvdF were concentrated and injected onto a Superdex 200 gel filtration column (GE Healthcare) pre-equilibrated in 50 mM potassium phosphate pH 7.4. PvdF eluted at a molecular weight consistent with monomeric protein. The protein was concentrated with an Amicon® Ultracell® 30K centrifugal filter to 70 mg/mL as determined by Bradford assay, and stored at −80 °C. The purification protocol yielded 148 mg per liter of culture.</p><!><p>The K72A,K74A-PvdF expression plasmid was transformed into BL21(DE3) E.coli (New England BioLabs). The variant protein was expressed and purified in a similar manner to wildtype PvdF, except that the phenyl sepharose column was not required to attain high purity. Therefore, the protein eluted from the Source 30Q affinity column was directly injected onto the Superdex 75 gel filtration column. This preparation yielded 55 mg of protein per liter of cell culture.</p><!><p>Se-Met PvdF was produced according to the protocol by Van Duyne et al.27 with some modifications. M9 minimal media was augmented with 2 mM MgCl2, 0.1 mM CaCl2, 0.4% (w/v) glucose and 50 μg/mL kanamycin. Growth cultures (1L) were inoculated with 10 mL of overnight culture and incubated at 37 °C in a shaker incubator (225 rpm) until an OD600 of 0.5 was reached. The temperature was lowered to 25 °C and an amino acid mixture was added to inhibit methionine production and allow for selenomethionine incorporation (the amino acid mixture included: 100 mg each of lysine, phenylalanine, threonine; 50 mg each of isoleucine, leucine, valine; 60 mg of selenomethionine, per liter of culture). When the OD600 of the culture reached 1.0, IPTG was added to a final concentration of 0.2 mM and the culture was incubated for a further 16 hours with shaking. The SeMet protein purification was performed as for the native protein, with the exception that all buffers were supplemented with 2 mM dithiothreitol (DTT). The purified protein was concentrated with an Amicon® Ultracell® 30K centrifugal filter to 80 mg/mL as determined by Bradford assay and stored in −80 °C. The purification protocol yielded 100 mg per liter of culture.</p><!><p>The PvdA enzyme was expressed and purified as previously reported8, 10.</p><!><p>N5-hydroxyornithine was prepared by Garrett Moraski (Montana State University) according to the published protocol.28</p><!><p>Both fDDF and DDF were a generous gift from Dr. Carol Caperelli (University of Cincinnati). Both substrates were prepared according to the published protocols.29</p><!><p>The assay buffer contained 50 mM potassium phosphate buffer pH 7.4. The deformylation of fDDF is followed by the change in extinction coefficient at 295 nm (Δε=18.9 mM−1cm−1)30. The assay was performed using a TgK scientific stopped-flow instrument at 25 °C equipped with a mercury-xenon lamp. Enzyme (200 nM) with 11 mM N5-hydroxyornithine was mixed at 1:1 ratio with varied fDDF concentration (5 μM to 642 μM). The rate was measured for 30 sec and the rate dependence was fit to the Michaelis-Menten equation. When hydoxyornithine was varied, the reaction was performed in 96 well flat–bottom plate (Corning cat # 9107) using a Varian 50MPR Microplate Reader, with a total reaction mixture per well of 300 μL. Each well contained a final enzyme concentration of 100 nM. The hydroxyornithine concentration was varied from 5.8 mM to 100 mM. The reaction was initiated with addition of 150 μM fDDF; the highest concentration possible at the fDDF λmax, within the linear range of the instrument. Reaction progress was monitored at 295 nm in 1 sec cycle reads for 90 seconds. Data were fit to the Michaelis-Menten equation.</p><!><p>Pseudomonas aeruginosa ornithine hydroxylase (PvdA) was used to generate the substrate OHOrn. The standard assay buffer contained 50 mM potassium phosphate buffer pH 7.4. The reaction was performed in a 1.5 mL quartz cuvette using a Cary 50 Bio UV-visible spectrophotometer. The initial reaction (600 μL) contained 1 μM PvdA, 150 μM FAD, 10 mM ornithine. Varying concentrations of NADPH were added (133 μM to 1 mM) for PvdA to generate defined OHOrn concentrations. The progress of the reaction was monitored at 300 nm as a measure of NADPH turnover. When the change in absorbance at 300 nm ceased, 1 mM fDDF (final concentration) was added to the cuvette and the spectrometer was blanked. The transformylase reaction was initiated by the addition of 200 nM PvdF and monitored at 295 nm for 30 sec using a 0.5 cm pathlength. The initial rates were plotted versus [OHOrn], assuming that each NADPH consumed by PvdA produced one OHOrn molecule. The plot showed a non-hyperbolic velocity curve, with a decreasing rate at concentrations of OHOrn above 400 μM. Points lower than 133 μM were not obtained due to insufficient signal-to-noise.</p><!><p>Progress curves were measured using the same buffer as the steady state assays, generating OHOrn with 1 μM PvdA, 150 μM FAD, 10 mM ornithine, 500 μM NADPH. After the reaction ceased to change at 300 nm, the PvdF reaction was initiated by the addition of fDDF (46–183 μM) and 200 nM PvdF. The reaction was monitored at 295 nm for 60 min using 0.5 cm pathlength quartz cuvette.</p><!><p>Samples from the PvdA-PvdF reaction were diluted 1000-fold with LC-MS grade water (Sigma-Aldrich), and 10 μL of each dilution was analyzed by LC-MS over 65 minutes on an LCMS-IT-TOF (Shimadzu Scientific Instruments) with a Shim-pack XR-ODS column. The mobile phase consisted of 95% of an aqueous 0.1% formic acid solution and 5% acetonitrile (Sigma Aldrich), with a total flow rate of 0.2 mL/min. An ESI source was used, and acquisition was performed in scan mode from 120–550 m/z for both positive and negative ion modes. A 10 msec ion accumulation time was used, and event time was set to 100 msec. A three stage gradient was run as follows: 5% acetonitrile for 5 minutes, a linear gradient from 5% to 95% acetonitrile over 20 minutes, and 95% acetonitrile for another 20 minutes.</p><!><p>Purified SeMet protein was exchanged into 50 mM potassium phosphate buffer pH 7.4, 2 mM DTT and diluted to 40 mg/mL. A few flakes of powdered DDF were added to the protein solution, and the mixture was incubated on ice for 15 min. The protein solution was centrifuged (12000 × g, 30 sec, 4 °C). The protein was crystallized using the hanging-drop vapor diffusion method. Crystallization drops were prepared by mixing 1.5 μL protein solution with 1.5 μL precipitant solution containing 0.55 M sodium citrate, 0.1 M Tris-HCl pH 8.5. Rectangular-prism shaped crystals with dimensions 0.15 μm × 0.15 μm × 0.04 μm grew within 2 weeks. For data collection, crystals were soaked in a cryoprotectant solution containing the precipitant solution augmented with 20% ethylene glycol and flash cooled in liquid nitrogen.</p><!><p>A single wavelength anomalous dispersion (SAD) dataset was collected at the Stanford Synchrotron Radiation Laboratory (SSRL, Stanford, CA) beamline 12–2 using a wavelength of 0.9795 at 100 K. This wavelength was based on a selenium fluorescence scan which showed a strong signal, with an inflection point at 0.9795 Å. The software package Blu-Ice31 was used to collect 847 oscillation images (0.15 º per image). The exposure time per frame was 0.2 sec with a transmission of 3%, and the crystal-to-detector distance set at 400 mm. Diffraction data were processed using XDS32 to 2.3 Å with anomalous signal to 2.73 Å in the space group P21 with cell dimensions of a=128 Å, b=92.7 Å, c=128 Å, β=90.1 º. While the crystals appeared single, the diffraction pattern showed twinning (overlapping lattices) making space group determination problematic. Data were frequently auto-processed as P422, but had to be manually re-processed in P21 in order to obtain the solution. Despite a strong anomalous signal, the SAD data did not lead to a solution using PHENIX. Crank233 in the CCP4 program suite34 was used to determine the location of 24 Se atoms, providing initial phases to build eight monomers in the asymmetric unit. This solution had a figure of merit (FOM) of 0.782 and Rcomb of 0.351. XTRIAGE35 identified the twin fractions (-l, k, h; -h, -k, l; l, -k, h) with (-h, -k, l) showing the highest twin fraction of 0.49. This twin fraction was applied in subsequent rounds of model building and refinement using Coot36 and Phenix.Refine37. Water molecules were added automatically and inspected manually using Coot. Citrate molecules, derived from the crystallization conditions, were modeled manually using Coot. DDF molecules were built using LigandFit38, 39 with restraints generated using eLBOW40 and REEL41. Statistics for data refinement and analysis can be found in Table 1.</p><!><p>The final PvdF model contains eight monomers; however, the model is discontinuous with several chain breaks per monomer due to disorder. The amino acid summary can be found in Table 2. The model contains 336 waters, seven DDF molecules and eight citrate molecules. The DDF molecules are located at an interface between monomers. Four DDF molecules are present in 100% occupancy, whereas the remaining three were refined to 66–73% occupancy. Final Ramachandran analysis has been calculated with MolProbity42 with 96.5% in the favored regions and one outlier in an area of poor density. Root mean square deviation values were calculated using PDBeFold43 and protein interaction interfaces were calculated using PDBePISA44. Structures figures were prepared using Pymol45. Atomic coordinates and structure factors for SeMet PvdF were deposited into the Protein Data Bank, with the accession code 6CUL.</p><!><p>PvdF protein was heterologously produced in E. coli and purification was completed in three steps, using anion exchange, phenyl sepharose and gel filtration chromatography. The 31 kDa protein eluted from the gel filtration column at a molecular weight consistent with monomeric protein in solution (Supplemental Figure S1). The SeMet protein was purified using a similar protocol, with the addition of a reducing agent to all buffers. The SeMet protein was crystallized using citrate as the precipitant, and crystals only formed in the presence of the product analogue 5,8-dideazafolate (DDF). As defined in the Materials and Methods section, the structure determination was complicated by twinning. Despite the technical difficulties, initial phase estimates for PvdF were determined by single wavelength anomalous dispersion phasing using the selenomethionine-substituted form of the protein to 2.34 Å (Table 1). A representative electron density map of the refined structure is depicted in Figure 2A, with example density for the DDF in Figure 2B. The asymmetric unit contains eight monomers, arranged in two rings with 4-fold rotational symmetry (Figure 2C). Consistent with the gel filtration data for the protein in solution, the average interface area between monomers calculated by PDBePISA44 was 719 Å2, indicating that the four-fold symmetry is the result of the arrangement of the monomers in the crystal lattice and not indicative of an oligomeric state. Unexpectedly contacts within this monomer-monomer interface are mediated by the bound product analogue, 5,8 dideazafolate (DDF) which is not observed to bind in the putative active site, defined by the conserved catalytic triad (Figure 2D).</p><!><p>The core of the PvdF monomer shows the standard formyltransferase fold found in the N10-formyltetrahydrofolate dependent enzymes, with a central 7-stranded sheet surrounded by helices and loops (Figure 3A). The fold has been previously divided into two subdomains: an N-terminal subdomain for binding the folate substrate and a C-terminal subdomain for binding the substrate to be formylated18, 19. Glycinamide ribonucleotide transformylase (GART) is one of the best studied formyltransferases15, 17, and E. coli GART (EcGART) will serve as a frame of reference for our discussion (Figure 3B). The core of PvdF and EcGART are structurally conserved; however, PvdF has three major structural insertions. In the N-terminal subdomain, EcGART has a short loop between strand 1 and helix A. In PvdF, strand 1 is followed by a short helix, which is labeled helix a, and then an antiparallel β-sheet (strands i and ii), before helix A (grey in Figure 3). This is a total insertion of 23 amino acids (residues 12 to 34 in PvdF). In EcGART, the connection between the second β-strand and helix B is also a short loop. Helix B is followed by another short loop that connects to strand 3. PvdF has a large insertion at this connection (40 amino acids), beginning at residue 63. Helix B of the Rossman fold is replaced with helix b, which is roughly parallel to helix a (not packed against the central sheet). Helix b is followed by a long loop that does pack against the central sheet, structurally replacing helix B of EcGART. PvdF does have a short turn of a helix B before it rejoins the standard fold at residue 102 in strand 3. These changes are highlighted in yellow in Figure 3. Finally, in the C-terminal subdomain, PvdF has an insertion between helix E and strand 6, shown in orange in Figure 3. This insertion, residues 197 to 214 (18 amino acids), forms an antiparallel β-sheet with strands labeled iii and iv. EcGART continues after the F helix, with 2 additional strands forming a small sheet, structural elements not within the formyltransferase fold and not found in PvdF.</p><!><p>The closest structural homologues to PvdF are the GART proteins, which are found in the pathway for the de novo biosynthesis of purines. The root mean square deviation (rmsd) calculated for the comparison of PvdF to EcGART is 2.2 Å for 165 Cα residues (Figure 4A and 4B). PvdF is 275 amino acids in length, whereas EcGART is 209 amino acids. The three insertions listed above account for most of the differences in Cα comparison, with other more subtle changes found within other loops. Unlike PvdF, GART proteins are dimeric, though the interfaces for dimerization within the GART family are not conserved. The methionyl-tRNA formyltransferase (MTF), which formylates the primary amine of the methionine attached to the initiator methionyl-tRNA, is monomeric and also shares a similar fold with PvdF and EcGART. The enzyme from Yersinia pestis is shown in Figure 4C, which when compared to PvdF has an rmsd of 2.3 Å over 150 Cα residues. The sugar N-transformylases involved in production of modified sugars for incorporation into O-antigens also fall into this structural and functional class. Of these, VioF from P. alcalifaciens O30 showed the closest structural similarity to PvdF with an rmsd of 2.5 Å over 145 Cα residues and is dimeric using a structural feature not found in PvdF (Figure 4D). Many of the sugar N-transformylases include C-terminal domains with other catalytic activities or regulatory roles22–26. Recently, the structure of the gramicidin initiation module (LgrA) was determined, which includes a formyltransferase domain (Figure 4E). Interestingly, this formyltransferase domain has been proposed to be incorporated into the NRPS assembly line as the result of a gene duplication and horizontal transfer of a sugar N-transformylase46, potentially an evolutionary precursor of VioF. At the C-terminal end of the LgrA transformylase domain is a new structural element, a loop that includes an α-helix, that serves as a linker to the adenylation domain of the NRPS module46. Note that the secondary structure insertions of PvdF are unique among these enzymes (Figure 4), placing PvdF in a new structural subclass of N10-fTHF dependent transformylase enzymes.</p><!><p>Enzymes dependent on N10-formyltetrahydrofolate are sometimes identified by a folate binding motif, HxSLLPxxxG where x is any residue, in the C-terminal subdomain (despite the N-terminal domain being labeled the folate-binding domain) (Figure 5)22. The histidine in this sequence is one of three residues in the catalytic triad, discussed later. This sequence starts in strand 5 and continues through the loop that connects to helix E, forming a portion of the folate binding pocket. For many of the N-sugar transformylases and for PvdF, this sequence is not conserved. The initial histidine and final glycine residues are conserved in three dimensions in PvdF. However, the remaining residues are not conserved, and the loop is 5 amino acids longer. The resulting sequence is: 170-HxGVTRyyyyyxxxG-184, where y is any residue in the extended PvdF loop. The primary contacts with the folate as determined in the GART and N-sugar transformylases are through hydrophobic interactions and hydrogen bonds between the methylpterin rings and the protein backbone found in the loop connecting strands 4 and 5, and in the loop connecting strands 6 and 7. The loop connecting strands 4 and 5 in EcGART and most of the other transformylases includes a helix (labeled α1 in Figure 3B). In PvdF, this loop is similarly coiled, but does not make the hydrogen bonds requisite to define this as a helix. While the residues that interact with the methylpterin of the folates are not conserved between the GART and N-sugar transformylases, the shape, hydrophobicity of the pocket, and hydrogen bonding interactions are. In PvdF, the loop connecting strands 4 and 5 maintains the proper shape for interaction with the methylpterin rings; however, the loop connecting strands 6 and 7 is disordered, potentially because no folate is bound in the active site. This loop has been noted in other transformylases to be mobile18, 47, 48. We hypothesize that for catalysis, the folate binds in the analogous location in PvdF when compared to both the GART and N-sugar transformylases. PvdF crystals would not grow without inclusion of a folate analogue. However, the folates evident in the electron density map are found forming crystal contacts, as was described previously, rather than in the active site.</p><!><p>The substrates differ widely between N-transformylase groups, and include nucleotide precursors, sugar-nucleotides, tRNA, amino acids, and amino acids attached to a NRPS carrier domain through a phosphopantetheinyl linker, as documented in Figure 4. It is not surprising that the individual substrate binding interactions are specific, with the ultimate goal of presenting the amine group undergoing formylation within range of the formyl group of the N10-formyltetrahydrofolate. While there is no substrate bound in the PvdF structure, comparison to the holo-structures of transformylases previously determined suggests that the OHOrn substrate will bind in an analogous location (Figure 5A). The loop between strand 1 and helix A forms important hydrogen bonding and ionic interactions with the phosphates of the substrates in the GART and N-sugar transformylases (grey in Figure 5B)15, 18, 23, 24. In PvdF, the loop that contains these amino acids is not present, and this is instead the location of the first major insertion. Indeed, helix a, which is in the same three-dimensional space, has two sidechains, from Asn14 and Asp18, that point into the cavity and may form interactions with the backbone of the OHOrn substrate. The PvdF helix a is placed more interior in the active site than the GART or N-sugar transformylase 1-A loop, and the Asn14 and Asp18 sidechains would place the substrate deeper in the active site, potentially accounting for the considerably shorter OHOrn substrate.</p><p>The loop in the tRNA transformylase connecting β-strand 2 and helix B has been considered important for binding of the tRNA substrate20, 21. This loop corresponds with the second major insertion in PvdF; however helix b is in the comparable three dimensional location (yellow in Figure 5). Helix b is unlikely to play a direct role in hydroxyornithine binding in PvdF, being too distant from the putative substrate binding site. Finally, helix F has a proline forming a kink in these homologues, not necessarily at the same turn in the helix, but still suitable to promote a conformation in which the N-terminus of the helix is bent toward the substrate binding cavity. In the GART and N-sugar transformylases, charged and polar residues of helix F form hydrogen bonds with the substrates. The analogous residue in PvdF is Arg252, which may serve a similar role. In all, the location for substrate binding is likely analogous to that seen in the GART and sugar transformylases, but the residues that promote binding may be contributed at least in part by the structural features that are unique to PvdF.</p><!><p>The model of PvdF has a well-ordered citrate bound by three arginine residues (13, 68, and 111) from helix a, helix b, and helix C in every monomer (Figure 5A). The citrate molecule is derived from the mother liquor, which required greater than 0.5 M citrate for protein crystal formation. The binding of citrate in this site is undoubtedly a crystallization artifact. Nevertheless, the citrate is in close proximity of N10-fTHF binding site, and is bound with a free carboxylate less than 3 Å from the putative location for the folate glutamate tail (Figure 5B). This suggests that the citrate from the crystallization conditions, in large molar excess, prevented binding of folate in the active site. If this is correct, Arg 111 or potentially Arg 115 (nearby but not bound to citrate) may be involved in binding the glutamate tail of N10-fTHF; however, a new crystal form with folate bound in the active site would be necessary to establish this. It is important to note that there are no comparable binding interactions for binding the glutamate tail within the previously determined N10-fTHF dependent transformylase domains: the tail is frequently found to be disordered or having high B-factors in structures where folate is bound15, 19, 23, 24.</p><!><p>The loop connecting strands 6 and 7 (black in Figure 4 and Figure 5) has been named both the active site loop18 and the folate binding loop19. As noted before, this loop has been documented in other transformylases to be mobile, so it is not surprising that this loop is disordered in PvdF in the absence of folate. This loop harbors an aspartic acid that is one of three residues in a proposed catalytic triad. The other two residues, a histidine and an asparagine, are located in strand 5. These three residues are conserved in PvdF: Asn168, His170, and Asp229 (purple in Figure 5). In the GART and sugar transformylases, the binding pose of the folate is such that the formyl group on N10-fTHF is positioned at the center of the triad17, 22, 23. The proposed mechanism for the GART enzymes, and by extension all enzymes of this class, suggests that the amino group of the substrate performs a nucleophilic attack on the carbonyl of the formyl group of N10-fTHF, generating a tetrahedral intermediate. The catalytic triad residues are proposed to serve as general acid-general base residues to promote intermediate formation and resolution of the catalytic cycle17, 22.</p><!><p>N10-fTHF dependent transformylases, such as those from purine biosynthesis, have been successfully assayed using the analogue 10-formyl-5,8-dideazafolate (fDDF)30, 49. When the formyl group is removed from fDDF, there is an increase in absorbance at 295 nm, allowing for a convenient continuous spectroscopic assay. When OHOrn is held in excess, and the varied substrate is fDDF, kinetic parameters are readily determined: Km = 60 ± 10 μM, kcat = 1.7 ± 0.1 sec−1 (Figure 6, red curve). However, the converse reaction, with fDDF in excess and OHOrn as the varied substrate, yielded a physiologically improbable kinetic constants (Supplemental Figure S2). This is likely due to difficulties with the assay: the experiment was not repeatable with each subsequent experiment showing an increase in Km and a decrease in kcat. The Km effect can be rationalized as OHOrn is known to be unstable50–52, and so the effective substrate concentration was diminishing with time. We hypothesized that a solution to this problem was to have the preceding enzyme of the biosynthetic pathway, the ornithine hydroxylase PvdA, generate the necessary substrate in situ. PvdA is a flavin-dependent enzyme that must be reduced by NADH with each catalytic cycle8–10. The appropriate concentration of OHOrn was produced from PvdA by varying and limiting the concentration of NADH. The steady state plot produced in this manner, showed a non-hyperbolic velocity curve (blue in Figure 6). The curve does not fit well to a Michealis-Menten or a substrate inhibition model. Instead, the data suggest a random-sequential bireactant mechanism in which the pathway for fDDF binding first is preferred for product generation, a model that as has been previously described for other enzymes53, 54.</p><!><p>The absorbance assay described above indicates the loss of the formyl group from fDDF, but not necessarily for formation of the product fOHOrn. To confirm fOHOrn production, the PvdA-PvdF reaction was analyzed by LCMS, monitoring for fOHOrn (m/z=177.1), fDDF (m/z=466.2) and DDF (m/z=438.1) (Figure 7). fOHOrn and DDF were observed, and the fDDF decreased in reactions containing both enzymes and all necessary substrates (flavin, NADH, Orn, fDDF), whereas controls that did not contain one of the enzymes (PvdA or PvdF) did not show production of the fOHOrn or DDF products.</p><!><p>We hypothesized that the observed binding site for DDF, outside the active site and 22 Å distant from the catalytic triad, is a crystallization artifact (Figure 2D). As mentioned previously, the crystals only grew in the presence of DDF, so potentially this binding promoted the formation of an oligomerization interface that promoted crystallization. Despite >10 years of effort, these twinned crystals were the best to date and the only ones that produced a refined structure. However, we now have the benefit of a refined structure in which we can analyze crystallization contacts. We generated a variant, K72A,K74A-PvdF. These two lysine residues flank the DDF binding site. In monomers C and D, K72 directly hydrogen bond with the glutamate tail of DDF. K74 of one monomer is proximity of E65 of the next monomer in the ring, and in two of the eight cases, these residues form a hydrogen bond. The double K→A variant did not crystallize, and the protein was active as shown in full progress curves (Figure 8). Therefore, the catalytically relevant binding of the folate is not seen in this PvdF structure. Instead, we hypothesize that in the catalytic complex, the folate will bind such that the formate to be transferred (attached to N10) will be adjacent to the catalytic triad, as seen in all other homologues of this family.</p><!><p>PvdF is the formyltransferase that converts N5-hydroxyornithine (OHOrn) to N5-formyl-N5-hydroxyornithine (fOHOrn) so that fOHOrn can be incorporated into the siderophore pyoverdin by a nonribosomal peptide synthetase assembly line (Figure 1). An N10-fTHF-dependent hydroxyornithine transformylase involved in siderophore biosynthesis has been functionally characterized previously12. The enzyme, rhodochelin formyltransferase, or Rft, is involved in the biosynthesis of the mixed catecholate-hydroxamate siderophore rhodochelin by Rhodococcus jostii RHA1, a gram positive soil bacteria. Like pyoverdin, this siderophore includes two formylhydroxyornithine residues for iron chelation, and is assembled by a nonribosomal peptide synthetase. Using an HPLC-MS assay, the authors showed conversion of OHOrn to fOHOrn, but there are no structural data for Rft. Sequence comparisons indicate that Rft is not a close structural homologue of PvdF. Instead, Rft is likely to be structurally similar to either the tRNA transformylase (FMT) or the N-sugar transformylase ArnA, an E. coli enzyme involved in lipid A modification that transformylates UDP-4-amino-4-deoxy-L-arabinose12. Rft, FMT, and ArnA all have the conserved HxSLLPxxxG motif for binding the folate co-substrate that PvdF lacks, and they all lack the major insertions highlighted in Figure 3 that are specific to PvdF. Also unlike PvdF, they all have a C-terminal domain that provides additional functionality (to enhance substrate binding or to provide an additional catalytic activity). Finally, Rft is an allosteric enzyme showing positive cooperativity, and is proposed to be a tetramer in solution12. PvdF is a monomer in solution and in the crystals. The 4-fold ring structure seen in Figure 2B is the result of crystal packing. Therefore, Rft is more functionally and likely structurally similar to the N-sugar transformylases and the tRNA transformylases than to PvdF.</p><p>The initiation module of the NRPS for the biosynthesis of the antibiotic gramicidin (LgrA) includes a formyltransferase that has been structurally characterized. The LgrA N10-fTHF-dependent formyltransferase domain transfers a formate to the backbone amine of a valine while the amino acid is covalently attached to the peptidyl carrier domain by a phosphopantethienyl tether46. This is in contrast to PvdF, which is a stand-alone accessory enzyme that formylates the sidechain amine of the free OHOrn substrate before the product fOHOrn is activated by an NRPS adenylation domain and attached to the carrier domain of the PvdI or PvdJ proteins2. LgrA has been hypothesized to be an evolutionary descendent of an N-sugar transformylase46. Both LgrA and N-sugar transformylases are considerably shorter in length, and lack the secondary structure insertions of PvdF, again suggesting that PvdF belongs to a distinct structural class of N10-fTHF dependent formyltransferase.</p><p>PvdF maintains the 7-stranded β-sheet core in a formyltransferase fold common to the N10-fTHF dependent transformylases, and the catalytic triad characteristic for this class of enzymes is conserved. The structurally unique features of PvdF are likely responsible for interaction with the smaller amino acid substrate. The mechanism previously defined for the GART enzymes and hypothesized to be conserved in the class likely holds for PvdF. A detailed steady state kinetic analysis of human GART demonstrated an ordered-sequential kinetic mechanism in which the folate binds first55. Subsequent, pre-steady state kinetic experiments for the E. coli GART defined a random sequential kinetic mechanism in which folate and GAR bind in no obligatory order, but for which the apoenzyme has higher affinity for fDDF than GAR49. For PvdF, varying the concentration of the substrate analogue fDDF (OHOrn in excess) generated Michaelis-Menten kinetics with kcat and Km values in keeping with an enzyme from secondary metabolism. N5-hydroxyornithine was a difficult substrate with which to work, and when used as the varied substrate, the data were not reproducible and provided kinetic values that were not physiologically relevant. However, when the ornithine hydroxylase (PvdA) of the same biosynthetic pathway was used to generate the OHOrn in situ, reproducible data could be generated. Interestingly, the curve was nonhyperbolic and not well fit by Michealis-Menten nor a substrate inhibition model (Figure 6). Instead, these data represent a model defined by Ferdinand53 and later echoed by Segel54 in which the bireactant system shows random binding of the two substrates, but favors binding of the folate over the OHOrn, very similar to the kinetic models for the GART proteins.</p>

PubMed Author Manuscript

Poly(Ionic Liquid) Based Chemosensors for Detection of Basic Amino Acids in Aqueous Medium

Naked-eye detection of amino acids (AA) in water is of great significance in the field of bioanalytical applications. Herein, polymerized ionic liquids (PILs) with controlled chain length structures were synthesized via reversible addition–fragmentation chain-transfer (RAFT) polymerization and post-quaternization approach. The AA recognition performance of PILs with different alkyl chain lengths and molecular weights was evaluated by naked-eye color change and ultraviolet-visible (UV–vis) spectral studies. These PILs were successfully used for highly sensitive and selective detection of Arg, Lys, and His in water. The recognition performance was improved effectively with increased molecular weight of PILs. The biosensitivity of the PILs in water was strongly dependent on their aggregation effect and polarization effect. Highly sensitive and selective detection of AA was successfully accomplished by introducing positively charged pyridinium moieties and controlled RAFT radical polymerization.

poly(ionic_liquid)_based_chemosensors_for_detection_of_basic_amino_acids_in_aqueous_medium

Introduction<!>Materials<!>Instruments<!>Preparation and characterization of PILs<!>FTIR characterization<!>DLS measurement<!>GPC measurement<!>General UV–vis spectral measurements<!>Synthesis and characterization of PILs<!><!>Molecular recognition ability of PIL for detecting amino acid in ethanol<!><!>Molecular recognition ability of PIL for detecting amino acid in ethanol<!><!>Molecular recognition ability of PIL for detecting amino acid in pure water<!><!>Molecular recognition ability of PIL for detecting amino acid in pure water<!><!>Molecular recognition ability of PIL for detecting amino acid in pure water<!><!>Conclusions<!>Author contributions<!>Conflict of interest statement

<p>Amino acids (AA) play an important key role in many physiological processes (Mohr et al., 1998; Shahrokhian, 2010). With increasing attention paid to human health, including diagnosis and treatment of diseases, scientists have devoted a lot of energy in exploring new methods for amino acid analysis (Vychytil et al., 2003; Zhou and Yoon, 2012). The traditional detection method mainly comprises the introduction of some electrophilic groups and structures like aldehyde that react with amine group (Xu et al., 2017). However, the prevalent problem is not only the solubility of AA in water but also the occurrence of weak intermolecular interaction with recognition receptors in water (Riikka et al., 2014; Samantha and Francesca, 2014). Hence, it is urgently needed to develop new amino acid sensors that are not only highly sensitive but also capable of molecular recognition in aqueous system that will aid in the enhancement of bioanalytical applications (Ooyama et al., 2013).</p><p>In recent years, molecule-based ion sensors are getting considerable attention because of their interaction with Lewis basic substrates in water, resulting in marked color and/or luminescence changes (Li et al., 2013; Ding et al., 2015; Wan et al., 2017). Gold nanoparticles modified with p-sulfonato-1, 3-dialkoxycalix[4]arene thiol have been reported to probe AA in aqueous solution (Patel and Menon, 2009). Phosphonate cavitands have been known as powerful receptors that possess molecular recognition properties toward AA (Roberta et al., 2016). In the process of amino acid detection, the calix[n]arenes have aroused many researchers' concern and interest. However, the traditional sensors have some disadvantages requiring time-consuming design and more synthetic steps (Sasaki et al., 2002). Recently, water-soluble conjugated polymers with charged groups as novel biosensors have shown tremendous potential for protein recognition (Huang et al., 2005; Chen et al., 2009, 2011). All the analytical processes were performed mainly through electron transfer (ET) and electrostatic interaction (Vilkanauskyte et al., 2002). However, such polymer sensor was limited to the conjugated structure with ET center (Tan et al., 2012), whereas no other work is reported using polymer without ET center for the biosensing applications (Zhou et al., 2011).</p><p>Polymer ionic liquid (PIL) is the polymer that contains repeating units of ionic liquid. Polymerized ionic liquids (PILs) possess unique features that belong to both polymer and counter intrinsic properties of IL with anion and cation (Zhang et al., 2013; Mahsa et al., 2017). In recent years, PILs have been widely applied in organic dispersant, nano composite materials, electrochemical, adsorbent, and separation techniques (Reeca et al., 2006; Mecerreyes, 2011; Li et al., 2017). Although PILs have extensive application values, their application as molecular probes for detection of AA in water has not been reported yet.</p><p>In this paper, we introduce a kind of PILs with well-controlled architectures via reversible addition-fragmentation chain-transfer polymerization (RAFT polymerization) and post-quaternization approach (Yuan et al., 2011; Mori et al., 2012). The AA recognition performance of PILs with different alkyl chain lengths and molecular weights was evaluated by ultraviolet-visible (UV–vis) spectral studies and color changes in ethanol and in pure water. The experimental results and theoretical analysis showed that PILs detected amino acid in ethanol by the strong ionic interactions and hydrogen bonding interactions. These PILs were successfully used for highly sensitive and selective detection of Arg, Lys, and His in water, whose recognition was dependent on the aggregation effect and electrostatic of PILs and basic AA. To the best of our knowledge, no work has been reported so far on the polymeric ionic liquid that can detect AA in water in highly sensitive manner without any special recognition groups.</p><!><p>All the AA and alkyl halides were purchased from Aladdin and AA standards are of 98–99% purity. 4-Vinylpyridine (4-VP) and N, N-Dimethylformamide (DMF) were purified by distillation after pressure. The chain transfer agent cumyl dithiobenzoate (CDB) was synthesized according to literature method (Li et al., 2015). The initiator 2, 2′-azobis (2-methylpropionitrile) (AIBN) was purchased from Sigma-Aldrich and purified by recrystallization in ethanol before use. The other reagents were used as received.</p><!><p>A Cary 1E UV–vis spectrophotometer was used to measure absorption spectra at 25°C. 1H NMR spectral analysis was carried out using Bruker AV-400 NMR spectrometer. FT-IR spectra were performed on a Nicolet NEXUS Fourier transform infrared spectrometer. Dynamic light scattering (DLS) was performed on Nano Particle Analyzer (Zetasizer Nano ZS90, Malvern Instruments). Gel permeation chromatography (GPC) was performed on a Waters 1515 apparatus.</p><!><p>4-VP (3.85 mmol), AIBN (3.1 mg, 0.019 mmol), CDB (20.7 mg, 0.076 mmol), and DMF (5 mL) were added into a 10 mL round-bottom flask. A clear solution was obtained after stirring for 10 min. After five freeze pump thaw cycles for degassing in order to remove the oxygen from the system, the flask was sealed then and placed at 65°C for 48 h. Thereafter, the product was subsided in ether, centrifuged, and washed with ether to remove the unreacted 4-VP monomer. The polymer product, P4VP1, obtained was dried in a vacuum at 40°C.</p><p>P4VP2 was prepared by using the similar method: 4-VP (7.61 mmol), AIBN (3.1 mg, 0.019 mmol), CDB (20.7 mg, 0.076 mmol), and DMF (5 mL) were added into a 10 mL round-bottom flask. The sample is treated and purified in a same way as P4VP1.</p><p>P4VP1 or P4VP2 (0.1 g, 0.95 mmol) was added into 3 mL chloroform, followed by the addition of bromoethane (9.5 mmol). The reaction mixture was heated at 60°C for 48 h and then kept for cooling at room temperature. The solid was separated out and washed with chloroform for five times to remove the unreacted P4VP. The resulting solid was filtered and vacuum-dried at 40°C to produce the P4VP1-Br2 and P4VP2-Br2. P4VP1-Br4 and P4VP1-Br8 were synthesized by the reaction of P4VP1 with n-butyl bromide or n-octyl bromide, respectively.</p><!><p>FT-IR spectra were recorded on a Nicolet NEXUS Fourier transform infrared spectrometer, and the samples were prepared as follows: the PIL sample was dissolved in ethanol and then dripped on a potassium bromide sheet to make it dry perfectly. P4VP1-Br2 solution (4.3 mM in ethanol) with addition of 2 × 10−4 M L-Arg was also prepared using the same method. The spectra were performed in a Biorad FTS 165 spectrophotometer with a resolution of 4 m−1 (32 scans).</p><!><p>The average hydrodynamic sizes of PIL in water were determined by DLS using a Nano Particle Analyzer with a He Ne laser and 90° collecting optics operating at λ = 660 nm at 25°C. The samples were prepared as follows: P4VP1-Br2 solution (4.3 mM in water) with the addition or no addition of 2 × 10−4 M L-Arg. The solution was filtered using a 0.22 μm filters.</p><!><p>The molecular weights and molecular weight distribution (PDI = Mw/Mn) of the synthesized polymer samples (5 mg/mL DMF solution) were determined by gel permeation chromatography (GPC) equipped with a Waters 1515 apparatus and a PD2020 light scattering detector, using DMF as eluent. The flow rate was 1.0 mL.min−1, and polystyrene samples were used as standards.</p><!><p>In the titration part, PIL solution (4.3 mM) was prepared in ethanol or water. UV–vis spectra were also obtained in ethanol or water solution. Other AA (0–2 × 10−4 M) and chemicals were prepared in deionized water or ethanol. The limit of detection (LOD) for Arg is calculated according to a signal-to-noise ratio of 3 (ΔI = I–I0), where I and I0 are the wavelength intensities of the PIL in the presence and absence of Arg, respectively (Lu et al., 2017).</p><!><p>The synthetic procedure of PILs with different alkyl chain lengths was outlined in Figure 1. The neutral polymers P4VPs with different molecular weights were obtained through RAFT polymerization, azo-diisobutyronitrile (AIBN) as initiator agent, and CDB as RAFT chain transfer agent. GPC analyses demonstrated that P4VPs with narrow molecular weight distribution were successfully obtained (Table 1). In the two cases, the molecular weights from GPC are far smaller than the values obtained from 1H NMR (Figure S1). This may be attributed to the different aggregates in solution or the deviation from the polystyrene samples (Schilli et al., 2004). The neutral polymers P4VP1 and P4VP2 were converted to corresponding P4VP1-Br2 and P4VP2-Br2, which were achieved by stirring the polymer with bromoethane in CHCl3 at 60°C for 2 d. Such a post-quaternization approach is highly valuable regarding the easy purification and further characterization of neutral polymer products. Meanwhile, P4VP1-Br4 and P4VP1-Br8 were also obtained by reacting the P4VP1 with n-butyl bromide and n-octyl bromide, respectively.</p><!><p>The synthesized method of polymer ionic liquid.</p><p>Polymer molecular weight data of P4VP.</p><p>GPC was used to measure molecular weights of the polymers.</p><p>Determined by 1H NMR spectroscopic analysis in CDCl3, by comparing the proton peak intensity of P4VP (δ 8.2–8.6 ppm, 2H) and the RAFT end-group (δ 6.8–7.9 ppm, 5H).</p><!><p>In order to evaluate the recognition ability of PIL, 8 different AA (2 × 10−4 M) were added to P4VP1-Br2 ethanol solutions (4.3 mM). A few seconds later, the colorless solutions containing Arg, Lys, His, Pro, Trp, Phe, and N-Boc-Pro were found turning to blue color (Figure 2). At the same time, UV-vis spectral analysis showed the appearance of a new absorption peak at 600–650 nm on the addition of these AA (Figure 3A). However, the color or adsorption spectrum of P4VP1-Br2 was found to have no effect on the addition of acidic AA (e.g., Asp). In addition, basic AA (Arg and Lys) showed more obvious recognition than the neutral AA and acidic AA, which are the evidences of occurrence of electrostatic interaction between the PIL and AA with carboxyl groups. The recognition performance for N-Boc proline suggests that the key factor for recognition is the presence of carboxyl groups. But the stable identification for AA depends on the presence of amino groups because the color and UV-vis spectral changes of the ethanol solution for P4VP1-Br2 and N-Boc Proline disappeared after placing 24 h at room temperature (Figures S2, S3).</p><!><p>Photographic images of P4VP1-Br2 ethanol solutions containing different amino acids.</p><p>UV-visible spectra of 4.3 mM P4VP1-Br2 responding to 1 × 10−4 M amino acids in ethanol solutions (A), and different PILs ethanol solutions of 4.3 mM P4VP1-Br2 with the addition of 1 × 10−4 M L-Arg (B).</p><!><p>Control experiments of 4-VP monomer and P4VPs with the addition of Arg, Lys, or His showed no peaks in the range of 600–650 nm and also no color change was observed. This demonstrated that the recognition performance of PIL might be due to its aggregated and polarized nature (Wang et al., 2011).</p><p>In addition, we have also studied the effect of different alkyl chain lengths on the recognition performance of PILs. There were similar changes in the UV-vis spectra of P4VP1-Br4 and P4VP1-Br8 induced by the addition of Arg (Figure 3B). It was noticed that with increase in the alkyl chain length, the recognition performance decreased. P4VP1-Br2 with the shortest alkyl chain length was found to exhibit the best molecular recognition performance. For pyrrolidinium ionic liquids, the length of alkyl chain on the cation was observed to have a significant influence on their physicochemical properties (such as the polarity and conductivity). The decreased recognition performance of P4VP1-Br4 and P4VP1-Br8 might be due to the steric effects causing the increasing distance between the PIL and amino acid (Lee and Prausnitz, 2010).</p><p>To understand the geometry of the PIL-amino acid 1:2 complex, the density functional theory (DFT) calculations were carried out using Gaussian 09 package (Rohit et al., 2007; Rega et al., 2009; Yan et al., 2012). The B3LYP/6-31+G(d)-optimized structures for the complexes of receptor with Arg are represented in Figure S4. The theoretical results indicated that the strength of interactions among the ionic interactions and the hydrogen bonding interactions between different polymer chains and amino acid groups might be the important factor responsible for better recognition performance of PILs than their corresponding monomer counterparts.</p><p>Furthermore, the FTIR spectra of P4VP1-Br2 before and after amino acid recognition are presented in Figure 4. There were some common peaks appeared at around 3,040, 1,640, and 1,470 cm−1, which were the typical characteristic peaks of P4VP1-Br2. After combination with amino acid in solution, the new peaks at 1,630 and 1,340 cm−1 were obtained that attributed to ionic bond interactions between the two moieties (Zhou et al., 2012). After PIL reacted with the carboxylic acid group of amino acid, a new structure was formed that marked the significant recognition performance of PIL.</p><!><p>FTIR of P4VP1-Br2 (A), L-Arg (B), and P4VP1-Br2 after combining with L-Arg (C).</p><!><p>Limited by the insolubility of other PILs in water, we investigated the molecular recognition ability of P4VP1-Br2 in pure water. Different AA (2 × 10−4 M) were added into P4VP1-Br2 aqueous solution (4.3 mM). After a short period, the colorless solutions containing Arg, Lys, and His were found turning green that could be readily visualized by the naked eyes (Figure 5). However, the P4VP1-Br2 solution containing the other AA showed no color or adsorption spectral changes, which indicated that the P4VP1-Br2 responded selectively to Lys, Arg, and His. In Figure 6, it is clear that there is a significant change in absorbance intensity at 610 nm upon addition of Lys, Arg, and His. With the addition of Arg (0–2 × 10−4 M), absorption peak at 375 nm showed great increase in intensity along with the appearance of a new peak at 610 nm in UV–vis spectra. Interestingly, the peak at 375 nm, which was assigned to π-π stacking of PIL, increased with the addition of Arg. This result demonstrated that the interactions between PIL and Arg might be due to the aggregation effect. It is clear that the absorbance intensity at around 610 nm showed a significant change with increased concentration of these three AA. Binding constant of 6.1 × 108 for 1:2 binding of PIL:Arg was obtained. The binding constants for Lys and His were calculated as 5.1 × 108 and 2.9 × 107, and the linear fitting constants were found as 0.999, 0.997, and 0.991 for Arg, Lys, and His, respectively (Table 2). Particularly, the addition of other AA (Phe, Proline, Asp, and N-Boc-Pro) had no effect on the color or UV–vis adsorption spectrum for P4VP1-Br2 and Arg (Figure 7), which showed the selective recognition ability of PIL for basic AA. In water solution, on one hand, due to high polarity of water, part of the hydrogen bond was damaged and the hydrogen bonding interaction between AA and PIL was greatly reduced. On the other hand, the interaction of positive and negative charges promoted the predetermined selectivity of PILs for basic AA. In order to understand the crucial interactions involved in recognition, we also did the following experiments: a. P4VP1-Br2+Gnd.HCl (guanidine.HCl) in water and b. P4VP1-Br2+NH4Cl in water. However, no response was observed in the UV absorption and color changes in these experiments. The results indeed demonstrated the very high selectivity of the synthesized quaternary PIL, thereby highlighting the importance of amino acid functionality in addition to basic side chain group in the recognition.</p><!><p>Photographic images of P4VP1-Br2 water solutions containing different amino acids.</p><p>UV-vis spectra of P4VP1-Br2 (4.3 mM) in water in the presence of 2 × 10−4 M different amine acids (A) and Evolution of UV-vis spectrum of P4VP1-Br2 (4.3 M) on addition of L-Arg = 0–2 × 10−4 M in water. The inset showed the changes of absorbance at 610 nm against the added L-Arg (B).</p><p>Ka and R obtained from the titration of PIL with the interacting Amino Acids (aa).</p><p>UV-vis spectrum of P4VP1-Br2 (4.3 mM) on the addition of L-Arg (1 × 10−4 M) in water (the concentration of the other amine acid is 1 × 10−4 M).</p><!><p>Furthermore, we have studied the effect of molecular weights of polymer on the recognition performance. Interestingly, the recognition performance of P4VP2-Br2 with the same alkyl chain length but different molecular weights compared with P4VP1-Br2 for L-Arg appeared to increase in water. The binding constant for Arg was 7.1 × 108, and the coefficient of the linear fit was 0.993 (Table 2). Here, the detection limit of Arg was 0.026 μM, which was close to the minimum detection limit for the determination of Arg among all the developed approaches as shown in Table 3. PIL as amino acid sensor shows the high sensitivity in water. P4VP2-Br2 also showed the better recognition performance for Lys and His than P4VP1-Br2. This might be due to increased aggregation phenomenon with the increase of molecular weight, and hence, recognition performance increased substantially. It is advantageous that the recognition performance of the polymer sensor can be simply regulated by varying the molecular weights accordingly.</p><!><p>Detection limit of Arg by various detection methods.</p><!><p>Dynamic light scattering (DLS) experiments were used to measure the hydrodynamic diameters of P4VP1-Br2 in the presence of Arg (Figure 8). DLS results indicated that Arg-involved assembly structures comprise a much smaller volume in such complex structures. The polarity of cation of P4VP1-Br2 was found effective for the electrostatic interaction between polymer and the carboxyl group of Arg. It was noticed that Arg and PIL tend to aggregate and assemble more easily than the other AA with PIL (Patel and Menon, 2009). The further detailed analysis is currently under investigation.</p><!><p>Dynamic light scattering of P4VP1-Br2 4.3 mM before and after treatment with 2 × 10−4 M Arg.</p><!><p>In summary, a water-soluble polymer ionic liquid (PIL) sensor is synthesized using RAFT polymerization and post-quaternization approach. The PIL acted as a highly efficient colorimetric sensor and selectively detected the basic AA in an aqueous environment. Further research proved that the recognition performance of PIL in ethanol depends on the strength of intermolecular electrostatic interactions and hydrogen-bond interactions between PIL and the AA. The selective recognition for basic AA in water could be owing to the aggregation effect and polarization effect. Deeper recognition mechanisms are being further studied. The new sensor was realized by simply introducing the positively charged pyridinium moieties and controlled RAFT radical polymerization that paved the new method for designing highly sensitive and highly selective biosensor for special AA recognition.</p><!><p>XL and KW contributed to the experimental studies and manuscript preparation. NM contributed to the theory calculations. XJ contributed to the study design, manuscript revision, and final version approval.</p><!><p>The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.</p>

PubMed Open Access

Toward a more step-economical and scalable synthesis of spongistatin 1 to facilitate cancer drug development efforts\xe2\x80\xa0

An efficient, step-economical, and scalable synthesis of a diene-bearing AB spiroketal fragment of spongistatin 1, and a demonstration of its efficient coupling to an aldehyde derived from silylformylation of a homopropargyl alcohol to produce the entire complex C(13)\xe2\x80\x93C(17) linker region are described. The scalability of the synthesis of the AB spiroketal fragment was demonstrated by the preparation of 34.5 grams by one chemist in ~60 workdays, and more than 40 grams overall. With this material in hand and having established a method for its efficient coupling to the CD fragment, we have set the stage for the rapid synthesis and evaluation of a series of analogs of the CD spiroketal.

toward_a_more_step-economical_and_scalable_synthesis_of_spongistatin_1_to_facilitate_cancer_drug_dev

Introduction<!>Study design<!>Results and discussion<!>Conclusions

<p>Natural products – by virtue of their structural complexity and variety – provide a rich forum for reaction design and chemical invention and innovation. When they are possessed of truly extraordinary biological activity and at the same time are available in significant quantity only through total chemical synthesis, they provide much more than that, and it would be difficult to identify a natural product that more clearly exemplifies this than spongistatin 1. This extraordinarily complex and exceedingly precious anti-mitotic agent was first reported nearly simultaneously by three research groups in 1993,1–3 and has been reported to have an average IC50 value against the NCI panel of 60 human cancer cell lines of 0.12 nM.4 Seven research groups have reported syntheses of spongistatin 1 and/or 2,5–11 and, notably, the Smith team ultimately produced 1 gram of fully synthetic spongistatin 1.7c–e Despite all of this excellent and pioneering synthetic chemistry, the possibility of developing a synthesis that could deliver the kinds of amounts of spongistatin 1 or an analog thereof that will be needed for clinical development and beyond still seems quite remote. Thus, while the world does not need an eighth synthesis of one of the spongistatins, it certainly does need better chemotherapeutics, and one important part of addressing the extremely daunting challenge of turning spongistatin 1 or more likely a designed analog thereof into an effective cancer drug will be the development of significantly more efficient, step-economical, and scalable synthetic chemistry.</p><!><p>More recently, a University of Pennsylvania and Eisai team led by Smith demonstrated that the CD spiroketal is most likely not directly involved in the binding of spongistatin 1 to β-tubulin by the synthesis of a "diminutive congener" wherein the CD spiroketal and C(13)–C(17) linker region were replaced with a simple tether and the subsequent demonstration that significant anti-mitotic activity was retained, albeit with a reduction in potency.12 Inspired by this sophisticated and elegant work, and, in part, because the synthesis of the CD spiroketal has been one of the most difficult challenges in this arena, we have initiated a program whose ultimate long-term goal is the preparation and biological evaluation of a series of CD spiroketal-modified analogs of spongistatin 1 (Figure 1). In addition to an efficient synthesis and large supply of the "EF Half" of spongistatin 1 (C(29)–C(51), not shown), this would entail the synthesis of a series of ABCD fragments (1) with the targeted CD spiroketal analogs incorporated. Since we were and remain quite sober about the likely necessity of preparing many such analogs in order to have a reasonable chance at identifying one that retained the sub-nanomolar potency of the natural product while being significantly easier to synthesize, we were convinced that the only way this would be feasible would be to first develop new synthetic chemistry with two main objectives: 1) a highly efficient and step-economical method for the rapid union of the AB spiroketal with a variety of CD spiroketal analogs in a way that would directly establish the entire complex C(13)–C(17) linker region between the spiroketals from simple precursors, and 2) a significantly more step-economical and scalable synthesis of the AB spiroketal than any yet advanced that could then be used to synthesize a large enough supply to support all of the CD spiroketal analog work and beyond. For the former objective, we envisioned application of our recently reported complex fragment coupling by crotylation methodology13 in combination with the silylformylation/crotylation/Tamao oxidation/diastereo-selective tautomerization methodology.13,14 For the latter objective, we envisioned the use of another recently developed method, asymmetric aldehyde isoprenylation,13 with aldehyde 4 followed by spiroketalization to produce 2, and an aldol coupling of fragments 5 and 6 to produce 4. Herein we describe the development of a synthesis of an AB spiroketal of type 2 that has allowed us to produce a large supply, and a demonstration of its efficient coupling with a model aldehyde derived from a homopropargyl alcohol of type 3 to directly produce the entire C(13)–C(17) linker region.</p><!><p>Our synthesis of the precursor to an aldehyde of type 5 commenced with conversion of 2-(trimethylsilyl)ethanol (TMSEOH) to β-ketoester 7 in 83% yield (Scheme 1a). The mixed Na/Li dienolate of 7 was alkylated with PhCH2OCH2Cl to give 8 in 65% yield (86% based on recovered 7).15 This set the stage for the Noyori hydrogenation16 which proceeded smoothly with 0.5 mol % of ((R)-BINAP)RuCl2 to give 9 in 96% enantiomeric excess (ee). After protection of the alcohol with triethylsilyl chloride (TESCl) to give 10, the benzyl (Bn) group was removed by hydrogenation, allowing the isolation of alcohol 11 in 72% overall yield from 8.</p><p>The synthesis of ketone 6 began with an application of our recently reported direct and enantioselective allylation of β-diketones,17 in this case acetylacetone (Scheme 1b). Allyl-silane (S,S)-12 (commercially available, and easily prepared on tens of grams scale from (S,S)-13 and allyltrichlorosilane) reacts smoothly with acetylacetone to provide 14. The only technical challenge we faced as we scaled this reaction up was the volatility of 14, which required the use of low boiling point solvents for the workup and purification. Once optimized, the reaction proved capable of delivering ~10 g (~70 mmol) of 14 in a single run in 65% yield and 89% ee. In addition, the diamine (S,S)-13 was recovered in 92% yield, and recycled. This reaction greatly simplifies the synthesis of the AB spiroketal. The ketone that is to be used in the aldol reaction and that becomes the spiroketal (see Scheme 2) is purchased in the form of a commodity chemical, acetylacetone, and is never protected or masked. At the same time the tertiary carbinol – one of the more challenging aspects of the target – is directly installed in a single step from acetylacetone. Protection of the tertiary alcohol as its TES ether proceeded smoothly in 97% yield, and provided the methyl ketone 6 in just two steps.</p><p>Swern oxidation18 of 11 provided access to aldehyde 15 and set the stage for the aldol coupling reaction (Scheme 2). For stereochemical induction, we turned to the use of a chiral boron enolate derived from (−)-B-chlorodiisopinocampheyl-borane ((−)-ipc2BCl), according to the procedure reported by Paterson.19 (−)-Ipc2BCl is inexpensive and its use on larger scales typically presents only one (indirect) technical problem: the oxidative workup creates two equivalents of ipcOH, which can be, and almost always is in practice, difficult to separate from the desired aldol product. In the present case, the aldol coupling of 6 with 15 proceeded smoothly to give 16 (with ≥10:1 diastereoselectivity as judged by 1H NMR spectroscopic analysis of the unpurified reaction mixture). Instead of struggling with a difficult separation at this stage, however, the mixture of 16 and ipcOH was simply acetylated. Separation of the desired product 17 from the ipcOAc was far more straightforward, and delivered 17 in 72% overall yield from alcohol 11. This 3-step sequence proved readily scalable, and indeed, could be and was used to prepare ~15 g of 17 per run. Ozonolysis of alkene 17 provided aldehyde 18 which was, without purification, subjected to an asymmetric isoprenylation reaction according to our recently reported procedure using (R,R)-19.13 The unpurified product (analysis of which by 1H NMR spectroscopy revealed a diastereoselectivity for the isoprenylation reaction of ≥15:1) was subjected to spiroketalization with camphorsulfonic acid (CSA), leading to the smooth production of 20 in 66% overall yield from 17. Finally, the liberated tertiary alcohol was trifluoroacetylated to give completed AB spiroketal fragment 21 in 97% yield.</p><p>The synthesis of AB spiroketal 21 thus proceeds with a longest linear sequence of 12 steps from TMSEOH in 18% overall yield (Figure 2). Because 6 is accessed in only 2 steps, the total step count is only 14 steps (this count does not include the preparation of 12 and 19). This comprises by far the most step-economical synthesis of a fully configured AB spiroketal ready for coupling to the CD spiroketal fragment yet reported. While step-economy can be valuable in and of itself (for example for the more rapid production of analogs), it is not the same thing as scalability. The most useful and efficient syntheses will combine step-economy with true scalability, and while scalability can be easily claimed, it is, in the end, best demonstrated. The synthesis described has thus far been used to prepare 43.5 g of 21. Of course, even this is not particularly informative without an accompanying description of how much time and effort was required of how many chemists. In that regard, we can report that 34.5 g of 21 was prepared in a single campaign carried out entirely by one of us (S.K.R.) in ~60 workdays. It is perhaps not unreasonable to suggest based on this that the route could serve as the starting point for the development of a process that could deliver hundreds of grams or more of 21 by an industrial process chemistry team, should a clinical candidate ultimately emerge.</p><p>Of course, none of this will mean very much without an efficient method to both couple AB spiroketal fragment 21 to the CD spiroketal fragment or analogs thereof, and at the same time establish the entire complex C(13)–C(17) array of a 1,1-disubstituted alkene, three stereocenters, and a ketone that comprises the linking region between the spiroketals. We have reported the development of a fragment coupling by crotylation method13,20 that was designed to accomplish exactly that as demonstrated with simplified model dienes, and given such a large supply of 21, we thought it prudent to evaluate its performance in the fragment coupling process. In that sequence, the diene is subjected to a Pd(PPh3)4-catalyzed hydrosilylation reaction with HSiCl3 according to Tsuji's protocol,21 and then to chiral diamine complexation. The resulting complex crotylsilane species is then reacted with an aldehyde partner in a Sc(OTf)3-catalyzed crotylsilylation reaction.22 We applied the first two operations to 21 to generate complex crotylsilane 22 (Scheme 3). This was split in half and used in the Sc(OTf)3-catalyzed crotylation reactions of two different aldehydes which contain the necessary functionality for the installation of the C(16) methyl-bearing stereocenter and the C(17) ketone. Aldehyde 23 represented the more conservative choice, and its crotylation reaction with 22 produced 24 in 59% yield and with >10:1 diastereoselectivity. We were at first concerned about the moderate efficiency of this reaction relative to the 73–79% yields that we had consistently achieved with model dienes,13 but quickly discovered that a significant amount of the product had undergone TES ether cleavage during the work-up (with n-Bu4NF•(H2O)3), a presumably fixable problem. Confirmation that diene 21 performs well was then provided by the fragment coupling crotylation reaction with aldehyde 25, which, as we have previously reported,13 arises very efficiently from the silylformylation of the corresponding homopropargylic alcohol (e.g. 3, Fig. 1). This reaction produced 26 in 74% yield and with >10:1 diastereoselectivity. Finally, subjection of 26 to our optimized conditions for a Tamao oxidation/diastereoselective tautomerization reaction13 resulted in the isolation of 27 as the major product of a 5:1 mixture of diastereomers (at C(16)) in 75% yield. These model fragment coupling transformations thus serve as convincing proofs of concept for the planned coupling of the AB spiroketal 21 with a variety of CD spiroketal analogs and concurrent establishment of the C(13)–C(17) linker region in a rapid and highly efficient fashion.</p><!><p>We have described the development of a highly step-economical and scalable synthesis of the AB spiroketal fragment of spongistatin 1. The scalability and the step-economy were demonstrated by the synthesis of 34.5 g of 21 by one chemist in a ~60 workday campaign. This unprecedented level of synthetic efficiency was achieved primarily through methodological innovations such as the β-diketone allylation, aldehyde isoprenylation, and fragment coupling by crotylation reactions that both facilitated efficient and scalable access to the AB spiroketal target and rendered the target simpler than the corresponding AB fragments used in other spongistatin syntheses. In addition, the suitability of this material for use in the fragment coupling by crotylation methodology was demonstrated in coupling reactions with two different model aldehydes. By taking the time to innovate and develop this chemistry, we think we have taken significant steps towards the ultimate goal of a synthesis of spongistatin 1 or an analog thereof that can deliver the kinds of amounts that would be needed for its clinical development. By leveraging the synthetic efficiency that derived from that innovation to build a "war chest" of more than 40 g of 21, and by developing the fragment coupling by crotylation chemistry we have further put ourselves in a strong position with regard to our ability rapidly to synthesize and evaluate a series of CD spiroketal analogs of the natural product, and subsequently and rapidly to produce significant quantities of any such analogs that showed promise for full pre-clinical evaluation. A correspondingly efficient and scalable synthesis of the EF fragment and some good ideas about which analogs to target are the other major pieces of the puzzle, and our efforts along those lines continue.</p>

PubMed Author Manuscript

Compact representation of continuous energy surfaces for more\nefficient protein design

In macromolecular design, conformational energies are sensitive to small changes in atom coordinates, so modeling the small, continuous motions of atoms around low-energy wells confers a substantial advantage in structural accuracy; however, modeling these motions comes at the cost of a very large number of energy function calls, which form the bottleneck in the design calculation. In this work, we remove this bottleneck by consolidating all conformational energy evaluations into the precomputation of a local polynomial expansion of the energy about the \xe2\x80\x9cideal\xe2\x80\x9d conformation for each low-energy, \xe2\x80\x9crotameric\xe2\x80\x9d state of each residue pair. This expansion is called Energy as Polynomials in Internal Coordinates (EPIC), where the internal coordinates can be sidechain dihedrals, backrub angles, and/or any other continuous degrees of freedom of a macromolecule, and any energy function can be used without adding any asymptotic complexity to the design. We demonstrate that EPIC efficiently represents the energy surface for both molecular-mechanics and quantum-mechanical energy functions, and apply it specifically to protein design to model both sidechain and backbone degrees of freedom.

compact_representation_of_continuous_energy_surfaces_for_more\nefficient_protein_design

2 Introduction<!><!>3.1 Preliminaries<!>3.2 Basic least-squares method<!>3.3 Modified least-squares method<!><!>3.3 Modified least-squares method<!>3.4 A fast algorithm for modified least-squares fitting<!>3.5 Sparse atom-pair energies (SAPE)<!>3.6 Attaining the required accuracy<!>3.7 Sampling to train and validate least-squares fits<!>3.8 Application in proteins design algorithms<!>3.9 Complexity of energy evaluations<!>3.10 Computational Experiments<!>3.11 Applications of EPIC to other algorithms<!>3.12 EPIC can accommodate higher-than-pairwise energies<!>4 Results<!>4.1 Application to protein designs<!>4.2 Quantum-mechanical energies<!>5 Conclusions<!>Software Availability

<p>Computational design algorithms are an effective approach to engineer proteins and discover new drugs for many biomedically relevant challenges, such as drug resistance prediction,1 peptide-inhibitor design,2 and enzyme design.3 Protein design algorithms search through large sequence and conformational spaces for sequences that will fold to a desired structure and perform a specific function. One of the key challenges in protein design is modeling and searching the many continuous conformational degrees of freedom inherent in proteins and other molecules. Protein design algorithms must estimate optimal values for all these degrees of freedom in order to optimize the sequence of the protein, or to optimize the chemical structure of the ligand if used for drug design. Molecular dynamics simulations can be used for this purpose if the protein sequence and ligand are known, because they can move all of the molecule's degrees of freedom,4 but these simulations are computationally expensive and must be run separately for each sequence or ligand chemical structure. Hence, this direct simulation approach is unsuitable for searching large combinatorial design spaces. For example, many protein design problems require searching over trillions of sequences—far too many for individual molecular dynamics runs.</p><p>To address the combinatorially large sequence spaces inherent to protein design, dedicated protein design algorithms efficiently choose an amino-acid type and conformation for each residue in a protein that, together, minimize some energy function.5 Since the sidechain conformations of each amino-acid type are generally found in clusters, known as rotamers,6 the protein design problem has often been treated as a discrete optimization problem. In this case, the output is a set of rotamer assignments (a rotamer, including amino-acid type, is assigned to each residue). The objective function is an energy function, which maps conformations to their energies. However, because proteins are continuously flexible and have backbone as well as sidechain flexibility, some of the protein's internal coordinates will likely have functionally significant variations from the rotamer's "ideal" value (at the center of the cluster). Clashing ideal rotamers can often be converted to favorable conformations by relatively small adjustments in the sidechain conformations.7,8 Small adjustments in the backbone conformation away from the wild-type backbone can also be functionally significant.9–11 As a result, modeling of continuous flexibility has been shown to dramatically improve the accuracy of structural modeling in designs,8,11 even using a limited set of degrees of freedom, and has led to designs that perform well experimentally.1–3,5,7,12 Furthermore, attempts to mimic this effect by discrete sampling at a finer resolution have been shown to either poorly approximate the continuous solutions, or to be computationally prohibitive.8 Modeling additional continuous degrees of freedom, with the goal of modeling all conformational variations that significantly impact protein function, is expected to increase the accuracy of designs further.</p><p>Modeling of continuous flexibility in protein design can still exploit our knowledge of rotamers, because rotamers provide an excellent prior estimate of where "energy wells" in the conformational space of the protein are likely to be. Residues' sidechains will usually be found in the region of conformational space fairly close (e.g., within 10–20° for sidechain dihedrals) to an ideal rotamer, even with a relatively small rotamer library.13 As a result, if by using a "minimization-aware" search process one can find the nearest ideal rotameric conformation to the true Global Minimum-Energy Conformation (GMEC) of a protein, the GMEC itself can generally be found by local minimization initialized to that ideal rotameric conformation. Thus, protein design can fully account for continuous sidechain flexibility while still functioning as a "minimization-aware" search7 over discrete rotamer space. This same paradigm can be extended to continuous backbone flexibility if ideal conformations that include backbone motions—"residue conformations" or RCs11—are included in the search.</p><p>Such "minimization-aware" search can take multiple forms. For example, the iMinDEE algorithm8 produces a gap-free, provably accurate list of rotamer assignments in order of lower bound on minimized energy. iMinDEE performs energy minimization on each of these rotamer assignments in that gap-free order until the lower bound exceeds the best minimized energy Eb enumerated so far. At this point, any subsequently enumerated assignments would be guaranteed to have higher minimized energies than Eb, so Eb is provably the global minimum energy. iMinDEE enumerates rotamer assignments efficiently using the A* search algorithm.14,15 Monte Carlo search over rotamer space can also incorporate minimization,16,17 but without any provable guarantees. A Monte Carlo search will be minimization-aware if continuous minimization is performed for sequences and conformations during (rather than after) the search and the minimized energy is used in the calculation of acceptance probabilities for new rotamer assignments, as in16 and the final phase of.17 Nevertheless, a method without provable guarantees will likely require more conformations and sequences to be minimized to obtain the same gain in accuracy as iMinDEE, because unlike in iMinDEE, the conformations being minimized are not guaranteed to be the most promising ones. Furthermore, there is no finite number n for a given protein design problem such that enumerating n conformations by Monte Carlo is guaranteed to yield the GMEC.</p><p>Any minimization-aware method, however, will require a large number of subroutine calls to local minimization. Continuous energy minimization is computationally expensive, even with molecular mechanics-type energy functions that prioritize speed over accuracy. This causes the minimization of the energy function to be the bottleneck in protein design with continuous flexibility.</p><p>This bottleneck becomes more severe when more sophisticated energy functions are introduced. Computational protein design is typically performed with energy functions that prioritize speed over accuracy. For example, they typically use simplified implicit solvation models, such as EEF1.18 Vizcarra et al. 19 have investigated the use of the Poisson-Boltzmann model, a much more accurate implicit solvation model, in protein design. They found it to be amenable to representation as a sum of residue-pair interactions—the form required for most protein design algorithms— but orders of magnitude more expensive than EEF1. Other methods to improve energy function accuracy are likely to face the same problem. For example, quantitatively accurate descriptions of most molecular interactions require computation of the electronic structure using quantum chemistry, but methods to do this are very computationally intensive.20 Methods to reduce the number of calls to an energy function needed in protein design could allow more accurate energy functions to be used, and thus yield more accurate results.</p><p>In protein design with only discrete flexibility, precomputation methods are typically used to reduce the number of energy function calls needed—that is, the number of conformations for which the energy must be evaluated. Before the design is started, the interaction energy of each pair of ideal, rigid rotamers at different residue positions is precomputed and stored in an energy matrix. Then, an ideal rotamer is chosen for each residue based on the energies in this matrix, and no further calls to the energy function are needed during the actual design calculation. The number of energy function calls required is thus quadratic in the number of residues in the system (that is, it scales as the number of pairs of residues). A precomputed energy matrix is, however, of limited use if we want to model continuous flexibility. No benefit in design is gained by performing post-hoc minimization on the best conformation found using ideal rotamers.8 This is true even if a high degree of flexibility is used for minimization, e.g., if molecular dynamics techniques are used, because the designed sequence is already determined before minimization is performed. In contrast, a minimization-aware search performs local continuous minimization for all rotamer assignments that might be optimal, in order to find the true GMEC.7</p><p>Thus, an analogous energy matrix precomputation method for continuous flexibility would be very useful. It would ensure a polynomial number of energy function calls for minimization-aware protein design, in contrast to the exponential number of calls that may arise in minimization of all possibly optimal rotamer assignments (since the number of such assignments may be exponential with respect to the number of residues modeled). The search for rotamer assignments itself is unlikely to admit a polynomial-time algorithm, because it is NP-hard even to approximate.21,22 But a method to precompute pairwise energies for continuously flexible design would change the overall time cost from</p><p>(a large rotamer search cost) times (the energy function cost)</p><p>to</p><p>(a large rotamer search cost) plus (the energy function cost).</p><p>The rotamer search cost will necessarily be exponential in the worst case, if one wants to obtain the GMEC or an approximation to the GMEC within a fixed error threshold. But the energy function cost will be merely quadratic in the number of residues, indicating that the pairwise energy precomputation shifts the bottleneck away from the energy function calls. This brings the same improvement to minimization-aware design that energy matrix precomputation brought to non-continuously-flexible design.</p><p>We now present a pairwise energy precomputation method that admits continuous flexibility: EPIC (Energy as Polynomials in Internal Coordinates). EPIC computes a representation of the pairwise energy for each rotamer pair, not just at the rotamers' ideal values of the internal coordinates, but for values within specified ranges around the ideal ones (Fig. 1). This computation is performed before the rotamer search computation is begun. This allows the rotamer search to substitute the new quickly evaluable representation for the original energy function. EPIC is implemented in the OSPREY7,23,24 open-source protein design package, which has yielded many designs that performed well experimentally—in vitro 1–3,25–28 and in vivo 1,2,25,26 as well as in non-human primates.25 EPIC provides a significant speedup when used with OSPREY's default AMBER29,30- and EEF118-based energy function, but is also shown to be suitable for representing quantum-mechanical energies.</p><p>This paper makes the following contributions:</p><!><p>A compact, closed-form representation of energy as a function of continuous internal coordinates of a protein system.</p><p>A modified least-squares method to compute this representation.</p><p>A modified implementation of the iMinDEE 8 and DEEPer11 protein design algorithms, integrated into the OSPREY1,3,7,23,24 open-source protein design package, that makes use of this representation to achieve substantial speedups. It is available online24 as free software.</p><p>Computational experiments showing that compact and accurate EPIC representations are possible both for the standard energy function in OSPREY and for energies obtained by quantum chemistry at the SCF and MP2 levels of theory.</p><p>Computational experiments showing that EPIC greatly speeds up minimization-aware protein design calculations, thus allowing designs to include not only more flexible residues, but also more conformational flexibility at those residues.</p><!><p>EPIC, like most previous protein design algorithms, is designed for pairwise energy functions. Pairwise energy functions are sums of intra+shell and pairwise terms. Intra+shell terms are functions of the amino-acid type and conformation of one residue, and pairwise terms are functions of the amino-acid types and conformations of two residues. Each pairwise term represents the interaction between a pair of flexible residues, while each intra+shell term represents the internal energy of a residue plus its interactions with non-flexible "shell" residues (those that are frozen in a single, fixed conformation throughout the entire calculation). EPIC could be easily modified to include some higher-order terms for defined combinations of more than two residues: these terms can also be represented as polynomials in their residues' degrees of freedom.</p><p>To find the GMEC, we must find an amino-acid type and conformation for each flexible residue such that the sum of intra+shell terms for all flexible residues, plus the sum of pairwise terms for all pairs of flexible residues, is minimized. This problem is referred to as conformational search. Conformational search can also comprise sequence search, by searching for the best conformation across many sequences' conformational spaces. Many algorithms are available for this problem, including iMinDEE8 and DEEPer,11 which solve it with provable accuracy. EPIC can be used in any conformational search algorithm that models continuous flexibility, because it provides polynomials that can be directly substituted for intra+shell and pairwise terms of the energy function.</p><p>The essence of EPIC is to exploit the fact that for each pairwise or intra+shell energy, the energy in the vicinity of the minimum can be described well by a relatively low-degree polynomial (usually quadratic total degree, sometimes higher; see Fig. 5B). This description is computed using a modified least-squares method. We will refer to the states to which a residue may be assigned as residue conformations (RCs; cf. DEEPer11). In the absence of backbone flexibility, each RC will correspond to a sidechain rotamer. Within a residue conformation, the residue's continuous energy variations can be described by a set of internal coordinates, which are subject to box constraints (i.e., bounds on each internal coordinate).</p><p>Herein, the word "polynomial" will be used in two very different senses in the description of EPIC below. First, EPIC is a polynomial representation of the energy, namely, a polynomial function with respect to the internal coordinates that is explicitly constructed by the EPIC algorithm. Second, a measure m of the computational complexity of an algorithm can be described as "polynomial"5,31 if, for input size n, m grows no faster than nd for a fixed exponent d. In this case, one can construct a polynomial with respect to the size of the input that will be an upper bound on the computational cost, no matter how large the input is (n). For this purpose, we will consider either the time or the number of energy function calls as m—these two measures of computational complexity are related to each other by a constant factor in current protein design algorithms with continuous flexibility, since energy function calls generally dominate the cost of these algorithms. For example, precomputation of an EPIC representation must be performed only once for every pair of RCs at different residues, and thus the number of polynomial fits (and thus the total time for precomputation of EPIC fits) does not grow faster than the square of the number of residues being modeled. On the other hand, protein design itself has been shown to be NP-hard,21,22 which means no polynomial-time algorithm is likely to exist for it. In other words, for every polynomial p(nr) in the number nr of residues, there are protein design problems of nr residues that are not expected to be solvable in time p(nr). A problem only solvable by exponential-time algorithms—those that take time scaling as bn, where b is a constant and n is the size of the input—would typically be considered NP-hard.</p><p>Energy function calls are typically the bottleneck in protein design algorithms that model continuous flexibility. EPIC, however, ensures that the number of energy function calls in a protein design calculation is linear in the number of RC pairs, and thus polynomial in the size of the input (Fig. 2).</p><!><p>Consider two interacting residues i and j. Let us start with the "well-behaved" case where there exists a low-degree polynomial representation of a pairwise energy throughout the allowed ranges of both residues' internal coordinates (of the form in Eq. 1).</p><p>We employ the notation introduced in the DEEPer algorithm.11 Suppose we have RCs ir and js with pairwise energy E(ir, js, x), where x is the vector of internal coordinates (for example, dihedrals) affecting residues i and j when they have the amino-acid types corresponding to ir and js. Let E⊖(ir,is)=E(ir,js,x0(ir,js))=minxE(ir,js,x), where the minimum is taken with respect to the internal coordinates over their allowed ranges for the current RCs. This definition of E⊝ is consistent with iMinDEE8 and DEEPer.11 x0(ir, js) is the set of internal coordinates that minimizes the pairwise energy.</p><p>We seek a multivariate polynomial pir,js(x) such that</p><p> (1)E⊖(ir,js)+pir,js(x-x0(ir,js)) is a good approximation to E(ir, js, x). This multivariate polynomial is approximately a finite, low-degree Taylor expansion about the minimum. However, we use least-squares fits because we have found that they perform much better than Taylor expansions that are based on numerical derivatives. The fits are performed using a training set with ten times as many samples as there are parameters (polynomial coefficients) in the fit; the sampling procedure is described in Section 3.7. The fits are cross-validated with an independent set of samples (Section 3.6). The constraint p(0) = 0 is applied, so the real energy and the polynomial will agree exactly at the minimum-energy point. This constraint is easily implemented by not including a constant term in the polynomial, and reflects the need for the highest accuracy to be attained for the lowest-energy, and thus most biophysically reasonable, conformations. As a result of this constraint, all values of the polynomial on its domain will be nonnegative. Fitting begins with a multivariate quadratic fit and then moves up to higher degrees as needed (see Section 3.6). Since polynomials are linear with respect to their coefficients, the fitting is a linear least-squares problem.</p><p>This method can be generalized without modification to intra+shell energies as well as to any continuous degrees of freedom, such as newly modeled backbone perturbations11 or rigid-body motions of ligands. In every case, the number of variables for the polynomial will be the number of continuous degrees of freedom that define the conformation of the residue or residue pair of interest. For example, in a pairwise energy computation for a rotamer of lysine and a rotamer of valine with only sidechain flexibility, the polynomial will be in five variables (the four dihedrals of lysine and one dihedral of valine). The polynomial coefficients are real numbers.</p><p>Let r be an RC assignment, represented as a tuple of RCs with one RC for each residue. Let ir be the rotamer in r at residue i. To approximate the minimized energy of an enumerated conformation r, instead of minimizing the full energy</p><p> (2)Er(x)=∑iE(ir,x)+∑j<iE(ir,jr,x) with respect to the system's continuous degrees of freedom x, we simply minimize the polynomial approximation</p><p> (3)qr(x)=∑iE⊖(ir)+pir(x-x0(ir))+∑j<iE⊖(ir,jr)+pir,jr(x-x0(ir,jr)) with respect to x. These least-squares approximations achieve high accuracy for the low-energy wells of rotamers and local backbone motions, i.e., the portions of conformational space where both the continuous degrees of freedom and the energy are relatively close to the local minimum of the pairwise energy. Higher energies may also be found close in conformational space to the local minimum, but these energies indicate strained conformations unlikely to be seen in nature. Thus, for a "well-behaved" energy term whose energy is unstrained throughout the bounds on continuous degrees of freedom that define our current RCs, EPIC simply performs a least-squares fit of the energy, to represent it as a multivariate polynomial with respect to the continuous degrees of freedom.</p><p>Many RCs do however contain both regions with feasible energies and regions with higher energies that represent biophysically inaccessible conformations such as steric clashes. These RCs present difficulties for the basic least-squares fit, but the following algorithmic modification avoids this problem.</p><!><p>To handle RCs with high-energy regions, we note that we do not necessarily need the polynomial to be a good approximation for the energy throughout the entire region allowed by the box constraints. We merely require that Eq. (2) be a good approximation for Eq. (3) when used with biophysically feasible, minimized values of x. In particular, we can expect that the optimal, minimized structure has no clashes or other particularly large local strains. We have the advantage that while interaction energies in proteins can rise steeply towards infinity in the case of steric clashes, there is no physical phenomenon that will cause interaction energies to decrease steeply towards negative infinity. So energies are relatively well-behaved in low-energy regions. We can thus effectively partition the conformational space into relatively smooth, low-energy regions that we approximate accurately, and high-energy regions that we can rule out.</p><p>Let us denote the energy relative to the minimum as E′(ir, js, x − x0(ir, js)) = E(ir, js,, x) − E⊝(ir, js) in the pairwise case, or E′(ir, x − x0(ir)) = E(ir, x) − E⊝(ir) in the intra+shell case. Our requirements for a "good approximation" of the energy can be defined rigorously in terms of two upper bounds b1 and b2 that we place on E′. For each intra+shell or pairwise energy term, we estimate an upper bound b1 on E′ that we expect to hold for all minimized conformations that we want to output (the GMEC, or the lowest-energy c conformations if we are computing a c-conformation ensemble). The algorithm will be able to check if b1 is valid or not, so we can try again with a higher b1 if needed. Additionally, we need a second, possibly looser upper bound b2 on E′ that we are confident will be valid for all minimized conformations that we compute during our search, whether they turn out to be the GMEC or not. The value of b2 must be the same for all intra+shell and pairwise terms (b1 can be term-specific, though in practice a single value for b1 is convenient).</p><p>If EPIC is being used with the iMinDEE algorithm for conformational search,8 we can provably obtain the GMEC without considering any conformations whose energies E′ exceed the pruning interval, 8 an upper bound computed by iMinDEE for the difference between the lowest conformational energy lower bound (based on pairwise minimum energies) and the GMEC. Thus, when running iMinDEE, we can set b2 equal to the iMinDEE pruning interval. When b2 is set equal to the pruning interval, we know it is a valid upper bound on E′ for all minimized conformations computed during the search, and thus our GMEC calculation is provable. We can also do this when running DEEPer,11 which is essentially a backbone-flexible version of iMinDEE. For other algorithms we may want to set b2 based on knowledge of the system being designed—setting b2 = 2b1 is likely to be an acceptable heuristic.</p><p>Our polynomial only needs to be a good fit to E′ for values of the internal coordinates where E′ ≤ b1. For E′ > b1, we will require that the polynomial lie above b1. This will ensure that when we enumerate conformations in order of minimized energy computed using polynomials, as long as the thresholds b1 are chosen correctly, we will obtain non-clashing conformations before conformations with clashes, and these non-clashing conformations' energies will be accurately represented by the polynomial fits. Furthermore, we will require that for b1 < E′ < b2, the polynomial should be a lower bound on E′ (Fig. 3). This will ensure that regardless of what thresholds b1 were used, we never overestimate a conformational energy that is below the threshold b2, and thus never exclude it from the enumerated list of conformations. The requirement to be a lower bound is easy to satisfy, because clashing van der Waals interactions are very steep and thus will tend to rise much more quickly than the polynomial fits. Thus, when we perform polynomial fits using thresholds, we know we will be getting a gap-free list of conformations in order of energy. If the thresholds b1 were chosen to be too low, some higher-energy conformations with underestimated energy might be included as well, but these will be limited to minimized conformations containing energy terms with E′ > b1. This condition can be checked easily. If desired, the run can be redone with increased b1 thresholds to eliminate this error. Thus, the choice of b1 affects the ultimate speed of the algorithm but not its correctness.</p><p>For our experiments in this work (Section 4), we have set b1 to 10 kcal/mol. This threshold was found to be sufficient for all experiments described in this work, and most other EPIC designs that we have tried. Physically, any pair of residues whose interaction energy is 10 kcal/mol worse than the optimal interaction for its current RC pair is likely in a highly strained conformation such as a steric clash. Thus a design requiring b1 greater than 10 kcal/mol is likely to be biologically infeasible. For example, the protein is likely to unfold or undergo a large and unexpected structural change rather than suffer this local strain.</p><p>Let us use z to denote a vector in the domain of our polynomial fit p. p is considered a good representation of the energy if, for some small ε > 0, the following conditions are satisfied:</p><!><p>For z such that E′(ir, js, z) ≤ b1, |pir,js(z) − E′(ir, js, z)| < ε.</p><p>For z such that b1 < E′(ir, js, z) < b2, b1 − ε < pir,js(z) < E′(ir, js, z) + ε.</p><p>For z such that b2 ≤ E′(ir, js, z), b1 − ε < pir,js(z).</p><!><p>These conditions are illustrated in Fig. 3. They can be achieved using a modified least-squares fit, using special "one-sided" penalties to enforce the inequalities in conditions 2 and 3, along with usual (two-sided) least-squares penalties to enforce condition 1. The objective function is the sum of terms from each sample in the training set. For a sample z such that E′(ir, js, z) ≤ b1, the objective function term is (pir,js(z) − E′(ir, js, z))2 (as is typical for least squares). A term of this form is also used if the lower-bounding condition is violated, i.e., if E′ < b2 but p > E′. Otherwise, the objective function term for z is (pir,js(z) − b1)2 for pir,js(z) < b1, and 0 for pir,js(z) ≥ b1.</p><p>If the modified least-squares method is applied to a set of samples that mostly have E′ > b1, then overfitting to the few points with E′ < b1 may occur no matter how many samples there are. As an extreme case, if all samples have E′ > b2, then almost any polynomial with very large values throughout its domain will give a 0 value for the objective function, but this may still provide a very poor description of the energy landscape. To avoid this situation, when a test set of n samples is being drawn and n/2 samples with E′ > b1 have been drawn already, then if more samples come up with E′ > b1, they are redrawn to ensure that a sufficient number of samples with E′ ≤ b1 is available (Section 3.7). Minimization-aware dead-end elimination pruning8 (both singles and pairs pruning) is performed before computation of the polynomial fits, since the pruned rotamers and pairs won't be needed during enumeration. This pruning usually eliminates the clashing rotamers and pairs, leaving rotamers and pairs that are well suited for simple polynomial representations.</p><p>This objective function can be optimized efficiently because it is convex with respect to the polynomial coefficients (see Section 3.4). But we found general-purpose convex minimizers to be rather time-consuming for the higher-order fits. To address this, the algorithm described in Section 3.4 was developed. It exploits specific properties of the objective function to obtain a more efficient and reliable fit than a general-purpose convex minimizer would be likely to obtain.</p><!><p>The following algorithm performs a modified least-squares fit, providing a useful polynomial for energy terms that include both low-energy regions, where an accurate polynomial representation of the energy surface is required, and high-energy regions that we must exclude from our search.</p><p>Let us represent our polynomial fit p(z) as p · y(z), where p is the polynomial's vector of coefficients, y(z) is the corresponding vector of monomials built from the degree-of-freedom values z, and · is the standard inner product. For example, if z consists of the two dihedrals z1 and z2 and we are performing a quadratic fit, then y(z) will have the elements 1, z1, z2, z12,z22, and z1z2. For each sample s in our training set of samples (see Section 3.7), let zs be the vector of degree-of-freedom values, and let ys = y(zs) be the corresponding vector of monomials. Let Es′ be the energy for the sample, where the minimum-energy point is defined to have zero energy. Then, a modified least squares fit consists of minimizing the objective function f to obtain best-fit polynomial coefficients pb: (4)pb=argminp∑s∣Es′≤b1(Es′-p·ys)2+∑s∣p·ys≥Es′,b1<Es′<b2(Es′-p·ys)2+∑s∣Es′>b1,p·ys<b1(b1-p·ys)2 where { s∣Es′≤b1} denotes the set of samples whose energies are less than or equal to b1. If we define P1 to be the set of sample points such that either</p><p> (5)Es′≤b1 or</p><p> (6)p·ys≥Es′,b1<Es′<b2 and we define P2 to be the set of sample points such that</p><p> (7)Es′>b1,p·ys<b1 then our objective function f becomes</p><p>Thus, if we know P1 and P2, minimizing the objective function is a basic least squares problem and can be solved analytically. Like basic least squares, this algorithm operates on a single "training" set of samples and provably minimizes the objective function (i.e., the error) for that training set.</p><p>We can show the objective function is convex with respect to p by noting that the contribution from each sample s is a function of the single linear combination u = p · ys of the elements of p. This contribution depends on Es′, but it is always convex (and piecewise quadratic). If Es′≤b1, the contribution is just the parabola (Es′-u)2. If b1<Es′<b2, it's the "truncated" or "flat-bottomed" parabola given by (b1 − u)2 for u ≤ b1, 0 for b1≤u≤Es′, and (Es′-u)2 for u≥Es′. Otherwise (if b2≤Es′), the contribution is the "one-sided" parabola given by (b1 − u)2 for u < b1 and 0 otherwise. Hence, the objective function is a sum of convex functions, making it convex itself. Thus, minimizing the objective function to find p is tractable, with any local minimum being the global minimum. As a result, we know that if for any sets of samples P1 and P2 we have coefficients p that minimize Eq. (8) and satisfy the conditions Eq. (5–7), then the coefficients p are globally optimal.</p><p>The algorithm finds P1 and P2 iteratively. As an initial guess, P1 can be initialized to s such that Es′≤b1, and P2 to be empty. (This corresponds to assuming that the one-sided restraints can all be satisfied perfectly.) This is followed by performing the basic least-squares computation of minimizing Eq. (8), which returns coefficients p, and recalculating P1 and P2 from p using the conditions Eq. (5–7). This procedure is then repeated using the new P1 and P2 until a self-consistent solution is found. Generally, only a small minority of the samples will be moved in and out of the least-squares problem at each iteration, so the least-squares matrix can be updated quickly at each step—this is useful because forming this matrix is the bottleneck. Typically, only a few iterations are needed.</p><p>This algorithm is actually a special case of Newton's method, because its estimate for the objective-function minimum at each iteration is the minimum of the local quadratic Taylor expansion of the objective function. This minimum can be found analytically because the local expansion is convex.</p><p>In our implementation of this algorithm, by far the bulk of its time cost is spent in forming the matrix for the first basic least-squares fit (with initial P1 and P2). The subsequent fits are much faster because they are only sparse updates. Thus, the modified least-squares fitting is only negligibly more expensive than the first basic least-squares fitting.</p><!><p>SAPE is a method to reduce the degrees of polynomials needed by EPIC by including some non-polynomial terms in the representation of the energy.</p><p>The need for higher-order polynomial fits is driven by large values of higher derivatives. These values are contributed primarily by a small number of van der Waals (vdW) terms between pairs of atoms that are very near each other. It is possible to obtain substantial time and memory savings by evaluating these terms explicitly and fitting the rest of the energy function to a polynomial. To select atom pairs whose vdW terms are to be evaluated explicitly, a cutoff distance (3 or 4 Å; see Section 3.6) is chosen. Then, an atom pair's vdW terms are evaluated explicitly if and only if the atoms can be found within that distance of each other within the bounds on internal coordinates for the given residue conformations. These terms are not polynomials in the degrees of freedom because they are inverse powers of distances between atoms, and the atom coordinates themselves are in general not polynomial functions of the degrees of freedom. For example, the expressions for atom coordinates in a sidechain in terms of the sidechain dihedral angles will include sines and cosines of those angles.</p><p>Once we decide to evaluate vdW terms for a given pair of atoms, it costs negligible extra time and memory to also calculate the electrostatic interaction between these atoms (since we already have the distance between the atoms).</p><!><p>We will now describe the methods used to choose polynomial degrees for EPIC fits and ensure that fits of sufficient accuracy are obtained.</p><p>Fit accuracy is checked and controlled using cross-validation. For cross-validation purposes, a mean-square error is computed, with absolute error used below E′ =1 kcal/mol and relative error above. This can be seen as a weighting of the error terms: the weight is 1 for E′ ≤ 1 and 1/min(E′, b1) for E′ ≥ 1 (this levels off at b1 to avoid excessive underweighting of the one-sided constraints). These weights, which are continuous with respect to E′, are also used during the least-squares fitting.</p><p>Cross-validation is used to select the degree of the polynomial that is fit. Low-degree polynomials save time and memory both during the A*/enumeration step and during the precomputation step, but may not provide a sufficiently good representation. Hence, we proceed through a sequence of increasingly expensive fits (Fig. 1A), and each time a fit is completed, it is cross-validated with an independently drawn set of samples. Like the training set, this cross-validation sample set has ten times as many samples as fit parameters. If the mean-square error is below a specified threshold, the fit is stored, and if it is above, we proceed to the next method. The default threshold value is set to 10−4. However, limited investigation suggests larger thresholds still tend to keep the errors in conformations' minimized energies small compared to thermal energy, and thus are likely acceptable as well. It is also useful to avoid doing fits with very large number of parameters, as these have enormous time and memory costs both in the enumeration and precomputation steps. Thus, OSPREY is currently set to refuse to do fits with over 2000 parameters—this way, computations that would have prohibitive time costs may still be satisfactorily completed with a slightly higher error threshold than usual.</p><p>Some of the fits use lower-degree terms for all degrees of freedom and higher-order terms for selected degrees of freedom. These selected degrees of freedom are eigenvectors vk of the Hessian from a modified least-squares quadratic fit (step 1 in the list of steps below). Letting λk be the eigenvalue corresponding to vk, we define</p><p> (9)Dq={vk|∣λk∣≥maxi∣λi∣q} for q > 0. Let us define fn to be a polynomial fit of total degree d (e.g., f2 is a quadratic fit); fd(Dq) to be a fit to a polynomial of total degree d in all degrees of freedom plus terms of total degree d + 1 and d + 2 in the degrees of freedom in Dq; and s(n, c) to be a polynomial fit of total degree d plus SAPE with a cutoff of c Å. Fits were tried in the following order: f2, s(2, 3), f2(D10), f2(D100), f4, s(4, 4), f4(D10), f4(D100), f6, and s(6, 4).</p><p>The Stone-Weierstrass theorem32 guarantees that a sufficiently high-degree polynomial can approximate any function on any closed and bounded portion of Cartesian space to any desired accuracy. In other words, it guarantees that any energy function can be represented by EPIC to arbitrary accuracy if we allow sufficiently high-degree polynomials. The basis of Bernstein polynomials can be used to construct such approximations with guaranteed convergence to any function.33 However, for the purpose of energy representation for protein design, modified least squares is likely to provide good approximations using much lower-degree polynomials than we would obtain using the Bernstein basis, because we do not need close approximations of the high energy in clashing regions. In these regions, we only need a reasonable lower bound that is much higher than the rotameric wells. This strategy keeps the polynomial degrees low enough to be practical.</p><!><p>Training and validation sets for EPIC fits consist of sample conformations of the residue(s) involved, specified as vectors of internal coordinates, drawn from throughout the allowed region of conformational space.</p><p>By default, samples for both training and validation sets were sampled uniformly (i.e., each degree of freedom was sampled uniformly and independently from the interval corresponding to the current rotamer or RC). Ten samples were always used in each of these training and validation sets for each parameter in a fit. However, if most of the samples corresponded to energies above the threshold b1 (see Section 3.3), then overfitting could result, because for such samples there are infinitely many polynomial values that yield zero error. To avoid this, we need sufficient samples from the set B of conformations with energies below b1; B is the set of conformations where the polynomial needs to be quantitatively accurate. We ensure sufficient samples from B by rejecting samples outside B whenever we desire n samples in total and we already have n/2 samples outside B, and thus drawing the rest of our samples uniformly from B by rejection sampling. If 10,000 samples are rejected consecutively, indicating that B is too small for efficient rejection sampling, then the Metropolis algorithm34 is used to sample from B.</p><p>We have confidence in the parameters obtained by fitting to the training samples for three reasons. First, a useful measure of the accuracy of a polynomial approximation to the energy surface is that there is a low probability that any region of the energy surface deviates significantly from the polynomial approximation (except for high-energy regions approximated by similarly high values of the polynomial). Since our cross-validation of each polynomial fit uses a large number of independent samples—ten times the number of parameters—we are left with a very low chance that our cross-validation samples will miss any such regions. Thus an insufficiently accurate polynomial surface will be detected upon cross-validation and remedied by an increase in polynomial degree. Second, errors in the minimized energies obtained using polynomial approximations are consistently low, as shown in our computational experiments (Table 1). Third, we expect the energy function to be relatively smooth in the vicinity of a minimum, since the gradient must be zero at the minimum, and thus we expect a polynomial of relatively low order (e.g., the Taylor series of the energy) to yield a good approximation in the vicinity of a minimum.</p><!><p>Once the polynomials are computed, they can be used in protein design algorithms wherever the energy function would ordinarily be called. The GMEC will simply be the set of rotamers for which the minimized value of Eq. (3) with respect to x has the lowest possible value.</p><p>The simplest method to provably find the GMEC using EPIC is to use a protein design algorithm that enumerates conformations in order of a lower bound, and then instead of minimizing the full energy (Eq. 2), merely minimizing the polynomial-based energy (Eq. 3) to compute the energy for each enumerated conformation (Fig. 2). For example, iMinDEE/A*8 can be used for this enumeration process, and we use this algorithm in our computational experiments (Section 3.10).</p><p>EPIC can also be applied in free energy calculations using the K* algorithm,7,12 which approximates binding constants as ratios of partition functions computed from low-energy conformations enumerated by A*. During these calculations, one can simply use the polynomials instead of the energy function to compute the partition function, given the enumerated RC assignments. This method gives a constant-time speedup, determined by the ratio of time to evaluate the energy function versus the EPIC energy.</p><p>An additional speedup is possible for branch-and-bound protein design algorithms (e.g., A*14,15) that use a tree structure for conformational search. These algorithms build nodes that each represent a subset of conformational space and are scored using a lower bound on the conformational energies in that space. In each node's conformational space, some residues are restricted to a single RC; these RCs are referred to as assigned to their respective residues. At each level of the tree, an RC is assigned to one more residue. One can use the EPIC polynomials to improve the lower-bound energy for each of these nodes. At each node, we need to compute a lower bound L for the conformational energy qr, which is defined in Eq. (3): that is, we compute L such that</p><p> (10)L≤qr(x)=∑iE⊖(ir)+pir(x-x0(ir))+∑j<iE⊖(ir,jr)+pir,jr(x-x0(ir,jr)) for all RC assignments r and all degree-of-freedom values x that are part of the node's conformational space. If r is known (i.e., if RCs are fully assigned at all residue positions), then a tight lower bound can be computed trivially by local minimization with respect to x. Otherwise, we let qr(x) = E⊝(r) + Ep(r, x), where Ep consists only of EPIC polynomials: (11)E⊖(r)=∑iE⊖(ir)+∑j<iE⊖(ir,jr) (12)Ep(r,x)=∑ipir(x-x0(ir))+∑j<ipir,jr(x-x0(ir,jr))</p><p>Now, if we compute lower bounds L⊝ and Lp such that L⊝ ≤ E⊝(r) and Lp ≤ Ep for all r, x in our conformational space, then L = L⊝ + Lp will satisfy Eq. (10), giving us a valid lower bound. Computation of L⊝ has been described previously, because lower bounds of this form are computed in iMinDEE8 and DEEPer.11 To compute Lp, we use the fact that EPIC polynomials are always nonnegative; thus, for any r and x and any subset S of the residues we are modeling,</p><p>If we let S be the set of residues with fully assigned RCs, then there is only one possible RC ir for each residue i ∈ S, and so we can find the minimum of Eq. (13),</p><p> (14)minr,x(∑i∈Spir(x-x0(ir))+∑j∈S,j<ipir,jr(x-x0(ir,jr))) exactly by local minimization with respect to x. Eq. (14) is a lower bound on Ep(r, x), and thus we set Lp equal to it, giving us a score for our node. We note that Lp is strictly nonnegative, because Eq. (13) and thus Eq. (14) are always nonnegative.</p><p>Because this continuous minimization with respect to x is more expensive and has to be performed separately at each node, it is evaluated in a lazy31 fashion in our A* implementation. Nodes are assigned the traditional, discrete lower bound L⊝ when they are generated; this bound is fast to compute. The A* priority queue contains nodes both with and without the polynomial contribution Lp included. When a new node is popped from the queue, we check if Lp is present or not. If it is, we expand the node, and if it is not, we compute Lp and insert the node back in the priority queue. This ensures that nodes come off the priority queue in order of their complete lower bound L⊝ + Lp, as is necessary for A* to function correctly. However, it also ensures that we do not waste time computing Lp for nodes whose L⊝ is high enough to preclude expansion. This method gives a combinatorial speedup, since a high polynomial contribution for a partial conformation can effectively prune an entire branch of the A* tree. In practice, though, the large constant speedup from EPIC minimization of fully assigned conformations tends to be more significant (Section 4.1, Table 2).</p><!><p>The speedup due to EPIC can be explained in terms of the asymptotic costs of EPIC polynomial evaluations compared to direct energy function calls. The cost of evaluating an EPIC polynomial scales as the number of terms in the polynomial. This cost is itself a polynomial (usually quadratic) in the number of internal coordinates of the residue pair (or single residue, in the intra+shell case) of interest. By contrast, the cost of evaluating a molecular mechanics-based energy function is generally quadratic in the number of atoms involved, since distances between all pairs of atoms need to be considered. EPIC achieves a marked speedup because most residues have far more atoms than significantly flexible internal coordinates. For example, most protein residues have two or fewer sidechain dihedrals, but over ten atoms. The remaining internal coordinates—bond length, angles, etc.—are relatively inflexible. Polynomial evaluations are also performed entirely by addition and multiplication, which are much faster than the more complicated elementary operations (trigonometric functions, square roots, etc.) needed to evaluate molecular mechanics energy terms.</p><p>When quantum-mechanical energy functions are introduced, all the electrons must be accounted for explicitly, and even fairly approximate quantum-chemical methods have time costs that are higher-order polynomials with respect to the number of electrons. For example, any method that accounts for repulsions between all atomic orbitals (e.g., Hartree-Fock and all post-Hartree-Fock methods) must calculate repulsion integrals for all quadruples of atomic orbitals, and there are at least as many atomic orbitals as electrons. And there are far more electrons than there are atoms, and far more atoms than internal coordinates, giving EPIC an extreme performance advantage. Yet EPIC can represent the same energy surface to a high degree of accuracy, once the EPIC polynomials have been precomputed.</p><p>Whether EPIC is used or not, these types of pairwise energy evaluations must be performed for every pair of residues in a design system. In general, this means the number of pairwise energy evaluations needed is quadratic with respect to the number of residues in the system. This number can be reduced if a cutoff is applied to remove interactions between distant residues. However, this speedup applies equally for EPIC and non-EPIC calculations.</p><!><p>Protein design calculations were performed in OSPREY1,3,7,23,24 with and without EPIC to investigate (a) what previously intractable systems become newly tractable with EPIC, (b) what speedups EPIC brings to conformational enumeration for previously tractable systems, and (c) what types of polynomial representations are needed for these purposes. EPIC runs were performed with SAPE and with conformational minimization for partially assigned conformations during A* search, and for comparison, runs with either one of these features omitted were also performed.</p><p>Times were compared for the A* search, including conformation enumeration and minimization, because this is the portion of the design that is not guaranteed to complete in polynomial time and thus is the bottleneck. As part of the EPIC runs, GMEC energies were also computed using the regular energy function to compare to the EPIC results, and the ratio of minimization times with and without EPIC was computed. For runs with multiple conformations very close in energy to each other (within the error range of EPIC, typically <0.1 kcal/mol), the time ratios were averaged.</p><p>All minimizations were performed using a cyclic coordinate descent minimizer, which is now included in OSPREY. Default OSPREY energy function settings were used where applicable: AMBER with EEF1 solvation and a distance-dependent dielectric constant of 6. Rotamers were determined using the Penultimate rotamer library.13</p><p>Test systems were chosen to evaluate both partition function and GMEC calculations, and to include all three types of continuous degrees of freedom used in OSPREY: sidechain dihedrals, backbone perturbation (shear and backrub) parameters,11 and rigid-body rotations and translations of strands. Some of the tests are intended to be within the scope of previous methods, allowing a quantitative comparison of running times, while others are intended to show EPIC can compute previously intractable GMECs and partition functions with provable accuracy.</p><p>For GMEC calculations (Table 1), the first set of systems used was taken from Gainza, Roberts, and Donald,8 and featured only sidechain dihedral flexibility. The structures for these correspond to PDB codes 2o9s, 2qsk, 2rh2, 2ril, and 3g36. The second set of systems was taken from Hallen, Keedy, and Donald,11 and included both sidechain and backbone flexibility. The structures' PDB codes were 1aho, 1c75, 1cc8, 1f94, 1fk5, 1i27, 1iqz, 1jhg, 1l6w, 1l7a, 1l7l, 1l7m, 1l8n, 1l9l, 1l9x, 1lb3, 1m1q, and 1mwq. Three variants of the 1aho system with more residues were tried as well. Finally, a GMEC calculation was performed for the complex of the HIV surface protein gp120 with the broadly neutralizing antibody NIH45-46 (PDB code 3u7y35).</p><p>To investigate the application of EPIC to partition function calculations (Table 2), we first chose systems with only sidechain dihedral flexibility from Gainza, Roberts, and Donald8 and calculated a partition function for the unliganded protein, with wild-type amino acids at all residue positions, to within 97% guaranteed accuracy. Partition function calculations such as these are the key operation in K*7,12 calculations. The structures for these correspond to PDB codes 2cs7, 2o9s, 2p5k, 2qsk, 2r2z, 2rh2, 2ril, 2wj5, 2zxy, 3a38, 3dnj, 3fgv, 3fil, 3g21, 3g36, 3hfo, and 3i2z. Furthermore, a K* run is presented for trypsin with a small-molecule inhibitor (PDB code 3pwc); the run is tractable with EPIC but fails to finish without it. Unlike the calculations for the other, monomeric structures, the K* run for trypsin involves calculation of three partition functions: one for the protein, one for the ligand, and one for the complex.</p><p>Each design was allowed 17 days of total runtime, after which those that had not finished were deemed to have exceeded the time limit and were terminated. A* times with EPIC ranged from 0.7 seconds to 4 days, and the speedups due to EPIC are shown in Tables 1 and 2.</p><p>In these experiments, fitting was performed without parallelization. However, the computation of the EPIC polynomial for each pair of RCs is an independent operation, so each can be done in parallel, meaning that parallelization to p processors will give a p-fold speedup as long as p does not approach the number of RC pairs. OSPREY currently supports computation of each residue pair in parallel, so the speedup holds as long as p does not approach the number of residue pairs. In practice however, for large systems, pruning and A* take longer than the polynomial fitting, so this parallelization may not be necessary. Additionally, once the EPIC fits have been computed for a system, there may be a large number of computations that can be performed using it—calculation of partition functions for many sequences, computation of GMECs for various subsets of the sequence space, etc. These extensive reuses of the fits may be especially desirable when designing a library of sequences for experimental testing—if one performs various optimizations with different assumptions and tests top sequences from each optimization, the results will be more robust to errors in the assumptions.</p><p>To investigate the ability of EPIC to represent quantum-mechanical energy functions, EPIC calculations were also performed on the aspartame dipeptide (extracted from PDB code 1a8j36) with the energies for EPIC samples evaluated using NWChem37 instead of using OSPREY's usual energy function. Calculations were performed at the SCF level of theory with STO-3G and with 6-31G** basis sets, and also at the MP2 level of theory with a STO-3G basis set.20 For each rotamer of each residue, dihedrals were sampled within the allowed range for the rotamer and the total energy of the dipeptide was fit to a polynomial.</p><!><p>For this study, EPIC was implemented in the context of the OSPREY protein design package OSPREY1,3,7,23,24 to run along with the algorithms (iMinDEE, DEEPer, and K*) and pairwise energy functions already implemented in OSPREY. However, EPIC would enable some other capabilities in different implementations.</p><p>First, EPIC can be applied in the context of other protein design algorithms. For example, one can apply it an iterative algorithm like FASTER38 or Monte Carlo34 that tries to find a suitably low-energy conformation by accepting or rejecting rotamer changes based on the energies of conformations with these changes. Whenever the energy is needed for a rotamer assignment, the EPIC energy for the protein can be locally minimized starting at the ideal internal coordinate values for that rotamer assignment. In this case, the matrix of EPIC polynomials substitutes directly for the matrix of pairwise rotamer energies commonly used to calculate conformational energies for these algorithms in the absence of continuous flexibility. EPIC could even be used for molecular dynamics, since most residue pairs in a molecular dynamics trajectory4 will spend most of their time in fairly relaxed conformations—the region of conformational space modeled by EPIC. In all these cases, EPIC energy evaluations would be markedly faster than regular energy function calls, particularly for expensive energy functions. Thus, EPIC would provide a substantial speedup for any algorithm whose bottleneck is energy function calls.</p><p>For docking algorithms39 that require sidechain optimization or local backbone optimization, EPIC can be used both for conformation scoring and for conformational optimization using any of the above algorithms.</p><!><p>EPIC was implemented in this work to handle pairwise energy functions (in the sense of a sum of 1-body and 2-body energies with no terms dependent on three or more residues' degrees of freedom), because these are currently typical for protein design and include the AMBER, CHARMM, and EEF1 energy functions we use in OSPREY. However, the true energy of proteins is not exactly pairwise decomposable, and EPIC could easily accommodate higher-order terms. EPIC simply requires that each energy term correspond to a set D of degrees of freedom, constrained to a region in which they are relatively well-behaved (e.g., dihedrals at each residue constrained to a single rotamer); we can sample D subject to the constraints and then fit the energies as a polynomial function with domain D. Thus, for any set R of more than two interacting residues, for each RC assignment to those residues, we can fit an EPIC polynomial, and thus describe the energy terms for R.</p><p>In practice, the number of sets of residues that can interact significantly is quite limited, because residues typically must be physically near each other to have significant higher-than-pairwise interactions. For example, if we have a Ramachandran-based potential, its terms each depend on the ψ and ϕ backbone dihedrals of a certain residue r, and thus depend on the conformations of the three residues r − 1, r, and r + 1. Likewise, the conformation of a residue i can induce polarization effects in a nearby residue j that will affect the interactions of j with another residue k, and this effect can be quantified using quantum chemistry, but i and j have to be physically very close to each other (≪1 nm) for this effect to be significant (and j and k have to be fairly close too—probably subnanometer as well, since the potential of an induced dipole falls off faster than 1/d2 with distance d). Hence, EPIC can be used to model any realistic energy function, by accounting for all sets of residues with significant energetic interactions.</p><!><p>Computational experiments were performed to measure what kinds of polynomial fits are necessary to accurately model different proteins with different degrees of freedom and energy functions, and what speedups EPIC brings to DEE/A* and K* calculations. The results demonstrate that EPIC brings a substantial speedup to design calculations when proteins are modeled as in previous OSPREY designs.1,3,8,11,23 They also show that EPIC efficiently represents energies calculated by quantum chemistry, and is a potentially decisive tool for using both realistic, continuous flexibility and quantum-mechanical energy functions in protein design.</p><!><p>First, computational experiments were performed to compare GMEC search with and without EPIC, as described in Section 3.10. Key portions of the design calculation were timed with and without EPIC to determine the speedup for these portions (Table 1). On average, minimization of fully enumerated conformations was 79-fold faster using EPIC than with traditional energy function calls. Overall A* speedups due to EPIC averaged 167-fold (Fig. 5A). The overall A* speedup is likely greater than the minimization speedup because of the way OSPREY's standard energy function is implemented. Each time the energy function is run on a new sequence, setup time (e.g., initialization of the energy function) is required to identify electrostatic, van der Waals, and solvation terms that will be necessary for that sequence. This setup time is eliminated by EPIC, and is not counted as part of the minimization time here, but it may be performed an exponential number of times without EPIC, since minimizations may be required for an exponential number of sequences. Runs that did not finish without EPIC are not included in these averages. 85% of the fits in these experiments were quadratic, with no SAPE needed (Figure 5B). GMECs from EPIC runs showed good agreement between energies from minimization of EPIC energies and energies from the actual energy function. The average energy difference was 0.04 kcal/mol, which is less than one-tenth of thermal energy at room temperature (0.592 kcal/mol, calculated as the universal gas constant times a room temperature of 298° K) and thus functionally insignificant.</p><p>Five of the 27 systems finished only with EPIC, demonstrating that EPIC allows design of larger and more diverse systems than were previously designable. For example, a redesign of the complex of HIV surface protein gp120 with the antibody NIH45-46 did not finish when run without EPIC, but finished with EPIC using about a day of A* time (Fig. 4). This redesign allowed the mutation of 16 residues all over the gp120 surface in the interface—five in the D-loop of gp120,40 which is central to the interaction with NIH45-46, and the other 11 scattered through other parts of the interface in various types of secondary structure. Redesigns of the gp120 surface to achieve specific binding to particular antibodies has been instrumental in the development of probes to isolate these antibodies from sera.28 Redesign of the antibody surface of a gp120-antibody complex has also been effective in optimizing antibody affinity,25 which is useful for passive immunization and immunogen design. Interestingly, the redesign of the NIH45-46 complex yielded 12 top conformations within 0.06 kcal/mol of each other—two from the top sequence and ten from a double mutant. This high density of favorable conformations suggests the complex is entropically favored, a result consistent with the observed high affinity of NIH45-46 for gp120, attained through extensive affinity maturation of the antibody.</p><p>Two variations of EPIC were also tried for these systems. It was found that minimization of partial conformations during A* (Section 3.8) provides a speedup (2.3-fold on average), though it is not nearly as great as the speedup from faster minimization of fully assigned conformations (Fig. 5). Furthermore, EPIC without SAPE was often effective; however, under some circumstances it was unable to provide accurate fits (as usual, trying only the polynomial degrees described in Section 3.6). In systems where EPIC without SAPE was effective, it averaged insignificantly (1.1-fold) slower than EPIC with SAPE for A*. However, there were four systems that required SAPE to give accurate results (out of 27; Table 1), including the HIV gp120 complex with antibody NIH45-46 (Fig. 4). There were also four systems that exceeded the time limit during fitting. These limitations on EPIC without SAPE do not indicate a fundamental theoretical barrier, because in principle the Stone-Weierstrass theorem guarantees an accurate fit if the polynomial degree is sufficiently increased (see Section 3.6). However, they do indicate that EPIC without SAPE may sometimes require polynomial degrees that are prohibitively time-consuming for typical protein designs, and/or a higher numerical precision than the double precision efficiently supported in Java and thus used in OSPREY.</p><p>Experiments were also performed to compare partition function calculations with and without EPIC (Table 2). To obtain a provably good approximation to the partition function,7,12 many more conformations must be enumerated and minimized than for GMEC calculations. As a result, marked speedups were achieved by EPIC (Fig. 5C). Out of the 19 systems for which EPIC finished, only three finished without EPIC (average speedup 2000-fold). The speedup in EPIC designs from minimization of partial conformations was only modest (1.4-fold).</p><p>With an A* speedup of 2–3 orders of magnitude, designs that would previously take years can be performed with EPIC in days. This will allow many designs that would otherwise be considered intractable to be completed using EPIC.</p><!><p>In addition to classical mechanics-based energy functions, we also used EPIC to fit conformational energies of the aspartame dipeptide calculated using quantum-mechanical models of electronic structure. EPIC fits for aspartame showed that quantum-mechanical energies and AMBER and EEF1 energies can be represented by polynomials of very similar degree, i.e., energy surfaces from quantum chemistry are just as polynomial-like as energy surfaces from molecular mechanics. SAPE was not found to significantly increase the accuracy of the fits, and thus were not included, though it is likely that reparameterized van der Waals and/or electrostatic terms (or other specially fit functions of the atom-pair distance) would be able to improve fit quality. This discrepancy indicates that the atom-pair energies used in SAPE are a poor approximation to the interactions between the same atom pairs predicted by quantum mechanics, and thus that energies returned by quantum-mechanical and molecular-mechanics methods are substantively different.</p><p>For Phe 2 of aspartame, the same types of polynomial fits were needed for Hartree-Fock with a STO-3G basis set, Hartree-Fock with a 6-31G** basis set, and the usual AMBER/EEF1 energy function (Fig. 6). These were quadratic fits for three rotamers and a quadratic fit plus quartic fits on D10 for the fourth. MP2 with a STO-3G basis set also required quadratic fits for the first three rotamers, and required a quadratic fit plus quartic fits on D100 for the fourth.</p><p>For Asp 1, AMBER/EEF1 required quadratic terms plus quartic terms on D10 for three rotamers and on D100 for two. Both Hartree-Fock and MP2 with a STO-3G basis set required slightly simpler fits: only quadratic for one rotamer, quadratic plus quartic terms on D10 for three, and quadratic plus quartic terms on D100 for one.</p><p>These results show that quantum and polarization-type effects can be represented effectively by polynomial fits, and the polynomial degrees needed are essentially the same as for AMBER/EEF1.</p><!><p>EPIC eliminates the current bottleneck in minimization-aware protein design by performing energy function calls only in a precomputation step. It thus opens several avenues for more accurate and efficient design calculations. More residues can now be mutated, and more ligands can be tested. Additional continuous conformational degrees of freedom (e.g., in the backbone) can be modeled, and minimization can be performed over a greater range for each degree of freedom when appropriate. In this sense, EPIC helps protein design algorithms emulate the extensive continuous flexibility of molecular dynamics algorithms, while searching an exponentially large sequence space that would be intractable for molecular dynamics-based design. Furthermore, more accurate but (previously) slower energy functions can be incorporated without any asymptotic increase in computation time.</p><p>The polynomial representation of energy provided by EPIC could also allow dedicated algorithms for polynomials to be used in energy calculations. For example, since exact derivatives of polynomials are trivial to compute, EPIC is very amenable to rapid calculation of energy derivatives with respect to internal coordinates, which are used in many minimization algorithms.41,42 Note that the gradient of the EPIC polynomials may not approximate the gradient of the energy (i.e., forces) to the same degree of accuracy as the polynomials approximate the energy, because the polynomials are fit to the energies rather than the forces. Thus, differentiation of the fit polynomial may amplify noise. However, when we are numerically minimizing the polynomial approximation to the energy, we require derivatives of the polynomials themselves.</p><p>EPIC fits are tractable, accurate approximations that provide a new understanding of the energy landscape of proteins in the vicinity of ideal rotamers, and more generally in the vicinity of energy minima. This is useful because the high dimensionality of conformational space makes direct visualization difficult.</p><p>By enabling better modeling both of conformational space and of conformational energies, EPIC moves us closer to the goal of algorithms that can produce reliable predictions for our biomedically and biologically important protein and drug design problems. EPIC is thus offered to the protein design community both as an immediate speedup in designs and as an enabling technology for future improvements.</p><!><p>Our implementation of EPIC, as part of the OSPREY1,3,7,23,24 open-source protein design software package, is available for free download at http://www.cs.duke.edu/donaldlab/osprey.php.</p>

PubMed Author Manuscript

Hyperoxia-mediated transcriptional activation of cytochrome P4501A1 (CYP1A1) and decreased susceptibility to oxygen-mediated lung injury in newborn mice

Hyperoxia contributes to the development of bronchopulmonary dysplasia (BPD) in premature infants. In this study, we tested the hypothesis that newborn transgenic mice carrying the human CYP1A1-Luc promoter will display transcriptional activation of the human CYP1A1 promoter in vivo upon exposure to hyperoxia, and that these mice will be less susceptible to hyperoxic lung injury and alveolar simpliFIcation than similarly exposed wild type (WT) mice. Newborn WT (CD-1) or transgenic mice carrying a 13.2 kb human CYP1A1 promoter and the luciferase (Luc) reporter gene (CYP1A1-luc) were maintained in room air or exposed to hyperoxia (85% O2) for 7\xe2\x80\x9314 days. Hyperoxia exposure of CYP1A1-Luc mice for 7 and 14 days resulted in 4- and 30-fold increases, respectively, in hepatic Luc (CYP1A1) expression, compared to room air controls. In lung, hyperoxia caused a 2-fold induction of reporter Luc at 7 days, but the induction declined after 14 days. The newborn CYP1A1-Luc mice were less susceptible to lung injury and alveolar simplification than similarly exposed wild type (WT) CD-1 mice. Also, the CYP1A1-Luc mice showed increased levels of hepatic and pulmonary CYP1A1 expression and hepatic CYP1A2 activity after hyperoxia exposure. Hyperoxia also increased NADP(H) quinone reductase (NQO1) pulmonary gene expression in both CD-1 and CYP1A1-Luc mice at both time points, but this was more pronounced in the latter at 14 days. Our results support the hypothesis that hyperoxia activates the human CYP1A1 promoter in newborn mice, and that increased endogenous expression of CYP1A1 and NADP(H) quinone reductase (NQO1) contributes to the decreased susceptibilities to hyperoxic lung injury in the transgenic animals. This is the first report providing evidence of hyperoxia-mediated transcriptional activation of the human CYP1A1 promoter in newborn mice, and this in conjunction with decreased lung injury, suggests that these phenomena have important implications for BPD.

hyperoxia-mediated_transcriptional_activation_of_cytochrome_p4501a1_(cyp1a1)_and_decreased_susceptib

1. Introduction<!>2.1. Animal studies and treatment protocol<!>2.2. Analysis of lung morphometry<!>2.3. Quantitative real time PCR assays<!>2.4. Enzyme assays<!>2.5. Electrophoresis and western blotting<!>2.6. Statistical analyses<!>3.1. Effect of hyperoxia on transcriptional activation of human CYP1A1 and endogenous mouse CYP1A1 gene expression<!>3.2. Effect of hyperoxia on lung injury and alveolar simplification in newborn WT (CD-1) and human CYP1A1-Luc mice<!>3.3. Effect of hyperoxia on endogenous CYP1A1/1A2 activities and contents<!>3.4. Effect of hyperoxia on NQO1 mRNA and protein expression<!>4. Discussion

<p>Hyperoxia is commonly encountered in premature infants with pulmonary insufficiency [1,2]. Studies link hyperoxia exposure to the development of bronchopulmonary dysplasia (BPD), which is a major risk factor for mortality and morbidity in premature infants [1–6]. Hyperoxia causes lung injury in animal models [2,7,8], and reactive oxygen species (ROS) such as superoxide anion, hydroxyl radical, and hydrogen peroxide are likely candidates in eliciting lung injury [9–13]. The cytochrome P450 (CYP) enzymes belong to a superfamily of hemoproteins that participate in the metabolism of numerous endobiotics and xenobiotics [14]. The CYP1A family comprises of two proteins, i.e. CYP1A1 and 1A2, which are induced by polycyclic aromatic hydrocarbons (PAHs), such as benzo[a]pyrene or 3-methylcholanthrene [15]. Earlier work in rat and mouse models has shown that hyperoxia induces CYP1A enzymes [9,12,13,16–18], but the underlying mechanisms are not fully understood. We showed earlier that in adult mice, hyperoxia-mediated induction of CYP1A enzymes involves the arylhydrocarbon receptor (AHR) [18], which is the transcription factor that plays a pivotal role in the regulation of CYP1A genes [19–23].</p><p>Whereas induction of CYP has been implicated in hyperoxic lung injury [9,12,13,16,17,24], several groups, including ours have shown that CYP1A1 may play a protective role. Pretreatment of rats [25,26] with inducers of CYP1A enzymes attenuates hyperoxic lung injury. Also, adult mice lacking Cyp1a1 [27], or 1a2 [28], and newborn mice deficient in Cyp1a1 [29] are more susceptible to lung injury by hyperoxia, compared to similarly exposed wild type (WT) mice. The mechanisms of induction of human CYP1A1 by hyperoxia are not well understood. It is also not known if transgenic newborn mice expressing human CYP1A1 promoter will display altered susceptibility to hyperoxic lung injury. Therefore, in this study, we tested the hypothesis that newborn transgenic mice expressing the human CYP1A1-Luc promoter will display transcriptional activation of the human promoter in vivo upon exposure to hyperoxia, and that these mice will be less susceptible to hyperoxic lung injury and alveolar simplification than similarly exposed wild type (WT) mice.</p><!><p>Transgenic mice expressing the human 13.2 Kb CYP1A1 promoter to drive luciferase expression (CYP1A1-Luc) on CD-1 background were obtained from Xenogen Corporation (Alameda, CA) [30–32]. Newborn wild type (WT) (CD-1) or CYP1A1-Luc mice were placed in plexiglass chambers within 12 h of birth, and were either maintained in room air or were exposed to hyperoxia (85% O2) for 7 or 14 days [29,33]. Details of hyperoxia exposures were as reported previously [29]. The dams were rotated between room air and hyperoxia-exposed litters every 24 h to prevent oxygen toxicity in the dams. The mice were sacrificed on PND 8 or 15. The animals were anesthetized with sodium pentobarbital (200 mg/kg i.p.) and euthanized by exsanguination while under deep pentobarbital anesthesia. Six animals from each group were used for lung histopathology. Lung and liver tissues were harvested and stored at −80 °C for subsequent for mRNA and protein isolation and analyses. This study was approved and performed in accordance with federal guidelines for the humane care and use of laboratory animals by the Institutional Animal Care and Use Committee of Baylor College of Medicine.</p><!><p>Lungs were inflated through an intratracheal catheter with buffered zinc formalin (10%) and fixed in with the same solution at a constant pressure of 25 cm H20 for at least 10 min. Lungs were sectioned at 4 μm thickness on a rotary microtome and were stained with hematoxylin and eosin (H&E). Alveolar development was evaluated at PND8 and 15 (n = 6/group) by radial alveolar counts (RAC) [34] and mean linear intercept (MLI) [35] as described before [36]. Fifteen randomly chosen areas were photographed (200x magnification). Fields containing large airways and vessels were not included. Analysis of each section was carried out in a blinded fashion.</p><!><p>RNA extraction from liver and lung tissues was according to previously published protocols [29]. Real time quantitative PCR was conducted according our previous publication [29]. The cDNA (2 μl) and the real-time PCR primers for CYP1A1 (Forward 5′-GGTTAAC-CATGACCGGGAACT-3′ Reverse 5′-TGCCCAAACCAAAGAGAGTGA-3′) and NQO1 (Forward 5′-GGAAGCTGCAGACCTGGTGA-3′ Reverse 5′CCTTTCAGAATGGCTGGCA-3′) were used in final 20 μL qPCR reaction with a SYBR-green master mix (Qiagen-Cat#2014143). Real-time qPCR was performed in an ABI-Prism7700 sequence detection system. Data was analyzed by ΔΔ Ct method; the expression of the target genes such as CYP1A1 and NQO1 were normalized to β-Actin as an endogenous control.</p><!><p>Ethoxyresorufin-O-deethylase (EROD) (CYP1A1) and methoxyresorufin O-deethylase (MROD) activities in the liver and lung whole proteins were assayed essentially as described previously [9,17,18,29]. The luciferase assays in snap frozen lung and liver tissues were performed using a kit purchased from Promega (Cat# E1501), and the assay protocol was according to the manufacturer's instructions. Tissue luciferase activity was measured with a Luminometer (Spectra Max M3, Molecular Devices, California, USA) and recorded as relative light units (RLU).</p><!><p>Liver or lung tissue total proteins (~20 μg protein) were subjected to SDS-polyacrylamide gel electrophoresis in 10% acrylamide gel followed by Western blotting to detect CYP1A1, and NQO1 proteins, as reported in the recent articles from our laboratory [9,17,18,29]. The primary monoclonal antibody to CYP1A1 was a gift from Dr. P.E. Thomas (Rutgers University, Piscataway, NJ).</p><!><p>The data were analyzed by using two-way analyses of variance (ANOVA) followed by modified t-tests, which were used to assess significant differences arising from exposure to hyperoxia and room air in WT and CYP1A1-Luc mice. *, P-values < 0.05 were considered significant.</p><!><p>Hyperoxia elicited significant induction (~30-fold) of hepatic CYP1A1-Luc expression after 14 days of hyperoxia, compared to the room air group (Fig. 1A). In lung, hyperoxia caused a 2-fold increase in CYP1A1-Luc expression after 7 days, but the induction declined to control levels after 14 days (Fig. 1B). Hyperoxia did not alter the levels of endogenous hepatic CYP1A1 mRNA levels in WT, but it elicited about 60% induction in CYP1A1-Luc mice at the 7 but not the 14 day time point (Fig. 1C). In lung, similar to liver, hyperoxia did not alter mouse CYP1A1 levels in WT mice, but elicited a marked induction (~4-fold) of CYP1A1 mRNA in CYP1A1-Luc mice at 7 days (Fig. 1D).</p><!><p>As shown in Fig. 2, WT mice in room air at postnatal day 15 (14 days of room air exposure) showed normal lung architecture (Fig. 2A), but after hyperoxia, these mice showed impaired alveolarization (Fig. 2B). In transgenic mice maintained in room air, lung architecture was normal (Fig. 2C), while mice exposed to hyperoxia showed alveolar simplification (Fig. 2D), but this was less pronounced than in WT (CD-1) mice (Fig. 2B).</p><p>Quantitative analyses showed no changes in lung weight to body weight (LW/BW) ratios in WT mice exposed to hyperoxia for 7 days, compared to room air (Fig. 2E), but there was higher pulmonary edema after 14 days of hyperoxia compared to room air animals (Fig. 2E). In CYP1A1-Luc mice there were no changes in LW/BW ratios after 7 days of hyperoxia compared to room air (Fig. 2E). Interestingly, these mice showed lesser LW/BW ratios after 14 days of hyperoxia compared to air-breathing animals (Fig. 2E).</p><p>Lung morphometric studies showed significantly lower (~62%) radial alveolar counts (RAC) in WT mice after 7 as well as 14 days of hyperoxia, compared to the room air group (Fig. 2F). While RACs were not significantly different among room air and hyperoxic groups at 7 days in the CYP1A1-Luc mice (Fig. 2F), at the 14 day time point the air-breathing animals showed higher RACs than hyperoxic animals (Fig. 2F). The mean linear intercept (MLI) was significantly higher (~40–50%) after 7 and 15 days of hyperoxia in the WT mice compared to room air (Fig. 2G). On the other hand, the MLI was only slightly increased, albeit statistically significant, after hyperoxia in the CYP1A1-Luc mice (Fig. 2G).</p><!><p>There was no significant effect of hyperoxia on liver EROD (CYP1A1) activities in WT mice (Fig. 3A), but there was a significant increase in the activity of this enzyme in CYP1A1-Luc mice at the 14 day time point (Fig. 3A). A similar effect was seen in MROD (CYP1A2) activities (Fig. 3B). In lung, hyperoxia did not alter EROD activities in WT mice, but it did elicit significant increase compared to room air controls at both the time points in CYP1A1-Luc mice (Fig. 3C). Western blotting analyses (Fig. 3D–F) showed that hyperoxia caused significant induction of pulmonary CYP1A1 protein expression in WT mice at 14 days and CYP1A1-Luc mice at both time points, compared to room air controls (Fig. 3E and F).</p><!><p>Hyperoxia exposure led to decrease in hepatic NQO1 mRNA in WT mice at both time points (Fig. 4A), but in CYP1A1-Luc mice, it elicited a significant induction of NQO1 gene expression at 14, but not 7, day time point (Fig. 4A). In lung, hyperoxia caused induction of NQO1 mRNA in WT and CYP1A1-Luc mice at both time points, but this was most pronounced in CYP1A1-Luc mice at 14 days (Fig. 4B). Western blotting analyses (Fig. 4C and D), followed by densitometric analyses (Fig. 4E), showed that hyperoxia induced NQO1 protein expression in WT at 14 days, and at both time points in the CYP1A1-Luc mice (Fig. 4E).</p><!><p>In this study, we tested the hypothesis that hyperoxia leads to transcriptional activation of the human CYP1A1 promoter in vivo. Induction of luciferase expression in liver (Fig. 1A) and lung of newborn transgenic mice expressing CYP1A1-Luc (Fig. 1B) was probably due to transcriptional activation of the corresponding human promoter. These findings were consistent with our earlier study showing induction of luciferase in adult transgenic mice harboring the human CYP1A1-Luc or mouse CYP1A2-Luc [32]. Our observation that hyperoxia elicited induction of endogenous mouse Cyp1a1 gene expression in liver (Fig. 1C) and lung (Fig. 1D) of CYP1A1-Luc, but not WT mice, was intriguing, and could have been due to activation of mouse Cyp1a1 gene due to incorporation of the human CYP1A1-Luc promoter into the mouse genome.</p><p>Our results (Fig. 2A,B, E–G) showing increases in lung injury and alveolar simplification in WT (CD-1) mice following hyperoxia (85% O2) was consistent with our previous studies in newborn C57BL/6J mice [29,37] and Fisher 344 rats [38] that showed similar lung phenotype after prolonged hyperoxia. Our observations showing decreased susceptibility of newborn transgenic mice carrying the human CYP1A1-Luc promoter to lung injury and alveolar simplification supported the hypothesis that mice harboring the human CYP1A1 promoter display beneficial effects probably by modulating the endogenous expression of mouse CYP1A and phase II antioxidant enzymes such as NQO1.</p><p>The finding that hyperoxia caused significant increases of hepatic EROD (Fig. 3A) and MROD activities (Fig. 3B) at the 14 day time point of newborn CYP1A1-Luc mice indicated induction of CYP1A1 and 1A2 activities, respectively [9,17,18]. That the transgenic newborn mice displayed induction of pulmonary EROD after 7 or 14 days of hyperoxia (Fig. 3C) supported the hypothesis that hyperoxia induced functional endogenous CYP1A1 expression. The increased apoprotein contents of CYP1A1 in lungs of CYP1A1-Luc suggested that the increase in gene expression contributed to the augmented expression of the corresponding proteins (Fig. 3F).</p><p>The decrease in hepatic gene expression of NQO1 in WT mice by hyperoxia and the increase in the expression of this gene in CYP1A1-Luc mice (Fig. 4A) suggested that introduction of CYP1A1-Luc transgene altered the regulation of NQO1 via an as yet identified mechanism. This phenomenon may have in part contributed to the decreased susceptibility of CYP1A1-Luc mice to hyperoxia compared to WT mice. The observation in lung of hyperoxia-mediated induction of NQO1 to a greater extent in CYP1A1-Luc compared to that observed in WT mice (Fig. 4B) lends further credence to the hypothesis that increased NQO1 expression contributed to the protective effects of CYP1A1-Luc in the newborn transgenic mice. The increase in NQO1 gene expression in WT as well as CYP1A1-Luc mice by hyperoxia was accompanied by a concomitant increase in the corresponding protein levels (Fig. 4C–E).</p><p>Taken together, our results suggest that the beneficial effects observed in CYP1A1-Luc mice against hyperoxic lung injury were due to a combination of mechanisms entailing induction of endogenous CYP1A enzyme as well as NQO1 by hyperoxia, and the elevated CYP1A and NQO1 expression protect against lung injury by catalyzing the detoxification of ROS-mediated metabolites (e.g., lipid hydroperoxides, quinones). Our recent studies in newborn [29] as well as adult [27,28] showing that mice lacking the genes for CYP1A1 [27] or 1A2 [28] are more susceptible to hyperoxic lung injury than WT mice lend credence to this hypothesis. Interestingly, adult mice carrying the human CYP1A1-Luc or mouse Cyp1a2-Luc were more susceptible to lung injury than WT (CD-1) mice [32], but the fact that the adult hCYP1A1-and and mouse Cyp1a2-Luc displayed decreased expression of CYP1A enzymes may have contributed to the increased lung injury in these mice. It is also possible that differences in the mechanisms of lung injury in adult and newborn by hyperoxia may have in part contributed to the differential phenotype in adult and newborn mice [39]. Further studies on the molecular regulation of CYP1A and NQO1 genes by hyperoxia could lead to much needed novel strategies [40] for the prevention and/or treatment of BPD in premature infants.</p>

PubMed Author Manuscript

Enhanced CO evolution for photocatalytic conversion of CO2 by H2O over Ca modified Ga2O3

Artificial photosynthesis is a desirable critical technology for the conversion of CO 2 and H 2 O, which are abundant raw materials, into fuels and chemical feedstocks. Similar to plant photosynthesis, artificial photosynthesis can produce CO, CH 3 OH, CH 4 , and preferably higher hydrocarbons from CO 2 using H 2 O as an electron donor and solar light. At present, only insufficient amounts of CO 2 -reduction products such as CO, CH 3 OH, and CH 4 have been obtained using such a photocatalytic and photoelectrochemical conversion process. Here, we demonstrate that photocatalytic CO 2 conversion with a Ag@Cr-decorated mixture of CaGa 4 O 7 -loaded Ga 2 O 3 and the CaO photocatalyst leads to a satisfactory CO formation rate (>835 µmol h −1 ) and excellent selectivity toward CO evolution (95%), with O 2 as the stoichiometric oxidation product of H 2 O. Our photocatalytic system can convert CO 2 gas into CO at >1% CO 2 conversion (>11531 ppm CO) at ambient temperatures and pressures.

enhanced_co_evolution_for_photocatalytic_conversion_of_co2_by_h2o_over_ca_modified_ga2o3

C<!>Results and discussion<!>Methods

<p>arbon dioxide (CO 2 ) concentrations in the atmosphere have increased drastically over the past few centuries owing to the combustion of carbon-rich fossil fuels such as coal, oil, and natural gas. As a major anthropogenic greenhouse gas, these ever-increasing CO 2 emissions are detrimental to the environment and will affect both ecosystems and the global climate 1 . Therefore, there is a critical requirement of mitigating CO 2 emissions to achieve sustainable development. Since the pioneering work on the photocatalytic conversion of CO 2 into formic acid (HCOOH) and methyl alcohol (CH 3 OH) over semiconductors reported by Halmann and Inoue et al. 2,3 , the photocatalytic conversion of CO 2 into other valuable feedstocks at ambient temperatures and pressures has attracted considerable attention from the scientific community as a feasible strategy for CO 2 storage and conversion [4][5][6][7][8] .</p><p>In general, the photocatalytic conversion of CO 2 over an excited semiconductor-based catalyst involves three main steps. First, CO 2 molecules are adsorbed on the photocatalyst surface [9][10][11] . Second, the photogenerated electrons react with the adsorbed CO 2 species and protons (H + ) to yield products such as carbon monoxide (CO), methane (CH 4 ), CH 3 OH, and HCOOH. Among these possible reduction products, CO is one of the most useful because it is widely combined with H 2 to provide synthetic gas for use in many chemical processes, such as methanol synthesis 12,13 and the industrial Fischer-Tropsch process that produce various chemicals and synthetic fuels 14,15 . Third, the products are desorbed from the photocatalyst surface. However, as the H/H 2 redox potential (−0.41 V vs. NHE at pH 7) is more positive than that for CO 2 /CO (−0.52 V vs. NHE at pH 7), the generation of H 2 from H + is preferable for the photocatalytic conversion of CO 2 into CO, where H 2 O acts as the electron donor [16][17][18] . Moreover, because of the high thermodynamic stability of the linear CO 2 molecule, the fixation and activation of CO 2 are also immense challenges in the photocatalytic conversion of CO 2 by H 2 O 4,19 . Thus, although various heterogeneous photocatalysts have been reported for the photocatalytic conversion of CO 2 into CO with H 2 O as the electron donor [20][21][22][23][24] , the photocatalytic activity for CO evolution remains limited to a few micromoles, while the photocatalytic conversion rate of CO 2 into CO is <0.15%.</p><p>Based on the processes involved in the photocatalytic conversion of CO 2 described previously, we deduce that the photocatalytic activity of the photocatalyst for CO 2 conversion can be improved by increasing CO 2 adsorption, charge separation, and product desorption. Due to the fact that CO 2 acts as a Lewis acid that bonds easily with Lewis bases 25 , many studies have focused on improving CO 2 adsorption by modifying the photocatalyst surface with a CO 2 adsorbent, such as NaOH 26 , amino groups 27 , and rare earth species 28 , to increase the photocatalytic activity and selectivity for CO 2 conversion by H 2 O. Our group reported that modifying the photocatalyst surface with alkaline earth metals (e.g., Ca, Sr, and Ba) enhanced the conversion of CO 2 and the selectivity toward CO evolution 29 . Moreover, we found that a Ag@Cr core/shell cocatalyst suppresses the backward reaction from CO and O 2 to CO 2 , and enhances the adsorption of CO 2 , resulting in a highly selective photocatalytic CO 2 conversion 30,31 .</p><p>In this study, we exploited the above techniques and successfully fabricated a Ag@Cr-decorated mixture of CaGa 4 O 7 -loaded Ga 2 O 3 and CaO photocatalyst, which exhibits a high CO formation rate (>835 µmol h −1 ) per 0.5 g of catalyst, in addition to high selectivity toward CO evolution (>95%) with the stoichiometric production of O 2 as the oxidation product of H 2 O during the photocatalytic conversion of CO 2 by H 2 O. Approximately 1.2% of the CO 2 in the gas phase was transformed into CO (11531 ppm) as a product. The results reported in this study represent almost an order of magnitude higher than most previously published results, as summarized in Supplementary Table 1.</p><!><p>Photocatalytic reduction of CO 2 by H 2 O. Table 1 shows the formation rates of CO, H 2 , and O 2 , selectivity toward CO evolution, and the balance between consumed electrons and holes over the bare Ga 2 O 3 , Ag-modified Ga 2 O 3 (Ag/Ga 2 O 3 ), Ag@Crmodified Ga 2 O 3 (Ag@Cr/Ga 2 O 3 ), and Ag@Cr-modified Caloaded Ga 2 O 3 (Ag@Cr/Ga 2 O 3 _Ca) photocatalysts during the photocatalytic conversion of CO 2 by H 2 O. No liquid products were detected in the reaction solutions in these photocatalytic systems, and H 2 , O 2 , and CO were detected as gaseous products. As no reduction products other than H 2 and CO were generated, the selectivity toward CO evolution and the balance between the consumed electrons and holes were calculated as follows:</p><p>where R CO and R H2 represent the formation rates of CO and H 2 , respectively. If H 2 O acts as an electron donor, the value of e − /h + should be equal to 1.</p><p>We obtained stoichiometric amounts of H 2 and CO as reduction products in addition to O 2 as the oxidation product, indicating that H 2 O serves as the electron donor. Bare Ga 2 O 3 exhibited a particularly low selectivity toward CO evolution (4%) as the electrons generated by charge transfer were not consumed in the reduction of CO 2 , but rather in the production of H 2 from H + . Modifying Ga 2 O 3 with a Ag cocatalyst enhanced the selectivity toward CO evolution (29%); however, this was not sufficient to obtain a selectivity >50%. In contrast, we succeeded in the selective photocatalytic conversion of CO 2 by H 2 O over Ag@Cr/Ga 2 O 3 . A relatively high CO formation rate (499.6 µmol h −1 ) was achieved with 77% selectivity toward CO evolution. The photocatalytic reaction for the conversion of CO 2 by H 2 O over Ag@Cr/Ga 2 O 3 and Ag@Cr/Ga 2 O 3 _Ca was carried out for at least four times, and errors in the product formation rates (H 2 , O 2 , and CO) were smaller than 5%. Controlling both, the bulk and surface of the photocatalysts, is highly important for achieving a considerably high CO formation rate and selectivity toward CO evolution. We found that the amount of Ca species significantly affected the H 2 and CO formation rates (for the product formation rates and selectivity over various Ag@Cr/Ga 2 O 3 _Ca photocatalysts see Supplementary Fig. 1). The formation rate of CO increased first and then decreased as the Ca content increased (Supplementary Fig. 1a-g).</p><p>In contrast, the formation rate of H 2 over the Ag-Cr/ Ga 2 O 3 _Ca_x samples increased monotonically with increasing amount of Ca species. The Ag-Cr/CaGa 4 O 7 photocatalyst was only active for H 2 evolution derived from water splitting (Supplementary Fig. 1h). The Ag@Cr/Ga 2 O 3 _Ca photocatalyst exhibited the highest CO formation rate (794.2 µmol h −1 ), and the selectivity toward CO evolution was approximately 82%. Additionally, CO production from the photocatalytic conversion of CO 2 after photoirradiation for 15 h over Ag@Cr/Ga 2 O 3 _Ca was more stable than that over Ag@Cr/Ga 2 O 3 (for the product formation rates for 15 h see Supplementary Fig. 2), which indicates that the presence of Ca species is not only beneficial for improving the photocatalytic activity and selectivity, but also for improving stability during the photocatalytic conversion of CO 2 to CO.</p><p>Various control experiments were carried out to confirm the source of CO during the photocatalytic conversion of CO 2 by H 2 O, the results of which are shown in Supplementary Fig. 3. We did not detect any appreciable amounts of products under dark conditions or in the absence of a photocatalyst. In addition, H 2 was the main product formed when Ar gas was used instead of CO 2 or in the absence of NaHCO 3 . The control experiments confirmed that the evolved CO originated from the CO 2 gas introduced into the samples and not from carbon contaminants.</p><p>Photocatalyst characterization. The actual amounts of the Ca species loaded into Ga 2 O 3 at different CaCl 2 concentrations were measured using inductively coupled plasma optical emission spectrometry (ICP-OES) (Supplementary Table 2). We found that almost all the Ca species were loaded into the Ga 2 O 3 photocatalyst when the CaCl 2 concentration was <0.001 mol L −1 . However, not all the Ca species could be loaded into Ga 2 O 3 at higher CaCl 2 concentrations. Note that even when no CaCl 2 was added during the preparation of Ga 2 O 3 , trace amounts of Ca were detected in Ga 2 O 3 , which is likely due to Ca impurities present in the experimental vessels or precursor reagents. Hereinafter, we refer to the Ca-loaded Ga 2 O 3 photocatalysts as Ga 2 O 3 _Ca_x (x = 0.32, 0.62, 1.1, 1.6, 2.1, 3.3 mol%) based on the Ca/Ga molar ratio determined by ICP-OES. Figure 1a shows the X-ray diffraction (XRD) patterns of the bare Ga 2 O 3 , Ga 2 O 3 _Ca_x, and CaGa 4 O 7 photocatalysts. As indicated, gradual changes in the diffraction peaks assigned to the (020), (311), (400), (002), and (330) facets of CaGa 4 O 7 (JSPDS 01-071-1613) were observed as the amount of Ca species was increased. In general, a high Ca loading is favorable for the formation of CaGa 4 O 7 . We observed no distinct shifts in the diffraction peaks for the Ga 2 O 3 _Ca_x samples compared with those of bare Ga 2 O 3 . As the ionic radius of Ca 2+ (0.099 nm) 32 is larger than that of Ga 3+ (0.062 nm) 33 , the unshifted XRD peaks imply that Ca 2+ does not act as a dopant in the bulk Ga 2 O 3 lattice. However, there was a clear increase in the peak intensity at 2θ = 30.1°and an apparent decrease in that at 2θ = 30.5°with increasing amount of Ca species (Fig. 1b), which are possibly ascribed to the formation of CaGa 4 O 7 species on Ga 2 O 3 . The increased intensity of the Ca 2p X-ray photoelectron spectroscopy (XPS) peak (Fig. 1c) also indicates that the amount of Ca species on the Ga 2 O 3 surface increased with increasing Ca levels. In addition, the XPS peak locations in the Ca 2p spectra of the Ga 2 O 3 _Ca_x photocatalysts are similar to those of CaGa 4 O 7 , but different from those of CaO. The Ca 2p XPS profiles suggest that a thin CaGa 4 O 7 layer forms on the Ga 2 O 3 surface and that the amount of CaGa 4 O 7 increases as the amount of Ca is increased. We further confirmed the morphological changes in the Ga 2 O 3 _Ca sample by field-emission scanning electron microscopy (SEM), as shown in Fig. 1d. Both ends of the Ga 2 O 3 nanoparticles gradually sharpened and their surfaces became smoother as the amount of Ca species increased, especially when the Ca amount was higher than 1.1 mol%. This smoothing of the Ga 2 O 3 surfaces with increasing Ca/Ga molar ratio resulted in a decrease in the Brunauer-Emmett-Teller (BET) specific surface area of Ga 2 O 3 _Ca_x (Supplementary Fig. 4), which is attributable to the modification of CaGa 4 O 7 , as we confirmed from the XRD patterns and the XPS results that a CaGa 4 O 7 layer was formed on the Ga 2 O 3 surface.</p><p>The close linkage between CaGa 4 O 7 and Ga 2 O 3 on the Ga 2 O 3 surface was confirmed by field-emission transmission electron microscopy (TEM) and high-resolution TEM (HRTEM) (Fig. 2). The marked lattice spacings (0.296 and 0.255 nm) in Fig. 2b correspond to the (130) and (111) planes of CaGa 4 O 7 and Ga 2 O 3 , respectively. The core-shell-structured Ag@Cr cocatalyst was successfully loaded onto the Ga 2 O 3 _Ca surface using the photodeposition method (Fig. 2c, d), as reported previously by us 31 .</p><p>Role of the Ca species. Figure 3 shows the Fourier transform infrared (FTIR) spectra of the CO 2 -adsorbed samples after introducing CO 2 at ~0.2 Torr. When CO 2 was introduced into the Ga 2 O 3 sample, three absorbance peaks were observed at 1634, 1432, and 1225 cm -1 , which can be ascribed to asymmetric CO 3 stretching vibrations [ν as (CO 3 )], symmetric CO 3 stretching vibrations [ν s (CO 3 )] of monodentate bicarbonate species (m-HCO 3 -Ga), and OH deformation vibrations [δ(OH)], respectively [34][35][36] . The absorbance peaks at 1699 and 1636 cm -1 for the CO 2 -adsorbed CaO sample can be attributed to bridging carbonate stretching and asymmetric CO 3 stretching vibrations [ν as (CO 3 )] of the bicarbonate species, respectively. The broad structureless absorbance peaks between 1480 and 1318 cm -1 can be attributed to the symmetric and asymmetric CO 3 stretching of unidentate carbonate, as well as the symmetric CO 3 stretching [ν s (CO 3 )] of bicarbonate [37][38][39][40][41] . When the Ga 2 O 3 surface was modified with a small amount of Ca species, absorbance peaks attributable to CO 2 adsorption by both Ga 2 O 3 and CaO were observed after CO 2 was introduced into the Ga 2 O 3 _Ca_1.1 sample. However, when the Ga 2 O 3 surface was modified with large amounts of Ca species, the absorbance peaks attributed to CO 2 adsorption on Ga 2 O 3 had low intensity and mainly corresponded to the broad peaks derived from the adsorption of CO 2 on CaGa 4 O 7 . Supplementary Fig. 5 shows the FTIR spectra of CO 2 -adsorbed Ga 2 O 3 , Ga 2 O 3 _Ca_1.1, Ga 2 O 3 _Ca_3.3, and CaGa 4 O 7 samples after introducing the same amount of CO 2 at various pressures in the 0.1-40.0 Torr range. CO 2 was adsorbed significantly more on the Ga 2 O 3 _Ca_1.1 surface than on the Ga 2 O 3 surface due to its adsorption at both Ga and Ca sites. However, the CaGa 4 O 7 surface was not conducive to CO 2 adsorption; therefore, CO 2 adsorbed less onto the Ga 2 O 3 _Ca_3.3 surface than the Ga 2 O 3 _Ca_1.1 surface.</p><p>Figure 4 shows the FTIR spectra of the adsorbed CO 2 species on Ga 2 O 3 _Ca_1.1 after different durations of photoirradiation. As the photoirradiation time increased from 0 to 106 h, the bands at 1225 [δ(OH)-Ga] and 1408 cm -1 [ν s (CO 3 )-Ca] decreased and vanished after 104 h. At the same time, new bands gradually appeared at 1581, 1388, and 1353 cm -1 (asymmetric CO 2 stretching [ν as (CO 2 )], CH deformation [δ(CH)], and symmetric CO 2 stretching [ν s (CO 2 )] assigned to formate species (HCOO-Ga/Ca), respectively) [34][35][36] . As the photoirradiation continued, the formate species were consumed and gaseous CO (fundamental vibration band at 2143 cm −1 ) 42 was formed simultaneously. This result indicates that the bicarbonate species is the intermediate during the photocatalytic conversion of CO 2 , and the formates transform into CO with photoirradiation, which is consistent with our previous results 43,44 . It is worth mentioning that in addition to the presence of intermediate species on the Ga 2 O 3 surface ([δ(OH)-Ga]), the modification by Ca species further increased the amount of intermediate on Ga 2 O 3 _Ca_1.1. As the photocatalytic conversion of H + into H 2 and the conversion of CO 2 into CO are two competing processes in an aqueous solution, the high adsorption of CO 2 at the base site leads to high photocatalytic activity and selectivity toward CO evolution during the photocatalytic conversion of CO 2 by H 2 O.</p><p>In order to demonstrate that the presence of CaO on the Ga 2 O 3 surface enhances the photocatalytic activity and selectivity during the photocatalytic conversion of CO 2 into CO, we investigated the photocatalytic performance during the conversion of CO 2 by H 2 O over various Ag@Cr/CaO/Ga 2 O 3 photocatalysts, the results of which are shown in Fig. 5. We found that the Ag@Cr/ Ga 2 O 3 _Ca_1.1 photocatalyst (with a low amount of CaO generated on the Ga 2 O 3 surface) significantly enhanced the rate of CO formation during the photocatalytic conversion of CO 2 by H 2 O compared with bare Ag@Cr/Ga 2 O 3 (Fig. 5a, b). However, no significant change in the rate of CO formation and selectivity toward CO evolution was observed for the sample labeled "Ag@Cr/(1.1 mol%CaO/Ga 2 O 3 )" (in which 1.1 mol% CaO was physically loaded onto Ga 2 O 3 by grinding before loading Ag@Cr cocatalyst onto the CaO/Ga 2 O 3 surface) as compared to bare Ga 2 O 3 (Fig. 5c). Because uncalcined CaO-loaded Ga 2 O 3 easily dissolves in H 2 O, we increased the CaO loading on the Ga 2 O 3 surface to 30 mol% using the same grinding method (labeled "Ag@Cr/(30 mol%CaO/Ga 2 O 3 )"), which resulted in an increased rate of CO formation and a decrease in H 2 formation (Fig. 5d).</p><p>However, no improvement in photocatalytic activity and selectivity was observed when 30 mol% CaO was mixed with the prepared Ag@Cr/Ga 2 O 3 and ground together (Fig. 5e) or when they were directly mixed in the reaction solution (Fig. 5f). These results clearly reveal that the addition of CaO on the Ga 2 O 3 surface enhances the rate of CO formation and suppresses that of H 2 during the photocatalytic conversion of CO 2 by H 2 O. In addition, the tight junction between Ga 2 O 3 , CaO, and the Ag@Cr cocatalyst is crucial for the superior photocatalytic activity and selectivity of the photocatalyst for the conversion of CO 2 into CO. In our previous work, we confirmed that Ag acts as an active site while the Cr(OH) 3 •H 2 O layer exterior to the Ag core increases CO 2 adsorption 30,31 . Hence, the Ag@Cr cocatalyst should be loaded at the CaO/Ga 2 O 3 interface in order to facilitate contact between the CaO-adsorbed CO 2 species and the Ag active sites.</p><p>Notably, although CaGa 4 O 7 exhibited high selectivity toward H 2 evolution, the H 2 formation rate for CaGa 4 O 7 was significantly lower than that for Ga 2 O 3 _Ca_3.3 (for the product formation rates over Ag-Cr/Ga 2 O 3 _Ca_3.3 see Supplementary Fig. 6). This indicates that the presence of CaGa 4 O 7 on the Ga 2 O 3 surface enhances the overall photocatalytic efficiency during the photocatalytic reaction, including CO 2 conversion and water splitting. The Mott-Schottky plot (Supplementary Fig. 7) and the absorption spectra converted from the diffuse reflectance spectra using the Kubelka-Munk equation (Supplementary Fig We expect that by exploiting the high CO 2 adsorption of CaO and the high photocatalytic efficiency of CaGa 4 O 7 /Ga 2 O 3 , we can further improve the photocatalytic activity and selectivity of the photocatalyst to maximize the conversion of CO 2 into CO by H 2 O. Figure 6a shows the formation rates of H 2 , O 2 , and CO during the photocatalytic conversion of CO 2 by H 2 O for the Ga 2 O 3 _Ca_3.3 photocatalyst physically mixed with 30 mol% of CaO and Ag@Cr as the cocatalyst. As indicated, a high formation rate of CO (>835 µmol h -1 ) was achieved, in addition to an excellent selectivity toward CO evolution (>95%), with a stoichiometric amount of evolved O 2 . Both 12 CO and 13 CO were detected using quadrupole mass spectrometry (MS), and the peaks at m/z = 28 and m/z = 29 were located at the same positions as those detected by gas chromatography (GC) during the photocatalytic conversion of 13 CO 2 (for the isotopic lead experiments see Supplementary Fig. 10). Indeed, our results indicate that the detected 12 CO was produced from the reduction of 12 CO 2 derived from the NaHCO 3 additive in the solution 43 . As shown in Fig. 6b, with the consumption of 12 CO 2 derived from NaHCO 3 , the amount of generated 12 CO gradually decreased, while the 13 CO content increased under continuous bubbling of 13 CO 2 . The total amounts of 13 CO and 12 CO detected by MS were consistent with the amount of CO detected by GC (Fig. 6c), which indicates that the CO was generated as the reduction product of either CO 2 introduced in the gas phase or from NaHCO 3 , rather than from any organic contaminants on the photocatalyst surface. The converted concentration of CO based on the CO formation rate was found to be 11,531 ppm, indicating that ~1.2% of CO 2 in the gas phase was transformed into CO (see Supplementary Information for the calculation details. The actual amounts of CO detected are shown in Supplementary Movie 1).</p><p>In our previous work, we had found that basic oxides and hydroxides such as Cr(OH) 3 31 , SrO 44 , and rare earth (RE) hydrates and oxides 28 function as good CO 2 storage materials by generating the corresponding (hydroxy)carbonate compounds (e.g., Cr(OH) x (CO 3 ) y and RE 2 (OH) 2(3−x) (CO 3 ) x ), and they improve the photocatalytic activity and selectivity toward CO evolution. Now, we propose a possible mechanism for the photocatalytic conversion of CO 2 by H 2 O over Ag@Cr/CaO/ CaGa 2 O 7 /Ga 2 O 3 , as shown in Fig. 7. During the photocatalytic conversion of CO 2 in an aqueous solution of NaHCO 3 , the Cr (OH) 3 •H 2 O and CaO species that are in close contact with Ag particles easily form (hydroxy)carbonate species (named M (OH) x (CO 3 ) y , M=Cr or Ca) 31 , which greatly increase the concentration of CO 2 -related species around the Ag active sites, thereby improving selectivity for the photocatalytic conversion of CO 2 into CO instead of water splitting. On the other hand, the Ga 2 O 3 /CaGa 4 O 7 heterojunction improves the efficiency for spatial separation of the photogenerated carriers, which also increases the photocatalytic activity for the conversion of CO 2 into CO. Moreover, while the Cr(OH) 3 •xH 2 O shell outside the Ag particle can be oxidized to Cr 6+ and dissolve into the solution during the photocatalytic conversion of CO 2 46 , the presence of CaO around the Ag active site compensates for the reduced activity from the dissolution of Cr species. As a result, Ag@Cr/ Ga 2 O 3 _Ca is photocatalytically much more stable than Ag@Cr/ Ga 2 O 3 .</p><p>Herein, we reported the photocatalytic conversion of CO 2 using a Ag@Cr/CaO/CaGa 4 O 7 /Ga 2 O 3 photocatalyst, in which a satisfactory CO formation rate (>835 µmol h −1 ) and an excellent selectivity toward CO evolution (95%) were achieved with the stoichiometric production of O 2 as the oxidation product of H 2 O. Through the use of various characterization techniques, we found that the CaO and CaGa 4 O 7 formed on the Ga 2 O 3 surface improved the adsorption of CO 2 at basic sites in addition to enhancing the total photocatalytic efficiency. In addition, the physical mixing of CaGa 4 O 7 /Ga 2 O 3 with CaO was a particularly simple and convenient technique for exploiting the high CO 2 adsorption ability of CaO and the high photocatalytic efficiency of CaGa 4 O 7 /Ga 2 O 3 . These results are of particular interest, considering that previously, only insufficient amounts of CO 2 reduction products were produced during artificial photosynthesis.</p><!><p>Ca-modified Ga 2 O 3 (Ga 2 O 3 _Ca) was prepared using the ammonia precipitation method reported by Sakata et al. 47 . In this method, Ga(NO 3 ) 3 •nH 2 O (12.6 g) was dissolved in 200 mL of deionized water or CaCl 2 solution in ultrapure water at various concentrations. Hydroxylation was carried out by dripping an ammonium hydroxide solution until the pH level reached 9.1. The obtained hydroxides were centrifuged and dried overnight. The Ga 2 O 3 _Ca sample was obtained by calcining the precursor at 1273 K for 10 h. Ag@Cr/Ga 2 O 3 _Ca was synthesized using the photodeposition method reported in our previous work 30 . In this method, the asprepared Ga 2 O 3 _Ca powder (1.0 g) was dispersed in ultrapure water (1.0 L) containing the necessary amounts of silver nitrate (AgNO 3 ) and chromium (III) nitrate (Cr(NO 3 ) 3 ). The suspension was purged with Ar gas and irradiated under a 400 W high-pressure Hg lamp with Ar gas flowing for 1.0 h, followed by filtration and drying at room temperature (~298 K). The Ag/Ga and Cr/Ga molar ratios were both 1.0 mol%.</p><p>Characterization. The as-prepared Ga 2 O 3 _Ca samples were characterized using the following techniques: XRD (Model: Multiflex, Rigaku Corporation, Japan) with Cu Kα radiation (λ = 0.154 nm); XPS (Model: ESCA 3400, Shimadzu Corporation, Japan) with Mg Kα radiation; SEM (Model: SU-8220, Hitachi High-Technologies Corporation, Japan); TEM (Model: JEM-2100F, JEOL Ltd, Japan); and UV-Visible spectroscopy (V-650, JASCO) with an integrated sphere accessory. The BET surface areas of the photocatalyst samples were determined from their N 2 -adsorption isotherms at 77 K using a volumetric gas-adsorption measuring instrument (Model: BELSORP-miniII, MicrotracBEL Corp. (formerly BEL Japan, Inc.), Japan). Prior to these measurements, each sample was evacuated at 473 K for 1 h using a sample pretreatment system (Model: BELPREP-vacII, MicrotracBEL Corp. (formerly BEL Japan, Inc.), Japan). ICP-OES (Model: iCAP7400, Thermo Fisher Scientific, USA) was used to determine the actual amounts of Ca modified on the Ga 2 O 3 surface. The FTIR spectra of the adsorbed carbon species were recorded using an FTIR spectrometer (Model: FT/IR-4700, JASCO International Co., Ltd., Japan) equipped with a mercury-cadmium-tellurium (MCT) detector and cooled with liquid N 2 in the transmission mode at 303 K. Each sample (~30 mg) was pressed into a wafer (diameter: 10 mm) and introduced into the instrument in a cylindrical glass cell with calcium fluoride (CaF 2 ) windows. The wafer was evacuated at 673 K for 30 min before being examined, followed by treatment with O 2 at ~40 Torr for 30 min, after which the wafer was evacuated for 30 min and cooled to 303 K. The data for each FTIR spectrum were obtained from 128 scans with a resolution of 4 cm −1 . The energy gap of the band structure and flat band potential of the Ga 2 O 3 _Ca samples were determined using the Davis-Mott and Mott-Schottky equations, respectively; the experimental details are provided in the Supplementary Information.</p><p>Photocatalytic reaction. The photocatalytic reduction of CO 2 was carried out using a flow system with an inner irradiation-type reaction vessel. The synthesized photocatalyst (0.5 g) was dispersed in ultrapure water (1.0 L) containing 0.1 M sodium bicarbonate (NaHCO 3 ). The CO 2 was bubbled into the solution at a flow rate of 30 mL min −1 . The suspension was illuminated using a 400 W high-pressure Hg lamp with a quartz filter, and the assembly was connected to a water-cooling system. The amounts of evolved H 2 and O 2 were detected using a gas chromatography system fitted with a thermal conductivity detector (TCD-GC, Model: GC-8A, Shimadzu Corporation, Japan) and a 5A molecular sieve (MS 5A) column, and Ar was used as the carrier gas. The amount of evolved CO was analyzed using a gas chromatography system fitted with a flame ionization detector (FID-GC, Model: GC-8A, Shimadzu Corporation, Japan), a methanizer, and a ShinCarbon ST column, and N 2 was used as the carrier gas. High-performance liquid chromatography (Model: LC-4000, JASCO, USA) was used to detect the presence of liquid products.</p><p>In the isotope experiment, 12 CO 2 was replaced by 13 CO 2 . The formation rates of H 2 , O 2 , 12 CO, and 13 CO under photoirradiation were detected using a quadrupole mass spectrometer (BELMASS, Microtrac BEL) combined with a TCD-GC detector.</p>

Nature Communications Chemistry

Differential Effects of Estrogen Exposure on Arylsulfatase B, Galactose-6-Sulfatase, and Steroid Sulfatase in Rat Prostate Development

Sulfatase enzymes remove sulfate groups from sulfated steroid hormones, including estrone-sulfate and dehydroepiandrosterone-sulfate, and from sulfated glycosaminoglycans (GAGs), including chondroitin sulfates and heparan sulfate. The enzymes N-acetylgalactosamine-4-sulfatase (Arylsulfatase B; ARSB) and N-acetylgalactosamine-6-sulfatase (GALNS), which remove sulfate groups from the sulfated GAGs chondroitin 4-sulfate (C4S) and chondroitin 6-sulfate, respectively, have not been studied in prostate development previously. In this report, the endogenous variation and the impact of exogenous estradiol benzoate on the immunohistochemistry and activity of ARSB and GALNS in post-natal (days 1\xe2\x80\x9330) ventral rat prostate are presented, as well as measurements of steroid sulfatase activity (STS), C4S, total sulfated GAGs, and versican, an extracellular matrix proteoglycan with chondroitin sulfate attachments on days 5 and 30. Findings demonstrate distinct and reciprocal localization of ARSB and GALNS, with ARSB predominant in the stroma and GALNS predominant in the epithelium. Control ARSB activity increased significantly between days 5 and 30, but following estrogen exposure (estradiol benzoate 25 \xc2\xb5g in 25 \xc2\xb5l sesame oil subcutaneously on days 1, 3, and 5), activity was reduced and the observed increase on day 30 was inhibited. However, estrogen treatment did not inhibit the increase in GALNS activity between days 5 and 30, and reduced STS activity by 50% on both days 5 and 30 compared to vehicle control. Sulfated GAGs, C4S, and the extracellular matrix proteoglycan versican declined between days 5 and 30 in the control, but these declines were inhibited following estrogen. Study findings indicate distinct variation in expression and activity of sulfatases, sulfated GAGs, C4S, and versican in the process of normal prostate development, and disruption of these events by exogenous estrogen.

differential_effects_of_estrogen_exposure_on_arylsulfatase_b,_galactose-6-sulfatase,_and_steroid_sul

Introduction<!>Animal Care and Treatment<!>ARSB and GALNS immunohistochemistry<!>Determination of ARSB, GALNS, and STS activity in prostate tissue homogenates<!>Measurement of total sulfated glycosaminoglycans<!>Immunoprecipitation of tissue lysates by chondroitin 4-sulfate antibody<!>Determination of versican by competitive ELISA<!>Statistics<!>Changes in sulfatase activity, total sulfated glycosaminoglycans, chondroitin-4-sulfate and versican between day 5 and day 30<!>Impact of estrogen exposure on sulfatase activity, total sulfated glycosaminoglycans, chondroitin-4-sulfate, and versican<!>Predominant localization of ARSB in stroma and of GALNS in prostate epithelium<!>Decline in ARSB positive immunostaining with delay in acinar development following exogenous estrogen<!>Discussion<!>Conclusions<!>

<p>Sulfatase enzymes comprise a group of cellular and extracellular enzymes that are key regulators of the degradation of sulfated glycosaminoglycans, including chondroitin sulfate, dermatan sulfate, keratan sulfate, heparin, and heparan sulfate, and of sulfated steroids, including estrone sulfate and dehydroepiandrosterone sulfate. Arylsulfatase B (ARSB; N-acetylgalactosamine-4-sulfatase) and N-acetylgalactosamine-6-sulfatase (GALNS) are enzymes that remove sulfate groups from the sulfated glycosaminoglycans (GAGs) chondroitin 4-sulfate (C4S) and chondroitin 6-sulfate (C6S), respectively. Deficiency of ARSB or GALNS leads to accumulation of sulfated glycosaminoglycans, resulting in the lysosomal storage diseases Mucopolysaccharidosis (MPS) VI from ARSB deficiency and MPS IVA from GALNS deficiency. Removal of the 4-sulfate group is required for the degradation of C4S, and removal of the sulfate group is required for activity of steroid hormones. Prior experiments in human mammary cell lines demonstrated that 1) estrone (100 pg/ml) and estradiol (200 pg/ml) exposure significantly reduced activity of steroid sulfatase (STS) and ARSB, but not of GALNS in MCF-7 and T47D cells; and 2) GALNS activity was significantly higher in primary mammary epithelial cells, whereas ARSB and STS activity were higher in primary myoepithelial cells [1]. In this report, we present in vivo data that expand on these prior in vitro observations of sulfatase activity and response to exogenous estrogen and report the endogenous sulfatase activity and expression and the impact of estrogen exposure on sulfatase activity in a rodent model of prostate development.</p><p>This study, by elucidating differences between activity and distribution of sulfatases in both the native prostate and following estrogen treatment, presents a novel approach to prostate morphogenesis, focusing on activity and expression of sulfatases that modify chondroitin sulfate and on the associated changes in total sulfated GAGs, C4S, and the ECM proteoglycan versican. Decline in ARSB activity has been shown in malignant prostate tissue, malignant colon tissue, malignant mammary cells, and metastatic colonic cell lines [1, 2–6]. In human prostate cells, decline in ARSB produced increases in total sulfated GAGs, C4S, and versican [2, 3]. Previously, both versican and chondroitin sulfate have been identified as biomarkers of more aggressive prostate cancer [7, 8]. Versican is a large, aggregating extracellular matrix proteoglycan with chondroitin sulfate attachments that interacts with multiple cell surface receptors and recruits signaling molecules to the cell surface, thus modulating signaling pathways and stromal-epithelial interactions [3, 9–11].</p><p>The profiles of ARSB, GALNS, and STS enzyme activity and localization and of total sulfated GAGs, C4S and versican in the developing rat prostate have not previously been addressed. In the rodent, prostate development occurs predominantly in early post-natal life, and high doses of exogenous natural estrogens caused developmental and differentiation defects in the adult prostate [12]. In this report, measurements of ARSB, GALNS, and steroid sulfatase (STS) activity, total sulfated GAGs, C4S, and the proteoglycan versican on days 5 and 30 of post-natal development in rat prostate tissue are presented. The impact of an intermediate dose of exogenous estradiol benzoate (25 µg) exposure on days 1,3, and 5 of post-natal life on these parameters and on ARSB and GALNS immunohistochemistry on post-natal days 1, 3, 6, 10, 15, and 30 is presented. Since decline in ARSB activity has been associated with prostate neoplasia [2, 3], insight into the interactions among estrogen, ARSB, GALNS, and STS in prostate development may lead to better understanding of the effects of steroid hormones on stromal-epithelial interactions and on mechanisms of prostate carcinogenesis.</p><!><p>All procedures were approved by the Institutional Animal Care and Use Committee (IACUC) of the University of Illinois at Chicago. 96 male Sprague-Dawley rats (Zivic-Miller, Pittsburgh, PA) were treated with subcutaneous injections of either 25 µl sesame oil alone (controls) or high-dose estrogen (25 ug estradiol benzoate in 25 µl sesame oil) on post-natal days 1, 3, and 5 as previously described [12, 13]. Twelve animals (6 control and 6 estrogen-treated) were euthanized on days 1, 3, 6, 10, 15, and 30, and the prostates were removed, formalin-fixed, paraffin-embedded and sectioned, as previously detailed [14–17]. In addition, prostate tissue from similarly estrogen-treated and vehicle control animals was obtained from rats sacrificed on day 5 (total n=12) and on day 30 (total n=12).</p><!><p>Sections of the estrogen-treated and control ventral prostate tissues were mounted on the same slide and immunostained. Antigen retrieval was done with pH 6.1 citrate buffer at 90°C for 40 minutes. For ARSB, the slides were incubated overnight with rabbit polyclonal antibody (1:50; Open Biosystems, Huntsville, AL), then for 1 hour with anti-rabbit IgG-HRP at 1:1000 dilution. Color was developed with 3,3'-Diaminobenzidine (DAB) for 5 minutes. For GALNS, slides were incubated overnight with rabbit polyclonal antibody (1:200; Open Biosystems, Huntsville, AL), followed by anti-rabbit IgG-HRP at 1:1000 dilution for 1 hour and DAB for 5 minutes and counterstained with hematoxylin. The sections were blocked with a serum-free Universal Blocking Solution from Biocare Medical (Concord, CA). Negative control used normal rabbit IgG diluted in buffer at the same dilution as used for the primary antibody (1:50 for ARSB; 1:200 for GALNS), and all other staining procedures were similar. Positive controls were sections of adult human prostate. Digitized images were obtained with QCapture software (QImaging, Surrey, BC, Canada) at 20X magnification. Background and brightness were modified with GIMP Portable software (Portable Apps, New York, NY) or with Adobe Photoshop (CS2). Qualitative assessments of intensity and distribution of ARSB and GALNS immunostaining were made by three study investigators (LF, GSP, JKT), who agreed on the descriptive findings that are reported.</p><!><p>Homogenates were prepared from ventral prostate tissue of estrogen-treated and control rats on days 5 and day 30 of post-natal development (total n=24). Arylsulfatase B (ARSB; N-acetylgalactosamine-4-sulfatase) activity measurements were performed using a fluorometric assay, as previously, with 20 µl of tissue homogenate, 80 µl of assay buffer (0.05 M Na acetate buffer, pH 5.6), and 100 µl of substrate [5mM 4-methylumbelliferyl sulfate (MUS) in assay buffer] in wells of a microplate [1, 18, 19]. The microplate was incubated for 30 minutes at 37°C, and the reaction was stopped by 150 µl of stop buffer (Glycine-Carbonate buffer, pH 10.7), and fluorescence was measured at 360 nm (excitation) and 465 nm (emission) in a microplate reader (FLUOstar, BMG, Cary, North Carolina). Activity was expressed as nmol/mg protein/hour, based on a standard curve for ARSB activity prepared with known quantities of 4-methylumbelliferyl at pH 5.6. Protein was determined by total protein assay kit (Pierce, Thermo Fisher Scientific, Inc., Rockford, IL).</p><p>For STS determination, six 20 µl prostate tissue homogenates were incubated with 80 µl of assay buffer (0.5M Tris-Cl Buffer, pH 7.5), and 100 µl of substrate (0.5 mM 4-MUS in assay buffer), as previously detailed [1, 18, 20]. The reaction mixture was incubated for 4 hours at 37°C, at which time 100 µl of stop buffer (1M Tris-Cl buffer at pH 10.4) was added, and fluorescence was measured in a microplate reader (BMG). Activity was expressed as nmol/mg protein/h, and was derived from a standard curve prepared with known quantities of 4-methylumbelliferyl at pH 7.5. Protein was determined by total protein assay kit (Pierce).</p><p>The measurement of Galactose-6-sulfatase (GALNS) activity was performed using 5 µl of tissue homogenate made in ddH2O by sonication with metal tip, combined with 5 µl 0.2% heat-inactivated BSA (or 10 µl of 0.2% heat-inactivated BSA for blank) and 20 µl of substrate [10 mM 4-methylumbelliferyl-β-D-galactoside-6-sulfateNH4 (MU-βGal-6S)] in substrate buffer (0.1M sodium acetate / 0.1M acetic acid at pH 4.3 with 0.1M NaCl, 5mM Pb-acetate (1.9 mg/ml) and 0.02% Na-azide) in wells of a microplate [1, 18, 21]. After incubation for 17 hours at 37°C, 5µl 0.9 M Na-Phosphate buffer at pH 4.3 with 0.02% Na-azide was added, as well as 10µl of 10 U β-D-Galactoside galactohydrolase (Sigma) / ml 0.2% heat-inactivated BSA. After incubation for 2 hours at 37°C, 200 µl of stop buffer (0.5M NaHCO3 / 0.5M Na2CO3 at pH 10.7 with 0.025% Triton-X-100) was added, and readings of fluorescence were taken at 360 nm and 465 nm in a plate reader (BMG). GALNS activity was expressed as nmol/mg protein/h, and protein was determined by total protein assay kit (Pierce).</p><!><p>Total sulfated glycosaminoglycans (GAG) in control and estrogen-treated rat prostate tissue were measured using 1,9-dimethylmethylene blue (Blyscan™, Biocolor Ltd., Newtownabbey, Northern Ireland), as previously detailed [5, 22]. The sulfated polysaccharide component of proteoglycans (PG) and protein-free sulfated GAG chains were detected, whereas disaccharides and hyaluronan were not detected. The reaction was performed in the presence of excess unbound dye. The cationic dye and sGAG at acid pH produced an insoluble dye-GAG complex, and the GAG content was determined by the amount of dye that was recovered from the test sample following exposure to Blyscan dissociation reagent. Absorbance maximum of 1,9-dimethylmethylene blue is 656 nm; sulfated GAG concentration was expressed as µg / mg protein of tissue extract.</p><!><p>Tissue lysates were prepared using RIPA buffer (50 mmol/l Tris-HCl containing 0.15 mol/l NaCl, 1% Nonidet P40, 0.5% deoxycholic acid and 0.1% SDS, pH 7.4). Antibody specific to native C4S (4D1, Abnova, Littleton, CO) was added to lysates in tubes at a concentration of 1 µg/mg of tissue protein, and tubes were rotated overnight in a shaker at 4°C. Next, 100 µl of pre-washed Protein L-Agarose (SCBT, Santa Cruz, CA) was added to each tube, tubes were incubated overnight at 4°C, and the Protein L-Agarose treated beads were washed three times with PBS containing Protease Inhibitor Cocktail, as previously [2, 4]. The precipitate was eluted with dye-free elution buffer and subjected to sulfated GAG assay.</p><!><p>Versican was measured by competitive ELISA (My BioSource, San Diego, CA), in which color development was inversely proportional to the versican content in the test samples. Standards ranging from 1 to 25 ng/ml, samples, and versican-horseradish peroxidase conjugate were added to wells pre-coated with versican antibody, incubated for 1 hour at 37°C, and washed three times. Color was developed by adding hydrogen peroxide / tetramethylbenzidine (TMB) substrate, the reaction was stopped by 2N H2SO4, and the color intensity was read at 450 nm in a plate reader (BMG). The concentration of versican in the samples was extrapolated from the standard curve and expressed per mg of total tissue protein, measured by protein assay (Pierce).</p><!><p>Unless stated otherwise, one-way ANOVA with Tukey-Kramer post-test was performed using InStat (GraphPad Software, Inc., La Jolla, CA), to compare the measurements between day 5 and day 30 samples and between oil-control and estrogen-treated samples. Six independent tissue samples were analyzed for each time point and for treated and control conditions. Measurements are presented as mean values ± standard deviation (S.D.). P-value of less than 0.05 is considered statistically significant, and * represents p≤0.05, ** represents p≤0.01, and ***represents p≤0.001.</p><!><p>In the control, oil-treated, rat ventral prostate tissue, the ARSB activity and the GALNS activity both increased significantly between day 5 and day 30, (p<0.001, 1-way ANOVA with Tukey-Kramer post-test, n=6 in each group) (Table 1; Fig. 1A, 1B). In contrast, the steroid sulfatase (STS) activity was similar at both time points (Fig. 1C). In association with the increases in ARSB and GALNS, the total sulfated glycosaminoglycans (GAGs) and the chondroitin-4-sulfate (C4S) both declined significantly (p<0.001, p<0.001) (Fig. 2A, 2B). The extracellular matrix proteoglycan versican, which has chondroitin sulfate attachments, also declined significantly between days 5 and 30, corresponding to the decline in sGAG and C4S (Fig. 2C) (p<0.001).</p><!><p>Following treatment with estradiol benzoate (25 µg / 25 µl sesame oil) on days 1, 3, and 5, the ARSB activity in the prostate tissue was ~14% less on day 5 (p<0.05) in the estrogen-treated than in the control tissue (Fig. 1A; Table 1). On day 30, ARSB activity was similar to the day 5 level in the estrogen-treated tissue, but was ~35% less than in the control tissue (p<0.001). In contrast, the increase in GALNS activity between days 5 and 30 that occurred in the control tissue also occurred following estrogen, and the day 30 GALNS activity was ~12% higher following estrogen than in the control tissue (p<0.01) (Fig. 1B). Following estrogen, steroid sulfatase (STS) activity declined to ~50% of the control level on both days 5 and 30 (p<0.001) (Fig. 1C), and was similar at both time points.</p><p>ARSB is the enzyme which regulates chondroitin-4-sulfate degradation, so a decline in ARSB activity is anticipated to be associated with increased total sGAG and C4S. Hence, following estrogen exposure and decline in ARSB activity, the normal maturation-associated declines in total sGAG and in C4S between day 5 and day 30 did not occur (p<0.001, p<0.001) (Fig. 2A, 2B). Levels of total sGAG and C4S were similar on days 5 and 30 following estrogen exposure suggesting that the maximum effect was achieved by day 5. The increase in C4S accounted for ~45% of the total increase in sGAG. In a similar manner, the content of versican in the estrogen-exposed ventral prostate tissue did not decline with maturation between days 5 to day 30. While the versican level in the estrogen exposed ventral lobe on day 5 was the same as levels in the control animal, it was 35% greater than the control value on day 30 (p<0.001).</p><!><p>The prostate arises in the urogenital sinus from solid epithelial cords, which are canalized in a proximal to distal direction, beginning on day 1 in the rat ventral prostate. Images of rat ventral prostate on days 1, 3, 6, 10, 15, and 30 present the evolving distinct localization of ARSB (brown) in the stroma and in the luminal membrane of the epithelial cells (Figs. 3A-3F). Reciprocally, GALNS appears more prominent in the developing nests of epithelium (Days 1, 3, 6, and 10), which are negative for ARSB on days 3 and 6 (Figure 3B,C) and positive for GALNS (brown in Figure 4A-4F). ARSB continues to be prominent in the stroma throughout days 1–30, whereas the GALNS staining in the stroma declines, with increased localization in the developing acinar epithelium. An overall increase in intensity and distribution of ARSB is evident between the early and late control images, consistent with the rise in ARSB activity between days 5 and 30 in the control tissue. Similarly, the epithelial GALNS staining intensity appears increased in the Day 30 images, consistent with increased GALNS activity.</p><!><p>Prostate development following estrogen exposure (estradiol benzoate 25 µg on days 1, 3, and 5) is presented in Figs. 5A-5F and 6A-6F. Delay in acinar development post estrogen exposure appears at 15 days (Fig. 5E vs. Fig. 3E) with ARSB immunostaining on day 30 (Fig. 5F) post estrogen resembling vehicle control on day 15 (Fig. 3E). With GALNS immunostaining, the lumen and acinar structures on day 15 without estrogen (Fig. 4E) appear more advanced than on day 15 following estrogen exposure (Fig. 6E). In the estrogen-treated ventral prostates, the stromal compartment has relatively reduced immunostaining for ARSB, evident on all days 1–30 (Fig. 5A-5F), compared to the day-matched vehicle control (Fig. 3A-3F), and consistent with the measured decline in ARSB activity between days 5 and 30. Negative ARSB immunostaining of epithelium is best seen in control on day 3 (Fig. 3B) and in estrogen-treated on day 6 (Fig. 5C), consistent with a developmental delay. Luminal membrane appears positive for both ARSB and GALNS on day 30 (Figs. 5F6F). These findings suggest that the decline in stromal ARSB delays the development of the acinar structures, indicating impaired stromal-epithelial interaction following exogenous estrogen.</p><!><p>This study presents new information about activity and expression of sulfatases in rat prostate development, and the impact of estrogen exposure on sulfatase activity, as well as on content of total sulfated glycosaminoglycans, chondroitin-4-sulfate, and the extracellular matrix proteoglycan versican. In the normal developing rat ventral prostate, ARSB activity and GALNS activity increased between days 5 and 30, with an associated decline in the total sulfated GAGs, C4S, and versican. In contrast, steroid sulfatase activity was similar at these two time points. Following exposure to estradiol benzoate (25 µg subcutaneous injection on days 1, 3, and 5), ARSB activity was 14% less than vehicle control on day 5 and 35% less than vehicle control on day 30. In contrast, GALNS activity increased in both control and estrogen-treated tissue between days 5 and 30, and was slightly higher in the estrogen-treated tissue. Steroid sulfatase activity was reduced by ~50% following estrogen treatment at both time points. Total sulfated GAGs, C4S, and the extracellular matrix proteoglycan versican were significantly increased at 30 days, compared to vehicle controls. Hence, neonatal exposure to this dose and preparation of estrogen inhibited the normal increase in ARSB, suppressed STS, and increased C4S, total sulfated GAGs, and versican. These responses suggest that some of the mechanisms by which estrogen produces its effects in prostatic tissue may be mediated through changes in sulfatases and chondroitin sulfate.</p><p>Since estrogen receptor alpha (ERα) is present in the stromal tissue of the developing prostate [23, 24], but not in the epithelial cells, the decline in ARSB is consistent with a response to estrogen in the stromal cells where ARSB predominates, as shown by immunostaining. In contrast, the increase in GALNS is not inhibited by estrogen, since the epithelial cells, where GALNS immunostaining predominates, does not have ERs. The reciprocal localization of ARSB and GALNS immunostaining seen in the developing rat ventral prostate indicates that these sulfatase enzymes may have an impact upon the interaction between stromal and epithelial compartments in prostate morphogenesis. Disruption of their normal coordinated increases by estrogen may contribute to disordered development of the prostate. Disruption of normal prostate development can lead to irreversible morphogenic defects which affect the function of the prostate throughout life, as evidenced by the impact of aberrant estrogenic exposures during early development, predisposing the prostate to chronic inflammation and neoplasia with aging [13, 16, 25].</p><p>The dose of estradiol used in the studies of this report is intermediate in the range of estradiol benzoate previously tested, which included 0.015 ug/kg body weight to 15 mg/kg body weight [26]. Changes in body weight or in relative weight of the ventral prostate lobe (ventral lobe in mg/ 100 mg body weight) did not vary in a dose-dependent manner when the rats were euthanized on day 35 or day 90, following subcutaneous injections on days 1, 3, and 5. No significant change in the relative weight of the ventral prostate lobe was evident with an estradiol benzoate (EB) dose of 150 ug/kg body weight on days 1, 3, and 5, which is comparable to the EB dose used in the current studies (25 µg / ~160 g body weight). Body weight was significantly reduced on day 35 (183.8 ± 5.59 g for vehicle control vs. 161.6 ± 7.2 g for EB dose of 150 µg/kg body weight, n=8 for each group; p<0.001) or on day 90 following EB, compared to vehicle control. At the low dose (0.15 µg/kg body weight), EB did not modify the presence or localization of androgen receptors, the number of basal cells, the number of differentiated luminal cells, or the characteristics of periacinar and perivascular smooth muscle cells. However, increasing concentrations of EB caused progressive decline in the number of androgen receptor positive cells, increasing numbers of basal cells, decline in luminal epithelial cells, and increased extension of the multilayered fibroblast sheath into the distal tips of the prostatic ducts, so that smooth muscle cells were no longer in contact with the epithelium [26].</p><p>Prostate gland development involves significant remodeling of tissue architecture that includes reorganization of stromal and epithelial cells, basement membranes and extracellular matrices (ECM) [12]. Neonatal estrogen exposure perturbs the normal stromal-epithelial development [14, 16, 25–28], such as suggested by the changes in sulfatases, sulfated GAGs, and versican in our studies. Following neonatal estrogen exposure in this study, the ARSB immunostaining in the stroma was less intense. When ARSB was reduced following estrogen exposure, C4S and sGAG increased, due to decline in their normal degradation which requires ARSB. The increases in total sGAG and C4S in the prostate tissue when ARSB was reduced may contribute to impaired morphogenesis, due to inhibition of the normal stromal-epithelial signaling of chemokines and growth factors by the increased sGAG content [29–31]. The mechanism by which estrogen exposure leads to reduced ARSB is unknown at this time, and the extent to which decline in ARSB (with the associated increase in C4S) and in STS activity contributes to the overall impact of estrogen on the prostate is not yet discernible. The decline in versican, however, is likely attributable to a transcriptional mechanism which follows from the decline in ARSB and the associated increase in chondroitin-4-sulfation. Galectin-3, a β-galactoside-associated lectin, binds less to more highly sulfated C4S when ARSB is reduced, and is more abundant in the nucleus, where it activates the versican promoter in association with AP-1 components c-jun and c-fos [3, 32].</p><p>The effects of ARSB and of GALNS in prostate development have not been reported previously, although changes in chondroitin sulfates (CS) and in versican have been studied in prostate development [33, 34]. When guinea pigs were given estradiol, the CS was increased compared with the post-pubertal level, similar to the present findings in the estrogen-treated neonatal rat prostate. In contrast, when circulating dihydrotestosterone increased at puberty, periglandular and fibromuscular stromal staining intensity for native total chondroitin sulfate (immunostained with CS56 antibody) decreased 11-fold in the guinea pig prostate. Similarly, chondroitin sulfate declined in castrated guinea pigs given DHT replacement. Sakko et al hypothesized that both chondroitin sulfate and versican expression were negatively regulated by androgen, and showed that versican in the periacinar fibromuscular stroma peripheral to the basal epithelial cells as well as in loose fibrovascular connective tissue between acini decreased during normal pubertal development in association with increased circulating levels of androgen [34].</p><p>Other proteoglycans, including small-leucine rich proteoglycans (decorin, biglycan, and lumican), syndecan-1, betaglycan, and perlecan are also present in the prostate [35]. The GAG attachments of the proteoglycans affect the binding of ligands, including fibroblast growth factor (FGF)2 and FGF10, which bind to heparan sulfate (HS). Perlecan, which has both chondroitin sulfate and heparan sulfate attachments, has been proposed as a likely candidate for the HS proteoglycan that is active during prostate development [35]. Other sulfatases, including iduronate 2-sulfatase and Sulf1, have been identified in prostate tissue. The enzyme iduronate 2-sulfatase, which removes 2-sulfate groups from iduronate residues of HS or heparin, was one of 13 genes overexpressed in 16-month-old rat ventral prostate tissue compared with three-month-old rat tissue, suggesting a role in aging prostate tissue [36]. In cultured urogenital sinus (UGS) tissues from mouse embryos at different stages of development, expression of Sulfatase 1 (Sulf1), the enzyme that catalyses the hydrolysis of 6-O sulfates from heparan sulfate, was significantly decreased in male compared to female UGS as development progressed [37, 38]. This decrease was correlated with morphological changes in prostatic epithelial bud outgrowth, and interruption of HS 6-sulfation by ectopic expression of Sulf1 partially inhibited the testosterone-induced ductal morphogenesis and impaired FGF10 signaling. This further supports the present data and indicates a specific role for multiple sulfatases in tissue remodeling necessary for the development of normal prostate gland architecture.</p><p>Future work will be designed to integrate the findings in the current report to other observations about the programmed development of the prostate and the role of sulfatases, GAGs, and proteoglycans. Further evaluation of the interrelationships among ARSB, GALNS, C4S, C6S, and versican with estrogens and testosterone are anticipated to provide new insights into stromal-epithelial interactions in the developing prostate and into the mechanisms that affect cell fate determinations in prostate carcinogenesis.</p><!><p>Study findings demonstrate significant differences between the localization and the response to estrogen of the chondroitin sulfate modifying enzymes ARSB and GALNS in the developing ventral rat prostate. Immunohistochemistry demonstrated predominance of ARSB in the cytoplasm of the stroma and of GALNS in the cytoplasm of the epithelium. Between days 5 and 30, in the vehicle-only controls, ARSB activity increased, GALNS activity increased, STS did not change, and total sulfated GAGs, C4S, and versican declined. Following estrogen exposure, ARSB activity declined significantly, GALNS increased slightly, STS declined by almost 50%, and total sulfated GAGs, C4S, and versican all increased significantly. These results indicate that ARSB and GALNS are important in the development of the normal prostate architecture, and that some of the mechanisms by which estrogen produces its effects in the prostate may be mediated through changes in sulfatases and chondroitin sulfates.</p><!><p>This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.</p>

PubMed Author Manuscript

Real-world use of sunitinib in Japanese patients with pancreatic neuroendocrine tumors: results from a post-marketing surveillance study

BackgroundSunitinib is approved for the treatment of progressive, well-differentiated pancreatic neuroendocrine tumors (pNETs) in patients with unresectable, locally advanced or metastatic disease. Safety and efficacy data in Japanese patients are limited. We report outcomes from a post-marketing surveillance study of sunitinib treatment in Japanese patients.MethodsSunitinib 37.5 mg once daily was orally administered in Japanese patients aged ≥ 15 years with pNETs. The primary endpoints included adverse events (AEs) occurring during the observation period of 168 days and objective response rate (ORR).ResultsSunitinib was administered in 62 patients with pNETs. The median duration of treatment was 165 days. At 168 days from the start of treatment, 31 patients were still receiving sunitinib treatment and treatment continuation rate was 50.0%. Of the 31 patients who discontinued treatment, 18 (58.1%) discontinued because of AEs and 16 (51.6%) patients discontinued due to insufficient clinical effect. Of the 18 patients who discontinued due to AEs, 10 did so within 42 days of treatment initiation. The most common all-grade AEs were platelet count decreased (33.9%), diarrhea (29.0%), neutrophil count decreased (27.4%), hypertension (24.2%), and palmar-plantar erythrodysesthesia syndrome (24.2%). In the 51 patients eligible for the efficacy analysis, ORR was 13.7% (95% confidence interval, 5.7–26.3) and clinical benefit rate was 70.6%.ConclusionsThere were no new safety concerns in real-world use of sunitinib in Japanese patients with pNETs. The short treatment duration likely led to low tumor response. Appropriate AEs management through dose interruption/reduction is essential for sunitinib treatment success in this patient population.

real-world_use_of_sunitinib_in_japanese_patients_with_pancreatic_neuroendocrine_tumors:_results_from

Introduction<!>Study design and treatment<!>Analysis plan<!>Patients and treatments<!><!>Patients and treatments<!><!>Patients and treatments<!>Safety<!><!>Safety<!><!>Efficacy<!><!>Discussion

<p>Pancreatic neuroendocrine tumors (pNETs) are considered rare, but their incidence worldwide is increasing annually [1–4]. In Japan alone, a 20% increase in the number of patients treated for pNETs was recorded in the year 2010 compared with 2005 [5]. The overall prevalence of pNETs in Japan in 2010 was 2.69 per 100,000 population [5].</p><p>Surgery is the standard of care for localized pNETs [6]; however, many patients are diagnosed at late stage with advanced or unresectable metastatic disease whereby surgery is not always an option [4, 7]. Prognosis for patients with pNETs is dependent on the histology and disease stage at diagnosis. For patients with well- or moderately-differentiated pNETs, 5-year survival rate is 79% for localized disease, 62% for regional disease, and 27% in patients with distant metastases [3]. The outcome in patients with poorly differentiated pNETs is worse, depending on the disease stage [3].</p><p>The clinical outcomes in patients with unresectable metastatic pNETs have improved with the introduction of new therapies, including everolimus (AFINITOR; Novartis Pharmaceuticals, East Hanover, NJ, USA), sunitinib (SUTENT; Pfizer Inc, New York, NY, USA), and streptozotocin (ZANOSAR; Teva Pharmaceuticals, North Wales, UK) [8–12]. All three agents are available for use in Japan.</p><p>Sunitinib is a potent inhibitor of multiple tyrosine kinase receptors, including the vascular endothelial growth factor receptors (VEGFR) and platelet-derived growth factor receptors (PDGFR) that are essential for pNETs proliferation and angiogenesis [13–15]. In a randomized phase III trial, sunitinib (37.5 mg once daily) improved progression-free survival compared with placebo in patients with advanced, well-differentiated pNETs [hazard ratio (HR) 0.42; 95% confidence interval (CI) 0.26–0.66; P = 0.0001] [8]. Commonly reported adverse events (AEs) were manageable and consistent with the known safety profile of sunitinib. An updated analysis of overall survival (OS) at 5 years after study closure favored sunitinib vs. placebo (HR 0.73; 95% CI 0.50–1.06; P = 0.094), despite that most patients (69%) in the placebo arm crossed over to sunitinib [16].</p><p>Sunitinib has also demonstrated antitumor activity and a manageable safety profile in Japanese patients with pNETs [11]. In a phase II trial in 12 Japanese patients, sunitinib demonstrated antitumor activity, with an objective response rate (ORR) of 50% (95% CI 21.1–78.9), a median PFS 16.8 months (95% CI 9.3–26.2), and a safety profile similar to that shown in the global trial [11, 17]. As a result, sunitinib was approved globally, including Japan, for the treatment of progressive, well-differentiated pNETs in patients with unresectable, locally advanced or metastatic disease [18].</p><p>At the time of sunitinib approval for treatment of pNETs in Japan, safety and efficacy data in Japanese patients were limited. Therefore, the Japan Pharmaceuticals and Medical Devices Agency requested the conduct of a post-marketing surveillance (PMS) study to expand the safety database and ensure appropriate use of sunitinib in the Japanese population. Here, we report the outcomes from this PMS study, including adverse drug reactions and efficacy associated with sunitinib in Japanese patients with pNETs, and how sunitinib treatment is managed in clinical practice in Japan.</p><!><p>This was a PMS study of sunitinib in patients aged ≥ 15 years with pNETs. Sunitinib was orally administered at a starting dose of 37.5 mg once daily (using 12.5 mg capsules). Dose increase (to a maximum of 50 mg once daily) or decrease were permitted depending on patient tolerance. The study was conducted between August 10, 2012 (the date of sunitinib approval in Japan) and February 2017 in 20 centers in Japan specializing in the treatment of pNETs.</p><p>The investigation consisted of two periods, registration and observation. The registration period continued until the target number of 60 patients was achieved or for 4 years, whichever was earlier. The observation period was 168 days (24 weeks) from the first day of sunitinib administration. Patients were observed until treatment completion or discontinuation. During the observation period, investigators were required to document information about sunitinib administration (daily dose and frequency, administration period, reasons for dose modifications) and discontinuation, with reasons for discontinuation.</p><p>Safety assessments included AEs (graded according to the National Cancer Institute Common Terminology Criteria for Adverse Events v4.0), laboratory assessments (hematology and biochemistry), blood pressure, pregnancy, and concomitant therapy. Efficacy was assessed by the investigators using the Response Evaluation Criteria in Solid Tumors revised (RECIST) version 1.1. This study was performed in compliance with Ministry of Health, Labour and Welfare (MHLW) Good Post-marketing Study Practice for drugs (MHLW Ministerial Ordinance No. 171, dated December 20, 2004). Patient data collected from this investigation were reported to MHLW according to the Pharmaceutical Affairs Law.</p><!><p>Safety analysis population included all eligible patients who received at least one dose of sunitinib. Efficacy analysis population included patients with at least one measurable lesion who underwent efficacy assessment. The primary safety endpoint of this analysis was the occurrence of adverse drug reactions during the 168-day observation period from the first administration of sunitinib. The incidence of notable adverse reactions with sunitinib was also examined. These notable adverse reactions included: (1) platelet count decreased, white blood cell count decreased, anemia, and other bone marrow suppression; (2) gastrointestinal disorders; (3) hypertension; (4) cutaneous symptom (hand–foot syndrome); (5) abnormal liver function; and (6) hypothyroidism.</p><p>The primary efficacy endpoint was ORR, defined as the percentage of patients in the efficacy analysis population who achieved complete response or partial response. Efficacy was assessed from initiation of sunitinib administration until the date of response determination. Subgroup ORR analyses by baseline factors were conducted and included: sex, age, weight, body surface area, body mass index, clinical symptoms of pNETs, presence/absence of metastasis (liver, lymph nodes, peritoneum, bone), general condition [Eastern Cooperative Oncology Group performance status (ECOG PS)], liver function disorder, renal impairment, history of previous treatment for pNETs, history of pharmacotherapy, and initial dose per day.</p><!><p>Between August 2012 and September 2015, 62 patients were registered in 17 clinical centers. All 62 patients were included in the safety analysis and 51 patients were included in the efficacy analysis. Ten patients were excluded from the efficacy analysis due to indeterminate efficacy evaluation (n = 10) or not meeting criteria for efficacy (n = 1). Patient demographics and baseline characteristics are summarized in Table 1.</p><!><p>Patient demographic and baseline characteristics</p><p>ECOG PS Eastern Cooperative Oncology Group performance status, pNETs pancreatic neuroendocrine tumors, SD standard deviation</p><!><p>Sunitinib was administered for > 56 and ≤ 168 days in 48 (77.4%) patients and ≤ 56 days in 14 (22.6%) patients. Median (range) duration of treatment was 165 (4–168) days, and the mean ± SD total administered dose of sunitinib was 2711.49 ± 1722.159 mg. At 168 days from the start of treatment, 31 patients were still receiving sunitinib treatment and treatment continuation rate was 50.0% [95% confidence interval (CI) 37.1–61.6; Fig. 1]. Fourteen (22.6%) patients discontinued treatment ≤ 56 days, and 17 (35.4%) patients discontinued treatment between > 56 and ≤ 168 days. Discontinuation by duration of administration period is presented in Table 2.</p><!><p>Duration of treatment</p><p>Discontinuation by duration of administration period</p><p>AE adverse event</p><p>aNo patient had dose interruption</p><p>bNo patient discontinued because of death or loss to follow up</p><p>cPatient could have had more than one reason for discontinuing</p><p>dThe end of observation period of 168 days (24 weeks)</p><!><p>Of the 31 patients who discontinued treatment, 18 (58.1%) discontinued because of AEs occurrence and 16 (51.6%) due to insufficient clinical effect; patients could have had more than one reason for discontinuing. Of the patients who discontinued due to AEs, 10 discontinued within 42 days of treatment initiation.</p><p>The initial sunitinib dose was 37.5 mg/day in 43 (69.4%) patients, 25 mg/day in 16 (25.8%) patients, and 12.5 mg/day in three (4.8%) patients. Dose reduction occurred in 26 of 62 patients (41.9%), which included 23 of 43 patients administered a starting dose of 37.5 mg and three of 16 patients with starting dose of 25 mg. Per prescribing information, none of the 62 patients had cytokine P450 3A4 inhibitors co-administered with sunitinib [18].</p><!><p>Overall, 300 all-grade AEs occurred in 59 (95.2%) patients and 57 grade ≥ 3 AEs occurred in 30 (48.4%) patients. The most commonly occurring all-grade AEs (Table 3) were platelet count decreased (33.9%), followed by diarrhea (29.0%), neutrophil count decreased (27.4%), hypertension (24.2%), and palmar-plantar erythrodysesthesia syndrome (24.2%). The most commonly (> 5% of patients) occurring grade ≥ 3 AEs (Table 3) were neutrophil count decreased (16.1%), platelet count decreased (14.5%), hypertension (6.5%), and white blood cell count decreased (6.5%).</p><!><p>Adverse events occurring in ≥ 15% of patients</p><p>AE adverse event, CTCAE Common Terminology Criteria for Adverse Events, MedDRA Medical Dictionary for Regulatory Activities, PPE palmar-plantar erythrodysesthesia syndrome</p><p>aAccording to MedDRA (version 18.1) coding dictionary and CTCAE (version 4.0)</p><p>bThere were no grade 5 AEs</p><!><p>AEs leading to permanent treatment discontinuation included: malaise (n = 3); decreased appetite, diarrhea, abnormal liver function, and platelet count decreased (n = 2, each); and influenza, hypothyroidism, heart failure, mitral insufficiency, aortic dissection, nausea, duodenal perforation, gastrointestinal perforation, vomiting, hepatic disorder, rash, fever, aspartate aminotransferase increased, alanine aminotransferase increased, and neutrophil count decreased (n = 1, each).</p><p>Notable adverse reactions were first observed within 42 days (especially from day 14 to day 42) after the initiation of sunitinib administration (Fig. 2). There was no characteristic adverse reaction with a late-onset tendency. There were 22 serious AEs that occurred in 13 (21.0%) patients. The only serious AE occurring in two or more patients was diarrhea (two patients, 3.2%), the outcome of both of these AEs was disappeared/resolved.</p><!><p>Time to onset of notable adverse reactions</p><!><p>In the 51 patients eligible for the efficacy analysis, the ORR was 13.7% (95% CI 5.7–26.3). Clinical benefit rate, defined as complete response plus partial response plus stable disease, was 70.6% (95% CI 56.2–82.5; Table 4). The ORR analyzed according to ECOG PS before the start of sunitinib administration were 15.6% for patients with ECOG PS 0, 11.8% for patients with ECOG PS 1, and 0% for patients with ECOG PS 2 (Table 4). The response rate was 16.7% (n = 3) among 18 elderly (≥ 65 years old) patients and 12.1% (n = 4) among 33 non-elderly (< 65 years old) patients. Analysis by baseline factors revealed no major tendencies in efficacy.</p><!><p>Best overall response, efficacy analysis set (N = 51)</p><p>CI confidence interval, CR complete response, ECOG PS Eastern Cooperative Oncology Group performance status, PD progressive disease, PR partial response, SD stable disease</p><p>aThere were no patients with ECOG PS 3–5</p><!><p>The results of this PMS study showed the safety profile of sunitinib in Japanese patients with pNETs treated in clinical practice was similar to the safety profile in patients treated in clinical trials, including the global phase III and IV trials and the phase II trial in Japanese patients [8, 11, 19]. Furthermore, there were no new safety concerns in sunitinib-treated Japanese patients with pNETs, and the commonly reported AEs were similar to those previously reported in clinical trials and surveillance studies of sunitinib in patients with metastatic renal cell carcinoma and gastrointestinal stromal tumors [20–24].</p><p>Notably, of the 31 patients who discontinued sunitinib treatment, 18 (58%) discontinued because of AEs; of these, 10 (56%) patients discontinued as early as within 42 days after the start of sunitinib administration. This suggests the Japanese clinicians decided on treatment discontinuation at the initial occurrence of AEs without considering temporary dose interruption or dose reduction to manage the AEs. As a result, the median duration of treatment was short (165 days) and ORR was low (13.7%). In fact, the median duration of treatment in this PMS study was shorter than in the global phase IV trial (356 days) or the real-world clinical setting of the Japanese clinical study (490 days) [19, 25]. The longer treatment duration in those trials likely led to the better outcomes: ORR was 24.5% in the global phase IV trial and 44% in the Japanese study [19, 25]. Keeping patients on treatment longer, by appropriately managing AEs at early stages, is essential for achieving better clinical outcome in sunitinib-treated patients with pNETs.</p><p>The most commonly reported AEs in this Japanese PMS study were similar to those reported in the phase III pivotal trial, the global IV study, and the phase II Japanese trial [8, 11, 19]. The most common AEs of bone marrow depression observed in this Japanese PMS study were lower than those reported in the global phase IV study and included neutrophil count decreased (neutropenia in the global phase IV) (27.4% vs. 53.8%) and white blood cell count decreased (leukopenia in the global phase IV) (19.4% vs. 43.4%), respectively [19]. One potential reason for the lower frequency of these AEs in this Japanese PMS study compared with the global phase IV trial is the shorter treatment duration (165 vs. 490 days, respectively) [19].</p><p>The AEs leading to treatment discontinuation included malaise, decreased appetite, diarrhea, abnormal liver function, and platelet count decreased, and most were of grade 1–2. Moreover, many of the AEs occurred between 14 and 42 days after the start of sunitinib administration. Therefore, anticipating the occurrence of AEs and their timing, and managing AEs by dose interruption/reduction could have resulted in longer duration of sunitinib treatment and better clinical outcome. In fact, as a result of AEs management by dose interruption/reduction in the Japanese phase II trial and Japanese real-world clinical setting, respectively, continued long-term administration of sunitinib (median, 298 days and 490 days) was achieved and ORR was high (50% and 44%) [11, 25].</p><p>In conclusion, there were no new safety concerns with sunitinib in this post-marketing study in Japanese patients with pNETs. Continuation of sunitinib administration for as long as possible leads to improved prognosis in patients with pNETs. Therefore, anticipation and management of AEs through appropriate dose interruption/reduction in actual clinical settings is essential for treatment success with sunitinib in patients with pNETs.</p>

PubMed Open Access

Pentamericthiophene based ligands that spectrally discriminate amyloid-\xce\xb2 and tau aggregates display distinct solvatochromism and viscosity induced spectral shifts

A wide range of neurodegenerative diseases are characterized by the deposition of multiple protein aggregates. Ligands for molecular characterization and discrimination of these pathological hallmarks are thus important for understanding their potential role in pathogenesis as well as for clinical diagnosis of the disease. In this regard, luminescent conjugated oligothiophenes (LCOs) have proven useful for spectral discrimination of amyloid-beta (A\xce\xb2) and tau neurofibrillary tangles (NFTs), two of the pathological hallmarks associated with Alzheimer\xe2\x80\x99s disease. Herein, the solvatochromism of a library of anionic pentamericthiophene-based ligands, as well as their ability to spectrally discriminate A\xce\xb2 and tau aggregates, were investigated. Overall, the results from this study identified distinct solvatochromic and viscosity-dependent behavior of thiophene-based ligands that can be applied as indices to direct the chemical design of improved LCOs for spectral separation of A\xce\xb2 and tau aggregates in brain tissue sections. The results also suggest that the observed spectral transitions of the ligands are due to their ability to conform by induced fit to specific microenvironments within the binding interface of each particular protein aggregate. We foresee that these findings might aid in the chemical design of thiophene-based ligands that are increasingly selective for distinct disease-associated protein aggregates.

pentamericthiophene_based_ligands_that_spectrally_discriminate_amyloid-\xce\xb2_and_tau_aggregates_d

Introduction<!>Solvatochromism of a library of anionic LCOs<!>Viscosity dependent excitation- and emission profiles of anionic LCOs<!>Synthesis and evaluation of three additional LCOs<!>Conclusion<!>Experimental Section

<p>Conjugated polymers have many unique photophysical properties which render them useful in a variety of applications within the fields of chemistry, molecular biology, and medicine. Luminescent conjugated polymers (LCPs) account for a growing number of developing sensors and probes as their unique properties make them useful reporters in the detection of ions, DNA, and proteins, to name just a few.[1–4] These sensors mainly employ the efficient light harvesting properties or the conformation-sensitive optical properties of the LCPs. The latter is particularly observed for LCPs with a repetitive flexible thiophene backbone, as conformational restriction of the thiophene rings leads to a distinct optical fingerprint.[5]</p><p>In recent studies, it has been shown that LCPs can function as target-specific chameleons that change color depending on the structural motif of the target molecule, even in complex samples such as tissue sections.[6–8] This intrinsic property of LCPs make them useful as selective probes for identifying and distinguishing protein deposits consisting mainly of fibrils with a repetitive cross-beta sheet structure. Accumulation of such proteinaceous deposits is the histopathological hallmark of several devastating diseases, including Alzheimer's disease (AD) and prion diseases.[9,10] In addition, novel chemically defined thiophene scaffolds, denoted luminescent conjugated oligothiophenes (LCOs), have been utilized as specific ligands for a variety of disease associated protein aggregates, as well as for optical in vivo imaging of protein aggregates in real time.[11–16]</p><p>Anionic LCOs have proven particulary useful for spectral discrimination of amyloid beta (Aβ) deposits and tau neurofibrillary tangles (NFTs), the two major pathological hallmarks of AD.[11,12,17] Ligands with molecular scaffolds other than thiophene showing selectivity towards tau or Aβ deposits have also been presented and these studies showed that minor chemical alterations of the molecular scaffold could influence the specificity toward either tau or Aβ aggregates.[18,19] Recently, aminonaphthalenyl 2-cyanoacrylate-based probes were also shown to fluorescently discriminate between different types of protein deposits in brain.[20] The discriminating capability of these dyes was due to the stabilization of the ground versus excited states of these probes as a function of the polarity of the binding pocket on the amyloid. Hence, although most protein deposits share a repetitive cross-beta sheet structure, possible differences in the binding pocket microenvironments should be considered when designing ligands towards distinct protein aggregates.</p><p>In previous comparisons of structurally related LCOs, it was suggested that for optimal spectral separation of Aβ deposits and tau tangles, an LCO-based ligand should comprise a conformationally flexible backbone consisting of five to seven thiophene units and terminal carboxyl groups extending the conjugated thiophene backbone.[12,17] Upon binding to protein aggregates, LCOs also exhibit decreased Stokes shifts, red-shifted excitation maxima and blue-shifted emission maxima compared to free dyes in solution.[11,12,17] In an effort to elucidate the specific structural features that contribute to enhanced spectral separation between protein aggregates and the photophysical behavior of LCOs bound to protein deposits in more detail, we herein investigated solvatochromism and the effects of solvent viscosity on a group of structurally similar, pentameric, anionic oligothiophene probes. LCOs that displayed spectral variations for Aβ deposits and NFTs also showed distinct solvatochromism and decreased Stokes shifts due to increased solvent viscosity. Hence, these photophysical assessments might aid in the design of LCOs for sensitive optical discrimination of Aβ and tau deposits.</p><!><p>Solvatochromic behavior of small amyloid ligands can be used to approximate the amyloid fibril binding site polarity, as well as the relative change in the dipole moment for individual ligands.[20–23] Therefore we tested the solvatochromic behavior of a subset of previously reported anionic pentameric LCOs (Figure 1).[11,12,17] These LCOs were chosen as their molecular composition varies in anionic substitution patterns and in backbone rigidity, as well as in terminal functional groups extending the thiophene backbone. To assess the solvent sensitivity of these LCOs, Lippert-Mataga plots (Stokes shift vs orientation polarizability) were used.[24] The relative slopes of the fitted lines allows for a comparison of the solvent sensitivity of each of the LCOs tested by means of the orientation polarizability, Δf, for each solvent, determined by the equation,</p><p> Δf=ε-12ε+1-n2-12n2+1 in which ε is the dielectric constant and n the refractive index of each solvent. The calculated Δfs are shown in Table S1 (Supporting information). The Stokes shifts, Δλ, were given by,</p><p> Δλ=λEX-λEM where λEX and λEM are the wavelengths corresponding to the excitation- or the emission maximum, respectively.</p><p>The results from the solvatochromism study are shown in Figure 1 and Table 1. p-FTAA and HS-72 displayed the highest slope values as these dyes have the highest degree of conformational freedom along the backbone and carboxyl groups extending the thiophene backbone. When the central thiophene ring was replaced with a selenophene (p-FTAA-Se), the relative slope was decreased and still further reduced through the introduction of a central phenyl (p-FTAA-Ph). A reduction in the relative slopes of the fitted lines was also observed when replacing the terminal carboxyl groups with hydrogen (p-HTAA). The combination of chromophore planarization and polarization occurs in natural processes, such as the chemistry of vision,[25,26] and earlier studies[27,28] have also shown that similar phenomena occurs in tetrameric oligothiophenes. Thus, conformational restrictions of the LCO backbone will most likely influence the polarization of the dye, since chromophore planarization and polarization are coupled processes. Likewise, substituting the polarizable carboxyl groups with hydrogens will influence the polarization of the molecule and the solvent sensitivity. Furthermore, the carboxyl groups can also acts as π-acceptors. A previous study,[17] comparing p-FTAA, HS-72, p-FTAA-Se and p-FTAA-Ph, has shown that p-FTAA and HS-72 are efficient for spectral discrimination of Aβ deposits and NFTs, whereas p-FTAA-Se was less effective, and p-FTAA-Ph completely lacked the ability to distinguish the two aggregated species. In addition, carboxyl groups extending the conjugated thiophene backbone have been shown to be an additional molecular determinant for achieving optimal spectral separation of Aβ deposits and NFTs.[12] Overall, the trend in solvatochromic behavior of the LCOs supports a correlation between the solvent sensitivity of the LCOs and their ability for spectral separation of Aβ deposits and tau tangles.</p><p>As solvatochromism Stokes shifts can provide insight regarding protein binding site polarity,[19,21] we next compared the Stokes shifts obtained from the solvatochromism experiments with the Stokes shifts from the LCOs bound to recombinant Aβ1-42 fibrils (Table 1). All the LCOs, except for p-FTAA-Ph, displayed considerably reduced Stokes shifts for Aβ1-42 fibrils compared to ethyl acetate, indicating that the Aβ1-42 binding pocket is substantially more non-polar than ethyl acetate (ε = 6.08). In a recent study,[20] using aminonaphthalene 2-cyanoacrylate dyes, the dielectric constants of the binding pocket of Aβ deposits in tissue was determined to be roughly similar to diethyl ether (ε = 4.27) and in a similar study[23] using Nile Red, the Aβ1-42 fibrils binding site polarity was predicted to have a dielectric constant lower than 8.</p><p>Previous studies indicate that molecules exhibiting abnormally low Stokes shift are most likely undergoing secondary effects, such as hydrogen bonding or binding induced conformational restrictions.[24] Thus, the low Stokes shifts observed for the LCOs bound to recombinant Aβ1-42 fibrils might also be due to such secondary effects. In fact, p-FTAA-Ph displayed a similar Stokes shift for Aβ1-42 fibrils as the solvents and this dye has less conformational freedom than the other LCOs so the dye is most likely prevented from undergoing additional conformational restriction upon binding to the fibrils. Furthermore, in contrast to the solvent shifts, Aβ1-42 binding also caused a significant bathochromic shift in the excitation spectra for all the LCOs except p-FTAA-Ph, indicating that a planarization of the ground state of the probes occurs upon interaction with the fibrils (SI, Figure S1).</p><!><p>As the considerably reduced Stokes shifts observed for LCOs bound to Aβ1-42 fibrils could be due to conformational restrictions within the binding pocket, we proceeded to investigate the role of such restrictions on Stokes shift upon fibril binding using a solvent viscosity model. Solvent viscosity can be used to assess probe behavior based on conformational restrictions,[24,29] since highly viscous solvents exhibit slow solvent reorientation, restrict conformational freedom, and reduce vibrational modes of relaxation, resulting in emission before complete solvent reorientation and a hypsochromic shift of the probe's emission spectra. In some cases, increasing viscosity can also stabilize the ground state, causing a bathochromic shift in the excitation spectra and further reducing the Stokes shift. By using two solvents of similar polarity, (ethylene glycol and glycerol), but with differing viscosities, we were able to simulate the conformational restrictions with minimal changes in polarity. For p-FTAA and HS-72, increasing the ratio of glycerol to ethylene glycol resulted in a red-shift of the excitation spectrum and blue-shift of the emission spectrum, (Figure 2). Hence, both of these dyes displayed a viscosity-dependent decrease in their Stokes shifts and considerable Stokes-viscosity slopes, when plotting the Stokes shifts versus percent of glycerol (Figure 1, Table 1). p-HTAA also displayed a substantial Stokes-viscosity slope, although the blue-shift in the emission spectrum was less pronounced. In contrast, p-FTAA-Se and p-FTAA-Ph, displayed minor Stokes-viscosity slopes. As mentioned above, p-FTAA-Ph with a central phenyl ring has restricted degrees of conformational freedom and for p-FTAA-Se, replacement of the sulfur atom with selenium in the central ring causes the central portion of the backbone to be more planar and less flexible. Compared to the solvatochromism experiments, the viscosity measurements better reflect Stokes shift changes upon conformational restriction of the molecule. Here it becomes more apparent that p-FTAA-Ph and p-FTAA-Se, displayed reduced chromic responsiveness due to inherent conformational restrictions of the conjugated backbone.</p><p>Although the effort of increasing the conformational restrictions via viscosity reduced the Stokes shift, the spectral changes that occur upon binding to Aβ1-42 fibrils cannot be reproduced by solvent-only restrictions (Figure 2, Table 1). The occurrence of vibronic peaks, the bathochromic excitation shift, and the hypsochromic emission shift all indicate that the LCOs are much more constrained in the ground state upon binding to Aβ1-42 fibrils, even when compared to the most viscous condition of 100% glycerol. These signature peaks in both the excitation and emission spectra when bound to Aβ1-42 fibrils can be observed for all of the probes with the exception of p-FTAA-Ph. In addition, similar to the fluorescent amyloid ligand thioflavin T (ThT), all the LCOs displayed an enhanced fluorescence upon binding to amyloid fibrils (SI, Figure S2). Recent studies indicate that the binding of ThT to amyloid fibrils are highly dependent on interactions with the aromatic and hydrophobic sides chains of the protein fibrils.[30,31] Thus, ThT fluorescence is highly sensitive to local interactions occurring when the dye is bound to amyloid fibrils. In addition, it was recently shown that efficient binding of the most conventionally used amyloid ligand, Congo Red, to amyloid fibrils is highly dependent on electrostatic interactions and hydrogen bonding.[32] Most likely such interactions are also relevant for the conformational restriction of the anionic LCOs upon binding to Aβ1-42 fibrils and the interplay of these interactions cannot be completely mimicked by the solvent-only models presented above. However, the solvatochromism and viscosity indices did correlate with the pentameric LCOs optical ability to distinguish Aβ and tau deposits. As reported earlier,[12,17] LCOs having carboxyl groups extending the pentamericthiophene backbone, as well as having greater conformational freedom, tend to perform better as probes for the detection of conformational differences between protein aggregates. As shown herein, such LCOs also displayed distinct solvatochromism and decreased Stokes shifts due to increased solvent viscosity. These fundamental photophysical assessments might thus be utilized to predict the LCOs ability to act as sensitive optical discriminators of Aβ and tau deposits.</p><!><p>In order to test our hypothesis that pentameric LCOs for optimal spectral discrimination of Aβ deposits and NFTs should display distinct solvatochromism as well as spectral shifts due to solvent viscosity, three novel anionic pentameric LCO analogues to p-FTAA were synthesized (Figure 3). Firstly, the terminal carboxyl groups were replaced by ketones, resulting in p-KTAA, an anionic pentamer with the same central trimer building block as p-FTAA and neutral polarizable π-acceptor groups (ketones) extending the thiophene backbone instead of negatively charged carboxyl groups. Secondly, the positions of the acetic acid side chains were altered on the trimer building block to render HS-84, an isomer to p-FTAA having the acetic side chains of the trimeric building block tail-to-tail instead of head-to-head. Thirdly, a pentamer (HS-42) lacking the terminal carboxyl groups extending the conjugated backbone, but displaying the same amount of net charge (-4) as p-FTAA was synthesized.</p><p>The new LCOs were synthesized in a similar fashion as previously reported LCOs.[11,12,17] Thiophenetrimer 1[33] was used as precursor for the synthesis of target compounds p-KTAA and HS-42 (Scheme 1). Electrophilic aromatic substitution on trimer 1 using N-bromosuccinimide in DMF gave dibrominatedthiophenetrimer 2 in 94% yield. Compound 2 was subjected to a Suzuki coupling with 5-acetyl-2-thienylboronic acid (3) using K2CO3 and the palladium-catalyst PEPPSI™-IPr. Due to solubility problems, after workup, the crude methylesterpentamer was subsequently hydrolyzed with 1 M aqueous NaOH in dioxane and water to give p-KTAA in an overall yield of 83 % over two steps. Compound 4[17] was coupled to compound 2 according to the above-mentioned Suzuki conditions affording pentamer 5 in 48 % yield. Hydrolysis with 1 M aqueous NaOH in dioxane and water gave HS-42 quantitatively (Scheme 1). The synthetic approach towards pentamericoligothiophene HS-84 required the dimericthiophene 8 (Scheme 2). This intermediate was prepared according to the same palladium cross-coupling conditions as described above using monomers 6[33] and 7, followed by esterification under acidic conditions using methanol as solvent and nucleophile in an overall yield of 71 % over two steps. Bromination of the intermediate 8 with N-bromosuccinimide in DMF afforded the key precursor 9 in 73 % yield. Following the previous procedure dibrominated dimer 9 was coupled to 2,5-thiophenediylbisboronic acid (10) yielding methyl ester pentamer 11 in 80 %. Final hydrolysis as for HS-42 and p-KTAA gave HS-84 quantitatively (Scheme 2). After synthesis and purification, the solvatochromism and viscosity-dependent spectral changes of the dyes were assessed as described above.</p><p>The solvatochromic Lippert-Mataga plot for the three novel anionic pentameric LCOs are shown in figure 3 and the slope values for the respective LCOs are summarized in Table 2. Similar to p-HTAA, HS-42 also displayed low solvent sensitivity due to the absence of terminal carboxyl groups extending the conjugated thiophene backbone. HS-84 showed a similar solvent sensitivity as p-FTAA, verifying that LCOs having terminal carboxyl groups reveal high solvent sensitivity. In addition, p-KTAA displayed an even higher slope value than p-FTAA, verifying that elongation of the conjugated thiophene backbone with polarizable π-acceptor groups other than carboxyl groups, are possible for achieving LCOs displaying high solvent sensibility. In the viscosity plot, HS-42 and p-KTAA showed high slope values, whereas HS-84 lacked the viscosity-induced changes of the Stokes shift (Figure 3, Table 2). Hence, it appears that HS-84 has less conformational freedom along the backbone than p-FTAA, indicating that changing the position of the acetic acid side chains of the trimeric building block from head-to-head to tail-to-tail might induce alternative intramolecular interactions, such as hydrogen bonding or sulphur-oxygen interactions,[34] resulting in backbone rigidity. In addition, it appears that the positioning of acetic acids moieties on adjacent thiophene rings, as in HS-42, induces an increase in steric hindrance between pendant groups without hindering the flexibility of the conjugated backbone. For all the newly synthesized LCOs, the spectral changes that occur upon binding to Aβ1-42 fibrils could not be reproduced by solvent-only restrictions (SI, Figure S3 and S4). Furthermore, all the three LCOs displayed an enhanced fluorescence upon binding to amyloid fibrils (SI, Figure S5). Overall, the results from the solvent experiments predict p-KTAA to be an efficient LCO for spectral discrimination between Aβ and NFTs, while HS-42 and HS-84 should display similar emission profiles for these aggregated species.</p><p>To verify the three new pentamers' abilities to distinguish between Aβ and tau deposits, the dyes were utilized for staining of human brain tissue with AD pathology (Figure 4). Indeed, p-KTAA, with the highest solvent sensitivity and the most dynamic range of conformational freedom based on viscosity slopes, proved to be most efficient in terms of ability to spectrally distinguish between immunopositive Aβ deposits and NFTs. p-KTAA bound to Aβ-deposits showed an emission maximum around 570 nm, whereas the p-KTAA spectrum from NFTs was red-shifted, having an emission maximum around 595 nm (Figure 4A). Both HS-84 and HS-42 displayed similar emission spectra for the two aggregated entities, verifying that LCOs that are efficient for spectral separation of Aβ deposits and NFTs need to display solvent sensitivity as well as viscosity induced Stokes shifts. Therefore, the results indicate that the indices of solvent sensitivity (solvatochromism) and conformational freedom (viscosity) can be utilized as predictive determinants for achieving superior LCOs for spectral separation of differing protein aggregates.</p><!><p>Herein, we show that LCOs that are able to spectrally distinguish Aβ-deposits and NFTs display distinct solvatochromism as well as viscosity-dependent optical transitions. In addition, the spectral transitions that arise from the LCOs interactions with specific protein aggregates are most likely due to differences in binding pocket polarities and the conformational restrictions of the respective protein aggregates. Overall, we have demonstrated that a combination of basic photophysical assessments can facilitate the chemical design of novel thiophene-based ligands that can distinguish different protein aggregate topologies. The results presented also underline that the microenvironment in the binding pockets of distinct protein aggregates differ and this should be considered when designing protein aggregate specific ligands.</p><!><p>Full experimental details including additional characterization data and NMR spectra of new compounds are given in the Supporting Information.</p>

PubMed Author Manuscript

Catalytic Mechanism of 4-Oxalocrotonate Tautomerase: Significances of Protein-Protein Interactions on Proton Transfer Pathways

4-oxalocrotonate tautomerase (4-OT), a member of tautomerase superfamily, is an essential enzyme in the degradative metabolism pathway occurring in the Krebs cycle. The proton transfer process catalyzed by 4-OT has been explored previously using both experimental and theoretical methods, however, the elaborate catalytic mechanism of 4-OT still remains unsettled. By combining classical molecular mechanics with quantum mechanics, our results demonstrate that the native hexametric 4-OT enzyme, including six protein monomers, must be employed to simulate the proton transfer process in 4-OT due to protein-protein steric and electrostatic interactions. As a consequence, only three out of the six active sites in the 4-OT hexamer are observed to be occupied by three 2-oxo-4-hexenedioates (2o4hex), i.e., half-of-the-sites occupation. This agrees with experimental observations on negative cooperative effect between two adjacent substrates. Two sequential proton transfers occur: one proton from the C3 position of 2o4hex is initially transferred to the nitrogen atom of the general base, Pro1. Subsequently, the same proton is shuttled back to the position C5 of 2o4hex to complete the proton transfer process in 4-OT. During the catalytic reaction, conformational changes (i.e., 1-carboxyl group rotation) of 2o4hex may occur in the 4-OT dimer model but cannot proceed in the hexametric structure. We further explained that the docking process of 2o4hex can influence the specific reactant conformations and an alternative substrate (2-hydroxymuconate) may serve as reactant under a different reaction mechanism than 2o4hex.

catalytic_mechanism_of_4-oxalocrotonate_tautomerase:_significances_of_protein-protein_interactions_o

Introduction<!>Computational Methods for QM/MM Simulations<!>Three Structural Models<!>MD Simulations of Enzyme Complex<!>QM/MM-MFEP Simulations of Proton Transfer Reaction<!>Ketonization mechanism in 4-OT depends on enzymatic models<!>One-proton transfer mechanism is supported in 3D3S and 3D6S<!>Half-of-the-sites occupation number is supported<!>Detailed reaction mechanism is revealed by the 3D3S model<!>2o4hex can have the alternative reactant geometry with different 1-carboxyl group orientation<!>The alternative reactant 2HM may adopt a water-mediated proton transfer mechanism<!>Conclusions

<p>4-oxalocrotonate tautomerase (4-OT), with an amino-terminal proline and a beta-alpha-beta fold, is an important enzyme in tautomerase superfamily1. 4-OT catalyzes the ketonization process of 2-oxo-4-hexenedioate (2o4hex), to its conjugated isomer, 2-oxo-3-hexadienedioate (2o3hex), through the dienol intermediate 2-hydroxymuconate (2HM)2 as shown in Figure 1. This proton transfer process is an essential part of degradative metabolism pathway to convert various aromatic hydrocarbons into their corresponding intermediates in the Krebs cycle3. 4-OT is a unique enzyme in terms of its biological and chemical significances: i) it is one of the smallest subunit enzymes (each subunit has only 62 amino acids4) with the catalytical efficiency around seven orders (the ketonization process has the rate of 3×103 s-1 in 4OT, while its rate is about 1.7×10-4 s-1 in aqueous solution with the reaction barrier around 23 kcal/mol); ii) the proton transfer catalyzed by 4-OT does not involve any co-factors or transition metals; iii) the catalytic residue is an N-terminal proline.</p><p>Experimental2,4-14 and theoretical15-21 studies on 4-OT have been carried out since 1990's. The crystal structure of apo-4-OT is composed of six identical subunits arranged into three dimers (i.e., a hexamer)12. The spatial positions of the six subunits labeled by A to F are shown in Figure 2 with six inhibitors (2-oxo-3-pentynoate). Each subunit contains one active site (i.e. the hydrophobic region surrounding the amino-terminal proline: Pro1). Mutation experiments and theoretical studies have clearly showed that no general acid is required during catalysis. Instead, the amino-terminal proline (Pro1) acts as a general base during proton transfer. Several residues in the active site, such as Arg11′, Arg39″, and Leu8′, play important roles to stabilize the reactant as well as promote the reaction via electrostatic interactions20,22. (Note that the residues with single prime reside in the neighboring subunit of the same dimer while the residues with double primes belong to the nearby subunit of the adjacent dimer.)</p><p>Although previous studies6,12,14-16,18-22 of 4-OT have been helpful to understand the proton transfer process in the tautomerase superfamily, some key issues of this reaction are still obscure. First, are protein-protein interactions between monomers important to the reaction? As shown in Figure 2, six active sites (in A to F) of 4-OT are arranged in a symmetric way. Adjacent active sites within two dimers such as A and D are approximately 10 Å away. The distance between two non-adjacent active sites of the same dimer such as E and F, is between 25 ∼ 30 Å. Previous theoretical studies have applied both hexamer model (i.e., six protein monomers)16,21 and dimer model (i.e., two protein monomers)15 to study the reaction mechanism. However, 4-OT can only catalyze the reaction as a hexamer under physiological condition23. In addition, since each substrate has two negative charges on two carboxyl groups, the electrostatic interactions between two substrates (for instance, in A and D) are not negligible to the catalytic process of 4-OT. Although all six active sites could be occupied by the singly charged inhibitors12 as shown in Figure 2, equilibrium mixture titration and NMR experiments14 suggest a half-of-the-sites binding mechanism due to negative cooperative effect between two adjacent substrates (such as substrates in monomers A and D): Only three out of six active sites should be occupied to achieve the maximum reaction efficiency. Hence, different enzymatic structure models should be scrutinized to simulate the reaction.</p><p>Second, in the 4-OT catalytic process with Pro1 as a general base, two proton transfer pathways have been proposed previously: one proton at the C3 site of the substrate is abstracted by Pro1 and the same proton is shuttled back to the C5 site of 2o4hex (i.e., one-proton transfer mechanism) 16-21; or one proton at C3 is first abstracted by Pro1 but the other proton of Pro1-N is transferred back to the substrate (i.e., two-proton transfer mechanism)15. Unfortunately, these two distinct proton transfer mechanisms have not been emphasized by the previous theoretical works, which cannot be easily clarified by experiments.</p><p>Third, other issues related to the substrate conformation and even the proper reactant forms have been debated in previous experimental and theoretical studies. For instance, Ruiz-Pernia et. al15 recently proposed that significant conformational changes of 2o4hex are coupled with proton transfer based on their dimer structure model. However, some studies based on the hexamer structure model16 support that the substrate orientation does not influence the reaction process significantly although the 2o4hex substrate is very flexible in solvent. Whether the substrate undergoes conformational changes during proton transfer or not is still an open question. In addition, a recent experimental study19 suggests that 2HM in Figure S10 may be the appropriate reactant form instead of 2o4hex. Since 2HM is the dienol form of 2o4hex and it inter-converts fast with 2o4hex in aqueous solution, either 2HM or 2o4hex, or even both, can be the reactant forms in the catalytic process.</p><p>In this work, we built three 4-OT models: 3 dimers 3 substrates (3D3S), 3dimers 6 substrates (3D6S), and 1 dimer 1 substrate (1D1S) as listed in Table 1 to characterize the significances of protein-protein interactions on the reaction process. Our results show that the reaction process should be characterized by the hexamer model. The half-of-the-sites occupation mechanism is supported by comparing the reaction barrier differences between 3D3S and 3D6S. During the reaction, the same proton of C3 is transferred from C3 to C5 of 2o4hex via Pro1 as a general base. Large conformation changes on substrate (i.e., 1-carboxyl group rotations) do not occur based on the hexamer model. 2HM, which was used in previous experiments as an alternative reactant, may adopt another different reaction mechanism through water in the first proton transfer process comparing with the 2o4hex reactant.</p><!><p>The QM/MM minimum free-energy path method (QM/MM-MFEP)24,25 was applied to optimize the geometries of reactant and product. The reaction barriers were computed by combining QM/MM-MFEP with path optimization methods of Nudged Elastic Bond (NEB)26. The key feature of QM/MM-MFEP is that all of calculations are performed on a potential of mean force (PMF) surface of the fixed QM subsystem conformation, which is defined by</p><p>where E(rQM, rMM) is the total energy of the entire system expressed as a function of the Cartesian coordinates of the QM and MM subsystems. The QM/MM interaction energy in E(rQM, rMM) includes the electrostatic interactions from classical point charges and Van der Waals interactions between QM and MM subsystems. The QM point charges are fitted from the QM electrostatic potential with the QM geometry at each optimization cycle. The boundary atoms between QM and MM subsystems are simulated by the pseudobond approach27 with the recently refitted parameters28.</p><p>The QM/MM MFEP with sequential sampling is an efficient and accurate approach to perform reaction path optimization using the NEB or Quadratic String Method (QSM)29. The detailed procedure was discussed in Refs.24,25,30. This sequential sampling method has been successfully applied to study several enzymatic reaction mechanisms31-34 and solution reactions35. This approach was further extended to compute accurate redox free energies for different solutes in aqueous solution36.</p><!><p>The crystal structure of 4-OT with six inhibitors was taken from Ref. 21. Considering the proton transfer stereochemistry6 and the protonation states of histidine residues16, three enzyme models were built as listed in Table 1. In the 3D3S model, three substrates occupy three active sites since the negative cooperative effect was observed in experiments. As such, the shortest distance between two substrates in 3D3S is ∼25 Å compared to 10 Å in 3D6S when all six active sites are occupied. Since each monomer has one negative charge and each substrate has two negative charges, the total charges for 3D6S, 3D3S, and 1D1S systems are -18, -12, and -4, respectively. 3D6S and 3D3S are solvated in a rectangular water box of 90×90×90 Å3, which contains 5,869 protein atoms and 21,404 water molecules. 1D1S is solvated in a 64× 64× 92 Å3 water box with 1,971 protein atoms and 11,417 water molecules. Protein and water molecules are described by CHARMM22 force fields 37,38 and TIP3P model 39, respectively.</p><!><p>Using force field parameters on substrates from our previous study21, three model systems were warmed gradually from 10 K up to 300 K with a series of restrained 120 ps MD simulations. Harmonic restraints were applied on all heavy atoms with a force constant of 40 kcal/mol/Å2, then reduced to 20 kcal/mol/Å2. Finally, only Cα atoms were restrained with a force constant of 10 kcal/mol/Å2. During the warming procedure, all the substrate structures were fixed. 2 ns MD simulations for three models were performed without any restraints to reach the equilibrium states. Subsequently, three independent 2.4 ns MD simulations were carried out for three models to characterize the dynamic behaviors of three protein models. The leapfrog algorithm (a modified version of the Verlet algorithm40) was employed with different integration step sizes: 1 fs for short range force, 4 fs for medium range force, and 8 fs for long range electrostatic force. The PME method was applied to take long range electrostatic interactions41 into account with the uniform grids (i.e., one grid point per Å in this work). Bonds in water molecules were constrained by the SHAKE algorithm42. A 9-15 Å dual cut off method was employed to generate the nonbonded pair list, which was updated every 16 fs. The NVT ensemble was used in all molecular dynamic simulations with T=300K, which was maintained by the Berendsen thermostat 43 with 0.05 ps relaxation time.</p><!><p>Since the reaction catalyzed by 4-OT is the proton transfer process with Pro1 as a general base, the active site described by QM contains the deprotonated 2o4hex, protonated Pro1, and the boundary atom Ile2 Cα between Pro1 and Ile2. Only one substrate is computed by QM during the QM/MM simulations while the other substrates are depicted by the fitted MM force fields21. Therefore, the total number of QM atoms is 33 including the boundary atom, which were calculated by B3LYP/6-31+G(d)44,45. All the geometries for reactants, intermediates, and product states were optimized by the QM/MM-MFEP approach. The bond length difference between Pro1-N-H and C3 of the substrate is used as the driving coordinate to generate the initial path from reactant to intermediate state while the bond length difference between Pro1-N-H and C5 of the substrate is chosen as another coordinate to obtain the initial path from intermediate to product states. The MM MD sampling time for single point geometry optimizations with the QM/MM-MFEP approach in the coordinate driving procedure is 40 ps. Based on the foregoing initial reaction paths, NEB was employed to optimize the reaction path in association with QM/MM-MFEP. During the NEB path optimizations, the MD sampling time was initially taken as 40 ps at each optimization cycle, and was increased to 80 ps later. The 160 ps MD sampling was also performed to verify the convergence of the path optimizations. The computational details for MM MD samplings in the QM/MM-MFEP calculations are same as the classical MD simulations discussed before.</p><!><p>The root mean square deviations (RMSDs) of the alpha carbon atoms on the protein backbone computed from 4.8 and 2.4 ns MD simulations are shown in Figure 3 for 1D1S, 3D3S, and 3D6S using their initial structures as the references, respectively. The substrate form is chosen as the deprotonated 2HM, which is a stable intermediate state during proton transfer process. Pro1-N is protonated as well. The overall RMSD values for both 3D6S and 3D3S hexamer models are less than 1.5 Å, which suggests that the global structures of 3D6S and 3D3S are very stable. However, the 1D1S model exhibits large structural fluctuations, in which the RMSD value can be larger than 3 Å.</p><p>The analysis of the MD trajectory in 1D1S (protein structure in shown in Figure S1) suggests that the substrate cannot bind tightly to the surrounding protein residues from one dimer of 4-OT. To further support this understanding, we monitored key substrate-protein hydrogen bond distances. Particularly, two hydrogen bond distances between Arg39 and the substrate and another two distances between Arg61′ and the substrate, as shown in Figures S2 and S3, fluctuate dramatically. The hydrogen bonds between Arg61′ and the deprotonated 2HM undergo large fluctuations and can be broken during the MD simulations while Arg39 rotate its side-chain to stabilize the substrate by breaking the initial hydrogen bonds (see Figure S2 for details) and plays a similar role as Arg39″ in other two hexamer models, the active site is highly exposed to the solvent and exhibits large fluctuations during the MD simulations. In contrast, the geometries of active sites in both 3D3S and 3D6S models are maintained as well as the entire protein structures. Based on the entire hexamer structure in Figure 2, the occupied active site in one dimer such as AB is encompassed with the beta sheet and alpha helix from subunits C and D. These protein-protein interactions between dimers can stabilize the beta-hairpin and beta-strand at the end of subunit A. As such, the hydrophobic environment surrounding the active site of subunit A is produced and is maintained during MD simulations. More importantly, the essential Arg39″ in the two hexamer models (note that the critical Arg39″ is completely missing in 1D1S) interacts with the 1-carboxyl group of the substrate through two hydrogen bonds, which can stabilize the substrate in 4-OT and promote the proton transfer reaction.</p><p>In the previous study by Ruiz-Pernia et al15 using the one dimer model with both active site occupied, they observed the proton transfer process is coupled with 1-carboxyl group rotation of the substrate. Since two active sites in one dimer are more than 20 Å away, our 1D1S model with one active site occupied is similar to Ruiz-Pernia's one dimer model. As such, we performed QM/MM MFEP reaction path optimization for the second proton transfer step with our 1D1S model (i.e., proton transfer from the protonated Pro1 to C5 of the deprotonated 2HM). The optimized reaction path for this second proton transfer step is shown in Figure S4 with the activation barrier around 7.0 kcal/mol, which agrees with the previous study in Ref. 15, which has the barrier around 5.0 kcal/mol. We found that after the second proton transfer is accomplished, the substrate undergoes the 1-carboxyl group rotations to further relax itself to a more stable conformation state with the barrier around 8.0 kcal/mol (see Figure S4). Although our calculations on the 1D1S model focus on the second proton transfer process, our results suggest that the proton transfer processes in 1D1S of 4-OT may involve the substrate conformation changes as shown in Ref. 15. Note that the dimer form of 4-OT can only exist when the pH value is less than 4.8 and the enzymatic activity is then completely lost23. Therefore, to reveal the subtle reaction mechanism of 4-OT in physiological conditions, we use the hexamer model (i.e., 3D3S and 3D6S) in the following sections."</p><!><p>The reaction process catalyzed by 4-OT includes two proton transfer steps through Pro1 as a general base and requires one definite intermediate (i.e., deprotonated 2HM) between the two steps (see Figure 1). Experiments alone cannot determine whether or not the same proton is involved for two sequential proton transfer steps. Herein, two mechanisms have been investigated in previous theoretical studies15,21: one-proton transfer mechanism, in which the same proton is transferred twice; and two-proton transfer mechanism, in which two different protons participate in two separate transfer steps.</p><p>To determine the preference of either one- or two- proton transfer mechanism in 3D3S and 3D6S, we found that the Pro1 orientation with respect to the substrate conjugate plane is a structural metric. As shown in Figure 4, the relative positions between Pro1 and 2o4hex can be characterized by the dihedral angle formed by Pro1:N, Pro1:CA, Pro1:CD, and 2o4hex:C4 shown in the inset of Figure 4. When the five-ring plane of Pro1 is perpendicular to the conjugate plane of 2o4hex (i.e., θ=0° or 180°), the two-proton transfer mechanism is favorable: one proton is abstracted by Pro1 and the other one on Pro1-N is transferred to the substrate. When the dihedral angle between two planes is close to zero (i.e., θ=90°), the proton on Pro1 resides in the opposite side of the Pro1 plane that suggests one-proton transfer mechanism. Based on the 2.4 ns MD simulations with the deprotonated 2HM, the average dihedral angles were 62 and 60 degree in 3D3S and 3D6S, respectively (see Table S1). Therefore, the Pro1 plane is always parallel with respect to the substrate plane, which indicates that one-proton transfer mechanism is preferred in the hexamer models.</p><p>Since the substrates were described by the MM force fields in our MM MD simulations, we prepared one new system directly generated from the 4-OT crystal structure with the inhibitor (PDB ID: 1BJP)12. Note that the initial dihedral angle is close to 180°. Five independent direct QM/MM MD simulations (see details in SI-IV) with three 2o4hex substrates were performed up to 32 ps. (Note that only one substrate is characterized by QM and the other two are still computed by the MM force fields.) The average dihedral angles for five MD simulations were 92, 104, 98, 72, and 76 degree in Table S2. This result is consistent with the classical MM MD simulations. Overall, both MM and QM/MM simulations support that the Pro1 plane is parallel with respect to the substrate plane during the reaction. (Note that we have also carried out QM/MM-MFEP simulations to study the two-proton transfer mechanism in both 3D3S and 3D6S. We found that the active site is completely distorted due to the orientation of Pro1 plane and the reaction path cannot be obtained.) One-proton transfer mechanism is preferred because of this active site structural feature.</p><!><p>The free energy profiles of the optimized reaction paths for 3D3S and 3D6S using one-proton transfer mechanism is illustrated in Figure S5 and a simple free energy diagram about the reaction is shown in Figure 5. For both structural models, the reaction involves two sequential proton transfer steps as discussed in the next section. The rate-limiting step for the 3D3S and 3D6S models is the first proton transfer step with the reaction barriers of 18.5 and 27.1 kcal/mol, respectively. The barriers for the second proton transfer step are 15.2 and 23.0 kcal/mol, respectively in 3D3S and 3D6S. Since the only structural difference between 3D3S and 3D6S is the number of substrates in the 4-OT hexamer, the corresponding barrier difference indicates that the occupation of all active sites by six substrates in 3D6S impair the catalytic activity of 4-OT, which agrees with the experimental observations of negative cooperative effect14. Therefore, our theoretical calculations on 3D3S and 3D6S support the experimentally observed half-of-the-sites occupation mechanism in 4-OT.</p><p>In Figure 5, the free energy difference between intermediate and reactant states is only 9.8 kcal/mol in 3D3S compared to 17.8 kcal/mol in 3D6S. To explain why the deprotonated 2HM needs higher energy to reach in 3D6S than in 3D3S, the geometric differences between 3D3S and 3D6S around the active site were scrutinized using classical MD simulations with the intermediate state (i.e., deprotonated 2HM). We found that several distance patterns play important roles to stabilize the intermediate state, including key hydrogen bonds between the substrate (S) and nearby residues (i.e., Arg39″, Arg61′, Arg11′, and Leu8′) and one distance between Arg39 CZ atom and substrate C1 atom (Arg39-S). The measured average distances from MM MD simulations are listed in Table II. For 3D3S and 3D6S, Arg39″-S and Arg61′-S are nearly identical, which suggests that Arg61′ and Arg39″ bind tightly the head part (i.e., 1-carboxyl and 2-keto group) of the deprotonated 2HM through hydrogen bonds regardless of the occupation number in the six active sites. However, the bond distances of Arg11′-S, Leu8′-S, and Arg39-S in 3D3S are much shorter than in 3D6S. In fact, these shorter hydrogen bond distances between Leu8′/Arg11′ and the tail part of substrate in 3D3S help stabilize the intermediate state. In addition, Arg39-S in 3D6S becomes 1.4 Å longer than in 3D3S because Arg39 in 3D6S must bind another substrate in the adjacent active site. We further carried out Non-Covalent Interaction (NCI) analysis46,47 for a typical snapshot from 3D3S and 3D6S MD simulations to visualize and qualitatively compare non-covalent interactions strength48 between the substrate and the surrounding protein amino acids. As shown in Figure S6 (see details in SI-III), Arg39 forms a hydrogen bond to the deprotonated 2HM in 3D3S but not in 3D6S. In addition, the densities at the NCI critical point of interaction for the hydrogen bonds between Arg39″ and the deprotonated 2HM (list in Table II) are larger in 3D3S (0.052 and 0.057) than in 3D6S (0.031 and 0.045). This indicates that the hydrogen bond between Arg39″ and the deprotonated 2HM is stronger in 3D3S than in 3D6S. Overall, the NCI analysis qualitatively agrees with our previous structural analysis. As such, 3D3S can form more compact hydrophobic environments due to the half-of-the-sites occupation to stabilize the intermediate state and lower the activation barrier for the first proton transfer step.</p><!><p>Based on the QM/MM-MFEP simulations of 3D3S, the proton transfer process catalyzed by 4OT includes two sequential proton transfer steps: the proton on the substrate C3 atom is abstracted by the nitrogen (Pro-N) atom on neutral Pro1 to reach the intermediate state; after the substrate slides slightly to pose the C5 atom of the substrate above Pro1-N, and then this proton is shuttled back to C5 on the Si face. The rate-limiting step is the first proton transfer step with the activation barrier 18.5 kcal/mol, which agrees with the previous theoretical studies15-17,21 and the estimated experimental value (13.8 kcal/mol)2.</p><p>The optimized geometries of the reactant (index 5 in Figure S5), intermediate (index 23), and product states (index 43) are shown in Figure 6. (Note that the intermediate state has some iso-energy conformations caused by the substrate sliding during the reaction. All these structures are chosen based on the lowest free energy points along the optimized reaction path.) Since the surrounding residues are crucial to stabilize these geometries, the ensemble averaged values of distances between the QM active site and surrounding residues in MM subsystems were calculated in QM/MM-MFEP simulations. All the key distances are listed in Table III for the reactant, first transition state (TS1), intermediate, second transition state (TS2), and product. The values of Arg39″-S, Arg11′-S, and Leu8′-S indicate that the Arg39″, Arg11′, and Leu8′ residues bind the substrate strongly in all five states during the reaction through the stable hydrogen bonds. The distance of Arg61′-S becomes shorter to stabilize TS1, intermediate, and TS2. Furthermore, the shorter distance of Arg39-S in the intermediate state suggests that Arg39 binds the intermediate state tightly to form a compact active site structure and stabilize the intermediate state. In contrast, Arg39 in 1D1S serves a similar role as Arg39″ in 3D3S, which stabilizes the 1-carobxyl group of 2o4hex and facilitates the proton transfer reaction. These structural observations are consistent with previous theoretical studies17,18,21. In contrast to the recent study15 based on one dimer model for 4-OT, no large conformation changes of the substrate occur during the reaction in 3D3S. Instead, the alternative reactant geometry of 2o4hex may exist during the reaction as discussed below.</p><!><p>As shown in Figure 6, the head 1-carboxyl group is stabilized by Arg61′, Arg39″, and Arg39 while the tail carboxyl group is anchored by hydrogen bonds with Arg11′ and Leu8′. However, 2o4hex itself is a flexible substrate in aqueous solution. In fact, we observed two possible reactant geometries with two different orientations of the 1-carboxyl group as shown in the insets of Figure S7 "straight" and "bent". The free energy difference between the straight orientation and the bent one in 4-OT was computed by the QM/MM MFEP approach using the dihedral angle defined by atoms C1, C2, C3, and C4 of the substrate as the initial driving coordinate. Figure S7 indicates that the free energy difference between "straight" and "bent" conformations is less than 1 kcal/mol. (Note that the activation barrier between "straight" and "bent" conformations only demonstrates that such conformation changes are not possible after 2o4hex docks to the protein binding pocket.) Energetically, both "straight" and "bent" conformations are possible as the reactant. The specific reactant conformation in 4-OT depends on the docking process between the substrate and 4-OT. Furthermore, the first proton transfer step with the "bent" reactant in the 3D3S model was simulated by QM/MM-MFEP and the reaction profile was shown in Figure S8. The reaction barrier of the first proton transfer step is about 16.0 kcal/mol, which is comparable with the result from the "straight" reactant. This suggests that both "straight" and "bent" conformations of 2o4hex can be the reactant geometry.</p><p>The recent theoretical study by Ruiz-Pernia et al15 observed that the substrate conformation (i.e., the orientations of 1-carboxyl group) undergoes large changes during the reaction using their dimer model of 4-OT. In their works, the reactant 2o4hex with the "bent" conformation is first changed to the "straight" one in the intermediate state and the 1-carboxyl group rotates itself back to generate the "bent" product 2o3hex. In our 1D1S model, we also observe the substrate orientation change in the second proton transfer process, which supports that substrate conformational change might be necessary in a dimer model. However, in 3D3S and 3D6S models, we do not observe these large conformation changes. The substrate always remains either "straight" or "bent" during the reaction in the hexamer models.</p><p>Based on our simulation results of 1D1S, 3D3S, and 3D6S models, we found that the substrate conformation change depends on the enzymatic model used. In the dimer model (i.e., 1D1S and Ruiz-Pernia's models), Arg39 and Arg61′ cannot bind the substrate tightly. As such, the active site structure is flexible during the reaction. The rotation of the 1-carboxyl group of 2o4hex is required to promote the proton transfer process. In the hexamer model (i.e., 3D3S and 3D6S), the substrate is confined by surrounding amino acids including Arg39″, Arg61′ and Arg11′. The 1-carboxyl group of 2o4hex cannot be rotated easily and its rotation is not necessary to transfer the proton due to the docked structure (see Figure 6-(a)). In addition, the one-proton or two-proton transfer mechanism also depends on the enzymatic model. In the dimer model, since the active site is exposed to the solvent and endures large structural fluctuations, the Pro1 orientation might be perpendicular with respect to the substrate plane. Hence, the two-proton transfer mechanism is plausible in the dimer model, but not in the hexamer model.</p><!><p>Based on the original experimental study2, 2o4hex has been used as the reactant form in most of theoretical studies. However, some recent experimental studies19,49 proposed that 2HM may serve as the reactant. (Note that more details about the reaction channels among 2o4hex, 2HM, and 2o3hex are discussed in SI-V.)</p><p>2HM has two different states: protonated or deprotonated. The deprotonated state in Figure 1 is the intermediate state of the 4-OT reaction. Since 2o4hex and protonated 2HM can interconvert quickly in aqueous solution, it is very challenging for experimental approaches to identify the correct reactant form. Herein, we performed 32 ps direct QM/MM MD simulations with 2o4hex and 2HM substrates using the "bent" conformation. The key reaction coordinates, i.e., the distance between Pro-R proton (stereo-chemically different from Pro-S proton) in 2o4hex (or 2-hydroxyl proton in 2HM) and Pro-N, were monitored during MD simulations. As listed in Tables S3 and S4, the average distances in 2o4hex are 3.40, 4.87, 3.74, 3.52, and 3.26 Å for five independent simulations. In contrast, the average distances in 2HM are 5.24, 5.28, 4.69, 5.07 and 4.74. The longer distance for the first step proton transfer in 2HM indicates that 2HM needs extra free energies to promote the first proton transfer. Further QM/MM-MFEP reaction path simulations on the first step, as shown in Figure S9, demonstrate that the reaction barrier for the first step using 2HM as reactant is increased 27.6 kcal/mol. Compared to the computed barrier 18.5 kcal/mol using 2o4hex in 3D3S, 2HM is not likely to be a suitable reactant when the direct proton transfer mechanism is applied for the first proton transfer process.</p><p>However, previous experiments19 demonstrated that 2HM (reaction barrier: 12.92 kcal/mol) can be catalyzed by 4-OT more efficiently than 2o4hex (reaction barrier: 13.77 kcal/mol). The discrepancy between our computed reaction profile and experiments indicates that 2HM may adopt a different reaction mechanism rather than direct proton transfer for the first step. We carried out three independent 640 ps classical MD simulations for 2HM, using both straight and bent conformations as the initial structures, respectively (substrate force field is generated using MATCH program50 and more simulation details are provided in SI-VI). The monitored distance between 2HM 2-hydroxyl proton and Pro-N is 4.17, 4.14 and 4.10 Å for three simulations with the bent conformation, and 5.60, 5.17 and 5.15 Å for three simulations with the straight conformation. The structural information suggests that the direct proton transfer from 2HM 2-hydroxyl proton to Pro-N is not favored. However, in all three simulations with the bent conformation, one water molecule forms a persistent hydrogen bond with the 2HM 2-hydroxyl proton. Hence, we proposed that the water molecule could participate in the first proton transfer process. The optimized reactant structure including this water molecule, surrounding with Arg 11′, Arg 61′, Arg39″ side chains, the bent 2HM, and Pro1 by our QM/MM-MFEP simulations is shown in Figure S11. Using this optimized structure, we conducted two-dimension potential energy surface (2D-PES) scan using two bond distances (r1, represents distance between 2-hydroxyl proton and water oxygen; r2 represents distance between water hydrogen and Pro nitrogen, see calculation details in section SI-VI) with Gaussian09 program51. The 2D-PES results in gas phase and implicit solvent model as shown in Figure S12 indicate that the water-mediated proton transfer process has a low activation barrier less than 10 kcal/mol. Therefore, 2HM may adopt a new reaction mechanism involving a water molecule to facilitate the first proton transfer. Further studies using QM/MM simulations on this new mechanism will be investigated.</p><!><p>Although the mechanism of proton transfer reaction catalyzed by 4-OT has been studied in the last ten years, several reaction details are not clear. In this work, we applied classical MD and ab initio QM/MM-MFEP simulations to explore the subtle reaction mechanism of the ketonization process. We found that 4-OT, as a hexamer, contains the strong protein-protein interactions needed to maintain the stable structures of six active sites. As a consequence, a real hexamer model is appropriate to study this enzyme. The proton transfer pathway only involves one single proton for the two sequential steps: the Pro-R proton on the substrate C3 atom is first abstracted by Pro-N to reach the intermediate state; and then this proton is shuttled back to C5 of the substrate on the Si surface to fulfill the ketonization process. By comparing the barrier difference between 3D3S and 3D6S, we demonstrate that this reaction achieves the optimal efficiency when three out of six active sites are occupied by three substrates. This is consistent with the negative cooperative effect observed in experiments. The substrate conformation changes of 2o4hex are not observed in our hexamer models. Two possible conformations of 2o4hex may exist as the reactant, which may be determined by the docking process between 2o4hex and 4-OT. We further showed that protonated 2HM may serve as reactant using a water-mediated reaction mechanism for the first proton transfer process. Overall, our work clarifies several important issues about the reaction mechanism of 4-OT and reveals the concrete proton transfer pathway in the ketonization process.</p>

PubMed Author Manuscript

Development and validation of derivative UV spectroscopic method for simultaneous estimation of nicotinamide and tretinoin in their binary mixtures and pharmaceutical preparations

An accurate, precise, sensitive, and simple spectroscopic method was developed and validated for simultaneous quantification analysis of tretinoin (TRT) and nicotinamide (NCT) with a ratio of 1:40 (TRT: NCT) in a synthetic mixture from dermal pharmaceutical preparations (solution and cream). Wavelengths were chosen in the first and second-order derivatives which are valid for the determination of NCT with the existence of TRT and excipients of the tested pharmaceutical preparations. Wavelength 253 nm was picked for the first-order derivative. Wavelengths 245 and 269 nm were picked for the second derivative. All previous wavelengths were zero-crossing points for TRT and its pharmaceutical preparations. Zero-order spectroscopy was used to determine TRT at the wavelength 348 nm, where no interference with NCT or any substance in the previous pharmaceutical preparation. The linearity range was studied and found to be 20–120 μg/mL and 0.5–5.0 μg/mL for NCT and TRT respectively. The correlation coefficient was 0.9995–0.9999 for NCT and 0.9998–0.9999 for TRT. The limit of detection (LOD) and the limit of quantification (LOQ) of NCT were 1.510 μg/mL and 4.590 μg/mL respectively at the wavelength 269 nm of the second-order derivative.

development_and_validation_of_derivative_uv_spectroscopic_method_for_simultaneous_estimation_of_nico

<!>Introduction<!>Instruments<!>Solvents and chemicals<!>Standard solution of nicotinamide<!>Standard solution of tretinoin<!>Preparation of standard mixture of nicotinamide and tretinoin<!>Solution formula<!>Cream formula<!><!>Stability<!><!>Robustness<!>Application of the proposed method for pharmaceutical preparations<!><!>Dermal solution preparation<!>Dermal cream preparation<!>Conclusion<!>

<p>The chemical structures of nicotinamide (A), nicotinic acid (B), and tretinoin (C)</p><!><p>Many multivitamins and supplementary pharmaceutical preparations contain NCT. It is also found in many dermal preparations, such as solutions, creams, and gels whether as a single active ingredient or with other active ingredients. NCT has concentrations of 3, 4, or 5% (w/w%) in the previous preparations, which are used for many skin conditions including acne [3].</p><p>Tretinoin (TRT) is an all-trans-retinoic acid, as shown in Fig. 1-C [4]. It is a first-generation carboxylic form of retinoid, which is a derivative of vitamin A. TRT has a molecular weight of 300.4 g/mol. It is a yellow powder. Furthermore, it is sensitive to light, heat, and oxygen in the air, especially in solutions. TRT is used for several skin conditions, such as acne (as a first-line treatment), psoriasis, and photoaging. It exists as a single active ingredient or with other active ingredients. Its concentration in solution, cream, and gel preparations is 0.025, 0.050, or 0.100% [5–7].</p><p>TRT can be found in combination with clindamycin or benzoyl peroxide [7]. However, there is no marketed international combination of NCT and TRT spread worldwide. Such a combination may only be found locally, like in the United States, for example, it is found as a compounded drug, which is marketed and distributed by Sincerus® Florida, LLC [8]. According to a study, a combination containing NCT and retinol (an alcoholic form of vitamin A) has good therapeutical potentials for acne [9]. Therefore, we developed an ultra-violet (UV) spectroscopy method to estimate NCT and TRT simultaneously in synthetic mixtures and pharmaceutical preparations.</p><p>There are many UV spectroscopic methods [10–13] in addition to high-performance liquid chromatography (HPLC) [14–19] and electrochemical methods [20–22], to determine NCT or TRT in combination with other active ingredients. However, no previous studies have developed a method to estimate NCT and TRT simultaneously in a binary mixture or in a pharmaceutical preparation.</p><p>The first and second-order derivative methods are simple and accurate for the direct determination of more than one ingredient in mixtures and pharmaceutical preparations, such as TRT with clindamycin [23].</p><p>NCT is freely soluble in water, ethanol, and methanol, while slightly soluble in diethyl ether. On the other hand, the solubility of TRT is limited as it is only soluble in dimethyl sulfoxide; slightly soluble in polyethylene glycol 400, octanol, and ethanol; and practically insoluble in mineral oil, glycerin, and water. It is also slightly soluble in methanol, which is used in many developed methods in previous studies since it overcomes ethanol in terms of TRT solubility [11, 24–26]. Thus, methanol is used as a solvent in this suggested method.</p><p>The aqueous and alcoholic solutions of NCT, such as methanol, are transparent. NCT does not absorb visual light. Instead, it has a sharp absorbance peak in the UV domain at 262 nm [12, 27]. TRT absorbs visual light with wavelengths less than 440 nm. It has a wide peak around 348 nm (basically between 340 and 353 nm) [11, 23, 28]. According to the previous studies, the spectrum of TRT and its minor absorbance in the rest of the UV domain overlaps NCT. In contrast, NCT has no absorbance at wavelengths higher than 300 nm; thus, NCT does not interfere with the signal of TRT.</p><!><p>The ultra-violet spectrophotometric instrument is T80 + UV/V Spectrophotometer Instrument Ltd (UK). It is connected to a computer. The cells used are 1-cm width quartz cells. The weighing device is an analytical balance (Sartorius, model 2474, Germany). Other devices and instruments used to achieve the work are an ultrasonic bath (Power sonic, model 405, Korea), a centrifuge device (90-1 Centrifuge, Shanghai Surgical Instruments Factory, China), a porcelain mortar, volumetric dark flasks, and several scales of glass pipettes.</p><!><p>Standard active pharmaceutical ingredients are nicotinamide powder 99% (BDH Laboratory Supplies, England), tretinoin powder ≥ 98% gifted by Rama Pharma Co, Aleppo, and methanol of analytical grade (Merck, Germany).</p><!><p>First, 20 mg of NCT was weighed. Then it was transferred into a 20-mL flask and diluted with methanol to the mark to obtain a standard stock solution of NCT with a concentration of 1000 µg/mL. Then, six quantities of 0.2, 0.4, 0.6, 0.8, 1.0, and 1.2 mL were pipetted out to 10-mL flasks and diluted with methanol to prepare a series of standard solutions of NCT with concentrations of 20, 40, 60, 80, 100, and 120 µg/mL.</p><!><p>After weighing 20 mg of TRT, it was transferred into a 20-mL flask and diluted with methanol to obtain a standard stock solution of TRT with a concentration of 1000 µg/mL. Then 0.5 mL was pipetted out to a 20-mL flask to obtain a stock solution of TRT with a concentration of 25 µg/mL. Then, ten quantities from 0.2 to 2.0 mL were pipetted out to 10-mL flasks and diluted with methanol to prepare stock solutions of TRT with concentrations of 0.5, 1.0, 1.5, 2.0, 2.5, 3.0, 3.5, 4.0, 4.5, and 5.0 µg/mL.</p><!><p>The synthetic mixture of NCT and TRT was made in the ratio of 1:40 (TRT: NCT). The same procedure was followed in Sects. 2.3.1 and 2.3.2. Except that the same final 20-mL flask was used for both NCT and TRT with concentrations of 1000 µg/mL and 25 µg/mL respectively.</p><!><p>Locacid®, a dermal solution formula which is produced by Universal Pharma Co, Damascus. It is labeled to contains 1000 µg/mL of TRT (0.1% w/v%). The volume of the container is 30 mL. Two quantities of 0.5 mL of the dermal solution were pipetted. Each one was transferred to a 20-mL flask. A quantity of 20 mg of NCT was weighed and transferred to one of the previous flasks. Both flasks were diluted with methanol to obtain two work solutions. The first one (A) contains a synthetic mixture of; the dermal solution formula, which includes excipients and TRT, plus NCT with a ratio of 1:40 (TRT: NCT) (the same ratio as if the formula also contains 4% of NCT). The second one (B) only contains the formula of the dermal solution which includes excipients and TRT. This procedure was done to study the absorbance of all ingredients of the formula and compare it to the absorbance of the standard solution of TRT. Next, two quantities of 0.4 mL were pipetted from A and B. Each of them was transferred to 4 of 10-mL flasks to obtain 8 solutions. Four of them (A series) contain TRT, excipients, and NCT, while the others (B series) contain only TRT and excipients. Standard additions were added before diluting with methanol. Standard additions of 0%, 50%, 100%, and 150% of NCT and TRT were added to the A series to obtain final concentrations of 40, 60, 80, and 100 µg/mL and 1.0, 1.5, 2.0, and 2.5 µg/mL of NCT and TRT respectively. Standard additions of 0%, 50%, 100%, and 150% of TRT were added to the B series to obtain final concentrations of 1.0, 1.5, 2.0, and 2.5 µg/mL of TRT. The flask with a concentration of 1.0 µg/mL of TRT which has no standard addition was given according to the labeled 0.1% concentration of the formula.</p><!><p>Retinoram®, a dermal cream formula which is produced by Rama Pharma Co, Aleppo. It is labeled to contain 1000 µg/g of TRT (0.1% w/w%). The net weight of the cream is 30 g. Two quantities of 500 mg of the cream were weighed. Each one was transferred to a mortar with 10 mL of methanol for trituration then transferred to a 20-mL flask. Next, each of them was sonicated for 10 min to ensure the maximum disintegration of the cream and full dissolution of TRT. After sonication, a quantity of 20 mg of NCT was weighed and added to one of the previous flasks. Both were diluted with methanol. Then, each of them was centrifuged to obtain two work solutions. The first one (A) contains a synthetic mixture of; the cream formula, which includes excipients and TRT, plus NCT with a ratio of 1:40 (TRT: NCT) (the same ratio as if the formula also contains 4% of NCT). The second one (B) only contains the formula of the cream which includes excipients and TRT. Next, two quantities of 0.4 mL were pipetted from A and B. Each of them was transferred to 4 of 10-mL flasks to obtain 8 solutions. Four of them (A series) contain TRT, excipients, and NCT, while the others (B series) contain only TRT and excipients. Standard additions were added before diluting with methanol. Standard additions of 0%, 50%, 100%, and 150% of NCT and TRT were added to the A series to obtain final concentrations of 40, 60, 80, and 100 µg/mL and 1.0, 1.5, 2.0, and 2.5 µg/mL of NCT and TRT respectively. Standard additions of 0%, 50%, 100%, and 150% of TRT were added to the B series to obtain final concentrations of 1.0, 1.5, 2.0, and 2.5 µg/mL of TRT. The flask with a concentration of 1.0 µg/mL of TRT which has no standard addition was given according to the labeled 0.1% concentration of the formula.</p><!><p>The zero-order UV absorption spectrogram in methanol shows: (1) nicotinamide (80 µg/mL) with 262 nm peak, (2) tretinoin (2 µg/mL) with 348 nm peak, and (3) their binary mixture (NCT 40 µg/mL + TRT 1 µg/mL)</p><p>The zero-order UV absorption spectrogram in methanol shows (A) nicotinamide (30 µg/mL), and its 262 nm peak is overlapped by (B1–B10) tretinoin standard series (B1 = 0.5 µg/mL, B2 = 1.0 µg/mL, B3 = 1.5 µg/mL. B10 = 5.0 µg/mL) with 348 nm peaks</p><p>The zero-order UV absorption spectrograms show spectra of nicotinamide (A) and pharmaceutical preparations formulas of tretinoin (B) in methanol</p><p>The first-order derivative spectrograms show nicotinamide spectrum 30 µg/mL (A), nicotinamide series (A10–A80), tretinoin standard series (B1–B3), solution formula (S), and cream formula (C)</p><p>The second-order derivative spectrograms show nicotinamide spectrum 10 µg/mL (A), nicotinamide series (A1–A6), tretinoin 2 µg/mL (B), solution formula (S), and cream formula (C)</p><p>Analytical performance data for the proposed method</p><!><p>The stability was studied at room temperature (25 ℃), at low relative humidity, and in a dark flask for both NCT and TRT using methanol as solvent. However, NCT is already known for its good stability in both aqueous and alcoholic solutions. NCT 1 mg/mL in methanol can be purchased and transported worldwide by many suppliers. Our study shows that NCT is stable in methanol in storage at room temperature for more than 1 week [32].</p><!><p>Data of the TRT Stability Study</p><p>Data of the Accuracy Study of NCT and TRT for the Proposed Method</p><p>aThree repetitions of each concentration were analyzed with nine total repetitions</p><p>bFrom all 9 repetitions of each wavelength</p><p>Data of the Precision Study of NCT and TRT for the Proposed Method</p><p>aFor each category: there are three concentrations and three repetitions for each concentration with a total of 27 repetitions</p><p>Data of the Specifity Study of NCT and TRT for the Proposed Method</p><!><p>A robustness study was achieved for TRT and NCT in only D2 245 nm for three different instrumental parameters changes and three repetitions for a minor change in each parameter independently. First, the scan speed was changed from medium to fast. Second, the scan range was changed from 200–400 nm to 190–350 nm. Third, the bandwidth of the light beam was changed from 0.2 nm wide to 0.1 nm wide. The developed method appeared to be robust for TRT with the three tests but only robust with two of them for NCT. Recovery varied when the bandwidth of the spectrometer was changed. That was probably due to dependence on its strict method of zero-crossing point. NCT had recovery ranges of: 98.13–101.91% for scan speed parameter, 100.26–101.91% for scan range parameter, and 102.56–116.10% for bandwidth parameter. TRT had recovery ranges of: 100.16–101.42% for scan speed parameter, 99.67–100.84% for scan range parameter, and 98.33–99.35% for bandwidth parameter.</p><!><p>The pharmaceutical preparations were analyzed for their content of TRT and the content of the added NCT from standard solutions. In this study standard additions method has been used for both NCT and TRT with three repetitions. All amounts and concentrations that were prepared for this test are mentioned in Section "Preparation of synthetic mixture from pharmaceutical formulas".</p><!><p>Data of the analysis of nicotinamide and tretinoin in the synthetic pharmaceutical preparation with the Proposed Method</p><p>aIn solution preparation</p><p>bIn cream preparation</p><!><p>TRT in pharmaceutical preparations has an accepted assay range of 90–120% in several pharmacopeias like the British Pharmacopeia (BP) and the United States Pharmacopeia (USP) because the concentration of TRT in the preparations is 0.1% or less [33, 34].</p><!><p>The recovery% and RSD% of both TRT and NCT for each selected wavelengths are shown in Table 6. The studied cream has no zero-crossing point around 245 nm D2 as we mentioned before in Section "First and second-order derivative spectrophotometric optimal wavelength" thus NCT couldn't be studied at that point.</p><!><p>It could be concluded from the results obtained in the present paper that the developed method for simultaneous determination of NCT and TRT in binary mixtures is simple, accurate, robust, precise, and rapid. This method can be used for routine quality control tests to directly determine NCT and TRT individually or simultaneously in a binary combination of either solution or cream formula with their excipients in the mixture without any prior separation. Moreover, it's recommended to analyze TRT within 48 h after being dissolved in methanol because TRT is stable within this period according to the stability test.</p><!><p>British pharmacopeia</p><p>Zero-order spectrum</p><p>First-order derivative</p><p>Second order derivative</p><p>High-performance liquid chromatography</p><p>Limit of detection</p><p>Limit of quantification</p><p>Nicotinamide</p><p>Pellagra preventing substance</p><p>Correlation coefficient</p><p>Retinoic acid receptors</p><p>Relative standard deviation</p><p>Tretinoin</p><p>United States pharmacopeia</p><p>Ultra-violet</p><p>Wavelength with maximum absorbance</p><p>Publisher's Note</p><p>Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.</p>

PubMed Open Access

Approaches for enhancing the analysis of chemical space for drug discovery

Chemical space is a powerful, general, and practical conceptual framework in drug discovery and other areas in chemistry that addresses the diversity of molecules and it has various applications. Moreover, chemical space is a cornerstone of chemoinformatics as a scientific discipline. In response to the increase in the set of chemical compounds in databases, generators of chemical structures, and tools to calculate molecular descriptors, novel approaches to generate visual representations of chemical space in low dimensions are emerging and evolving. Such approaches include a wide range of commercial and free applications, software, and open-source methods. Herein, the current state of chemical space in drug design and discovery is reviewed. The topics discussed herein include advances for efficient navigation in chemical space, the use of this concept in assessing the diversity of different data sets, exploring structure-property/activity relationships for one or multiple endpoints, and compound library design. Recent advances in methodologies for generating visual representations of chemical space have been highlighted, thereby emphasizing open-source methods. It is concluded that quantitative and qualitative generation and analysis of chemical space require novel approaches for handling the increasing number of molecules and their information available in chemical databases (including emerging ultra-large libraries). In addition, it is of utmost importance to note that chemical space is a conceptual framework that goes beyond visual representation in low dimensions. However, the graphical representation of chemical space has several practical applications in drug discovery and beyond.

approaches_for_enhancing_the_analysis_of_chemical_space_for_drug_discovery

Introduction<!>State-of-the-art applications of chemical space<!>Navigation of chemical space: selection of compounds from existing libraries<!>Molecular diversity<!>Structure-property (activity) relationships<!>Compound library design<!>Meta-analysis of applications of chemical space<!>Novel implementations for visualization<!>Expert opinion

<p>Chemical space occasionally referred to in the literature as the "chemical universe" [1] is a concept that has become significant in chemoinformatics as an independent theoretical discipline [2]. Chemical space refers to all possible molecules and multi-dimensional conceptual spaces representing their structural and functional properties. In other words, chemical space is a contraction of the "chemical descriptor vector space" defined by the numerical vector D encoding molecular structure and/or property aspects as elements of the descriptor vector D. Therefore, and in contrast to cosmic space, chemical space is not a physical space and is not unique, because anyone is free to customize its vector space based on structural and functional properties. Indeed, structural and functional representation is arguably the most relevant feature in virtually all chemoinformatics or computational studies [3].</p><p>Applications of chemical space concept have progressed from drug discovery to other areas in chemistry, including organic synthesis, food chemistry, and material sciences, to name a few examples reviewed in the literature [4][5][6]. A key distinction between the different types of the systematic representations of chemical spaces in compound datasets lies in the type of properties or descriptors that are used to represent the compounds of interest. For instance, the nature of the descriptors used to represent small organic molecules is typically different from that describing chemicals with applications in material sciences. In some instances, the qualitative concept of chemical space is actively used to guide drug discovery projects; however, developing a consistent method to visually represent chemical space remains elusive because of the challenge in generating a consistent manner of representing chemical structures. A typical method employed in this area includes analyzing the chemical space of metalcontaining compounds [7].</p><p>Initially, in drug discovery, chemical space concept proved useful to understand and generate knowledge of the pharmacokinetic properties and molecular diversity of biologically relevant compounds [8,9]. As the number of chemical compounds and their information in databases increased, more sophisticated molecular descriptors and visualization techniques were developed to expand their applications. For instance, explorations of chemical space have considerably improved our comprehension of biology and led to the development of several tools for investigating structure-property and structure-activity relationships (SPR, SAR, and SP(A)R) [10]. In addition, this concept has raised interesting questions regarding the estimated size of chemical space, and has motivated several research groups to enumerate large libraries of virtual compounds [11,12]. Recently, the availability of software libraries and the rise of artificial intelligence (AI) [13] have led to the emergence of several tools that integrate machine learning (ML) methods as versatile tools to design, generate, and visualize the chemical space of small molecules [14].</p><p>Most chemoinformatics tools use two discrete procedures to represent chemical space: (i) calculation of molecular descriptors and (ii) projection from descriptor space into a two-dimensional (2D) plane or three-dimensional (3D) volume using one of the several known techniques [15]. The descriptors can be selected from the structure (constitution, configuration, and conformation) or properties (physical, chemical, and biological) of the molecules present. The types of descriptors guide the interpretations and predictions that can be made [16]. Therefore, descriptors based on physicochemical properties have been widely used to encode absorption, distribution, metabolism, and excretion properties that play an important role in determining the characteristics of therapeutic agents, such as absorption, solubility, and permeability through the membrane [17]. Other commonly used molecular representations are fingerprint-based descriptors in which the Molecular Access System (MACCS) Keys [18] and Extended Connectivity Fingerprints (ECFPs) [19] are among the most widely used methods to assess the structural diversity of small organic molecules. To improve the visual representation of chemical space and expand its application to larger compounds such as peptides, oligonucleotides, and complex carbohydrates, Capecchi et al. recently proposed the MAP4 (MinHashed Atom-Pair fingerprint up to four bonds) molecular fingerprint that, in principle, can encode compounds of virtually any size [20]. MAP4 combines substructure and atom-pair concepts to capture global and specific characteristics of the molecular size and shape, which are captured by the bond distance information encoded into the MAP4.</p><p>To generate graphical representations of chemical space, coordinate- [16] and cell-based [21] approaches have been developed. Recently, molecular networks have been recommended for addressing the dimensionality problem [22,23]. Because it is complicated to visualize multidimensional spaces, coordinate-based approaches usually rely on dimensionality reduction techniques to transform high-dimensional data into two or three dimensions. Over the past two decades, several research groups have implemented different dimensionality reduction techniques to analyze chemical space. Such advances were extensively reviewed in a previous study [16]. The most common techniques include principal component analysis (PCA), t-distributed stochastic neighbor embedding (t-SNE) [24], and selforganizing map (SOM) [25]. Previous studies have discussed the exploration of SPR in the context of chemical spaces [10].</p><p>The objective of this manuscript is to review recent advances in methodologies for generating lowdimensional visual representations of chemical spaces. We emphasize on freely available and opensource methods. Despite the concept of chemical space having broad applicability in several areas of chemistry, including in organic and inorganic molecules (for instance, metallodrugs used in drug discovery [7]), this review focuses on the development and applications of chemical space to small organic compounds. It is expected that some of these methods can be extended or adapted to explore chemical space of other types of compounds. Using an analogy with the concept of a multiverse in cosmology, regions in the universe detached from one another exhibit distinct properties [26], and the systematic description of different types of chemical compounds with varying properties (metalcontaining molecules, larger chemical compounds relevant in polymers, material science, and biochemistry) can be increased to chemical multiverses.</p><!><p>The concept of chemical space has several practical applications. In this study, we organized the applications into four categories: selection of molecules from existing compound libraries, analysis of molecular diversity, SP(A)R, and library design (i.e., to assist the expansion of the chemical libraries).</p><!><p>The identification of biologically relevant starting points within a vast chemical space is a particularly relevant task in designing compound collections and selecting compounds from existing libraries for computational and/or experimental screening. Although it is not an easy task, it is possible to utilize the fact that the physicochemical and biological properties of molecules are associated with their molecular structures. This is known as "chemical similarity principle," which states that if two molecules share similar structures, then they will likely have similar bioactivities. Thus, the distribution of the compounds in chemical space guides the search for compounds with a specific set of properties. The choice of descriptors to define chemical space is crucial, however, it is not unique; different from cosmic space, chemical space is not invariant. Therefore, molecular representation is the cornerstone of chemical space (and basically any other computational approach).</p><p>In this context, different cartographic methods have been proposed to efficiently navigate chemical spaces once a set of descriptors has been selected [27]. Most navigation methods involve positioning a reference query molecule and scanning a large database to identify the adjacent molecules, which are molecules with properties significantly similar to those of the reference structures. Notably, the adjacent molecules to the reference compound can be identified using the full set of descriptors that define chemical space, and this can be performed independently of the visualization method to project the fulldimensional space into a 2D/3D graph.</p><p>ChemGPS-NP was one of the first chemographic models used to comprehensively describe chemical space of natural products (NPs) using physicochemical properties and has proven to be useful in various applications [28]. ChemGPS-NP is a PCA-based model of physicochemical properties, defined by a training set of carefully selected compounds that act as "satellites" or reference structures with extreme properties. ChemGPS-NP projects or "positions" new molecules into the chemical space by comparing their physicochemical properties with those of the reference structures. Although PCA-based mapping is fast and easy to compute, it omits nonlinear interactions and some map regions are overloaded with data. Some non-linear algorithms that have been implemented for chemical space visualization are t-SNE [24], and more recently, uniform manifold approximation and projection (UMAP) [29]. These types of algorithms effectively visualize clusters or groups of data points and their relative proximities. Another frequently used method is SOM [30], a grid-based method that has been used to support lead discovery efforts and target prediction. Examples of the latter include SOM-based prediction of drug equivalence relationships [31] and target inference generator [32].</p><p>Generative topographic mapping (GTM) represents a probabilistic alternative to SOMs [33]. This approach has been applied to visualize, analyze and model large collections of data sets for drug design and was also successfully used for large-scale SAR scanning [34].</p><p>As discussed in the Introduction section, within non-coordinate-based approaches, chemical space networks (CSNs) were proposed by Bajorath et al. to address the problem of dimensionality [22,23].</p><p>CSNs transform a multidimensional chemical space into a graph with the nodes representing chemical compounds and edges connecting compounds within a specific similarity boundary. These graphs provide immediate visualization that can be easily interpreted. CSNs can also be adequately characterized and compared using generally applicable statistical measures from network science. However, visualization becomes increasingly difficult as the number of compounds increases. Therefore, this method is not directly designed for diversity analysis. Recently, networks have been used as the basis for developing chemical library networks (CLNs) that can be used to explore the diversity of large and ultralarge molecular libraries. In general, representations of a tree-like nature, such as Tree MAP (TMAP), are more suitable for analyzing and interpreting large datasets [35].</p><p>To navigate through chemical and biological spaces more intuitively, several researchers have developed methods that seek to improve the interpretation by representing molecules beyond individual data points. An example is the scaffold tree approach that graphically represents chemical space as a tree, where the leaves represent individual chemical compounds and the intermediate nodes represent scaffolds and sub-scaffolds [36]. These representations allow a more consistent scaffold analysis in an SAR/SPR context and facilitate the identification of analog collections [37]. To facilitate the visualization of large analog series Constellation plots have been proposed (see section 2.3) [38].</p><!><p>In drug design, the concept of chemical similarity (or chemical diversity) has been addressed using different approaches, and its applications are mainly found in ligand-based design, for instance, in identifying bioactive compounds when some active compounds are known. Similarly, chemical similarity/diversity analysis provides useful information for projects that seek to prioritize the selection of potentially active compounds for experimental evaluation. Another application is the profiling and selection of compound collections with chemically diverse structures to increase the probability of identifying new scaffolds that can lead to specific biological targets [39]. Similarity and diversity analyses have also been integrated into de novo design strategies to evaluate the structural and molecular novelty of chemical libraries, which play an important role in fairly comparing generative approaches [40].</p><p>Several studies reported thus far focus on the use of chemical space as an approach to assess the diversity of different datasets and explore the relationships between compound collections, from which valuable conclusions or interpretations have been obtained. For instance, the chemical space of natural compounds has been compared with other collections of compounds such as drugs approved for clinical use, synthetic molecules, and food chemicals [41]. In general, NPs are characterized by covering a region of chemical space more extensively than synthetic compounds and approved drugs, and they also populate areas in the chemical space that are generally not synthetically accessible [41][42][43]. The structural uniqueness and complexity of NPs have encouraged the continued use of these compounds to identify bioactive compounds for further development, optimization, or inspire the synthesis of compounds with unique scaffolds [44,45].</p><p>Recent representative molecular diversity studies include the analysis of novel libraries, such as compounds applied in the food industry [5,46], peptides [47], focused libraries [48], de novo virtual libraries [49], and commercially available fragments libraries for medicinal chemistry [50]. The results are summarized in Table 1. For these analyses, new molecular representations and visualization techniques were implemented. For instance, the chemical space of food compounds stored in FooDB was analyzed using ChemMaps, an approach based on reference or "satellite" compounds, that is, molecules whose distance (or similarity) to all other molecules in the chemical space yield sufficient information to produce a visual representation of the space [51,52]. In principle, it is possible to generate a 3D visual representation of chemical space using satellite structures. TMAP was used in the global analysis of the peptide chemical space, whereas MAP4 was employed as the molecular representation of peptides [47]. A similar approach was used to visualize the chemical space of NPs in the public domain [53]. To assist the processes of decision-making and selecting compound libraries for further virtual screening or compound acquisition for high-or medium-throughput screening for epigenetic drug discovery, Flores-Padilla et al. reported a comprehensive analysis of 11 commercial libraries of varying sizes focused on epigenetic targets (with 53,443 compounds in total) [48]. Analysis of the chemical diversity and coverage of chemical space was conducted with Constellation plots based on the chemical core scaffolds and CLNs [54]. The latter is based on structural fingerprints and facilitates the visual representation of the chemical space of compound datasets with a significant number (millions) of compounds in an efficient manner. The analysis highlighted a commercial library with an extensive coverage of chemical space (despite low intra-molecular diversity) and identified compound collections that cover unique regions of the chemical space not populated by other epigenetic-focused libraries.</p><p>As previously discussed, diversity analysis of chemical space can be used to evaluate and compare different generative approaches. For instance, Arús-Pous et al. used PCA plots of molecular quantum number (MQN) fingerprints to assess the quality of the training process in generative models [49]. In that study, MQN PCA plots allowed the following up and improvement of the comprehension of the varying architectures of molecular generative models. Another recent and representative example of the use of chemical space to analyze diversity was performed with more than 400,000 purchasable building blocks (PBBs) provided by eMolecules (Zabolotna et al. 2021). Visualization of the chemical space of these PBBs using GTM allowed the identification of the most represented and underrepresented classes of PBBs. The results can be focused to improve PBB libraries in a way that allows efficient synthesis in a relevant medicinal chemistry space.</p><!><p>As mentioned in the Introduction section, one of the major practical applications of visual representation of chemical space in drug discovery is SAR analysis [55] where the concept of chemical space provides a solid and consistent framework for representing the structural data. When activity data are added (e.g., mapped) into a visual representation of chemical space, it is possible to navigate through the chemical space and exploring (qualitatively or quantitatively) variations in activity upon changes in chemical structures. The massive amount of data stored in chemical databases, including incomplete chemogenomic data or activity data obtained at single concentrations, makes visualization SAR difficult; however, is can be aided by the power of visualization tools. Previous studies highlighted advances in methodologies that explore SAR of compound data sets and screening collections [10,55].</p><p>Recent developments in analyzing SP(A)R include constellation plots. Briefly, constellation plots are 2D graphs that combine the clustering of compound datasets based on chemical scaffolds (in particular, analog series) and the distributions or mutual relationships of analog series based on fingerprint representations. Recently constellation plots were used to analyze the SAR of a large dataset of small molecules tested in a panel of cell lines using high-throughput screening. The authors identified a consistent cell-selective analog series of chemical compounds and proposed statistics to quantify cell promiscuity and consistency [56].</p><p>In a separate and recent analysis, constellation plots were used to uncover a promising analog series of inhibitors of tubulin-microtubules. In that study [57], the authors analyzed the SAR of a curated dataset of 851 compounds with anticancer activity targeting tubulin-microtubules. In particular, the constellation plots identified at least six analog series of compounds with high average activity (known as "bright regions" in chemical space). The plot also indicates an analog series with predominantly inactive molecules ("dark regions" in chemical space). In recent developments, constellation plots have been implemented in DataWarrior [58] such that the user can explore the chemical space interactively.</p><p>Another recent example of the application of chemical space to SP(A)R analysis lies at the interface of drug discovery and food chemistry [46]. Bayer et al. explored the associations between the chemical structures of 133 compounds with known biological activities and extra-oral bitter taste receptors, which belong to the superfamily of G-protein-coupled receptors. As part of the analysis, the authors represented the chemical space of the compounds using t-SNE as a visualization tool; the compounds were represented using MACCS key fingerprints. It was observed that the visual representation of chemical space grouped chemical compounds with similar functional groups, even though the compounds can belong to different classes (depending on the type of receptors they are related to).</p><!><p>Over the last few decades, medicinal chemistry has made major breakthroughs in increasing the accessible chemical space, which is estimated to contain approximately 10 63 molecules [11,59]. In this context, having access to more regions of the chemical space can, in principle, augment the probability of finding something "interesting" and valuable. Thus, algorithms and methods to augment and search these spaces can focus on the generation of new molecules to compounds with desirable properties for drug design or discovery projects. In this regard, it remains to determine the medicinally relevant chemical space as the number of therapeutic targets is evolving [60]. A related challenge is to establish the intersection of chemical space with the biological space. These questions are being addressed by computational chemogenomics and have been noted as one of the major challenges in computer-aided drug design [61].</p><p>Computational approaches to facilitate the design of functional molecules include the development of de novo algorithms that explore chemical spaces to generate new compounds. For instance, the de novo design algorithm for exploring chemical space scans the space and generates structures in a specific area on a user-selected pane [62]. Similarly, Capecchi et al. developed the peptide design genetic algorithm (PDGA), a computational tool that generates highly-similarity analogs of bioactive peptides with various peptide chain topologies in a chemical space defined by the macromolecule extended atom pair fingerprint [63]. Recently, Aspuru et al. proposed the superfast traversal, optimization, novelty, exploration, and discovery (STONED) algorithm to perform exploration and interpolation in chemical space to obtain novel molecules [64]. STONED uses self-referencing embedded strings [65], a molecular representation that is more suitable for ML. This algorithm reduces the long training times, large datasets, and handcrafted rules.</p><p>In general, deep generative models can operate over large spaces of molecular structures and embed the chemical properties of these structures into a vector space. These models can generate new and previously unidentified chemical compounds by decoding from this 'latent' space of chemical structures. Recent reviews of the de novo design have examined progress in generative model architecture and evaluated their efficiency with reference to experimentally validated test cases in the literature [66][67][68].</p><!><p>To understand the evolution of the concept of chemical space and its applications, a meta-analysis of the literature has been performed using the search terms "chemical space" and "drug design" in PubMed (https://pubmed.ncbi.nlm.nih.gov/). In total, the search yielded 1538 articles (November 2021) that were analyzed using VOSviewer [69]. The results of the meta-analysis indicated that the main concurrent terms associated with the keywords used were SAR analysis and small-molecule library design (Figure 1a). Visualization of chemical space has been used frequently to support the analysis of antineoplastic agents (76 articles), protein kinase inhibitors (60), antibacterials (51), antimalarials (28), and antiviral compounds (21). Similarly, a notable number of articles related to the concept of chemical space are associated with drug repurposing (20). In particular, using network-based representations to predict drug-target interactions and more complex interactions, including drug-disease, protein-disease, and drug-side effect associations, to name a few [70].</p><p>According to the author's keywords (Figure 1b), the most recent articles (see color scale) are focused on ML methods such as deep learning. It is also highlighted that the concept of chemical space has had recent applications in Alzheimer's disease and in emerging diseases such as COVID-19. Particularly for COVID-19, chemical space visualization proved to be a fast way to analyze and describe the huge chemical space of known antiviral compounds [71,72]. For instance, GTM is one of the methods used to represent the chemical space of compounds obtained from medicinal chemistry efforts against coronaviruses (CoVs) [72]. In particular, GTMs helped highlight the structural relationship between antivirals of different categories, predict their polypharmacological profiles, and emphasize frequently encountered chemotypes. Similarly, chemical space concept was very helpful in finding attractive compounds for repositioning [73] and guiding the identification of potent and selective scaffolds with anti-COVID activity [74]. Advances in AI and availability of software libraries have resulted in ML methods, such as deep learning and versatile tools for exploring chemical space for drug discovery applications [14]. Table 2 summarizes the novel approaches using ML methods, some of which have been mentioned previously.</p><p>Recent advances have focused in identifying molecules with desirable properties in large chemical spaces. To this end, genetic algorithms (GAs) [75,76], methods using variational autoencoders (VAEs) [77,78], recurrent neural networks (RNNs) [79,80], and generative antagonistic networks (GANs) [81,82] have been developed. In several instances, these algorithms are associated with the generation of new molecules and have exhibited the ability to traverse chemical space more effectively, reaching optimal chemical solutions while considering fewer molecules than allowed by the brute-force screening of large chemical libraries. Similarly, several evolutionary and RNN selection mechanisms have proven successful in multi-objective optimization problems [83,84]. Emerging approaches in chemical enumeration incorporate chemical reactions into ML-based generation to design novel compounds in a synthetically accessible chemical space [85].</p><p>As mentioned, similarity-based compound networks such as CSNs allow the visualization of SAR patterns. To increase the number of practical applications of network-based chemical space representations and decrease biases in ML, it is necessary to incorporate amounts of data from chemical interactomes. In this regard, it is also necessary to improve network visualization to obtain reasonable representations of networks containing thousands of nodes. Addressing these difficulties will be useful in SAR analysis and drug repurposing.</p><p>Another application of the field of neural networks has been to solve address problems related to big data and visual representation of datasets with a large number of compounds [54]. It is anticipated that more researchers will integrate ML methods to speed up chemical space analysis and realize more efficient outcomes. • Chemical library networks (CLN) [24,34,35] [54]</p><p>Structure-property/activity relationships • Chemical space networks (CSNs) [22,23] • Constellation plots [38] Design novel compound libraries • Chemical reactions in ML-based generation</p><p>• Multi-objective optimization algorithms [12,85] [83,84]</p><!><p>The interactive visualization of 2D and 3D representations of chemical spaces, in particular of large and ultra-large data sets, has been an active area of investigation. The interactive visualization of chemical space was performed using an open-source code and is freely available on websites.</p><p>Web servers available in the public domain for enabling interactive visualization of chemical spaces have been reviewed recently [10]. This review includes classical and early developments such as Chem-GPS (vide supra), a significant set of public tools developed by Reymond et al. such as Ferun, and PDB Explorer. In the past few months, progress has been made in the interactive analysis of chemical space.</p><p>A notable example is the "magic rings" developed by Ertl: a freely available web page with an interactive clustering of rings and Bemis-Murcko scaffolds present in compounds in ChEMBL [86] (28 release) with a biological activity value of 10 microM [87]. The interactive clustering available at https://bit.ly/magicrings enables users to quickly identify the main substructures of the major target classes of relevance in drug discovery.</p><p>Another recent development is the NP navigator [88], which further bridges the application of cheminformatics in NP research [89,90]. The NP navigator, publicly available at https://infochm.chimie.unistra.fr/npnav/chematlas_userspace/, is an implementation of the visualization algorithm GTM maps that explores interactively the chemical space of COCONUT (Collection of Open Natural Products database) [91] (currently the largest collection of NPs in the public domain), bioactive molecules in ChEMBL, and purchasable compounds from the ZINC database [92]. Interactive navigation can be used to explore chemical compounds based on different representations such as physicochemical properties, scaffold distribution, commercial availability, and biological activity.</p><p>In a recent study, Chávez-Hernández et al. implemented an interactive visualization of the chemical space of a newly generated library of HIV-1 viral protease inhibitors assembled from NP fragments.</p><p>Visual representation of the chemical space was based on TMAPs [20] and molecular fingerprints. The interactive representation of the chemical space enables the user to navigate through a synthetic compound library of pseudo-NPs [93] designed de novo.</p><!><p>Chemical space is a core concept in chemoinformatics with several practical applications in drug discovery and other areas in chemistry. Typically, chemical space is used for selecting specific sets of compounds for further computational or experimental screening, diversity, and SP(A)R analysis, and to guide the design of novel molecules. The latter application is intended such that the newly generated compounds are at the intersection of the biologically relevant chemical space. In any application, compound representation is a key variable in qualitative or quantitative chemical space analysis (including visual representation); it has to be in line with the objective of the study as it will guide the interpretation of the analysis.</p><p>Currently, ML methodologies continue to open new possibilities for generating hundreds and thousands of new molecules from an exhaustive search in chemical space. To perform the search in the chemical space faster and more efficiently, in particular for large data sets, the visualization methods should scale well with the number of molecules ("haystack size"); find the most relevant compounds (e.g., find the "needle," irrespective of the size of the haystack); and be affordable to run on standard hardware.</p><p>In recent years, with a significant an increasing number of molecules to be analyzed, novel methods to generate visual representations of chemical space have been developed. While interpreting such visualizations, one should consider that they are approximations and that the "true" chemical space is defined by the complete set of descriptors used. Because it is challenging to select the appropriate method according to the expected qualities of the visualization, it is advisable to complement the visual (e.g., qualitative) analysis of chemical space with a quantitative analysis considering the entire multidimensional space. In this regard, it is advisable to consider consensus approaches: multiple representations of chemical space (at least more than one), because each visualization will capture part of the "true" chemical space.</p><p>As part of the progress in method development, there have been notable developments in the implementation of freely available online resources. In this manner, the user can interactively explore the chemical space of compound datasets.</p><p>There are still challenges in exploring the chemical space for drug discovery, such as developing consistent representations of metal-containing compounds. Other challenges include consistently representing the chemical space of non-traditional small-and medium-sized biologically relevant compounds such as peptides, macrocycles, and metal-containing clinical candidates.</p>

ChemRxiv

Functional role of an unusual tyrosine residue in the electron transfer chain of a prokaryotic (6–4) photolyase

Cryptochromes and photolyases form a flavoprotein family in which the FAD chromophore undergoes light induced changes of its redox state. During this process, termed photoreduction, electrons flow from the surface via conserved amino acid residues to FAD. The bacterial (6-4) photolyase PhrB belongs to a phylogenetically ancient group. Photoreduction of PhrB differs from the typical pattern because the amino acid of the electron cascade next to FAD is a tyrosine (Tyr391), whereas photolyases and cryptochromes of other groups have a tryptophan as direct electron donor of FAD. Mutagenesis studies have identified Trp342 and Trp390 as essential for charge transfer. Trp342 is located at the periphery of PhrB while Trp390 connects Trp342 and Tyr391. The role of Tyr391, which lies between Trp390 and FAD, is however unclear as its replacement by phenylalanine did not block photoreduction. Experiments reported here, which replace Tyr391 by Ala, show that photoreduction is blocked, underlining the relevance of Tyr/Phe at position 391 and indicating that charge transfer occurs via the triad 391-390-342. This raises the question, why PhrB positions a tyrosine at this location, having a less favourable ionisation potential than tryptophan, which occurs at this position in many proteins of the photolyase/ cryptochrome family. Tunnelling matrix calculations show that tyrosine or phenylalanine can be involved in a productive bridged electron transfer between FAD and Trp390, in line with experimental findings. Since replacement of Tyr391 by Trp resulted in loss of FAD and DMRL chromophores, electron transfer cannot be studied experimentally in this mutant, but calculations on a mutant model suggest that Trp might participate in the electron transfer cascade. Charge transfer simulations reveal an unusual stabilization of the positive charge on site 391 compared to other photolyases or cryptochromes. Water molecules near Tyr391 offer a polar environment which stabilizes the positive charge on this site, thereby lowering the energetic barrier intrinsic to tyrosine. This opens a second charge transfer channel in addition to tunnelling through the tyrosine barrier, based on hopping and therefore transient oxidation of Tyr391, which enables a fast charge transfer similar to proteins utilizing a tryptophan-triad. Our results suggest that evolution of the first site of the redox chain has just been possible by tuning the protein structure and environment to manage a downhill hole transfer process from FAD to solvent.

functional_role_of_an_unusual_tyrosine_residue_in_the_electron_transfer_chain_of_a_prokaryotic_(6–4)

Introduction<!>Experimental methods<!>Computational methods<!>Photoreduction of PhrB and mutants<!>Molecular dynamics simulations<!>Site energy and electronic coupling<!>Charge transfer simulations<!>Discussion and conclusion<!>Conflicts of interest

<p>Cryptochromes and photolyases are homologous proteins with a central avin adenine dinucleotide (FAD) chromophore that full different biological functions, which are most oen triggered by light. During a process termed photoreduction, oxidised FAD of cryptochromes or photolyases takes up one or two electrons to convert to the semiquinone or the fully reduced state, respectively. In cryptochromes, which oen function as photoreceptor proteins, FAD adopts the oxidised form in darkness. In these proteins the photoreduction is regarded as the rst step of a signal transduction cascade. 1 Cryptochrome of migrating birds functions as molecular compass, due to the radical pair formed in the semiquinone state. In photolyases, which are light triggered enzymes that repair UV-damaged DNA, the chromophore assumes the fully reduced form, FADH À , in vivo but converts to the semireduced or oxidised form under aerobic conditions in vitro. Photoreduction is important to ensure a high level of reduced FADH À , which is required for DNA repair. [2][3][4] Because FAD is embedded in the centre of the protein, a direct transfer of electrons from solution is not possible.</p><p>Crystal structures and site directed mutagenesis identied amino acids that constitute an electron transfer (ET) cascade. Most oen, replacement of relevant Trp or Tyr by Phe results in a slower or blocked photoreduction. The interpretation of mutant results can however be hampered by the possibility of parallel pathways. [5][6][7] Three Trp residues at position 306, 382 and 359 of E. coli photolyase constitute the rst identied ET cascade of the cryptochrome-photolyase family (PDB 1DNP). 8 These Trp residues are conserved in all other members of class I CPD photolyases, to which E. coli photolyase belongs, in class III CPD photolyases, CRY-DASH proteins, plant cryptochromes, animal cryptochromes and eukaryotic 6-4 photolyases, but not in class II CPD photolyases 9 or FeS-BCP proteins (see Fig. 1 for a phylogenetic tree of photolyases and cryptochromes). In class I and class III CPD photolyases, additional Trp residues have been shown to be involved in photoreduction. For example, in the class III CPD photolyase PhrA, the Trp residue of the classical triad that is closest to FAD is linked to a second, less conserved electron pathway comprising two Trp residues. 5 The class II CPD photolyases have another Trp triad, which is conserved among this group. 10 Trp side chains are chemically ideally suited for ET processes due to their aromatic cycles, low redox potential (around 0.6 V at pH 7) 12 and stable radical state for the deprotonated form. However, in several CPD class I and II photolyases, Tyr residues are also involved in ET. In the class I CPD photolyase from Anacystis nidulans, a Tyr radical is formed within 50 ms aer FAD excitation, as detected by ultrafast spectroscopy. 13 In Methanosarcina mazei CPD class II photolyase, a Tyr residue is required for full photoreduction. In the Xenopus laevis (6-4) photolyase the involvement of a Tyr residue in photoreduction was shown by electron paramagnetic resonance. 14 The conservation of Tyr or Trp for electron transfer shows that this selection is not a random process. Factors such as the chemical environment (other nearby amino acids or water) certainly play a signicant role. Electron transferring Tyr residues are usually located close to the protein surface, surrounded by water and/or located close to deprotonating amino acids. 15,16 Electron transferring Trp residues can occur at the periphery or in the centre of a protein.</p><p>The group of FeS-BCP proteins is a phylogenetically distinct group of (6-4) photolyases (Fig. 1) with unique properties such as an Fe-S cluster and a 6,7-dimethyl-8-ribityllumazine (DMRL) antenna chromophore. 17 The ET in FeS-BCP proteins differs from all other cryptochrome and photolyase groups in several ways. In the FeS-PCB members PhrB from Agrobacterium fabrum (PDB 4DJA) 17 and CryB from Rhodobacter sphaeroides (PDB 3ZXS) 18 photoreduction proceeds via Trp390 and Trp342 (PhrB numbering), as shown by site directed mutagenesis. These residues are highly conserved in FeS-BCP members (see ESI Fig. S1 †). In both proteins, the Tyr391 side chain is directly located between Trp390 and FAD (see also Fig. 2) suggesting that this Tyr must be part of the ET chain. However, when the Tyr was replaced by Phe, the photoreduction rate of PhrB was not affected, 19 and only slightly affected in CryB. 18 In about 30% of FeS-BCP proteins, a Phe is placed at this position (see Fig. S1 †).</p><p>These observations raise several questions about Tyr391:</p><p>is this residue involved in photoreduction as the rst electron donor of FAD, as proposed by its spatial position?</p><p>if yes, how can the electrons be transmitted via this Tyr residue and why is it a Tyr residue, whereas in other groups of photolyases and cryptochromes a Trp residue serves as electron donor for FAD?</p><p>-Phe is usually used to interrupt electron chains. Why does the replacement of Tyr391 by Phe not interrupt the electron chain?</p><p>FAD photoreduction involving a Trp triad has been widely studied by computational approaches in Escherichia coli photolyase, [21][22][23] Arabidopsis thaliana cryptochrome, [24][25][26][27][28][29] Synechocystis sp. CRY-DASH protein 30 or Xenopus laevis (6-4) photolyase. 31 These studies highlighted the role of Trp in the classical triad [27][28][29] or of additional residues 30,31 at an atomistic level, as well as the crucial role of the environment 24,25,29 or the quantum effects in these ultrafast charge transfers. 26 In our group, we have established a quantum mechanics/classical mechanics (QM/MM) scheme based on fragment orbital tight-binding density functional theory (FODFTB) coupled to MM molecular dynamics (MD). [22][23][24]32 Our approach allows direct simulations of the charge propagation along a Trp triad. Previously we established our method by successful reproductions of ET rate constants in E. coli photolyase 23 and Arabidopsis cryptochrome 24 and also highlighted the role of environment in the downhill charge transfer process. Applying these techniques, we are able to shed light onto the molecular evolution of ET pathways in different proteins belonging to the cryptochrome-photolyase family.</p><p>In the present work, we investigate the role of Tyr391 in PhrB by experimental and QM/MM approaches, comparing PhrB wild type (WT) with its Y391F, Y391W and Y391A mutants. We regard that the analysis of the Tyr to Trp replacement is important for an understanding of the evolution of photolyases and cryptochromes, which most oen have a Trp triad as electron transmission. Due to the difference in ionization potentials, it is expected that a Tyr residue slows down or blocks charge transfer, and that the replacement of Tyr by Trp would result in increased ET rates. However, all forward and backward charge transfers along the triad and subsequent possible charge recombination with FAD have to be considered. The comparison between PhrB and other members of the cryptochromephotolyase family shed light on protein ne-tuning, enhancement of FAD photoreduction rates and avoidance of charge recombination.</p><!><p>Site directed mutagenesis, protein expression and purication. A PhrB E. coli expression vector based on pET21b was used for recombinant expression of PhrB in ER2566 cells. The vectors for WT and the Y391F mutant are described in earlier publications. 19,33 To obtain the Y391A and Y391W mutants, site directed mutagenesis was performed according to the Quik Change mutagenesis kit (Agilent) using a pair of complementary primers (Table S1 †) with the desired mutation in the middle for initial polymerase reactions. Mutagenesis success and correctness of the sequences were conrmed by DNA sequencing. Expression and purication followed the procedure described in ref. 11 for WT and mutants. In brief, E. coli cells from agar plates were used for the inoculation of 3 l LB containing ampicillin. Following specic induction of recombinant expression with IPTG and subsequent incubation over night at 28 C, all purication steps were carried out at 4 C. Cells were harvested by centrifugation, suspended in 50 ml extraction buffer (50 mM Tris/HCl, 5 mM EDTA, 300 mM NaCl, 10% glycerol, pH 7.8) and extracted with a French Press (America Instrument Company) at 1000 bar. Following centrifugation and precipitation of soluble protein by ammonium sulfate (93% saturation), the protein pellet was suspended in EDTA free buffer. Soluble protein was puried by Ni affinity chromatography followed by size exclusion chromatography. The nal buffer was 50 mM Tris/HCl, 5 mM EDTA, 300 mM NaCl, 10% glycerol, pH 7.8.</p><p>Photoreduction measurements by UV/vis spectroscopy. PhrB WT and mutant proteins were diluted to a nal concentration of ca. 10 mM. The samples were incubated at 4 C in darkness in saturated oxygen solution. During this treatment, reduced FADH À is converted to oxidised FAD, although spectral analyses revealed that the fraction of oxidised FAD differed among the different proteins. Thereaer, 10 mM 1,4-dithiotreitol were added to the protein solution. UV/vis spectra were recorded using a Jasco V550 photometer with temperature control adjusted to 10 C. Aer the rst recording, the sample was illuminated with blue light emitting diodes (l max ¼ 470 nm) with a light intensity of 55 mmol m À2 s À1 at the position of the cuvette. Subsequent spectra were recorded at a series of time points as given in the results section. For data evaluation, complete spectra, 450 nm or 580 absorbance values, which stand for FAD in the oxidised or semireduced forms respectively, were presented.</p><p>Cofactor detection and repair assay. For detection of FAD and DMRL, 85 mM protein was denatured by 95 C incubation for 5 min. The insoluble protein and the soluble chromophores were separated by 15 000 Â g 10 min centrifugation and 10 mL supernatant were analysed by HPLC (Agilent system with a Gemini C18 column (50 Â 4.60 mm, 110 Å, Phenomenex)). The HPLC buffer conditions were: 5% acetonitrile (ACN) in 0.1% formic acid for 0-5 min; 5-75% ACN in 0.1% formic acid for 5-25 min. The ow rate was set to 0.75 ml min À1 and the column temperature to 25 C. Elution was monitored at 260 nm and 400 nm.</p><p>The photorepair reaction mixture contained 5 mM of the puried (6-4) photoproduct of t-repair_1 (Table S1 †) and 8.5 mM protein in repair buffer (50 mM Tris-HCl, pH 7.0, 1 mM EDTA, 100 mM NaCl, 5 mM MnCl 2 , 5% (w/v) glycerol, 14 mM 1,4dithiothreitol). Aliquots were irradiated with 400 nm light emitting diodes (250 mmol m À2 s À1 ) for 3 min. Thereaer, the reactions were stopped by heating to 95 C for 10 min. Samples were centrifuged at 15 000 Â g for 10 min and the supernatants analysed by HPLC (same column and system as above). The buffer conditions were: 7% acetonitrile (ACN) in 0.1 M triethylamine acetate (TEAA) (pH 7.0) for 0-5 min; 7-10% ACN in 0.1 M TEAA (pH 7.0) for 5-35 min. The ow rate was set to 0.75 ml min À1 and the column temperature to 25 C.</p><!><p>Model structures and molecular dynamics simulations. The structural model of PhrB WT has been derived from the X-ray crystal structure of Zhang et al. (PDB ID 4DJA). 17 The Y391F, Y391A and Y391W mutants have not been crystallized. For the setup of the model structures, we suppose that the mutation of Tyr391 does not affect the structure of the remainder of the protein. Starting from the PhrB-WT model structure, we replaced the aromatic cycle of Tyr391 by a phenyl, a methyl or an indol ring. Two conformations of the indol ring are allowed by the protein structure, but steric hindrance prevents rotation from one to the other (see Fig. 5). In the rst conformation, the Trp side chain can orient toward FAD being in a closer contact than in the second conformation. The conformations are termed Y391Wp (for proximal) and Y391Wd (for distal) in the following.</p><p>All mentioned simulations were performed with the GRO-MACS 5.0.4 package 34,35 using the AMBER-SB99-ILDN force eld. 36,37 The force eld parameters for neutral (oxidised) FAD and negatively charged cofactor FADc À were taken from ribo-avin and adenosine diphosphate (ADP) models developed in previous studies. 23,24 The GAFF parameters 38,39 were used for the DMRL antenna chromophore. The DMRL atomic charges were calculated by restrained tting on the electrostatic potential (RESP) 40,41 at HF/6-31G* (ref. 42 and 43) level with Gaussian 09 package. 44 Bonded parameters of the cubic FeS-cluster were taken from ref. 45 and the charges were taken from ref. 46.</p><p>The loop region from residues 180 to 182, which might impact the DNA binding abilities, 17 was not structurally resolved in the X-ray structure of WT PhrB. It was reconstructed using the MODELLER program. 47 WT and mutated proteins (Y391F, Y391A and Y391W) were solvated in a 106.24 Å3 cubic box lled by TIP3P water molecules. 48 Twelve sodium ions were added to create a neutral system.</p><p>Equilibration of the solvated proteins (WT and mutants) starts with a minimization step, followed by 100 ps MD in the NVT ensemble and 100 ps in the NPT ensemble. 100 ns of production NPT MD simulations were performed aerwards. Nose-Hoover thermostat 49 was used to keep a constant temperature at 300 K and Parinello-Rahman barostat 50 to keep the pressure at 1 atm. Covalent hydrogen bonds were xed on a constant length by the use of the LINCS algorithm. 51 The time step for the MD simulations was 2 fs.</p><p>Site energy and electronic coupling calculations. To treat the charge transfer processes, a quantum mechanical treatment of the active site has to be included via a so called combined Quantum Mechanics/Molecular Mechanics (QM/MM) scheme. The structural part of interest for charge transfer, which is treated at QM level, contains the side chains of amino acids involved in the called triad (A, B and C, see Fig. 2) and the isoalloxazine ring of FAD. The remaining atoms are treated classically using force elds (MM) and affect the QM zone by electrostatic interactions. Hydrogen link atoms 52 are inserted at the QM/MM boundary, namely in the C a -C b bond of A, B and C side chains or in the C1-C2 bond of the FAD D-ribitol tail.</p><p>To compute the electronic properties along the classical MD simulations, we use the semi-empirical Tight-Binding Density Functional theory (DFTB) method, 53 which is derived from density functional theory (DFT) but roughly 2-3 orders of magnitude faster than standard GGA-DFT methods with medium sized basis sets. Running a fragmentation of the QM region into several functional parts speeds up the calculation signicantly and allows to systematically correct for errors well known in DFT-GGA, like self-interaction error (for a detailed discussion see ref. 54 and 55). Each fragment is represented by only one frontier orbital and the chosen i-th fragment orbital (FO) of the fragment m is expressed in an atomic basis set c m and determined by FO coefficients c m i : 32,56</p><p>This FO-DFTB approach has been extensively evaluated and tested in previous publications 32,54,57 and has been so far successfully applied to describe the charge transfer in photolyase, 23 cryptochrome 24 and DNA 58,59 where the HOMO's of the fragments were used to calculate the evolution of electronic couplings and site energies.</p><p>The Hamiltonian H mn matrix is built out of the FO coefficients and the Hamiltonian H mn in the atomic-orbital-like basis function of the fragments:</p><p>The diagonal elements of the Hamiltonian matrix correspond to the site energies 3 m :</p><p>The HOMO energies for different molecules show nonsystematic errors, which can be corrected adding a constant energy shi depending on the chemical identity of the fragment. This correction was evaluated in ref. 58 for the relative energies of Trp and Tyr and expanded in this work to Phe and FAD relative to Trp (see Table S2 †). The energy gap between two sites can be related to the driving force of a charge transfer as described in the Marcus theory.</p><p>The off-diagonal elements of the Hamiltonian correspond to the direct electronic coupling H DA between two sites.</p><p>The FO-DFTB electronic coupling calculations were also validated in comparison with higher theoretical level results on set of organic stacked molecules. 56,60 These matrix elements (H DA ) can be used to compute two different charge transfer regimes. When large energy barriers occur along the charge transfer pathway, tunnelling matrix elements (T DA ) can be computed ('super-exchange tunnelling'), for small barriers, a 'direct' hopping-type mechanism may appear, which we treat using a QM/MM non-adiabatic MD technique, also based on the same computed matrix elements H DA .</p><p>Super-exchange tunnelling. In case of super-exchange tunnelling mechanism for a bridged charge transfer, the n fragments of the bridge B must be included in the electronic coupling T DA calculation. The system is thus divided into the donor/acceptor (D/A) and bridge subspace and an effective Hamiltonian is calculated in which off-diagonal elements correspond to T DA (for detailed discussions see ref. 54 and 61):</p><p>where b DA describes the direct electronic interaction between D and A, b Di and b jA the electronic interactions between D or A and the i or j-th fragment of the bridge. The tunnelling energy 3 tun corresponds to the average of the eigenvalues of the effective Hamiltonian and G ij is an element of the Green's function matrix, describing the probabilities for the electron to tunnel through the i-j space of the bridge. Direct electron transfer. We established in our previous work a multi-scale method to describe charge transfer reactions in cryptochromes and photolyases. 23,24 This non-adiabatic charge propagation scheme allows simulations where the charge transfer occurs on the same timescale as environment response charge state.</p><p>Nonadiabatic QM/MM MD simulations directly calculate the population of the charge which is located at a specic site of the QM zone. 62 This method doesn't force a specic charge transfer process; however it can describe a hopping or even a band-like conduction. 32,57 The wave-function associated with the transferred charge and atomic coordinates are simultaneously propagated solving time-dependent Schrödinger equation and classical Newton equations at every MD step, respectively. The excess charge corresponds to a second order perturbation of the neutral system Hamiltonian while environment interacts with the QM part as point charges. To take into account charge transfer, the partial charge of the QM part is calculated at every MD step to model the interaction of the moving charge with the charge distribution of the environment. More detailed description can be found in ref. 32, 57 and 62. As for the electronic coupling calculations, charge propagation simulations were performed using an in-house GROMACS 4.6 version. 63 The charge transfer between the triad and the isoalloxazine ring is initiated by the excitation of the FAD cofactor. The rst transfer continuing the excitation was studied in some photolyases to happen within one picosecond. 64 These two ultra-fast events were excluded in the previous studies 23,24 to reduce complexity and to focus on charge transfer along the Trp triad. To keep consistency, the same exclusion was also applied here and the QM part consists in A, B and C.</p><p>Charge propagation simulations start with an electronic state where FADc À is a radical anion and the rst site A a radical cation. The hole on the rst site is then propagated along the triad and moves in the reverse direction compared to the electron. These simulations were performed on PhrB WT and the Y391W mutant. Charge occupation of each site, varying from 0 (neutral state) to 1 (fully oxidized state) is followed during 1 ns QM/MM MD simulations. We randomly chose 20 to 25 starting structures from classical MD trajectories to guarantee well equilibrated systems while sampling different initial conditions. A kinetic model 23,24 was used to t the average occupations (from the individual simulations) of each site and thus determine the different reaction rate constants.</p><!><p>For photoreduction studies, we generated the Y391F, Y391W and Y391A mutants of PhrB. All mutants and WT PhrB were expressed in E. coli and puried by Ni chromatography and size exclusion chromatography. Whereas protein yields of Y391F and Y391A are comparable to those of WT, the yield of Y391W is ca. 10 times lower. Absorbance spectra of Y391A and Y391F in the oxidised FAD state are comparable with WT (Fig. 3), although detailed analyses reveal different chromophore to protein ratios and/or different fractions of reduced FAD at starting time. The absorbance of the Y391W in the blue spectral range is very weak (Fig. 3). Chromophore analyses show that this mutant contains only (1.3 AE 0.1)% FAD and (1.8 AE 0.2)% DMRL as compared to WT. These values are (94 AE 1)% and (87 AE 3)% for FAD and DMRL of Y391F, respectively, and (53 AE 2)% and (53 AE 2)% for FAD and DMRL of Y391A, respectively. We propose that the replacement of Tyr 391, which is located close to FAD, by bulky Trp in Y391W results in opening of the FAD pocket and loss of FAD binding capacity. The DMRL pocket is formed by amino acids of the N-terminus and more distant from the mutation. The loss of DMRL results therefore probably from FAD depletion. The partial loss of both chromophores to equal percentages in the Y391A mutant supports this idea. The 410 nm peak in the spectrum of the Y391W mutant is assigned to the iron sulphur cluster.</p><p>During blue light irradiation, the spectra of WT PhrB and the Y391F mutant change in a characteristic manner. The transient increase at 580 nm, the maximum of the protonated FAD semiquinone, and the loss of absorbance at 450 nm, characteristic for the loss of oxidised FAD (Fig. 4), are comparable to data published earlier. 33 Here, the relative A 450 nm decrease at t ¼ 90 min in the Y391F mutant appears smaller than in WT. This can be due to slower photoreduction or smaller fraction of oxidised vs. total FAD in the mutant. Both decay curves can be tted with monoexponential decay functions which yielded time constants of 36 AE 1 min and 32 AE 1 min for WT and Y391F, respectively. Thus, the rate of overall photoreduction is not affected by the Tyr to Phe replacement, but the oxidation state of Y391F was incomplete at the start of the photoreduction experiments. Formation and decay of the semiquinone intermediate absorbing at 580 nm is slower in Y391F (rise and decay times of 6 AE 0.3 min, 110 AE 30 min for WT and 7 AE 0.3 min, 200 AE 110 min for Y391F, respectively). This results shows that the role of Tyr or Phe differs in the rst and second electron transfer. In summary, the present and published data 19 show clearly that the replacement of Tyr by Phe into the proposed electron path does not block photoreduction. We do not observe any light induced absorbance changes in the Y391A mutant (Fig. 3). This result suggests that position 391 is critical for photoreduction, as proposed above. DNA repair in the presence of Mn 2+ (ref. 11) is complete aer 5 min for WT PhrB and Y391F mutant, whereas no repair activity is observed for Y391A and Y391W mutants under these conditions. When the repair time is prolonged to 120 min, Y391W repairs about (8.7 AE 0.7) % of damaged DNA, whereas with Y391A still no repair is observed.</p><!><p>We performed 100 ns classical MD simulations of PhrB WT and the Y391F, Y391A and Y391W mutant, based on the crystal structure of PhrB. In these simulations, no major conformational change of the overall protein structures is observed. The residues A, B and C (see Fig. 2) involved in the triad occupy similar positions in WT and the Y391F or Y391A mutants. In the model structure of the Y391A mutant, water molecules ll the space let by the replacement of Tyr to Ala (see Fig. S2 †). In WT, Y391F and Y391A, a water molecule, present in the crystallographic structure, interacts with isoalloxazine O4 and A backbone (see Table S3 and Fig. S3 †). The experimental ndings for the Y391W mutant, which has lost both chromophores, suggests a more drastic impact on the protein structure which would require simulation protocols dedicated to protein folding and FAD docking. Our simulations, which follow the dynamics of Y391W on relatively short time-scales, however, allow us to investigate the theoretical role of an amino acid independent on large structural changes that might occur on much longer time-scales. Such simulations help us to compare the relationship between environment and the rst site in different members of the cryptochrome-photolyase family. In the end, we have to combine experimental and theoretical results to obtain the maximum information for the role of each amino acid. The simulations for the Y391W mutant revealed two conformations of the Trp391 side chain (A) (see Fig. 5), denominated as Y391Wd and Y391Wp. In Y391Wd, A stays continuously close to the second Trp side chain B whereas in Y391Wp, A stays most of the time close and parallel to FAD isoalloxazine ring, but moves closer to B during a few nanoseconds (see Fig. 5, S4 and S5 †).</p><p>As shown in Table 1, the distances between all neighbouring sites (FAD-A, A-B, B-C) in WT and mutant structures are between 5 and 8 Å. These are typical nearest neighbour distances reported in our MD simulations of Arabidopsis cryptochrome and E. coli photolyase [22][23][24] which allow for sufficiently large electronic couplings in order to enable fast charge transfer. The larger distances between FAD and the second neighbour between 12 and 13 Å (FAD-B) suggest that a direct electron transfer from B to FAD is unlikely because the electronic coupling decreases exponentially with distance in absence of charge transfer bridge. We also report the distance between FAD or B and another tyrosine, Tyr395 (see also Fig. 6), which is close to the charge transfer chain. Mutations have no impact on the position of Tyr395; it stays at around 8 Å from FAD and slightly less than 7 Å from B.</p><!><p>Upon excitation of FAD by light or by energy transfer from the excited antenna chromophore, this site is reduced to the negatively charged FADc À species by an ET from the triad, which then in turn becomes positively charged. In a rst step, we investigate the electronic structure of the neutral FAD-triad system by computing site energies and electronic couplings. In a second step, we compute the changes in the electronic structure due to the charge transfer. In previous work, we have studied Arabidopsis cryptochrome and E. coli photolyase in a similar way. 23,24 Both proteins show a highly exergonic and fast charge transfer on a picosecond time scale. To investigate the effect of the mutation at site A, we compare the relevant parameters for charge transfer in the present study with these two reference systems.</p><p>The charge transfer parameters have been evaluated for oxidised FAD and a neutral triad along the 100 ns classical MD simulations on the neutral state of the protein using the FO-DFTB/MM scheme as discussed above. The orbital energies of A, B and C are called site energies and are a direct measure of the (relative) ionization potentials (IP) for the residue in the protein. The electronic couplings between the different partners are a measure for the charge transfer probability, i.e. they can be directly related to the prefactors in Marcus theory. 65 Average values of energies and couplings are reported in Fig. 7, the error bars indicate the associated standard deviation. According to these data, site energies of Trp390 and Trp342 (B and C) are similar, independent of the chemical nature of A. The electronic coupling values associated with the charge transfer between these Trp's are about 4.1-4.2 meV for WT and the Y391F mutant, 5.9 meV for the Y391A mutant and between 7.0 and 9.0 meV for the two rotamers of the Y391W mutant. These values are comparable with those found in Arabidopsis cryptochrome and E. coli photolyase. The small increase of electronic coupling from WT to Y391W corresponds to a small decrease in B-C distance in the mutant (Table 1). For reference, we also show the location of the HOMO level of neutral FAD, which is the electron acceptor aer excitation of one of its electrons.</p><p>The site energy of A as an aromatic residue (Trp, Tyr or Phe) inside the protein follows the same order as the HOMO energies in gas phase (Table 2, HOMO z ÀIP): HOMO(Trp) > HOMO (Tyr) > HOMO(Phe). In the protein, A has a value of about À5.8 eV for Trp, which is substantially decreased to À6.3 eV for Tyr and to around À7 eV for Phe. In the Y391W mutant, the energy of A is similar to energies of B or C, leading to a small energy gap between the charge transfer partners (within about 0.3 eV). Electronic couplings are twice as large as in PhrB WT or Y391F. These results support the possibility of a charge migration involving a Trp triad in the Y391W mutant. The distal conformation Y391Wd facilitates B / A charge transfer by decreasing the A-B energy gap by 0.08 eV and the average A-B distance by 0.64 Å compared with Y391Wp (Table 1 and Fig. 7). On the contrary, in the proximal conformation, the aromatic ring of A and the isoalloxazine ring are in a close position which must enhance A / FAD charge transfer.</p><p>As expected, the lower site energies of Tyr in PhrB WT and Phe in Y391F may present a barrier for hole transfer from FAD, as clearly seen in Fig. 7. While our previous studies on Arabidopsis cryptochrome or E. coli photolyase showed a hopping type mechanism along the Trp-triad, one may expect in WT and Y391F a tunnelling mechanism to be in operation, which would lead to less efficient charge transfer. We calculated the electronic coupling between FAD and B which involves A as a bridge for the WT and Y391F by a previously published method. 54 Our calculation shows a coupling of 0.06 and 0.02 meV for WT and Y391F which is ten-fold more than the direct (assuming no bridge residue) coupling between FAD and B (see ESI Table S4 †). Electron tunnelling through the A side chain seems possible, as also indicated by the experiments on Y391F. Tunnelling involving a Phe and protein backbone has been also described for the E. coli photolyase. 67 Therefore, in the case of Y391F, where a barrier of more than 1 eV is apparent, we denitely have to consider tunnelling, for the smaller barrier resulting from the presence of Tyr391 in WT this is not necessarily the case. This has been discussed for charge transfer in DNA, 68 where small charge transfer barriers up to 0.4 eV may easily be overcome, especially for short bridges due to molecular uctuations. The hole transfer between FAD and A falls in this range, allowing a direct hopping mechanism (Fig. 7). Electron hopping via Tyr would involve a transiently positively charged side chain, which must be followed by deprotonation. However, a nearby proton acceptor is missing in the PhrB structure, ruling out the deprotonation mechanism of Tyr391. Oxidised Tyr cannot be stabilised while the electronic coupling between A and B is strong (see Fig. 7). If FAD / A transfer occurs, the following hole transfer between A and B must be very fast.</p><p>Other charge transfer pathways can also be considered from the crystallographic structure. A charge transfer chain via Tyr399 and Tyr40 has been suggested in our previous experimental study. 19 Spectral changes related to FAD reduction in Y399F mutant are slightly slower compared to WT. However, the absence of FAD reduction in both W390F and W342 mutants clearly indicates that Trp390 and Trp342 are essential in the charge transfer process, which cannot be compensated by a transport via Tyr399 and Tyr40. Consequently, the Tyr399-Tyr40 charge transfer pathway was rejected.</p><p>Another possibility could be a transfer between FAD and Trp390 via Tyr395 (see Fig. 6). This residue could substitute site A (Tyr391), since it has similar distances to FAD and site B, as shown in Table 1. Interestingly, theses distances do not change upon mutation at site A, i.e. the FAD binding pocket and the connection via Tyr391 is very similar in all variants. According to the computed electronic couplings and site energies (Table S5 †), this pathway seems to be possible as well. However, our photoreduction studies for the Y391A mutant clearly indicate the absence of any charge transfer, i.e. the pathway does not seem to be a possible alternative. Although it has been shown that our scheme for the calculation of charge transfer couplings is very reliable for all couplings between the oxidisable residues, 60 the couplings between FAD and site A respective Tyr391 are very small, i.e. a small change in geometry could lead to a different result. For example, the PhrB structure in solution (experimental photoreduction measurements) could differ from the crystal structure used for calculations. Further, the couplings are computed for neutral FAD in the ground state, whereas the active state is an excited state, which may have some impact on the value of the coupling. As a consequence, within the accuracy of the calculations reported here, the rather small value reported for the FAD-Tyr395 coupling (Table S5 †) could as well turn out to vanish. Therefore, we believe that under the conditions the experiments are performed, the pathway via Tyr395 seems to be impeded. However, since an analogous alternative pathway has been reported for the PhrB homologue CryB, 18 it could be interesting to investigate under which conditions this pathway could be activated in PhrB. This, however, would require further investigations which are beyond the scope of the present work. For the main focus of this work, it is very interesting to note that for this alternative pathway a tyrosine is the primary electron donor of FAD, i.e. highlighting the role of tyrosine in the charge transfer cascade, which is the main emphasise of this work.</p><p>To estimate the effect of a water molecule bridging the FAD and the backbone of site A (see ESI †), we also applied the pathways model from Beratan and co-workers. 69 The couplings for these pathways are very small, therefore we did not consider these pathways further (see Table S4 and Fig. S6 †).</p><p>We now discuss the energetics and couplings for the case, where FAD and one of the residues of the triad are charged. The energetic landscape is changed drastically when a charged FAD and a charged A are considered (Fig. 8). As discussed recently, the fast charge transfer in cryptochromes and photolyases can be explained by a steep downhill energetics, which results from the interaction of the charge with the protein and solvent environment. 23,24 The rst ET results in negatively charged FADc À and a positively charged side chain on one site of the triad. Charge separation has a sizable effect on the site energies. The positive charge on A, B or C leads to a strong polarization of the environment. This polarized environment in turn leads to a stabilization of the charge at the respective site. In Marcus theory, this effect is called outer-sphere reorganization energy; in our simulations it manifests itself by the lowering of the site energies with regard to the neutral states. For each charge state on A, B and C, we compute the site energies along 1 ns MD simulations containing the charge on the respective site during the MD simulations, as shown in Fig. 8. We have discussed this effect in some details in our previous work, showing that the solvent has a distinct impact, in particular on those sites which are more solvent exposed, 24 like site C which is located on the protein surface.</p><p>In previously studied Arabidopsis cryptochrome or E. coli photolyase, the positive charge stabilization follows a downhill scheme, with an increasing energy gap between neutral and charge residue from A to C and where neighbouring sites have an energy difference of about 0.5 eV (Fig. 8). This leads to the fast charge transfer in a picosecond regime.</p><p>In PhrB, a similar stabilization on C occurs, this results from the solvent exposure. However, surprisingly also site A is massively stabilized, for Tyr even more than for both Trp rotamers. In the latter, A and B site energies are similar and the charge transfer from B to A no longer follows the downhill scheme described for the WT, E. coli photolyase or Arabisopsis cryptochrome. In total, the energy difference between site C and A in Y391W is only half of the value compared to the other two systems (Fig. 8), which has a drastic effect on the charge transfer equilibrium.</p><p>There is an obvious energetic difference for the site A Trp in Y391W, E. coli photolyase and Arabidopsis cryptochrome. In Y391W, A is nearly isoenergetic to B, while in the other proteins, positively charged B is more stable than A by about 0.4 eV. In our previous work, we have analysed structural reasons for this energy gap between A and B. In the E. coli photolyase or Arabidopsis cryptochrome, A is buried in a pocket with more than 5 Å distance from any water molecule, as documented by calculation of water distribution functions along MD simulations (Fig. 9). On the contrary, water molecules can easily move toward A in PhrB: the radial distribution function of water around Tyr391 presents a peak around 5 Å, while the rst peak around B is observed at 7 Å (Fig. 9). Moreover, a stable hydrogen bond network, involving Thr373, a water molecule, Tyr385, Thr358 and the backbone of Val354 (Fig. 6) can also participate in Ac + stabilization in our simulation. In E. coli photolyase and Arabidopsis cryptochrome, sites B and C are stabilised by solvent interactions, leading to a larger solvent reorganization energy. This explains the downhill energetics as shown in Fig. 8. On the contrary, water molecules close to the FAD-A complex in PhrB help to stabilize the charge-separated RP-A (FADc À -Ac + ) state, and compensate the unfavourable intrinsic IP of Tyr. Likewise, this could impact the charge transfer efficiency in the Y391W mutant by increasing the probability to localize the charge on the rst member of the triad.</p><!><p>To study the implication of the energetic landscape on the charge transfer dynamics, we performed unbiased simulations of the charge transfer through the residues A, B and C in WT protein and Y391W rotamers. Y391F and Y391A are not considered in this part as Phe and Ala cannot be oxidised. These simulations indicate the formation and lifetime of each radical pair state: RP-A (FADc À -Ac + ), RP-B (FADc À -Bc + ), RP-C (FADc À -Cc + ), summarized in Fig. 10. We present the averages over 25 simulations for WT and 20 simulations for the two Y391W mutants; more details and charge transfers movies are given in ESI. † The charge is considered to be on a specic site when the occupation of the site is larger than 50%. The time dependence of the average site occupations have been tted using a kinetic model previously described 23,24 which allows to obtain rate constants (Table 3) corresponding to the following steps:</p><p>In WT, the population of the rst radical state RP-A drops within a few ps and no back transfer is observed. The second state, RP-B, is transiently occupied by 60% of the total positive charge before decay and formation of the third state RP-C. Aer 25 ps simulation time, the charge distribution remains stable, with roughly 85% of the charge on the third state, while 15% remains on the second state. Due to the averaging over several simulations, these numbers show a statistical distribution rather than a charge delocalization over different sites. The positive charge is therefore well stabilized on the solvent accessible Trp342. No backward charge transfers from B to A occur during 1 ns of simulation time. The very fast rst charge transfer from Tyr391 to Trp390 shows a signicantly larger transfer rate than that calculated for plant cryptochrome. 1 PhrB k 23 and k 32 are consistent with values from Arabidopsis cryptochrome and E. coli photolyase, with a backward transfer 10 fold smaller than the forward one.</p><p>The Y391W mutants show a different behaviour. On average, the hole remains on RP-A during the rst 50 ps in Y391Wp and during the rst 21 ps in Y391Wd, respectively. The nal stabilization of 70-80% of the charge on site C occurs aer 70 ps for Y391Wp and 33 ps for Y391Wd. For the A-B transfer, the forward and backward rate constants are very similar in each rotamer simulation, as shown in Table 3. In Y391Wd, the k 21 rate is also close to k 23 . For the transfer from B to C, the forward rate is 10-fold higher than the backward rate. All rate constants are in the same order of magnitude as those of Arabidopsis cryptochrome or E. coli photolyase. The main difference between the Y391W mutant and Arabidopsis cryptochrome is the strong back transfer rate from B to A (k 21 in Table 3).</p><p>Indeed, during all charge transfer simulations of our different cryptochromes and photolyases proteins, we observe several backward transfers to A. No backward transfer is present in PhrB WT simulations, as the rst residue is a Tyr. We compare the number and the stability of backward transfers in Table 4 for different systems: Y391Wp/d and E. coli photolyase. In Y391W, we observe about 30 crossings in which the positive charge moves back to A and forms a stable RP-A for at least 500 fs which is ten-fold more than in E. coli photolyase. The difference between the two conformations of Y391W is related to the A-B distance: in Y391Wd, A is closer to B, which facilitates backward and forward charge transfer between them. On the contrary, in Y391Wp, A is closer to negatively charged FAD which contributes to stabilize the positive charge on A. Furthermore, a charge recombination on the isoalloxazine ring becomes more likely. Nevertheless, in 100 ns MD simulation of Y391Wp, motion of A to a distal conformation is observed and can also contribute to an enhanced charge transfer between the two Trp residues. In both conformations, the charge is more oen back transferred to A, but stays less time for one transfer than in E. coli photolyase.</p><!><p>In most members of the cryptochrome-photolyase family, the central ET pathway contains a triad of Trp residues, from the surface of the protein to the FAD chromophore. The mechanism commonly accepted for the FAD photoreduction consists of three successive hole transfer steps: FAD* / A, A / B and B / C. 31 Like other members of the cryptochrome-photolyase family, PhrB is able to reduce FAD upon light absorption via a long range ET involving aromatic residues. Site directed mutagenesis experiments have shown that two Trp, Trp390 and Trp342, are essential for the reduction process. The distance between the isoalloxazine ring and Trp390 is roughly 12-13 Å (Table 1) and thus too far from FAD for direct ET. There is no other Trp in the structure that can complete the triad, which raises the question of the role of the closest residue to FAD involved in charge transfer.</p><p>The presence of Tyr391 is intriguing: it is situated between FAD and Trp390 in a suitable place to take part in the ET, but its oxidation appearsat rst sightnot required for FAD reduction: (i) mutation of Tyr391 to redox inert Phe residue neither blocks FAD photoreduction nor DNA repair; (ii) in 464 PhrB homologs, this residue is either Tyr or Phe (Fig. S1 †). However, experimental mutation of Tyr391 to Ala blocks FAD photoreduction and DNA repair, underlining its relevance in electron transfer process. This observation rules out other alternative pathways e.g. through Tyr395 or water molecules.</p><p>In both WT and Y391F mutant we nd an energy barrier at site A. Therefore, we considered a super-exchange tunnelling process for both residues, which is a viable hypothesis to explain the experimental charge transfer between FAD and Trp390 in Y391F mutant where a large energy barrier occurs, but the computed couplings are sufficient to allow charge transfer according to this mechanism. 70 We observe similar electronic couplings between FAD and B when a Tyr or Phe aromatic cycle is included as a bridge. One can notice that in WT and Y391F, the Tyr391 or Phe391 aromatic cycle is parallel to the isoalloxazine ring of FAD and in a suitable conformation for pp orbitals interaction.</p><p>Hole transfer via tyrosines is usually not favourable due to the high ionisation potential, which is 0.6 eV larger than for tryptophans (Table 2). The oxidised state of tyrosines can be stabilized by proton transfer, but there is no proton acceptor in the neighbourhood of Tyr391. Therefore, an efficient hopping mechanism, where Tyr391 is transiently oxidized, seems improbable at rst sight. The energy of the Tyr391 HOMO, however, is signicantly reduced in PhrB. The Tyr391 oxidized state is stabilized by the protein environment instead, in the immediate environment a hydrogen-bond network (Fig. 6) can attract the proton from Tyr391 and allow the O-H bond elongation to compensate an electronic density decrease on the cycle. Therefore, the charge transfer from FAD to B via a transiently oxidized tyrosine seems to be possible. In addition to tunnelling, a much more efficient hopping regime seems to be enabled in PhrB due to the specic protein environment of site A. Mutation of Tyr391 to Phe disables this efficient pathway, but does not impede tunnelling as shown by both, experimental and theoretical results. Mutation to Ala, however, blocks the pathway, as shown in experiments and calculations, and thereby blocks FAD photoreduction due to vanishing electronic couplings.</p><p>Most photolyases and cryptochromes carry a Trp triad. In these cases, the strong exothermicity of the charge transfer results from stabilization of the positive charge by the solvent, as discussed in detail in our previous work. [22][23][24] The question arises therefore, why PhrB makes use of a Tyr at site A instead of a Trp residue.</p><p>Electron transfer can be described by an energy landscape as shown schematically in Fig. 11. It is clear that a Tyr substitution would introduce signicant barriers into the charge transfer pathway when it would be placed into the photoreduction pathway of e.g. E. coli photolyase. The barrier, when estimated using the gas-phase ionisation potentials (Table 2) is in the order of 0.6 eV, which surely would block an efficient hopping type charge transfer, as found for proteins with a Trp-triad. In PhrB, however, both Tyr and Trp at site A show a stronger stabilization, due to interactions with water and the protein environment, when compared to other members of the cryptochrome-photolyase family which we have already simulated. 23,24 Since site A in PhrB is very close to water molecules in the binding pocket, the intrinsic IP of Tyr and Trp are substantially Table 4 Average number n of backward transfer from B to A and average total time s of occupied A. The charge is considered to be on A when the occupation of the site is larger than 50%. A transfer is counted in s when the charge stays more than 500 fs on site A. stabilized by this polar environment. This stabilization results in a downhill hole transfer from FAD to C in PhrB WT but in similar energies for site A and B in Y391W mutant. The calculations suggest a reason for the presence of Tyr instead of Trp in PhrB. The forward transfer is less efficient for Tyr compared to Trp due to the slightly higher IP, however, the back-transfer seems to be too efficient in Y391W mutant, which may lead to unproductive cycles. The Y391W mutation, because of the higher intrinsic IP of Trp, disrupts the downhill energetics and allows charge recombination of FAD. The protein may therefore trade a slightly less efficient forward transfer for blocking backtransfer when using a tyrosine at site A.</p><p>Absorbance spectra showed that the Y391W mutant has lost both FAD and the antenna chromophore DMRL almost completely. The loss of FAD could be due to a modication of the chromophore pocket by Trp, which is larger than Tyr. The small percentage of DNA repair by the Y391W mutant suggests that a small fraction is still able to bind FAD. Quantitative comparisons between photoreduction of Y391W and WT are not possible experimentally. However, our simulations of the Y391W mutant (where loss of FAD was not considered) provide valuable insight of the importance of the Tyr residue in comparison with previously studied cryptochromes and photolyases. 23,24 The radical pair state of RP-A is stabilized in our Y391W simulations due to strong electronic couplings and small energy gaps between Trp390 and Trp391 and charge is transferred to Trp342 within 33-70 ps (Fig. 10). Although rate constants for the hopping mechanisms in the Y391W mutant of PhrB are comparable with the Trp triads in E. coli photolyase and Arabidopsis cryptochrome, 23,24 backward charge transfers to A occurs more frequently in the PhrB mutant than in these two proteins. Such back transfer increases the risk of charge recombination on FAD and hence a more likely inefficient charge transfer mechanism. If a Trp residue corresponded to the rst site and if the FAD binding were efficient, the environment would stabilize the RP-A state in PhrB obviously more than in Trp triads of other members of the cryptochromephotolyase family.</p><p>Taken together, our experimental and theoretical results indicate the following: the protein environment is quite different in PhrB from other groups of photolyases and cryptochromes. Residues bigger than Tyr at position A result in loss of FAD binding. Solvent can come closer to A, stabilizing the FADc À -Ac + state due to reorganization energy. The presence of a Tyr residue instead of a Trp at this site preserves the structure, the energetics, and therefore the function of PhrB.</p><!><p>There are no conicts to declare.</p>

Royal Society of Chemistry (RSC)

Mass Spectrometric Detection of Neuropeptides Using Affinity-Enhanced Microdialysis with Antibody-Coated Magnetic Nanoparticles

Microdialysis (MD) is a useful sampling tool for many applications due to its ability to permit sampling from an animal concurrent with normal activity. MD is of particular importance in the field of neuroscience, in which it is used to sample neurotransmitters (NTs) while the animal is behaving in order to correlate dynamic changes in NTs with behavior. One important class of signaling molecules, the neuropeptides (NPs), however, presented significant challenges when studied with MD, due to the low relative recovery (RR) of NPs by this technique. Affinity-enhanced microdialysis (AE-MD) has previously been used to improve recovery of NPs and similar molecules. For AE-MD, an affinity agent (AA), such as an antibody-coated particle or free antibody, is added to the liquid perfusing the MD probe. This AA provides an additional mass transport driving force for analyte to pass through the dialysis membrane, and thus increases the RR. In this work, a variety of AAs have been investigated for AE-MD of NPs in vitro and in vivo, including particles with C18 surface functionality and antibody-coated particles. Antibody-coated magnetic nanoparticles (AbMnP) provided the best RR enhancement in vitro, with statistically significant (p<0.05) enhancements for 4 out of 6 NP standards tested, and RR increases up to 41-fold. These particles were then used for in vivo MD in the Jonah crab, Cancer borealis, during a feeding study, with mass spectrometric (MS) detection. 31 NPs were detected in a 30 min collection sample, compared to 17 when no AA was used. The use of AbMnP also increased the temporal resolution from 4\xe2\x80\x9318 hrs in previous studies to just 30 min in this study. The levels of NPs detected were also sufficient for reliable quantitation with the MS system in use, permitting quantitative analysis of the concentration changes for 7 identified NPs on a 30 min time course during feeding.

mass_spectrometric_detection_of_neuropeptides_using_affinity-enhanced_microdialysis_with_antibody-co

Introduction<!>Reagents<!>Animals<!>Microdialysis Supplies<!>In vitro MD Experiments<!>In Vivo Microdialysis<!>UPLC-MS and UPLC-MS/MS Analysis and Data Processing<!>In vitro recovery enhancement<!>In vivo recovery enhancement<!>Conclusions

<p>Microdialysis (MD) is a sampling technique that allows collection of signaling molecules from an animal while it is alert and behaving, with minimal disturbance to the animal. In this technique, a MD probe is implanted into the tissue of interest and perfused with liquid at a flow rate in the range of 0.1–10 μL/min. The tip of this MD probe consists of a dialysis membrane, having pores with a defined molecular weight cutoff (MWCO). Molecules below this MWCO near the tip of the probe passively diffuse into the probe and are then carried by the slowly-moving liquid out of the probe, through a length of tubing, and finally to a sample collection vial or analysis system. This technique has been used successfully to collect a variety of different molecules from a number of tissues in several species, and has provided important insights into the action of compounds in vivo in a minimally perturbed animal.1,2</p><p>MD is of great utility in neuroscience, in which time-resolved changes in neurochemistry during the performance of a behavior or exposure to a stimulus are of interest. Continual collection of neurochemicals without disturbing the animal to obtain the samples allows the experimenter to determine the molecular underpinnings of neuronal activity related to these events, in the absence of any sampling-induced neuronal changes. MD has been used successfully to monitor small molecule neurotransmitter (NT) changes in vertebrate animals under a variety of different conditions, and has contributed greatly to our understanding of the effects of NT release on behavior.1,3–5</p><p>One area that is particularly challenging for MD sampling is the analysis of larger molecules, such as neuropeptides (NPs), which are below the MWCO of the probe but are in the mass range of 500–10,000.5–9 A number of complex factors make recovery of NPs difficult. One such reason is the lower relative recovery (RR) of NPs in comparison to small molecules due to their larger size hindering passage through the dialysis membrane. The RR is calculated by taking the concentration of an analyte collected through MD divided by the concentration outside the probe, and is usually expressed as a percentage. This RR is inversely related to the mass of the molecule, with larger molecules typically having RRs of less than 50%. Another challenge is the low endogenous concentration of these compounds. NPs are present in vivo at the nM - pM concentration range. 2 Therefore, the concentration collected, governed by the laws of passive diffusion, is reduced compared to NTs due not only to their low endogenous concentrations but also their reduced RR. Furthermore, RR is also governed by the amount of time the liquid is in contact with the membrane (the MD flow rate, FR), with lower flow rates leading to greater RRs.2,8,10 If increased amounts of analyte are desired, a longer collection time can be employed. If a short collection time is desired, an experiment will detect NPs reliably only if the RR is improved by other means,7 or a more sensitive detection technique is employed.</p><p>Progress has been made in using highly sensitive and specific detection methods for NPs in samples obtained via microdialysis, relying mainly on mass spectrometry (MS).1,3,4,11–13 Some of these studies use MS for surveys of NP content and identity.14–23 Other studies use MS for quantitation of identified NPs in microdialysate, mostly with selected reaction monitoring (SRM) of daughter or granddaughter ions.2,10,24–30 Finally, microdialysates can be analyzed via MS-based techniques for NP discovery combined with less precise quantification methods commonly used in proteomics.31–36 In addition to MS-based analysis of dialysates, other sensitive techniques, including those that rely on immunochemical or spectrophotometric detection, have been used for quantitation of NPs in dialysates, but these methods lack specificity.1,3,13 Although MS instruments are highly sensitive, not all perform adequately in the concentration range at which NPs are present in vivo, and other methods to increase sensitivity must be investigated.</p><p>An important method to increase the sensitivity of NP detection in MD is to increase the RR. The relative recovery can be increased by adding affinity agents (AAs) to the liquid perfusing the probe. The analytes form interactions with the AAs and lead to reduction of free concentration of analytes in the dialysate, thus increasing the concentration gradient for analytes which drives mass transport to allow additional analytes to diffuse into the probe.7 This technique is termed affinity-enhanced microdialysis (AE-MD), and has been used by a number of different researchers with a wide variety of compounds. 5–7,17,34,35,37–52 Stenken and colleagues, among others, have achieved success in improving the recovery of cytokines in vitro and in vivo 7,39,51,53,54 and neuropeptides in vitro,43 using free antibody, cyclodextrins, and micron-sized beads coated with antibodies or heparin.</p><p>AE-MD is not yet optimal, as saturation of the beads can occur, leading to non-linear recovery enhancement. Clogging or settling of the beads in solution is also a major concern. The technique has also not yet been applied to study NPs in vivo (although the cytokine CCL2 has been studied in rats using AE-MD54), nor have smaller beads been employed as affinity agents. In this work, several AAs are tested for enhancement of NP recovery. Nanoscale magnetic beads are developed for use as AAs, with the advantages of reduced settling rate and greater binding capacity. They enhance recovery of 4 out of 6 NP standards tested in vitro. They are also employed in vivo to study the time course of NP release following feeding in the Jonah crab, Cancer borealis. This new affinity agent for AE-MD greatly increases the utility of this technique for monitoring peptide secretion during behavior.</p><!><p>Peptide standards (bradykinin (BK), somatostatin-14 (SMT), substance P (SP), Homarus americanus FMRFamide-like peptide I (FLP I), H. americanus FMRFamide-like peptide II (FLP II), and FMRFamide) were purchased from American Peptide (Sunnyvale, CA, USA) and used without further purification. C18 silica microparticles (C18SμP) were purchased from Varian (now Agilent Technologies, Santa Clara, CA, USA) and were 5μm in size with 300Å pore size. They were used in perfusate at a concentration of 0.2mg/mL, or 3.04 × 103 beads/μL. C18 magnetic microparticles of 1μm diameter (C18MμP) were purchased from Varian at a stock concentration of 2 × 106 beads/μL and used in perfusate at 3.3 × 104 beads/μL Magnetic microparticles pre-coated with protein G were purchased from New England Biolabs (Ipswich, MA, USA), with a stock concentration of 3.11 × 104 beads/μL and a final perfusate concentration of 518 beads/μL. Magnetic nanoparticles of 100nm diameter pre-coated with protein G were purchased from Chemicell GmbH (Berlin, Germany) with a stock concentration of 1.8 × 1010 beads/μL, and thus perfusate concentrations of 3.0 × 108 beads/μL and 1.8 × 109 beads/μL as indicated below. Bovine serum albumin (BSA) and formic acid (FA) were purchased from Sigma-Aldrich (St. Louis, MO, USA). Polyclonal rabbit anti-FMRFa antibody was purchased from Abcam (Cambridge, MA, USA). All other chemicals were purchased from Fisher Scientific (Pittsburgh, PA, USA) at ACS reagent-grade and used without further purification. ACS reagent-grade solvents and Milli-Q water were used for sample preparation. Optima grade solvents were used for operation of the UPLC-QTOF. C18-coated magnetic beads and antibody-linked beads were prepared and used as recommended by the manufacturers. Details of preparation, unbinding, and in vitro bead binding assays can be found in the supplementary information.</p><!><p>Jonah crabs (Cancer borealis) were purchased from Ocean Resources, Inc. (Sedgwick, ME, USA) and The Fresh Lobster Company (Gloucester, MA, USA). These crabs were wild-caught and shipped overnight packed on ice. The crabs were then maintained in an artificial seawater tank at 10–12°C, with crushed gravel as a substrate. Details of animal housing procedures are included in the supplementary information.</p><!><p>CMA/20 Elite probes with 4 mm membranes of polyarylether sulfone (PAES) were purchased from CMA Microdialysis (Harvard Apparatus, Holliston, MA, USA). All MD probes were rinsed with water prior to use. Several pumps were used, including a CMA/102, a KD Scientific 100 (KD Scientific Inc., Holliston, MA, USA), and a Harvard 22 (Harvard Apparatus, Holliston, MA, USA). When required, additional FEP (CMA) or PEEK (Upchurch-Scientific, Idex Health and Science, Oak Harbor, WA, USA) tubing was connected to the tubing of the probe by flanged connectors from CMA, and BASi (West Lafayette, IN, USA). BD (Franklin Lakes, NJ, USA) plastic syringes were typically used. Flanged connectors were used to connect 21 gauge Luer-lock needles (included with CMA 20 series probes) blunted by grinding with a rotary tool (Dremel, Robert Bosch LLC, Farmington Hills, MI, USA) to the probe tubing.</p><!><p>For in vitro experiments, the tip of the probe was immersed into a vial with a home-built apparatus to hold the probe in place. Typically, 3 different probes' tips were immersed in the solution in the vial concurrently to provide multiple experimental replicates. The vial contained microdialysis medium, which consisted of a phosphate buffered saline (PBS) solution with neuropeptide standards of interest dissolved in it at known concentrations, in the 1–5 micromolar range, except for C18 silica particles, which used 50μM. The vial with probe holder was placed on an orbital rotating platform to produce constant mixing. A sample of the medium was taken prior to starting microdialysis and following the experiment. The probe was allowed to equilibrate at the flow rate of the experiment (0.5 μL/min) for 30 min before starting dialysate collection. Technical replicates were taken as consecutive 30 min samples of the liquid flowing out of the tubing. Samples from the medium taken before and after the experiment were used to determine the relative recovery percentage. A minimum of three technical replicates were obtained per experiment, and a minimum of three experimental replicates were obtained, each coming from either a different probe or a different instance of setting up and conducting the experiment. Medium samples and samples containing no AA were placed immediately in a 96-well sample plate for UPLC-QTOF analysis. For AE-MD, NPs were unbound from AAs as recommended by the manufacturer (see Supplemental Information for details) and combined with the liquid portion of the sample in a 96-well plate for analysis. The percent of beads passing through the probe was determined by counting on a hemacytometer for micron-sized beads, and by comparing the dry mass of particles for nanoscale beads.</p><p>When affinity agent was used, a clean steel ball bearing of appropriate size (1/8 inch, Wheels Manufacturing, Louisville, CO, USA) was added inside the barrel of the syringe delivering microdialysate. The pump was placed into a rocking platform shaker with the syringe placed at an angle to the axis of rotation of the shaker. The rolling of the ball bearing inside the syringe kept the affinity agent in solution.55 For the affinity agent perfusate, the equivalent of 50μL of bead solution was diluted to 3mL with PBS (a dilution of 1:60), with one exception. Although the concentrations of beads in mg/mL varied, it was determined that they had equal activity per mL, as the manufacturers' protocols recommended the same ratio of beads to sample, i.e. 50μL beads with 0.5 mL cell lysate, a 1:10 ratio. In one set of experiments, a higher concentration of affinity agent was used, as it was possible to increase bead concentration without adverse experimental effects. This trial is noted as 6× AbMnP, containing six times as many nanoparticles per unit volume (50μL of bead solution diluted to 0.5mL of perfusate).</p><!><p>The procedure for implantation of a MD probe was adapted from previous publications.14,15 A detailed description of the implantation and modifications are presented in the supplementary information. The probe was surgically implanted in the crab 2 days prior to the first feeding experiment, and the last feeding experiment was conducted 8 days after surgery. This time window was chosen to avoid effects from surgery (stress of anesthesia and being out of water, trauma to the hypodermis) and tissue growth over the probe's active membrane. Artificial crab saline (440 mM NaCl;11 mM KCl; 13 mM CaCl2; 26 mM MgCl2; 10 mM HEPES acid, pH 7.4, adjusted with NaOH) was used as the basis for perfusate. The flow rate was 0.5 μL/min, supplied by a programmable syringe pump (KD Scientific Model 100, Holliston, MA, USA) and samples were collected every 30 min with a refrigerated fraction collector (BASi Honeycomb, Bioanalytical Systems, Inc. Indianapolis, IN, USA). For each feeding trial, a 30min sample was acquired prior to feeding the animal but after allowing the probe to equilibrate for 30min, and this was used as the baseline sample. This experiment was conducted 3 times on the crab under normal MD conditions. For AE-MD in vivo, a 1:10 dilution of nanobeads was used (equal to the high AbMnP concentration for in vitro AE-MD studies). Upon collection, 1.5 μL of formic acid was added to each sample to improve NP stability2 and unbind NPs from the antibody-coated nanoparticles, and an internal standard (bradykinin, 1μM) was added for quantitation. Samples collected without affinity agent were directly injected onto the UPLC-MS system, and magnetic beads were removed from AE-MD samples prior to addition of internal standard and MS analysis.</p><!><p>In vitro MD samples were analyzed via a UPLC-MS approach. A Waters nanoAcquity UPLC system (Waters, Millford, MA, USA) was used in conjunction with a home-packed capillary column (360 μm OD, 75 μm ID, 10cm long, Magic C18 particles (Michrom, Auburn, CA, USA), 3 μm diameter, 100Å pore size) with integrated laser-pulled (approx. 7 μm diameter, with a Sutter Instruments P-2000 (Novato, CA, USA)) ESI emitter tip. Details of the UPLC-MS parameters and data analysis methods can be found in the supplemental information, including a Supplemental Table S1, which enumerates the retention times of the peptides using a reversed phase separation (H2O/ACN/0.1% formic acid) with gradient from 95% aqueous to 95% organic over 30 min. Statistical significance was determined using JMP statistical software (Version 9.0.2 SAS Institute, Inc., Cary, NC, USA). Graphs illustrating this data were generated in Microsoft Excel 2010 (Microsoft Corporation, Redmond, WA, USA).</p><p>In vivo MD samples were analyzed using the same instrumentation platform but with different MS and LC specifications. The LC gradient was 60 min long and a larger MS window was monitored for quantitative analyses. Data analysis is detailed in the supplemental section.</p><!><p>Due to previous work using column packing materials as AAs,34,35,38,39,43,45,49 initial experiments employed C18 silica microparticles as a generic, easily obtained AA (Supplemental Figure S1). These experiments led to a modest increase in NP recovery, but problems in their implementation, including bead settling/clogging (only ~25% pass through the probe and tubing) and the need to use additives (BSA) to improve bead dispersion made them impractical for in vivo use. In order to obtain non-specific affinity enhancement of NPs, another type of particles that had C18 surface functionality but also magnetic cores for simplified sample handling and other surface modifications for increased water solubility were employed, C18 magnetic microparticles (C18MμP). These 1 μm diameter particles are commonly employed for removal of salts from biological samples prior to analyses that are sensitive to salt, such as mass spectrometry. Results obtained using C18MμP as affinity agents are presented in Fig. 1 and Tables 1 and S2. The C18MμP significantly enhanced the recovery of 4 peptides—FLP I, FLP II, SP, and SMT. Recovery was at least doubled, with final RRs of several compounds at 50% or higher. However, settling and clogging were still observed due in part to the propensity of the particles to attract each other via magnetism. The settling observed was less than the C18 silica particles due to surface modifications of these particles for improved aqueous solubility.</p><p>Based on previous work employing antibody-coated microspheres 6,7,37,39,41–43 to improve the recovery of cytokines and neuropeptides, antibody-coated microparticles were also employed for AE-MD. Commercial magnetic immunoprecipitation (IP) kits and a commercially available anti-FMRFa antibody were used to create antibody-coated beads. Traditional agarose bead-based IP kits contain beads of ~140 μm diameter, which is unsuitable for passage through the probe and tubing. Magnetic IP kits employ smaller beads (diameters 1–5 μm), and have the advantage of simpler separation. Magnetic microparticles coated with protein G were linked to rabbit polyclonal anti-FMRFa as recommended by the manufacturer to create antibody-coated magnetic microparticles (AbMμP) and added to the perfusate.</p><p>The relative recoveries obtained with and without AbMμP in the perfusate are enumerated in Fig. 2 and Tables 1 and S2. No significant increases in RR were obtained for the NP standards, with the exception of SMT, whose RR was enhanced significantly (p=0.002) by about 2.5-fold. These results do not mirror findings of in vitro bead-binding assays (Supplemental Figure S2), in which SMT bound poorly to AbMμP and other NPs had a high degree of binding. Bead settling and tube clogging were observed to a similar extent as observed with the magnetic C18 particles, and this could explain these contrary results—beads with bound NPs may have remained stuck in the tubing. Approximately half of the AbMμP passed through the probe and tubing.</p><p>Magnetic nanoparticles of 100 nm diameter were the final affinity agent tested in this study. These particles are also coated with protein G and were conjugated to the same anti-FMRFamide antibody (AbMnP). Particles of this size do not have permanent magnetic fields and thus are not attracted to each other in the absence of an external magnetic field.56 This greatly reduces settling of the beads in the syringe and tubing. The smaller size of these particles also reduces settling, and it was observed that around 100% of the particles pass through the tubing. Two different concentrations of nanoparticles were used; one that was equal to that used for the magnetic microparticles and C18 beads, and one that was 6 times concentrated (6× AbMnP), possible due to the reduced settling and lack of magnetic interaction between nanoparticles.</p><p>RR enhancements are shown in Fig. 3 and Tables 1 and S2. Statistically significant (p<0.05) recovery enhancements were obtained with AbMnP for 3 of the 6 NPs. When the concentration of nanobeads was increased, 4 out of 6 NPs had significantly enhanced RRs. FMRFa showed a strong trend toward enhanced recovery, but this trend did not meet the statistical significance threshold. The recovery of bradykinin was not enhanced by any of the affinity agents, likely due to its small, hydrophilic character and lack of an amidated C-terminus. From the data obtained here for several NPs with different sequences, it is fair to assume that the antibody used primarily recognizes C-terminal amidation, followed by hydrophobic and basic amino acids. As a side note, the promiscuity of this antibody provides additional support for the use of mass spectrometry as an unbiased technique for NP analysis.</p><p>One concern that thus becomes important due to non-specific binding to antibodies is saturation of the beads' binding capacity. High-concentration components of a complex biological sample (or fragments thereof), such as albumin in mammals and cyanin proteins in crustaceans, could fully occupy these sites in vivo, and thus binding of low-concentration biological molecules of interest to the beads will be non-linear and unreliable for accurate representation of in vivo concentration changes. Thus, antibody-linked beads in microdialysis perfusate should be used with caution when attempting to accurately determine the concentration changes of analytes in the extracellular environment.</p><!><p>Several proof-of-principle tests to determine the suitability of affinity-enhanced microdialysis for in vivo application were conducted. Jonah crabs (Cancer borealis) were implanted with microdialysis probes following a modified version of a published technique.14 The probes of several crabs were perfused with crab saline or antibody-linked nanobeads in crab saline solution.</p><p>Representative data for baseline NP content in microdialysis samples obtained with and without AbMnP in the perfusate is shown in Fig. 4. Here, extracted ion chromatograms (XICs) for two FMRFamide-like peptides (FLPs) in samples obtained under baseline conditions with and without AbMnP are displayed. These samples were obtained from the same crab at the same time period on different days, with the AbMnP experiment conducted on the last day, so probe fouling is not a factor. They were also analyzed on the same day after storage under acidic conditions, which have been shown to stabilize MD samples2. XICs are plotted with the same y-axis scale after smoothing and baseline subtraction. FLP peaks from samples obtained without AbMnP have intensities that are 20 to 40% as intense as those obtained with AbMnP. Only very slight retention time shifts are observed, likely due to changes in ambient temperature. The UPLC column was kept at room temperature, which varies several degrees throughout the day. A similar qualitative enhancement of peak intensity was observed in two other AE-MD experiments, conducted on different crabs, the results of which are not presented here.</p><p>Quantitative analysis also indicates that AbMnP improve NP recovery in vivo. Table S3 enumerates NPs previously detected in MD and whether these compounds were also detected in samples obtained during a feeding experiment with or without AbMnP in the perfusate. Compared with previous work in which only 35 NPs were detected in samples collected over 4–18 hours and concentrated ~100-fold prior to analysis,14 the detection of 31 NPs in a sample collected over only 30 min and analyzed without preconcentration is a great enhancement in sensitivity. Microdialysate from the same crab obtained with simple crab saline as perfusate only contained an average of 25 NPs, with 17 of those detected in all three feeding trials. Therefore, the increased NP detection sensitivity is due not only to enhancements in UPLC-MS sensitivity that have occurred since the previous work was published, but also to the affinity-enhanced recovery by AbMnPs.</p><p>For many trials without affinity agent in the perfusate, it was possible to detect NPs with reasonable reproducibility across trials, but their abundance was so low that reliable quantity changes could not be observed. In other words, the NPs were present at their lower limit of detection (LLOD), which is below their lower limit of quantification (LLOQ). Addition of affinity agents increased the concentration of the NPs for collection and detection to a higher level, and thus reliable quantitation could be conducted. While concerns about AA active site saturation should still be addressed by validation of observed NP trends by more sensitive techniques, affinity enhancement improves MD from a mostly qualitative technique to a quantitative technique by increasing the concentration of NPs to above their LLOQ. The LLOQ, LLOD, and linear range of this detection method when used with BK as an internal standard are illustrated in Fig. S3. For the LC-MS system employed, the LLOQ of these peptides is around 45 μM, and the LLOD is ~5 μM. Thus, NPs must be enriched for reliable quantitation.</p><p>AE-MD was employed to enrich NPs collected in vivo. Results are indicated in Figs. 5 and S4, which describe relative concentration changes in seven identified crab NPs following feeding. All data points were obtained from the same crab; three feeding trials were conducted without AA, and one was conducted with AbMnP. The means and SEMs are plotted for the no AA trials. The NPs shown had low variance between the three no AA trials, and were detected in all time points. However, their concentrations do not appreciably change as they cannot be reliably quantified. Dynamic changes in these peptides are observed when AbMnPs are added to the perfusate, increasing the concentration of the NPs to a level at which they can be quantitated. This result has been replicated in a separate C. borealis feeding trial (unpublished data).</p><p>Most of the NPs appear to increase in concentration following feeding; reaching peak levels around 105 min after feeding. It is possible, therefore, that these NPs are neurochemicals involved in feeding. In the authors' experience, crabs start feeding immediately after food is presented and stop 15–30 min. later, while still holding on to the food. They then recommence eating a few minutes later, around 60–90 min. after food is presented. This may occur as a result of the crab filling its stomach with food, allowing that food to be processed by the stomach and passed on to more distal parts of the digestive tract, and followed by repetition of the process. Since some of the NPs gradually increase before reaching peak concentration (RYLPT, SDRNFLRFa, YSFGLa, and pERPYSFGLa), these peptides might be continually released into the hemolymph while the animal is eating, and reaching a certain peak level could be an indicator of satiety. Other NPs that have a more dramatic spike in concentration levels (GPRNFLRFa, I/LNFTHKFa, and NFDEIDRSGFGFN) could be released after satiety is reached. Although these results are preliminary and further investigation will be required, they demonstrate how AE-MD can be used to observed dynamic changes in NPs that would otherwise appear unchanged in typical MD experiments, due to their presence below their LLOQ.</p><!><p>AE-MD has been conducted with a variety of AAs. AbMnP provided the greatest enhancement in neuropeptide recovery and could be used at higher concentrations than AbMμP due to their non-aggregation. This permitted statistically significant recovery enhancements for 4 out of 6 NP standards tested. The recovery enhancement was not specific to compounds with sequence similarity to FMRFamide, the compound against which the antibody was generated, and thus the mechanism of recovery enhancement is non-specific. This is likely due to the high cross-reactivity of most NP antibodies, and although it could be detrimental if one desires to enrich a single NP for analysis due to specific interest in that compound, it will provide greater success in NP discovery and survey experiments, in which multiple and/or unknown NPs are of interest. These AbMnPs were used in experiments in the crab to determine the roles of circulatory NPs in feeding. They increased NP identification and made quantitation of NPs possible during a dynamic process. This will enable assignment of putative function for several NPs in feeding. AE-MD with AbMnP is a technique that has great potential to enrich NPs in microdialysate for correlation of function with molecular identity.</p>

PubMed Author Manuscript

Formation of todorokite from “c-disordered” H+-birnessites: the roles of average manganese oxidation state and interlayer cations

BackgroundTodorokite, a 3 × 3 tectomanganate, is one of three main manganese oxide minerals in marine nodules and can be used as an active MnO6 octahedral molecular sieve. The formation of todorokite is closely associated with the poorly crystalline phyllomanganates in nature. However, the effect of the preparative parameters on the transformation of “c-disordered” H+-birnessites, analogue to natural phyllomanganates, into todorokite has not yet been explored.ResultsSynthesis of “c-disordered” H+-birnessites with different average manganese oxidation states (AOS) was performed by controlling the MnO4−/Mn2+ ratio in low-concentrated NaOH or KOH media. Further transformation to todorokite, using “c-disordered” H+-birnessites pre-exchanged with Na+ or K+ or not before exchange with Mg2+, was conducted under reflux conditions to investigate the effects of Mn AOS and interlayer cations. The results show that all of these “c-disordered” H+-birnessites exhibit hexagonal layer symmetry and can be transformed into todorokite to different extents. “c-disordered” H+-birnessite without pre-exchange treatment contains lower levels of Na/K and is preferably transformed into ramsdellite with a smaller 1 × 2 tunnel structure rather than todorokite. Na+ pre-exchange, i.e. to form Na-H-birnessite, greatly enhances transformation into todorokite, whereas K+ pre-exchange, i.e. to form K-H-birnessite, inhibits the transformation. This is because the interlayer K+ of birnessite cannot be completely exchanged with Mg2+, which restrains the formation of tunnel “walls” with 1 nm in length. When the Mn AOS values of Na-H-birnessite increase from 3.58 to 3.74, the rate and extent of the transformation sharply decrease, indicating that a key process is Mn(III) species migration from layer into interlayer to form the tunnel structure during todorokite formation.ConclusionsStructural Mn(III), together with the content and type of interlayer metal ions, plays a crucial role in the transformation of “c-disordered” H+-birnessites with hexagonal symmetry into todorokite. This provides further explanation for the common occurrence of todorokite in the hydrothermal ocean environment, where is usually enriched in large metal ions such as Mg, Ca, Ni, Co and etc. These results have significant implications for exploring the origin and formation process of todorokite in various geochemical settings and promoting the practical application of todorokite in many fields.Graphical abstractXRD patterns of Mg2+-exchanged and reflux treatment products for the synthetic “c-disordered” H+-birnessites.

formation_of_todorokite_from_“c-disordered”_h+-birnessites:_the_roles_of_average_manganese_oxidation

<!>Background<!>Preparation of “c-disordered” H+-birnessites with different Mn AOS values<!>Transformation of “c-disordered” H+-birnessite to todorokite<!>Characterization of “c-disordered” H+-birnessites and their products<!><!>Transformation from “c-disordered” H+-birnessite to buserite and todorokite<!><!>Transformation from “c-disordered” H+-birnessite to buserite and todorokite<!>The effect of interlayer Na+ and K+ on the formation of tectomanganates<!>The role of Mn AOS in transformation of “c-disordered” H+-birnessite into todorokite<!>Conclusions

<p>XRD patterns of Mg2+-exchanged and reflux treatment products for the synthetic "c-disordered" H+-birnessites.</p><!><p>Todorokite, a 3 × 3 tunnel structure with corner-sharing triple chains of MnO6 octahedra, is a naturally occurring manganese (Mn) oxide found in terrestrial Mn ore deposits, weathering products of manganese-bearing rocks, and marine Mn nodules [1–3]. Due to the superior characteristics of todorokite in ion exchange, specific surface area, thermal stability and molecule-sized tunnels [4], it has many potential industrial applications as molecular sieves, lithium-manganese-oxide cathode materials, heterogeneous catalysts and sensors [4–8], and it also plays an important role in cleaning up natural water and controlling the concentrations of heavy metals in soil solution [9].</p><p>Todorokites are often obtained by hydrothermal treatment or refluxing process from triclinic birnessite with a layered structure [4, 10, 11]. Oxidation of the Mn(II) located above or below vacant sites facilitates the formation of a stable todorokite-like structure during marine diagenesis and hydrothermal process [12]. Liu et al. [13] proposed that some octahedral Mn(III) from the layers of buserite could migrate into the interlayer and become corner-sharing octahedra that assist in the formation of the "wall" of the tunnels during hydrothermal heat treatment. Cui et al. [8, 14] also reported that the formation of todorokite under reflux conditions is closely related to the Mn(III) content and the content of Mn(III) in birnessite depends partly on the Mn AOS values. The degree of transformation of triclinic birnessite samples to todorokite decreases with the increasing of their Mn AOS values [8].</p><p>Recently Atkins et al. [15] found that "c-disordered" H+-birnessite with a poorly crystalline, hexagonal layer symmetry can also be transformed into todorokite under a mild reflux procedure. In addition, in the natural environment the microbial oxidation products of Mn(II) are mainly birnessite-like phases exhibiting poor crystallinity and hexagonal symmetry [16–19]. Thus, the formation of todorokite and poorly crystalline phyllomanganates are closely associated in natural marine environments. However, the effect of the preparative parameters on the transformation of "c-disordered" H+-birnessites into todorokite has not yet been explored.</p><p>Here we prepared "c-disordered" H+-birnessites with different Mn AOS by controlling the MnO4−/Mn2+ ratio in low-concentrated NaOH or KOH media. The transformation of "c-disordered" H+-birnessite to todorokite was achieved under a mild reflux condition, elucidating the roles of Mn AOS and the interlayer K+ and Na+ of the "c-disordered" H+-birnessite in the transformation. The results are expected to provide clues for understanding the transformation of poorly crystalline phyllomanganates with hexagonal symmetry to todorokite and shed light on the mechanisms of todorokite formation in earth-surface environment.</p><!><p>Synthesis of "c-disordered" H+-birnessite was performed following the method of Villalobos et al. [17]. A solution of 10 g of KMnO4 (0.1977 M) in 320 mL of distilled deionized water (DDW) was added slowly to a solution of 7.33 g of NaOH (0.51 M) solution in 360 mL of DDW while stirring. Subsequently, a solution of 23.26 g MnCl2 (0.3673 M) in 320 mL of DDW was added dropwise to the mixture while stirring vigorously to form precipitate. The suspension was left to settle for 4 h, following the supernatant was discarded. The pH value was 3.1. The remaining slurry was subsequently centrifuged at 10,000 rpm (Neofuge 23R) for 30 min and the resulting supernatant was discarded. The product was subjected to NaCl wash or Na+ pre-exchange following the procedures reported by Atkins et al. [15]. The centrifuged paste was mixed with 1 M NaCl, shaken for 1 h and centrifuged, the supernatant was discarded. This process was repeated 5 times. For the last 1 M NaCl wash the pH was adjusted to pH 8 via the drop-wise addition of 1 M NaOH and the suspension was shaken overnight. After centrifuging, the resulting paste was combined with DDW, shaken for 1 h and centrifuged. This DDW wash cycle was repeated 10 times. Following the final wash, the suspension was dialyzed for 3 days in 43 × 27 mm cellulose dialysis tubing. This sample is named 0.52-Na-H-Bir in this paper (0.52 was the ratio of Mn(VII)/Mn(II)).</p><p>The synthesis of 0.42-Na-H-Bir and 0.67-Na-H-Bir was similar to that of 0.52-Na-H-Bir, but the amounts added of MnCl2·4H2O were 30 g (0.4737 M) and 18.785 g (0.2966 M), respectively (the ratio of Mn(VII)/Mn(II) was 0.42 and 0.67). The synthesis of 0.52-H-Bir did not include the Na+ pre-exchange. The synthesis of 0.52-K-H-Bir was also similar to that of 0.52-Na-H-Bir, but used KOH solution as the material for synthesis and pH adjustment. The NaCl wash or Na+ pre-exchange was replaced by KCl wash or K+ pre-exchange.</p><!><p>Transformation of "c-disordered" H+-birnessite into todorokite was following a method adapted from Feng et al. [11, 20]. Approximately 20 g of wet "c-disordered" H+-birnessite slurry was dispersed in 1.5 L of 1 mol/L MgCl2 solution. After being fully exchanged for 24 h, the birnessites were converted into buserites (Bus). Birnessite and buserite are all the layer structure Mn oxides (phyllomanganates), and structurally related to each other. The former exhibits a 0.7 nm basal plane spacing due to the presence of single layer of water molecules and interlayer cations between the layers. When hydrated in the suspension or exchanged with larger hydrated cations (for instance, Mg2+, Ca2+), birnessite can be transformed into buserite with expanded basal plane spacing of 1.0 nm due to double layers of water molecules in the interlayers. In turn, upon drying or heating, buserite can be dehydrated and converted back into birnessite due to loss of one layer of water molecules. Then buserite solids were collected by centrifugation and re-suspended in 400 mL 1 mol/L MgCl2 solution and transferred to a 1 L triangular flask connected with a condensation device, then heated to reflux at 100°C and maintained at reflux conditions under magnetic stirring. Suspension aliquots were sampled at time intervals of 3, 6, 9, 12, 24, 48, 72 h, and 1 month. The produced minerals were washed until the conductance of the supernatant was below 20 μS/cm and freeze-dried.</p><!><p>XRD data were collected using Ni-filtered Cu Kα radiation (λ = 0.15418 nm) on a Bruker D8 Advance diffractometer equipped with a LynxEye detector. The diffractometer was operated at a tube voltage of 40 kV and a current of 40 mA with a scanning rate of 10°/min at a step size of 0.02°. Both buserite and todorokite have a basal d-spacing of ~1 nm. The former is not stable and can be transformed into the 0.7 nm phase (birnessite) during heating or dehydrating, whereas the latter has relatively high thermal stability [21]. To eliminate the interference of diffraction peaks of buserite on the identification of todorokite, the oriented samples spread on glass slices for the refluxed products were heated for 10 h at 140°C before XRD analysis [11].</p><p>The chemical composition of the samples was determined as follows: 0.1000 g of sample was dissolved in 25 mL of 0.25 mol/L NH2OH·HCl. The contents of Na and K were analyzed by a Varian Vista-MPX ICP-OES and the contents of Mn and Mg were measured using atomic absorption spectrophotometer (AAS). The average oxidation state (AOS) of Mn was determined using the oxalic acid-permanganate back-titration method.</p><p>Transmission electron microscopy (TEM) and high-resolution transmission electron microscopy (HRTEM) analyses of the crystal particles were conducted using a JEM-2100F electron microscope (JEOL, Japan) at an accelerating voltage of 200 kV. The samples were dispersed into absolute alcohol and ultrasonically vibrated prior to deposition on a holey carbon film, and then air-dried at room temperature.</p><p>The IR spectra of samples were recorded on a Bruker Vertex 70 FTIR spectrometer by making pellets with KBr. Samples were scanned 128 times between 4,000 and 400 cm−1 at a resolution of 4 cm−1.</p><!><p>XRD patterns of the synthetic "c-disordered" H+-birnessites with different Mn(VII)/Mn(II) ratios pre-exchanged with Na+ or K+ or not.</p><p>Chemical analyses and average manganese oxidation states of the synthetic "c-disordered" H+-birnessites</p><p>HR-TEM images of "c-disordered" Na-H-birnessites with different Mn(VII)/Mn(II) ratios. a, b 0.42-Na-H-Bir; c, d 0.52-Na-H-Bir; e, f 0.67-Na-H-Bir.</p><p>FTIR spectra of "c-disordered" Na-H-birnessites with different Mn(VII)/Mn(II) ratios (a) and their products after Mg-exchange and reflux treatment for 72 h (b).</p><p>XRD patterns of Mg2+-exchanged intermediates and reflux products for the synthetic "c-disordered" H+-birnessites at different times. a 0.42-Na-H-Bir, b 0.52-Na-H-Bir, c 0.67-Na-H-Bir, d 0.52-K-H-Bir, e 0.52-H-Bir. Tod todorokite, Bir birnessite, Man manganite, Ram ramsdellite.</p><p>Elemental compositions of the synthetic "c-disordered" H+-birnessites after exchanged with Mg2+ for 24 h</p><p>* Equivalent charge of metal ions per Mn.</p><!><p>Over the course of the reflux treatment of 0.42-Na-H-Bus, the characteristic diffraction peaks of todorokite appear at 3 h in the XRD pattern (Figure 4a). With reflux time prolonged to 48 h, the peaks of birnessite disappear leaving only strong todorokite XRD reflections, indicating that the 0.42-Na-H-Bus has been completely converted to todorokite. With a prolonged reflux time of 1 month, a very weak diffraction peak of manganite at 0.34 nm appears besides slightly strengthened those of todorokite (Figure 4a). For the 0.52-Na-H-Bus sample under reflux treatment, broad todorokite peaks and a very weak birnessite peak are present in the XRD patterns at reflux times of 48–72 h. As the reflux time is extended to 1 month, the birnessite peaks disappear and the characteristic peaks of todorokite become much stronger, while peaks of low-valence manganite (γ-MnOOH, JPCDS 8–99) are observed (Figure 4b). When the Mn(VII)/Mn(II) ratio increases to 0.67, only the characteristic peaks of birnessite are observed at a reflux time of 72 h, indicating that todorokite is not formed. At a reflux time of 1 month, the weak characteristic peaks of todorokite are visible with similar intensity to those of birnessite, suggesting that part of the 0.67-Na-H-Bus has been converted to todorokite (Figure 4c).</p><p>Furthermore, it is also observed that 0.52-K-H-Bus and 0.52-H-Bus after reflux treatment are partially transformed into todorokite but to a lesser extent compared to 0.52-Na-H-Bus, the corresponding Na+ pre-exchanged precursor. The intensities of the characteristic peaks of todorokite and birnessite in the XRD pattern of 0.52-K-H-Bus after reflux for 72 h (Figure 4d) are close to those of 0.52-Na-H-Bus refluxed for 48 h (Figure 4b). This suggests that 0.52-K-H-Bus is less favorable for the transformation than 0.52-Na-H-Bus. When 0.52-H-Bus (without Na+ or K+ pre-exchange treatment) is refluxed for 3 h, the characteristic peaks of ramsdellite (1 × 2 tunnel-structured γ-MnO2, JPCDS 44-0142) appear, while very weak XRD reflections of todorokite can be discerned after 12 h. With increasing reflux time, the intensities of the peaks of todorokite and those of ramsdellite remain nearly unchanged (Figure 4e).</p><!><p>HR-TEM images of "c-disordered" Na-H-birnessites with different Mn(VII)/Mn(II) ratios after Mg-exchange and reflux treatment for 72 h. a, b 0.42-Na-H-Bir; c, d 0.52-Na-H-Bir; e, f 0.67-Na-H-Bir.</p><!><p>FTIR spectroscopy can conclusively distinguish between layer-type birnessite and tunnel-type todorokite [15, 34], because the broad peak at ~760 cm−1 in FTIR of Mn oxides is typically assigned to an asymmetrical Mn–O stretching vibration, corresponding to corner-sharing MnO6 octahedra, which is absent in phyllomanganate Mn oxides [34]. Figure 3b shows FTIR spectra of "c-disordered" Na-H-birnessites after Mg-exchange and reflux treatment for 72 h. Besides the common bands of Mn oxides at 438, 517, and 558 cm−1, the characteristic peak of todorokite at ~762 cm−1 can be observed with decreasing intensities from 0.42-Na-H-Bus 72 h reflux to 0.52-Na-H-Bus 72 h reflux to 0.67-Na-H-Bus 72 h reflux (Figure 3b). This result suggests that these "c-disordered" Na-H-birnessites are converted into todorokite to different extents after reflux treatment, and low Mn AOS is favorable to the transformation. This conclusion agrees very well with the XRD and TEM analyses (Figures 4, 5).</p><!><p>It is believed that a small amount of cations such as Ca2+, Mg2+, Ni2+, Li+, Na+, K+, NH4+, or H3O+ is required to stabilize the tunnels in the formation of tectomanganates [35–37]. The order of hydrated radii of interlayer ions is K+ (3.31 Å) < Na+ (3.58 Å) < Mg2+ (4.28 Å). After the "c-disordered" H+-birnessite samples had been exchanged with Mg2+ for 24 h, most of the interlayer Na+ was replaced by Mg2+ (Table 2). Only part of the interlayer K+ was replaced by Mg2+ in the 0.52-K-H-Bus sample, probably caused by stronger electrostatic interaction between K+ with smaller hydrated radius and negative MnO6 octahedral layers compared with Na+ and Mg2+. The transformation from 0.52-K-H-Bus to todorokite is slower than that from 0.52-Na-H-Bus, which can be attributed to the fact that the interlayer K+ of 0.52-K-H-Bus cannot be completely exchanged with Mg2+, and the remaining 12.6% of interlayer K+ inhibits transformation of buserite to todorokite.</p><p>Under reflux treatment at 100°C, 0.52-H-Bus is transformed into ramsdellite, while Na/K-H-buserites are transformed into todorokite. This significant difference may be ascribed to different contents of interlayer ions in their precursors. The levels of Mg2+ and K+ in the interlayer of 0.52-H-Bus are 0.110 and 0.012 in molar ratios of Mg/Mn and K/Mn, or 0.233 equivalent charge of cations per Mn atom (Table 2). These values are much lower than the levels in 0.52-Na-H-Bus (0.273 and 0.005, or 0.554) and 0.52-K-H-Bus (0.234 and 0.033, or 0.501). Given the close MnO6 layer structures of 0.52-Na-H-Bus, 0.52-K-H-Bus, and 0.52-H-Bus, it can be inferred that a substantial part of the negative layer charge of 0.52-H-Bus is also electrostatically balanced by H+, either structurally bound or free in interlayers, apart from Mg2+ and K+ ions in the interlayers. Because of the smaller hydrated radius of H+ (2.82 Å) compared to K+, Na+, and Mg2+, the interaction between H+ and birnessite layers is supposed to be the greatest, which can account for the lower Mg2+ content of 0.52-H-Bus in comparison with 0.52-K-H-Bus and 0.52-Na-H-Bus (Table 2). Ramsdellite has a 1 × 2 tunnel structure, which is constructed of double chains of octahedra linked with single octahedral units by sharing corner oxygen atoms to form a rectangular cross-section [9]. Waychunas [38] reported that the tunnels of ramsdellite are normally empty but trace amounts of water, Na, and Ca might be present in the tunnels. Todorokite has a 3 × 3 tunnel structure with corner-sharing triple chains of MnO6 octahedra that are saturated with different larger cations such as Mg2+, Ca2+, and Ni2+ [3]. The large tunnels (6.9 × 6.9 Å) make todorokite materials possess a high ion-exchange capacity [4]. Therefore, lower contents of interlayer metal ions can account for the transformation of "c-disordered" H+-birnessite to ramsdellite with a smaller tunnel size of 1 × 2 rather than todorokite.</p><!><p>The results show that from 0.42-Na-H-Bus to 0.52-Na-H-Bus to 0.67-Na-H-Bus, increasing reflux treatment time is required for both todorokite appearance and complete formation (Figure 4). In other words, "c-disordered" H+-birnessites with small Mn AOS values are readily transformed into todorokite, which is consistent with triclinic birnessite transformation to todorokite [8]. When Mg2+ exchanged "c-disordered" H+-birnessite is refluxed at 100°C, the Mn(III) species may easily move from the layers of buserite to the interlayer, leaving more vacancies in the buserite framework. Subsequently more interlayer Mn(III) species can gradually bond through covalence via condensation and dehydration reactions with each other and structural rearrangement to construct the "walls" of the tunnels. At this time, if the content and type of interlayer cations are sufficient to stabilize the 1-nm interlayer spacing, todorokite with 1-nm tunnel size will gradually form. Otherwise if the interlayer cations possess small hydrated radii, such as Na+, K+, or are present in insufficient amounts, such as in the case of 0.52-H-Bus, the 1-nm interlayer spacing will collapse and the shorter tunnel "walls" may be built instead. This leads to formation of tectomanganates with small size tunnels, for instance, cryptomelane with 2 × 2 size [39], ramsdellite with 1 × 2 size, or even pyrolusite with 1 × 1 size [40].</p><p>In addition, a sufficient amount of large interlayer cations does not necessarily facilitate the formation of todorokite if the Mn AOS of the precursor birnessite is high enough or there is insufficient Mn(III) species. After Mg2+-exchange for 24 h, the Mg2+ content of Na-H-Bus increases with increasing Mn(VII)/Mn(II) ratio in the synthesis of "c-disordered" H-birnessite (Table 2). This is probably because the Na-H-birnessite with large Mn AOS contains more vacancy sites [41, 42], and so more negative layer charge will develop and more Mg2+ is needed to reach charge equilibrium. However, the extent and rate of transformation into todorokite sharply decreases with increasing Mg content from 0.42-Na-H-Bus to 0.52-Na-H-Bus to 0.67-Na-H-Bus (Figure 4; Table 2).</p><p>It should be noted that "c-disordered" H+-birnessite is more difficult to convert to todorokite compared to triclinic birnessite with similar Mn AOS value under the same reflux conditions. For instance, triclinic birnessite with Mn AOS of 3.58 (0.33-Na-bir) can be almost completely converted into todorokite within 24 h [8], while it will take 48 h for complete conversion of 0.42-Na-H-Bir with Mn AOS of 3.55. One explanation for this could be that greater structural adjustment may be required for "c-disordered" H+-birnessite, a type of hexagonal birnessite, to be converted into todorokite relative to triclinic birnessite. Our other experiments showed that hexagonal birnessites, either δ-MnO2 or acid birnessite, with high Mn AOS of 3.85 are barely converted into todorokite after Mg2+ exchange and reflux treatment (data not shown). After reacting with low-concentration aqueous Mn(II), these hexagonal birnessites can gradually transform into birnessites with a orthogonal layer symmetry, and the produced birnessites can be completely converted into todorokite after 24 h of reflux treatment. This result confirms that triclinic birnessite with orthogonal layer is more likely than hexagonal birnessite to be converted into todorokite. Thus, triclinic birnessite had been exclusively used to prepare todorokite in the laboratory in the previous literature until recently Atkins et al. [15] reported successful synthesis of todorokite using "c-disordered" H+-birnessite as the precursor.</p><p>Ideal hexagonal birnessite hardly contains Mn(III), i.e. it has high Mn AOS but a large number of vacancy sites (16.7%), contrarily ideal triclinic birnessite barely possesses vacancy sites but does have a large amount of Mn(III) (33.3%) [17, 23, 43]. In this study, with decreasing Mn AOS the layer structure of "c-disordered" birnessite may adjust gradually towards that of triclinic birnessite, in other words, "c-disordered" birnessite with low Mn AOS may modify its structure, at least partially, towards triclinic birnessite through layer symmetry adjustment. These will definitely facilitate their transformation to todorokite. Further study is needed to explore the pathway of todorokite formation and whether triclinic birnessite is the necessary intermediate during the transformation of hexagonal birnessite into todorokite.</p><p>Therefore, Mn(III) together with content and type of interlayer metal ions plays a crucial role in the transformation of "c-disordered" H+-birnessite with hexagonal symmetry to todorokite, which is also the case for the triclinic birnessite transformation [8]. These results provide a further explanation for the common occurrence of todorokite in the hydrothermal ocean environment [32, 44], where is usually enriched in large metal ions such as Mg, Ca, Ni, Co and etc., and plenty of Mn(II) released from hydrothermal activity. This Mn(II) can easily cause Mn(III) formation in birnessite, either with orthogonal or hexagonal layer symmetry.</p><!><p>The transformation of a layered birnessite to a tunnel-structure todorokite is a complex process. From the above results it is evident that the phase transformation of layer-structure "c-disordered" H+-birnessite to tunnel-structure todorokite is greatly affected by the amount of Mn(III) and the type and content of interlayer cations in the birnessite. Under reflux conditions, the degree of transformation of "c-disordered" H+-birnessite to todorokite decreases significantly as the Mn AOS values of Na-H-birnessite increase from 3.58 to 3.74. The transformation from K-H-birnessite to todorokite is slower than that from Na-H-birnessite, because the interlayer K+ of birnessite cannot be completely exchanged with Mg2+, and so interlayer K+ inhibits transformation of birnessite to todorokite. "c-disordered" H+-birnessite contained lower levels of Na/K is transformed into ramsdellite rather than todorokite.</p>

PubMed Open Access

AN ANGIOTENSIN II TYPE 1 RECEPTOR ACTIVATION SWITCH PATCH REVEALED THROUGH EVOLUTIONARY TRACE ANALYSIS

Seven transmembrane (7TM) or G protein-coupled receptors constitute a large superfamily of cell surface receptors sharing a structural motif of seven transmembrane spanning alpha helices. Their activation mechanism most likely involves concerted movements of the transmembrane helices, but remains to be completely resolved. Evolutionary Trace (ET) analysis is a computational method, which identifies clusters of functionally important residues by integrating information on evolutionary important residue variations with receptor structure. Combined with known mutational data, ET predicted a patch of residues in the cytoplasmic parts of TM2, TM3, and TM6 to form an activation switch that is common to all family A 7TM receptors. We tested this hypothesis in the rat Angiotensin II (Ang II) type 1 (AT1) receptor. The receptor has important roles in the cardiovascular system, but has also frequently been applied as a model for 7TM receptor activation and signaling. Six mutations: F66A, L67R, L70R, L119R, D125A, and I245F were targeted to the putative switch and assayed for changes in activation state by their ligand binding, signaling, and trafficking properties. All but one receptor mutant (that was not expressed well) displayed phenotypes associated with changed activation state, such as increased agonist affinity or basal activity, promiscuous activation, or constitutive internalization highlighting the importance of testing different signaling pathways. We conclude that this evolutionary important patch mediates interactions important for maintaining the inactive state. More broadly, these observations in the AT1 receptor are consistent with computational predictions of a generic role for this patch in 7TM receptor activation.

an_angiotensin_ii_type_1_receptor_activation_switch_patch_revealed_through_evolutionary_trace_analys

1. INTRODUCTION<!>2.1 ET analysis<!>2.2 Ligands<!>2.3 Recombinant DNA plasmids<!>2.4 Cell culture<!>2.5 Whole-cell competitive radioligand binding assay<!>2.6 Inositol phosphate accumulation<!>2.7 ERK1/2 phosphorylation by Western blotting<!>2.8 Cell surface/whole cell ELISA<!>2.9 Immunocytochemistry and confocal microscopy<!>2.10 Receptor Selection and Amplification Technology (R-SAT\xc2\xae)<!>3.1 ET analysis<!>3.2 Agonist binding properties of receptor mutants<!>3.3.1 G protein activation<!>3.3.2 ERK1/2 activation<!>3.4 Receptor surface expression, localization, trafficking, and \xce\xb2-arrestin2 co-localization<!>3.5 R-SAT\xc2\xae assay<!>4. DISCUSSION<!>The predicted activation switch patch in a helical model of the rat AT1a receptor<!>Effect of receptor point mutations on Ang II affinity tested by competitive ligand binding<!>Signaling properties of mutant receptors determined by inositol phosphate accumulation and ERK1/2 phosphorylation<!>Cell surface expression and trafficking properties of mutant receptors<!>R-SAT\xc2\xae assay basal activity, normalized to WT<!>Residues of the rat AT1a receptor activation switch patch identified by ET Analysis<!>pKi values for Ang II competition binding<!><!>Ang II dose response curves, carried out in R-SAT\xc2\xae assay<!>

<p>7TM receptors constitute a large family of cell surface receptors all characterized by a conserved motif of seven transmembrane spanning alpha helices. The receptors are involved in numerous processes in the human body with important roles in many physiological processes [1].</p><p>The molecular mechanisms underlying 7TM receptor activation have not been completely resolved, but they most likely involve similar or even conserved concerted rearrangements of the helical bundle with an outward movement of TM6. This is summarized in the global toggle switch activation model [2]. One of the frequently applied methods to identify molecular interactions involved in receptor activation is the study of constitutively activated receptor mutants (CAMs). The elevated basal activity of these mutants is attributed to removal of restraints normally maintaining the inactive state of the receptor [1].</p><p>Thus, an extensive amount of information currently exists regarding receptor regions involved in conversion between inactive and active states now exists. Some of these regions have been verified by mutational studies in several family A 7TM receptors and an extensive amount of mutational data is available. Evolutionary Trace (ET) analysis is a computational method that identifies functional sites in protein families by looking at the evolutionary conservation of the residue positions in homologues sequences [3]. The analysis ranks all the residue positions in a protein family according to how well their side chain variations correlate with major evolutionary divergences of the protein family [3]. Top-ranked residues have important proteome-wide biological properties: they cluster structurally [4]; and these clusters predict functional sites [5]. Thus ET may be used to identify functional sites [6] and to guide mutations that selectively alter these sites and hereby separate singular functions [7] or to exchange functions [8, 9]. Specifically, for family A 7TM receptors this analysis has led to the prediction of a universally conserved activation mechanism consistent with mutations in rhodopsin [3, 10] and in the β-adrenergic receptor [11].</p><p>The Angiotensin II (Ang II) type 1 (AT1) receptor is the primary effector of the peptide hormone Ang II. The receptor is coupled to Gq/11, leading to inositol phosphate pathway induction, which in turn triggers an increase in intracellular calcium and protein kinase C activation [12]. However, G protein independent, β-arrestin dependent signaling also has been demonstrated for the receptor [13]. The receptor has frequently been applied as a model receptor for 7TM receptor activation. Previous studies on the molecular determinants of AT1 receptor activation support the general model for 7TM receptor activation with special attention to interactions between and movements of TM2, TM3, TM6, and TM7 (for review: Petrel and Clauser 2009 [14]). Studies have shown that constitutive AT1 receptor activity sometimes reveals itself as constitutive internalization of the receptor or promiscuous activation [15–17]. Combined with the reports of G protein-independent signaling from this receptor, this suggests that testing of more than one signaling pathway is advisable when studying AT1 receptor activation.</p><p>In this study, we combined Evolutionary Trace analysis with mutational data from the GPCR database to predict a patch of six amino acids in the cytoplasmic core of the rat AT1a receptor involved in conversion between inactive and active state. These residues were mutated individually and mutant receptors were characterized for expression, ligand binding and G protein-dependent as well as G protein-independent signaling outcomes. Five out of six mutant receptors showed signs of a changed activation state in one or more assays hereby supporting a role for the patch in AT1 receptor activation. We find that our analysis supports the general ET model of 7TM receptor function thus displaying the strength of combining an unbiased computational approach with mutational history to guide mutational studies in 7TM receptors and that it highlights the importance of testing multiple signaling endpoints to ensure accurate characterization of mutation phenotypes.</p><!><p>To identify key residues that may form part of an activation switch in the AT1 receptor (rat AT1a), we followed a computational approach first described by Madabushi et al.[10], which is based on the ET method. The underlying hypothesis is that residues predicted to be important among a large group of 7TM receptors most likely take part in functions common to the entire group of 7TM receptors rather than functions specific to their ligands. A comparison with the large body of mutational data described in the literature can further narrow down this set of key functional residues to some more closely associated with ligand sensitivity, conformational switching, or G protein-coupling.</p><p>In this study, we have focused on residues involved in conformational switching from the inactive to the active state of the receptor. Thus, we compared the ET data to mutational data where mutations caused constitutive activity. This led to the identification of six amino acids in TM2, TM3 and TM6. To test the role of this patch in the conformational switch mechanism, we then predicted disruptive mutations most likely to induce constitutive activity.</p><p>Specifically, a BLAST [18] search retrieved 129 visual opsins, 69 bioamine, 58 olfactory, and 82 chemokine family A receptor sequences from NCBI data base (See supplementary material table 1). An alignment of each receptor family was generated by ClustalW [19]. 7TM receptors are most conserved in the transmembrane helices whereas the loops are highly divergent. Therefore the alignments were performed on seven transmembrane helix residues only. Specifically, the analyzed residues of the rat AT1a sequence are: 27–55 (TM1), 63–88 (TM2), 99–132 (TM3), 145–167 (TM4), 194–222 (TM5), 238–265 (TM6), and 282–307 (TM7). We then performed an Evolutionary Trace on the concatenated gapless TM segments. This produced a relative ranking of evolutionary importance of the transmembrane sequence residues.</p><p>To further narrow down our search to residues most likely to have a direct role in the conformational switch mechanism linked to activation, we picked a subset of residues implicated in constitutive activity or misfolding from the top 20th percentile of importance residues. These choices were either based on mutational data (sixteen residues) or on structural proximity (two others). We then selected six top-ranked residues from this group (15th percentile or better) by ET located in TM2, TM3 and TM6 and with a clustering z-score of 2.8 in the rhodopsin structure 1F88, details are listed in Table 1 and below. Four of these, L67, L119, D125, I245 in the rat AT1a receptor (Ballesteros # in order is 2.43, 3.43, 3.49, 6.40), had cognate residues mutations causing constitutive activity in luteinizing hormone receptor, muscarinic acetylcholine, rhodopsin, and anaphylatoxin chemotactic receptors [20–23]. One residue L70 (2.46) was a novel prediction that was found to cause constitutive activity in rhodopsin [10], and F66 (2.42) was not yet tested for constitutive activity to our knowledge.</p><p>To pick substitutions likely to induce constitutive activity for each of these residues, we followed several specific criteria. First, we favored substitution that were already implicated in constitutive activation in the literature; second, we rejected substitutions if these naturally occurred in the family of Ang II receptor sequences; and third, we favored substitutions that were thought to be more disruptive substitutions by traditional criterion (i.e. positive charge to negative charge residue – although we recognize that the meaning of disruptive is statistical rather and may not stringently apply at any one specific site). For example, the L67R mutation switches a non polar sequence position into a positively charged. It also mimics two similar Arg susbstitutions at the cognate position that led to constitutive activity in Glucagon and Vasoactive Intestinal peptide 1 receptors [24, 25]. Finally, no Ang II receptor sequences have an Arg at that position. Likewise, for other mutations such as L119R, D125A and I245F: L119 (3.43) is well reported to be a conformational switch residue; D125 (3.49) is part of the TM3 DRY motif frequently reported to have mixed effects on conformational activation, mechanistic effects and G protein uncoupling; and I245 (6.40) has been reported to induce constitutive activation of receptor in rhodopsin and anaphylatoxin chemotactic receptors [20, 22]. We also note that L70R mutation partially goes along with our previous study [10] showing that a mutation to Ala at this residue position causes constitutive activity in rhodopsin. And finally, F66A is a novel change with no previous mutational data on constitutive activity from the literature, but the Phe to Ala mutation is a highly disruptive change. A detailed evolutionary information and mutation recommendation of above residues is listed in Table 1.</p><!><p>Ang II and Sar1- Thr8 (ST) Ang II were from Sigma-Aldrich (St. Louis, MO). Sar1 - Ile4 - Ile8 (SII) Ang II was synthesized at the Cleveland Clinic, Lerner Research Institute (Cleveland, OH, USA). Telmisartan was from Boehringer Ingelheim (Ingelheim am Rhein, Germany)</p><!><p>Mutations were generated using the Quickchange mutagenesis protocol (Stratagene, La Jolla, CA) with the rat AT1a or Myc-tagged human AT1 receptor as template. Constructs were subcloned into a modified pCDNA3.1vector containing the M1 Flag-tag inserted after a hemagglutinin signal peptide (rat AT1a constructs; [26]) or into the pSI vector (human AT1 constructs, see Hansen et al, 2004 [27]). Mutations were verified by sequencing at Eurofins MWG Operon (Ebersberg, Germany). The hAT1 N111A construct was described in Hansen et al, 2008 [28]. GFP tagged β-arrestin2 was a gift from Dr. Marc Caron [29].</p><!><p>Human embryonic kidney (HEK) 293 cells (American Type Culture Collection, Manassas, VA, USA) were grown in Dulbecco's modified Eagle's medium (DMEM) with 4.5 g/L glucose and L-Glutamine (Invitrogen, Carlsbad, CA) or supplemented with 0.03% L-Glutamine (Substrate Department at Panum Institute, Copenhagen, Denmark), both supplemented with 10% fetal bovine serum (HyClone, Logan, UT or BioChrome AG, Berlin Germany). N-terminally M1 FLAG-tagged constructs were stably expressed in HEK293 cells. For generation of clonal stable cell lines, single colonies were selected and propagated in the presence of selection-containing media (Zeocin, 0.2ug/mL (Invitrogen, Carlsbad, CA)).</p><!><p>Binding assay was conducted as described in Hansen et al., 2004 [27] with minor modifications. In brief, HEK293 cells stably expressing AT1 receptor constructs were seeded in poly-L-lysine (Sigma-Aldrich, St. Louis, MO) coated 48-well dishes at 180 000 cells/well. On the following day, cells were kept at 4°C for 30–60 mins., washed once in cold Hanks Balanced Salt Solution (HBSS) with 20 mM Hepes supplemented with 0.9 mM CaCl2 and 1.05 mM MgCl2 (HBSS+). This was followed by 3h of incubation at 4°C with 0.2 mL of HBSS+ containing radioligand ~ 4.5 pM of 125I - Sar1- Ile8 (SI) Ang II (Perkin Elmer, Waltham, MA) and increasing amounts of unlabeled Ang II (triplicate measures for each point). Cells were washed twice in ice cold HBSS+ and lysed in 0.5 mL of lysis buffer (1% TritonX, 50 mM Tris-HCl pH 7.5, 100 mM NaCl, and 5mM EDTA) at room temperature (RT) for 30–60 mins. Well content was transfered to scintillation vials containing 4mL Ultima Gold scintillation liquid (Perkin Elmer, Waltham, MA). Total radioactivity was measured in a TriCarb 2800 TR Liquid Scintillation analyzer (Perkin Elmer, Waltham, MA). Results were analyzed in GraphPad Prism and Excel. Curves were fitted using the non-linear regression analysis (one-site competition) in GraphPad Prism. Kd values for SI Ang II were identified by homologous competition binding using unlabeled SI Ang II as competitor (data not shown), Kd = IC50-[radioligand]. Ki values for Ang II were then calculated using the formula Ki=IC50/[1+([radioligand]/Kd)]. Statistical significance was evaluated by Student's t-test (two tailed, paired) on pKi values in Excel.</p><!><p>Cells were seeded in poly-L-lysine coated 96 well dishes at 100 000/well in 100 μL DMEM without I-inositol with Glutamax (Invitrogen, Carlsbad, CA) supplemented with 10% fetal bovine serum and 2 μCi H3myo inositol (GE Healthcare, Buckinghamshire, UK) pr. mL medium. On the following day, cells were preincubated for 25 min at 37°C with either inverse agonist or vehicle (DMSO). After incubation, a wash was performed in Hanks Balanced Salt Solution (HBSS) with) and 20mM Hepes supplemented with 0.9 mM CaCl2, 1.05 mM MgCl2, 10 mM LiCl (HBSS++ vehicle and thereafter incubated 20 mins at 37°C with HBSS++ and inverse agonist for Telmisartan condition to ensure steady state interaction. This was followed by ligand stimulation for 45 mins at 37°C. Hereafter, HBSS++ was removed, and cells were incubated at 4°C with 20 mM formic acid for about 45 mins. 20 μL of the supernatant was transferred to a white 96 well plate (Perkin Elmer, Waltham, MA) and 80 μL SPA beads (GE Healthcare, Buckinghamshire, UK, freshly diluted in MilliQ H20 to final concentration of 1mg/well) was added. Plate was sealed with "Top Seal" cover (Perkin Elmer, Waltham, MA) and incubated for 30 min at constant shaking. Hereafter, SPA beads settled at RT for at least 8h before counted on Topcounter. Data was analyzed in GraphPad Prism and Excel. Statistical analysis was performed in Excel (unpaired Student's t-test, two-tailed) on un-normalized means.</p><!><p>Stable cell lines were seeded in poly-L-lysine coated 6 well dishes at 400 000/well and grown to confluency. Cells were then serum starved for 4h, stimulated with ligands, washed twice in ice cold PBS and frozen at −20°C. Lysis, SDS-PAGE, and immunoblotting were conducted as described in Theilade et al. with minor modifications [30]. Antibodies were diluted in Tris Buffered Saline with 0.2% Tween20, 0.1% sodium azide was added to primary antibody dilutions. For total ERK1/2 protein: anti p42/44 MAPK antibody (Cell Signaling, Danvers, MA, #9107) 1:2000; HRP conjugated anti-mouse antibody (GE Healthcare, Buckinghamshire, UK) 1:5000. For phosphorylated ERK1/2 protein: anti phospho-p42/44 MAPK antibody (Cell Signaling, Danvers, MA, #9101) 1:1000, HRP conjugated anti-rabbit antibody (GE Healthcare, Buckinghamshire, UK) 1:10 000. Primary antibody incubations were performed either overnight at 4°C or at 1h at RT. Secondary antibody incubation was performed at 1h at RT. Blots were developed using an enhanced chemiluminescence system (GE Healthcare, Buckinghamshire, UK) and visualized at a Fluor Chem HD2 reader (Alpha Innotech, San Leandro, CA). Blots were quantified with densitometric gel analysis in the program Image Gauge. Results were analyzed in Excel and Graphpad Prism. Representative images were rendered in the Irfanview program.</p><!><p>Cells were seeded for triplicate measures with and without primary antibody in poly-L-lysine coated 96 well dishes at 35 000/well in 100 μL DMEM supplemented with L-glutamine and serum. The following day, cells were preincubated for 90 mins at 37°C with either Telmisartan or vehicle (DMSO). In this assay, we used a concentration of 10μM of the inverse agonist. This was done, because higher concentrations are often needed to detect changes in secondary parameters such as beta-arrestin recruitment and internalization[31, 32]. Cells were subsequently placed on ice. For surface detection of the M1 Flag tag, cells were incubated with cold DMEM and anti M1 Antibody (Sigma-Aldrich, (St. Louis, MO); 1:2000) for 1h at 4°C (cells without primary antibody were control incubated in cold DMEM). Cells were washed first in cold DMEM, then in PBS, and fixed with 4% paraformaldehyde in PBS for 10 mins. Cells were washed twice in PBS and blocked for 30 mins in PBS with 1% Bovine Serum Albumin (blocking buffer). Cells were incubated with secondary antibody (HRP-conjugated anti-mouse IgG, GE Healtcare) diluted 1:3000 in blocking buffer. Cells were washed in PBS and developed using 3,3′,5,5′-tetramethylbenzidine (TMB) liquid substrate system for ELISA (Sigma-Aldrich, St. Louis, MO). Reactions were stopped with 1M hydrochloric acid after development of a blue color and absorbance at 450 nm was measured. For detection of total protein, cells were washed in PBS and fixed as described above. Blocking, primary, and secondary antibody conditions were conducted in blocking buffer with 0.1% Triton X-100 for permeabilization. Results were analyzed in GraphPad Prism and Excel. Triplicate means were normalized firstly by subtraction of values for cells without antibody, and secondly, by substraction of values for empty HEK293 cells.</p><!><p>Cells stably expressing M1 FLAG-tagged rat AT1 WT and mutant receptors were seeded in 6 cm dishes and grown to 80 – 90 % confluency. Cells were transfected with 3 μg GFP-tagged β-arrestin2 using Fugene 6 (Roche, Basel, Switzerland) according to manufacturer's protocol. The day after transfection, cells were plated on poly-D-lysine (Sigma-Aldrich, St. Louis, MO) coated coverslips and grown to about 50% confluency. On the assay day, media was changed on cells and 10 μM Telmisartan was added to "Telmisartan" well for 30 mins. Cells were then incubated with M1 antibody (IgG2b, Sigma-Aldrich, St. Louis, MO) directed against the FLAG tag (1:1000, 30 mins) and hereafter treated with 1 μM Ang II or vehicle (PBS) for 30 mins. Cells were fixed with 4% formaldehyde in PBS for 20 mins, then permeabilized in Blotto (50 mM Tris-Cl pH 7.5, incubated with Alexa 594 conjugated 1mM CaCl2, 3% milk, 0.1% Triton X-100) for 20 mins and (red) antibody (1:500, Invitrogen, Carlsbad, CA) for 45 mins. Coverslips were mounted antiIgG2b using mounting media from Vectashield (Vector laboratories, Burlingame, CA) containing 4′,6′-diamidino-2-phenylidole (DAPI). Hence receptor (red), β-arrestin2 (green) and cell nuclei (blue) could be individually visualized. Mounted slides were analyzed on a 510 LSM laser confocal microscope (Zeiss, Thornwood, NY) using the 63x oil objective.</p><!><p>The R-SAT® assay measures the ability of transiently expressed oncogenes, proto-oncogenes, and many 7TM receptors to confer partial or total transformation of NIH3T3 cells causing a loss of contact inhibition of these cells, hereby allowing them to proliferate when they would otherwise stop (Brauner-Osborne and Brann, 1996; Burstein et al., 2006). In R-SAT®, a reporter gene (in this case β-galactosidase) is co-transfected into the cells with the 7TM receptor of interest to quantify this proliferative response. The β-galactosidase reporter is constitutively expressed in this system and does not participate in driving the biological response but rather works as an indirect measure of proliferation [33, 34]. The R-SAT® assay was performed as described previously [27]. Briefly, NIH3T3 cells at 70–80% confluence were transfected with plasmids containing WT or mutant human AT1 receptor cDNA (25 ng of receptor and 20 ng of β-galactosidase reporter/well of a 96-well plate) using the PolyFect Reagent (QIAGEN, Valencia, CA) as described in the manufacturer's protocol. One day after transfection, ligands were added in DMEM supplemented with penicillin (100 U/ml), streptomycin (100 g/ml), and 25% Ultraculture synthetic supplement (Cambrex, Walkersville, MD) instead of calf serum to a final volume of 200 μL. After 5 days, the media was aspirated and cells were lysed, O-nitrophenyl-β-D-galactopyranoside was added, and the resulting absorbance was measured spectrophotometrically. Concentration response curves were performed in at least duplicates. Data was processed in GraphPad Prism and Excel. Statistical analysis was performed in Excel (unpaired Student's t-test, two-tailed).</p><!><p>To identify key residues involved in conformational switching from the inactive to the active state of the AT1 receptor (rat AT1a), we performed an ET analysis and compared this to mutational data where mutations caused constitutive activity (details are described in the methods section 2.1). This analysis led to the identification of six amino acids in TM2, TM3 and TM6. To test the role of this patch in the conformational switch mechanism, we predicted mutations most likely to induce constitutive activity (Figure 1, Table 1). To investigate the effect of the suggested mutations in the patch, we generated N-terminally FLAG tagged constructs of the WT and mutated AT1 receptors and generated HEK293 cell lines stably expressing each construct. These cell lines were characterized for ligand affinity and signaling properties. The cell line expressing L67R did not respond in ligand binding or signaling assays. This receptor shows very low cell surface expression and consequently is included only in select analyses (Figure 4 and in supplementary material Figure S1).</p><!><p>To assess the ligand binding properties of the AT1 receptor mutants relative to WT receptor, we performed a whole cell competitive ligand binding assay using the peptide antagonist 125I–SI Ang II as radioligand and unlabeled Ang II as competitor. The analysis showed an increased affinity for the agonist Ang II for all of the mutated receptors compared to WT receptor (Table 2; Figure 2). L119R shows a moderate increase in affinity about 3-fold compared to WT. L70R, D125A, and I245F show a 10–15 fold increase, while the F66A mutant shows a robust increase of about 70-fold compared to WT.</p><!><p>Gq/11-dependent signaling properties of WT and mutant receptors were analyzed assaying inositol phosphate accumulation in the HEK293 cell lines stably expressing the individual receptors. Partial agonists/neutral antagonists ST and SII Ang II were included to investigate if any of the mutated receptors showed promiscuous activation compared to WT receptor, as has been found for other AT1 CAMs [14]. Compounds were used at high doses to yield maximum responses (Ang II 0.1μM, SII Ang II 18.75 μM, ST Ang II 5 μM, Telmisartan 1 μM). The analysis revealed that receptor mutants L70R, L119R and D125A had moderate, yet significant increases in basal activity compared to WT (~2–3 fold, p<0.05) (Figure 3A). L119R and D125A mutants showed very robust activation by ST and SII Ang II. I245F had an activation profile very similar to WT. The Ang II response was similar to WT for mutants L119R, D125A and I245F (102 ± 8 %, 102 ± 14 %, and 79 ± 4 %, respectively) while reduced for L70R (56 ± 4 %) and very low for the F66A mutant (12 ± 1 %).</p><!><p>To characterize signaling of the AT1 receptor mutants through other pathways, ERK1/2 phosphorylation was assayed by immuno-blotting for total- and phospho-ERK1/2. To assess basal activity and Ang II induced ERK1/2 activation, HEK293 cell lines stably expressing receptors were stimulated with either 1μM Telmisartan for 20 mins or 0.1μM Ang II for 10 mins. Ang II induced responses in ERK1/2 phosphorylation assays (Figures 3B) were comparable to the inositol phosphate assay (Figure 3A). Based on densitometric quantification of the blots, the F66A and D125A showed a tendency to slight increases basal activity (basal values, percent of max WT: F66A: 13 ± 4, D125A: 16 ± 4 compared to WT: 8 ± 1).</p><!><p>Receptor signaling properties, especially the magnitude of the G protein-dependent signaling response, are usually positively correlated to receptor surface expression. Studies have shown that some AT1 CAMs are constitutively internalized. This constitutive internalization can be mediated by constitutive association to β-arrestin2 and inhibited by incubation with inverse agonists [14, 16, 35]. To compare the cell surface expression levels of WT and mutated receptors, we applied ELISA for quantitative evaluation and immunocytochemistry and confocal microscopy for qualitative evaluation of receptor localization and trafficking. In both assays, we also tested the effect of the inverse agonist Telmisartan, because previous studies have shown that incubation with an inverse agonist can increase cell surface expression of CAMs [16, 35].</p><p>In the ELISA assay, cell surface expression was determined by incubation with anti the M1 anti-FLAG antibody at 4°C followed by fixation and secondary antibody incubation, while the total amount of protein recognized by the M1 FLAG antibody in the cell was determined by permeabilizing cells prior to and during antibody incubation (Figure 4A). Noticeably, F66A and L67R mutants showed very low cell surface expression, but had total protein levels that were about 80% of WT receptor, suggesting that these mutant receptors were likely located intracellularly. The F66A mutant showed approximately 3 fold increased cell surface expression upon preincubation with 10 μM Telmisartan (Tel/noTel ratio 2.8 vs WT 1.1). Preincubation also yielded a moderate increase in I245F expression (Ratio 1.4). The L119R mutant was expressed at slightly higher levels than WT receptor (0.12 ± 0.02 vs. 0.086 ± 0.01 (WT)), D125A (0.073 ± 0.01) at similar levels as WT, while L70R and I245F showed expression levels at approximately half of WT (0.046 ± 0.01 and 0.047 ± 0.01, respectively).</p><p>For visual evaluation of receptor distribution and β-arrestin2 co-localization in basal conditions as well as in the presence of Ang II and Telmisartan, we performed an antibody feeding assay on cells stably expressing AT1 WT or mutant receptors that had been transiently transfected with GFP-tagged β-arrestin2 [29]. For this analysis, cells were incubated (fed) with the M1 anti-FLAG antibody recognizing the extracellular epitope tag while alive thereby labeling a population of receptors that have reached the cell surface. The distribution of these labeled receptors is then monitored in the presence or absence of exogenous ligand. As depicted in Figure 4B, WT receptor is expressed at the cell surface in the absence of ligand and does not colocalize with β-arrestin2. After 30 mins of 1 μM Ang II stimulation, the receptor co-localizes with β-arrestin2 intracellularly, as described previously [36]. Cells expressing WT receptor that were preincubated with 10 μM Telmisartan looked similar to the untreated condition. In agreement with the ELISA data (Figure 4A), the F66A mutant was located intracellularly, even in the absence of agonist ligand, thus, showing constitutive internalization (Figure 4C). Furthermore, Telmisartan preincubation of cells expressing F66A induced a redistribution of the F66A receptor mutant to the cell surface, also consistent with the increased ELISA signal in the presence of Telmisartan (Figure 4A). However, F66A mutants that were either constitutively internalized or internalized in the presence of Ang II, were not colocalized with β-arrestin2 (Figure 4C), unlike the WT receptor (Figure 4B). In this assay, we observed that overexpresssion of β-arrestin2 enhanced internalization of the WT receptor in response to Ang II, possibly because Ang II treatment appeared to be somewhat toxic for the cell lines, particularly WT, L119R, and D125A. The L67R mutant was localized intracellularly suggesting this mutant receptor, like F66A is also constitutively internalized. However, unlike F66A, the localization of L67R was not altered by Telmisartan (supplementary material Figure S1). The remaining mutants, L70R, L119R, D125A and I245F, were all primarily localized to the cell surface in the absence of ligand, and were co-localized intracellularly with β-arrestin2 upon Ang II treatment (Figure S1). As for the WT receptor, internalization of the L119R mutant by Ang II was also enhanced by β-arrestin2 overexpression, while the others (L70R, D125A, and I245F) were internalized in response to Ang II even in the absence of β-arrestin2 overexpression.</p><!><p>To measure effects on more downstream signaling pathways, the receptor mutants were tested in R-SAT®. R-SAT® assays are typically very sensitive to detecting constitutive activation, possibly due to the duration of the assay [34]. The N111A mutant was included as a positive control since this mutant has been shown previously to be constitutively active in R-SAT® [28]. Figure 5 depicts basal responses normalized to WT response. F66A, D125A and I245F show moderate constitutive activity, while the positive control N111A shows robust constitutive activity. pEC50 and efficacy values from Ang II dose response curves are depicted in Table 3. The F66A mutant had lower efficacy when compared to WT. The remaining mutants show efficacies similar to that of WT, thus, following the tendency of the inositol phosphate and ERK1/2 phosphorylation assays. The positive control N111A shows super efficacy for Ang II as reported previously [28].</p><!><p>In this study, we characterize six AT1 receptor mutants predicted computationally to play a role in the conformational switch linked to 7TM receptor activation and which form a patch localized to the cytoplasmic core of the helical bundle of 7TM receptors. Overall, the experimental data support the role for this patch in AT1 receptor conformational switching, since each mutant (except the poorly expressed L67R) shows either moderate constitutive activity or a change of state towards activation, detectable in certain assays. These findings sustain the notion that 7TM receptors, in our case the AT1 receptor, can adopt different conformations and that monitoring several different signaling endpoints is critical for accurate detection of changed activation states. Collectively, the data support a role for this patch in receptor activation and hereby illustrate the benefits of combining computational analysis and mutational data to generate insights to specific functional properties of family A 7TM receptors.</p><p>The receptor mutants show different signs of a changed activation state in the different assays. This is consistent with other mutational studies of molecular determinants of AT receptor activation that have also reported mutants that show only slightly elevated basal activity or increased EC50 values through the inositol phosphate pathway as reviewed recently by Petrel et al. [14, 15, 37] and also confirms the advantage of probing different pathways when looking at the AT1 receptor as mentioned in the introduction section. Notably, the WT AT1 receptor also shows no or low constitutive activity compared to untransfected cells indicating that the basal state of the receptor might be more tightly controlled than for certain other 7TM receptors [14, 38].</p><p>All of the mutants show increased binding affinity for Ang II in the competition binding assay. This has been associated with a shift towards the active conformation based on the theory that agonists preferably bind and thus stabilize the active conformation of the receptor [1]. Despite this only the L70R, L119R and D125A mutants showed increased basal activity compared to WT in the inositol phosphate assay whereas the I245F mutant showed a tendency towards increased basal activity in the R-SAT® assay. Signaling responses and receptor-surface expression levels are usually positively correlated. Although, this correlation is most likely not completely linear, a normalization of signaling responses to the surface levels of the receptor may provide certain insights about the receptor mutants and suggest tendencies to constitutive activity. If we normalize the inositol phosphate data to the surface expression, there is a tendency towards constitutive activity for all of the receptor mutants in the range of 2–4 fold of the WT receptor. Similarly, if ERK1/2 data are correlated to ELISA data, not only the F66A and D125A, but also the L70R mutant show tendencies towards increased basal activity.</p><p>F66A showed very low surface expression, constitutive internalization and responded poorly to Ang II stimulation in all signaling assays, which is most likely a reflection of the low expression levels. If we normalize the inositol phosphate accumulation response of F66A to its surface expression, it shows a basal activity of 42% of WT Ang II response whereas the WT receptor only shows a basal activity of 3% of WT Ang II response. Furthermore, the normalized F66A Ang II induced response is a 178% when compared to the WT response. Similarly, the basal activity in the ERK phosphorylation assay is 203%, and the Ang II response is 456% of WT Ang II for ERK1/2 phosphorylation. Although these calculations may indicate that F66A is substantially more constitutively active than it appears,, a linear correlation between signaling and surface expression is clearly unlikely when the difference in expression levels are this pronounced. Therefore, these calculations should be considered with caution.</p><p>Surface expression of the F66A mutant was increased both qualitatively and quantitatively by pre-incubation with the inverse agonist Telmisartan. This could suggest this internalization is dependent on an active receptor conformation, which has been observed for other constitutively activated AT1 receptor mutants [16]. Previous studies have shown that constitutive internalization can be mediated by constitutive association with β-arrestin2 [35]. However, we did not see co-localization with β-arrestin2 in the untreated, nor the Ang II stimulated condition. Importantly, it cannot be excluded that the low levels of receptor expression, validated by ELISA, complicate detection of recruitment by this method. The low cell surface expression of some of the receptor mutants compared to the WT could also result from structural instability of the receptors. Studies of constitutively active receptor mutants of the β2-adrenergic receptor also showed decreased stability compared to the WT receptor possibly reflecting a gain of flexibility due to removal of structural constraints by mutation [39, 40]. The expression of these mutants could be stabilized by the presence of either an inverse agonist or agonist.</p><p>The results on the F66A mutant clearly represent the most significant example of the importance of testing several endpoints in characterization of 7TM receptor mutants. Since most mutations reported to be non-signaling have only been tested for G protein dependent signaling endpoints, these mutants could in fact be constitutively active with respect other signaling pathways. Thus, a reevaluation of some of these mutants in more signaling pathways could potentially yield new insights in the molecular mechanisms behind receptor activation.</p><p>The identified activation switch patch of residues is located in the cytoplasmic part of TM2 (F66, L67, L70), TM3 (L119, D125) and TM6 (I245, location and standard nomenclature is reported in Table 1 and Figure 1). With the exception of the F66A mutation (2.42), which to our knowledge, has not been reported previously, all the other mutations in the ET patch were previously reported to confer constitutive activity in some 7TM receptors. Based on the general theories of 7TM receptor activation, which involves outward movements of the cytoplasmic part the helical bundle, it seems plausible that the patch mediates interactions that stabilize the inactive conformation [2]. The most extensively studied being the D125 (3.49) residue which is part of the conserved DRY motif. The residue has been proposed to interact with the neighboring Arg (3.50) residue and hereby aiding to keep the receptor in the inactive conformation [2]. The increasing number of available crystal structures of 7TM receptors published during recent years also provides interesting perspectives to the role of the patch in receptor activation. Here, residue positions from the patch have been found to be involved in hydrophobic interactions (2.43, 2.46, 3.43, and 6.40) [41] or participate in a hydrogen bonding network (2.43 and 6.40) [42]. Interestingly, in the structure of the Gα peptide bound opsin, position 2.42 (corresponding to the F66A mutation), which has not been previously characterized) was proposed to indirectly interact with the 3.49 residue through a water molecule (supplementary information to [43]). This suggests a role for this residue position in conversion between inactive and active states, though further studies will be needed to establish the exact nature of these interactions.</p><p>In conclusion, we were able to computationally predict a patch of conserved residues in the cytoplasmic core of the rat AT1a receptor important for receptor activation as confirmed by experimental studies. We illustrate how a combination of unbiased computational prediction and the knowledge from published mutational data provides an exciting tool to explore the molecular mechanisms of specific receptor functions. Due to the nature of the analysis, this approach will most likely be applicable to family A 7TM receptors in general. Our study support the current theories of 7TM receptor activation involving shared concerted movements of the transmembrane helices, but also show how the ability of 7TM receptors to adopt multiple conformations requires the need of testing several signaling endpoints to accurately detect and characterize a change in activation state for the individual mutant.</p><!><p>The residues F66, L67, L70, L119, D125, and I245 are depicted in a model of the helical region of the rAT1a receptor made using Swiss model's GPCR mode (http://swissmodel.expasy.org//SWISS-MODEL.html) (for details and standard nomenclature, see Table 1). The rhodopsin structure PDB: 1F88 was used for helical alignment. The helical bundle is viewed from the cytoplasmic side. Hydrophobic residues are shown in blue and D125 is shown in red. The figure was made using the Yasara software.</p><!><p>Whole cell competition binding assay was conducted on stable cell lines as indicated in Table 2. Normalized mean values ± S.E.M. from 3 independent experiments are reported. Statistical analysis is indicated in Table 2.</p><!><p>A: Inositol phosphate accumulation in response to full and partial AT1 receptor agonists was determined on stable HEK293 cell lines expressing mutant and WT receptors. Left to right bars represent: Basal, Ang II 0.1 μM, SII Ang II 18.75 μM, ST Ang II 5 μM, Telmisartan 1 μM. Data represent normalized mean values ± S.E.M. from at least 4 independent experiments. *, #, and ¤ indicate p<0.05 compared to WT basal, SII Ang II, and ST Ang II respectively in two tailed unpaired Student's t-test on un-normalized mean values. B: ERK1/2 phosphorylation was determined by Western blotting of lysates from stable HEK293 cell lines expressing WT or mutant receptors stimulated with 0.1 μM Ang II for 10 mins. or 1 μM Telmisartan for 20 mins. Top: To quantify ERK1/2 phosphorylation response, Phospho ERK1/2:Total ERK1/2 ratios were normalized to max WT response (WT Ang II) for each set of blots. Normalized mean values ± S.E.M. from at least 3 independent experiments are reported. Bottom: Representative blots from the experiments are depicted: B: Basal, A: Ang II, and T: Telmisartan. WT Basal and Ang II were included on every gel for normalization.</p><!><p>A. Cell surface expression levels were determined using an ELISA approach against the FLAG tag of the receptors. To assay total protein, cells were permeabilized during blocking and antibody incubation periods. Graph shows normalized mean values ± S.E.M., n=4–7 for either non-permeabilized cells with or without pre-incubation with 10 μM Telmisartan for 90 mins. (Surface and Surface plus Tel) or for permeabilized cells (Total Protein). Triplicate mean values were normalized firstly by subtraction of values for cells without antibody, and secondly, by subtraction of values for empty HEK293 cells. B (WT) and C (F66A) cell surface expression, internalization, and β-arrestin2 colocalization patterns were analyzed by antibody feeding assay on HEK293 cells stably expressing receptor variants and transiently transfected with GFP tagged β-arrestin2. Cells were pre-incubated with medium containing 10 μM Telmisartan or control treated with regular medium for 30 mins, fed with M1 anti Flag antibody for 30 mins, and then stimulated with 1 μM Ang II or control treated with PBS for 30 mins. Cells were subsequently fixed, blocked, permeabilized and stained with fluorescent secondary antibody. Slides were analyzed by confocal microscopy. For each figure (B and C), left panel shows staining for M1 Flag tag, middle panel shows GFP tagged β-arrestin2, and right panel shows overlay of M1 Flag tag (red) and GFP tagged β-arrestin2 (green). Pictures are representative of at least two independent experiments. Results for remaining mutant receptors can be found in supplementary material Figure S1.</p><!><p>on cells transiently transfected with hAT1 WT or mutated receptors. Normalized means and S.E.M. from at least 8 experiments are reported. Basal activities could be inhibited by Telmisartan, data not shown.</p><!><p>To predict residues involved in 7TM receptor activation, ET analysis was performed in combination with analysis of mutational data on constitutive activity from the literature. Mutations most likely to cause constitutive activity in the AT1 receptor were then predicted based on records of constitutive activity in the literature, natural variance at the sequence loci in Ang II receptors, and lastly more dispruptive substitution were favored. The resulting six amino acids and selected mutations are reported here. Residues are shown both as numbered from the N-terminal of the AT1 receptor and according to the Schwartz and Ballesteros standardized numbering schemes. Percent rank in ET analysis is reported along with mutation records on constitutive activity (CA) in the literature and variations in Ang II receptor sequences.</p><!><p>Whole cell competition binding assay was performed on stable HEK293 cell lines expressing WT or mutated AT1 receptors. 125I - SI Ang II was used as radioligand and competed with unlabeled Ang II. Average pKi ± S.D. for 3 independent experiments are reported. pKi values were calculated based on pKd values from experiments using unlabeled SI Ang II as competitor.</p><!><p>= p < 0.05 determined by two-tailed paired Student's t-test compared to WT.</p><!><p>on cells transiently transfected with human AT1 WT or mutated receptors. pEC50 values ± S.D. are reported,</p><!><p>= p<0.05 compared to WT in unpaired two-tailed Student's t-test. Efficacy in percent compared to WT ± S.E.M.</p>

PubMed Author Manuscript

A Promising Therapeutic Soy-Based Pickering Emulsion Gel Stabilized by a Multifunctional Microcrystalline Cellulose: Application in 3D Food Printing

The feasible application of additive manufacturing in the food and pharmaceutical industries strongly depends on the development of highly stable inks with bioactive properties. Surface-modified microcrystalline cellulose (MCC) shows the potential of being a useful particulate (i.e., Pickering)-type emulsifier to stabilize emulsions. To attain desired therapeutic properties, MCC can also be tuned with cationic antimicrobial compounds to fabricate an antimicrobial printable ink. However, due to the formation of complex coacervates between the two, the Pickering emulsion is very susceptible to phase separation with an insufficient therapeutic effect. To address this drawback, we reported a green method to produce antioxidant and antimicrobial three-dimensional (3D)-printed objects, illustrated here using a printable ink based on a soy-based particulate-type emulsion gel stabilized by a surface-active MCC conjugate (micro-biosurfactant). A sustainable method for the modification of MCC is investigated by grafting gallic acid onto the MCC backbone, followed by in situ reacting via lauric arginate through Schiff-base formation and/or Michael-type addition. Our results show that the grafted micro-biosurfactant was more efficient in providing the necessary physical stability of soy-based emulsion gel. The grafted micro-biosurfactant produced a multifunctional ink with viscoelastic behavior, thixotropic property, and outstanding bioactivities. Following the 3D printing process, highly porous 3D structures with a more precise geometry were fabricated after addition of the micro-biosurfactant. Dynamic sensory evaluation showed that the micro-biosurfactant has a remarkable ability to improve the temporal perceptions of fibrousness and juiciness in printed meat analogue. The results of this study showed the possibility of the development of a therapeutic 3D-printed meat analogue with desired sensory properties, conceiving it as a promising meat analogue product.

a_promising_therapeutic_soy-based_pickering_emulsion_gel_stabilized_by_a_multifunctional_microcrysta

Introduction<!>Material<!>Synthesis of Grafted MCC Conjugate<!>Fourier Transform Infrared (FTIR) Spectroscopy<!>Solid-State 13C NMR Spectroscopy<!>X-ray Photoelectron Spectroscopy (XPS)<!>X-ray Diffraction (XRD)<!>Water Contact Angle<!>Scavenging Activity on DPPH Free Radicals<!>Total Antioxidant Capacity<!>Antimicrobial Properties<!>Preparation of Soy Protein-Based Pickering Emulsion<!>Rheological Experiment<!>Creep and Creep-Recovery Test<!>Three Interval Thixotropy Test (3ITT)<!>Printing Process<!>Printing Performance Measurement<!>Temporal Dominance of Sensations (TDS)<!>Statistical Analysis<!>Characterizations of Grafted MCC Conjugate<!><!>FTIR Measurement<!><!>13C NMR Spectroscopy<!>XPS Measurement<!>XRD Pattern<!>Morphological Evaluation<!>Contact Angle<!>Antioxidant Activity and Reducing Power<!>Antimicrobial Activity<!>Flow Curve of Soy Protein-Based Pickering Emulsions<!><!>Flow Curve of Soy Protein-Based Pickering Emulsions<!>Strain Sweep of Soy Protein-Based Pickering Emulsions<!>Frequency Sweep of Soy Protein-Based Pickering Emulsions<!>Creep-Recovery of Soy Protein-Based Pickering Emulsions<!>3ITT of Soy Protein-Based Pickering Emulsions<!>Printing Performance<!><!>Printing Performance<!>Morphology of 3D-Printed Objects<!><!>Dynamic Sensory Evolution<!><!>Dynamic Sensory Evolution<!><!>Author Contributions<!>

<p>Three-dimensional (3D) printing is considered to be the process of material assembly to construct 3D structures in a layer-by-layer manner.1 This technology can contribute to waste reduction and to the significant advancement of environmental sustainability on account of the decreased transportation, storage requirements, and the decentralization of material preparation.2 In this sense, the importance of selecting and characterizing the precise materials to obtain the desired structural and rheological properties, as a key requirement to attain an effective 3D printing process, can be clearly appreciated.3 Emulsion templating has been considered a common approach in additive manufacturing as it can offer proper pseudoplastic- and viscoelastic-type inks, with a reversible network that can lead to thixotropic behavior. Nevertheless, 3D printing using such emulsions is hampered by emulsion instability during application and its poor self-supporting feature. The latter presents numerous challenges in attaining an acceptable level of shape control. To overcome these obstacles, the application of particle-stabilized emulsions (Pickering emulsions) to the 3D printing process has attracted considerable attention in recent years.4−6 The main advantages of Pickering emulsions are their much-enhanced stability against droplet coalescence, arising from the almost irreversible absorption of particles at the interfaces, and their favorable flow characteristics.</p><p>Microcrystalline cellulose (MCC) is a highly crystalline material developed with the partial acid hydrolysis of α-cellulose, which has an extensive application in the pharmaceutical and food sectors. It shows a unique functional property as a thickening agent with rheology-modifying ability.7 However, unmodified MCC possesses a high surface charge density with a hydrophilic nature and hence offers a poor affinity for adsorption at oil–water interfaces. This makes MCC, in its native form, an unsuitable emulsifier for stabilizing O/W emulsions. Controlled physicochemical grafting modifications, through introducing hydrophobic groups, can remedy this shortcoming and confer innovative features to biopolymers without suppressing their desirable inherent properties.8 There exist several techniques for grafting onto the surface of a crystal. These include chemical grafting, high-energy radiation modification, plasma-induced technique, and grafting of polymeric chains.9 Despite numerous attempted techniques in hydrophobic modification, many remaining challenges still prevent the large-scale implementation of such MCC-based emulsifiers. The expensive modifier agents and environmental concerns they pose (regarding chemical process), steric hindrance (concerning polymer chains grafting), degradation of the polymeric backbone (e.g., in high-energy radiation-based techniques), and hydrophobic recovery (in the case of plasma treatment) are a few examples of such issues, limiting the wider commercial uses of hydrophobically modified MCC.9,10</p><p>In recent years, fabrication and application of antioxidant–biopolymer conjugates through the grafting of phenolic compounds onto the biomacromolecules backbone, have received significantly more interest.11 Reportedly, the grafting of the phenolic compounds offers biopolymers with desired flow properties and improved emulsifying features. Moreover, the grafted phenolic conjugates could also provide reinforced bioactivity (i.e., antioxidant activity) in comparison with their free-phenolic counterparts.12−14 Gallic acid (GA), as a secondary metabolite in plants, is recognized to provide therapeutic features such as antioxidant, anticancer, and anti-inflammatory properties.12 The GA grafted onto a biopolymer (for instance, MCC) can also be reacted with a biologically derived amphiphilic tag to enhance the emulsification stability. Particularly, the cationic surfactants were revealed to interact by a negatively charged biopolymer (i.e., MCC). Lauric arginate (LAE), as a biologically derived surface-active compound, is obtained from L-arginine and lauric acid, showing excellent therapeutic effects. There are several published works on the therapeutic effects of LAE against many bacteria in the microbiological growth medium15 and food matrices16 when used alone17 or in combination with other antimicrobials.18 As a valuable method, grafting LAE on the MCC can then decrease intra- or intermolecular hydrogen linkages, increasing the surface hydrophobicity of MCC and thus improving its emulsification property and biological activities of the pristine MCC.11</p><p>The accelerated growth of human population in the world and the subsequent impacts that this has on the consumption of natural resources are likely to lead to a lack of availability of proteins with a high biological significance. Recently, soy protein has received much attention as a promising resource in the development of meat alternatives due to its ability to provide suitable gel-like structures. However, soy protein alone is unable to offer a proper fibrous and layered matrix exhibited by common meat. In general, the soy-based emulsion is an unstable system, with soy protein having poor emulsifying capability compared to milk and other animal-derived proteins. When soy is employed to develop an O/W emulsion, only a small portion of this protein actually absorbs the droplets. This is due to its high molecular weight, compact globular structure, and low solubility. It was stated that the utilization of Pickering emulsions and emulsion gels in the formulation of meat analogues can improve its textural features.19 The gel-like structure produced with this multisystem can provide the essential consistency in the meat alternative products, ascribing the fibrous perception. On the other hand, the rising health awareness and increasing demand for reduced-fat meat alternatives drive researchers to focus on reducing the original fat in these products without the deterioration of the sensory quality. Once again, the application of Pickering emulsions and emulsion gels could offer a more effective technique for the replacement of the oil phase in the fat-substitute meat alternatives, while also preserving the textural and sensory features of the products.19</p><p>The main objective of the current work was to fabricate a reduced-fat 3D-printed meat analogue having a therapeutic property using the functionality of a modified bioactive MCC. To this end, a novel sustainable process for the modification of MCC was introduced. By grating GA onto the MCC backbone, an antioxidative MCC-g-GA was synthesized. This compound was in situ reacted with LAE through Schiff-base formation and/or Michael-type addition to offer therapeutic behavior to MCC-g-GA. Next, different levels of the grafted MCC conjugate were introduced to the soy-based emulsion to partially replace its oil phase. Finally, the prepared emulsion was processed with an extrusion-type printing system, and changes in the printing performance, morphology, and dynamic sensory features of 3D-printed objects were evaluated.</p><!><p>The soy protein isolate (SPI) (moisture: 4.83%, fat: 0.32%, protein: 92.88%, ash: 3.40%, pH: 7.09, and viscosity of 1 wt % solution: 10.0 cP) was obtained from Archer Daniels Midland Company (ADM, Decatur, IL). Microcrystalline cellulose (MCC) Avicel PH-101 was purchased from Sigma (Sigma-Aldrich GmbH, Sternheim, Germany). The lauric arginate (LAE, C20 H41 N4 O3 Cl, ≥98% purity, contained 20% LAE ± 1 in propylene glycol) with a commercial name of CytoGuard LA 2X was obtained from A&B Ingredients (Farfield, NJ). 4-(2-Hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES) was purchased from Sigma-Aldrich (Steinheim, Germany). Free radical 2,2-diphenyl-1-picrylhydrazyl (DPPH), gallic acid (GA), and 2,4,6-tripyridyl-S-triazine (TPTZ) were supplied from Sigma-Aldrich (St. Louis, MO). Beet juice extract was purchased from Nature's Bounty (Winnipeg, Manitoba, Canada). All other reagents used were of analytical grade and used without further purification.</p><!><p>To modify the surface of MCC, 3.0 g of MCC with an average particle size of 30 μm was homogeneously dispersed in 100 mL of deionized water and stirred for 30 min. Then, HEPES (0.5 g) was introduced into the MCC suspension and pH was adjusted to 8.5 by the addition of NaOH. This was followed by the sonication of the MCC suspension using an ultrasonic cleaning device (Bandelin 400, Berlin, Germany) at 10 kHz for 2 min. Next, the sonicated suspension was vigorously stirred at ambient temperature overnight. Afterward, the GA (0.3 g) was introduced into the MCC suspension (pH = 8.5) and stirred constantly by a magnetic stirrer. The reaction was done for 18 h at ambient temperature and under atmospheric pressure conditions. In the final step, LAE (0.3 g) was added to the MCC/GA suspension. This was continuously stirred for a further 18 h under ambient conditions. The product was then centrifuged (Eppendorf centrifuge 5417R, Hamburg, Germany) at 1409 G-force for 5 min and washed four times with deionized water, followed by four more times with ethanol. The MCC/GA/LAE particles were oven-dried overnight at 40 °C and stored for further characterization.</p><!><p>The transmission infrared spectra of the pristine MCC, MCC/GA, and grafted MCC conjugate were identified with an FTIR spectrometer (Jasco FT/IR6200, Tokyo, Japan) to ensure the presence of the grafting process. The solid samples needed for the FTIR assay were obtained in the pellet form by blending 10 mg of each sample with 100 mg of dry KBr. Next, the samples were transferred to pellets to scan the spectral area at the wavenumber range of 400 and 4000 cm–1, whereby 50 scans were recorded at 1 cm–1 resolution.20</p><!><p>To further confirm the grafting of the MCC backbone upon the introduction of GA and LAE, solid-state 13C NMR was employed using a Bruker spectrometer (AvanceIII 500, Bruker, Ettlingen, Germany). The device was linked to a 4 mm MAS (magic angle spinning) probe, where the frequencies of carbons and protons were 75.46 and 300.13 MHz, respectively. The external reference was glycine which was utilized aimed at the 13C spectra and in order to set the Hartmann–Hahn matching condition in the cross-polarization experiments. The spectrum of each sample was obtained with the ramp {1H} → {13C} cross-polarization (CP)/MAS pulse sequence using the proton decoupling upon acquisition. The recycling period of 10 s and a contact time of 3 ms during CP were adjusted for all experiments. The SPINAL64 (small phase incremental alternation with 64 steps) sequence was employed for heteronuclear decoupling upon acquisition with a proton field H1H satisfying ω1H/2π = γHH1H = 62 kHz. The spinning rate for all samples was 10 kHz.5</p><!><p>The XPS assay was also conducted to further verify the grafting modification of MCC. The experiments were conducted through a Kratos Axis spectrometer (Ultra Kratos Analytical, Manchester, U.K.) via a monochromatic Al Kα source and a 180° hemispherical electron energy analyzer, working at a pass energy of 65 eV. The step size was 0.1 eV with a dwell time of 1000 ms. The analyzed zone was adjusted at a region of 300 × 700 μm2. A Shirley baseline was employed aimed at the subtraction of the background, and Gaussian/Lorentzian (70/30) peaks were applied for spectral decomposition. Each spectrum was analyzed through Vision software supplied by Kratos (Vision 2.2.2, Ultra Kratos Analytical, Manchester, U.K.).</p><!><p>The XRD diffractogram was obtained by an X-ray diffractometer (Shimadzu XRD 7000, Tokyo, Japan) with Cu Kα irradiation. The samples were exposed to the X-ray beam at 2θ angles ranging from 2 to 60° running at 45 kV and 40 mA, employing Cu Kα radiation (λ = 1.541 Å) at a rate of 2° min–1. To evaluate the relative crystallinity degree (RCD), the total curve area (It) and the area under the peaks (Ip) were determined using the software offered by Shimadzu, and RCD was then calculated according to eq 1(21)1</p><!><p>The contact angle (CA) was obtained by an OCA 20 contact angle meter (Dataphysics Instruments GmbH, Filderstadt, Germany) using the sessile drop approach in natural light. A uniform thin film was fabricated using a KW-4A spin-coater (CHEMAT Technology Northridge, CA). This was achieved by spin coating of 2.0 wt % pristine and modified MCC suspensions (in toluene) onto silicon wafers at a shear rate of 210 s–1 (5690 G-force) for 1 min. This was then followed by heat treatment at 90 °C overnight. The resulting films were sectioned 4 × 6 cm2 and placed on a horizontal movable stage. A drop (5 μL) of deionized water with a syringe (10 μL, Hamilton, Switzerland) was deposited centrally on the surface of the films. The data were analyzed by the software offered by the Dataphysics Instruments.22</p><!><p>The stock DPPH solution was obtained by introducing 5.0 mg of DPPH in methanol (100 mL). The aqueous suspensions/solutions of pristine MCC, GA, LAE, and grafted MCC conjugate were individually prepared by dispersing 50.0 mg of each sample in 100 mL of deionized water and stirred for 60 min. Then, the DPPH solution (2 × 10–4 M, 100 μL) was blended with these aqueous suspensions/solutions (100 μL). The resulting mixtures were shaken vigorously and incubated at ambient temperature in the dark for 1 h. Next, the reactants were centrifuged (Eppendorf centrifuge 5417R, Hamburg, Germany) at 4000 G-force for 5 min. The absorbance was then measured at 517 nm using a spectrophotometer (UVIDEC-50, JASCO Corporation, Tokyo, Japan). The scavenging effect of DPPH radical was obtained as follows2where A0 is the absorbance of the control (using deionized water instead of the sample), As is the absorbance of the samples mixed with reaction solution, and Ab is the absorbance of the sample under the same condition as As, but where ethanol was used instead of the ethanol solution of DPPH.</p><!><p>The ferric reducing antioxidant potential was used to determine the total antioxidant capacity. The ferric reducing antioxidant potential working solution was provided by mixing 100.0 mL of sodium acetic buffer (0.3 M, pH 3.6), 10.0 mL of TPTZ (10 mM, dissolved in 40 mM HCl), and 10.0 mL of FeCl3 solution (20 mM).23 The mixture containing 100 μL of sample and 200 μL of the potential ferric reducing antioxidant solution was incubated at room temperature for 20 min, and the Ab and As values were detected at 593 nm.3</p><p>where, once again, A0 is the absorbance of the control (using deionized water instead of the sample), As is the absorbance of the samples mixed with a working solution, and Ab is the absorbance of the sample under the same condition as As, but with ethanol used instead of the working solution.</p><!><p>Three bacteria cocktails containing identical populations of five trial strains/serovars were utilized in the antimicrobial assessment, including (1) Salmonella enterica cocktail: Salmonella montevideo, Salmonella gaminara, Salmonella agona, S. Michigan, and S. Saint Paul; (2) Escherichia coli O157:H7 cocktail: H1730, K3995, F4546, 658, and 932; and (3) Listeria monocytogenes cocktail: LM1, LM2, 310, Scott A, and V7. Tryptic soy broth (TSB) was employed for S. enterica and E. coli O157:H7, and Tryptic soy broth supplemented with yeast extract (TSBYE) was applied for the growth of L. monocytogenes.24</p><p>A disk diffusion technique was applied to assess the antimicrobial features of the samples. First, the MCC film variants were fabricated by the spin coating method (see the Water Contact Angle section). Then, the tryptic soy agar (TSA) or TSA supplemented with yeast extract (TSAYE, for L. monocytogenes) plates were spread with 100 mL of culture containing 105 CFU mL–1 of bacteria cocktail. Two circular disks of each film specimen (10 mm) were transformed into each plate. After incubation for 24 and 48 h at 32 °C (L. monocytogenes) or 37 °C (E. coli O157:H7 and S. enterica), the diameter (mm) of inhibition zones was determined using a ruler. The mean values of inhibition zone diameters from the two films, with two disks each (n = 8), are reported.</p><!><p>The SPI aqueous dispersion was made by dispersing SPI powder (25.0 g) into part of the citrate phosphate buffer (pH 5.6, 60 mL), with the rest of the water being used for the grafted MCC conjugate. Next, salt (0.1 g) and beet juice extract (0.3% v/v) were introduced into the SPI-based dispersion. This was followed by gentle stirring of the SPI-based dispersion at 40 °C for 60 min, using a magnetic heater stirrer. At the same time, 10% (v/v) sunflower oil was incorporated into the SPI-based dispersion with a burette. The obtained emulsions were stirred by an Ultra-Turrax at a speed of 210 s–1 (5690 G-force) for 5 min. Separately, a stock suspension of grafted MCC conjugate was made by dispersing a weighed amount (70 wt %) of the grafted MCC powder into the same buffer (pH 5.6, 40 mL). This was mixed with a high-speed rotor-stator device (Ultra-Turrax T25D IKA, Germany) at a shear rate of 400 s–1 (20664 G-force) for 10 min at an ambient temperature. This grafted MCC suspension was gently stirred overnight at room temperature. The pH of the suspension was then adjusted back to 5.6.</p><p>An O/W emulsion was prepared by blending sunflower oil 10 and 90 wt % aqueous SPI-based dispersions (25.0 wt % SPI, pH 5.6) using a high-speed blender (Ultra-Turrax T25D IKA, Germany) for 5 min. This coarse emulsion was homogenized by a two-stage high-pressure Microfluidizer processor (M110-PS, Microfluidics international Corp., Newton, MA) with 1800 psi at the first stage and 700 psi at the second stage. The full-fat stabilized emulsion, regarded as control hereafter (10 wt % sunflower oil, 25.0 wt % SPI, pH 5.6), was employed for comparison with reduced-fat emulsions. Hence, 20, 40, 60, and 80% reduced-fat emulsions were prepared with grafted MCC stock suspension (70 wt % grafted MCC, pH 5.6) to obtain the Pickering emulsions with the following compositions: 8 wt % sunflower oil and 1.4 wt % grafted MCC (SGM1), 6 wt % sunflower oil and 2.8 wt % grafted MCC (SGM2), 4 wt % sunflower oil and 4.2 wt % grafted MCC (SGM3), and 2 wt % sunflower oil and 5.6 wt % grafted MCC (SGM4).</p><!><p>The rheological behavior of ink samples was characterized by an AR 2000ex rheometer (TA Instruments, New Castle, DE) using a parallel-plate geometry (diameter of 40 mm, gap size of 1 mm). The oscillatory strain sweep (0.1–100%, frequency = 1 Hz) was performed to attain the limit of the linear viscoelastic region (LVR). Apart from this, the frequency sweep test (0.1–100 Hz) was accomplished within the LVR (γ = 1%). All measurements were performed at 25 °C.19,25</p><p>To evaluate the steady rheological properties, the shear stress was measured as a function of increasing shear rate (γ̇) from 0.1 to 100 s–1. The best constituent rheological equation was selected via statistical analysis, and the rheological variables were measured using fits to the optimum model.26 Hence, the consistency index, flow behavior index, and yield stress values were obtained by fitting the Herschel–Bulkley model to the data (eq 4).4where τ is the shear stress (Pa), τ0 is the yield stress (Pa), K is the consistency index (Pa sn), γ̇ is the shear rate (s–1), and n is the flow behavior index.</p><!><p>The creep and creep-recovery measurements were performed to evaluate the compliance level during the creep and recovery stages via an AR 2000ex rheometer (TA Instruments, New Castle, DE). First, a stress sweep (1 Hz, 0.1–10 Pa) was accomplished (data not shown) to evaluate the oscillatory yield stress (G′(τ) = G″(τ)), and then the obtained values were considered as being about 50% of the yield stress.5,19 The inks were moved to a parallel-plate geometry with a diameter of 40 mm and a gap size of 1 mm, maintained at 25 °C. The creep measurement included the prompt application of a constant shear stress within the LVR area, lasting from 0 to 500 s while evaluating the sample deformation during these time intervals. Regarding the recovery phase, the applied stress was rapidly removed (τapplied = 0.0 Pa) and the recovery values were recorded for an additional 500 s at the same temperature as that in the creep phase.19 The calculated strain and the recovery are considered to be the creep compliance and creep-recovery compliance (J) (eq 5). The creep-recovery percentages of inks were then obtained according to eq 656where J(t) (Pa–1) is creep compliance at time t, γ is the measured strain, τ0 is the constant applied shear stress, Jm (Pa–1) is the maximum creep, and Je (Pa–1) is the equilibrium creep compliance after recovery.</p><!><p>3ITT involved a three-step shear rate test, where the first one comprised a steady shear rate to recognize the ink reference stage without interrupting the microstructure. This was accomplished with a fixed shear rate of 1 s–1 for 400 s. In the second interval, a steady shear rate of 80 s–1 was applied for 200 s and used to terminate the microstructure of the ink. The third interval included a similar assessment condition as the first interval, allowing the partial (or full) recovery of the original structure of the Pickering emulsion gels19 (speed and degree of recovery).</p><!><p>The 3D printing process of prepared soy protein-based inks was conducted using an extrusion-based system (nScrypt-3D-450, nScrypt, Orlando, FL), coupled with a syringe pump (PHD Ultra; Harvard Apparatus Holliston, MA). With the use of computer-aided design software (AutoCAD; Autodesk, Inc., San Rafael, CA), a special circle and butterfly-shaped objects were modeled and converted to an STL file. The print paths were provided through the creation of the G-code files to control XYZ direction instruction for the printer, developed by the open-source CAM software Slic3r (slic3r.org, consulted on December 2020) from the STL file. The printable soy protein-based inks were poured into a stainless-steel cartridge (10 mL) and stirred with a Vortex mixer (Fisher Scientific, Ontario, Canada) for 10 min, thus removing any air bubbles from the ink. The layer height was set at 1 mm, indicating that the nozzle tip was elevated by this value upon completion of the fabrication of each layer and before commencing the construction of the next layer. The process was continued until the suitable 3D architecture was printed in each case. The number of deposited layers was 8, and the width of the tip was 1 mm (Supporting Information, Table S1).</p><!><p>Each 3D-printed object was transferred into a specific chamber 20 × 20 × 20 cm3 to be photographed using a digital camera (α 7M3 E-Mount, Full-Frame Mirrorless, 24.2 MP, Sony, Tokyo, Japan). The printing performance of 3D-printed architectures was accomplished by determining the line width and layer number through a digital caliper (Mitutoyo, Absolute Digimatic, Tokyo, Japan).6</p><!><p>Ten assessors (five females and five males, aged: 20–35 years) were selected to evaluate variances in the dynamic sensory profile of 3D-printed meat analogue through the TDS method.27 In this regard, six selected attributes were listed as being firmness, juiciness, oiliness, graininess, fibrousness, and chewiness (Supporting Information Table S2). The TDS evaluation was performed over three sessions to run three replicates. The 3D-printed objects 5 × 5 cm2 were provided to the assessors by the randomized complete block design in a monadic order. The panelists were then presented with a list of six attributes on the computer screen, each associated with an unstructured scale, ranked from weak to strong. FIZZ software (Version 1.9, Biosystems, Counternon, France) was utilized to obtain the TDS plots. The TDS graphs were smoothed by MATLAB software (R2016a, MathWorks, Inc., Natick, Ma) to generate TDS curves.28</p><!><p>All instrumental experiments were carried out in triplicate with the mean and standard deviation of the data calculated and reported. Analysis of variance (ANOVA) was utilized for the determination of the main effects and to examine the independence or the interactions between various factors on the instrumental and sensory data. Duncan's multiple range test was applied to separate means of data when significant differences (P < 0.05) were observed.</p><!><p>The pristine MCC powder was purchased as a freeze-dried powder, which was simply redispersed in water with the shearing treatment (Figure 1). The introduction of GA into the MCC dispersion upon the sonication process resulted in the coated MCC (MCC/GA) that preserved the colloidal stability of MCC albeit with a somewhat yellow discoloration (Figure 1). On the other hand, the development of the grafted conjugate in the presence of LAE caused an increase in turbidity. This could imply an increase in the amphiphilicity of the system upon the grafting process and also the formation of complex coacervates between MCC and LAE.10 It was stated that the interaction of LAE with other negatively charged molecules leads to the formation of complex coacervates.29 This finally decreases the efficiency of LAE and therefore reduces the minimum inhibitory concentration of the LAE in certain food and pharmaceutical applications. Hence, there is a need to decrease the coacervate development and therefore enhance the efficiency of LAE. In the current study, we used grafting GA onto the MCC backbone to reduce the possible formation of large aggregates that leads to low turbidity and large particle size. The possible mechanism for the surface grafting of MCC as affected by introducing GA and LAE is schematically shown in Figure 1. It has been stated that interaction among biopolymers and oxidized polyphenols leads to the development of new inter- or intermolecular interactions including covalent and hydrogen linkages, metal chelation, and π effects.30 In the first phase, GA is reacted with MCC at pH 8.5 since it is recognized that under such circumstances, polyphenol oxidation and oligomerization can proceed. This is especially the case when dissolved oxygen is present. The process leads to the development of a high-molecular-weight compound with reduced solubility. Apparently, the solubility reduction of GA or its intrinsic affinity to MCC results in its grafting on the surface of MCC.31 In this situation, quinones are produced and, in the subsequent reaction phase, react by −NH2 of LAE through the Schiff-base and/or Michael-type addition reactions (Figure 1).</p><!><p>(Left) Products produced from the LAE reacting with a Schiff-base reaction and Michael addition. The dotted line signifies GA grafting onto the MCC via covalent and/or strong intermolecular interactions, as reported for other substrates (10). (Right) Photographs displaying the presence of modified MCCs in different phases regarding the one-pot water-based grafting modification.</p><!><p>Figure 2a shows the FTIR spectra of unmodified MCC, MCC/GA, and MCC/GA/LAE products. A characteristic peak of cellulose I, centered from 2500 to 3750 and 700 to 1800 cm–1, was detected for pristine MCC. Moreover, the stretching of the hydroxyl group appeared around 3380 cm–1. There are also some representative bands at 2953 cm–1 (assigned to C–H stretching), 1640 cm–1 (related to asymmetric stretching of a carboxyl group), 1430 cm–1 (associated with methylene symmetrical bending), 1080 cm–1 (allocated to cellulose C–O–C bridge), and a peak around 880 cm–1 that corresponds to β-linked glucose moieties.5 The FTIR spectrum of MCC/GA evidently shows the emergence of a new typical carbonyl stretching vibration around 1869 cm–1. Moreover, the band induced by −COOH groups shifted to the higher wavenumbers (1687 cm–1). However, the main difference between the pristine MCC and MCC/GA was the appearance of a band associated with C–O stretching, occurring around 1340 cm–1 for GA-coated MCC.32 In addition, also a new peak appeared around 791 cm–1, resulting from the distortion of the vibrations of MCC in the benzene rings, and a further band at about 1510 cm–1 caused by the stretching vibration of the C–C aromatic ring. These changes are suggestive of the interactions between MCC and GA. A relevant FTIR pattern was also stated by Hu et al.31 The FTIR spectrum of the grafted MCC/GA/LAE conjugate exhibiting the magnitude of the stretching band of hydroxyl groups (−OH) was wider and less pronounced (Figure 2a). This is likely because of the higher consumption of −OH groups of MCC, resulting from the grafting process. The grafted MCC conjugate also displayed the presence of a band around 1340 cm–1 that was thought to be the result of C–O stretching. At the same time, the asymmetrical/symmetrical stretching of methylene results in bands at 2705 and 2810 cm–1. This suggests the likely grafting of LAE to the MCC surface. Furthermore, the typical secondary N–H bending around 1490 and 1603 cm–1 confirmed the occurrence of Michael addition.</p><!><p>Characterization of grafted MCC conjugate: (a) FTIR, (b) 13C NMR, (c) XPS, (d) XRD, (e) transmission electron microscopy (TEM), (f) water contact angle, and (g) bioactive properties. In the case of bioactive features, the means inside each column with various letters (a–d) are significantly different (P < 0.05) according to Duncan's test.</p><!><p>The grafting treatment of MCC was qualitatively further verified by 13C NMR spectroscopy (Figure 2b). A characteristic band of cellulose I (i.e., C1: 110.9 ppm, C4: 98.1 ppm, C4′: 95.3 ppm, group C2–C3–C5: 70.2–80.1 ppm, C6: 69.1 ppm, and C6′: 64.9 ppm) was detected for pristine MCC.5 After GA grafting onto MCC, no obvious new peaks emerged in the NMR spectrum compared to pristine MCC. However, a relatively wide band appeared at about 145.3 ppm, which associates with the resonance of the carbonyl ester (C=O). This result was in accordance with the carbonyl stretching vibration around 1869 cm–1 detected by FTIR assay. In contrast, the changes of the NMR spectrum of grafted MCC/GA/LAE conjugate were more marked compared to GA-coated MCC. As the grafting reaction progressed, new strong resonances emerged in the NMR spectrum, as seen by the appearance of distinguished signals at 8.2, 13.2, 24.3, 34.3, 41.8, and 45.5 ppm. These changes could be caused by the terminal methyl carbons, i.e., the secondary carbon (−CH2−) and the primary carbon groups (−CH3). The obtained data further supported the effective grafting process of GA and LAE onto the MCC backbone. Moreover, the crystallinity degree measured by NMR demonstrated that only a small alteration in the crystallinity index of MCC (about 85%) occurred upon the grafting treatment (∼86%) (Supporting Information, Section S.1.1).</p><!><p>The XPS patterns of the pristine MCC, GA-coated MCC, and grafted MCC conjugate are compared in Figure 2c, providing yet further support for the indicated grafting mechanism. As expected, the XPS spectrum obtained from the pristine MCC identified oxygen and carbon. After the grafting treatment, a nitrogen peak also emerged on the XPS pattern of the grafted MCC conjugate. This latter peak results from LAE. The resolving of the nitrogen peak (396.3 eV) led to the identification of binary compounds, N1 (395.5 eV) and N2 (397.1 eV). They were likely induced by aromatic C=N and aromatic C–N, respectively.33 The presence of both aromatic C=N and C–N bands denotes the formation of both Schiff-base reaction and Michael addition between MCC/GA and LAE.33 Additionally, the peak intensity at 266.3 and 267.1 eV was increased, which could also be related to the aromatic C=N and aromatic C–N, respectively. This then also implies the successful LAE grafting on the GA-coated MCC backbone. Per the XPS patterns, some general interpretations regarding atomic ratios could be made. The theoretical proportion of O to C in pure cellulose is known to be 0.83.34 In the current study, the obtained O/C ratios were 0.79, 0.91, and 0.42 for pristine MCC, GA-coated MCC, and grafted MCC conjugate, respectively. The obtained data for the grafted MCC conjugate was not surprising due to the presence of R–NH2 chains in the LAE structure with no oxygen, where the O/C proportion reduced significantly. The discrepancy in O/C proportion regarding MCC and GA-coated MCC might also be ascribed to slight contamination.31 It is worth noting that the GA layer on the GA-coated MCC was relatively thin since if the XPS beam only interacted with GA (with a certain penetration depth of ∼10 nm), a much lower O/C ratio of 0.73 (O/C ratio for neat GA) would be anticipated.</p><!><p>The diffractogram of pristine MCC exhibited a crystalline structure with the dominance of cellulose type I in the MCC.5 In this case, there are some strong reflections around 2θ = 14.0° (d001 = 5.4 Å), 2θ = 22.9° (d001 = 4.8 Å), and 2θ = 34.9° (d001 = 4.0 Å) (Figure 2d), whose RCD was measured to be 78%. This crystallinity index was slightly lower than the crystallinity degree of about 84% obtained from the NMR experiment. The diffractogram of GA-coated MCC presented that the GA grafting onto the MCC resulted in the fading of the pronounced peak located around 2θ = 14.0°, which signified the interaction of GA with MCC in the interhelical backbone. The RCD of this sample was also dropped to 71%. The obtained result signifies a reduction in the magnitude of the crystalline reflections together with a decline in the magnitude of the pronounced diffraction peak of MCC. Moreover, the characteristic peak of MCC shifted from 2θ = 22.9° to 2θ = 22.2°, which indicates that the gallery spacing from d001 = 4.8 Å (2θ = 22.9°) increased to d001 = 4.9 Å (22.2°). It has been reported that there is a slight change in the spatial structure and unfolding of the cellulose upon the interaction of functional groups of GA with MCC.10 As Figure 2d reveals, the intensity of diffraction peaks decreased more upon the development of a grafted MCC/GA/LAE conjugate, whose RCD dropped down to 58%. The XRD diffractogram showed that the typical MCC peak (2θ = 22.9°) altered to a more diffused peak, showing a reduction in the crystallinity. Furthermore, the diffractogram of the MCC/GA/LAE conjugate displayed the appearance of a new reflection around 2θ = 8° (d001 = 6.1 Å). This could indicate that a new crystalline phase was developed in the amorphous area of cellulose through the grafting process.</p><!><p>Figure 2e shows the transmission electron microscopy (TEM) photograph of unmodified MCC, MCC/GA, and grafted MCC/GA/LAE products (For an interpretation of the reference for the preparation of samples for TEM analysis, the reader is referred to Supporting Information Sections S2). In the case of unmodified MCC, the particles are rod-shaped and are well separated from each other, with no noticeable agglomeration. Compared to pristine MCC, the size of particles in the GA-coated MCC and grafted MCC conjugate had increased, but nonetheless, they preserved their rod-shaped morphology and remained as well-dispersed individual particles, without any significant levels of agglomeration. Furthermore, there was an enhancement of the contrast dark coating on the surface of modified MCC. It has been stated that an increased contrast of dark coating in the modified MCC is probably because of the existence of aromatic groups in the phenolic compounds.31</p><!><p>The model MCC thin film was made by the spin coating approach, and a sessile drop water contact angle experiment was performed. The pristine MCC film presented a high hydrophilicity character, showing a low water contact angle of θ = 26.1° (Figure 2f). In this case, the water droplet was completely absorbed into the MCC film after 1 min. The MCC backbone is quite wettable due to the presence of a large number of hydrophilic groups.35 The contact angle measurement also exposed that the grafting of GA onto the MCC backbone did not alter the hydrophobicity of the MCC matrix (θ = 25.8°), specifying the lack of development of a hydrophobic surface/matrix in the GA-coated MCC. As Figure 2f shows, a huge enhancement in the contact angle by a value of 59.3° was detected after the formation of grafted MCC conjugate (θ = 85.4°). This could be related to the progress of intermolecular interactions between MCC, GA, and LAE due to the grafting reaction. This provides fewer hydrophilic sites on the surface of films, but with more matrix rigidity.31</p><!><p>The antioxidant behavior of a compound is related to several mechanisms, including breaking the radical chains, hampering metal-catalyzed initiation reaction, and inhibition of the introduction of initiating radicals.36 To determine the antioxidant capacity, different in vitro chemical-based experiments, including assays of radical scavenging activity on DPPH free radicals and ferric reducing antioxidant power, were performed. Among all evaluated samples, the unmodified MCC presented the lowest DPPH scavenging effect (Figure 2g). The scavenging activity of GA alone on the DPPH free radicals was also measured. This illustrated that the scavenging effect was appreciably higher than the pristine MCC. GA is considered a strong apoptosis-inducing agent and an effective antioxidant agent.10,37 Thus, the introduction of GA to MCC could reasonably induce an efficient antioxidant property with a promising therapeutic effect. Compared to the free GA compound, the GA-coated MCC and grafted MCC/GA/LAE conjugate provided a stronger quenching of DPPH radicals in a dose-dependent manner (P < 0.05). It was reported that polyphenol oxidation and oligomerization are developed after the grafting of GA onto the MCC backbone.10 This enlarges the size of the conjugated compound of GA, contributing to an increase in the electron-donating groups. Thus, the GA-coated MCC and MCC/GA/LAE conjugate have more biological stability than the free GA. We assume then that these products, with their swollen supramolecular structure, could capture the free radicals more successfully compared to the small GA molecules.</p><p>Another important parameter to evaluate the antioxidant activity of a compound is the ferric reducing antioxidant power.10 The reducing power activity of different samples is shown in Figure 2g. The reducing power of all samples, except unmodified MCC, was increased in a dose-dependent manner. The unmodified MCC showed a weak reducing antioxidant power. The GA-coated MCC, by contrast, offered stronger reducing power compared to unmodified MCC (P < 0.05) and was similar to the neat GA (P > 0.05). This is expected since GA is established for its excellent antioxidant feature via the active hydrogen-donating groups.10,37 The reducing power activity had enhanced even further for the grafted MCC/GA/LAE conjugate (Figure 2g). The obtained data suggest that the interaction of MCC by both GA and LAE improved the antioxidant behavior of unmodified MCC. In this regard, high-molecular-weight species are developed upon the grafting process that provided a biological stable compound, strongly mopping up the free radicals more efficiently compared to free GA molecules on their own.10 Thus, the idea to reinforce the antioxidant activity by grafting GA and LAE onto the MCC backbone seems a sensible one.</p><!><p>The antimicrobial activities of free LAE compound, pristine MCC, GA-coated MCC, and grafted MCC/GA/LAE conjugate were assessed by the disk diffusion experiment upon different incubation conditions (Figure 2g). The pristine MCC did not show an inhibitory effect against any of the evaluated microorganisms, where some bacteria growth was detected under the MCC films' disks. In contrast, a slightly larger inhibition zone diameter was observed for the disks of GA-coated MCC (P < 0.05), albeit the difference was not significant after 24 and 48 h (P > 0.05). A great inhibition zone diameter was also detected for disks of grafted MCC conjugate, with the differences now being significant after 24 and 48 h (P < 0.05). This demonstrated a continuous inhibitory effect of film disks of the grafted MCC/GA/LAE conjugate. The antimicrobial effect of the LAE component29 in the grafted system was also in accordance with the inhibitory effect of LAE alone, as conducted in the present study (Figure 2g). The presence of a positive charge on the protonated guanidine group of LAE can disrupt the cell membranes of microorganisms without triggering the cell lysis. However, this mechanism might also act toward various intracellular bacterial pathogens, which can also prove lethal to the bacteria.24,29 In the current work, the antimicrobial MCC offered a great inhibitory effect against L. monocytogenes compared to S. enterica and E. coli O157:H7, which was also reported by Pattanayaiying, Aran, and Cutter.38 Following 24 h of incubation, the film disks of grafted MCC/GA/LAE conjugate presented meaningfully higher inhibition zones to inhibit all three types of bacteria studied here when these were compared to GA-coated MCC film disks (P < 0.05). This shows a promising efficiency of grafted MCC/GA/LAE conjugate to enhance product safety in possible industrial applications. Compared to the grafted MCC/GA/LAE conjugate film disks, the GA-coated MCC film disks showed significantly lower inhibition zones (P < 0.05). The obtained data might have resulted from the lower inhibitory effect of GA in comparison to LAE.</p><!><p>To be appropriate for 3D printing, a biopolymeric ink must simply be squeezed out during the process, with reduced viscous flow forces inside the nozzle at the time of its application.3,5,10 However, the ink must also become viscous enough to keep the 3D-printed shape once the printing process is completed. Therefore, it is obvious that such an ink must exhibit a significant degree of pseudoplasticity. Figure 3a presents the flow curves of soy-based Pickering emulsions. Each plot was fitted to the Herschel–Bulkley equation, showing a high correlation coefficient above 0.97 (Supporting Information Table S3). As given in Table S3, the oil replacement by MCC/GA/LAE conjugate (micro-biosurfactant) led to a lower flow behavior index, suggesting that the Pickering emulsions have become more shear thinning. Generally, the inks formulated with the higher levels of grafted MCC conjugate, i.e., SGM3 and SGM4, provided lower flow behavior indices, with a clear pseudoplastic behavior (0.455 < n < 0.542) (Supporting Information Table S3). The presence of the less aggregated droplets/particles and more efficient dispersed particles in the inks including higher contents of grafted MCC/GA/LAE conjugate could be that the aggregates become less strong and break more easily at the lower shear forces, thus providing rapid shear thinning.5,19</p><!><p>(a) Flow curve of soy protein-based emulsions. (b) Amplitude sweep and (c) frequency sweep of soy protein-based emulsions, where G′ is denoted by solid symbols and G″ is denoted by open symbols. (d) Creep and creep-recovery curves and (e) 3-ITT of prepared Pickering emulsion variants. (f) Droplet size, (g) particle size distribution, and (h) ζ-potential (For more insight into these tests, see Supporting Information Section S3) of different soy protein-based inks. (i) Confocal laser scanning microscopy (CLSM) image of control emulsion (For more insight into the CLSM test investigation, see Supporting Information Section S4).</p><!><p>The viscosity, which corresponds to the consistency index of an emulsion, offered an appreciable increase when the grafted MCC conjugate was replaced by oil in the range of 4.2–5.6 wt%, i.e., SGM3 and SGM4 (Figure 3a). The SGM4 ink presented the highest viscosity (at a shear rate of 10 s–1) with an initial value of 41.7 Pa s, followed by SGM3 and SGM2 with a viscosity (at a shear rate of 10 s–1) of 38.5 and 34.5 Pa s, respectively (Supporting Information Table S3). The greater viscosity value with grafted micro-biosurfactant is likely because of the development of intermolecular interaction among the hydrophobic regions of LAE and oil on the one hand and more polar cellulose chains with the hydrophilic domains of soy protein on the other. Moreover, the higher solid volume fraction in the system upon the addition of grafted MCC may increase the particle–particle interactions, leading to the formation of weak but large clusters and decreased mobility that in turn causes an increase in viscosity.5</p><p>The yield stress difference is considered an important property of printable inks. It is associated with the reinforcement of ink printing performance in addition to the development of well-defined 3D structures.3 There was an increase in the yield of soy protein-based ink upon the addition of grafted micro-biosurfactant (Supporting Information Table S3). Compared to control ink, replacement of the lowest level of grafted MCC conjugate did not significantly alter the yield stress (P > 0.05), while at higher ratios, it provided a noticeable increase in this parameter (P < 0.05). The introduction of GA and LAE to the MCC backbone presents the possibility for linking a diverse range of polymeric chains, via the development of phenolic dimers/trimers and/or even polyphenol oxidation and oligomerization. The new linkages might result in greater molecular weight and longer chain lengths, requiring higher stress to make the ink flow. Thus, it was assumed that the high yield stress of the inks containing higher levels of grafted MCC conjugate was ascribed to the cross-linking junction zones obtained by the oxidation of GA followed by in situ reacting LAE via Schiff-base formation and/or Michael-type addition. The obtained results are imperative since the desired range of yield stress prevents the deformation of the extruded layers and the collapse of the 3D-printed structures.</p><!><p>The dynamic rheological behavior of inks was taken into consideration as it can offer valuable insights into the viscoelastic parameters of biopolymeric inks, which directly affects the printability and resolution of deposited layers.3 The results of the strain sweep test are depicted in Figure 3b. All emulsions presented a G′(γ) higher than G″(γ) for a wide range of strain amplitude, exhibiting the gel-like character of the Pickering emulsion gel. The G′ modulus shows a sign of the system rigidity; then, the inks with a greater G′ offer a more desirable structural strength and may aid to form a well-defined geometry upon the printing process. Compared to the control ink, the inks containing grafted MCC conjugate presented greater viscoelasticity throughout the entire period of the strain sweep. Besides, the G′(γ) values increased at higher ratios of grafted micro-biosurfactant. Moreover, the amphiphilic chains with several hydrophobic regions/groups along the backbone tend to become simultaneously adsorbed on the surface of two nearby droplets, thus resulting in the bridging flocculation of droplets. If droplets remain stable against coalescence, then the resulting network formed by these oil droplets exhibits strong viscoelastic behavior.5,19</p><p>Additionally, the linear viscoelastic region (LVR), where moduli remain independent of the amplitude of oscillations, was also evaluated to confirm that the experiment was carried out within LVR. The extent of LVR can reflect the strength of any formed network in the system. Stronger structured emulsion gels could remain within the LVR over a higher level of deformation compared to the weak gels.39 As Figure 3b depicts, the control emulsion possessed a short LVR extent, showing the lower critical strain (2.3%) in comparison with other samples, whereas the critical strain values of SGM1, SGM2, SGM3, and SGM4 inks were determined to be higher at 3.8 5.6, 8.2, and 9.6%, respectively. As a result, compared to the lower contents of the grafted MCC conjugates, the higher ratios of the grafted micro-biosurfactant offered a greater G′LVR with a less restricted LVR, thus also a higher critical strain. This specifies that the reduced-fat ink variants behaved more like a solid with elastic-like features, signifying the development of more structured systems under nondestructive conditions.</p><p>The loss modulus (G″) is an index of the dynamic viscous character that signifies the dissipation of energy linked to the unrecoverable viscous loss. As illustrated in Figure 3b, all ink variants showed a constant G″(γ) trend within LVR, as expected. However, at higher amplitude sweeps (>7%), G″(γ) plots crossed over with those of G′(γ), indicating a shift from more elastic behavior to a more viscous liquid-like one. The crossover points of the reduced-fat inks were transitioned to the higher strain amplitudes compared to control the ink. This suggests the development of a stronger gel-like matrix with a more structured behavior.5 As an outcome, the enhancement of viscoelastic properties of the Pickering emulsion gels resulting from the addition of a high content of grafted micro-biosurfactant may better preserve their shapes upon printing and enhance the printability and also the geometrical retention of 3D architectures.</p><!><p>Figure 3c shows the viscoelastic moduli as a function of oscillatory frequency (ω) inside LVR (γ = 1%). Concerning the control and SGM1 inks, G′(ω) dominated over G″(ω) at low frequencies (<1 Hz), demonstrating the dominance of quasi-solid behavior. However, as the frequency proceeded, G″(ω) gradually approached more closely the value of G′(ω) until a crossover point (ωc), i.e., G′(ω) = G″(ω). On the other hand, the moduli of the SGM2, SGM3, and SGM4 inks all presented a relatively linear rise by increasing the frequency. In these cases, the intensity of G′(ω) was also considerably higher as the ratio of grafted micro-biosurfactant was increased, with the stronger gel-like feature. It is proposed that the development of GA-coated MCC contains the oxidation of phenolic group for producing o-quinones or o-semiquinones.10,30 The orthoquinone then in situ reacted with the amino groups of LAE through Schiff-base development and/or Michael-type addition, which improves the hydrophobicity contact angle for the surface of MCC. MCC particles with higher hydrophobicity tend to aggregate to form stronger networks. This most likely includes MCC particles adsorbed on the surface of droplets, hence incorporating droplets into the MCC gel network, as active fillers.</p><p>The reinforcing effects of the grafted MCC/GA/LAE conjugate were also reflected on the G″(ω) modulus. In this case, G″(ω) linearly increased as the micro-biosurfactant level was increased. This offers the strengthened structure of soy protein-based emulsion and a mechanically stable matrix, showing an elastic gel-like character. As Figure 3c depicts, the slope of G″(ω) for the reduced-fat inks was higher than G′(ω) at higher frequencies (>10 Hz), where the crossover point (ωc) was observed. Moreover, a shift in ωc to higher frequencies was also noticed with increasing content of grafted micro-biosurfactant. The formation of viscoelastic Pickering emulsion systems offered by the grafted MCC conjugates could be a valuable way for the realization of the soy protein-based ink. This helps to keep the printed shape upon extrusion force during 3D printing and reinforces the mechanical strength and shape fidelity.</p><!><p>From Figure 3d, the creep compliance levels of SGM3 and SGM4 inks were lower than those of SGM1 and SGM2 samples. As the creep-recovery data presented, the SGM3 and SGM4 showed a maximum creep compliance, which are about 33-fold (J(t) = 0.026 Pa–1) and 96-fold (J(t) =0.009 Pa–1) lower than the control ink (J(t) = 0.86 Pa–1), respectively. Compared to SGM3 and SGM4, the SGM1 (J(t) = 0.75 Pa–1) and SGM2 (J(t) = 0.48 Pa–1) inks presented a relatively weaker structure, caused by a larger peak strain level (Figure 3d). Therefore, the greater ratios of grafted MCC conjugate could result in a greater improvement of the elastic component of the viscoelasticity. This might be explained by the existence of higher amounts of hydrophobic/hydrophilic regions in Pickering emulsion gels.10 In this situation, a strong interaction could be developed between comparable groups in the adjacent droplets/particles.5</p><p>The recovery area of the creep test can be considered as the amount of decline in the compound deformation upon the removal of the applied stress.5,10 A material with higher elasticity and a more gel-like matrix offers a greater relative recovery.40 The recovery phase data revealed that the relative recovery property of soy protein-based ink was enhanced with increasing level of the grafted MCC conjugate. In this regard, the recovery percentages of SGM3 (∼70%) and SGM4 (∼75%) inks were much higher than the control ink (∼45%), hence signifying an appreciable reinforcement of the recovery properties of the ink system. A poor relative recovery, by contrast, was noticed for SGM1 and SGM2 with a recovery percentage of around 45 and 50%, respectively. This indicated weaker elasticity and mechanically a less stable structure. The creep-recovery data point to the incorporation of a higher level of micro-biosurfactant that formed a strengthened ink structure. According to the creep-recovery test, a more reversible network matrix in the reduced-fat inks was induced after introducing the micro-biosurfactant, which affects the elastic or viscous behavior of the viscoelastic properties of the soy-based printable ink.</p><!><p>The 3ITT measurement was used to evaluate if the Pickering emulsions are prone to rapid recovery upon being sheared at large deformations. As a criterion, when the peak viscosity of a material in the third interval recovers to at least 70% of its initial value, it has shown an ideal thixotropic behavior.40Figure 3e presents the changes in the viscosity of printable ink variants as a function of shear rate and time. Compared to control ink, the viscosity of reduced-fat inks, formulated by grafted MCC conjugates, presented greater values as may be expected. This signifies an improvement of the structural strength. Concerning the initial shearing interval, SGM3 and SGM4 inks displayed a considerably higher viscosity value compared to SGM1 and SGM2. The greater viscosity, attributed to a greater amount of micro-biosurfactant, can be ascribed to the more hydrophobic nature of MCC/GA/LAE conjugate, leading to stronger bonds holding particles together in the formed colloidal gel networks. Concerning the third stage, a lower viscosity was detected for all inks than that of the first stage. Regarding the 3ITT, the final viscosity transients expose the changes in the structural levels of the system once the shear rate is suddenly stepped up or down.10 Thus, a change in the viscosity of Pickering emulsion gel, after application of a stepwise shear rate, is related to its viscoelasticity and recovery compliance resulting from its structures.5 In the second interval (with a steady shear rate of 80 s–1), a marked reduction in viscosity was observed due to the breaking up of the high levels of interconnected structures in the Pickering emulsion gels (Figure 3e). Once shear was removed (i.e., the third interval), the reduced-fat inks showed a reversible restructuration toward their initial structure. As displayed in Figure 3e, the SGM3 and SGM4 inks presented an outstanding level of structural recovery, with initial viscosity recoveries were determined to be 76 and 84%, respectively. This follows the expected trend of structuring of the Pickering emulsion gel resulting from the micro-biosurfactant addition, thus allowing a higher resistance to the application of a rapid strain. In contrast, the lower extents of viscosity recoveries of SGM1 (38%) and SGM2 (44%) can be attributed to rather an irreversible structural failure. These showed an inferior mechanical strength owing to the existence of the less structured gel-like system and weakly connected networks.5,19 In accordance with these data, the steady and oscillatory rheological tests revealed that a less structured ink showed a weaker elastic property. Therefore, an excellent dynamic network with a desirable reversible structure was fabricated in the Pickering emulsion inks, especially those with a higher ratio of grafted micro-biosurfactant.</p><!><p>To perform an effective 3D printing process, a high level of pseudoplasticity and viscoelasticity offers ink high printability.3,10 Under this condition, the inks can easily be extruded out during the 3D printing process. Furthermore, a more reversible ink having smaller droplets/particles, which are uniformly dispersed in the emulsion, can offer an effective 3D printing process concerning the shape stability and resolution of the printed structures.3 In the present study, the results obtained above already prove that the SGM3 and SGM4 Pickering emulsions can be characterized as forming superior ink in terms of their rheological and structural properties. In these cases, they showed a higher degree of shear-thinning behavior, greater elastic modulus, and a strong thixotropic feature with lower creep compliance. Moreover, as outlined in the Supporting Information, and specifically in Section S.3.1, the SGM3 and SGM4 inks also presented the smallest particle size (also see Figure 3f) with a monomodal particle size distribution (also see Figure 3g) and higher ζ-potential (also see Figure 3h), denoting a more structured Pickering emulsion.</p><p>Figure 4 illustrates the photographs of 3D-printed soy protein-based architectures containing different ratios of grafted MCC conjugates, which were printed in a layer-by-layer manner. The 3D structures involving SGM4 ink, with 5.6 wt % grafted micro-biosurfactant, preserved their shape during the printing process, while also resulting in a fine resolution with a smooth surface (Figure 4, fifth column). The shape produced with SGM3 ink, having 4.2 wt % grafted micro-biosurfactant, was printed without failure upon the printing process. However, after the printing process, the supporting structure in the underside layer was weakened and somewhat deformed (Figure 4, fourth column). The SGM2 ink, involving 2.8 wt % grafted micro-biosurfactant, was moderately easily extruded out from the nozzle but deformed without keeping the structure at the curved area of the designed shape (Figure 4, third column). Finally, the control (with no grafted micro-biosurfactant) (Figure 4, first column) and SGM1 (with only 1.4 wt % grafted micro-biosurfactant) (Figure 4, second column) were stacked upon printing and then crumbled in the middle of the process. These latter two inks could not provide sufficient support, and the stacking became untenable in the upper portion of the design. As a consequence, it was confirmed that the increasing level of micro-biosurfactant could enhance the layer resolution, leading to a high geometrical accuracy. An increase in the content of grafted MCC/GA/LAE reinforced the mechanical strength of the inks, where the emulsion gels showed a higher structural strength of the internal linkages. This enhances the spatial resolution of the resulting 3D-printed objects.3,6,10</p><!><p>Printing performance photographs of different 3D-printed structures. The scale bar is 5 cm.</p><!><p>To assess the structural strength and printing precision, the layer number (assigned to the structural strength) and line width (related to the printing accuracy) of 3D-printed architectures were evaluated (Supporting Information Table S4). The layer numbers and line widths of the printed control, SGM1, and SGM2 were statistically similar to each other (P > 0.05). On the contrary, the printed SGM3 and SGM4 objects offered the 3D structures higher layer numbers and thinner line widths, representing a desired structural strength and well-defined shape. The manufacture of 3D constructs with a precisely controlled structure caused by the inclusion of higher micro-biosurfactant ratios could be elucidated due to the development of a structured matrix with a strong elastic matrix and reinforced connected network. As viscoelastic data show, the SGM3 and SGM4 inks provided a higher value of the elastic property with enhanced structural recovery. This could offer the Pickering emulsion gel resistance against the rapid shear stress and deformation during the 3D printing process.</p><!><p>Figure 5 shows the VP-SEM microstructure of the printed objects developed with different levels of grafted micro-biosurfactant. The microstructure of printed control seemed to be rugged and irregular with noticeable agglomerated pieces. Besides, there is no obvious pore structure inside its matrix. As Figure 5 depicts, the incorporation of 1.4 wt % grafted micro-biosurfactant could not induce a positive effect on the microstructural properties. This means that the microstructure of printed SGM1 had some level of unevenness with the existence of the aggregated particles on the surface. Moreover, its structure seemed to be denser than that of other reduced-fat printed objects. Regarding printed SGM2, there is the appearance of a limited number of pores with small size and more spherical shape. The VP-SEM micrograph also presented pores in the printed SGM2 randomly dispersed throughout the matrix. In the case of printed SGM3 and SGM4, there was a large number of pores with smaller sizes that were regularly distributed within the matrix. In this context, the highest ratio of grafted MCC conjugate, i.e., printed SGM4, could enhance the porosity with an ordered pore, while this was not the case for printed SGM3. Then, the changes in the printing performance and structural stability of 3D structures could be due to differences in their microstructure and therefore the properties.6</p><!><p>VP-SEM photomicrographs of different meat analogues variants.</p><!><p>To accomplish the temporal perceptions of different sensory attributes, the TDS technique is considered an effective tool to compare the sensations perceived simultaneously in a complex food product.27,28Figure 6 shows the TDS curves of 3D-printed meat analogue variants, where the significance and chance levels are specified. Based on a binomial distribution and in view of 30 evaluations, the chance (20%) and significance (34%) levels, obtained from six traits, were determined. Figure 6 displays that the graininess trait was a dominant attribute in the 3D-printed control, SGM1, and SGM2 with a maximum dominance rate (max. DR%) of 63.5, 60.1, and 50.9%, respectively. This trait prevailed during the whole period of the sensory evaluation (P < 0.05). Based on the VP-SEM experiment, a high level of irregularity with some aggregated micro-sized particles (except SGM2) were detected on the surface of these printed samples. Contrary, the grainy texture in the 3D-printed SGM3 and SGM4 was not perceived as significantly dominant at any time of the sensory assessment (P > 0.05). An even and homogeneous structure could be possibly developed in these samples as revealed by microstructure investigations, which showed a smaller pore size distribution with an extremely porous matrix.</p><!><p>Temporal profile of dominant sensations in standardized time with specific attributes in meat analogue samples.</p><!><p>In the case of printed SGM3 and SGM4, the chewiness and firmness traits were also dominant at the middle of the consumption time. In the temporal profile of printed SGM2, the chewiness attribute also created a trivial peak, albeit significant, at the middle of TDS evaluation (P < 0.05). By referring to the instrumental texture measurement, the printed SGM3 and SGM4 meat analogues proposed a firmer matrix compared to 3D-printed SGM1 and SGM2 (Supporting Information Section S.5.1). These reduced-fat printed samples presented a strong gel-like matrix; thus, they rationally needed a greater force required to chew their matrix. Similarly, the 3D-printed SGM3 and SGM4 meat analogues showed a juicy attribute, especially at the middle of the mastication period, with greater dominance regarding printed SGM4 (max. DR = 45.6%). A promising sensory result was detected for the fibrous attribute, where the printed SGM3 and SGM4 showed the greater dominance of fibrousness with a max. DR of 46.8 and 51.4%, respectively. However, the fibrous sensation in the printed SGM1 and SGM2 was not perceived as significantly dominant throughout the consumption time (P > 0.05).</p><p>In summary, the newly developed bioactive soy-based Pickering emulsion gel, stabilized by multifunctional microcrystalline cellulose, established feasibility and efficiency for applications in 3D food printing. As gallic acid is well known for its antioxidant ability and lauric arginate shows an interfacial stabilizing effect with excellent antimicrobial activity, we attempted to improve the interfacial activity of microcrystalline cellulose via interfering with intra- and intermolecular hydrogen linkages and hydrophobic interaction and also reinforced its antioxidant and antimicrobial activities through grafting of gallic acid and lauric arginate onto the surface of microcrystalline cellulose. After introducing multifunctional microcrystalline cellulose to a soy-based dispersion, a Pickering emulsion gel showing pseudoplastic, viscoelastic, and thixotropic properties was obtained, having a monomodal particle size distribution. The soy protein-based Pickering ink was also effectively processed through an extrusion-type printing system to produce a fibrous meat analogue with a high degree of shape fidelity. The printing performance results, obtained from the inks containing the grafted microcrystalline cellulose, offered the enhancement of layer resolution with a high geometrical precision. A promising result regarding the sensorial properties of the 3D-printed meat analogues was the evidence of a fibrous sensation, which was confirmed by dynamic sensory evaluation. Such introductory results in 3D printing showed how this technique could further generate plant-based meat with the desired texture for enhanced eating experiences. The novel applications in food texture modification to potential fabrication of printable emulsions with a flexible interfacial behavior are promising viewpoints. Considering the obtained results, the food-grade particulate-type inks are not only attractive from an academic look but would have an excessive effect on the construction of suitable emulsion systems in industrial applications.</p><!><p>13C NMR (Section S.1); crystallinity index obtained by 13C NMR (Section S.1.1); transmission electron microscopy (TEM) (Section S.2); measurement of droplet size and electrical charges (Section S.3); droplet size and ζ-potential (Section S.3.1); confocal laser scanning microscopy (CLSM) (Section S.4); textural properties (Section S.5); textural evaluation (Section S.5.1); toxicity (Section S.6); printing settings expressed as Slic3r terms (http://slic3r.com) (Table S1); references used to acquaint the panelists with each attribute (Table S2); obtained consistency index, flow behavior index, and yield stress of soy protein-based ink variants (Table S3); and summary of printing performance results and textural properties of 3D-printed meat analogues (Table S4) (PDF)</p><p>jf1c05644_si_001.pdf</p><!><p>The manuscript was written through the contributions of all authors. All authors have given approval to the final version of the manuscript.</p><!><p>Open access funding provided by the University of Natural Resources and Life Sciences Vienna (BOKU).</p><p>The authors declare no competing financial interest.</p>

PubMed Open Access

Regulation of neuronal bioenergy homeostasis by glutamate

Bioenergy homeostasis is crucial in maintaining normal cell function and survival and it is thus important to understand cellular mechanisms underlying its regulation. Neurons use a large amount of ATP to maintain membrane potential and synaptic communication, making the brain the most energy consuming organ in the body. Glutamate mediates a large majority of synaptic transmission which is responsible for the expression of neural plasticity and higher brain functions. Most of the energy cost is attributable to the glutamatergic system; under pathological conditions such as stroke and brain ischemia, neural energy depletion is accompanied by a massive release of glutamate. However, the specific cellular processes implicated in glutamate-dependent bioenergy dynamics are not well understood. We find that glutamate induces a rapid and dramatic reduction of ATP levels in neurons, through reduced ATP genesis and elevated consumption. ATP reduction depends on NMDA receptor activity, but is not a result of neuronal firing, gap junction-mediated leaking or intracellular signaling. Similar changes in ATP levels are also induced by synaptic glutamate accumulation following suppression of glutamate transporter activity. Furthermore, the glutamate-induced ATP down-regulation is blocked by the sodium pump inhibitor ouabain, suggesting the sodium pump as the primary energy consumer during glutamate stimulation. These data suggest the important role of glutamate in the control of cellular ATP homeostasis.

regulation_of_neuronal_bioenergy_homeostasis_by_glutamate

1. Introduction<!>2.1. Primary cortical neuron culture<!>2.2. Glia culture<!>2.3. ATP assays<!>2.4. Western blotting<!>2.5. Neuronal cell death detection<!>2.6. Data analysis and statistics<!>3.1. Glutamate induces rapid reduction in neuronal ATP<!>3.2. Glutamate regulates ATP production and utilization independent of neuronal firing<!>3.3. Involvement of glutamate receptors and signaling cascades in glutamate-induced ATP reduction<!>3.4. Synaptic glutamate regulates ATP in a receptor activity-dependent manner<!>3.5. Sodium pump is involved in glutamate-induced ATP reduction<!>3.6. Glutamate incubation does not cause ATP leaking via gap junction<!>3.7. Glia are not the major target in glutamate-induced ATP regulation<!>4. Discussion

<p>The brain is the most energy consuming organ in our body. It accounts for 2% of our body weight, but uses more than 20% of the oxygen and ATP supply (Magistretti and Pellerin 1997; Magistretti et al. 1999; Attwell and Laughlin 2001; Raichle and Gusnard 2002; Rao et al. 2006). Therefore, neurons are extremely sensitive to bioenergy homeostasis. Within five minutes, the brain will cease to function if the oxygen supply stops. In contrast to peripheral tissues, neurons depend almost solely on glucose for ATP production (Attwell and Laughlin 2001). Among many neural processes, the excitatory glutamatergic system uses the most energy (Sibson et al. 1998; Shen et al. 1999; Raichle and Gusnard 2002), most probably due to postsynaptic events (Jolivet et al. 2009). Neuronal activity, manifested by fluctuation in membrane potential, represents a major ATP-consuming process. Other glutamate-related events, including glutamate synthesis, vesicle filling and release, transporter uptake and recycling, as well as receptor trafficking and signaling, are also energy consuming.</p><p>In the brain, the glutamatergic system is responsible for the majority of excitatory synaptic activity. In presynaptic terminals, glutamate is enriched in synaptic vesicles, powered indirectly by a proton pump on the vesicle membrane, at a concentration of 100 mM. During synaptic transmission, the release of a single vesicle can cause a rapid rise of glutamate in the cleft. At its peak, it can reach 1 mM (Danbolt 2001). Under normal conditions, ambient glutamate at the extracellular environment is maintained by the constant activity of glutamate transporters at the plasma membrane of both neurons and glia (Danbolt 2001; Tzingounis and Wadiche 2007). Transporters at the glia, which often surround synapses to ensure efficient uptake of released transmitter and to prevent glutamate spillover, are believed to play a major role in ensuring minimal ambient glutamate.</p><p>Glutamate functions through its specific receptors, which include ionotropic AMPA (AMPAR), NMDA (NMDAR) and kainate receptors (KR), as well as metabolic receptors (mGluRs) (Man et al. 2000; Collingridge et al. 2004; Newpher and Ehlers 2008). AMPARs are the major components responsible for synaptic transmission, whereas NMDARs play an essential role in the formation of synaptic plasticity, mainly via regulation of AMPAR trafficking and synaptic localization. More importantly, the high permeability of NMDAR to calcium enables the receptor to initiate a series of calcium-dependent signaling cascades, including energy-dependent protein modification and metabolic regulations (Kim et al. 2005). mGluRs are G-protein coupled receptors that, upon glutamate binding, regulate neuronal functions via PKA, PLC and PKC signaling cascades (Anwyl 2009; Gladding et al. 2009; Olive 2009).</p><p>Under normal conditions, high glutamate concentrations only occur at the synaptic cleft. The ambient glutamate concentration is maintained at very low levels (Herman and Jahr 2007). However, during traumatic brain injuries or stroke, massive glutamate release can lead to a marked increase in extracellular glutamate and hyperactivity of the overall glutamate system, causing additional acute and delayed neural pathology. A reduction in ATP is one of the first cellular abnormalities in stroke. Lack of energy will eventually cause the failure of the sodium pump, leading to a collapse of ion/electrical gradients and a complete loss of neuronal function. In addition, dysfunction in energy metabolism has been implicated in multiple neurological disorders including Alzheimer's disease, Parkinson's disease and Huntington disease (Ferreira et al.; Mochel et al.; Young-Collier et al.; Blass et al. 1988; Wallace 1994; Mattson et al. 1999; Parihar and Brewer 2007; Amato and Man). However, the cellular mechanisms by which glutamate negatively regulates ATP homeostasis remain less clear.</p><p>To this end, we examined the effect of glutamate application on cellular energy status in cultured cortical neurons. We find that glutamate triggers a rapid, significant reduction in neuronal ATP levels. The glutamate effect on ATP is dependent on the activity of NMDAR and the sodium pump, and can be triggered by synaptic glutamate accumulation. In contrast, neither firing of action potentials nor gap junction-mediated leaking is responsible for the ATP reduction. These findings suggest a process in which sodium influx via glutamate receptor channels stimulates sodium pump activity, leading to elevated energy consumption and ATP reduction in neurons.</p><!><p>Primary neuronal cultures were prepared as described earlier (Zhang et al. 2009; Lin et al. 2011; Jarzylo and Man 2012). Briefly, rat brain cortices from E18 rat embryos were digested with papain (0.1 mg/ml in HBSS, 37°C for 20 min), washed and triturated with a serological pipette. To ensure high-quality cell adhesion and growth, 6-well plates were coated with poly-L-lysine (Sigma, 0.1 mg/ml) overnight and washed before being kept in plating medium (MEM containing 10% fetal bovine serum (FBS), 5% horse serum, 31 mg cystine, 5 mM glutamax and 1% P/S) until cell plating. Neurons were counted and plated onto 6-well dishes (1×106 per well). 24 hrs after plating, the culture medium was replaced with feeding medium (Neurobasal medium supplemented with 1% HS, 2% B-27, 2 mM glutamax and 1% P/S). Thereafter, cortical neurons were fed twice a week with 2 ml feeding medium/dish until experiments. To keep a low level of glial cells in the culture, the mitotic inhibitor 5-FDU (5 μM) was added to DIV 7 cultures to inhibit glial cell proliferation.</p><!><p>Cells from rat cortices were prepared as above, except that medium containing high serum (15%) was used for plating. Neurons do not survive high serum concentrations and die out gradually. Glia cultures were used for assays when cell confluency reached above 80%.</p><!><p>2 wk-old cultured cortical neurons grown in 6-well plates were used for ATP assays. Cells were treated with glutamate alone, or together with other drugs for 1 hr. On ice, culture medium was removed, and cells were washed 3 times with ACSF containing (in mM) 140 NaCl, 3 KCl, 1.5 MgCl2, 2.5 CaCl2, 11 Glucose, 10 HEPES, pH 7.4. After washing, 400 μl HEPES (20 mM in H2O) lysis buffer was added to each well, and cells were collected after thoroughly scraping the culture at the well bottom. Cell lysates were further triturated 5–6 times by repeated pippetting, boiled at 95°C for 10 min, sonicated in cold room for 2 s and followed by ATP assays using a Luminometer according to manufacturer's instructions (ATP Determination Kit, Invitrogen).</p><!><p>Cultures were rinsed with cold PBS and resuspended in 200 μl modified RIPA lysis buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 1% NP40, 1% SDOC and 0.1% SDS) containing mini cOmplete protease inhibitor (Roche). Lysates were further solubilized by sonication and 10 min incubation on ice followed by centrifugation for 10 min at 13,000 × g to remove insolubilities. Supernatants were mixed with 2x sample buffer and denatured on a 95°C heat block for 10 min. Lysate samples were separated by SDS-PAGE and transferred to PVDF membranes. The blot was then blocked at room temperature with 5% nonfat dry milk in TBST for one hr and incubated with primary antibody diluted in TBST overnight at 4°C, followed by incubation with HRP-conjugated secondary antibodies. The blot was developed using enhanced chemiluminescence (ECL) detection methods (Amersham).</p><!><p>Cultured cortical neurons on coverslips were treated with 50 μM glutamate for 60 min at 37°C. Some treated cells were then washed and incubated at 37°C in normal drug-free medium for an additional 3 hr for recovery in order to detect possible delayed cell death. During the final 20 min of 1 hr glutamate treatment or of the 3 hr recovery, respectively, propidium iodide (1 μM) and Hoechst (1 μM) were added to the medium. After incubation, cells were washed 3 times with cold ACSF on the rocker. Following fixation, cells were mounted for cell death examination. Cells demonstrating nuclear condensation and positive PI staining were counted as cells undergoing cell death.</p><!><p>All values were reported as mean ± SEM. Statistical analysis was performed using the two-population student's t test. N indicates the number of independent experiments. For westerns, the film was scanned and the optical intensities of the protein bands were quantified using NIH Image-J software.</p><!><p>To examine the effect of glutamatergic activation on neuronal bioenergy homeostasis, 2 wk-old primary cortical neurons were incubated with glutamate (50 μM) in culture medium for varied periods of time. Neurons were then rinsed with ACSF, lysed in HEPES buffer and boiled at 95°C for 10 min. Equal amount of lysates were used for the luciferase-based ATP assay. As a control, a standard curve showed a linear relationship between ATP concentration and fluorescence intensity (Figure 1A). A 10 min glutamate treatment caused a 30% reduction in ATP level (0.74 ± 0.04, n=3). At 60 min, the ATP level was further reduced to 60% of control (0.59 ± 0.04, n=3) (Figure 1B). Next, we examined the dose response of glutamate incubation. 10 μM glutamate induced a modest, but significant ATP reduction (0.85 ± 0.04, n=3). A more drastic effect was obtained by a higher concentration, but the effect reached a plateau at 30–50 μM (Figure 1C). For the rest of the experiments, 50 μM glutamate for 60 min was adopted as a standard treatment protocol.</p><!><p>The reduction in ATP could result from suppression in energy production or enhancement in ATP consumption, or both. To test these possibilities, we used potassium cyanide (KCN) to block mitochondrial oxidative phosphorylation and ATP genesis. Cultured neurons were incubated with KCN for 15 min, then supplemented with glutamate for 60 min. Indeed, KCN alone in the same incubation time period (75 min) caused a drastic reduction in ATP (0.41 ± 0.02, n=5). When ATP synthesis was blocked by KCN, the addition of glutamate was able to induce a further reduction in ATP levels (0.25 ± 0.002, n=3) (Figure 2A), indicating enhanced ATP utilization by glutamatergic activation. However, glutamate caused only a 16% further reduction from that of KCN alone, much less than the total effect of glutamate alone (about 50%). Therefore, the glutamate effect likely resulted from a combination of inhibited production and facilitated consumption of cellular ATP.</p><p>Application of glutamate should cause membrane depolarization and an increase in neuronal firing of action potentials, leading to elevated levels of network activity. To test whether the drop of ATP is caused by higher neuronal activation, cultured neurons were treated with KCl (40 mM) to trigger depolarization and firing. As expected, a 1 hr KCl incubation induced a significant drop in ATP. To further confirm the involvement of firing in ATP consumption, we applied tetrodotoxin (TTX, 1 μM) to block sodium channels and action potentials. Interestingly, no changes in the ATP level were detected by applying TTX alone, probably due to low levels of basal activity. Indeed, in our cultures, patch-clamp recording revealed that 2 wk-old neurons spike at a low frequency of about 0.5 Hz at resting condition (Hou et al. 2011). Consistent with the role of firing, we found that KCl-induced reduction in ATP abundance was completely abolished by TTX (KCl, 0.78 ± 0.04, n=3; KCl + TTX, 0.95 ± 0.06, n=3) (Figure 2B), confirming the contribution of cell firing in the KCl effect. By contrast, we found that in the presence of TTX, the glutamate treatment remained able to reduce ATP levels (Glutamate 0.46 ±0.02, n=3; glutamate plus TTX 0.51 ± 0.04, n=3) (Figure 2B), excluding cell firing as a major factor in the glutamate effect. This suggests that other cellular processes than neuronal electrical activity are responsible for the observed glutamate-dependent ATP depletion.</p><!><p>Glutamate affects neuronal activity via its receptors including ionotropic AMPARs and NMDARs, as well as metabotropic mGluRs. To examine the involvement of glutamate receptor activity, we mixed CNQX (20 μM), APV (50 μM) and MCPG (500 μM) to form an antagonist cocktail (AC) to block all glutamate receptors. In the presence of AC, glutamate failed to induce a change in ATP levels, indicating a requirement of glutamate receptor activation (Figure 3A). To dissect the receptor subtype(s) that mediates glutamate effect, a receptor antagonist was applied individually to selectively block AMPAR/KR (CNQX), NMDAR (APV) and mGluRs (MCPG), respectively. Glutamate effect on ATP was blocked by APV, but not by CNQX or MCPG (Figure 3A), indicating an important role of NMDARs. Consistently, when neurons were incubated with a specific agonist to activate individual receptor subtypes, we found that only NMDA, but neither AMPA nor DHPG, induced a reduction in ATP (Figure 3B). These data strongly indicate that the glutamate effect is mediated primarily via NMDARs.</p><p>Next, we wanted to confirm that the ATP reduction is not a result of non-specific deleterious effects of glutamate on cell conditions, causing a failure of ATP synthesis. First, we tested whether glutamate could induce ATP reduction in other cell types. We found that when ATP was measured in HEK cells, glutamate treatment did not show any effect (Glutamate 1.12 ±0.05, n=3; glutamate plus AC 0.95 ± 0.12, n=3) (Figure 3C), indicating neuronal specificity in glutamatergic regulation of ATP homeostasis. Second, we performed cell death assays in glutamate treated neuronal cultures. Hoechst and propidium iodide (PI) dyes were added to the medium immediately after glutamate treatment, or after 1 hr recovery following glutamate treatment. Nucleus condensation and fragmentation in PI staining, as well as positive fluorescence signals in PI staining would indicate cell death. Examination showed a minimal level of cell death in non-treated controls (0%, n=850 cells). We found no significant change in cell death immediately after 1 hr glutamate treatment (0%, n=861 cells), and only a modest increase in cell death after 1 hr recovery (9.2%, n=1201 cells). In contrast, as a positive control, a typical cell death protocol (100 μM glutamate for 1 hr, and 5 hr recovery) induced a drastic rate of cell death (42%, n=935 cells). Thus, alterations in cell health condition did not seem to contribute significantly to the glutamate effect on ATP.</p><p>We then examined the involvement of multiple major signaling pathways related to glutamate activation including protein kinase A (PKA), PI3-kinase (PI3K), calcium/calmodulin-dependent kinase II (CaMKII), and tyrosine kinase. In the presence of antagonists specific to each of the individual kinases, we found that glutamate induced a reduction in ATP to levels comparable to glutamate alone (Figure 4), excluding these signaling molecules from a role in bioenergy regulation.</p><!><p>Application of glutamate in the medium should activate all receptors at the cell surface, including those localized at synaptic and non-synaptic sites. We wondered whether an increase in glutamate selectively at the synaptic sites also regulates cellular energy homeostasis. At basal conditions, glutamate release can be triggered by action potentials, or can occur spontaneously. The presynaptically released glutamate is rapidly removed by diffusion and activities of excitatory amino acid transporters (EAATs) (Gegelashvili et al. 2000; Danbolt 2001; Nieoullon et al. 2006; Tzingounis and Wadiche 2007). Inhibition of EAATs slows down transmitter clearance, leading to glutamate synaptic accumulation, AMPAR internalization and degradation and alterations in synaptic transmission (Danbolt 2001; Tzingounis and Wadiche 2007). To examine the role of synaptic glutamate in ATP status, we incubated neurons with TBOA, a non-transportable glutamate transporter inhibitor. Following 1 hr TBOA treatment (100 μM), the ATP assays revealed a dramatic reduction in ATP amount. Similar to the effect of glutamate application, the TBOA effect on ATP abundance was completely blocked by the glutamate receptor antagonist cocktail (TBOA 0.62 ± 0.05, n=3; TBOA plus AC 0.94 ± 0.01, n=3) (Figure 5). Interestingly, glutamate application in the presence of TBOA resulted in a drop in ATP to a level similar to that of glutamate alone or TBOA alone (Glutamate 0.50 ± 0.05, n=2; TBOA 0.62 ± 0.05, n=3; TBOA plus glutamate 0.57 ± 0.04, n=3) (Figure 5), indicating a saturation and/or shared processes on ATP regulation by synaptic vs. global glutamate stimulation.</p><!><p>During glutamate application, the activity of glutamate receptors, transporters and voltage-gated sodium channels permit a large amount of sodium into neurons. The maintenance of ion gradients, which are necessary for the proper functioning of neurons, depends on the activity of the Na+/K+-ATPase (NKA), or sodium pump. For each cycle, NKA takes in 2 K+ and brings out 3 Na+, powered by the hydrolysis of ATP (Hernandez 1992; Schoner and Scheiner-Bobis 2007). Through intracellular sodium, sodium pumps may be involved in other physiological functions including synaptic transmission, and channel and receptor expression (Nanou et al. 2008; Zhang et al. 2009, Desfrere, 2009 #3681). Therefore, glutamate treatment should stimulate the NKA, leading to an enhanced consumption of ATP. To test this hypothesis, neurons were incubated with glutamate in the presence of NKA inhibitor ouabain (50 μM). Consistent with energy consumption under basal conditions, ouabain caused a modest, but significant increase in ATP amount (Figure 6). Importantly, in the presence of ouabain, glutamate incubation failed to induce a decrease in ATP abundance (Glutamate 1.0 ± 0.01; Glutamate 0.69 ± 0.03; Ouabain 1.35 ± 0.08; Ouabain + glutamate 1.26 ± 0.01; n=4) (Figure 6). This data indicates that NKA activity is essential in glutamate-dependent down-regulation of ATP homeostasis.</p><!><p>Gap junctions are molecular tunnels that directly connect neighboring cells, through which ions and small molecules with a molecular weight below 1000 Da are allowed to be exchanged (Zoidl et al. 2008; Dobrowolski and Willecke 2009). Gap junctions exist in glia (Seifert and Steinhauser; Giaume and Venance 1998), and to a lesser extent, in neurons (Bennett and Zukin 2004; Connors and Long 2004; Hughes and Crunelli 2006). Studies have shown that cellular ATP can be released from glia via gap junctions (Garre et al.; Orellana et al.; Frenguelli et al. 2007; Blum et al. 2008; Gourine et al. 2010). It is thus possible that glutamate stimulation opens the gap junction channels as a consequence of membrane depolarization or signaling cascades, causing diffusion of intracellular ATP into the extracellular solution. To this end, we treated cultured neurons with the gap junction inhibitor carbenoxolene (100 μM) for 1 hr. With the blockade of the gap junctions, glutamate still caused a significant decrease in ATP amount (Glutamate 0.55 ± 0.04; Carb 1.17 ± 0.08; Glutamate+Carb 0.64 ± 0.01; n=3 each) (Figure 7). We then measured ATP levels in the medium of 2 wk-old cultured neurons under basal conditions and following glutamate treatment. The ATP assay revealed that in control culture medium, only background levels of signal were detected, and no difference was found following glutamate treatment (data not shown). These results indicate that gap junctions are not implicated in ATP reduction during glutamate incubation.</p><!><p>The primary cell cultures used in this study contain both neurons and glia, in order to ensure normal maturation of neuronal connections (Christopherson et al. 2005). In the brain, glia play a critical role in energy metabolism, which accounts for approximately 30% of oxygen consumption in brain cortext (Dienel and Cruz 2004; Hertz et al. 2007; Hertz 2011). In this study, it is not clear whether the glutamate-induced ATP reduction occurs only in neurons or in both cell types. Given that glutamate transporters and glutamate receptors are expressed in glia (Zhou et al.; Conti et al. 1996; Verkhratsky and Steinhauser 2000; Ge et al. 2006; De Biase et al. 2011), glutamatergic stimulation could be detected by glia to regulate their energetic metabolism (Prebil et al.). To examine the possible involvement of glia in glutamate-dependent bioenergy regulation, we eliminated neurons from the culture by incubating the newly-plated cells with medium containing high serum, resulting in cultures of pure glia. Western blots showed that the regular mixed cultures of neuron and glia expressed low levels of the glial marker protein GFAP (glial fibrillary acidic protein) and high levels of AMPARs, indicating a dominant amount of neuronal cells. In contrast, glial cultures expressed much higher GFAP, but only a negligible amount of AMPARs (Figure 8A). This result indicated an efficient elimination of neurons from the culture. Using this glia culture, we found that glutamate stimulation remained able to cause a significant reduction in ATP (Glutamate 0.78 ± 0.03, n=3) (Figure 8B). However, compared to the effect on normal mixed culture (up to 50% ATP reduction), the effect on glia was markedly weaker (20% ATP reduction). Further, since glia growth was routinely inhibited with FDU in our regular neuronal culture, the relative proportion of total cellular ATP in glia was much lower than that of neurons. These results suggest that neurons, rather than glia, are the major target of the glutamate effect in terms of ATP regulation. Similar to neuronal culture, the glutamate effect was blocked by NMDAR antagonist APV (Glutamate+CNQX 0.80 ± 0.02, n=3; Glutamate+APV 1.26 ± 0.04, n=3), but not affected by AMPAR antagonist CNQX (Figure 8B), indicating that in glia, glutamate-caused ATP reduction is also mediated by NMDARs.</p><!><p>In this study, we found that neuronal bioenergy levels are controlled by glutamatergic activity. Application of glutamate induces a rapid and dramatic reduction in cellular ATP abundance. This regulation is independent of neuronal firing, and is not a result of gap junction-mediated ATP transmembrane diffusion. We find that glutamate regulates ATP in both neuronal and glial cells. Neurons produce and utilize most of the ATP in the brain (Magistretti et al. 1999; Pellerin and Magistretti 2003), but a more active role for glia in brain energetics has been increasingly appreciated (Hertz et al. 2007; Hertz 2011). In our study, since glial cells were maintained at a minimal level by routine application of the mitotic inhibitor 5-FDU in the medium, the observed ATP reduction by glutamate incubation seems mainly attributable to the neuronal component. In addition, the effect of glutamate on bioenergy homeostasis is not a non-specific side effect since glutamate treatment does not cause significant cell death, and incubation of HEK cells with the same concentration of glutamate does not change ATP levels.</p><p>There are three types of ionotropic glutamate receptors including AMPARs, NMDARs and KRs. Activation of these receptors will lead to membrane depolarization, firing of action potentials and thus elevated neuronal network activity, which are all energy consuming processes. Glutamate can also stimulate metabotropic mGluRs, triggering intracellular signaling pathways. The glutamate induced ATP reduction in neurons as well as in glia was blocked only by NMDAR antagonist APV, indicating that the glutamate effect is mediated solely by NMDARs.</p><p>In addition to receptors, glutamate also binds to its transporters for removal. To date, five EAATs (EAAT1–5) have been identified in glia and neurons. The glial transporters EAAT1–2 are primarily localized to the plasma membrane of specialized domains in astrocytic processes that wrap the synapse and are believed to be the major players in glutamate removal during synaptic transmission (Rothstein et al. 1994; Chaudhry et al. 1995). The distribution of neuronal transporters shows cell type specificity. EAAT3 is expressed in most neurons, including hippocampal and cortical neurons, whereas EAAT4 is mainly localized in cerebellar Purkinje cells and EAAT5 only in retinal ganglion neurons (Rothstein et al. 1994). Given spontaneous neuronal activation and synaptic activity, EAAT suppression leads to glutamate accumulation in the synaptic cleft. We found that blockade of EAATs results in a decrease in ATP amount, which can be completely blocked by the glutamate receptor antagonist cocktail, indicating that local glutamate stimulation at synaptic sites causes similar ATP regulation as global glutamate application. Since the transporter blockade can cause glutamate spillover and activation of parasynaptic NR2B-containing NMDARs, the TBOA effect could also result from the stimulation of non-synaptic NMDARs (Tzingounis and Wadiche 2007; Scimemi et al. 2009; Chalifoux and Carter 2011; Jarzylo and Man 2012). Also, TBOA and glutamate effects are not additive, because in the presence of EAAT inhibitors, glutamate application fails to induce further reduction in ATP, suggesting the utilization of share cellular processes.</p><p>The sodium pump is a major energy user in neurons. During neuronal activity, a large amount of sodium fluxes into the cell via multiple routes, but mainly through glutamate receptors and sodium channels. At steady states, the intracellular sodium concentration is about 10 mM. In hippocampal neurons one action potential can increase spine sodium to 35–40 mM, and a typical protocol for the induction of long-term potentiation (100 Hz stimulation for 1 s) increases sodium in the spine to more than 100 mM (Rose and Konnerth 2001). The frequent rises of cellular sodium must be efficiently rebalanced by the activity of the sodium pump, accompanied by a large amount of ATP consumption. We found that NKA suppression abolished glutamate-induced ATP reduction, supporting a hypothesis that glutamate-dependent sodium rises lead to elevated sodium pump activity and unbalanced ATP consumption. Sodium influxes mainly via ionotropic glutamate receptors and voltage-gated sodium channels. However, blockage of sodium channels by TTX had no effect on ATP reduction, indicating glutamate receptors, especially the NMDARs, are the primary source of intracellular sodium. Given that AMPARs contribute the most in synaptic currents, it seems a surprise that NMDARs, rather than AMPARs, are responsible for presumably sodium/sodium pump-dependent ATP reduction. Although NMDARs show high permeability to calcium and are often mistakenly considered a calcium channel, more than 80% of NMDA currents are actually carried also by sodium (Skeberdis et al. 2006). Further, due to its long-lasting time course, albeit with a lower amplitude compared to that of AMPAR, NMDA current can confer a large amount of sodium during synaptic activities. Alternatively, the requirement of NMDARs may be, at least partially, due to calcium-dependent regulatory events.</p><p>Glucose is the sole source for ATP production in neurons (Attwell and Laughlin 2001). Therefore, glutamate might alter neuronal energy balance through an inhibition of glucose uptake. It has been shown that glutamate treatment leads to an increase in glucose uptake in glia (Loaiza et al. 2003), but a decrease in neurons (Porras et al. 2004). However, the reduction of neuronal glucose uptake is AMPAR-dependent, whereas the glutamate-induced ATP reduction is independent of glutamate receptor activity, suggesting a negligible role for changes in the cellular supply of glucose.</p><p>Energy depletion has been implicated in glutamate-induced neurotoxicity (Baltan et al.; Nicholls and Budd 1998; Del Rio et al. 2007; Nicholls et al. 2007) and neurological disorders (Ferreira et al.; Mochel et al.; Blass et al. 1988; Wallace 1994; Mattson et al. 1999; Parihar and Brewer 2007; Amato and Man). Glutamate stimulation causes more severe cell death when cellular energy homeostasis is impaired (Del Rio et al. 2007). In the brain, massive exposure to glutamate often occurs during neural trauma and stroke. Under these circumstances, a drop in cellular energy supply is one of the first pathological events. A lack of sufficient ATP undermines a large number of energy-dependent cellular processes including kinase/enzymatic activity, proteasomal protein turnover, and the operation of the sodium pump, all leading to a collapse of cellular functional integrity and deterioration of cell conditions. In stroke, the decrease in energy level is believed to be the consequence of a ceased supply of oxygen and glucose; the role of glutamate in the control of bioenergy homeostasis has not been fully appreciated. Findings of current work suggest that a massive release of glutamate during the early stages of stroke is a crucial factor leading to energy disturbance. Therefore, early intervention in glutamate accumulation should be important in preventing a disruption in bioenergy homeostasis and subsequent cell death.</p>

PubMed Author Manuscript

Potential SARS-CoV-2 Nonstructural Protein 15 (NSP15) Inhibitors: Repurposing FDA-Approved Drugs

Purpose: Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has caused millions of deaths worldwide, pushing the urgent need for an efficient treatment. Nonstructural protein 15 (NSP15) is a promising target due to its importance for SARS-CoV-2's evasion of the host's innate immune response.Methods: Using the crystal structure of SARS-CoV-2 NSP15 endoribonuclease, we developed a pharmacophore model of the functional centers in the NSP15 inhibitor's binding pocket. With this model, we conducted data mining of the conformational database of FDA-approved drugs.The conformations of these compounds underwent 3D fingerprint similarity clustering, and possible conformers were docked to the NSP15 binding pocket. We also simulated docking of random compounds to the NSP15 binding pocket for comparison.Results: This search identified 170 compounds as potential inhibitors of SARS-CoV-2 NSP15.The mean free energy of docking for the group of potential inhibitors were significantly lower than for the group of random compounds. Twenty-one of the compounds identified as potential NSP15 inhibitors were antiviral compounds used in the inhibition of a range of viruses, including MERS, SARS-CoV, and even SARS-CoV-2. Eight of the selected antiviral compounds in cluster A are pyrimidine analogues, six of which are currently used in a clinical setting. Four tyrosine kinase inhibitors were identified with potential SARS-CoV-2 inhibition, which is consistent with previous studies showing some kinase inhibitors acting as antiviral drugs. Conclusions:We recommended testing of these 21 selected antiviral compounds for the treatment of COVID-19.

potential_sars-cov-2_nonstructural_protein_15_(nsp15)_inhibitors:_repurposing_fda-approved_drugs

INTRODUCTION<!>MATERIALS AND METHODS<!>Pharmacophore model creation and search of drugs database<!>Computational docking<!>Molecular dynamics simulations<!>CONCLUSION<!>ACKOWLEDGEMENTS

<p>Coronavirus disease-2019 (COVID-19) is a respiratory disease caused by SARS-CoV-2. As of August 1st, 2021, SARS-CoV-2 has cumulatively infected over 198 million people and killed over 4 million individuals in almost 200 countries and regions (https://coronavirus.jhu.edu). The serious threats to global public health and the economy presented by SARS-CoV-2 create an urgent need to identify novel tools to provide new pharmacologic leads that can improve survival for those already infected. SARS-CoV-2 is a positive-sense, single-stranded, RNA betacoronavirus with a genome size of approximately 30kb. The genomic RNA contains a 5'-cap structure and a 3'-poly(A) tail. During infection, the genome is translated to generate viral polyproteins and transcribed to generate negative-sense RNA and subgenomic RNAs. The SARS-CoV-2 genome contains 14 open reading frames (ORFs) that encode 29 proteins, including nonstructural proteins (NSPs), structural proteins, and accessory proteins. The two main units, ORF1a and ORF1b are located at 5'-terminus and produce 16 NSPs through proteolytic cleavage by two viral proteases: the 3Clike protease (3CL pro ) and the papain-like protease (PL pro ). NSPs are essential for RNA transcription, replication, translation, and suppressing the host antiviral response [1][2][3] .</p><p>Targeting viral proteins to disrupt replication is an important approach in developing a therapeutic treatment against SARS-CoV-2 infection. Ideally, one can target highly conserved viral proteins that are unlikely to acquire resistance as the outbreak progresses. Recent studies report SARS-CoV-2 genomic variations in over 10% of isolated sequences, with the most frequent mutations being P323L in NSP12 and D641G in the spike protein 4,5 . In contrast, NSP15, an RNA uridylate-specific endoribonuclease (with a C-terminal region homologous to EndoU enzymes), is highly conserved, making it an attractive target for drug development. NSP15-like endoribonucleases are found in all coronavirus family members, suggesting its endonuclease function is critical for the viral life cycle. The amino-acid sequence alignment of NSP15 from SARS-CoV and SARS-CoV-2 showed 88% sequence identity and 95% sequence similarity 6 .</p><p>NSP15 recognizes uracil and cleaves single stranded RNA through an Mn 2+ requiring transesterification reaction 7 . Recent studies indicate that NSP15 is not required for viral RNA synthesis; rather, NSP15 suppresses the host protective immune response through evasion of host dsRNA sensors 8 . Most recently, NSP15 was reported to participate in viral RNA processing by degrading viral polyuridine sequences. This may prevent the host immune sensing system from detecting viral RNA via cell pathogen-recognition receptors, which subsequently inhibits both direct and indirect antiviral effects 9 . These mechanisms are important for normal coronavirus infection of host cells. In the absence of NSP15 activity, viral replication is slowed significantly, and therefore NSP15 remains an attractive target for addressing SARS-CoV-2 infection 10 . NSP15 is only active as a hexamer, which is formed as a dimer of trimers. The NSP15 monomer contains three domains: a N-terminal domain responsible for oligomerization, a middle domain, and a C-terminal domain, which contains the catalytic domain 11 . Binding sites of each of the catalytic domains are accessible despite hexamerization. A recent publication showed the first two crystal structures of SARS-CoV-2 NSP15 with 1.90 Å and 2.20 Å resolution 6 . In the Cterminal catalytic domain of SARS-CoV-2 NSP15, the active site carries six key residues: His235, His250, Lys290, Thr341, Tyr343, and Ser294. Among of these residues, His235, His250, and Lys290 are suggested to constitute the catalytic triad for its nuclease activity. His250 acts as a general base to activate the 2'-OH of the ribose while His235 functions as a general acid to donate a proton to the leaving 5'-OH the ribose 6,11 . Ser294 together with Tyr343 determine uridine specificity. Ser294 is a key residue to recognize uracil and is assumed to interact with the carbonyl oxygen atom O2 of uracil, while Tyr343 orients the ribose of uridine for cleavage by van-der-Waals interactions 11 . In the crystal structure of the NSP15 citrate-bound form, the citrate ion forms hydrogen bonds with active site residues including His235, His250, Lys290, and Thr341 6 . In the crystal structure of NSP15 complexed with uridine-5'-monophosphate (5'-UMP), 5'-UMP was found to interact with all six active site residues. The uridine base of 5'-UMP interacts with Tyr343 through van der Waals and forms hydrogen bonds with the nitrogen atom of Ser294, Lys290, and His250 12 . This structural information is important for exploring binding of uridine analogues as a potential SARS-CoV-2 NSP15 inhibitors.</p><p>Tipiracil, an uracil derivative, is a thymidine phosphorylase inhibitor. It is an FDA-approved drug used with trifluridine to treat metastatic colorectal and gastric cancer. Previously, tipiracil has been reported to form hydrogen bonds with SARS-CoV-2 NSP15 active site residues Ser 294, Lys345, and His250 12 . Tipiracil suppresses RNA nuclease activity of NSP15 and modestly inhibits SARS-CoV-2 virus replication in vitro without affecting viability of host cells most likely through competitive inhibition 12 . Moreover, recent in-silico-based approaches have identified other potential NSP15 inhibitors that await further structural and biochemical validation 13,14 . The current COVID-19 pandemic brought attention to the repurposing of existing drugs and the rapid identification of candidate compounds. In this study, we use structure-based pharmacophore model and molecular docking to identify potential inhibitors of NSP15 by screening FDA approved drug database.</p><!><p>The crystal structure of SARS-CoV-2 NSP15 endoribonuclease (PDB ID: 6WXC) in complex with the ligand tipiracil (5-chloro-6-(1-(2-iminopyrrolidinyl)methyl)uracil) was downloaded from the RCSB protein data bank. Using Molecular Operating Environment (MOE; CCG, Montreal, Canada), we analyzed the key binding site residues that are responsible for interaction between the NSP15 and tipiracil and employed a structure-based approach to construct our pharmacophore model of NSP15. The default forcefield is Amber 10: EHT with R Field solvation. Our pharmacophore model was created with seven features and excluded volume R= 1.6 Å. It had 1 donor, 3 acceptors, 1 cationic atom&donor, and 2 hydrophobic centroids. Based on this developed pharmacophore, we conducted a pharmacophore search on our conformational database of 2356 FDA approved drugs. Pharmacophore partial match was used for a 5 of 7 features search.</p><p>For multi-conformational docking of the selected compounds, we prepared the NSP15 structure with Protonate 3D application, isolated the ligand and pocket, visualized the space available for docked ligands, defined the binding pocket based on the known key residues for its nuclease activity and uridine specificity, and generated ligand conformations using the bond rotation method. The compounds were docked into the pocket using the Triangle Matcher Method and London dG scoring for placement; and the Induced Fit Method and GBVI/WSA dG scoring for refinement. Poses were ranked by GBVI/WSA binding free energy calculation in the S field. The 56 random control compounds were selected from the FDA drug database.</p><p>To further analyze ligand interactions for some of the above models, the structures were divided into ligand and protein pdb files. The separate structures were protonated: the protein with VMD (Visual Molecular Dynamics, v1.9.4) and the ligand with Avogadro v1.2.0. VMD was used to generate a psf (NAMD protein structure file) file for the protein and the Ligand Reader and Modeler from charmm-gui.org, was used to generate the psf and prm files for the ligand. VMD was then used with the CHARMM36 forcefield to re-combine the ligand and protein, thus solvating the structure and generating the required psf and pdb files 15,16,17 . NAMD v2.14 was used to run 100 steps of minimization followed by 100 ns of dynamics with 2fs/step (50,000,000 iterations). The simulation conditions were rigid bonds involving hydrogen (rigidbonds set to "all"), a splitting distance of 12A between the short range and PME long range potential, Langevin dynamics at 310K with hydrogen atoms excluded (Langevin hydrogen set to "off"), and Periodic Boundary Conditions 15,16,17 .</p><!><p>A recent publication of the crystal structure of SARS-CoV-2 NSP15 endoribonuclease in complex with the ligand tipiracil provides detailed information regarding key residues responsible for the catalytic activity of NSP15 and its interactions with the potential ligands 6 .</p><p>Based on the binding information for these key residues, we generated a pharmacophore model with potential functional centers that bind to the residues in the pocket (Figure 1A). The pharmacophore search with a partial match 5 of 7 centers identified 803 compounds. We selected 170 compounds from the search based on the number of H-bonds and hydrophobic interactions in the best docking pose. Minimum three H-bonds and two hydrophobic interactions was the criteria for selection. Then we clustered the selected compounds using the Similarity Clustering of the MOE Database Viewer with a fingerprint GpiDAPH3 and similarity-overlap parameter SO = 45%. The search identified three major hit clusters, containing ten or more compounds, along with several clusters containing less than ten compounds (from nine to two) and 36 single clusters with just one compound (Table 1). The two largest clusters (A and B) contain 16 and 35 compounds respectively, clusters C, D, E, F, G and H contain 11, 9, 7, 7, 5, and 5 compounds respectively; clusters I, J and K contain 4 compounds each, clusters L to V contain 2 to 3 compounds each, and there are 36 not clustered single compounds (Table 1). 1 and 2).</p><!><p>For docking the selected compounds, we used the crystal structure of SARS-CoV-2 NSP15 endoribonuclease (PDB ID: 6WXC), which was imported into MOE. After the structure preparation and the model's binding pocket was defined, based on known key residues for its nuclease activity and uridine specificity, ligand conformations were generated using the bond rotation method. These were then docked into the site with the Triangle Matcher method and ranked with the London dG scoring function. The retain option specifies the number of poses (30) to pass to the refinement, which is for energy minimization in the pocket, before rescoring with the Induced Fit method and GBVI/WSA dG scoring function. To validate docking, 56 random control compounds were selected from the FDA drug database, using a random number generator without repetitions.</p><p>The values of docking free energies of the selected and random compounds are shown in Figure 3. The means of the selected and random compounds are −6.50 kcal/mol and −5.79 kcal/mol, respectively. Furthermore, the p value of one tail for selected vs random compounds is 1.31 E-06.</p><p>Energies of interaction with the NSP15 active site are shown in Table 2 and Table S1. Figure 3. Free energies of docking interaction of selected and random compounds with SARS-CoV-2 NSP15. The means of the selected and random compounds are −6.50 and −5.79 kcal/mol, respectively. The p value of one tail is 1.31E-06. Table 2. List of selected compounds sorted by their energies of interaction with SARS-CoV-2 NSP15 in the docked positions. All compounds shown have an energy less than -7. DFE: Docking free energy.</p><!><p>We selected the top three compounds in docking energies to further analyze stability of ligand interactions: cefmenoxime, cefotiam, and ceforanide. The final configuration of the compoundprotein complexes resulting in this MD simulations are shown in Figure 4 and Table 3.</p><p>Cefmenoxime (Figure 4A and 4D) had 6 major ligand interactions with NSP15, the shortest distance of which was 2.73 with the residue Lys290. Cefotiam (Figure 4B and 4E) had 4 major ligand interactions with NSP15, the shortest distance of which was 2.693 with the residue Leu246. Finally, ceforanide (Figure 4C and 4F) had 2 major ligand interactions with NSP15, the shortest distance of which was 2.70 with the residue Lys290. Figure 5 shows the measures though MD distances between the NZ atom of LYZ290 of protein with the geometric center of these compounds. One can see that these distances pretty stable during MD simulations. 4).</p><p>Table 4. List of selected compounds with known antiviral activity.</p><p>According to the DrugVirus.info database 18 , 13 of the antiviral compounds selected by the pharmacophore-based search showed activity against a total of 40 viruses in cell-culture, animal, and clinic models (Figure 6). The other eight antiviral compounds were not in the database. A previous study did not identify any of these compounds as potential NSP15 inhibitors, and their top selected drugs did not show antiviral activity 14 . Differences in methodology may explain these discrepancies of results. Specifically, Chandra and co-authors used NSP15 PDB ID 6W01 structure with a citrate ion 14 ; we used crystal structure of NSP15 in complex with tipiracil that binds to NSP15 uracil site. We assume that the pharmacophore model generated on this protein structure includes the key features responsible for ligand interaction with residues in NSP15 active site. We did notice that tipiracil, the positive control, did not have a low free energy.</p><p>However, an in-vitro study confirmed that tipiracil can inhibit uracil binding to the NSP15 active site presumably through competitive inhibition and modestly suppress SARS-CoV-2 viral replication in cellular assays 12 . Cluster A includes six pyrimidine analogues that are currently used as viral inhibitors: HIV reverse transcriptase inhibitors-zidovudine and stavudine, HBV DNA polymerase inhibitor-telbivudine, and HSV DNA polymerase inhibitors-brivudine, edoxudine, and trifluridine (Table 4 and Figure 6). The other two drugs in cluster A, tipiracil 12 and floxuridine 19 , are anticancer drugs that have antiviral properties. All these pyrimidine analogues are polymerase inhibitors, which is a major class of antiviral drugs. These results support using NSP15's pharmacophore features to identify potential antiviral compounds containing a pyrimidine-like scaffold and further development of nucleotide-like drugs with higher affinity for the active site of NSP15. Recent studies demonstrated that tyrosine kinase inhibitors have antiviral potential through inhibition of key kinases required for viral entry and reproduction 20,21 . Thus, repurposing receptor tyrosine kinase inhibitors is an effective strategy in the fight against COVID-19 22 . Our pharmacophore model successfully identified four tyrosine kinase inhibitors with antiviral activity in cluster G, the binding affinities of which are high. Dasatinib, an approved drug for chronic myelogenous leukemia (CML), has activity against both MERS-CoV and SARS-CoV in vitro and possible protection against SARS-CoV-2 infection 23,24 . EGFR inhibitor gefitinib has demonstrated in vitro activity against HCV, BKV, CMV, and VACV (Figure 6). Lapatinib was just recently found to potently inhibit SARS-CoV-2 replication at clinical doses, strongly supporting our screening result 25 .</p><p>Promising antiviral drugs from cluster U includes HIV proteinase inhibitor amprenavir.</p><p>Specifically, amprenavir has a free energy of −7.29 kcal/mol and modestly inhibits replication of SARS-CoV-2 in vitro 26 . Outside of clusters A, G, and U, other antiviral drugs include influenza neuraminidase inhibitors peramivir and oseltamivir, HIV non-nucleoside reverse transcriptase inhibitor doravirine, and HCV NS5B polymerase inhibitor sofosbuvir, which displayed activity against SARS-CoV-2 27 .</p><p>Interesting to note that some of randomly selected FDA-approved drugs had free energies below −7.00 kcal/mol, namely gadoxetate (−8.31 kcal/mol), iohexol (−7.45 kcal/mol), and chlortetracycline (−7.11 kcal/mol) (Table S1). These compounds also can be potential inhibitors of NSP15.</p><!><p>Given the severity of the COVID-19 pandemic, we need a fast way of finding treatment.</p><p>Identification of FDA-approved drugs to inhibit SARS-CoV-2 could lead to advances in this field. Though this study is limited due to only using computer-based screening, the implications of the 170 compounds is a key step in finally finding a treatment. Twenty-one of these drugs have known antiviral properties, some of which have demonstrated inhibition of SARS-CoV-2 in vitro. We recommended testing of selected compounds for the treatment of COVID-19, especially those in clusters A, G, and U.</p><!><p>We would like to thank the people of San Diego Supercomputer Center and CCG (Montreal, Canada).</p>

ChemRxiv

A Structural Model for Glutathione-Complexed Iron-Sulfur Cluster as a Substrate for ABCB7-Type Transporters

Glutathione-complexed [2Fe-2S] cluster is shown to significantly stimulate the ATPase activity of an ABCB7-type transporter in both solution and proteoliposome-bound forms (KD ~ 68 \xce\xbcM). The cluster is a likely natural substrate for this transporter, which has been implicated in cytosolic Fe-S cluster protein maturation. A possible substrate-binding site is identified on a new structural model for the active transporter.

a_structural_model_for_glutathione-complexed_iron-sulfur_cluster_as_a_substrate_for_abcb7-type_trans

<p>Iron-sulfur clusters are essential cofactors in many biological pathways. Several functionally discrete biosynthetic pathways for bacterial iron-sulfur cluster biogenesis have been described (Isc, Nif and Suf) and each has been studied extensively.1, 2 Eukaryotic cluster assembly involves a pathway based on proteins in the bacterial ISC operon, and it is generally believed that both cytosolic and nuclear iron-sulfur clusters are dependent on mitochondrial iron sulfur cluster assembly.2, 3 Details of how the mitochondrial and cytosolic iron-sulfur cluster assembly pathways are connected remain unclear, but have been the subject of intense scrutiny with multiple proteins implicated, even if their roles are not unequivocally defined.4–7 Studies have shown that Atm1p/ABC7 deficiency leads to impaired cytosolic iron-sulfur cluster protein activity and iron accumulation in mitochondria, but there is no impact on mitochondrial iron-sulfur cluster protein activity.3, 8 In humans, natural mutants of the transporter have been identified in patients affected with X-linked sideroblastic anaemia and cerebellar ataxia,9 and definition of the substrate and a structural model for the protein are important first steps toward understanding the molecular basis for these disease states</p><p>Although the Atm1p/ABC7 membrane spanning protein appears to be the exporter required for cytosolic cluster biosynthesis,3, 7, 8 the substrate for the transporter is unknown. In this paper we present evidence that a novel glutathione complexed [2Fe-2S] cluster10,11 is a plausible transporter substrate,10, 11 and discuss this finding in the context of a new structural model that we have defined for the heretofore structurally uncharacterized ABC7-type transporters. Definition of the pathway for mitochondrial cluster export is a crucial step to understanding the biogenesis and regulation of cellular iron-sulfur cluster cofactors.</p><p>Atm1p/ABC7 proteins are ATPase-driven pumps that drive active transport.3 Previously it has been shown that both reduced and oxidized glutathione stimulate the ATPase activity of Atm1p/ABC7,12 indicative that the thiol is not a key contributor to the stimulatory mechanism. A role for glutathione in mediating mitochondrial cluster export is supported by the observation that glutathione depletion impairs the maturation of cytosolic iron-sulfur cluster proteins, but has no effect on mitochondrial cluster proteins, consistent with a close genetic relationship between ATM1 and GSH1.13 It is therefore clear that glutathione is intimately involved in iron-sulfur cluster export.</p><p>The involvement of glutathione in both cellular iron chemistry14 and iron-sulfur cluster biosynthesis has previously been evidenced by the characterization of several glutaredoxin proteins with glutathione-coordinated [2Fe-2S] clusters that mediate cluster transfer chemistry,15–19 and by the fact that human glutaredoxin can exchange its [2Fe-2S] cluster with the scaffold protein ISU.20 This glutathione-coordinated iron-sulfur cluster complex is stable under physiological conditions in the presence of physiological concentrations of glutathione, and undergoes cluster exchange with the ISU scaffold protein.10 Since neither a bare cluster core, nor a protein-bound cluster are likely substrate candidates for this class of exporter (on ligand and size grounds), and given the additional evidence implicating glutathione in cluster export, we viewed such a cluster complex as a viable substrate candidate for the ABC7-type transporter. Herein we present results of investigations that further support the idea that [2Fe-2S](GS)4 is a substrate for mitochondrial cluster export, and identify a possible substrate-binding on a new structural model for the active transporter.</p><p>It is generally observed that substrates for ABC transporters stimulate the ATPase activity of the transporter.21, 22 To test the hypothesis that [2Fe-2S](GS)4 is a substrate for the transporter, yeast Atm1p protein was cloned, expressed and purified (SI). Its activity was confirmed by ATPase assay measurements, yielding standard Michaelis-Menten parameters (KM ~ 54.6 ± 0.4 μM and kcat ~1.93 ± 0.03 min−1) (Figure S9) in good agreement with other ABC transporters and Atm1p.23 Varying concentrations of the complex were incubated with the transporter in the presence of physiological glutathione concentrations and the ATPase activity of the transporter was followed (Figures 1 and 2). With 10 mM glutathione, but no cluster complex present, the rate of phosphate formation increases, which is consistent with previous studies that have shown that glutathione can stimulate the ATPase activity of Atm1p.12 With the same concentration of GSH, cluster complex significantly stimulates the ATPase activity of Atm1p at low μM concentrations (Figure 1 and S8). The dependence of activity on cluster concentration (Figure 2) was fit to a nonessential activation model (equation 1),24 where νmax is the maximum initial ATPase activity of Atm1p in the absence of cluster, [S] is the concentration of substrate Mg-ATP, [A] is the concentration of cluster stimulant, KD is the binding constant of the cluster to Atm1p, KM is the binding constant of Mg-ATP to Atm1p, α accounts for the modification of KM by cluster, and β accounts for νmax stimulation by cluster. The data illustrated in Table 1 demonstrates the glutathione cluster complex to serve as a modifier that increases the velocity of Atm1p-catalyzed phosphate formation by 1.9-fold and decreases the KD 0.6-fold, which supports the hypothesis that the glutathione iron-sulfur cluster is a likely substrate for this transporter in a manner consistent with previous genetic interaction and knock-out studies.13 The cluster complex shows saturation binding to the transporter with a measured KD of 68 μM.</p><p>The stimulation of transporter ATPase activity by glutathione iron-sulfur cluster complex was further studied in a proteoliposome system. The proteoliposome was constructed by reconstituting purified yeast Atm1p protein into liposome made of a mixture of 1:1:1 DOPE, DOPC and DOPG.25 Similar to the results noted above, [2Fe-2S](GS)4 was found to stimulate the ATPase activity of reconstituted proteoliposome (Figures S7 and S8).</p><p>The relative KD's for [2Fe-2S](GS)4 and glutathione indicate a much higher affinity for the cluster complex (68 μM versus > 689 μM, respectively). Prior observation of very modest levels of stimulation of Atm1p/ABC7 ATPase activity by glutathione are consistent with a glutathione cluster as a natural transporter substrate, with the more modest levels of stimulation reflecting weaker intrinsic binding to the transporter (Table 1), as a result of partial occupation of some of the contact sites on the transport protein occupied by the full tetrameric glutathione complex cluster. Nevertheless, the link between glutathione and iron-sulfur cluster transport remained unclear until the successful synthesis and characterization of the stable glutathione Fe/S cluster complex.10, 11 In our previous studies, we were able to show that this complex is stable in the presence of physiological glutathione concentration, and that this complex is labile enough to exchange cluster with iron-sulfur cluster scaffold protein,10, 11 making this a perfect cluster carrier in a cellular environment.</p><p>Recent crystallographic advances have resulted in determination of the structure of a mitochondrial ABCB10 transporter, which shows the transporter in a functional dimeric state in a closed conformation, and with Mg-ATP bound to classical Walker motifs. This protein shows ~ 30% identity and 50% sequence similarity to the ABCB7 transporter (partial homology shown in Figure S10), and is of value in efforts to understand the structural mechanism of ABCB7 transport. By use of the ABCB10 structure (PDB: 3ZDQ) as a template in SWISS-MODEL, we generated model structures for both yeast Atm1p and human proteins. Electrostatic surface maps of both model structures were calculated using APBS and each showed two positively-charged pockets at the bottom of the transmembrane segment, and close to nucleotide binding domain. These represent possible binding sites for the negatively-charged {[2Fe-2S](GS)4}2− complex (Figure 3).</p><p>Two positively-charged patches were noted. One lies between the two transmembrane helix bundles, presumably facing inward to the channel once the dimeric transporter is formed. This positively-charged patch is formed on one side by a conserved arginine-rich area, Arg313, Arg315, Arg317, Arg319 of the human ABCB7 protein and Arg280, His282, Arg284, Arg285 of yeast Atm1p (Figure S12). On the opposite side, Arg432 and Arg435 of the human ABCB7 protein, and Arg397 and Lys400 of Atm1p complete a positively-charged pocket that is ready to bind the negatively charged iron-sulfur cluster complex (Figure 3). We speculate that in its native dimeric state, these two discrete sites of Atm1p/ABC7 may function as complementary domains to create a positive binding pocket for [2Fe-2S](GS)4 cluster(s). Significantly, no stimulation of ATPase activity by cluster was observed when Arg284 was substituted with Glu (Figure 2 and Table 1), although full ATPase activity was retained.</p><p>In this report, we have shown evidence in support of [2Fe-2S](GS)4 as a likely iron-sulfur cluster substrate for the Atm1p/ABC7 transporter in both solution and proteoliposome-bound forms, and identify a likely substrate binding site on the transporter. The mitochondrial cluster export pathway can be considered in four steps (Figure S13). First, mitochondrial glutathione abstracts the [2Fe-2S] cluster core from the mitochondrial ISC machinery (most likely ISU),10, 11 forming a glutathione iron-sulfur cluster complex. Such complexes are then transported through the mitochondrial membrane, driven by ATP hydrolysis. Finally, the exported complexes are delivered to the CIA (cytosolic Iron-sulfur cluster Assembly) machinery, possibly by transfer to the cytosolic ISU. This provides the essential link between mitochondrial cluster biosynthesis and the rest of the cell, as well as providing a test bed for understanding human disease states that stem from natural mutants of this transporter.9 Future studies will be focused on detailed evaluation of the substrate binding and transport mechanisms, the connection to the CIA pathway, and the activity of disease-causing protein point-substitutions.</p>

PubMed Author Manuscript

Tracers for non-invasive radionuclide imaging of immune checkpoint expression in cancer

AbstractImmunotherapy with checkpoint inhibitors demonstrates impressive improvements in the treatment of several types of cancer. Unfortunately, not all patients respond to therapy while severe immune-related adverse effects are prevalent. Currently, patient stratification is based on immunotherapy marker expression through immunohistochemical analysis on biopsied material. However, expression can be heterogeneous within and between tumor lesions, amplifying the sampling limitations of biopsies. Analysis of immunotherapy target expression by non-invasive quantitative molecular imaging with PET or SPECT may overcome this issue. In this review, an overview of tracers that have been developed for preclinical and clinical imaging of key immunotherapy targets, such as programmed cell death-1, programmed cell death ligand-1, IDO1 and cytotoxic T lymphocyte-associated antigen-4 is presented. We discuss important aspects to consider when developing such tracers and outline the future perspectives of molecular imaging of immunotherapy markers.Graphical abstractCurrent techniques in immune checkpoint imaging and its potential for future applications

tracers_for_non-invasive_radionuclide_imaging_of_immune_checkpoint_expression_in_cancer

<!>PD-1 imaging<!>PD-L1 imaging<!>Imaging CD28 and CTLA-4 and their ligands CD80 and CD86<!>CTLA-4 imaging<!>CD80 and CD86 imaging<!>IDO and TDO imaging<!>CD276 imaging<!>A2aR imaging<!>OX40 imaging<!>Further targets of interest<!>Discussion<!>Future prospects<!>Conclusion<!>

<p>Overview of nuclear imaging tracers for immune checkpoints. Only tracers that have been published and used in at least preclinical in vivo studies are described in the tables below</p><p>L-5-[18F]fluoro-tryptophan</p><p>and D-5-[18F]fluoro-tryptophan</p><p>1-(2-[18F]fluoroethyl)-l and d-tryptophan</p><p>(1-L-[18F] FETrp and 1-D-[18F]FETrp)</p><p>Immune checkpoint expression and main interactions on cell types which predominantly express them. Depicted are immune checkpoints for which tracers have been developed. Not all immune checkpoint interactions are known nor are all interactions displayed. For further reading on immune checkpoints, we refer to De Sousa Linhares et al. (2018) PD-L1: Programmed Death-ligand 1, PD-1: Programmed death-1, CTLA-4: Cytotoxic T lymphocyte associated antigen-4, A2aR: Adenosine 2a receptor, IDO: Indoleamine 2,3-dioxygenase</p><!><p>The potential of PD-1 imaging has been demonstrated in several preclinical and clinical studies. A copper-64 (64Cu) labeled anti-mouse PD-1 Ab was developed by Hettich et al. (Hettich et al. 2016). Studies in naïve and PD-1−/− mice showed specific uptake in lymphoid organs (lymph nodes and spleen) of naïve mice, which was significantly lower in PD-1 knock-out mice at 24 h after injection, confirming the physiological expression of PD-1 in different immune compartments. In naïve mice with B16-F10 melanoma tumors that received PD-L1 and Cytotoxic T lymphocyte-associated protein 4 (CTLA-4 or CD152) directed ICI therapy, high tracer uptake was observed in the tumor upon irradiation, suggesting that PD-1 PET can be used for ICI treatment monitoring. Natarajan et al. injected a 64Cu-labeled anti-mouse PD-1 Ab in Foxp3+ LuciDTR4 mice, a mouse model that contains high expressing PD-1 Foxp3+ regulatory T cells, bearing B16-F10 tumors, to detect PD-1 expressing tumor infiltrating lymphocytes (TILs) (Natarajan et al. 2015). PET images obtained at 1–48 h after injection indicated tracer uptake mainly in the tumor and spleen, which was both found to be PD-1 specific T cell-mediated uptake confirmed by bioluminescence, IHC, and concurrent blocking of uptake by co-injection of an excess unlabeled tracer. In a subsequent study, Natarajan et al. developed a tracer against human PD-1. Pembrolizumab, a clinically approved anti-PD-1 Ab, was radiolabeled with 64Cu or zirconium-89 (89Zr) to detect tumor infiltration of adoptively transferred human peripheral blood mononuclear cells (hPBMCs) in NSG mice xenografted with human A-375 melanoma tumors (Natarajan et al. 2017). 89Zr-pembrolizumab uptake was observed in the tumor and spleen of hPBMCs engrafted mice at 24 h after injection. Specificity for PD-1 was demonstrated by the reduced 64Cu-pembrolizumab tumor uptake in mice coinjected with an excess of unlabeled pembrolizumab. In a final study by Natarajan et al., the 64Cu human PD-1 tracer was evaluated in NSG mice xenografted with 293 T/hPD-1 stable non-cancer cells and in NSG mice with adoptively transferred hPMBCs xenografted with A375 human melanoma cells (Natarajan et al. 2018a). PET imaging demonstrated specificity of the tracer towards PD-1 in 293 T/hPD-1 tumors in mice that did not receive an excess of unlabeled pembrolizumab compared to mice that did receive an excess of unlabeled tracer which showed significantly less tumor uptake. Moreover, ex vivo biodistribution indicated a statistically significant different tumor uptake between these two groups at 1, 24, and 48 h after injection. Studies with the adoptively transferred hPBMCs xenografted with A375 human melanoma mouse model indicated clear 64Cu-pembrolizumab uptake in tumors, suggesting infiltration of hPBMCs into the tumor microenvironment. Others evaluated the use of 89Zr-labeled nivolumab, a clinically used anti-human PD-1 Ab, in humanized (engrafted with peripheral blood mononuclear cells) or non-humanized NSG mice bearing subcutaneous A549 human lung tumors (England et al. 2018). PET imaging revealed higher 89Zr-nivolumab tumor uptake in humanized mice compared to non-humanized mice from 72 h onwards. Experiments comparing the uptake of non-specific 89Zr-IgG versus 89Zr-nivolumab in the humanized A549 tumor bearing mice confirmed the specificity of 89Zr-nivolumab tumor uptake. Interestingly, specific salivary gland uptake was observed that was mainly attributed to homing of lymphocytes due to graft versus host disease in this specific mouse model. These preclinical studies indicate that PD-1 imaging with PET might be a useful tool to image the presence of PD-1 expressing lymphocytes in the tumor microenvironment before ICI treatment or to image PD-1 expressing TILs during ICI therapy for treatment monitoring.</p><p>Recently, in a first-in-human clinical study by Niemeijer et al., 89Zr-nivolumab tracer uptake was evaluated in non-small cell lung cancer patients prior to nivolumab ICI treatment (Niemeijer et al. 2018). As expected from mouse models, tracer accumulation was observed in the spleen because of PD-1 expression on lymphocytes and Fc-receptor mediated uptake (Arlauckas et al. 2017) as well as in the liver because of tracer catabolism. More interestingly, a correlation between PD-1 expressing TILs by IHC in a primary tumor biopsy and 89Zr-nivolumab uptake was observed. Moreover, higher 89Zr-nivolumab uptake prior treatment was observed in responding tumor lesions than in non-responding tumor lesions after 3 months of nivolumab treatment. This study demonstrated that PET imaging can be used to quantify and monitor PD-1 expression non-invasively over time before ICI therapy.</p><!><p>For PD-L1 imaging different tracer moieties have been explored, ranging from peptides, adnectins, up to full mAbs, labeled with various radionuclides for both PET and SPECT imaging of murine and human PD-L1 for mechanistic and translational purposes, respectively.</p><p>The first PD-L1 imaging agent was developed by Heskamp et al. who employed an 111In-labeled murine Ab directed against human PD-L1 and successfully imaged human xenografts in athymic mice with different PD-L1 expression levels (Heskamp et al. 2015). In a subsequent study, they investigated whether changes in PD-L1 expression on tumors could be visualized after radiotherapy using a mAb directed against murine PD-L1. Colon carcinoma (CT26) and Lewis lung carcinoma (LLC1) syngeneic mouse tumors showed significant increased tumor uptake after a single dose of 10 Gy external beam irradiation (Heskamp et al. 2019) which correlated to the increased IHC PD-L1 expression levels. Kikuchi et al. also investigated the effect of radiotherapy on the expression of PD-L1. A 89Zr-labeled mAb against mouse PD-L1 was used to show increased tracer uptake in a syngeneic head and neck tumor model after fractionated radiotherapy (Kikuchi et al. 2017). Other research teams have also successfully employed mAbs for imaging of PD-L1 using different radionuclides, including 89Zr, 111In, 64Cu and 131I (Hettich et al. 2016; Chatterjee et al. 2016; Josefsson et al. 2016; Nedrow et al. 2017a, 2017b; Lesniak et al. 2016; Li et al. 2018; Moroz et al. 2018; Truillet et al. 2018; Pang et al. 2018).</p><p>Next to mAbs, other moieties such as nanobodies, affibody molecules, adnectins and peptides have been tested in preclinical tumor models. Broos et al. developed a 99mTc-labeled anti-mouse PD-L1 nanobody (also known as single domain antibody or VHH) (Broos et al. 2017). Experiments with immunocompetent mice bearing syngeneic myeloma TC-1 tumors showed specific physiological uptake in lungs, heart, spleen, thymus, lymph nodes and brown fat, as well as moderate tumor uptake. Others have shown successful studies using a 64Cu-labeled peptide that binds human and mouse PD-L1 and demonstrated high uptake in PD-L1+ xenograft models (Chatterjee et al. 2017). To achieve high-contrast images at earlier time points, they also optimized their peptide for a 68Ga-labeling. These results showed less tracer accumulation in xenograft tumors, but higher tumor-normal-tissue contrast in the PET images. However, high kidney and liver uptake was also observed (De Silva et al. 2018). Gonzalez et al. have shown results of a 18F-labeled affibody molecule which specifically targets human tumor PD-L1, but also demonstrates high renal and bone uptake (Gonzalez Trotter et al. 2017). Others have also published positive results with smaller targeting agents radiolabeled with 99mTc, 18F, and 68Ga (De Silva et al. 2018; Donnelly et al. 2018; Ingram et al. 2017; Kumar et al. 2018; Mayer et al. 2017).</p><p>Niemeijer et al. performed the first in human PD-L1 imaging study with a PD-L1 targeting 18F-labeled adnectin (fibronectin binding domain 3 or monobody) (Niemeijer et al. 2018). Imaging with this low molecular weight tracer enabled same day imaging and illustrated the heterogeneous nature of PD-L1, both within and between patients with non-small-cell lung carcinoma (NSCLC). Most striking was the comparison of biopsied material against PD-L1 positive lesions on PET, where multiple cases of biopsy negative but scan-positive patients were observed. Furthermore, they showed therapy response was correlated with tracer uptake but not with biopsy findings. It must be noted that PD-L1 expression in these scan positive lesions was not confirmed with further biopsies and therefore we cannot conclude that these areas are indeed PD-L1 positive. However, these findings do suggest that PD-L1 can be expressed heterogeneously within and between tumors lesions and biopsies provide only limited information. Although the number of patients were limited, these results are encouraging for the future of immune checkpoint imaging in humans.</p><p>Bensch et al. performed patient imaging of PD-L1 with a clinically approved therapeutic mAb, 89Zr-labeled atezolizumab (anti-human PD-L1) (Bensch et al. 2018). Patients with metastatic bladder cancer, NSCLC or triple negative breast cancer being treated with atezolizumab were included. This study also found distinct heterogeneity of PD-L1 expression within and between patients on PD-L1 IHC. Furthermore, they showed a strong predictive value of PD-L1 PET imaging on progression-free survival as well as overall survival. When comparing PD-L1 PET to different PD-L1 IHC assays, they found that IHC could not predict treatment response and survival. This showcases the distinct value of imaging in patient stratification for immunotherapy, utilizing the therapeutic agent as a tracer.</p><p>Xing et al. evaluated a sdAb labeled with 99mTc to visualize PD-L1 status in NSCLC patients on SPECT (Xing et al. 2019). They demonstrated safety and of their imaging compound and showed acceptable dosimetry when using 99mTc. Furthermore, they were able to visualize PD-L1 positive tissues (spleen, liver and bone marrow) as well as tumors at 2 h post injection. As well as in the studies by Bensch and Niemeijer, heterogeneous uptake was found between primary tumors and nodal or bone metastases.</p><!><p>The first molecule expressed by T cells in their activation cascade and required for their survival, is CD28, which binds both CD80 and CD86 present on antigen presenting cells (APCs). This stimulatory interaction can be inhibited by Cytotoxic T lymphocyte-associated antigen-4 (CTLA-4) which is also expressed by T cells and has a significantly higher affinity for CD80 (also B7–1) and CD86 (also B7–2) than CD28 (see Fig. 1). In a normal situation this inhibitory signal dampens T cell responses thereby avoiding collateral damage to healthy tissues (Acuto and Michel 2003; Zhao et al. 2018). So far several radiotracers for CTLA-4, CD80 and CD86 have been reported.</p><!><p>CTLA-4 is present on activated T cells and constitutively expressed by regulatory T cells as well as some types of tumor cells (Contardi et al. 2005). Counteracting the immune inhibitory effect of CTLA-4, the FDA-approved CTLA-4-inhibitor ipilimumab shows great anticancer efficacy in a wide range of cancer types. Despite the multitude of clinical therapeutical studies, more than 300 ongoing clinical studies for this mAb alone (ClinicalTrials.gov 2019), only four publications on imaging of CTLA-4 have been published. Higashikawa et al. (2014) used a 64Cu-labeled anti-murine CTLA-4 mAb to visualize CTLA-4 in CT26 tumor-bearing BALB/c mice. Tumor uptake was significantly higher in mice that received radiolabeled anti-CTLA-4 compared to mice that received radiolabeled non-specific IgG. Polymerase chain reaction (PCR)-analysis on CT26 tumor tissues from BALB/c and T cell lacking BALB/c nu/nu indicated that the CTLA-4 expression was T cell dependent and therefore the tracer could be used to image CTLA-4 positive T cells. In 2017, Ehlerding et al. reported uptake of 64Cu-labeled ipilimumab by CTLA-4 expressing human NSCLC xenografts (A549, H460, and H358) (Ehlerding et al. 2017). In vivo tumor tracer uptake correlated with in vitro CTLA-4 expression levels of these tumor cells, with the highest uptake in the A549 cell line 48 h post infusion. Furthermore, antigen specificity was evaluated by administration of excess unlabeled Ab to tumor bearing control mice. In a recent study, Ehlerding et al., validated the same anti-CTLA-4 tracer and a 64Cu-labeled IdeS protease fragmented ipilimumab F (ab')2 in human peripheral blood lymphocytes engrafted NBSGW mice, a model that does not need full body irradiation to engraft human peripheral blood lymphocytes (PBLs) (Ehlerding et al. 2019). Both tracers showed targeting in salivary glands which upon IHC analysis showed activated CTLA-4 positive lymphocytes involved in a graft versus host disease. The F (ab')2 tracer showed increased clearance, and thereby an increased salivary gland to blood ratios. Furthermore, tracer specificity was confirmed with non PBL engrafted NBSGW mice and radiolabeled nonspecific IgG isotype controls (Ehlerding et al. 2019). Ingram et al. developed H11, an anti-CTLA-4 VHH that can be 18F or 89Zr-functionalized (Ingram et al. 2018). In vivo targeting of 18F-H11 in mice bearing T cell containing B16F10 tumors showed tracer uptake above background. A 89Zr-labeled 20 kD PEG conjugated H11 (H11PEG20) showed a substantially improved signal to noise ratio in the tumor and uptake in the GVAX injection site, a tumor model specific immune stimulatory cancer vaccine that was applied before tumor inoculation, indicating activated T cell specific targeting. These results indicate the feasibility of using radiolabeled anti-CTLA-4 agents for assessment of TIL or tumor CTLA-4 expression levels. Currently, a first clinical CTLA-4 imaging study is ongoing, where tumor lesion uptake and biodistribution of 89Zr-labeled ipilimumab will be assessed at the start of ipilimumab therapy and 3 weeks post start of therapy. Furthermore, this study is designed to determine a possible correlation between tumor uptake and therapy responses, uptake in normal tissues and to assess a correlation between 'on-target off-tumor' targeting and toxicity (Philips and Atkins 2015).</p><!><p>CD80 and CD86 are expressed mainly by APCs and their binding to CD28 and CTLA-4 stimulates and inhibits immune responses, respectively. These molecules are also expressed by some types of myelomas, lymphomas and carcinomas (Flörcken AaJ et al. 2017). Conditionally being either immune stimulatory or inhibitory, imaging CD80 or CD86 expression in tumors might be used to predict response to CTLA-4 targeted therapies. Alternatively, in case of CD80/CD86 negative tumor cells, imaging of these targets could be used to non-invasively measure APC infiltration. Furthermore, if the ongoing therapeutical clinical trial with CD80/CD86 targeting CAR-T cells turns out successful, nuclear imaging could aid in patient selection for this treatment approach (ClinicalTrials.gov 2019). Meletta et al. investigated in vivo tumor targeting of 111In-labeled belatacept (Meletta et al. 2016). This tracer, a fusion protein consisting of a human IgG1 Fc fragment linked to the extracellular domain of CTLA-4, showed higher uptake in CD80+/CD86+ Raji tumors (a Burkitt's lymphoma) compared to uptake in double negative NCI-H69 tumors. Furthermore, co-injection with excess unlabeled belatacept in Raji tumor-bearing mice showed significantly decreased tracer uptake, indicating specific receptor targeting of the tracer. Furthermore, Meletta et al. have developed a 11C-labeled pyrazolocinnoline derivative AM7 to target CD80 positive cells (Meletta et al. 2017). In vivo uptake of this tracer was low in atherosclerotic plaques rich with CD80 expressing macrophages compared to [18F]FDG. Tracer specificity for its target was confirmed by in vitro autoradiography and IHC (Muller et al. 2014). Although no studies have been performed in patients, these studies suggest the feasibility of assessing CD80 and CD86 expression with radiolabeled tracers.</p><!><p>Decreased extracellular tryptophan levels and increased kynurenines levels (tryptophan metabolites) inhibit T and NK cell proliferation and activation, therefore tryptophan is a metabolic immune checkpoint (see Fig. 1). Some tumors exploit this immune modulating mechanism by over-expressing tryptophan degrading enzymes indoleamine 2,3-dioxygenase (IDO1 and IDO2) and tryptophan 2,3-dioxygenase (TDO) (Platten et al. 2014; Munn and Mellor 2007). Several inhibitors for the rate limiting enzyme IDO1 (such as Epacadostat, Indoximod and Navoximod) have been developed and are currently undergoing evaluation in clinical trials as adjuvants (Prendergast et al. 2017).</p><p>Multiple tracers for assessing expression levels of IDO1/TDO have been developed. All tracers, except an 18F-labeled epacadostat analog developed by Huang et al. (2017), are 18F or 11C-labeled derivatives of either L or D isomers of tryptophan (Table 1). Although already developed in 1988, the first clinical study with the IDO1 tracer α-[11C]methyl-L-tryptophan ([11C]AMT) was performed in 2006. In patients with brain tumors, [11C]AMT-PET demonstrated increased tracer uptake in the tumor compared to normal cortex (Juhasz et al. 2006). Moreover, multiple clinical [11C]AMT-PET trials demonstrated prolonged tracer retention and high uptake in NSCLC lesions (Juhasz et al. 2009), invasive ductal breast carcinoma lesions Juhasz et al. (2012) and meningiomas Zitron et al. (2013). Further clinical studies with [11C]AMT-PET demonstrated the ability to differentiate radiation necrosis from recurrent gliomas (Alkonyi et al. 2012), and showed a strong association of high tumor [11C] AMT uptake parameters (SUVmax, SUVmean and tumor-to-background ratio) to a significantly decreased 1-year survival (Kamson et al. 2014). Although a recent clinical [11C]AMT-PET study in 3 patients failed to demonstrate objective clinical responses to IDO inhibitor therapy, it did show heterogeneous intratumoral tracer uptake potentially reflecting IDO activity (Lukas et al. 2019). Furthermore, these results indicate that [11C]AMT-PET might be used for stratification of true progression versus pseudoprogression. In a recent preclinical study, in vivo tumor targeting of a newly developed IDO1 tracer, 1-(2-[18F]-fluoroethyl)-L-tryptophan ([18F]FETrp), was compared with [11C] AMT and increased SUVs were observed for [18F] FETrp compared with [11C] AMT in lung, breast, and brain xenografts (Michelhaugh et al. 2017). Target specificity of [18F] FETrp has also been demonstrated in preclinical prostate, lung, breast, and glioma tumor models by Xin et al. with a significantly higher tumor uptake than in low IDO1 expressing healthy tissues, and a correlation of in vitro cell binding and in vivo [18F] FETrp tumor uptake. So far only [11C] AMT imaging agent is being investigated clinically to evaluate whether IDO1 imaging could serve as a predictive marker for immunotherapy. This study, [18F] FDG PET 48 h before pembrolizumab and [11C] AMT imaging 24 h before pembrolizumab, will investigate a possible association of the SUVmax with objective PD-1 between tracer uptake, PD-1 inhibitor therapy response and IHC (Philips and Atkins 2015).</p><!><p>CD276 (also known as B7-H3) is presented at low levels on healthy tissues but highly expressed by APCs and macrophages and by a range of solid tumors. Although its target is still unknown, ample evidence is present that show its involvement in inhibiting T cell function (see Fig. 1) (Dong et al. 2018). This is substantiated by the antitumor effects of currently preclinically and clinically investigated CD276 inhibitors or targeted therapies (ClinicalTrials.gov 2019; Lee et al. 2017). To aid patient selection for CD276 inhibitor therapy, tracers are being developed. A humanized anti-B7-H3 mAb, 89Zr-labeled 5573a, has been used in immunodeficient CD276+ MDA-MB-231 tumor-bearing mice for non-invasive imaging of CD276 (Burvenich et al. 2018). PET/MRI and biodistribution analysis showed significantly higher tumor uptake compared with mice that were co-injected with an excess unlabeled tracer, indicating CD276 specificity. This tracer demonstrates the potential of in vivo CD276 imaging. However, further preclinical evaluation in immunocompetent mice is warranted in order to better understand tracer behavior and to translate findings to patients.</p><!><p>The metabolic inhibitory immune checkpoint Adenosine 2a receptor (A2aR) is expressed by neurological synapses, certain tumor cells, and a wide range of immune cells (e.g. macrophages, T cells and monocytes). Binding of adenosine to tumor-expressed A2aR promotes tumor cell proliferation and metastasis, whereas ligation of immune cell-expressed A2aR suppresses immune function (see Fig. 1) (Sek et al. 2018). Together with increased extracellular adenosine levels caused by inefficient ATP production by tumor cells, a wide range of tumors exploit this pro tumorigenic and immune suppressive mechanism by expressing extracellular adenosine level increasing enzymes (CD39 and CD73) (Gao et al. 2014). Various A2aR and CD39/CD73 antagonistic therapies have been developed and show encouraging preclinical anti-tumor results warranting the multiple ongoing clinical studies (ClinicalTrials.gov 2019). Noninvasive nuclear imaging could aid in the evaluation of new antagonists and patient selection for A2aR and CD39/CD73 therapy. So far, all imaging studies have focused on A2aR, probably because adenosine targeting tracers risk altering the autoimmunity preventive effects of circulating adenosine. Furthermore, high adenosine levels in plasma could also interfere with targeting of adenosine in the tumor. Papers describing preclinical and clinical A2aR imaging studies have shown impressive results for [11C] Preladenant, [11C] TMSX and 18F-labeled SCH442416-analogs in a number of applications, most of which for intracranial A2aR imaging (Noguchi et al. 1998; Zhou et al. 2017a, 2017b, 2017c, 2017d; Rissanen et al. 2013; Ramlackhansingh et al. 2011; Sakata et al. 2017; Mishina et al. 2007, 2011; Naganawa et al. 2007, 2014; Lahesmaa et al. 2018; Khanapur et al. 2014, 2017; Ishibashi et al. 2018; Ishiwata et al. 2000a, 2002, 2000b, 1997, 2000c). No studies with these tracers have yet been performed to asses tumor expression of A2aR, but as some have already demonstrated to be safe for in human use, translation on short term to the field of oncology should be achievable.</p><!><p>Binding of the 'second wave' co-stimulatory receptor OX40 (also known as CD134) to its ligand OX40L promotes T cell activation (see Fig. 1). Several agonistic biologicals for this stimulatory immune checkpoint have already shown objective responses in phase I clinical trials as mono or combination therapy (Infante et al. 2016; El-Khoueiry et al. 2017). To complement these therapies, nuclear imaging of the T cell activation marker OX40 might be used to predict OX40 agonist responses or to follow treatment responses that focus on T cell activation. One study describing the development and in vivo characterization of a tracer binding OX40 has been published. In 2018, Alam et al. showed that the 64Cu-labeled murine Ab AbOX40 could be used to image OX40 noninvasively and longitudinally (Alam et al. 2018). In this study, dual A20 lymphoma-bearing mice received either an immune stimulant (Cytosine phosphodiester Guanine-oligodeoxynucleotides (CpG-ODN), microbial signature DNA fragments) or vehicle only intratumorally in one tumor, the second tumor was untreated. Two days post treatment, PET imaging demonstrated increased 64Cu-AbOX40 uptake in CpG-ODN treated tumors and their draining lymph nodes compared to vehicle treated and untreated tumors. Furthermore, a notable increased tracer uptake was observed in the tumor draining lymph nodes and spleen 9 days after treatment. These findings demonstrate that anti-OX40 Abs are suitable for in vivo PET imaging of activated T cells.</p><!><p>Besides the targets mentioned above, we believe there are other immune checkpoints which are potentially important for future immune checkpoint inhibition therapies and could become important targets for imaging as well. For example, LAG-3 and TIM-3 play a major role in the activation, proliferation and homeostasis of T cells. Multiple clinical trials are currently ongoing to evaluate the safety and efficacy of pharmaceuticals targeting LAG-3 and TIM-3 and a preclinical study is investigating potential of an imaging tracer directed against LAG-3 (Vivier et al. 2019). Another target of interest is V-domain Ig suppressor of T cell activation (VISTA). Antagonists blocking its inhibitory functions, resulted in increased immune activation in multiple mouse models. Expression of VISTA in immune checkpoint blockade therapy resistant patients could open up alternative treatment options. Finally, preclinical studies have demonstrated that Inducible T cell costimulatory (ICOS or CD278) agonistic therapy improve the efficacy of other immune checkpoint therapies, and imaging could therefore be of interest to predict response or stratify patients for combination therapy.</p><!><p>Current studies demonstrate the wide variety of successful immune checkpoint imaging approaches. As is the case with all radiotracers, high target affinity, stable in vivo behavior, and adequate specificity with minimal uptake in target negative tissue are desired when performing imaging studies. As opposed to imaging of (often) highly upregulated tumor related targets, immune checkpoints are mostly expressed by highly mobile immune cells, or in some cases at physiological levels by tumor cells, and can be dynamic in their expression levels over time. Because of this, determining optimal imaging timepoints during the course of disease is more critical than in imaging studies exploiting the constitutively highly expressed tumor targets. Also, available immune imaging targets in tumor lesions can be low compared to physiological expression levels in other immune-related organs in the body e.g. the spleen. Finally, because of the inherent immunological functions of immune checkpoints, there is a need to verify the effects tracers can exert on their target cell population; e.g. 'on-target off-tumor' immunogenicity or target cell differentiation or even depletion. Thus, to develop tracers for immune checkpoint imaging with favorable characteristics some issues require special attention.</p><p>First, determining in vivo targeting specificity is vital. To asses specificity, blocking studies can be performed. Here, the addition of an excess unlabeled tracer is used to block the labeled tracer from binding to its target in a concentration dependent manner, remaining tracer uptake must then be attributed to aspecific EPR effects. However, this method does not take into consideration dose-dependent biodistribution effects (for instance: sink organs) and specific Fc-mediated uptake, which may also reduce upon injection of an excess of unlabeled antibody. Alternatively, knockout mice can be used; by missing the target protein of interest all eventual uptake can be considered as non-target specific. The limitation here is that due to the genetic modification, developmental changes may occur and the resulting animal might not be sufficiently comparable to the baseline strain. Finally, to asses aspecific uptake it is also possible to use scrambled peptides, isotype control antibodies or other relevant controls depending on tracer moiety.</p><p>Second, because of their physiological role in regulating the immune response, immune checkpoints can be expressed by a wide variety of tissues, including cancer cells, subsets of immune cells, but also non-lymphoid tissues such as activated endothelial cells, brown fat, or duodenum (Heskamp et al. 2019). The physiological target expression throughout the body leads to 'on-target-off-tumor' distribution of tracer. Moreover, as with PD-1, the same immune checkpoint can be expressed by immune cells with immune stimulatory, suppressive or effector functions. Alternatively, certain immune checkpoints, like CD80 and CD86, can both stimulate or inhibit immune responses. Therefore, choosing the animal model and conditions that will yield relevant results is essential. Knowledge of the expression levels on different immune cells and their location is not only essential in order to accurately interpret the acquired PET or SPECT image, but also to find optimal dosing levels. For example, PD-L1 is highly expressed by splenic cells and therefore the spleen acts as a sink organ. As a consequence, at low tracer doses, all injected tracer will accumulate in the spleen, resulting in rapid blood clearance and minimal targeting to other PD-L1 positive tissues like the tumor (Heskamp et al. 2019; Nedrow et al. 2017b). By increasing the tracer dose, spleen uptake can be saturated, resulting in restored circulation time and increased targeting to tumor and other PD-L1 positive tissue. However, this is not the case for all immune checkpoints or their ligands. For example, expression levels of PD-1 are much lower and there is no sink organ affecting the biodistribution of PD-1 targeting tracers, therefore lower doses should be used to prevent saturation of all PD-1 and to obtain high contrast images (Hettich et al. 2016).</p><p>Third, when targeting immune checkpoints expressed on immune cells, the number of immune cells in a tumor may be below detection limit. Therefore, in order to be able to detect these low number of cells, it is essential that the tracer demonstrates high and specific uptake and can be labeled with sufficiently high specific activity. IgG based imaging moieties will result in high absolute uptake in tissue of interest but also a long circulation time, leading to imaging timepoints of 24 h and upward with low signal to background contrast at early time points post injection. They are however widely available and can be easily coupled with various SPECT and PET isotopes. The presence of the Fc region in antibodies has a large influence on its in vivo distribution, for example Fc-mediated recycling and degradation by endothelial and immune cells leads to a long circulation time and high liver uptake. Furthermore, antibodies are potentially immunogenic, depending on the antibody isotype. This can cause anti-target immune activation, complement activation or the formation of anti-drug antibodies, which could lead to target cell depletion, altered tracer pharmacokinetics and serious adverse events when there is cross-recognition of normal proteins; also it limits repetitive imaging in animals. To prevent this, the glycosylation in the Fc domain can be modified, resulting in reduced FcRy mediated uptake (Vivier et al. 2019). Despite these drawbacks, many therapeutic immune checkpoint targeting agents are antibodies and by radiolabeling these agents, they can be used for theranostic purposes. Thus, molecular imaging with therapeutic antibodies is still of great value. However, the use of smaller tracers (nanobodies, minibodies, affibodies, peptides, adnectins etc.) does have some advantages over intact antibodies. For example, it can result in higher target-to-background signals. Because of the rapid pharmacokinetics of smaller tracers, imaging is usually possible within 30 min up to a few hours post injection, while lowering radiation exposure to the patient. Due to the absence of the Fc-region, there is no Fc-mediated (re-)activity. However, these molecules need to be optimized in terms of affinity and specificity in order to result in favorable in vivo behavior. The optimum combination of different conjugations, PEGylation, spacers, chelator combinations, and an isotope matched to the biological half-life of the tracer is paramount to develop a good imaging moiety. PET may also be the preferred imaging method over SPECT because of the higher sensitivity and resolution, especially in the clinical setting. Although in the preclinical setting, SPECT has a higher resolution which can be used to study the heterogeneity within a tumor in more detail.</p><p>Fourth, immune checkpoint-targeting radiotracers should be carefully designed so that they do not interfere with nor alter normal functioning of target immune cells. For example, upon binding of the tracer, irradiation might damage the immune cells and different subsets of immune cells demonstrate differences in radio sensitivity. Potentially this could alter their function, although literature reports T cell labeling with 89Zr up to 0.5 MBq/106 cells without negative effect on viability over a period of 7 days (Bansal et al. 2015; Charoenphun et al. 2015). Moreover, terminally differentiated effector cells have a short biological lifespan. Furthermore, high doses of immune-cell targeting Abs might result in Fc receptor mediated depletion of these immune cells. This can be circumvented by using micro doses of Ab, or by developing tracers with a modified Fc domain, or lacking an Fc domain (Tavare et al. 2014, 2015).</p><p>Finally, in the preclinical setting it is essential to select the right animal model. It is known that different immunocompetent mouse strains show varying immune responses when encountering the same antigen. Many radiotracers described in literature have been tested in immunodeficient mice engrafted with human tumors, and often the radiotracers do not cross react with the murine immune checkpoint. Although this allows a first characterization of the radiotracers, it might be difficult to translate the findings to the clinical setting as immune checkpoints are also expressed on normal tissues. The use of humanized mice (mice transplanted with human immune cells) will overcome some of the limitations. However, these animals still do not have a fully functioning immune system, for example they lack mature T cells, or when T cells are present, graft-versus-host responses might occur. This can be used to collect additional proof of target specificity as observed by Ehlerding et al. in their CTLA-4 imaging studies. However, due to the inherent instability and ultimately lethal complications of such a transplanted foreign immune system, longitudinal studies are challenging. Finally, although these humanized mice develop human immune cells, the other tissues are still completely murine, therefore these models are not relevant for determining pharmacokinetics, biodistribution, immunogenicity or depletion effects. Therefore, many researchers make use of tracers specifically developed to detect murine immune checkpoints as this allows for experiments in animals with a fully functioning immune system.</p><!><p>To realize the full potential of immune checkpoint imaging, it is essential that novel immune checkpoint tracers are developed according to the requirements that serve their (pre) clinical application. The first step to achieve this is to select the optimal tracer for preclinical or clinical research, keeping in mind the specific criteria discussed in the previous section. Once these tracers have been validated, they can play an important role in early drug development, as it can provide information about pharmacokinetics, tumor targeting, and potential off-tumor accumulation which could lead to adverse side-effects of immune checkpoint inhibitors. Longitudinal molecular imaging will also aid in elucidating dynamic expression and interactions of immune checkpoints, and thus aid in better understanding of tumor immunology and providing new insights for therapeutic interventions. When a novel ICI has shown promising anti-tumor effects, imaging can be used to facilitate fast acquisition of preclinical and clinical experimental data on multiple treatment combinations to ultimately design the most rational treatment plan for a specific tumor immunological phenotype. Correct analysis of acquired images hinges on understanding the immune checkpoint expression in relation to therapy. As already shown by Niemijer et al., patient stratification based on whole-body PET imaging is a viable tool to facilitate treatment individualization (Niemeijer et al. 2018). However, in order to achieve this in clinical practice, high sensitivity and specificity, as well as access to practical preparation of the radiopharmaceutical are necessary.</p><p>Given the wealth of unique data that can be derived from in vivo checkpoint imaging prior to and during novel immune therapeutic (combination) strategies, complementary to tissue sampling; it should be stimulated to apply molecular imaging tools in early stages of drug development. Collaborative approaches between pharmaceutical industry and academic partners have shown their impact in studies using currently approved radiolabeled checkpoint inhibitors; and it is envisioned that this fosters similar studies with next generation of immune therapies. In particular, when new clinical trials combining different immune checkpoint therapies are being initiated. With only limited number of patients available to include in these trials (Tang et al. 2018), strict patient screening for treatment eligibility should be applied.</p><!><p>The introduction of immune checkpoint therapies initiated a new era of effective immune therapy. The decision to treat patients with immune checkpoint therapies currently depends on immune checkpoint expression and infiltration of immune cells detected with IHC, which requires invasive biopsies. In this review we have discussed the potential of molecular imaging for immune checkpoint therapy drug development, patient selection, and therapy individualization. For PD-L1 and PD-1, the strength of PET imaging in immunotherapy was recently underlined by first-in-human trials correlating uptake of immune checkpoint tracers with immunotherapy outcome (Niemeijer et al. 2018; Bensch et al. 2018). Since these studies have shown that PET imaging of immune checkpoint expression in tumor lesions is safe and feasible, the road is open for future clinical trials to validate the use of PET as a complementary diagnostic tool to IHC for patient stratification prior to ICI therapy, for treatment monitoring during therapy, and as a tool in early cancer drug development (Graphical abstract). Ultimately, imaging with these new immune checkpoint tracers could aid to speed up development and implementation of effective mono- and combination therapies, treatment of only patients that potentially benefit and preventing severe side-effects in patients that will not benefit from ICI therapy.</p><!><p>Adenosine 2a receptor</p><p>Cluster of Differentiation</p><p>Cytotoxic T lymphocyte associated antigen 4</p><p>Food and Drug Administration (U.S.)</p><p>Fludeoxyglucose</p><p>Forkhead box p3</p><p>Human peripheral blood mononuclear cell</p><p>Immune checkpoint inhibition</p><p>Indoleamine-pyrrole 2,3-dioxygenase 1</p><p>Immunohistochemistry</p><p>Monoclonal antibody</p><p>Magnetic resonance imaging</p><p>NOD.Cg-KitW-41J Tyr+ Prkdcscid Il2rgtm1Wjl/ThomJ</p><p>Non small-cell lung carcinoma</p><p>NOD scid gamma</p><p>Peripheral blood lymphocyte</p><p>Programmed cell Death 1</p><p>Programmed cell Death Ligand 1</p><p>Positron emission tomography</p><p>Single domain antibody</p><p>Single photon emission computed tomography</p><p>Standardized uptake value</p><p>Tumor infiltrating lymphocyte</p><p>Publisher's Note</p><p>Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.</p><p>Peter Wierstra and Gerwin Sandker contributed equally to this work.</p>

PubMed Open Access

Significant Cation Effects in Carbon Dioxide–Ionic Liquid Systems

Carbon dioxide–ionic liquid systems are of great current interest, and significant efforts have been made lately to understand the intermolecular interactions in these systems. In general, all the experimental and theoretical studies have concluded so far that the main solute–solvent interaction takes effect through the anion, and the cation has no, or only a secondary role in solvation. In this theoretical approach it is shown that this view is unfounded, and evidence is provided that, similarly to the benzene–CO2 system, dispersion interactions are present between the solute and the cation. Therefore, this defines a novel site for tailoring solvents to tune CO2 solubility.

significant_cation_effects_in_carbon_dioxide–ionic_liquid_systems

1. Introduction<!><!>1. Introduction<!>Computational Methods<!>2. Results and Discussion<!><!>2. Results and Discussion<!><!>2. Results and Discussion<!><!>2. Results and Discussion<!><!>3. Conclusions

<p>Among their numerous potentially advantageous properties,[1–5] ionic liquids (ILs) exhibit unique properties in CO2 absorption.[6] Although they dissolve CO2 much better than other gases, as shown by Brennecke and co-workers,[7] they are practically insoluble in supercritical CO2, which makes them perfect candidates not only for capturing CO2 from industrial waste gases,[8,9] but also for gas separations, extraction processes,[7] and bi- or multiphase catalysis involving CO2.[6] For the improvement of these applications, an understanding of the solubility of CO2 is required through the identification of the CO2–IL interaction sites.[10] Accordingly, several experimental studies were performed to compare the Henry's law constants for different ILs,[6,8,11–13] and based on the observed trends a picture of CO2 solvation in ILs was established, which could be justified by the corresponding theoretical investigations.[6,11–14] The general wisdom of these studies is that while the anion plays a crucial role in the solute–solvent interplay, the cation–CO2 interaction is rather limited to small contributions from the side chain,[12] and so far no significant direct effect of the cationic head group has been reported. Accordingly, the formation of a hydrogen-bond-like[15–17] interaction in imidazolium-based ILs between the CO2 oxygen atoms and the cationic ring hydrogen atoms was excluded, since neither the Henry's law constants in the experiments changed, nor were any discrepancies noticed in the microscopic structure of the solvent in classical molecular dynamics (MD) simulations through the exchange of the most acidic (thus, most likely interacting) H2 atom by a methyl group,[11] thus inferring a certain unimportance of the cation.</p><p>The anion–CO2 interaction can be described as a Lewis acid–base reaction, and accordingly, by the increasing basicity of the anion, this interaction becomes stronger.[10] Interestingly, in the presence of basic anions the formation of carbenes may also occur by proton transfer from the cation to the anion,[18–21] and since carbenes are known to react with CO2 yielding imidazolium carbonates,[22] in the case of sufficiently basic anions the formation of such structures is expected. In agreement, the chemical absorption of CO2 in 1,3-dialkylimidazolium acetates has been suggested based on the significantly increased solubility of CO2 in these ILs,[23] and Rogers and co-workers[24] (and later several other groups)[25–28] recently revealed the formation of 1,3-imidazolium carboxylates in the same system. According to the above information on IL–CO2 systems and carbene formation, it is reasonable to assume the mechanism depicted in Figure 1: physical absorption of CO2 in the 1,3-dialkylimidazolium acetate, followed by reaction of the solute with the carbene that is accessible in these ILs. However, to improve and to exploit this reaction more effectively, a more detailed mechanistic insight is required for each step of the process.</p><!><p>Mechanistic picture of CO2–[CnC1Im][OAc] systems (Im=imidazolium). The system investigated herein is highlighted by a thicker frame.</p><!><p>In this theoretical study we investigate the initial step, the physical absorption of CO2 in 1-ethyl-3-methylimidazolium acetate ([C2C1Im][OAc]), as the first of a series of investigations on this apparently interesting but rather complex system (Figure 1). Moreover, due to the higher basicity of the acetate anion, increased anion–CO2 interactions are expected.[10] Therefore, in this system the role of the cation in the solvation of the CO2 should be even lower, which allows a careful view in revisiting the presence of cation–CO2 interactions in imidazolium-based ILs in general.</p><!><p>Ab initio molecular dynamics (AIMD) simulations[29–31] were carried out with periodic boundary conditions, which—in contrast to classical MD simulations based on a force field—allow the monitoring of unforeseen changes in the electronic structure. Given that the bending of the CO2 is of high importance in the anion–CO2 interaction[32] (note that CO2 is usually kept linear in force fields),[12] there is a need for the description of the electronic structure in extreme molecular interactions, and thus the advantage of AIMD is clearly indicated.</p><p>The simulated system was built by inserting a single CO2 molecule into the simulation box, which was obtained in a series of previous simulations by our group on the neat IL, and successfully reproduced many of its experimental physical properties.[33] The resulting system of 36 ion pairs and one CO2 molecule was equilibrated for 5 ps in an NVT ensemble employing a massive Nosé–Hoover thermostat, and then simulated for 68 ps at 350 K in an NVT ensemble by applying a regular Nosé–Hoover thermostat, by the CP2k program package,[34] and by using the BLYP-D functional, the MOLOPT-DZVP-SR-GTH basis sets, and GTH pseudopotentials. The applied functional—in significant difference to previous AIMD studies on IL–CO2 systems[35,36]—also includes Grimme's most recent dispersion correction (D3),[37,38] which is essential in IL systems.[37,39–41] The analysis of the trajectories was performed with TRAVIS.[42]</p><p>Static quantum chemical calculations were carried out by applying the BLYP-D/def2-TZVPP, BLYP/def2-TZVPP, and (RI)MP2/def2-TZVPP methods and basis sets by the TURBOMOLE 6.0[43] (applying increased convergence criteria on the optimization of 10−4 a.u., and on the SCF of 10−8 hartree) and SNF[44] program packages, and M06-2X, B97-D, B3LYP, and MPW1K DFT with the 6-311+G** basis set by the Gaussian 09 program package.[45]</p><!><p>On the basis of the radial distribution functions (RDFs), the acetate oxygen–CO2 carbon distances are the shortest (2–300 pm), providing a very pronounced peak (black line in Figure 2 B) similar to that found before in other ILs.[11,14] However, our results show noticeable deviations compared to a previous AIMD study on the same IL containing 50 mol % CO2.[35] Here, the C(CO2)–O([OAc]−) distances are longer (black line in Figure 2 B) and also the CO2 bond angles are larger, although the bending is still more pronounced than that in the gas phase (Figure 2 D). These differences may originate from the different molar ratios (1:1[35] vs. 1:36), the different simulation temperature (298[35] vs. 350 K), or the much shorter simulation time (12[35] vs. 68 ps) and the lack of proper account for dispersion interaction in the previous AIMD study.[35] In full agreement, by static calculations on isolated acetate–CO2 assemblies lacking dispersion correction we observed, for example, the shortening of the distances between the aforementioned two atoms (by ca. 10 pm, see the Supporting Information), which clearly affects the outcome of the AIMD simulations as well. Nevertheless, despite these differences, the entries in the lower left part of the combined distribution function (CDF) in Figure 3 A clearly indicate that whenever the anion's oxygen atom is close to the CO2's carbon atom, the bending of the CO2 is increased, which—together with the observed short anion–CO2 distances—points to the importance of the anion–CO2 interactions.</p><!><p>Radial distribution functions, g(r), between centers of mass (COMs) (A), measured from the C atom (B) and from the O atom (C) of the CO2, and the angular distribution of CO2 in the gaseous phase and in the IL (D). mim=methylimidazolium, term=terminal.</p><p>Combined distribution function showing the CO2 bond angle against the depicted distances.</p><!><p>Surprisingly, the cationic centers of mass (COMs) are at similar distances to the solute as the anionic ones (Figure 2 A), while the corresponding peak is higher, thus showing that the cation also contributes to the solvent–solute interactions by providing more neighbors (ca. five versus the ca. one anion). Interestingly, although such pronounced peaks have previously been observed in cation–CO2 pair correlation functions, they were related to "packing effects" rather than to solute–solvent interactions. However, by comparing the spatial distribution functions (SDFs) of the two ions, a different viewpoint can be obtained (Figure 4). The interaction with the anion is clearly directed to the CO2's carbon atom; thus, the acetate ions are located mainly in a thin specific ring around the solute. The cations can be observed in a similarly structured manner around the CO2, but these regions of interaction cover its whole surface; thus, a picture of a cation cage emerges (Figure 4 B). This high local structuring of the ions around the solute is in contrast to the picture that CO2 solely occupies already existing voids in the IL.[6,11]</p><!><p>Spatial distribution of the anionic (A) and the cationic COMs (B), and the terminal carbon atom of the cationic ethyl group (C) around the CO2.</p><!><p>Given that the approach of the acetate anion toward the solute polarizes the CO2 by bending it into a negatively charged carboxylate group, one may infer that this bending strengthens the interaction with the cations, as was found in an analogous reaction between amines and CO2 in imidazolium-based ILs.[46] Surprisingly, the CDF in Figure 3 B clearly shows that the closer the solute is to the cation, the less bent it is, as for the lower C(CO2)–C2 distances there are no entries corresponding to lower O–C–O angles of the solute. Thus, instead of cooperation, competition is indicated between the anion and the cation for interacting with the CO2. The finding that despite this competition the aforementioned cation cage is formed clearly shows the significance and strength of the cation–CO2 interactions.</p><p>Although there is a large peak in the RDF between the H5 and the CO2's oxygen atoms, the large (above 200 pm) distances between any ring hydrogen atoms and the solute oxygen (Figure 2 C) support the previous findings[11] in pointing to the lack of hydrogen bonding with CO2 in such systems. These substantial distances in the H2 RDF (black line in Figure 2 C), together with the lack of any significant peaks in it, also perfectly explain why the methylation at position 2 has no effect on the CO2 solubility.[11] Similarly to Costa Gomes and co-workers,[12] a pronounced side-chain CO2 peak was obtained (dashed line in Figure 2 B), which suggests that this moiety also has some impact. However, the SDF of the terminal side-chain carbon around the solute exhibits significantly less structuring than that of the cationic COM (Figure 4 C), whereas the C2([C2mim]+)–C(CO2) distances (dotted line in Figure 2 B) show that the CO2 molecule is, in fact, similarly close to the cationic ring.</p><p>Furthermore, according to the CDFs shown in Figure 5, the CO2 is strictly above the ring of the nearby cations, and oriented mostly in a parallel fashion to the ring plane, although perpendicular conformers can also be observed. This on-top arrangement of the CO2 around the nearby imidazolium cations has been observed before,[47] and was related to the competition between the anion and the solute for interacting with the H2 atom. Clearly, this competition has an influence; however, we would like to point out that these findings also indicate the presence of a dispersion interaction with the cationic π system, which is analogous to that in the benzene–CO2[48] and pyridine–CO2[49] systems. The similar ring–CO2 distances (328.6 pm for benzene at the MP2/aug-cc-pVTZ level,[48] and ca. 360 pm in the present simulation) are also noteworthy. As mentioned above, the interaction with the cation is apparently enhanced by the linearity of the CO2; thus, the lack of a proper dispersion description in the simulations may result in the overestimation of the CO2's bending. Although this picture provides a possible explanation for the deviations from the previous study,[35] it should also be kept in mind that the different molar ratios may alter the number of available interacting cations.</p><!><p>Combined distribution functions representing the orientation of the CO2 with respect to the cationic ring, based on the depicted geometrical measures.</p><!><p>To further analyze the interaction between carbon dioxide and the imidazolium π system, static quantum chemical calculations were carried out by a number of different theoretical methods (see Computational Methods) on the CO2–1,3-dimethylimidazolium cation model system. The geometry of the obtained three minima (Figure 6) further stresses the analogy with the aforementioned benzene–CO2 interplay.[48,50] The most stable minimum (1) possesses the CO2 molecule in the ring plane, apparently in interaction with the H2 atom. The lack of this structure in the present AIMD trajectory, and also in the previous MD simulations, is due to the competition between the anion and the solute for this position (cf. with the neat IL).[33] The two other structures (2 and 3) are about 3 and 6 kJ mol−1 less stable, with the CO2 positioned approximately 320 pm above the cationic ring in either a perpendicular (2) or a parallel (3) fashion. The Bader analysis[51] of both 2 and 3 supports the presence of an interaction between the CO2 and the cationic π system, by exhibiting unprecedented bond critical points between the cation's nitrogen atoms and the CO2's oxygen atoms. The bond critical points between the methyl hydrogen atoms and the solute oxygen atoms allow concluding interactions with the methyl groups of the cation. The relative energies are comparable in all methods applied, but the importance of the dispersion's proper treatment was again observed, as during the geometry optimizations by the BLYP and B3LYP functionals either the rearrangement of 2-like and 3-like structures to 1 was observed, or the CO2–cation distance increased to 1300 pm (for more data, see the Supporting Information). Although the cation–CO2 interaction energies are somewhat lower than those for the anion–CO2, the cationic cage around the solute suggested by the AIMD calculations makes it necessary to consider the effect of these π interactions.</p><!><p>Obtained structures for the system composed of a 1,3-dimethylimidazolium cation and a CO2 molecule.</p><!><p>In this theoretical study the interactions between CO2 and imidazolium-based IL cations have been investigated by AIMD simulations and static quantum chemical calculations, on the one hand to provide insight into the first step of CO2 absorption in 1,3-dialkylimidazolium acetates, and on the other hand to revisit those results in the literature in which the main solute–solvent interaction in IL–CO2 systems in general takes effect through the anion.</p><p>Undeniably, there is a strong anion effect and a moderate side-chain effect on CO2 solvation in ILs, as was proposed previously by experimental (Henry's law constants) and theoretical (classical MD) studies. However, even in the case of such a strong anion–CO2 interplay as that with the acetate anion, the occurrence of an attractive interaction between the cationic π system and the solute has been evidenced in the study reported herein. Although nonaromatic cations may form other kinds of interactions as well,[52] and the corresponding interaction energies may therefore be similar, our results, and the fact that imidazolium-based ILs dissolve more CO2 than pyrrolidinium ones,[13] indicate that boosting the CO2–aromatic interactions may indeed increase CO2 solubility in ILs. This knowledge may allow not only a deeper understanding of the solubility of CO2 in imidazolium-based ILs, but also may provide novel perspectives in tailoring[52] of ILs by incorporating aromatic units into the ions, for example, by using aromatic anions or aryl-functionalized side chains. Such modification may allow the improvement of nonreactive CO2 capture processes, and may also open paths to the development of ILs that are soluble in supercritical CO2.</p>

PubMed Open Access

Metabolomic profiling reveals a role for CPT1c in neuronal oxidative metabolism

BackgroundCarnitine Palmitoyltransferase-1c (CPT1c) is a neuron specific homologue of the carnitine acyltransferase family of enzymes. CPT1 isoenzymes transfer long chain acyl groups to carnitine. This constitutes a rate setting step for mitochondrial fatty acid beta-oxidation by facilitating the initial step in acyl transfer to the mitochondrial matrix. In general, neurons do not heavily utilize fatty acids for bioenergetic needs and definitive enzymatic activity has been unable to be demonstrated for CPT1c. Although there are studies suggesting an enzymatic role of CPT1c, its role in neurochemistry remains elusive.ResultsIn order to better understand how CPT1c functions in neural metabolism, we performed unbiased metabolomic profiling on wild-type (WT) and CPT1c knockout (KO) mouse brains. Consistent with the notion that CPT1c is not involved in fatty acid beta-oxidation, there were no changes in metabolites associated with fatty acid oxidation. Endocannabinoids were suppressed in the CPT1c KO, which may explain the suppression of food intake seen in CPT1c KO mice. Although products of beta-oxidation were unchanged, small changes in carnitine and carnitine metabolites were observed. Finally, we observed changes in redox homeostasis including a greater than 2-fold increase in oxidized glutathione. This indicates that CPT1c may play a role in neural oxidative metabolism.ConclusionsSteady-state metabolomic analysis of CPT1c WT and KO mouse brains identified a small number of metabolites that differed between CPT1c WT and KO mice. The subtle changes in a broad range of metabolites in vivo indicate that CPT1c does not play a significant or required role in fatty acid oxidation; however, it could play an alternative role in neuronal oxidative metabolism.

metabolomic_profiling_reveals_a_role_for_cpt1c_in_neuronal_oxidative_metabolism

Background<!>Animals<!>Western blot analysis<!>Metabolomic measurements and profiling<!><!>Statistical analysis<!>Carnitine Palmitoyltransferase-1c KO mice<!><!>Fatty acid oxidative metabolites show no difference in overall trend in CPT1c KO mice<!><!>Fatty acid oxidative metabolites show no difference in overall trend in CPT1c KO mice<!>Loss of CPT1c results in decreased levels of endogenous endocannabinoids<!><!>Loss of CPT1c results in increased levels of glutathione<!><!>Role of CPT1c in behavior and physiology<!>Endocannabinoid regulation of food intake<!>Glutathione and redox metabolism<!>Conclusion<!>Abbreviations<!>Competing interests<!>Authors’ contributions

<p>Although the mammalian brain is lipid rich and mutations in lipid metabolizing enzymes result in debilitating neurological disease, neurons are generally not thought to rely on mitochondrial fatty acid beta-oxidation for bioenergetic requirements. Neurons instead mainly utilize the oxidation of glucose for most of their bioenergetic needs, although, during prolonged fasting, ketone bodies (i.e. acetoacetate and beta hydroxybutyrate) can also be used [1]. Most neurons have a low amount of the rate-setting enzymes in mitochondrial long chain fatty acid catabolism, namely, the malonyl-CoA sensitive Carnitine Palmitoyltransferase 1 (CPT1a and CPT1b) enzymes which limit most neurons potential for mitochondrial fatty acid beta-oxidation [2].</p><p>Carnitine acyltransferases are enzymes that catalyze the exchange of acyl groups between carnitine and Coenzyme A (CoA) to facilitate the transport acyl chains between the cytoplasm to the mitochondrial matrix [3]. CPT1 isoenzymes (EC 2.3.1.21) preferentially are positioned on the outer mitochondrial membrane and transfer long chain acyl groups from CoA to carnitine. CPT1a and CPT1b are malonyl-CoA sensitive and therefore inhibited when malonyl-CoA levels are high (e.g. during high glucose flux). The malonyl-CoA insensitive CPT2, on the other hand, is located in the mitochondrial matrix and reversibly transfers the acyl chain back to CoA to facilitate beta-oxidation. Although neurons have a relative dearth of CPT1a and CPT1b [2], they express a CPT1 homologue, CPT1c [4].</p><p>CPT1c has a high primary amino acid sequence similarity and identity to the canonical CPT enzymes. Therefore, it was surprising that definitive acyltransferase activity or enhanced oxidation of fatty acids could not be shown for CPT1c [4-6]. CPT1c KO mice exhibit both behavioral and metabolic deficits [6-9]. Over-expression of CPT1c in the brain of developing transgenic mice results in microencephaly [10]. Therefore, it is clear that CPT1c plays an important role in brain function. Although there were several metabolites identified that have been altered after over-expression [10,11] or knockout of CPT1c [7], the reaction that CPT1c catalyzes has remained elusive.</p><p>Here we used an unbiased metabolomic approach to broadly understand the consequence of CPT1c deletion to gain insight into the biochemical and physiological roles of CPT1c function. Similar to previous work in heterologous systems, we did not see changes consistent with a role for CPT1c in long chain fatty acid beta oxidation. However, there were changes in several fatty acid derived metabolites including endocannabinoids, which may explain the suppressed food intake in these models. Also, some of the most abundant changes were in redox biochemistry consistent with several models of CPT1c function recently proposed.</p><!><p>Mice with a targeted knockout of exons 1 and 2 of the cpt1c gene were propagated and genotyped as previously described [5,6]. Mice were fed a standard lab chow (Harlan 2018) after weaning. All procedures were performed in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and under the approval of the Johns Hopkins Medical School Animal Care and Use Committee.</p><!><p>A polyclonal rabbit antibody against CPT1c was used as a primary antibody for CPT1c detection in WT and CPT1c KO mice [5,6]. Anti-rabbit horseradish peroxidase (HRP) was used as a secondary antibody, and the blots for CPT1c were developed using ECL reagent. Mouse monoclonal anti-HSC70 (Santa Cruz biotech) and mouse monoclonal anti beta-actin (Sigma) was used as primary antibodies for loading control. Cy3 conjugated fluorescent secondary antibody was used for both HSC70 and beta-actin antibodies.</p><!><p>Unbiased metabolomics analysis of whole brain samples from WT and CPT1c KO mice (n=8/group) that were fasted overnight was performed using liquid chromatography/tandem mass spectrometry (HPLC/MS/MS2) and gas chromatography/mass spectrometry (GC/MS) platforms. The platform was able to screen and identify several metabolites in multiple classes, such as amino acids, lipids, and nucleotides. A complete list of the metabolites identified in this study is given in Tables 1, 2, 3 and 4. General platform methods about metabolomic measurements and profiling are described in the metabolomic study done by Eckel-Mahan et al. [12]</p><!><p>Biochemicals involved in lipid metabolic pathways</p><p>Biochemicals in the amino acid and peptide pathways</p><p>Biochemicals from the carbohydrate and energy pathways</p><p>Biochemicals in nucleotide, cofactors and vitamins, and xenobiotic Pathways</p><!><p>Pair-wise comparisons between CPT1c WT and KO were performed using Welch's two-sample t-tests. From the p-values, any value below the significance level of 0.05 was interpreted as statistically significant.</p><!><p>Although CPT1c is widely expressed in transformed cells and tumors [13], we have only been able to reliably detect CPT1c in neurons in vivo. To understand the endogenous function of CPT1c, we performed metabolomic profiling on brains of CPT1c KO mice and their littermate controls. Therefore, we collected and snap froze the brains of CPT1c KO and WT littermate sex matched adult mice after an overnight fast. Western blot analysis of WT and CPT1c KO mice showed that KO mice were indeed completely deficient of CPT1c (Figure 1A). These samples were then homogenized and the small organic metabolites were extracted and analyzed by a mixture of GC-MS and LC-MS/MS by a commercial supplier of metabolomic analyses (Figure 1B). Below, we detail the changes in steady-state biochemicals between WT and KO brains that were identified through an unbiased metabolomic screen.</p><!><p>CPT1c KO mice and metabolomic profiling. (A) CPT1c protein from homogenized brains of WT and CPT1c KO mice were analyzed by western blot using the anti-CPT1c antibody. Hsc70 and Actin were used for loading controls. (B) A schematic pathway of metabolomic profiling for KO and WT brains. A commercial supplier of metabolic analysis homogenized 8 brain samples from independent mice to extract organic metabolites for performing unbiased metabolomic analysis using a mixture of GC-MS and LC-MS/MS.</p><!><p>Given the high primary amino acid homology of CPT1c to other CPTs, it would follow that CPT1c may be involved in fatty acid beta oxidation or at least in long chain acyl-CoA metabolism. If CPT1c was involved in fatty acid oxidation, we would expect that the deletion of CPT1c would decrease the level of acyl-carnitines and potentially increase the levels of other long chain acyl-CoA dependent biosyntheses. A broad range of lipid species were identified in the metabolomic screen (Table 1). No changes were seen in oleoyl-carnitine, beta-hydroxybutyrate, or acetyl-carnitine, as we would have expected (Figure 2A). However, the metabolomic analysis did show that free carnitine, 3-dehydrocarnitine, glutaroylcarnitine, and betaine were significantly changed (Figure 2A).</p><!><p>Loss of CPT1c results in decreased free carnitine and no change in fatty acid oxidative metabolites in the brain. (A) Biochemicals involved in carnitine, amino acid, and fatty acid metabolism from WT and CPT1c KO brains were compared through metabolomic analyses, revealing a statistically significant change in levels of free carnitine (p=0.084), 3-dehydrocarnitine (p=0.0103), glutaroylcarnitine (p=0.0244) and betaine (p=0.0383). (B) Schematic of biochemical pathways altered in CPT1c KO mice. Based on this schematic pathway, glutaroyl carnitine and betaine may affect the level of free carnitine, since these biochemicals play a role in carnitine biosynthesis.</p><!><p>Among the metabolites that showed a statistically significant difference, only 3-dehydrocarnitine increased in CPT1c KO mice while glutaroyl carnitine, betaine and free carnitine decreased. Glutaroyl carnitine and betaine are biochemicals that are involved in carnitine biosynthesis (Figure 2B; Table 2). Glutaroyl carnitine is involved in lysine metabolism, which is one of the amino acids that is used to synthesize carnitine. In the carnitine biosynthesis pathway, betaine takes the form of butyrobetaine to synthesize L-carnitine [14]. As a result, it is possible that the decrease in glutaroyl carnitine and betaine could have caused free carnitine levels to decrease in CPT1c KO mice. Previous studies also tested hypothalamic and cortical explants from WT and CPT1c KO mice for their ability to oxidize fatty acids, but there was no evidence that unique properties in neurons existed to allow activation of fatty acid oxidation by CPT1c [5]. CPT1c over-expressed in heterologous cells in vitro also did not show a change in fatty acid oxidation [5]. Therefore, our results remain consistent with previous findings that CPT1c, although it is highly homologous with its isoforms CPT1a and CPT1b, does not participate substantially in neuronal mitochondrial fatty acid oxidation.</p><!><p>Several studies have investigated the neurological role of endocannabinoids on food intake [15]. A study investigated the role of endocannabinoids in regulating food intake in the tongue, gut and different brain regions, suggesting that the cannabinoid system plays a role in modulating the activity of neural pathways that regulate food intake and energy expenditure [15]. The brain cannabinoid system, as shown in Figure 3B, regulates food intake through the interaction of endogenous ligands and cannabinoid receptors. From our metabolomic analyses, there was a significant decrease in palmitoylethanolamine and a trend for a decrease in 2-oleolylglycerol in CPT1c KO mouse brains compared to WT mouse brains (Figure 3). There was no significant difference between WT and CPT1c KO mice for free nonesterified fatty acids (Table 1). Among the metabolites shown in Figure 3A, eicosapentaenoate and palmitoylethanolamine showed a significant decrease in CPT1c KO mice with a p-value of 0.0236 and 0.0331, respectively. There was also a slight increase in ethanolamine between WT and CPT1c KO mice, and decrease in 2-oleoylglycerol (p=0.0769), an endogenous cannabinoid (CB) CB-1 agonist (Figure 3A).</p><!><p>Loss of CPT1c results in decreased endocannabinoids in the brain. (A) Biochemicals involved in fatty acid biochemistry from WT and CPT1c KO mouse brains were compared to determine if metabolomic analyses showed any statistically significant changes. There was an overall decreasing trend in endocannabinoids in CPT1c KO mice. Specifically, eicosapentaenoate (p=0.0236) and palmitoylethanolamine (p=0.0331) significantly decreased in CPT1c KO mice. (B) A schematic of how a decrease in endocannabinoids can induce a decrease in food intake by interacting with CB1 and CB2 cannabinoid receptors.</p><!><p>The oxidized form of GSH (GSSG) and 5-oxoproline, biochemicals involved in the gamma-glutamyl redox cycle, resulted in a statistically significant difference in CPT1c KO mice (Table 2). GSSG and cysteine-glutathione disulfide levels increased while 5-oxoproline levels decreased in CPT1c KO mice (Figure 4A). Based on the schematic redox pathway shown in Figure 4B, our results suggest that CPT1c may play a role in oxidative metabolism. This is consistent with findings in cancer metabolism. Zaugg et al. depleted the levels of CPT1c in MCF-7 cells to determine whether these cells were sensitive to oxidative stress. Hypoxia was used as a stress inducer, and they found that CPT1c depletion caused an increased sensitivity to oxidative stress, implying that CPT1c may play a crucial role in protecting the cells from stress from the environment [13]. Furthermore, the loss of CPT1c resulted in an increase in ceramides [7,8], a key mediator of oxidative stress [16,17]. However, the mechanism and role of CPT1c in oxidative metabolism remains unknown.</p><!><p>Loss of CPT1c results in elevated oxidative demands in the brain. (A) In a comparison of biochemicals involved in redox homeostasis in WT and CPT1c KO mouse brains, GSSG and 5-oxoproline were statistically significant. GSSG levels increased in CPT1c KO mice with a p-value of 0.0307, while 5-oxoproline decreased in KO mice (p=0.0291). The biochemicals shown displayed an overall increasing trend in CPT1c KO mice. (B) A schematic of the gamma-glutamyl redox cycle. Based on the pathway, an increase in the biochemicals from Figure 4A may cause the cells to become more sensitive to oxidative stress.</p><!><p>Carnitine acyltransferases are enzymes that catalyze the exchange of acyl groups between carnitine and CoA to facilitate the transport of acyl groups from the cytoplasm to the mitochondrial matrix. Carnitine acetyltransferase (CRAT) and carnitine octonyltransferase (CROT) facilitate transport short- and medium-chain acyl-CoA, while CPT1 facilitate transports long chain acyl-CoA to the mitochondria. CPT1 enzymes are encoded by three genes in mammals that are localized in different tissues and have different properties. CPT1a, which is enriched in the liver, has been heavily studied due to its crucial role in β-oxidation and human fatty oxidation disorders (OMIM #255120) and is lethal when knocked out in mice [18]. CPT1b is localized mainly in the muscle and is a regulator for the use of fatty acids in muscle and is also lethal when knocked out in mice [19]. These two enzymes, which are present on the outer mitochondrial membrane, play a critical role in regulating and facilitating fatty acid beta-oxidation.</p><p>The brain specific CPT1c is highly homologous to its closely related genes, CPT1a and CPT1b [4]. However, despite its high homology, CPT1c does not catalyze acyl transfer from long chain acyl-CoA to carnitine [4-6]. Other distinguishing properties of CPT1c include a longer C-terminus and localization in the endoplasmic reticulum (ER) instead of the mitochondria [11]. Although it does not facilitate acyl transfer in the cell, CPT1c most likely remains sensitive to the endogenous allosteric CPT1 inhibitor, malonyl-CoA, binding with a similar affinity as CPT1a [4,6]. Moreover, while other isoenzymes are expressed in a broad range of organisms, CPT1c seems to have risen late in evolution, raising the question whether CPT1c has a specific role in mammalian brain function.</p><p>Several studies used CPT1c knockout (KO) and CPT1c transgenic mice to investigate the role of CPT1c in the CNS. Knockout studies showed that loss of CPT1c did not affect the viability or fertility of the mice, but resulted in a suppression in food intake and decrease in body weight when they were fed a normal or low-fat diet [6,9]. Paradoxically, when high fat diet was given to CPT1c KO mice, they exhibited diet-induced obesity which ultimately resulted in a diabetic phenotype [5,6]. Even though fatty acid oxidative metabolites showed no significant change based on the metabolomic analysis, due to a decrease in peripheral energy expenditure CPT1c KO mice were more susceptible to obesity and diabetes when fed a high fat diet. This suggests that CPT1c has a hypothalamic function in protecting the body from adverse weight gain when the mice were fed a high fat diet. Transgenic CPT1c mice (CPT1c-TgN), on the other hand, which allowed conditional expression of CPT1c in a tissue-specific manner via cre-lox recombination, showed enhanced expression of CPT1c and they were protected from diet-induced obesity even on a high-fat diet [10].</p><p>CPT1c KO mice also showed impaired spatial learning [7]. Cpt1c deficiency was shown to alter dendritic spine morphology by increasing immature filopodia and reducing mature mushroom and stubby spines. Compared to WT mice, CPT1c KO mice showed a higher escape latency, implying that they had a delay in the acquisition phase [7]. Based on this study, CPT1c deficiency interfered with consolidating new information but did not affect retaining information or motor behavior. As a result, there may be other physiological roles of CPT1c in addition to regulating food intake and energy expenditure consistent with its broad expression throughout the nervous system [7].</p><!><p>Endocannabinoids are endogenous ligands that bind to cannabinoid receptors to regulate many aspects of physiology and behavior. Specifically, the brain endocannabinoid system regulates food intake via the hypothalamus, where it activates necessary mediators to induce appetite after a short-term food deprivation. CB1 receptor KO mice showed reduced food intake, similar to CPT1c KO mice [20,21]. Based on our results, CPT1c could be interacting with the cannabinoid system, causing an overall decreasing trend in endocannabinoids in CPT1c KO mice. In this context, the loss of CPT1c could have influenced the endocannabinoid system and its function to regulate food intake and body weight, which may explain the suppressed food intake in CPT1c KO mice [5,9]. Therefore, a decrease in endocannabinoids based on metabolomic profiling may suggest a putative role of the endocannabinoid system in suppressing food intake in CPT1c KO mice. However, it is unclear if CPT1c affects endocannabinoid metabolism directly or more likely indirectly by altering neuronal specific fatty acid metabolism.</p><!><p>Neurons are particularly sensitive to oxidative stress and damage caused by reactive oxygen species (ROS). On the cellular level, there are many endogenous metabolic stress inducers, such as ROS produced from the mitochondria and cytosolic enzymes, such as cyclooxygenase and lipoxygenase. There are also various exogenous conditions that can also promote the level of ROS species to increase, such as H2O2 and hypoxia, that induces irreversible cellular damage or cell death. As shown by the pathway in Figure 4B, reduced glutathione (GSH) and oxidized glutathione (GSSG) are tightly regulated in order to maintain cellular redox homeostasis and to protect the cells from oxidative damage [17]. Carrasco et al. showed that CPT1c expression correlated with ceramide production and loss of CPT1c resulted in reduced ceramide levels. [7]. A recent study on the role of CPT1c in cancer cells in response to metabolic stress showed that CPT1c could participate in protecting cells from stress. In addition, they postulated that metabolic stress could alter regulation of the CPT1c gene, reducing ATP production and increasing sensitivity towards metabolic stress [13]. Here, we showed that CPT1c deficiency results in an increased oxidative environment. This may indicate that although CPT1c does not contribute in large part to beta-oxidation, it may be involved in other neuron specific oxidative metabolism. Alternatively, CPT1c may need to be activated in a yet to identified stress-induced manner. Barger et al. [22] showed that CPT1c was required for leukemia growth under low glucose conditions. Therefore, CPT1c may have a context dependent role in fatty acid catabolism. Although here we show that CPT1c could play a role in oxidative stress, the precise role of CPT1c in relation to oxidative stress remains unknown.</p><!><p>Unbiased metabolomic profiling of steady-state metabolites in WT and CPT1c KO brains revealed subtle changes in a broad range of metabolites in vivo. The metabolic alterations are not consistent with CPT1c playing a role in beta-oxidation or a large non-redundant role in bioenergetics.</p><!><p>WT: Wild-type; KO: Knockout; CPT1: Carnitine Palmitoyltransferase 1; CPT2: Carnitine Palmitoyltransferase 2; CoA: Coenzyme A; CB: Cannabinoids; GC: Gas chromatography; MS: Mass spectrometry.</p><!><p>The authors declare that they have no competing interests.</p><!><p>MJW conceived of the project, collected samples and aided in writing. JL interpreted results and wrote the manuscript. All authors read and approved the final manuscript.</p>

PubMed Open Access

Discovery of potent, novel Nrf2 inducers via quantum modeling, virtual screening and in vitro experimental validation

Nuclear factor erythroid 2-related factor 2 (Nrf2) is the master transcription factor of the antioxidant response element (ARE) pathway, coordinating the induction of detoxifying and antioxidant enzymes. Nrf2 is normally sequestered in the cytoplasm by Kelch-like ECH associating protein 1 (Keap1). To identify novel small molecules that will disturb Nrf2:Keap1 binding and promote activation of the Nrf2-ARE pathway, we generated a quantum model based on the structures of known Nrf2-ARE activators. We used the quantum model to perform in silico screening on over 18 million commercially available chemicals to identify the structures predicted to activate the Nrf2-ARE pathway based on the quantum model. The top hits were tested in vitro and half of the predicted hits activated the Nrf2-ARE pathway significantly in primary cell culture. In addition, we identified a new family of Nrf2-ARE activating structures that all have comparable activity to tBHQ and protect against oxidative stress and dopaminergic toxins in vitro. The improved ability to identify potent activators of Nrf2 through the combination of in silico and in vitro screening described here improves the speed and cost associated with screening Nrf2-ARE activating compounds for drug development.

discovery_of_potent,_novel_nrf2_inducers_via_quantum_modeling,_virtual_screening_and_in_vitro_experi

Introduction<!>Summary of Constellation software<!>Neuronally enriched and mixed neuron/astrocyte culture preparation<!>Dopaminergic Neuronal Culture preparation<!>Viability Assays<!>hPAP activity assays and histochemistry<!>NQO1 histochemistry<!>Tyrosine Hydroxylase Histology<!>Statistics<!>Identification of novel molecules<!>Secondary Screen on 822 family molecules<!>Activation is Nrf2-dependent<!>Astrocytes and neurons respond independently to 822-dependent Nrf2-ARE activation<!>Nrf2-ARE dependent protection from H2O2-induced neurotoxicity<!>Protection of dopaminergic neurons from MPP+ toxicity in primary midbrain cultures<!>Discussion

<p>Nuclear factor erythroid 2-related factor 2 (Nrf2) is a basic leucine zipper transcription factor that binds the cis-acting antioxidant response element (ARE) regulatory sequence(1-3). Binding of Nrf2 to the ARE results in a coordinated upregulation of various phase II detoxifying and antioxidant enzymes, including those involved in glutathione synthesis (glutamine cysteine ligase), redox regulation (catalase and peroxiredoxin), and quinone recycling [NAD(P)H dehydrogenase (quionone 1) (NQO1)](4, 5).</p><p>Activation of the Nrf2-ARE pathway has been shown to be protective in a variety of disorders, including in mouse models of neurodegenerative disorders (6-8). This is of special interest in the development of therapeutics for Parkinson's disease (PD), which is characterized by high levels of oxidative stress (9-11). Unfortunately, current PD therapeutics target only the symptoms and not the mechanisms responsible for disease progression. Identification of novel molecules that could cross the blood-brain-barrier and activate Nrf2 may have a valuable role in neuroprotection in Parkinson's disease.</p><p>Under basal conditions, Nrf2 is quarantined in the cytoplasm through binding to kelch-like ECH associated protein 1 (Keap1), with binds both Nrf2 and cytoskeletal elements in the cell, restricting Nrf2 in the cytoplasm (12), and also promoting the ubiquitination and degradation of Nrf2 through its role as a E3 ubiquitin ligase adaptor protein (13). The specific features of the Nrf2:Keap1 interaction, as well as the necessary structural features for small molecules to disrupt Nrf2:Keap1 binding to allow nuclear translocation of Nrf2, are still being discovered. However, it is clear that Keap1 acts as a redox sensor for the complex: the human Keap1 protein has 27 cysteine residues as opposed to only 6 in the human Nrf2 protein. Of the cysteines in Keap1, research has confirmed that the primary electrophilic sensors are C273, C288 and C151 (14-16).</p><p>Novel small molecule activators are continually being developed and tested for Nrf2-ARE activation. Two are currently being tested in clinical trials: an oral formulation of dimethyl fumarate marketed by Biogen Idec, Inc. is in clinical trials for treatment of multiple sclerosis, while an oral formulation of 2-cyano-3,12-dioxooleana-1,9-dien-28-oic acid-methyl amide (CDDO-MA) marketed by Reata Pharmaceuticals, Inc. is being tested for kidney disease in type II diabetes patients. Small molecule activators of the Nrf2-ARE pathway have been shown to reduce nigrostriatal degeneration in mouse models of Parkinson's disease (7, 17). However, identification of such potential compounds through conventional high-throughput screening is slow, extremely expensive and prone to failure in secondary assays. Another confounding factor is that different small molecule activators of the Nrf2-ARE pathway act through different mechanisms (18, 19). Finally, there is great structural diversity in the known Nrf2 activating molecules making it difficult to predict or intelligently design novel small molecule activators of the Nrf2-ARE pathway.</p><p>To overcome this obstacle, we used Constellation, an innovative quantum-based computational process for drug discovery developed by Gradient Biomodeling LLC, as recently described (20). Constellation has been shown to identify novel compounds with significant potency against high-profile targets ranging from single macromolecules [renin, glutamate carboxypeptidase II], to complex cellular pathways [Galanin Receptor 2 agonists], to complete organisms [P. falciparum] with high accuracy and efficiency [Sullivan et al.; unpublished data Gradient Biomodeling LLC].</p><p>As Constellation does not require any prior, explicit knowledge of the macromolecular targets, the approach allows for simultaneous interrogation of multiple targets within a pathway or multiple pathways within a sub-system (or a system), integrating various exploration of systems behavior, or "phenotypes," at all levels of structural and functional complexity.</p><p>Herein, we report on the experimental in vitro results obtained from this computational in silico prediction approach. A novel chemical class of Nrf2 pathway activators has been identified and validated in primary neuronal cultures. Activation is comparable to that of the existing Nrf2 activators and protects against oxidative stress induced neurotoxicity in primary cortical neuronal culture as well as MPP+-induced dopaminergic neuron death in primary midbrain culture. Thus, the increased throughput of compound screening afforded by this method of quantum drug discovery and design will ultimately result in the de novo synthesis of novel Nrf2 activating molecules with the potential to treat PD as well as other neurodegenerative diseases.</p><!><p>This technique has recently been described (20). Briefly, the metric modeling technology considers the compounds of interest as quantum objects, without explicit dependence on their particular chemical structure. On a theoretical level, these quantum properties serve as powerful descriptors for molecular modeling, compound identification, optimization and de novo design. The computational platform determines essential, rigorous, easily computable molecular attributes related to chemical activity. These attributes are derived from a special representation of quantum fields. Their well-defined mathematical characteristics afford systematic theoretical treatment and property prediction with methods that would otherwise be computationally impossible. Specialized machine-learning algorithms with fuzzy decision-making protocols(21, 22) are applied to identify both active compounds and the corresponding quantum features of chemical and biological interest. Combined with the underlying modeling architecture, the algorithms also provide mechanistic hypothesis for the modeled interactions. Since structurally dissimilar compounds can be similar on a quantum level, our process is particularly good at identifying chemically novel compounds that have significant potency against a known target (protein or biological pathway).</p><p>Here, the Constellation computational software developed by Gradient Biomodeling, LLC., previously described in (20), was used to identify novel activators of the Nrf2-ARE. Briefly, a predictive quantum filter was designed using a training set of known small molecule activators of the Nrf2-ARE pathway, including tBHQ, 3H-1,2-dithiole-3-thione and sulforaphane and in vitro data for over 10,000 additional compounds. A series of fuzzy algorithms were developed to classify ideal target properties. The resulting quantum filters were used to identify a set of quantum components for property prediction of prospective molecules (Fig. 1). Although there is no direct theoretical correlation between these scores and compound activity, in practice this is often the case: the activity of a compound greatly increases with the number of interaction constraints it satisfies. Approximately 18 million commercially available structures were screened using sources including Enanmine, ChemBridge, LifeChemicals, ChemDiv, TimTec and the National Cancer Institute. The compounds were rank-ordered by the model and 14 high-ranked hits were procured and tested in vitro.</p><!><p>Cultures were derived from ARE-hPAP reporter (23), Nrf2 wildtype (WT), or Nrf2 knockout (KO)(24) mice as previously described (Kraft et al., 2004). Briefly, cortices from E15 mouse pups were pooled in 10 mL ice-cold Ca2+ and Mg2+ free HBSS (Life Technologies, Carlsbad, CA). Tissue was minced, centrifuged and digested in 0.05% trypsin without EDTA in HBSS for 15 minutes at 37°C. Following trypsinization, cells were rinsed 3 times with HBSS. Cells were then washed with CEMEM (minimum essential media with Earle's salts; Life Technologies, Carlsbad, CA), 2mM glutamine, 1% penicillin/streptomycin, and 10% each of heat inactivated fetal bovine serum and horse serum (Atlanta Biologicals, Inc., Lawrenceville, GA) and triturated to a single-cell suspension and strained through a 70μM cell strainer (BD Biosciences, San Jose, CA). Cell were counted and assayed for viability using trypan blue and plated at a density of 3×105 cell/cm2 on poly-D-lysine coated plates. Cells were maintained in CEMEM for 45 minutes, followed by media change. After 45 minutes, media for neuronally-enriched cultures (>98% neurons) was changed from CEMEM to NBM (Neurobasal media; Life Technologies, Carlsbad CA) supplemented with B27 with antioxidants and 2mM glutamine. For mixed cultures (~ 40% astrocytes and 60% neurons), media was changed from CEMEM to NBM at 2 days post plating. All cells were left for at least 48 hours in NBM prior to initiating experiments. Cells were incubated at 37°C in a tri-gas incubator with 5% O2, 5% CO2, and 90% N2. Preparations for astrocytes were similar except cells were prepared from the cortices of P1 pups, plated at a density of 3×104 cell/cm2 on collagen coated plates, and maintained in CEMEM throughout with a complete media change every 2-3 days; experiments were performed when the cells were confluent (one week).</p><p>Compounds were dissolved in 100% DMSO and administered to cells for 48 hours (final concentration of DMSO was 0.1%). For H2O2 challenge experiments, 48 hours after dosing medium was changed to NBM with B27 minus antioxidants and hydrogen peroxide (H2O2) was administered to the cells for an additional 24 hours.</p><!><p>Dopaminergic cultures were prepared as in (25) with slight modifications. All chemicals were purchased from Sigma (St. Louis, MO) or Fisher Scientific (Lafayette, CO). Briefly, two litters of P3 ARE-hPAP pups were decapitated, skulls were removed, and the midbrain was dissected out in sterile ice-cold dissociation solution (90mM Na2SO4, 30 mM K2SO4, 6mM MgCl2, 0.25mM CaCl2, 10mM HEPES, 20mM D-glucose, pH 7.4). Midbrains were minced and incubated in papain (Worthington Biochemical Corporation, Lakewood, NJ) activated in 10mL dissociation dilution for 20 minutes at 37°C. Tissue blocks were washed twice with sterile trituration solution (NBM-A (Life Technologies), 1% penicillin/streptomycin, 1% glutaMAX™ (Life technologies), 2% B27, 10% heat inactivated fetal bovine serum, 1mg/mL trypsin inhibitor, 1mg/mL BSA, 10mM HEPES, pH 7.4). Cells were triturated in trituration solution, and gently dropped onto centrifugation solution (NBM-A, 1% penicillin/streptomycin, 1% glutaMAX™, 2% B27 with antioxidants, 10% heat inactivated fetal bovine serum, 10mg/mL trypsin inhibitor, 10mg/mL BSA, 10mM HEPES, pH 7.4) and centrifuged. Cells were resuspended in 500 uL trituration solution, counted and assayed for viability with trypan blue, and plated at 6×104 cells/cm2 in a mixture of 2/3 NBM-A+ (1% penicillin/streptomycin, 1% glutaMAX™, 2% B27 with antioxidants, 10% heat inactivated fetal bovine serum) and 1/3 MEM+ [1% penicillin/streptomycin, 1% glutaMAX™, 10% heat inactivated fetal bovine serum, 0.2mM glucose, 0.5mM sodium pyruvate, 0.05% MITO+ (BD Biosciences, San Jose, CA)]. One day after plating, cultures were treated with 2.5uL/mL of 5-fluoro-2′-deoxyuridine (Sigma) to prevent cell proliferation. Experiments were performed within a week. Cells were treated with 25μM 822 or tBHQ for 48 hours followed by 2.5μM MPP+ (Sigma) for 48 hours, fixed with 4% paraformaldehyde, and stained for tyrosine hydroxylase.</p><!><p>Cell viability was assayed using the MTS (3-(4,5-Dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium salt) assay from Promega (Madison, WI) following the manufacturer's suggested protocols. For TUNEL staining, cells were fixed for one hour with 4% paraformaldehyde, permeabilized and treated with the Roche (Indianapolis IN) In Situ Cell Death Detection kit according to manufacturer's instructions. Apoptotic cells are identified by fluorescein-labeling of DNA strand breaks with the TdT enzyme. Cells were counter stained with Hoescht and imaged using a Zeiss microscope.</p><!><p>The hPAP activity assay protocol was based on the protocol described in (23). Briefly, cells were lysed in TMNC lysis buffer (50mM Tris, 5mM MgCl2, 100mM NaCl, 1% 3-[(3-cholamidopropyl)dimethylammonio]-1-propanesulfonate (CHAPS)) and freeze-thawed at −20°C. Extracts were incubated with 200mM diethanolamine (DEA) buffer at 65°C to inactivate endogenous alkaline phosphatase activity. hPAP activity was quantified in 200mM DEA with 0.8mM CSPD [disodium 3-(4-methoxyspiro (1,2-dioxetane-3,2′-(5′-chloro)tricycle(3.3.1.1 3,7)decan)-4-yl)phenyl phosphate) (Life technologies), 2× Emerald and 5mM MgCl2]. Luminescence was measured on a Berthold Orion microplate luminometer with one second integration. Baseline signal from hPAP negative control culture samples was subtracted from all value.</p><p>To visualize hPAP activity, cells were fixed with 4% paraformaldehyde for 20 minutes and incubated in TMN buffer (50mM Tris, 5mM MgCl2, 100mM NaCl, pH 9.5) at 65°C for 25 minutes to heat inactivate endogenous alkaline phosphatase activity. Cells were stained for hPAP by incubating at 37°C in TMN buffer plus 1mg/mL NBT and 1mg/mL X-phosphate (5-bromo-4-chloro-3-indoyl-phosphate) (BCIP) (EMD Chemicals, Gibbstown, NJ) and counterstained with eosin.</p><!><p>To visualize NQO1 activity, cells were prepared as described in (26). Briefly, cells were fixed in 4% paraformaldehyde for 20 minutes, washed with PBS and incubated in reaction buffer (25mM Tris, 0.8% Triton X-100, 2mg/mL BSA) with 100μM NBT (Calbiochem), 1mM NADPH and with or without 100μM of the NQO1 substrate LY83583. The reaction was incubated at 37°C for an hour.</p><!><p>For immunohistochemistry, cells were fixed in 4% paraformaldehyde, then incubated for one hour with a blocking solution (PBS with 10% goat serum, 0.4% Triton-X-100, and 0.005mg/mL BSA), then incubated with tyrosine hydroxylase (TH) antibody (Millipore, Billerica, MA) at 1/1000 dilution overnight. Secondary detection was performed with fluorescently labeled antibodies. The Hoescht stain was used to identify nuclear DNA.</p><!><p>All values are represented as the mean ± SEM. The n number of values used is presented in the individual figure legends. Significance was determined using an unpaired Student's t-test (p<0.05) or a one-way ANOVA (p<0.05) followed by posthoc analysis to determine significant paired comparisons (p<0.05) based on experimental design.</p><!><p>The Constellation computational software developed by Gradient Biomodeling, LLC., previously described in (20), was used to identify novel activators of the Nrf2-ARE. All of the identified hits contained multiple quantum components [Fig. 1, quantum component (QC) 1] that correlated with Nrf2-ARE activation. Some compounds had multiple quantum components: for instance, structure 71 contained an additional component (QC2) not found on 193 and 822 (Fig. 1, C,D). The two oxygens at the end of the tail on 71, but not at the beginning of the tail on 822, create the additional quantum component.</p><p>The resulting efficacy of Nrf2-ARE activation was assessed via ARE-hPAP reporter activation (Fig. 2A, structures shown in Fig. 2B). Dose curves were run for all compounds to determine toxicity using the MTS assay (data not shown). The data presented in Fig. 2A represents the fold increase in hPAP activity for the highest non-toxic dose of each compound (ranging from 0.75-50μM). Of the 14 tested compounds, 7 showed significant activation relative to vehicle treated cells at non-toxic doses (822, 071, 193, 797, 186, 543, 881). Of these 7 compounds, 4 had hPAP activity of greater than 2-fold (822, 071, 797, 193). In fact, the strongest activator, 822, had activation that was higher than the positive control used, tert-butyl hydroquinone (tBHQ) at the same dose, although the difference was not statistically significant at this dose. Overall, this is a vast improvement on available screening methods; 50% of the predicted activators were validated in vitro to show Nrf2-ARE activation.</p><!><p>Interestingly, two of the best activators of the Nrf2-ARE pathway in this quantum screen were also structurally similar (Fig. 1C, 2B, 822 and 071). Since this structure has never previously been reported to activate the Nrf2-ARE pathway, a family of similar structures was obtained and screened for Nrf2-ARE activation using cultures derived from the ARE-hPAP reporter mice. An initial dose response analysis revealed that all compounds were non-toxic at 25μM (MTS assay; data not shown). This dose was used for further studies comparing all compounds. The structures, hPAP activity (fold change), and hPAP histochemistry for this set of compounds are shown in Fig. 3. Compounds are arranged in order of their partition coefficient or LogP, a measure of hydrophobicity, with the lowest, 302 having a LogP of 3.2 and the highest, 106, having a LogP of 5.7. All of the compounds except the base structure (302) show comparable or greater increases in hPAP activity relative to tBHQ. At this dose, 822 exhibited significantly higher activation than tBHQ.</p><!><p>To determine whether the activation of Nrf2-ARE proteins was Nrf2-dependent, changes in the Nrf2-dependent gene NQO1 were evaluated by histochemical staining. 822 was used as a representative compound because it exhibited the greatest activation. Nrf2 WT cultures were compared to Nrf2 KO cultures treated with 822, tBHQ or vehicle. NQO1 histochemical staining was dramatically increased by both 822 and tBHQ in Nrf2 WT cultures (Fig. 4, A-C). In contrast, NQO1 staining of Nrf2 KO cultures was not changed by treatment with either 822 or tBHQ relative to the vehicle treated control (Fig. 4, D-F). Pretreatment with the antioxidant n-acetyl cysteine resulted in approximately a 50% reduction in hPAP activity in 822 treated mixed cultures, but there was no change in hPAP activity in n-acetyl cysteine pretreated tBHQ treated cortical cultures (data not shown).</p><!><p>Since all initial screening was done in mixed neuron/astrocyte co-cultures and tBHQ has been found to selectively activate Nrf2 in astrocytes (23, 27), we prepared enriched cultures (> 98% of cells) of astrocytes and neurons to determine if 822 could also activate Nrf2 in neurons as well as astrocytes. Neuron or astrocyte cultures were incubated for 48 hours with 822. Neuronal-enriched cultures treated with 822 exhibit an increased histochemical staining for both hPAP (Fig. 5A,B) and NQO1 (Fig. 5C,D) relative to vehicle treated cells. hPAP activity also showed a significant increase in neurons following 822 with no observed toxicity (Fig. 5E,F). Similarly, astrocytes had a significant increase in hPAP activity with no associated toxicity (Fig. 5G,H).</p><!><p>To determine whether the activation of the Nrf2-ARE pathway by 822 treatment was sufficient to coordinate a protective response against oxidative stress-induced neurotoxicity, mixed neuron/astrocyte cultures were treated with 822 or tBHQ for 48 hours followed by challenge with H2O2 for 24 hours. Significant protection was observed in cells treated with 822 or tBHQ relative to vehicle in WT cells (Fig. 6A). No protection was observed in Nrf2 KO cells treated with 822 or tBHQ relative to vehicle (Fig. 6B). In addition, TUNEL staining was evaluated after treatment with tBHQ or 822 for 48 hours followed by treatment with 20μM H2O2 for 24 hours. There was a modest increase in TUNEL staining in KO cells relative to WT cells in the vehicle-treated group (Fig. 6 C-E vs H-J). In WT cells treated with H2O2 , there was a dramatic reduction in TUNEL staining associated with 822 or tBHQ treatment (Fig. 6, E-G), while no reduction in TUNEL staining was observed in the H2O2-treated KO cells (Fig. 6, K-M).</p><!><p>Primary midbrain cultures were stained for tyrosine hydroxylase, a marker for dopaminergic neurons, to verify the presence of dopaminergic neurons (Fig. 7A). Cultures were treated with vehicle, 822 or tBHQ for 48 hours, followed by a 48 hour administration with MPP+. MPP+ is a mitochondrial complex I inhibitor that is selectively taken up by the dopamine transporter, which is highly expressed on dopaminergic neurons, and thus useful as an in vitro model for dopaminergic neuronal death that is the cause in PD. Following treatment, the cultures were fixed and stained for tyrosine hydroxylase and surviving tyrosine-hydroxylase positive cells were counted. There was a significant loss to dopaminergic neurons associated with MPP+ treatment. There was also a significant reduction in dopaminergic neuronal loss in cultures treated with either 822 or tBHQ (Fig. 7B).</p><!><p>The Nrf2-ARE pathway is a viable therapeutic target for protection against oxidative stress in neurodegenerative disease. Many Nrf2-ARE activating small molecules have been identified (as recently reviewed in (28)) and most have a highly reactive chemical moiety that can be grouped broadly into families of isothiocyanates, Michael acceptors, organosulfur compounds, electrophilic compounds or heavy metal containing compounds (28). However, the complexity of the protein interactions in the Nrf2-ARE pathway makes the identification of novel activators difficult. For instance, it has been shown that the activators of Nrf2 function through different mechanisms: tBHQ treatment is associated with the formation of a Keap1 dimer and increased Keap1 ubiquitination(19); whereas sulforaphane appears to modulate the Nrf2-ARE pathway primarily through blocking the interaction between Keap1 and the Cul3, the ubiquitin ligase adaptor protein that is associated with Nrf2 ubiquitization (13). Other compounds have been shown to activate the Nrf2-ARE pathway in a phostphatidyl-inositol 3 kinase PI3K dependent mechanism (23, 27, 29, 30).</p><p>The multiple interactions possible in the Nrf2-ARE pathway makes intelligent design of novel activators of the Nrf2-ARE pathway especially difficult. A previous article detailed de novo development of Nrf2-ARE activators and found a modest 2-3 fold activation in cell lines (31), which was independently confirmed in our lab using primary cortical cultures (unpublished data). The difficulty of identifying novel Nrf2-ARE activators is further evidenced by the results of several recent chemical screens. In each, a single hit with activity comparable to that of tBHQ was identified out either 117 (32) or 9400 (33) screened compounds. In contrast, an initial in vitro screen in this work identified a compound with greater activation than that of tBHQ out of only 14 initial molecules, greatly improving the screening efficacy. Also, half of the initial screened compounds showed significant activation in vitro in this work. Thus, the in silico screening method described here was able to identify far more potent hits from existing chemical libraries that should translate to the eventual synthesis of novel potent Nrf2 activating compound based on the final quantum model.</p><p>Similarly, work modifying known activators has limited success, with most modified compounds showing no improvement in activity relative to the parent compound, and only a few modifications showing some improvement in activation over the parent structure (34, 35). In contrast, 7 of the 8 related family members of the initial hit identified here activated the Nrf2-ARE pathway similarly to that of tBHQ.</p><p>Here, we showed that the compounds were activating the Nrf2-ARE pathway through comparison to Nrf2 knockout cells. However, we have not yet identified the mechanism of compound interaction with Nrf2 and/or Keap1. We hypothesize that the mechanism of action is similar to that of a previously published structure, as the chemical features present in the 822 family are similar to those observed in a screen published previously (32). Specifically, the authors identified that a halide group on a hexane ring was critical for ARE activation. In addition, the presence of esters and ketones rather than amides on the tail promotes Nrf2-ARE activation. Further analysis of the ARE-activating compound, AI-1, suggests that the halide group in AI-1 interacts with C151 in the Cul3 interacting region on Keap1. This suggests a proposed sulforaphane-like mechanism of Nrf2-ARE activation via Cul3-Keap1 disruption. The family of compounds identified here have both halide groups and ester/ketone groups; these features are also evident on many of he structures identified in the first screen. This suggests that a similar mechanism may be promoting Nrf2-ARE activation in both AI-1 and these compounds; however, additional work will be required for characterization of the mechanism of Nrf2-ARE activation by the new chemicals identified here.</p><p>The data generated will significantly refine the quantum model and subsequent work will add quantum properties associated with the ability to cross the blood-brain-barrier and reduce first pass metabolism. Once optimized, novel Nrf2 activating compounds will be synthesized and tested for neuroprotective properties in mouse models of neurodegenerative diseases. The increased ability to generate potent activators of Nrf2 will increase the possibility of finding a new therapeutics with the potential to slow or halt the progression of PD and other neurodegenerative diseases.</p>

PubMed Author Manuscript

Development and screening of a series of antibody-conjugated and silica coated iron-oxide nanoparticles for targeting the Prostate Specific Membrane Antigen

The Prostate Specific Membrane Antigen (PSMA) is an established target for the delivery of cancer therapeutic and imaging agents due to its high expression on the surface of prostate cancer cells and within the neovasculature of other solid tumors. Here we describe the synthesis and screening of antibody-conjugated silica-coated iron oxide nanoparticles for PSMA-specific cell targeting. The humanized anti-PSMA antibody, HuJ591, was conjugated to a series of nanoparticles with varying densities of polyethylene glycol and primary amine groups. Customized assays utilizing iron spectral absorbance and Enzyme-Linked Immunoassay (ELISA) were developed to screen microgram quantities of nanoparticle formulations for immunoreactivity and cell targeting ability. Antibody and PSMA-specific targeting of the optimized nanoparticle was evaluated using an isogenic PSMA-positive and PSMA-negative cell line pair. Specific nanoparticle targeting was confirmed by iron quantification with inductively coupled plasma mass spectrometry (ICP-MS). These methods and nanoparticles support the promise of targeted theranostic agents for future treatment of prostate and other cancers.

development_and_screening_of_a_series_of_antibody-conjugated_and_silica_coated_iron-oxide_nanopartic

<!>Surface coating and antibody conjugation of MIONs<!>Immunoreactivity assay<!>Cell based ELISA assay<!>ICP-MS

<p>Iron-oxide nanoparticles (NPs) have tremendous potential as theranostic agents because of their biocompatibility and responsiveness to magnetic fields, which enables imaging (as MRI contrast agents) and therapy (through hyperthermia or drug delivery).[1] Making such NPs 'targeted' by decorating the surface with affinity ligands can increase the efficiency of NP delivery and retention in target cells and tissues.[2] However, the surface factors that lead to successful targeting are numerous and may be unpredictable, thus a great deal of optimization is required to generate each desired particle.[3] For example, recent active NP targeting studies with anti-HER2/neu affibodies found that an intermediate ligand density provided superior targeting efficiency when compared to higher ligand densities.[4] Similar studies with folate and ICAM (Intercellular Adhesion Molecule-1) as targeting ligands on other NPs have also underscored the benefit of optimized surface coatings and ligand densities for targeting.[5] It is therefore beneficial to generate and screen multiple NP prototypes, in order to achieve optimal targeting.</p><p>The Prostate Specific Membrane Antigen (PSMA) is a cell surface protein that is highly over-expressed in aggressive prostate cancer tumors[6] and within the tumor neovasculature of other solid tumors.[7] Unlike Prostate Specific Antigen (PSA), which is a secretary protein and serum biomarker, PSMA is a cell surface displayed integral membrane protein.[8] It is the target of the only FDA approved prostate-specific imaging agent[9] and has been studied as a target for the delivery of prostate cancer therapeutic agents for over a decade.[10] In light of the promising characteristics of PSMA as a cancer target, and iron oxide NPs as theranostic agents, we sought to generate an optimized PSMA-targeted magnetic iron oxide NP (MION) (Scheme 1). Silica was applied for NP surface coating because it is a low-cost and chemically inert material for coating that provides high colloidal stability in aqueous solution. Further, silica has not yet been explored as a coating or NP component for PSMA-targeted agents. Below we describe the synthesis, screening, and validation of these novel PSMA-targeted NPs.</p><p>Previously developed citrate stabilized MIONs [11] were coated with a silica shell of approximately 20 nm thickness by using a modified Stöber growth process.[12] In this process, various ratios of tetraethyl orthosilicate (TEOS) to (3-aminopropyl) triethoxysilane (APTES) were used to vary primary amine density on the surface of the MIONs (Table 1). Thus five MIONs having the same core, but differing only in surface amine density were generated in parallel. Each of the five silica-coated MIONs were then PEGylated using an excess mixture of 10% maleimide-PEG24-NHS and 90% non-functionalized PEG. Control particles containing 100% non-functionalized PEG were also generated for each coating condition. The maleimide-functionalized NPs were then conjugated to TCEP-reduced humanized anti-PSMA HuJ591 antibody (Ab). [13] HuJ591 was selected as the targeting agent because of its previous history as a PSMA-targeting ligand in clinical studies.[14] The zeta potential and hydrodynamic diameter of the five MION systems were measured before and after conjugation with the Ab. From Table 2 it is evident that the zeta potential of the silica-coated MIONs (with or without Ab) were nearly neutral (+/− 10 mV), which is desired to avoid non-specific charge-based interactions with the cell surface.[15]</p><p>To evaluate the immunoreactivity of the resulting Ab-targeted MIONs, we developed a custom binding assay which utilized the high spectral absorbance of iron. Specifically, individual wells of a 96 well-plate were coated with two Ab-specific (anti-human IgG or Protein A) or non-specific capturing agents (Bovine Serum Albumin, BSA or anti-GFAP Ab). The wells were blocked, washed, and then incubated with Ab-conjugated or non-conjugated MIONs. After washing, the captured MIONs were quantified by multiwell plate spectrophotometry at an absorbance of 350 nm. Serial dilution of MIONs demonstrated linear detection in low microgram range, with sensitivity to approximately 1 μg of MION (Figure 1, inset).</p><p>All of the J591-conjugated MIONs were captured by the anti-IgG coated wells, when compared to non-conjugated control MIONs or to control coated BSA or anti-GFAP wells (Figure 1, black bars versus grey bars). These results indicate that J591-conjugation was successful for all five coating conditions and that the spectral absorbance assay provided sufficient sensitivity and specificity for evaluating HuJ591-conjugated MION immunoreactivity. The second Ab capturing agent, Protein A, also demonstrated specificity for Ab conjugated MIONS when compared to non-conjugated controls (Figure 1, white bars versus grey bars). However, Protein A only captured MION-1.20-J591, MION-1.40-J591 and MION-1.80-J591 formulations. It is not clear why Protein A and IgG wells had differential MION capture efficiency. We speculate that this difference may be attributed to differential binding affinity or epitope availability of protein A versus anti-IgG. Alternatively, it could reflect differences in Ab conformation on the different MION surfaces. Nevertheless, it is notable that the MIONs having the lowest density of amine (1:80 APTES: TEOS, Table 1), rather than those having the highest, were captured most efficiently by both Protein A and anti-IgG. This result is surprising because MIONs having higher amine densities can be expected to have a greater number of conjugated antibodies and therefore a greater potential for capture. One possible explanation for this unexpected observation is that increased PEG (in MIONs with higher APTES: TEOS) sterically hindered the antigen-Ab interaction, leading to lower immunoreactivity. Conversely, low-to-intermediate J591 surface density may yield a higher percentage of antibodies having correct orientation or reduced steric impedance, thereby enhancing measured immunoreactivity. Regardless, it can be concluded from Figure 1 that J591 was successfully conjugated to the surface of several MIONs and having an orientation that enabled recognition by anti-human IgG and Protein A.</p><p>To further characterize the immunoreactivity of each Ab-conjugated MION, we applied a modified ELISA detection method for PSMA-specific cell binding. An isogenic pair of PSMA expressing and non-expressing LMD-MDA-MB-231 cells was used for MION capture. These two cell lines are genetically identical, except one has been engineered to express PSMA at levels comparable to that of prostate cancer cell lines (Figure 2). This model allows assessment of ligand-specific cell binding in a well-controlled manner.</p><p>The two cell lines were separately plated to equal density in 96 well-plates. Free J591 Ab or J591-targeted and non-targeted MIONs were then added to each well to allow for cell binding. After rigorous washing of each well, J591 and J591-targeted MIONS were detected by ELISA with HRP-conjugated Anti-human IgG. PSMA specific detection was demonstrated with the positive control, free J591 Ab, with linear sensitivity in the low nanogram range. (Figure 3, inset). The non-conjugated MIONs (J591 -) should not be specifically detected by this assay, although they were included as controls for determining anti-human IgG specificity. Three of the five J591-conjugated MION formulations, MION-1.20, MION-1.40, and MION-1.80, demonstrated PSMA-specific cell targeting (Figure 3, grey bars versus white bars). These results are consistent with Protein A capture in Figure 1 and indicate that the lower levels of active amine provided superior PSMA-specific immunoreactivity. Interestingly, the level of non-specific cell binding by the J591-conjugated MIONs to the PSMA-negative cells indirectly correlated with the level of functional amines (Figure 3, white bars). This may reflect the lower levels of PEG conjugation and reduced potential for MION surface shielding. Notably, this non-specific binding or "stickiness" was only detected by the cell binding ELISA assay, and was not reflected in the Ab capture spectrophotometric screens of Figure 1. It is likely that reflects the greater level or diversity of non-specific ligands on the surface of a cell when compared to homogenous MION capture by purified proteins.</p><p>Collectively, the Protein A capture spectral absorbance assay (Figure 1) and the isogenic cell-based ELISA assay (Figure 3) reveal complementary results for ligand-specific MION targeting. These data indicate that the top candidate particle for PSMA-specific targeting is MION-1.20-Ab, due to its strong PSMA-specific binding and minimal non-specific cell binding. To verify this Ab-specific and ligand-specific targeting we quantified PSMA-specific MION-1.20-Ab cell binding by inductively coupled plasma mass spectrometry (ICP-MS). MION binding, as quantified by iron-content (in ppb), was approximately five fold greater for the PSMA positive cells treated with MION-1.20-Ab, when compared to PSMA negative cells (Figure 4, grey bars versus white bars). In the absence of conjugated Ab, the MIONs bound similarly to PSMA-positive and negative cells.</p><p>These studies summarize a successful strategy for generating and evaluating a series of Ab-conjugated iron oxide NPs which target PSMA. There have been previous reports of PSMA-targeted iron oxide nanoparticles with differing surface chemistries and targeting ligands.[16] Here we selected silica for NP shelling because it is inert, biocompatible, easily modified, and thermally stable. Moreover, silica provided a rigid and stable coating that maintained NP solubility and stability after routine manipulations such as vigorous pipetting or centrifugation. A sequential NP synthesis approach was applied so that a single iron-oxide NP core could be similarly shelled with silica containing variable amine densities for antibody conjugation. Two custom assays were developed to screen the resulting particles for PSMA-targeting ability. The application of multi-well plate spectrophotometry assay enabled evaluation of small μg quantities of modified MIONs for immunoreactivity with low volumes of reagents and standard laboratory equipment. The ELISA-based cell binding assay applied an isogenic cell line pair provided for a well-controlled system for assaying ligand-specific MION-binding. This assay and provided insights into non-specific cell binding or off-target effects of some NP formulations. Both of these assays can be applied in a high-throughput manner to assess the optimal target specificity of a series of nanoparticle in short time. It is notable that the plates can coated and fixed in advance and stored for future use. Interestingly, our results indicate that increased density of PEG and targeting ligands on the NP surface does not necessarily enhance MION targeting. Rather, there is an optimized combination of surface chemistry, PEGylation and ligand density that determines the targeting potential of a NP. These techniques may be applicable to the development of other iron oxide NPs. Future work involving tumor specific imaging and/or therapy of the MION-1.20-Ab formulation will be necessary to determine the in vivo targeting potential of the optimized MIONs.</p><!><p>Silica shelling of MIONs was performed using standard silica shelling conditions followed by PEGylation with NHS-PEG-mal. TCEP reduced Ab was reacted with the maleimide group in order to synthesize Ab conjugated MIONs. The details of the process are described in supporting information.</p><!><p>Protein A, Anti-GFAP, BSA and Anti-IgG Ab were diluted to 0.2 mg/mL in PBS and 50 μL of each stock was added to each well of a 96 well plate. After incubating for one hour at 37º C, the solutions were removed and washed 3 times with PBS. Plates were then blocked with 100 μL of StartingBlock solution (Pierce Biotechnology, Rockford, IL) for one hour. After washing, 50 μL of various MION (+/− J591) samples in 1:10 dilutions of StartingBlock solutions were added, incubated for one hour at 37º C, washed three times with PBS, and the iron content measured by absorbance at 350 nm using a Molecular Devices Spectramax 250 plate spectrophotometer. Assay sensitivity and linearity were determined using serially diluted MIONs in 100 μL of PBS.</p><!><p>The PSMA positive LMD-MDA-MB-231-PSMA and parental LMD-MDA-MB-231 cells were plated in a 96 well plate at densities of 50,000–60,000 cell per well. The next day the confluent layer of cells were first fixed with 4% para-formaldehyde and then blocked with 3% BSA. After washing with PBS, 50 μL of different MION (+/− J591) or J591 alone (in different concentrations) were added in triplicate to various wells. The plate was kept at 37º C for 45 min. Sample plates were washed three times with PBS/0.1% tween-20. Anti-human HRP conjugated Ab (1: 20,000 dilutions in PBS) was then added to all wells and kept at 37º C for 45 min. After three PBS washes, 100 μL of Pierce OPD substrate was added to each well and incubated for 20 minutes at room temperature. After 20 minutes, absorbance at 450 nm was recorded for each well in a Molecular Devices Spectramax 250 spectrophotometer.</p><!><p>Following the ELISA, plates were washed with PBS (3X) and 20 μL of concentrated nitric acid was added to each well. After incubating at room temperature overnight, samples were transferred to a 15 mL conical flask. The wells were rinsed with 20 ppb EPA 200.7 water and transferred to the same conical. The total mass of each sample was adjusted to 2.00 g and samples were analyzed by Thermo Scientific X series ICP-MS</p>

PubMed Author Manuscript

AP-Net: An atomic-pairwise neural network for smooth and transferable interaction potentials

Intermolecular interactions are critical to many chemical phenomena, but their accurate computation using ab initio methods is often limited by computational cost. The recent emergence of machine learning (ML) potentials may be a promising alternative. Useful ML models should not only estimate accurate interaction energies, but also predict smooth and asymptotically correct potential energy surfaces. However, existing ML models are not guaranteed to obey these constraints. Indeed, systemic deficiencies are apparent in the predictions of our previous hydrogen-bond model as well as the popular ANI-1X model, which we attribute to the use of an atomic energy partition. As a solution, we propose an alternative atomic-pairwise framework specifically for intermolecular ML potentials, and we introduce AP-Net-a neural network model for interaction energies. The AP-Net model is developed using this physically motivated atomic-pairwise paradigm and also exploits the interpretability of symmetry adapted perturbation theory (SAPT). We show that in contrast to other models, AP-Net produces smooth, physically meaningful intermolecular potentials exhibiting correct asymptotic behavior. Initially trained on only a limited number of mostly hydrogen-bonded dimers, AP-Net makes accurate predictions across the chemically diverse S66x8 dataset, demonstrating significant transferability. On a test set including experimental hydrogen-bonded dimers, AP-Net predicts total interaction energies with a mean absolute error of 0.37 kcal mol −1 , reducing errors by a factor of 2-5 across SAPT components from previous neural network potentials. The pairwise interaction energies of the model are physically interpretable, and an investigation of predicted electrostatic energies suggests that the model 'learns' the physics of hydrogen-bonded interactions.

ap-net:_an_atomic-pairwise_neural_network_for_smooth_and_transferable_interaction_potentials

I. INTRODUCTION<!>II. METHODOLOGY A. Symmetry Adapted Perturbation Theory<!>B. Pairwise Energy Partition<!>C. Features<!>Atom-Centered Symmetry Functions (ACSFs)<!>Atom Pair Symmetry Functions (APSFs)<!>D. Network Architecture, Training, & Implementation<!>E. Training and Testing Datasets<!>A. NMA Dataset<!>FIG. 4.<!>B. H2O Dimer<!>C. S66x8<!>D. Pairwise Partition Analysis<!>IV. CONCLUSIONS<!>SUPPLEMENTARY MATERIAL<!>DATA AVAILABILITY

<p>Recent advances in the field of machine learning (ML) offer an exciting new perspective on the perpetual costaccuracy trade-off of quantum chemistry. 1,2 Models like neural networks (NNs), which are flexible universal function approximators, can be used to predict a variety of chemical properties. For a target molecular system, a parameterized (or trained) model can estimate a property in a small fraction of the time needed to compute it with quantum methods. These models come with some caveats, however. The accuracy of a prediction is highly dependent on the amount of data used to train the model and the similarity of the training data to the molecule to be predicted on. Also, it is nontrivial to design and train an ML model. Specific choices made in the architecture of a neural network, the choice of features, hyperparameters, etc. can have a large impact on the model's accuracy and transferability.</p><p>Over the past few years, ML models have been developed for seemingly every chemical property that can be computed with an ab initio method. The chemistry community has investigated electrostatic multipoles 3 , rate constants 4 , chemical shifts 5 , etc. However, a significant amount of attention has been dedicated to ML potentials (i.e. energy prediction). [6][7][8][9][10][11][12][13][14][15][16][17][18] These ML potentials possess one commonality, which is that their development and application have emphasized covalently bound systems and accurate total energy predictions. Less attention has been paid to noncovalent interactions (NCIs) and interaction energies, which are of fundamental importance to drug binding, liquid structure, biomolecular structure, molecular crystals, etc. In many of these applications, obtaining accurate interaction energies is a more important goal than total energies. This is not to say that existing ML potentials totally neglect NCIs. The ANI-1X model, for example, was trained using a dataset containing many molecular dimers. 14 Also, accurate total energy predictions can be used to obtain accurate intermolecular energy predictions. For a dimer, the interaction energy (∆E int ) is defined as:</p><p>where E AB , E A , and E B are total energies of the dimer and two monomers. For any ML potential, dimer interaction energies can be evaluated with this so-called 'supermolecular' approach. One challenge for NCIs is that to obtain accurate interaction energy, one needs good cancellation of errors in all three component energies, E AB , E A , and E B . Alternatively, more accurate models of interaction energies might estimate ∆E int directly. Creating such an intermolecular ML potential poses some unique challenges, as many standard representations of molecular systems for ML are not necessarily applicable, and special care must be taken in generating useful training data. These and other concerns are discussed in our recent pilot study of hydrogen-bonding interactions, which to our knowledge is the first purely ML intermolecular potential designed to work on an entire class of chemical systems. 19 In that work, neural networks were trained to predict symmetry adapted perturbation theory (SAPT) interaction energies, resulting in a model which we refer to here as SAPT-ML. The SAPT-ML intermolecular potential was designed in the same spirit as many of the popular total potentials, including the use of an atomic energy partition, discussed in more detail in Section II B. On a dataset of crystallographic and artificially constructed hydrogen bonding dimers, SAPT-ML predicted interaction energies with a mean absolute error of 1.2 kcal mol −1 , approaching quantitative accuracy, while using a relatively small training dataset for deep learning tasks.</p><p>The primary evaluation metric for any potential, SAPT-ML included, is generally a summary of the error distribution such as the mean absolute error, max absolute error, etc. preferably computed for a large and representative test dataset. A useful ML potential will make predictions with small errors, but these summary statistics alone are not sufficient criteria for usefulness. Of arguably equal importance is the smoothness of the predicted potential energy surface (PES). Jagged PESs yield inaccurate forces, which is a particularly problematic concern for molecular dynamics simulations, a target application of ML potentials. Smoothness is also a requirement for geometry minimization and transition state searches. For an intermolecular PES, an ML potential should adhere to additional asymptotic constraints. The predicted interaction energy should be approximately zero at large separations and strongly repulsive at near separations. The predicted PES should not only be locally smooth, but also exhibit minima with approximately the same energies and locations as the true PES.</p><p>The importance of smoothness for ML potentials has been acknowledged by ML approaches that use energy gradients in the training procedure, 20 but to our knowledge global PES smoothness has not been studied in detail, particularly in the context of intermolecular interactions. In the pilot hydrogen-bonding study, we did not explicitly examine smoothness nor asymptotic convergence of the SAPT-ML model. Here, we re-examine our SAPT-ML potential and also investigate the popular ANI-1X potential. We find that both models produce intermolecular PESs with unphysical irregularities. This appears to be a fundamental weakness of atomic neural network potentials, which predict an atomic partition of the target property. To overcome this deficiency we propose an alternative atomic-pairwise interaction paradigm, and we introduce AP-Net, a corresponding atomic-pairwise neural network intermolecular potential. To illustrate the advantage of this new approach, we test AP-Net on the same hydrogen-bonding task as SAPT-ML and report up to a 5-fold reduction in errors across the SAPT interaction energy components. The AP-Net model, trained only on a modest number of mostly hydrogen-bonded dimers, is also tested on the chemically diverse S66x8 benchmark dataset. 21 This atomic-pairwise models exhibits surprising transferability, making reasonable estimates of intermolecular PESs dominated by π−π stacking and dispersion interactions for systems where ANI-1X generates incorrect potentials. Overall, we observe that AP-Net uniquely displays correct asymptotic behavior and makes smooth predictions along intermolecular coordinates, both necessary aspects of intermolecular potentials. Lastly, we examine the individual atomicpair energies predicted by AP-Net and find good agreement with chemical intuition. This suggests that AP-Net predicts interaction energies by 'learning' some physically meaningful chemical representation.</p><!><p>In order to train an intermolecular ML potential, reference interaction energies (or labels) are needed from some quantum chemistry calculation. In principle, any wavefunction or density functional theory (DFT) method could be used in conjunction with Equation 1 to obtain supermolecular reference interaction energies. An alternative theoretical approach for computing the interaction energy is symmetry adapted perturbation theory (SAPT). [22][23][24] In its wavefunction-based formulation, SAPT accounts for the interaction between two monomer Hartree-Fock wavefunctions through a triple perturbation series in monomer A correlation, monomer B correlation, and intermonomer interaction. SAPT has a few advantages over the supermolecular approach. Firstly, SAPT is formally correct in the limit of the perturbation series, and it recovers the full configuration interaction solution to the time-independent Schrödinger equation. However, a major appeal of SAPT is the accuracy of early truncations of the series. One of the most popular and economical SAPT methods, 0 th order intramonomer correlation and 2 nd order intermonomer interaction, is commonly referred to as SAPT0. Combined with an appropriate jun-cc-pVDZ basis set, SAPT0 has been shown to be surprisingly accurate, in part due to fortuitous error cancellation. 25 At only O(N 5 ) in cost, SAPT0 is is affordable enough to probe interactions in systems with hundreds of atoms, large enough to be of biological interest. 26 The most useful feature of SAPT for an ML potential, however, is the resulting physically meaningful decomposition of the interaction energy. The individual terms of the perturbation series correspond to standard interpretations of intermolecular interactions: interactions between permanent charge distributions of opposite monomers (electrostatics, E elst ), induction or polarization of a charge distribution on one monomer by a charge distribution on the other (induction, E ind ), simultaneous correlation between charge distributions on opposite monomers (dispersion, E disp ), and finally fermionic Pauli exchange between electrons on opposite monomers (exchange, E exch ):</p><p>These components are often exploited in modern force field development, where separate physically-motivated functional forms are developed and parameterized for each term. [27][28][29][30][31][32][33][34] An analogous approach can be taken in an intermolecular ML potential by structuring a model to predict these individual components, therefore allowing the model to exploit the interpretability of SAPT. Detailed equations for the exact specification of individual SAPT0 components and their efficient implementation through density-fitting techniques are presented in Refs 35 and 36.</p><!><p>Nearly every published ML potential follows the same general formulation. For a given molecular system, there exist many arbitrary partitions of the total energy (E) into atomic energies (E i ):</p><p>It is assumed that a partition exists such that E i is a learnable and transferable function of the local environment around atom i. Regression models such as neural networks are used to predict E i for each atom. No ab initio E i labels exist, as atomic partitions of the energy are not generally computable with quantum mechanical methods. Instead, the sum of predicted E i 's is constrained to match E, which in most ML frameworks is easily specified. Thus, the predicted total energy is estimated through a learned atomic energy partition. The various ML approaches primarily differ in how the regression model is parameterized to predict the partitioned energy E i of each atom i.</p><p>The atomic nature of this type of potential has an important consequence. Because the regression is performed at the atomic level, models are transferable to many different molecules, provided that the local environments of atoms in the molecules are similar. For example, an atomic potential trained on conformations of butane and hexane should provide reasonable estimations for the PES of a pentane molecule. The transferability of atomistic models of this kind can be contrasted with traditional system-specific potentials, which must be created and parameterized separately for each molecular system of interest.</p><p>A weakness of atomic ML potentials is their poor performance in capturing long-range interactions. 10,13 This is because the enforced locality of the atomic partition is at odds with the long-range nature of NCIs. Most ML potentials limit an atom's environment to include neighboring atoms within some distance cutoff that can be as short as 5 Å. This is insufficient in capturing small but chemically important long-range electrostatics, and to a lesser degree, van der Waals effects. If one chose to use spatially large atomic environments, the transferability and computational cost of the resulting ML model would suffer. To address this issue, some machine-learning potentials simply add a classical force-field model (like Grimme's D3 dispersion correction) to their predictions in order to describe distant interactions 9,13 . Other models are parameterized to predict atomic charges, which are used to evaluate a separate long-range electrostatic energy. 17 In an attempt to explore the modeling of NCIs with flexible neural network models, we recently adapted the atomic partitioning approach to the explicit prediction of dimer interaction energies. 19 The resulting model, SAPT-ML, was trained to predict the four SAPT0 components for a set of hydrogen-bonded dimers. A consequential design choice of the SAPT-ML model was that the interaction energy could also be partitioned into individual atomic contributions, each of which is a learnable function of an atom's local environment:</p><p>where a and b index the atoms of monomers A and B respectively. Eq. 4 is a straightforward intermolecular adaptation of Eq. 3, the standard formulation of ML potentials.</p><p>The current work is motivated by the observation that NCIs between molecules are well understood to be approximately a sum of interactions between pairs of atoms. Atomic-pairwise additivity corresponds to a different partition of the interaction energy:</p><p>This paradigm is fundamental to NCIs and dates at least as far back as the original Lennard-Jones potential. 37 Pairwise additivity is still the basis of the many popular classical force fields like AMBER, CHARMM, and OPLS. [38][39][40] Advanced polarizable force fields such as AMOEBA add small, self-consistent corrections on top of a pairwise additive model. 41 The D3 dispersion correction of Grimme and coworkers captures the purely quantum mechanical phenomenon of dispersion with yet another pairwise model. 42 Our group has even developed postprocessing methods for partitioning a calculated SAPT0 interaction energy into both atom and functional group pairs. 43,44 In the design of machine learning models, incorporating prior knowledge about the nature of the function to be approximated is a fundamental priority. For example, major advances in the field of computer vision are a result of encoding locality and shift invariance in neural networks (through so-called "convolutional" layers). 45,46 Given the overwhelming amount of chemical intuition and empirical evidence supporting the pairwise additive nature of NCIs, it seems imperative to incorporate this information into the model. This insight is the motivation behind AP-Net, an atomic pairwise neural network model for interaction energies, developed in this work. Functionally, AP-Net is similar to SAPT-ML in that both models use a geometric description of a molecular dimer to predict an interaction energy. The important difference lies in the atomic-pairwise nature of AP-Net's architecture.</p><!><p>Regression problems require a careful choice of variables (or features/descriptors) from which a mapping is approximated (or learned) to some desired property (or label). The selection of features can greatly affect both the accuracy of the final model and its ability to generalize well to unseen data. These concerns are particularly relevant for chemical systems, since standard feature engineering techniques don't immediately apply to the unique graph-like structure of molecular geometries. Thus, the representation of chemical data is an essential component of machine learning potentials. Some aspects of useful features can be reasoned about a priori. For one, features should obey the same invariances as the predicted property. For potentials, this means that the features should be invariant to relabeling of identical atoms, molecular translation, and molecular rotation. While many features are used across various atomic potentials, the atomic-pairwise model described in this work necessitates the development of features that can describe a pair of atoms.</p><p>To begin with, the atomic pair paradigm has an immediate set of sensible and descriptive features not explicitly available in the original atomic paradigm: Z a , Z b , and r ab . The existence of simple, qualitatively correct, nonbonded force fields based on these variables suggests that they can be used to account for a large fraction of interaction energies, and we use them as inputs to AP-Net. However, these three variables alone are not sufficient to describe all of NCIs, as they don't contain information about individual atomic environments or their orientations with respect to the other monomers. Thus, additional descriptors are necessary. For this purpose, the popular atom-centered symmetry function (ACSF) 7 is reviewed and the new atom-pair symmetry function (APSF) is introduced.</p><!><p>The prediction of pseudo-atomic properties (like charge) and atomic partitions of molecular properties (like energy) are common applications of machine learning to chemistry. As such, much effort has been put into the design of features that encode local atomic environ-ments to use in ML models. In contrast, this work seeks to predict a partitioning of the interaction energy over pairs of atoms, not individual atoms. Nevertheless, a reasonable starting point for a feature to describe a pair of atoms is simply the concatenation of individual atomic features. A well-established descriptor for this purpose is the radial atom centered symmetry function (ACSF) of Behler and Parinello. 7 A radial ACSF of atom i (G rad i ) describes the atom's local environment in terms of the radial distribution of neighboring nuclei:</p><p>G rad i encodes the radial density of atom i's neighbors. The δ ZZj term filters only neighboring nuclei with a particular atomic number. The parameters µ and η define the radius and width of a spherical Gaussian shell upon which the neighbor density is projected onto. Lastly, the cutoff function f c ensures locality by removing contributions of neighbors outside of a chosen cutoff radius r c . In practice, many radial ACSFs are used to describe the complete radial environment around atom i (i.e. a collection of G rad i with varying µ, η, and Z). We denote such a collection or vector of ACSFs for a single atom with different {(µ, η, Z)} as G rad i . Figure 1 depicts individual elements of G rad i for an example system. In the dimer picture, we further indicate that the ACSF vector is formed from only intramonomer neighboring nuclei as G rad a or G rad b . It is important to note that, while we chose radial ACSFs to describe the intramolecular environment for their speed and simplicity, any other feature could be substituted or added, including the more expensive angular ACSFs. In particular, the increasingly popular message-passing framework 10,12,17,47 could be adapted to the dimer picture and used here.</p><!><p>The monomer ACSFs G rad a or G rad b as defined above encode an atom's local environment within its respective monomer, containing no information about the identities of, distances from, or orientations to atoms in the other monomer. However, when these monomer ACSFs are combined with Z a , Z b , and r ab , we can achieve a nearly complete description of a pair of atoms in opposite monomers, lacking only information about intermonomer orientation. A model developed to use only these features would therefore result in an isotropic description of atom-atom interactions. This can be a poor approximation, since the electron density around an atom may be anisotropic (e.g. higher along a bond axis or in the direction of an electron lone pair). ). Each µ is associated with a direction from the hydrogen, drawn as a blue ray. The dependence of this direction on the HO intermolecular axis is indicated in red.</p><p>Here, we introduce angular atom-pair symmetry functions (APSFs) as a way to account for orientation of atomic environments within the atom-pair paradigm. The angular APSF of an atom a in monomer A with respect to an atom b in monomer B is defined as:</p><p>where index a runs only over atoms in the same monomer as atom a. The angular APSF closely resembles the original radial ACSF, and the two features are compared in Figure 1. In the APSF, atom a is still described by the spatial distribution of its neighboring nuclei within monomer A. The terms f c (r aa ) and δ ZZ a serve analogous roles as in ACSFs, enforcing spatial locality and filtering out nuclei by atomic number, respectively. The essential difference of this descriptor is the shape of Gaussian that the neighbor density is projected onto, which is now cone-like instead of spherical. The apex of the cone is at nucleus a and the cone is aligned with the ab axis. µ determines the apex angle of a cone (instead of the radius of a sphere), and η is still the Gaussian width. Different values of µ define cones with different angles. The value cos θ a ab is determined by the alignment of the vector r aa with the ab axis. The range of reasonable µ values is the same range as the cosine function, [−1, 1]. If an ACSF is understood to encode an atom's neighbor density as a function of distance, the APSF can be thought of as encoding an atom's neighbor density as a function of angle or orientation, where an atom in a different monomer is the third point in the angle.</p><p>The vector notation used for the ACSFs is also used for APSFs. Note that G ang a(b) is not equal to G ang b(a) . The former describes the environment of atom a in monomer A with respect to the orientation of atom b in monomer B while the latter describes the environment of b with re-spect to the orientation of a. The intent behind the APSF descriptor is to explicitly decouple intermonomer orientation from intermonomer distance, which is already captured by using r ab as a feature. Without the APSF, AP-Net could not be able to account for atomic anisotropy. It would be possible to use additional hyperparameters in the APSF, for example by including an additional spherical Gaussian shell to project r aa onto with its own µ and η, much like an ACSF. However, minimizing the number of necessary features is a worthwhile pursuit if AP-Net is to be used in applications such as high-throughput screening or molecular dynamics simulations.</p><!><p>A common practice in the design of neural network potentials under the atomic partition is to parameterize a separate neural network for each element (atom type):</p><p>This splitting out of networks has been seemingly necessary to achieve good performance, but is undesirable for a number of reasons. For one, it results in unwieldy implementations that scale in size with the number of treated atom types. The use of separate atomic networks is also not very data-efficient, since generalization across atom types by definition cannot be learned by independent networks. Data-efficiency is particularly important when some atom types are much less common in a dataset. In accordance with recent work on shared-weight models, 48 the proposed AP-Net architecture avoids this problem.</p><p>For each SAPT component energy (E comp ), a single neural network is trained to predict atomic pair partitions (E ab,comp ) of the component energy:</p><p>where the notation (•, •) represents concatenation. The vector Z i is a concatenation of the atom's atomic number and a binary variable (or one-hot) encoding of the atomic number. (For a model that accommodates N atom types, Z i has length N +1, and the last N elements are all zeros except for a one in the position corresponding to the current atom type.) The network is also symmetrized with respect to a and b by averaging the output over the two possible orders, ensuring that predictions are independent of the order of the monomers. An important final adjustment to the internal neural network architecture of AP-Net is made to encourage correct asymptotic behavior. Rather than using a raw neural network output as a prediction for E ab,comp , the output of the last network layer is scaled by r −1 ab for every component. This normalization guides predictions to have small magnitude at large interatomic distance. Thus, r ab is both an input to the model and an explicit part of the functional form.</p><p>AP-Net was developed with version 2.1.0 of the Ten-sorFlow library. 49 The model was constructed to handle the six atom types present in the datasets described in Section II E (H, C, N, O, F, and S). For the ACSF feature, η is fixed at 100.0 and µ varies from 0.8 Å to 5.0 Å in increments of 0.1 Å. For the APSF feature, η is fixed at 25.0 and µ varies from −1.0 to 1.0 in increments of 0.1. The cutoff radius, r c is fixed at 8.0 Å. Each network consists of three dense layers of 128 nodes and a final output layer of a single node. As discussed previously, the output layer is scaled by r −1 ab to form E ab,comp . Additionally, both ACSF and APSF vectors are preprocessed with separate dense layers of 100 and 50 nodes, respectively. All layers use the rectified linear activation function, except for the output layer which uses a linear activation function. Each network is trained for 200 epochs using the Adam optimizer with a learning rate of 1.0 × 10 −4 and a batch size of a single molecule. 50 While we previously used a multi-target loss function that balanced the accuracy of individual component energies and total interaction energies, here we chose to train networks separately, prohibiting explicit error cancellation. 19 The training procedure minimizes the mean squared error of the predicted component energy. The network weights resulting in the lowest error on a held-out validation subset of the training data over the 200 epochs are used as in the final model. Eight randomly initialized networks are trained per SAPT component, and AP-Net reports the average prediction of the eight networks. The variance of the ensemble predictions can also be used as an uncertainty metric.</p><!><p>The hydrogen-bonded dimer dataset previously developed with SAPT-ML is revisited. 19 This dataset features interactions between N-methylacetamide (NMA)a popular model system for peptide bonding and universal hydrogen-bond donor and acceptor-paired with other small hydrogen-bond acceptors and donors. The training data was created by selecting a number of these other molecules, placing them in a favorable hydrogenbonding orientation with NMA, and then procedurally varying the monomer separation and orientation. Intramonomer geometry was also sampled with small, random perturbations from equilibrium so that models trained on the dataset are capable of disentangling the effect of intramolecular geometry on the interaction energy. This was done for 92 small molecules expressing 149 chemically distinguishable donor and acceptor sites hydrogen bonded with NMA in 7784 different configurations.</p><p>The training data was further supplemented with the 2192 neutral dimers in the sidechain-sidechain interaction (SSI) dataset. 51 The testing data consists of NMA in complexation with donors and acceptors absent from the training set, and was either taken from crystallographic data, 52,53 or, in the case of isoquinolone as an acceptor, generated using the same sampling method as described above for the training data. A subset of these configurations is shown in Figure 2. We will refer to these two datasets of dimers as NMA-training and NMA-testing. Throughout this work, a randomly selected subset of 498 dimers from NMA-training is used for validation as described in Section II D.</p><p>Experiments are performed to assess AP-Net's ability to describe a large expanse of the intermolecular PES of a single hydrogen-bonded dimer. We examine the H 2 O dimer and the dependence of the interaction energy on intermonomer separation and orientation. We report improvements on the shape, asymptotic behavior, and smoothness of the neural network PES.</p><p>We also test the transferability of AP-Net to the diverse S66x8 benchmark dataset, which contains 66 small dimers each at 8 different geometries along the radial dissociation curve. 21 The dimers consist of small, closedshell, neutral molecules and span a wide range of interaction types, from hydrogen-bonding to π − π stacking. Of the eight configurations, one is the equilibrium geometry, five are slightly dissociated (×1.05, ×1.10, ×1.25, ×1.50, and ×2.00), and two are slightly compressed (×0.95 and ×0.9).</p><!><p>We compare the accuracy of AP-Net to the reported accuracy of the original SAPT-ML model. Both models are trained on exactly the same NMA-training dataset, down to the random subset of the data used for validation. The performance of the two models on the NMAtesting dataset is shown in Table I. AP-Net exhibits a significant improvement in the prediction of all four SAPT0 components as well as the total SAPT0 interaction energy. For 8 of the 10 dimers, the AP-Net error is lower for all four energy components and the total SAPT0 energy. One exception is NMA/cyclohexanone, for which only five configurations are present in NMA-testing. With such a small sample, this behavior appears to be an instance of SAPT-ML benefiting from some combination of randomness and fortuitous error cancellation. Also, both models predict this interaction with small errors. The other dimer, 3-methylbutan-2-one features improved predictions for each component, but is narrowly worse at total interaction energies.</p><p>Although AP-Net's predictions for all four SAPT0 components are an improvement over SAPT-ML, the relative magnitudes of errors between components are largely the same for the two models. Errors in the prediction of E elst remain the highest, followed by E exch , E ind , and finally E disp . The average component error is partially explained by the range of possible energies for each component. Figure 3 shows that for the NMA-testing dataset, E elst and E exch take on a wider range of values (approximately 20 and 30 kcal mol −1 respectively) than E ind and E disp (both approximately 10 kcal mol −1 ), a trend that the average errors roughly mirror. However this is not a full explanation, since, for example E elst errors are actually larger than those of E exch , despite the labels having a smaller range. The existence of relative difficulties in predicting different SAPT components may imply that the physical nature of these interaction types governs how easily they can be modeled. This would explain why AP-Net does a particularly good job at predicting E disp , with an weighted total MAE of only 0.03 kcal mol −1 , an approximately 10-fold improvement over SAPT-ML: dispersion is often well modeled by an atomicpairwise functional form. 42 The use of an individual neu-ral network per component allows the model to separately learn the different physics governing each of them, a unique advantage of training to reproduce SAPT components rather than the total interaction energy. Further tailoring of these individual networks to match known physics could yield additional improvements. Another similarity between AP-Net and SAPT-ML is the relative difficulty of predicting interactions of different monomers in the NMA-testing dataset. The largest errors are still in the prediction of uracil and benzimidazole, and the smallest errors in the prediction of isoquinolone. We note that the experimentally measured NMA/uracil hydrogen bond lengths are shorter than any of the hydrogen bond lengths in the theoretically gen- Figure 4 shows a training saturation curve for AP-Net, illustrating the incremental improvements in accuracy with more training data. The validation error consistently declines for all four SAPT0 components as up to 9000 dimers are used in training. The lack of a plateau is a positive sign, as it suggests that the predictive capability of the relatively simple AP-Net architecture is not yet saturated. Improved performance could be attained simply by adding more training data than is used in this work. The low-data limit is also encouraging. Using only 200 training dimers, AP-Net's predicted total SAPT0 in-teraction energies already reach "chemical accuracy" or 1 kcal mol −1 MAE. The ability to make reasonable predictions with little data illustrates the appropriateness of the atomic-pairwise paradigm.</p><p>Another interesting detail in the saturation curve is the existence of a crossover between errors of the electrostatic and exchange components. This can probably be attributed to the variable and long-range nature of electrostatic interactions, which makes prediction slightly more difficult even with many training dimers. The electrostatic energy is the only SAPT component that can be both attractive and repulsive, and its r −1 decay is slowest among the four components. This learning curve also further reinforces the observation that the dispersion functional form is well described by a pairwise additive model. The validation MAE for E disp reaches 0.02 kcal mol −1 by 5000 training dimers. This error is much smaller than the errors expected in the SAPT0 dispersion energies themselves, as they are computed using only second-order perturbation theory. 25 Lastly, the parallel behavior of the electrostatic component and total interaction energy errors is particularly striking. Further improvements in the prediction of the SAPT0 interaction energy will necessitate focusing on this component.</p><!><p>Saturation curve of AP-Net mean absolute error in the total SAPT0 interaction energy and components. The number of training dimers is varied from 100 to 9000. The mean absolute error is computed on a randomly selected set of 498 validation dimers from the same distribution as the training data. All models were trained using the same procedure and hyperparameters described in Section II D.</p><!><p>Next, we examine the performance of AP-Net at describing the hydrogen-bonded water dimer. This dimer is absent from the NMA-training dataset, so this experiment is partially a test of AP-Net's ability to generalize to a different intermolecular interaction. More importantly, the H 2 O dimer is of incredible practical relevance and captures the essential, minimal hydrogen bond. We scan the intermolecular dissociation and rotation coordinates of the dimer, illustrated in Figure 5. In generating these coordinates, calculations are performed every 0.01Å or 0.5 • so that we can assess not only the asymptotic predictions of the models, but also the smoothness of the ML potential. Predicting smooth intermolecular PESs is a necessary requirement for using AP-Net in molecular dynamics or searches for stable intermolecular configurations.</p><p>FIG. 6. The total SAPT0 interaction energy of the H2O dimer along an intermonomer radial dissociation coordinate is plotted. Predictions of AP-Net and SAPT-ML models are compared to the true SAPT0 values.</p><p>The first coordinate probes the dependence of the hydrogen bond on intermonomer separation. Note that the monomer geometries are kept rigid, so any change in AP-Net's prediction along the scanned coordinate must attributed to changes in the features r ab , G ang a(b) , and G ang b(a) ; the other features (Z A , Z B , G rad a , and G rad b ) are constant. The total SAPT0 predictions of AP-Net and SAPT-ML in Figure 6 show that only AP-Net correctly captures the shape of the SAPT0 potential. AP-Net predicts a minimum in the total interaction energy at approximately the correct intermonomer separation, although the strength of the hydrogen bond is underestimated by approximately 1 kcal mol −1 . The most striking difference between SAPT-ML and AP-Net, however, is the smoothness of the intermolecular PES. For a given component, neighboring points on the SAPT-ML potential curve fluctuate as much as an entire kcal mol −1 . We believe this noisiness to be a result of the intermolecular descriptors used by SAPT-ML. The availability of actual distances as a feature in AP-Net's atom-pair paradigm combined with the carefully designed APSF feature is a significant improvement over the intermolecular ACSFs used in SAPT-ML. The second coordinate, shown in Figure 7, is a particularly challenging test of an intermolecular potential, as it isolates the angular dependence of the hydrogen bond. Not only are intramonomer geometries rigid, but also r OH is constant value of 1.95Å for the hydrogenoxygen pair participating in the hydrogen bond interaction. Therefore, most of the change in predicted energy over this coordinate must be accounted for by the new APSFs. AP-Net is still able to correctly predict the trend of the total interaction energy through the rotation. The decline in prediction quality at the smallest angle is a result of a strong repulsive clash between the two oxygen atoms, which become unreasonably close if we follow the curve all of the way to 90 • ; this repulsive contact is unlike any in the training dimers. The strong curvature of the potential in Figure 6 is an important result, as it shows that AP-Net is able to account for molecular anisotropy, a necessary aspect of accurate NCI potentials. Predictions are still smooth, especially when compared to SAPT-ML.</p><!><p>We have shown AP-Net to be an effective general model for hydrogen-bonded dimers when trained on the NMA-training dataset of similar hydrogen-bonded dimers. However, a useful intermolecular potential must also accurately describe all types of NCIs. This could obviously be accomplished by including additional and diverse training data. Here, we examine transferability from an alternate perspective: How well can the AP-Net model, a model specialized at hydrogen-bonding, describe other kinds of interactions? This type of analysis is necessary for ML potentials, since we cannot always expect a target system to be well represented in the model's training data. Some combination of interpolation and extrapolation within chemical space will always be necessary.</p><p>AP-Net's transferability is assessed by examining performance on the popular S66x8 benchmark dataset for intermolecular interactions. As a comparison, we also test the ANI-1X ML potential on the same benchmark. ANI-1X is a neural network potential for organic molecules, trained on a dataset of 5.5 million molecular conformations. This dataset includes many noncovalent complexes, making ANI-1X one of the best candidates for a robust intermolecular ML potential. Because AP-Net and ANI-1X are trained to match different levels of theory (SAPT0/jun-cc-pVDZ and ωB97X/6-31G* respectively), it would be unfair to compare each model's predictions to a single approximate reference method. Instead, we show predicted interaction energies for each model alongside the reference interaction energy at the same level of theory used to parameterize that model.</p><p>Of the 66 different dimers, two pathological but representative cases are shown: the benzene-benzene sandwich dimer in Figure 8 and the cyclopentane-neopentane dimer in Figure 9. Predictions on the remaining 64 dimers can be found in the SI. The intermolecular PESs predicted by AP-Net are at least as accurate and reasonable as ANI-1X's on average. For the benzene dimer, both AP-Net and ANI qualitatively predict the required interaction energy trend-repulsive at close separation and near zero at dissociation, with a slightly attractive minimum somewhere in between. However, ANI-1X predicts a steeper repulsive wall and more separated minimum, potentially as a result of missing the tricky chargepenetration effects known to occur in π − π interactions. AP-Net closely matches the entire potential, even though this dimer is chemically unlike the NMA hydrogen-bond dimers that make up its training data. AP-Net's behavior is unlikely to be the result of a lucky guess, since the predicted potentials of other π − π interactions in S66x8 are similarly correct. Presumably, the AP-Net model learned a representation of aromatic carbons via secondary interactions present in hydrogen-bonded dimers such as NMA/benzene.</p><p>The cyclopentane-neopentane dimer is an even more extreme comparison, and a good example of the disadvantages of relying solely on error statistics. Although the ANI-1X prediction is fairly accurate in terms of mean absolute error, the shape of the potential is unphysical, containing a spurious minimum and maximum along the coordinate. These artifacts would severely hinder the practical use of this potential. While AP-Net's prediction has only a slightly better MAE than that of ANI-1X, the predicted potential is parallel to the correct potential, and could be reasonably used for molecular dynamics or a rigid monomer geometry optimization. This occurs despite the fact that the AP-Net model was trained on a much smaller and more chemically homogeneous dataset than ANI-1X. The physical appropriateness of the atomic-pairwise representation is uniquely responsible for the dramatic generalization ability of AP-Net. It should also be emphasized that this type of purely dispersion-bound interaction was not well represented in the NMA-training dataset.</p><!><p>So far, the atomic-pairwise energy predictions of AP-Net are not individually used. They are summed to produce a predicted dimer interaction energy, which is then compared with the ab initio interaction energy. However, the existence of pairwise energy predictions is a unique feature of the atomic-pairwise paradigm, and analyzing these energies can provide insight into the AP-Net model and its representation of chemical systems. For example, AP-Net's pairwise partition of the interaction energy could be compared with empirical force-field models. Pairwise energy predictions of the H 2 O dimer along the radial dissociation coordinate from Section III B are shown in Figure 10.</p><p>As discussed earlier, the predicted SAPT0 interaction energy of this dimer as well as the four SAPT0 components correctly decrease in magnitude along the coordinate. Here, we see that the individual atompair predictions are similarly distance-dependent. This distance-dependence is a good validation of the AP-Net model, since it matches our intuitive understanding of interactions-any atom-pair interaction should grow weaker as the atoms become farther apart. AP-Net's predictions also qualitatively match the known physics of intermolecular interactions. Electrostatics, the longestrange SAPT component, is correctly predicted to decay slowest of the four components. The pairwise exchange, induction, and dispersion energies have the correct sign. These components are by definition only repulsive or attractive, and we would expect the pairwise energies to match this. It is also interesting to note that the signs of individual electrostatic energy predictions. Interactions between hydrogen and oxygen are attractive, while interactions between two atoms of the same type are repulsive. These predictions reflect the partial-charge picture used in virtually every force-field, even though AP-Net is not trained to predict anything related to electron density. This phenomenon is a natural consequence of the pairwise framework and is an exciting result, since it suggests that AP-Net contains some fundamental representation of the actual physics occurring in intermolecular interactions. This analysis of pairwise energies shows that AP-Net's predictions are uniquely physically interpretable, and it works towards countering the longstanding "blackbox" criticism of ML models.</p><p>In Figure ??, the interpretation of pairwise energies is taken a step further by comparing AP-Net's predictions to those of an ab initio force field developed by Van Vleet et al. 32 . The force field uses atomic dispersion coefficients (C 6 , C 8 , C 10 , C 12 ) computed from H 2 O monomer frequency-dependent polarizability tensors. Because this force field is fit to DFT-SAPT while AP-Net is fit to SAPT0, complete agreement between the two predictions is not expected. Still, the total dimer dispersion energy predicted by AP-Net and the force field are exceedingly close across the entire dissociation coordinate. Although the two predictions agree on the value of E disp , they differ significantly in the atomic-pairwise partition of E disp . AP-Net attributes most of the interaction energy to the close oxygen-hydrogen pair, while the force field assigns an approximately equal split between the close oxygenhydrogen pair and the oxygen-oxygen pair. The two partitionings are not completely irreconcilable, as they both yield the same ordering of pairwise interactions by magnitude. The qualitative agreement between AP-Net and the force field further validates the physical grounding of the AP-Net model. The close correspondence of AP-Net to an ab initio force field is something very few (if any) ML models can claim. AP-Net's predictions could be used to extract a different set of dispersion coefficients. This comparison also illustrates an important lesson in using any sort of energy partition, which is that there is not a singular 'correct' partition.</p><!><p>The development of ML potentials is a rapidly evolving endeavor, as evidenced by increasingly technical model architectures, more exhaustively constructed datasets, and lower reported errors on common benchmark tasks. As the field progresses, assessing and developing ML potentials for practical use will become even more important. One such practical application is the quantitative description of of NCIs, where estimated PESs must be smooth and physically reasonable. Although our previously developed SAPT-ML intermolecular potential obtained average errors near chemical accuracy, an investigation of the model's predictions revealed serious shortcomings in the estimated PESs.</p><p>In order to address these concerns, here we propose a different formulation of ML potentials specific to the intermolecular case. Instead of the usual atomic partition of energy central to nearly all ML models, the new formulation substitutes an atomic-pairwise partition of interaction energies. Although the energy partition used in any ML potential is arbitrary, an atomic-pairwise partition is decidedly more physically motivated, and therefore it should improve the accuracy and generalizability of any ML potential that make use of it. To test this claim we introduce AP-Net, an atomic-pairwise neural network model for the prediction of interaction energies. New ML descriptors that efficiently represent of a pair of atoms are also developed for AP-Net, including an APSF feature that captures the orientation dependence of monomers, a key aspect of NCIs.</p><p>AP-Net is applied to the SAPT0 hydrogen-bonding task developed with the atomic SAPT-ML model, and we find that the new atomic-pairwise model yields dramatic improvements in both the accuracy of single-point energies and the smoothness of predicted potentials. On an experimental NMA hydrogen-bond test dataset, AP-Net predicts a smooth and relatively accurate intermolecular PES for the H 2 O dimer, correctly describing both the radial and angular dependence of a hydrogen-bond, after being trained on a dataset entirely absent of this dimer.</p><p>This hydrogen-bond specialized AP-Net model also shows a surprising ability to generalize across chemicalspace, achieving a mean absolute error of 1.1 kcal mol −1 on the entire S66x8 noncovalent interaction benchmark. This level of accuracy surpasses that of the general universal neural network potential ANI-1X. AP-Net predicts more physically reasonable potentials than the aforementioned ML potential on this benchmark, free of spurious optima, while using fewer and less diverse training data.</p><p>The ability of AP-Net to make reasonable predictions for disparate interaction types, i.e. extrapolation across chemical space, is an incredibly important characteristic of an ML potential. This behavior is a direct consequence of the atomic-pairwise framework, which provides the model with a physically motivated inductive bias. Put another way, AP-Net regresses over many atomatom interactions, while an atomic model like SAPT-ML regresses over fewer atom-monomer interactions. The atom-atom 'chemical-space' is much smaller than the atom-monomer 'chemical-space', making the regression problem simpler and the predictions more accurate under an atomic-pairwise energy partition.</p><p>Future work related to AP-Net will focus on using a larger, more chemically diverse dataset that samples a greater expanse of the intermolecular PES. Curating more accurate interaction energies than the SAPT0 labels used here is also pertinent, given that AP-Net's errors with respect to SAPT0 are on average smaller than the errors in the ab initio SAPT0 calculation. Lastly, we note that the general atomic-pairwise framework advocated for in this work could be easily adapted to work with other features, model architectures, and learning tasks.</p><!><p>The supplementary material contains additional analysis of AP-Net's predictions on the NMA-testing dataset. Predictions on all 66 radial coordinates in the S66x8 benchmark are also included.</p><!><p>Code to create an AP-Net model, all datasets used in this work, and the trained AP-Net model used in this work will be included with the final manuscript.</p>

ChemRxiv

Fundamental Analysis of Piezocatalysis Process on the Surfaces of Strained Piezoelectric Materials

Recently, the strain state of a piezoelectric electrode has been found to impact the electrochemical activity taking place between the piezoelectric material and its solution environment. This effect, dubbed piezocatalysis, is prominent in piezoelectric materials because the strain state and electronic state of these materials are strongly coupled. Herein we develop a general theoretical analysis of the piezocatalysis process utilizing well-established piezoelectric, semiconductor, molecular orbital and electrochemistry frameworks. The analysis shows good agreement with experimental results, reproducing the time-dependent voltage drop and H 2 production behaviors of an oscillating piezoelectric Pb(Mg 1/3 Nb 2/3 )O 3 -32PbTiO 3 (PMN-PT) cantilever in deionized water environment. This study provides general guidance for future experiments utilizing different piezoelectric materials, such as ZnO, BaTiO 3 , PbTiO 3 , and PMN-PT. Our analysis indicates a high piezoelectric coupling coefficient and a low electrical conductivity are desired for enabling high electrochemical activity; whereas electrical permittivity must be optimized to balance piezoelectric and capacitive effects. Piezocatalysis is a new approach toward enabling or enhancing electrochemical processes by making use of the strain state of a piezoelectric material 1 . Piezocatalysis is the product of an intimate interaction between the native electronic state of the piezoelectric material, the chemistry of the surrounding medium, and a strain induced piezoelectric potential. The action of mechanically deforming a piezoelectric material induces a perfuse electric field which augments the energetics of both free and bound charges throughout the material 1,2 . The thermodynamic feasibility and kinetics of electrochemical processes occurring at the surface of the piezoelectric material sensitively depend upon the electrochemical potential difference between charges on the piezoelectric's surface and in the surrounding medium 3-6 . Thus piezoelectric potential, which can dramatically affect the difference between these electrochemical potentials, is a new means of modulating the material's electrochemical activity via its strain state.Recently, a piezocatalysis process was demonstrated by our study of a strained ferroelectric Pb(Mg 1/3 Nb 2/3 )O 3 -32PbTiO 3 (PMN-PT) beam in a deionized (DI) water system, from which a strong dependence of hydrogen evolution from the aqueous surroundings on the material's piezoelectric potential was observed 1 . In addition, numerous recent works have confirmed the correlation between electrochemical activity and piezoelectric or ferroelectric polarization in a broader sense. For example, a study conducted using ferroelectric poly(vinylidene fluoride) (PVDF) has demonstrate that in-situ piezopotential can influence lithium battery charging behavior 7 . Electrochemical deposition was found to be selectively activated by the ferroelectric domain polarization [8][9][10][11][12][13][14][15] . However, to date there is no general theoretical analysis of the piezopotential's effect on electrochemical activities, e.g. how one piezoelectric material's activity differs from another; the influence of metallic electrodes as compared to bare piezoelectric surface on reaction output, and how free charge in piezoelectric material systems affects the piezocatalysis process. In this paper, we address the piezocatalysis system in generality to elucidate the details underlying these open questions and illuminate trends for further experimental study.In order to clearly understand the piezocatalysis process, a conventional electrocatalysis process is discussed first, where the application of electrical potential from an external power source is a typical means of driving the electron transfer reactions (Fig. 1a). The applied potential can result in one of the following two processes: (1) lowering electronic energy levels of unoccupied states within the electrode to a magnitude less than that of the highest occupied molecular orbital (HOMO) in solution; (2) raising occupied states within the electrode above the lowest unoccupied molecular orbital (LUMO) in solution. Under the first condition, electrons will leave the HOMOs in solution and transfer to the unoccupied states within the electrode -oxidizing the solution

fundamental_analysis_of_piezocatalysis_process_on_the_surfaces_of_strained_piezoelectric_materials

<!>Discussion<!>Methods

<p>(center panel of Fig. 1a). Under the second condition, electrons will leave the occupied states in the electrode and transfer to the LUMOs in solution -reducing the solution (right panel of Fig. 1a). When the electrode is a non-metallic solid, the eQ HOMO and ew LUMO logic extends to the materials valence (eQ VB ) and conduction band (eQ CB ) edges, respectively.</p><p>For piezocatalysis, the external power source is replaced by piezoelectric potential which results from the piezoelectric polarization (P PZ ). A perfectly insulating piezoelectric material is the most ideal and simplest case and it will be analyzed first. For such a material, mechanical deformation creates a perfuse electric field inducing a total energy shift across the material, V p max , given by:</p><p>where T k is an applied stress in the k dimension, d xk are the piezoelectric moduli, e 0 is the electrical permittivity of free space, and e r,x is the relative permittivity in the x dimension and W x is the width of the piezoelectric material in the x dimension. In a one-dimensional case, x and k are equal and the subscripts are dropped. This piezoelectric potential, which changes the energetics of the valence band (VB) and conduction band (CB) across the piezoelectric material, can have a dramatic effect on the material's interaction with its environment (Fig. 1b). In this model it is assumed that stain does not change the magnitude of the band gap. If the eQ VB approaches ew LUMO , it becomes energetically favorable for electrons to leave the VB and enter the LUMO (center panel of Fig. 1b). If the eQ CB approaches eQ HOMO , it becomes energetically favorable for electrons to leave the HOMO and enter the CB (right panel of Fig. 1b). Placing metal electrodes between the piezoelectric and solution, with their continuous density of states about their Femi energies, simplifies this situation: the piezopotential now acts as a bias, lifting and lowering the metal's Fermi energy eQ M (Fig. 1c).</p><p>Piezoelectric materials, which are the source of bias in the case of piezocatalysis, are capable of achieving extremely high potentials (tens to hundreds of volts) when subjected to moderate to severe strain 16 . Under such circumstances (i.e. where electrode potentials versus standard hydrogen electrode (SHE) exceeding ,3 volts) many chemical species in contact with the piezoelectric will be thermodynamically capable of undergoing reduction or oxidation reactions.</p><p>In order to deduce the maximum quantity of oxidation or reduction reactions possible for a given deformation, it is necessary to quantify the interaction between charge exchange at the piezoelectric/electrolyte interface and the effect of charge exchange on the piezopotential. A characteristic of piezocatalytic systems is the piezopotential drop that takes place during charge transfer to and from its surfaces (or electrodes). In this way the system acts as a capacitor. The rate of piezopotential drop is dependent upon both the piezoelectric material's properties and the nature of the solution in which it is submerged. In the case where the concentration of reactive species in solution is low, e.g. when pure Milli-Q water is used for the hydrogen reduction reaction, and the power available for driving electron-transfer reactions is sufficiently high, the rate of electrochemical reactions at the electrode is dominated by the diffusion rate of redox species to the electrode. In the case of Milli-Q water, the redox species are protons, hydroxide and impurity species. Mass transfer to the electrode is described by the Nerst-Planck equation, which in its one-dimensional form along the x-axis is written as:</p><p>where J i (x) is the flux of species i at distance x from the surface, D i is the diffusion coefficient, LC i (x) Lx is the concentration gradient at distance x, F is the Faraday constant, R is the gas constant, T is absolute temperature, LQ(x) Lx is the potential gradient, C i is the concentration of species i, and n(x) is the velocity with which a volume element in solution moves along the axis 17 . The expression describes the contributions of diffusion, migration and convention, respectively, to the flux of species i.</p><p>Neglecting migration and convective phenomena, the flux of species i is dependent upon diffusion alone. In the diffusion limited regime, the maximum rate of electron transfer from the piezoelectric is equal to the rate of reactant diffusion to the piezoelectric, resulting in a current density j given by:</p><p>where n i is the number of electrons per reaction event with species i 17 .</p><p>Applying appropriate boundary conditions to equation ( 3) and taking account of the capacitive nature of the piezoelectric bias source, yields an expression of the piezopotential V p as a function of time (a variant on the Cottrell equation):</p><p>where c i is the bulk concentration of species i, f i is a parameter taking on a value between 0 and 1 that toggles the kinetics of the electrodes' reactivates with species i, and t is time 18 . Under conditions of low reactant and electrolyte concentrations and high positive electrode potential (ew LUMO .2eV vs SHE), a kinetics parameter (f i ) less than 1 causes a semi-diffusion controlled regime to form where charged reactant species (e.g. protons and metal ions) capactively couple to the electrode's surface, effectively reducing the surface potential before they are electrochemically reduced or oxidized (Fig. 2a). Under the application of a large positive electrode potential, a dilute (e.g. Milli-Q water) system's capacitance is well approximated by the Helmholtz model 17 . The voltage drop expected in time t by both electron transfer reactions and capacitive effects is given by integrating equation (4):</p><p>where z i is the charge sign (1 for cation, 21 for anion) of species i, w H is the thickness of the Helmholtz layer, and l is the number of different species in solution. The positive potential of the electrode dictates that a positive charge in solution will contribute to a positive capacitive current, lowering the piezopotential. The first term in equation ( 5) describes the piezopotential change as redox reactions proceed, the second term describes the potential change due to capacitive coupling of not yet reacted species at the piezoelectric material's surface. Equation ( 5) governs how the potential on the piezoelectric's wall should drop from the time (t 5 0) of initial mechanical deformation, when the piezopotential can be a value of tens or hundreds of volts, to a time (t 5 t p ) when the potential has been reduced to a reasonable value about the SHE (e.g. 62 V vs SHE). Between time 0 , t , t p , the electrode potential is sufficiently high as to reduce or oxidize species within solution in an effectively nonselective manner. At t . t p , the piezoelectric potential falls within the range of typical eQ HOMO and ew LUMO values of ions in solution resulting in a more selective reduction and oxidation process (Fig. 2b).</p><p>The processes described above pertain to both surfaces of the piezoelectric material; on one surface oxidation dominates while reduction dominates on the other. The full voltage change, as measured across the piezoelectric material, is a combination of both of these processes happening simultaneously. An electrode potential above (in the case of oxidation current) or below (reduction current) the selective potential window of ,62 V versus SHE here are treated identically. This results in a total voltage reduction across the piezoelectric material of twice the expected magnitude from a single wall (equation ( 5)).</p><p>As the piezopotential reaches a modest range about 0 V (t 5 t p ), the electrochemical activity of some ions is effectively switched off while others will remain highly active. Those ions which are no longer electrochemically active are still free to participate in capacitive effects (Fig. 2b and second term in equation ( 5)). The voltage decrease across the piezoelectric from t p $ t $ ' is thus due to a separate set of electrochemical and capacitive effects conducted by a subset (i...m) of the total chemical species present (i…l). By applying equation ( 5) to both the selective and unselective regime we can describe the magnitude of piezopotential at all times by the following derived expression:</p><p>where H(x) is the Heaviside function defined as 0 when x is less than 0, and 1 when x is equal to or greater than 0. e H is the electrical permittivity of the Helmholtz layer. The third and fourth terms in equation ( 6) acts to add continuity between the selective and unselective regimes. Equation ( 6) is a simplification of the activity occurring around the piezoelectric as the piezopotential drops to within the vicinity of eQ HOMO and ew LUMO of species in solution. As the energetic advantage for electron transfer diminishes, the current density j begins to show its exponential dependence upon driving force (eQ HOMO 2 ew LUMO ). The current density exponentially depends upon applied potential about E e :</p><p>where j 0 is the exchange current density and is equal to:</p><p>n is the number of electrons transferred in the reaction, k 0 a and k 0 C are the rate constants for the anodic and cathodic current, respectively, C R and C O are the concentrations of the reduced and oxidized species, respectively, a a is the anodic transfer coefficient, E e is the redox potential of a species in solution, and g is the applied bias (i.e. Fermi energy difference between eQ m and E e ) 18 . For the purposes of simplification, the effects of equation ( 7) will be approximated by utilizing two different kinetics constants acting in equation ( 6) for 1 $ i $ m: f 1,i and f 2,i for the nonselective time interval 0 # t # t p and the selective interval t p $ t $ ', respectively.</p><p>To elucidate the behavior of the system, Fig. 2c is a plot of equation ( 6) as it is applied to the case of a piezoelectric material (PMN-PT) strained in Milli-Q water. We take a value of electrical permittivity e PMN{PT,x 5 8000 and a thickness of PMN-PT w th 5 .23 mm 19 . The value of the Helmhotlz layer thickness (w H ) was fixed to 2.75A ˚with a e H 5 80. Under standard conditions, Milli-Q water has a resistivity value of 18 V?cm due to the contribution of l ''impurities'', including hydroxide and hydronium (c 1,2 5 ,0.100 ppb?mol?cm 23 , D 1,2 9:3 Ã 10 {5 cm 2 sec), organics (c 3 5 1.106 ppb?mol?cm 23 , D 3 ~0:52Ã 10 {5 cm 2 sec), silicates (c 4 5 0.5534 ppb?mol?cm 23 , D 4 ~1:28Ã 10 {5 cm 2 sec), and heavy metal ions (calcium, sodium, chloride, etc; c 5 5 0.1106 ppb?mol?cm 23 , D 5 ~1:2 Ã 10 {5 cm 2 sec) 20 . Based on experimental measurements discussed in the following section, t p is chosen at 0.042 s and f 1i is 0.715 (between 0 §t.t p , 71.5% of all species immediately undergo redox reactions upon their contact with the electrode, 28.5% of species first capacitively couple to the electrode).</p><p>The physical consequences of the voltage decrease described by equation (6) are: (1) the creation of reaction byproducts in the medium surrounding the piezoelectric material (e.g. hydrogen gas, oxygen gas, chlorine gas, oxidized organics, etc); (2) the buildup of a capacitive layer in solution alongside the piezoelectric material; and (3) a reduction in the electric field present inside the piezoelectric material. These effects will continue to change with time until the potential decrease (equation ( 6)) approaches the total potential originally generated (equation ( 1)). The rate of these effects depend sensitively upon the rate constants f 1i and f 2i , the concentration of reactive species and their diffusion coefficients. The natural driving force for these chemical changes is the piezopotential which, as described by equation ( 1), depends upon the electrical permittivity, mechanical strain, Young's moduli and piezoelectric moduli. This is true regardless of whether the electrochemically active medium is in contact with metallic electrodes or the piezoelectric material's bare surfaces. In the case of bare piezoelectric surfaces, the band gap and positions of ew CB and ew VB relative to eQ HOMO and ew LUMO also determine the nature of what happens during straining.</p><p>When comparing the piezocatalytic abilities of various piezoelectric materials, it is useful to choose a performance metric from equation ( 6) such as the quantity of a reaction byproduct, e.g. H 2 gas. This value depends sensitively upon the parameters we choose. Previous work reported the production of hydrogen gas via piezocatalysis between a PMN-PT single crystal slab (2 mm 3 10 mm in size) and Milli-Q water 1 . Using recursion we are able to fit equation (6) to the piezopotential verses time profile obtained from a working electrode on a PMN-PT surface under one strain (Fig. 2d). The results yield t p 5 0.042 s, f 1i 5 0.715, and f 2i 5 0.07 under the constraints that after t 5 t p only H 1 reduction and H 2 O oxidation were possible at the electrodes, whilst unreacted H 1 , OH 2 , silicate and heavy metal ions only acted capacitively. These low kinetics values likely resulted from a low concentration of electrolyte. The switch over at t p 5 0.042 s from one reaction regime to the other occurred at a potential (difference) value between the electrodes of approximately 2 V. A gold electrode in equilibrium with pH 7 Milli-Q water is in equilibrium with the H 1 reduction and H 2 O oxidation reactions. This places the Au electrode's potential (Q ElectrodeHOMO ) at 0.41 V vs. SHE (0.615 V below the eQ LUMO of H z and 20.615 above the eQ HOMO of H 2 O). A symmetric voltage shift applied to both Au electrodes of 1 V at t p means an overpotential (Q Electrode HOMO { Q LUMO ) of 20.385 V for the H z reduction reaction and an overpotential (Q Electrode HOMO {Q LUMO ) of 0.385 V for the H 2 O oxidation reaction. These are the potential values taken at which the system switches from a nonselective regime to a selective regime.</p><p>In order to derive a function of H 2 evolution dependency on strain, the minimum strain S min,1 capable of driving electrochemical reactions for an arbitrary piezoelectric is necessary and determined by augmenting equation ( 1):</p><p>where Y is the Young's modulus of the material. For metallic Au electrodes, the preceding analysis determined the potential necessary to drive electrochemical H 2 production under the observed conditions to be 0.615 (Q Electrode HOMO {Q LUMO ). In the case of a bare piezoelectric, (Q Electrode HOMO {Q LUMO ) becomes Q VB {Q LUMO ð Þ . S min,2 is denoted as the strain necessary to transition from the selective to nonselective regime, e.g. where (Q Electrode HOMO {Q LUMO ) in equation ( 9) becomes (Q Electrode HOMO {Q LUMO z0:385V). Using the value parameters fit from experiment in conjunction with the values of piezoelectric materials' parameters listed in Table 1, the H 2 generation capacity per straining event (H Metal,Total ) for a multitude of Au-electrode coated piezoelectric materials is depicted in Figure 3a and is given by a manipulation of equation ( 6):</p><p>H Metal,Total ~H(S{S min,1 )H Metal,SmallS H(S min,2 {S)z</p><p>H Metal,SmallS is the H 2 production for piezopotentials that fall within the selective regime and is given by:</p><p>where t 1 is the time required to exhaust the hydrogen production capability for a piezoelectric operating only within the selective regime (S min,1 vSvS min,2 ) and is given by:</p><p>where e H2O is the electrical permittivity of water. H Metal,LargeS is the H 2 production for piezopotentials that reach the unbiased regime (SwS min,2 ) and is given by:</p><p>where t 2 (equation ( 14)) is the time it takes for the piezopotential to fall from an arbitrarily large potential (V p max ) to that which brings it into the selective regime (Q LUMO 1 Q Op , where Q Op is the overpotential necessary to transition between reaction regimes, e.g. 0.385 V), and t 3 (equation ( 15)) is the time it takes for that same piezoelectric to reduce its potential from</p><p>and</p><p>At large strains (strain . .00025), PZT dominates H 2 gas production per unit strain with a value of 9.1 3 10 14 H 2 ?strain, PMN-PT and BTO have similar production rates of 5.4 3 10 14 H 2 ?strain and 4.5 3 10 14 H 2 ?strain respectively, while ZnO maintains a modest H 2 production rate of 79 3 10 13 H 2 ?strain. From equation ( 13) it can be determined that the H 2 production depends upon the relative weight of f 1,H z to f 2,H z, d xk and e r,x . When f 1,H z is much larger than f 2,H z, ffiffiffi ffi t 2 p dominates the expression where a large ffiffiffi ffi t 2 p value results from a larger V p max combined with a large e r,x . However e r,x is contained within the denominator of the expression determining V p max (equation (1)). Thus, while a large d xk is always valuable to the piezocatalysis process, a balance must be achieved with e r,x . This is why PZT, possessing both large d xk and moderately large e r,x , has the largest H 2 production rate. Physically this competition represents a balancing between the piezoelectric, voltage generating capabilities of the material, and the capacitive nature of the material. The inset in Figure 3a enlarges the small strain region (strain , 0.00004), demonstrating the dynamics of initiation H 2 production. Materials with the smallest e r,x d xk ratio, the most prominent being ZnO, begin H 2 production at strains far smaller than a material like PMN-PT. This is because the thermodynamic accessibility of electrochemical reactions depends upon the relative energetics of donor (eQ HOMO ) and acceptor (ew LUMO ) states and a high ratio of piezoelectric moduli to electrical permittivity strongly couples the strain state of the material to these state energetics (equation ( 9)).</p><p>The predicted H 2 production capacity is compared to experimental data obtained from an oscillating gold coated PMN-PT cantilever with a peak piezopotential of 20 volts (Fig. 3b) 1 . At 20 Hz, each straining action takes place during 0.025 seconds, which is insufficient for depleting all piezopotential and reaching a thermodynamic equilibrium. Nonetheless, an approximation of the experiment can be constructed by truncating the nonselective regime at 0.025 seconds and applying a window of acceptable values on the kinetics parameter f i from 0.07 to 0.715 (inset of Fig. 3b). Experimental data fall around the lower bound of predicted H 2 evolution rate, exhibiting good agreement. This indicates that there remains much room for experimentally improving the H 2 production rate by means of adding a proton reduction co-catalyst and thus increasing the f i value.</p><p>Several factors change when comparing a piezoelectric material coated with metal electrodes to a naked, insulating one. In general, a naked and perfectly insulating piezoelectric material does not necessarily possess a continuum of states about the redox potentials in solution. To achieve piezocatalysis (including H 2 production) with such a piezoelectric material, the conduction band and valence band must be moved sufficiently in potential so that their energies coincide with redox potentials under examination. Taking the production of H 2 as our prototypical case, the energetics of top-most valence band electrons (eQ VB ) must be lifted above the (ew LUMO ) of H 2 . The minimum strain, and thus potential, required to achieve this criteria is again given by equation ( 9), except in the case of a bare piezoelectric Q ElectrodeHOMO becomes Q VB . The values of potentials necessary to cause the reduction of H 1 and oxidation of H 2 O for various piezoelectric materials can be found in Table 1. The hydrogen production capacity of a bare piezoelectric material can be calculated by modifying equation ( 10):</p><p>H Insulator,Total ~H(S min,1 {S)H Insulator,SmallS H(S{S min,2 )z</p><p>where H Insulator,SmallS is the H 2 production for piezopotentials that fall within the selective regime (S min,1 vSvS min,2 ), H Insulator,LargeS is the H 2 production for piezopotentials capable to reaching the nonselective regime (SwS min,2 ).</p><p>Applying the values from Table 1 and the same kinetic parameters as in the metal electrode case (f 1i 5 0.715, f 2i 5 0.07, and Q Op 5 20.385 V), Fig. 3c shows the H 2 production rate for various naked piezoelectric materials. Subjected to large strains (strain..001), both naked and Au coated piezoelectric materials perform consonantly. Discrepancies between these two are more prominent under small strain (strain,0.0002), which are a result of the difference in the band structures of individual piezoelectric materials. Under small strain, each piezoelectric's H 2 -production's turn-on strain depends not only upon the e r,x d xk ratio but also upon the value of the individual piezoelectric material's (eQ VB {eQ LUMO ) value. Concerns over the stability of a piezocatalysis system comprising a bare piezoelectric material can be addressed by comparing piezocatalysis with a controlled corrosion process, where reduction reactions take place through the piezoelectric's oxidation and vice versa. The maximum piezoelectric charge density at the surface is on the order of 10 13 charge?cm 2 , two orders of magnitude lower than the average atomic density of a solid-state surface. With oxidation events of this magnitude, it is expected that a piezoelectric capable of withstanding photocatalytic reactions with water is also able to survive the piezocatalysis process.</p><p>In contrast to the previous analysis, most piezoelectric materials are not perfect insulators, mobile electrical charges inside the piezoelectric material respond to and rearrange according to its internal electric field. In bulk systems, these mobile charges act as extremely effective screening agents of the piezopotential, resulting in an effective piezopotential given by a manipulation of the Gouy-Chapman capacitance model:</p><p>where k is Boltzmann's constant,, and n 0 is the bulk concentration of free charge 17 . According to equation (17), a highly insulating bulk piezoelectric is required to reach an appreciable potential (Fig. 4a).</p><p>Taking the values from Table 1 for various piezoelectric materials subjected to a strain of 0.002, the maximum concentration of free charges allowed in order for the piezoelectric materials' surfaces to obtain the H 2 production potential were 5.22 PZT respectively. The H 2 production capacity for these piezoelectric materials subjected to 0.2% strain was determined as a function of free charge by manipulating equation ( 16):</p><p>where H Highn0 is the expression for H 2 production while the piezopotential is within the selective regime and H Lown0 describes the H 2 production for a piezoelectric that has achieved a potential for accessing the nonselective regime. n max,1 is the maximum number of free charges allowed for a given piezoelectric material under 0.2% strain to reach the threshold potential Q VB {Q LUMO ð Þfor driving H 2 production. n max,2 is the charge concentration necessary to reach the potential threshold between the two reaction regimes</p><p>In equation (18) the dependence of H 2 production on n 0 comes exclusively from the fact that in the case of a semiconducting piezoelectric material, V p,max [equation ( 13) & ( 14)] becomes V Semi [equation (17)].</p><p>The effect of free charge on H 2 production for various piezoelectric materials is demonstrated in Fig. 4b, where the hydrogen production for a strain of 0.2% is shown as a function of mobile charge concentration, n 0 . Behavior of particular note is the dramatic increase in H 2 production that results from decreasing the mobile charge concentration ,5 orders of magnitude below the minimum concentration required to reach the H 2 potential for the piezoelectric materials. Additional mobile charge reduction beyond that initial 5 orders reduction has relatively little effect on H 2 evolution.</p><!><p>A number of factors have been shown to augment the efficacy of strain to induce electrochemical reactions. The relative energies of states within the electrode with respect to HOMO and LUMO energies in solution can both dramatically change the rate of chemical reactions and be a determining factor in which chemical reactions are allowed to proceed. These energy state positions depend directly on the magnitude of strain and piezoelectric coefficient while depending inversely upon the electrical permittivity. In the presence of freecharge, generated either by doping, photo or thermal excitations, the piezopotential can be markedly decreased with direct repercussions on reactivity. In order to increase the electrochemical activity of a strained piezoelectric, the value electrical permittivity needs to be optimized. The piezoelectric, semiconductor and molecular orbital frameworks discussed herein have historically and successfully been applied to solid state and electrochemical systems. It is expected that these frameworks can be extended to the piezocatalysis theory. The theoretical work presented here can provide fundamental guidance for designing and understanding the novel piezocatalysis phenomenon from broad electrochemistry and piezoelectric material systems.</p><!><p>All calculations were performed in Wolfram Mathematica 8. To calculate the total H 2 output from an insulating piezoelectric with metal electrodes, we first calculated the H 2 generated as the piezoelectric potential dropped from the maximum potential induced by strain [equation( 1)] to the potential present at time t P . Then, the H 2 generated after time t p was calculated using another value of the kinetic parameter. Summation of these two H 2 quantities gave the total H 2 output. The first time segment utilized kinetics parameter f 1 5 0.715; while the second segment utilized f 2 5 0.07. To accomplish this we used equation (10), using the variable values found within the paper. Species i 5 1, … m was designated as H 1 ions only, while i 5 m 1 1, … l included all other species in solution.</p><p>Calculations conducted on bare, insulating piezoelectric materials used the same methodology as those done for metal electrodes-covered insulating piezoelectric materials. The only difference is that the valence band and conduction band energies of the piezoelectric material were used in place of the metal's Fermi energy (equation 16 and Table 1). In the case of the semiconducting piezoelectric, the same procedure was followed the case of the insulating piezoelectric [equation (18)] except that V Semi [equation ( 17)) was used in place of V P max .</p><p>The experimental data was acquired by using a PMN-PT single crystal slab as the piezoelectric component for piezocatalyzed water splitting (from our previously publication 1 ). The piezoelectric cantilever architecture was constructed and placed within a sealed chamber with access ports for piezoelectric actuation and monitoring, environmental purging, and atmospheric sampling. Straining of the piezoelectric cantilever was achieved by external electronic actuators. The H 2 concentration was determined by an AMETEK Analyzer ta3000F H 2 gas analyzer. The piezoelectric potential was measured by a digital oscilloscope and a potentiostat, respectively. See reference 1 for more details.</p>

Scientific Reports - Nature

Recognizing chemicals in patents: a comparative analysis

Recently, methods for Chemical Named Entity Recognition (NER) have gained substantial interest, driven by the need for automatically analyzing todays ever growing collections of biomedical text. Chemical NER for patents is particularly essential due to the high economic importance of pharmaceutical findings. However, NER on patents has essentially been neglected by the research community for long, mostly because of the lack of enough annotated corpora. A recent international competition specifically targeted this task, but evaluated tools only on gold standard patent abstracts instead of full patents; furthermore, results from such competitions are often difficult to extrapolate to real-life settings due to the relatively high homogeneity of training and test data. Here, we evaluate the two state-of-the-art chemical NER tools, tmChem and ChemSpot, on four different annotated patent corpora, two of which consist of full texts. We study the overall performance of the tools, compare their results at the instance level, report on high-recall and high-precision ensembles, and perform cross-corpus and intra-corpus evaluations. Our findings indicate that full patents are considerably harder to analyze than patent abstracts and clearly confirm the common wisdom that using the same text genre (patent vs. scientific) and text type (abstract vs. full text) for training and testing is a pre-requisite for achieving high quality text mining results.

recognizing_chemicals_in_patents:_a_comparative_analysis

Background<!>Methods<!><!>Patent corpora<!><!>Chemical NER systems<!>Ensemble NER systems<!>Evaluation metrics<!>Text preprocessing<!>Results<!><!>Cross-genre evaluation<!><!>Comparison at instance level<!><!>Error distribution<!><!>Impact of simple chemical elements<!><!>Ensemble performance<!><!>Cross-text-genre to cross-corpus evaluation<!>Discussion<!>Difference between scientific articles and patents<!><!>Execution time analysis<!><!>Text-genre statistics<!>Difference between patent full texts and patent abstracts<!><!>Impact of different annotation guidelines<!>Conclusion<!>

<p>Patents are an economically important type of text directly related to the commercial exploitation of research results. They are particularly essential for the pharmaceutical industry, where novel findings, such as new therapeutics or medicinal procedures, result from extremely cost-intensive, long-running research projects, but often are relatively easy to copy or reproduce [1]. Accordingly, a large number of commercial services exists regarding the formulation and retrieval of patents [2], and large companies devote entire departments to the creation, the licensing, and the defense of their patent portfolio. Such services must be supported by proper computational tools, as the number of patents is increasing rapidly. For instance, the European Patent Office granted 614,850 patents since 2006 [3]; the size of the United States Patent and Trademark Office corpus currently is 6,718,054 patents with a yearly increase of roughly 300,000 over the last 5 years [4]. However, current tools for patent management mostly support keyword search [5–9], whereas only few projects exist that target the extraction of specific facts from patents [10, 11].</p><p>In this work, we study the identification and extraction of chemical names1 from patents. By extraction, we mean the identification of left and right borders of mentions in patents, a task usually referred to as Named Entity Recognition (NER). Extracting chemicals from scientific articles has been a topic of ample research over the last 15 years, leading to the creation of high quality tools like OSCAR [12] or ChemSpot [13] which focus on the particularities of chemical names when compared to other entities, such as genes or species [13–16]. However, the extraction of chemicals from patents has been neglected by the research community for long, mostly due to the difficulties in obtaining computer-readable patents at large—compared to the simple procedures necessary to download scientific articles from sources like PubMedCentral2—and the lack of properly annotated patents, i.e. gold standard corpora. It is tempting to apply tools and models developed for scientific articles on patents, but patent texts are quite different from scientific articles. They are typically much longer, yet have a lower word density [17]. Their writing is more difficult to understand as the protection of broad claims and a mild obfuscation of procedures are established means to increase patent value and decrease the likelihood of being reproduced [5]. Therefore, it is rather unclear whether tools developed for scientific articles perform equally well on patent data.</p><p>Since 2012, two gold standard full-text patent corpora have been published: the chapati corpus [18] and the corpus from the BioSemantics research group [19]. The field was further boosted by a recent international competition, the CEMP task at BioCreative V. For this task, two large corpora for training and development were prepared and used by 21 teams to develop patent-specific solutions [20–22], achieving the F-measure values of up to 89% (87% precision and 91% recall) using an ensemble approach. However, both corpora consist only of patent titles and abstracts, while commercially interesting applications critically depend on analyzing full texts, as a significant number of entities is not even mentioned in an abstract [17]. Furthermore, international challenges are important to make different approaches comparable and also provide a strong incentive for groups to enter a field [11], yet their performance results are difficult to extrapolate due to the relatively high homogeneity of training and test data within the competition. In contrast, real applications typically have to perform information extraction on diverse text collections without having accordingly diverse training data. To mimic such situations, cross-corpus evaluations can be used, where the performance of a tool trained on one corpus is measured on another corpus following different annotation guidelines [23].</p><p>In this paper, we take this idea one step further and perform a cross-text-genre evaluation by assessing the performance of chemical NER tools trained on scientific articles—a problem much better researched—on patent corpora. We choose tmChem [24] and ChemSpot [13], two state-of-the-art tools for chemical NER from scientific articles, and evaluate their performance (without retraining) on all four freely available gold standard patent corpora with annotations of chemical mentions. We put emphasis on the differences between evaluations on abstracts versus full texts, showing that the latter is a considerably harder task for current tools. We also compare results on the instance level, showing that, despite having similar performance numbers, tmChem and ChemSpot actually return quite different results. This makes the creation of ensembles attractive, two of which we evaluate on all four corpora. We also contrast our cross-text-genre results to those obtained after retraining a chemical NER tool on patent corpora, showing that taking away the text-genre difference significantly boosts results, i.e., that the different characteristics of patent versus scientific texts strongly impact chemical NER performance. Overall, our results emphasize the common wisdom that using the same text genre (patents vs. scientific articles) and text type (abstracts vs. full texts) for training and application is a pre-requisite for achieving high-quality text mining results.</p><!><p>In this section, we describe the characteristics of the four freely available gold standard patent corpora and present the two chemical NER systems utilized in our study. We also explain two ensemble approaches, our evaluation metrics, and the text preprocessing techniques employed.</p><!><p>The details of the gold standard patent corpora containing the annotations for chemicals</p><!><p>Note that these corpora were annotated using different annotation guidelines. The corpora from the CEMP task and BioSemantics group were annotated using two specific annotation guidelines, while the chapati curators considered all entities that could be automatically mapped to a CHEBI identifier. The annotation guidelines vary in several aspects. For instance, the IUPAC name "water" should not be annotated as a chemical in the CEMP corpora but it should be in the BioS corpus [19]. Additionally, simple chemical elements are annotated in the CEMP corpora but not in the BioS corpus.</p><p>Additionally, we compare our results on the patent corpora with those achieved on two corpora consisting of scientific articles: CHEMDNER and CRAFT. The CHEMDNER corpus [25] was developed for the CHEMDNER task at the BioCreative IV challenge. The corpus consists of scientific abstracts that were annotated using the same annotation guideline used for the CEMP task. In this work, we only used the test set, containing 3000 abstracts, since the training set and the development set were used for training tmChem. The CRAFT corpus [26] consists of 97 scientific full texts, yet only 67 of these have been publicly released to date. The chemical annotations of the CRAFT corpus are limited to terms from the CHEBI database.</p><p>For illustration (see "Text-genre statistics" section), we briefly analyzed differences between patents and scientific articles in terms of the average number of words per sentence (sentence length), the average number of words per document (document length), the average number of unique/non-unique TLAs7 in a document, and the average number of figures and tables per document. To this end, we used a collection of randomly selected full patent documents from European Patent Office8, and a set of randomly selected full journal articles from PubMedCentral. All texts were from year 2015; patents were selected after classification to ensure biomedical topics. We calculated the number of tables and figures by counting the number of their tags in xml format.</p><!><p>Details on the chemical NER tools in terms of training sets, databases to which the entities are normalized, classes of chemicals addressed, and tokenization methods</p><!><p>ChemSpot employs a hybrid approach, in which the results of a CRF model trained to recognize IUPAC entities are combined with dictionary matching to find other chemical names. TmChem uses ensembles of two CRF models, called Model1 and Model2, with different setups and configurations. As the implementation of the ensembles were not freely avaliable, we performed our experiments using the individual models. We limited our study to the results from Model1, noted as tmChem, as its performance was always very close to or higher than that of the Model2 [22, 24]. Both tmChem and ChemSpot build on BANNER as CRF implementation [30], but use different feature sets, tokenization methods, and training sets. Both tools were trained on scientific abstracts, but the training corpora comprise different articles and were annotated based on different annotation guidelines [31]. Note that the annotation guideline used to annotate the training set of tmChem is very similar to the one used to annotate the two CEMP patent corpora. Both tools normalize extracted entities. TmChem maps the entities to identifiers from CHEBI and MESH, whereas ChemSpot maps them to further databases, like InChI and DrugBank.</p><!><p>A comparison of the concrete set of entities returned by tmChem and ChemSpot (see "Comparison at instance level" section), respectively, showed significant divergence. Since ensembles of NER tools often outperform individual tools [28], we also measure the performance of two ensembles produced by merging the results of tmChem and ChemSpot. One ensemble system, called Ensemble-I, accepts a mention as a chemical name when both tmChem and ChemSpot recognize it as such. The second ensemble, noted Ensemble-U, considers a span as a chemical name when it is recognized by at least one of the two systems.</p><!><p>Performance values were computed in terms of precision, recall, F-measure, and true positive (TP), false positive (FP), and false negative (FN) counts; in all cases, only exact span matches were considered. Precision measures the ratio of correctly predicted chemical entities to all predicted entities; recall is defined as the ratio of correctly predicted entities to all annotated entities within a corpus. F-measure is the harmonic mean of the precision and the recall values.</p><p>We measured performance values using the conlleval11 script run over the prediction and reference annotation files in IOB format. We also compared the different methods with respect to the execution time on patents and scientific articles.</p><!><p>The different gold standard corpora were available in different text formats which we homogenized before running the NER tools. In a first step, each document was converted to plain text format and stored in a single file. Then we transformed these files to the input format defined by each NER tool. For evaluation purposes, we tokenized each prediction and gold standard annotation files as suggested by Klinger et al. [32]12, and represented each token in IOB format. We used the Stanford parser13 [33] to split the text into sentences and to parse a sentence.</p><!><p>We first provide evaluation scores for the models trained on the abstract of scientific articles when applied to patents. Then we present an analysis of entities frequently recognized incorrectly as chemicals or non-chemicals by the two systems. Afterwards, we describe results of the two ensemble systems. Finally we present the results of cross-corpus and intra-corpus evaluations to study the impact of the use of different text genres and text types as training and test sets in patent mining.</p><!><p>Evaluation scores in terms of precision, recall and F-measure values are measured for ChemSpot and tmChem NER tools over gold standard corpora</p><!><p>We first observe that the performances of both tools are much lower on the CRAFT corpus than on the other corpora. The reason for this discrepancy seems to be that the version of the CHEBI database used for annotating the CRAFT corpus is quite different from that used for training both NER tools [25]. On the CHEMDNER corpus, both tools, despite the use of different training data, have higher performance values than on the other corpora, indicating that they perform best on scientific abstracts—the type of texts they also were trained on. This observation indicates that models trained on scientific abstracts are not quite as capable of recognizing chemical entities from patents. Drawing a conclusion regarding the difference between scientific abstracts and full texts, however, is difficult due to the quite different scope of chemical annotations in the CRAFT corpus [25].</p><p>The results, obtained from both tools, also report a ~10% higher performance on patent abstracts compared to full patents, indicating that there is more similarity between patent abstracts and scientific abstracts than between scientific abstracts and full patents.</p><p>TmChem has higher F-measure values than ChemSpot (at least 5%) on the CHEMDNER and the CEMPs corpora, both of which follow a annotation guideline similar to that of the tmChem training set. In contrast, ChemSpot was trained on a corpus with a different annotation guideline. However, the F-measure values of both tmChem and ChemSpot were very close on full patent corpora, annotated using guidelines which differ from those of the systems' training sets. This means that the improvement obtained by tmChem on patent abstracts is likely due to the similarity of the annotation guidelines and not to the superiority of the method. Thus, we cannot conclude which tool is better suited for chemical NER on patents.</p><!><p>The top 10 entities with highest FP for each chemical NER tool on the four different corpora</p><p>Common mistakes are shown in italic</p><p>The top 10 entities with highest FN for each chemical NER tool on the four different corpora</p><p>Common mistakes are shown in italic</p><!><p>Counting the number of common errors, we find around 50% overlap between the top-10 entities with highest FP counts and nearly 70% overlap between entities with highest FN values between the tools. However, the individual error frequencies are very different. For example, the number of times that the entity "alkyl" is incorrectly recognized as a chemical entity by ChemSpot is around 6 times higher than that of tmChem (in patent abstracts). There are several similar cases in the corpora containing full patents. Similarly, there are also a number of common entities with diverging FN values. By excluding common entities with highly different frequencies, the overlaps between these two tools for patent corpora are reduced to around 40 (for FP counts) and 50% (for FN counts), which indicates the two tools perform rather differently.</p><!><p>Distributions of FP counts from high to low, for unique entities covering 25% of cases, obtained by tmChem and ChemSpot over all corpora. The x-axis represents the number of unique entities. The distributions are notably different for full patents compared to patent abstracts</p><p>Distributions of FN counts from high to low, for unique entities covering 25% of cases, obtained by tmChem and ChemSpot over all corpora. The x-axis represents the number of unique entities. The distributions are very similar for full patents and patent abstracts</p><!><p>The distributions over patent abstracts showed that 25% of FP counts are produced by around 90 unique entities for tmChem, and only 30 unique ones for ChemSpot. Moreover, the shapes of the distributions were quite different. Similarly, by comparing the distributions of FP counts over full patents, we observed that the number of unique entities leading to 25% of FPs for tmChem is around 20 and for ChemSpot, it is nearly 30. The shapes of tmChem and ChemSpot distributions were very similar over full patents, but they were different from the ones obtained for ChemSpot over patent abstracts. These results confirm that the distributions of FP counts are different over full patents and patent abstracts.</p><p>On the contrary, the shapes of the distributions drawn for FN counts were very similar for both systems over all corpora, but the number of unique entities leading to 25% of FNs over full patents is around 25 while it is 75 for abstracts.</p><!><p>The FP and FN counts of simple chemical elements normalized by the FP and FN counts obtained for the entire entities by tmChem and ChemSpot over all corpora</p><!><p>First, we computed the FP and FN counts for simple chemical elements normalized with the FP and FN counts of all entities extracted by each NER tool from each corpus as shown in Fig. 4. The results obtained by the two tools demonstrate that the normalized FP values of simple elements are higher than those of FN values for full patents, while they are approximately analogous for patent abstracts. It implies that simple chemical elements are frequently recognized incorrectly as chemicals on full patents.</p><!><p>Evaluation scores with regard to precision, recall and F-measure over recognized spans obtained by ChemSpot and tmChem NER tools over gold standard corpora. The results are provided by considering simple elements represented by "+" and without them noted by "−"</p><p>Evaluation scores with regard to precision, recall, and F-measure values over recognized spans obtained by ChemSpot, tmChem, the area of their intersection and union over gold standard corpora</p><!><p>We provide precision, recall, and F-measure values calculated for tmChem, ChemSpot, and for the intersection and the union of their outputs in Fig. 6. As expected, on all corpora, the highest precision is obtained by intersecting the results of the two tools, while the highest recall is provided by unifying the results of the two systems. The results also show that the Ensemble-U provides the highest F-measure value on full patents, while tmChem has the highest F-measure scores on patent abstracts. This can probably be attributed to the use of similar annotation guidelines for training and test sets.</p><!><p>Precision, recall and F-measure values of the models trained using different corpora on the CEMPs, chapati, and BioS patent corpora</p><!><p>The F-measure values of all models evaluated on chapati are nearly identical. Additionally, the F-measure scores of the model trained using the chapati corpus are very close on other corpora, perhaps for its small number of instances. Thus, we limit our analysis to the remaining corpora.</p><p>The results show that the performance values on the BioS corpus, containing full patents, are demoted when using models trained on CEMP_T or CEMP_D, both of which contain patent abstracts and were annotated using a different annotation guideline. Likewise, the performance values on CEMP_T and CEMP_D are higher when using models trained on CEMPs corpora, and not on BioS. However, we cannot conclude that the models trained on abstracts are not suitable to identify entities from full texts and vice versa, as the corpora are annotated using different annotation guidelines.</p><p>The precision, recall and F-measure values achieved in intra-corpus evaluation, shown in Fig. 7, indicate that the precision values on full patents are at least 10% lower than on patent abstracts. Although the recall value on BioS is close to the ones obtained on patent abstracts, its F-measure value is still lower than those of patent abstracts. These results imply that identifying chemical names from full patents is more difficult compared to that of patent abstracts.</p><!><p>We have empirically shown that significant differences exist between the results of chemical NER on patents and scientific articles and even between different types of patent texts. Our study has demonstrated that identifying chemical entities from patent full texts is more complex than from patent abstracts or scientific abstracts. In the following, we assess the complexity of this task on patents, especially on patent full texts.</p><!><p>The performance values attained in cross-text-genre evaluations show that the F-measure values of the models trained on the abstracts of scientific articles decrease by around 10% when tested on patent abstracts and by nearly 18% when applied to patent full texts.</p><p>The lower F-measure scores obtained by tmChem on patent abstracts compared to that of scientific abstracts, while both have annotation guidelines very similar to those of the tmChem training set, show that there are several chemical entities in patent abstracts that cannot be recognized by the models trained using scientific articles. This finding emphasizes the difficulty of the chemical NER task on patents.</p><p>The F-measure scores of ChemSpot trained on scientific abstracts annotated using a guideline different from the ones used for the patent corpora, indicate that these models are not adequate to recognize entities from patents, and accentuate the need for more annotated patent corpora for chemical NER.</p><!><p>The execution time, in seconds, of NER tools over 10 full patent documents and 10 journal articles</p><p>The execution time values of both systems are lower on scientific articles shown in italic compared with patents</p><!><p>ChemSpot required 149 s to complete the task on scientific articles, around four times faster than the time required for patents. Similarly, tmChem needed approximately 50% more time to finish the task on patents compared with scientific articles. The main reason is the difference in their lengths (see next section).</p><p>Then we estimated the execution time of the two systems that one would have to expect on 10 million patents and 10 million full scientific articles, assuming 8 parallel threads by extrapolating the above values. The results show that tmChem would take around 3 months over patents and 2 months over journals while ChemSpot would take approximately 2 years for patents and 7 months for journals. We conclude that large parallel systems are required for patent chemical NER.</p><!><p>Statistical measurements calculated over 17,000 patent documents and 17,000 journal articles</p><p>The largest values are represented in italic for each measurement</p><!><p>The average number of words per sentence is almost the same for both patents and journals. However, the average number of words of a patent document is approximately five times higher than that of a journal article, which is in agreement with the findings obtained by Aras et al. [34]. We also observed that the number of TLAs is four times higher in patents than in journal articles, on average. This huge number of TLAs per document makes the NER task on patents harder because of the inherent ambiguity of acronyms. Moreover, the number of tables and figures in patents are more than those in scientific articles. This also makes the extraction of entities from patent documents more difficult than from journal articles [35].</p><!><p>The intra-corpus evaluation scores obtained by retraining tmChem (see "Cross-text-genre to cross-corpus evaluation" section) show that the precision (F-measure) values on abstracts are at least 12% (6%) higher than those on full texts. Since both training and test sets contain documents of the same type (abstracts vs. full texts) annotated with the same annotation guideline, we can conclude that the NER task over patent full texts is more complex than that on patent abstracts.</p><p>Moreover, the comparisons at instance level indicate that the patterns of errors observed for FP counts are generally different for different types of patent texts, while they are nearly identical for FN counts. We also infer that filtering just a small number of cases correctly as non-chemicals could reduce the FP or FN values significantly. However, achieving such a filtering is difficult, as shown in the following section.</p><!><p>The full confusion matrix for the ambiguous entity "alkyl" calculated for ChemSpot and tmChem over CEMP_T and CEMP_D corpora</p><p>The full confusion matrix for the ambiguous entity "H" calculated for ChemSpot and tmChem over chapati and BioS corpora containing complete patent documents</p><!><p>By comparing the results obtained at the instance level shown in Tables 3 and 4, we noticed that some of the errors are produced due to the differences in the annotation guidelines of NER training sets and patent test sets (see "Patent corpora" section). For example, in these tables, the word "water" is not correctly recognized as a chemical entity by tmChem from BioS corpus or is incorrectly considered chemical by ChemSpot from CEMPs corpora due to the differences in the annotation guidelines.</p><p>Moreover, there are many simple chemical elements in the list of entities with high FP counts obtained by tmChem for BioS in Table 3, because simple elements are annotated as chemicals in the training set used for tmChem while they are not labeled as chemicals in BioS corpus. The impact of different rules for annotating simple chemical elements is also observed from the improvement obtained by tmChem in the precision of the BioS corpus after excluding simple chemical elements from both reference and prediction files in "Impact of simple chemical elements" section.</p><!><p>In this paper, we performed a cross-text-genre evaluation by measuring the tagging quality of the two NER baselines trained on the abstract of scientific articles when evaluated on patent corpora. We noticed that the results are significantly worse on patent corpora compared to scientific abstracts. Although intra-corpus evaluation has shown that training on patent corpora will improve the performance results, performance values are still below the ones achieved for scientific abstracts. Our findings clearly confirm that there are major differences in the NER task between patent and scientific abstracts, and emphasize the complexity of this task on patents.</p><p>Moreover, we compared patent abstracts and full texts and addressed the differences between them using various evaluation metrics such as intra-corpus evaluations, and comparison of errors observed at the instance level. We showed that the results on patent abstracts are not extendable to patent full texts which are more important in practice. Therefore, the preparation of more annotated patent full texts is a major requirement for further research in this area.</p><!><p>Abbreviated as "chemical" from now on.</p><p>http://www.ncbi.nlm.nih.gov/pmc.</p><p>See http://www.biocreative.org.</p><p>See https://www.epo.org/.</p><p>See https://www.ebi.ac.uk/chebi/.</p><p>See http://www.biosemantics.org/.</p><p>TLA is defined as any three-letter word with letters all in uppercase form.</p><p>See https://www.epo.org/.</p><p>Although the performance of ChER was very close to that of tmChem in the BioCreative IV challenge, we were not able to include it in our study as it is not freely available.</p><p>Note that ChemSpot was not evaluated in the BioCreative IV challenge.</p><p>See http://www.cnts.ua.ac.be/conll2000/chunking.</p><p>In this format, every non-letter and non-digit character, and all number-letter changes are split.</p><p>Version (stanford-parser 3.5) is available in http://nlp.stanford.edu/software/lex-parser.shtml.</p><p>See http://www.chemicalelements.com.</p>

PubMed Open Access

Integration of in silico methods and computational systems biology to explore endocrine-disrupting chemical binding with nuclear hormone receptors

Thousands of potential endocrine-disrupting chemicals present difficult regulatory challenges. Endocrine-disrupting chemicals can interfere with several nuclear hormone receptors associated with a variety of adverse health effects. The U.S. Environmental Protection Agency (U.S. EPA) has released its reviews of Tier 1 screening assay results for a set of pesticides in the Endocrine Disruptor Screening Program (EDSP), and recently, the Collaborative Estrogen Receptor Activity Prediction Project (CERAPP) data. In this study, the predictive ability of QSAR and docking approaches is evaluated using these data sets. This study also presents a computational systems biology approach using carbaryl (1-naphthyl methylcarbamate) as a case study. For estrogen receptor and androgen receptor binding predictions, two commercial and two open source QSAR tools were used, as was the publicly available docking tool Endocrine Disruptome. For estrogen receptor binding predictions, the ADMET Predictor, VEGA, and OCHEM models (specificity: 0.88, 0.88, and 0.86, and accuracy: 0.81, 0.84, and 0.88, respectively) were each more reliable than the MetaDrug\xe2\x84\xa2 model (specificity 0.81 and accuracy 0.77). For androgen receptor binding predictions, the Endocrine Disruptome and ADMET Predictor models (specificity: 0.94 and 0.8, and accuracy: 0.78 and 0.71, respectively) were more reliable than the MetaDrug\xe2\x84\xa2 model (specificity 0.33 and accuracy 0.4). A consensus approach is proposed that reaches general agreement among the models (specificity 0.94 and accuracy 0.89). This study integrates QSAR, docking, and systems biology approaches as a virtual screening tool for use in risk assessment. As such, this systems biology pathways and network analysis approach provides a means to more critically assess the potential effects of endocrine-disrupting chemicals.

integration_of_in_silico_methods_and_computational_systems_biology_to_explore_endocrine-disrupting_c

Introduction<!>Dataset<!>QSAR modeling<!>ADMET Predictor\xe2\x84\xa2<!>MetaDrug\xe2\x84\xa2<!>The Online Chemical Modeling Environment (OCHEM)<!>VEGA-QSAR<!>Endocrine Disruptome<!>QSAR modeling evaluations<!><!>Chemoinformatics and ontology enrichment analysis: carbaryl case study<!>Results and discussion<!>Androgen receptor and U.S. EPA tier 1 data set<!>Estrogen receptor and CERAPP data set<!>QSAR consensus approaches<!>Chemoinformatics and ontology enrichment analysis: carbaryl case study<!>Conclusions

<p>Endocrine-disrupting chemicals are exogenous compounds that alter the normal function of the endocrine system, potentially causing adverse health effects in humans and wildlife (McKinlay et al., 2008; Mnif et al., 2011). The endocrine-disrupting potential of many pesticides is a major health concern (Klotz et al., 1997; Bjorling-Poulsen et al., 2008; Mnif et al., 2011; Andersen et al., 2015; Ewence et al., 2015). Pesticide residues are found in the environment and in many food items as a result of the widespread use of pesticides. Humans are exposed to pesticides in many ways throughout life, knowingly and unknowingly. They might use pesticides at work, consume pesticide residue on foods they eat, or contact pesticides in air, soil, or water. Several studies have found pesticide residues and their metabolites in human tissues, demonstrating a global distribution (Wohlfahrt-Veje et al., 2011; Roca et al., 2014; Andersen et al., 2015).</p><p>Most endocrine-disrupting pesticides mimic estrogen function by acting as a ligand for the estrogen receptor, converting other steroids to active estrogen or increasing the expression of estrogen responsive genes, as shown by some organochlorines, organophosphates, carbamates, and pyrethroids. Antiandrogenic effects also have been reported for organochlorine and carbamate insecticides, and for triazines, a group of herbicides, through inhibition of binding natural ligand to androgen-binding receptors. Endocrine disruption also occurs through competitive inhibition of thyroid hormone receptors by organophosphates and inhibition of progesterone action by pyrethroids (McKinlay et al., 2008). The results of various transactivation assays using mammalian and yeast cells have indicated agonistic or antagonistic activity of pesticides toward aryl hydrocarbon receptors and some members of the nuclear receptor superfamily, including retinoic acid receptors, pregnane X receptors, and peroxisome proliferator-activated receptors.</p><p>Structure activity relationship (SAR) and quantitative structure-activity relationship (QSAR) approaches have been used by academia, pharmaceutical, agrochemical, food, and other industries, and by various advisory and regulatory government agencies as decision support tools to address data gaps and provide provisional data in the toxicity database of a chemical. These tools allow for the assessment of chemicals for which no data are available. SAR/QSAR approaches are increasingly being used as core prediction systems in toxicology and have proven to be powerful tools to increase our understanding of the potential harmful effects of chemicals on the environment and human health.</p><p>As the number and variety of potentially hazardous chemicals continues to increase, regulatory authorities have approached the challenge by applying these approaches as decision support to chemical risk-based approaches. Furthermore, the ever-increasing economic, social, and political call to reduce animal testing in toxicity evaluation has led to an expansion of the use of these tools. Many different SAR/QSAR models and platforms for predicting a wide range of toxicological endpoints have been documented. Many of these are commercially available (e.g., MultiCASE, DEREK, TOPKAT®, Leadscope), others are open source (e.g., OncoLogic™, ToxCast™, ECOSAR, OASIS, VEGA, CAESAR, TEST, OECD tool box), and some are proprietary in-house systems (e.g., U.S. Food and Drug Administration QSAR models). Although several SAR/QSAR computer-based models have been developed, only some have been routinely used in assessment of chemical toxicity. Their validation, reliability, and use protocols are crucial to further enhance their regulatory acceptance.</p><p>Government programs use structure–activity modeling to help protect human populations from exposure to environmental contaminants (Tong et al., 2004; Netzeva et al., 2005; Worth et al., 2007; Arvidson et al., 2010; Ruiz et al., 2011). Computational toxicology methods can identify chemicals that might pose an endocrin–disruption hazard and then prioritize those chemicals for additional in vitro or in vivo testing (Tong et al., 2004; Ruiz et al., 2013; Vuorinen et al., 2013; Zhang et al., 2013). Because of the diversity and complexity of endocrine disruption mechanisms and the limited data available for in silico modeling, most studies have focused on endocrine-disrupting chemicals that act via estrogen or androgen receptors (Tong et al., 2004; Li and Gramatica, 2010b, a; Ruiz et al., 2013; Zhang et al., 2013). These modeling approaches include QSAR modeling (Li and Gramatica, 2010b; Ruiz et al., 2013, 2016; Devillers et al., 2015; Ruiz et al., 2016a), molecular dynamics simulations (Celik et al., 2008; Tsakovska et al., 2011; Chen et al., 2014), and docking (Celik et al., 2008; Vedani et al., 2012; Kolsek et al., 2014; Plosnik et al., 2015). Regulators and industry stakeholders can select from several commercial and freely available computational tools to evaluate the potential toxicity of compounds. Comparative analysis of in silico methods and their interpretation on relevant data sets is of high interest to these stakeholders.</p><p>A lack of test data for many of the thousands (>84,000) of chemicals in the Toxic Substance Control Act (TSCA) inventory presents a challenge for regulatory decision making. One of the high-interest toxicity endpoints is endocrine-disrupting activity, which includes interference with several nuclear hormone receptors, such as estrogen, androgen, thyroid, glucocorticoid, and mineralocorticoid receptors that have been associated with cancers, diabetes, obesity, reproductive, or immune problems, and metabolic disorders. Recently, the U.S. Environmental Protection Agency (U.S. EPA) released its reviews of the ESDP Tier 1 screening assay results for a set of pesticides (U.S. EPA, 2015). Even more recently, the Collaborative Estrogen Receptor Activity Prediction Project (CERAPP) provided a training set data of 1677 chemicals derived from ToxCast and Tox21 programs (Mansouri et al., 2016).</p><p>The study provides an analysis into the scientific understanding and value of the practical application of these methods for classification and regulatory decision making. This work also aims to stimulate formation of new testable hypotheses to address some of the data gaps previously identified by in the U.S. EPA Endocrine Disruptor Screening Program (EDSP) Tier 1 assessment (U.S. EPA, 2015). Collaborations that focus on alternative methods and in silico studies with endocrine disruptors continue to evolve (e.g., CERAPP) (Mansouri et al., 2016). The data set from CERAPP is also used as a comparative reference with the defined regulatory Tier 1 data set. For estrogen and androgen receptor binding predictions, comparisons among two commercially available QSAR tools (ADMET Predictor™ and MetaDrug™), two publicly available QSAR tools (Virtual models for property Evaluation of chemicals within a Global Architecture [VEGA] and the Online Chemical Modeling Environment [OCHEM]), and an open source docking tool (Endocrine Disruptome) were performed (Roncaglioni et al., 2008; Sushko et al., 2011; Kolsek et al., 2014).</p><p>As part of the study, a computational systems biology approach was applied to assess the underlying mechanisms related to endocrine disruption; a case study using a well-study carbamate pesticide, carbaryl, which has been associated with endocrine-disrupting toxicity is presented. This carbaryl case study includes an in silico generation of metabolites and a biological pathway enrichment analysis, and the results of this analysis were compared with some known findings for carbaryl. The purpose of this study is not the development of a new in silico model. Instead, this work evaluates and integrates in silico methods and computational systems biology approaches to assess their usefulness in screening endocrine-disrupting chemicals.</p><!><p>U.S. EPA recently released the Tier 1 screening results for 52 chemicals included in its Endocrine Disruptor Screening Program (U.S. EPA, 2015). This screening is used to determine if chemicals have the potential to interact with the three hormonal pathways (estrogen, androgen, and thyroid) in the body's endocrine system. Eleven assays — five in vitro (cell systems) and six in vivo (live animal) — are used to determine if the chemicals interact with these three hormone pathways, and whether or not they are potential developmental and carcinogenic toxicants. The results of the 52 chemicals studied in the EDSP Tier 1 were used as a reference for androgen binding (Table 1S). The recently available training data set described by CERAPP (Mansouri et al., 2016), which consist of 1677 chemicals, was used for the estrogenicity endpoint.</p><!><p>For this study, the ability of five QSAR modeling packages to estimate the potential of these pesticide chemicals to interact with the estrogen and androgen receptor hormonal pathways was evaluated using ADMET Predictor™ 5.0 (Simulations Plus Inc.), MetaDrug™ with System Toxicology Module, VEGA, OCHEM, and the Endocrine Disruptome docking program.</p><p>In silico predictions for determining potential human metabolite profiles were performed using MetaDrug™ and ADMET Predictor. Chemical structures for all compounds in the dataset were drawn in MOL file format using ChemBioDraw Ultra 12.0 (CambridgeSoft), from which they could be imported into the QSAR packages. Qualitative assessment of the in silico predictions of potential metabolite structures was established by comparison with human in vivo metabolic profiles characterized in available biomonitoring studies from the literature.</p><!><p>ADMET Predictor™ is a state-of-the-art computer program designed to estimate certain absorption, distribution, metabolism, elimination, and toxicity properties of a chemical from its 2- and 3-dimensional molecular structures. The program uses molecular descriptor values as inputs to independent mathematical models (generally, nonlinear machine learning techniques) to generate estimates for each of the ADMET properties. Numerous studies have been gathered in U.S. EPA's Distributed Structure-Searchable Toxicity (DSSTox) database, which was used to train ADMET Predictor's models of endocrine receptor binding.</p><p>ADMET Predictor™ estimates qualitative and quantitative models of several measures of toxicity. It includes a qualitative assessment of estrogen receptor toxicity in rats (TOX_ER_filter), together with a quantitative measure of estrogen receptor toxicity in rats (TOX_ER [IC50 (estrogen)]) that is applied only for compounds classified as "toxic" by the previous quantitative assessment model.</p><p>Two neural network ensemble models are used to assess a compound's likelihood of binding to the estrogen receptor. The first is a straightforward prediction of whether the molecule will have a detectable affinity for the receptor. A result of "nontoxic" indicates that the compound is unlikely to cause endocrine toxicity by binding to the receptor. A result of "toxic" indicates only that the compound is likely to bind detectably to the receptor. The degree of toxicity is not specified.</p><p>The second model predicts the degree of binding for those compounds that are identified by the filter as "toxic." This model displays the relative binding affinity (%RBA) of a molecule determined by a competitive binding assay. The %RBA is a dimensionless number expressed as the percent ratio: 100% * (50% of the maximum inhibitory concentration (IC50) for 17 β-estradiol/IC50 for the chemical in question). Higher values indicate greater binding affinity and likelihood for endocrine-related toxicity.</p><p>Likewise, two neural network ensemble models were built for the androgen receptor: qualitative TOX_AR_Filter (toxic/nontoxic) and quantitative TOX_AR. Unlike the corresponding estrogen receptor models referenced by a natural estrogen, the reference ligand here was 17R-methyl-[3H]-methyltrienolone (R1881), a synthetic substrate that binds more strongly than testosterone. The binary classification of TOX_AR_Filter was based on the LC25 value of R1881 displacement assay. Compounds with LC25 ≤ 10 μmol/L were labeled "toxic" (likely to bind). Those >10 μmol/L were considered "nontoxic."</p><p>The TOX_AR model predicts the degree of binding for those compounds identified by the filter as "toxic." In this case, IC50 measurements were used from R1881 displacement assays. The % RBA is defined in a way analogous to TOX_ER.</p><p>ADMET Predictor tool automatically determines whether a given compound is within the scope (applicability domain) of each of these models. The phrase "within the scope" is defined in terms of molecular descriptor space, not in terms of the relative value of an ADMET property. Let X denote an ADMET model with known training set and let {xi}i=1N be a descriptor set on which model X is based. For each descriptor, xi(i = 1, …, N), its minimum and maximum values, ximinandximax, are determined over the training set of X. A new compound C is said to be within the scope of X if the value of each relevant descriptor ci(i = 1, …, N) calculated for C is contained within the corresponding interval 〈ximin, ximax〉 with tolerance equal to 10% of the interval length. Such a compound has its X value typed in black bold font. Otherwise, the compound in question is outside the scope of X and its X value is marked by magenta font.</p><!><p>MetaDrug™ includes more than 70 QSAR models calculated with ChemTree™ (Golden Helix, Bozeman, MT, USA) using a recursive partitioning algorithm. Overall, these models cover the full spectrum of absorption, metabolism, distribution, excretion, and toxicology (ADMET) properties for a given chemical compound. QSAR models are built based on two types of training sets, those taken directly from the literature and those manually annotated from the MetaCore™/MetaDrug™ database. Every molecule in the training set is carefully analyzed for duplicates, stereoisomers, salts, etc.</p><p>A QSAR model built with ChemTree™ is passed through internal validation (Myshkin et al., 2012). The coefficient of determination (R2) and root mean squared error are calculated for real vs. predicted values of the training set. Activity predicted from the corresponding QSAR model for the uploaded compound constitutes its QSAR value. The QSAR value has to comply with two QSAR thresholds. One corresponds to the negative logarithm of activity value of the most active compound of the training set defining the predictability limit of the model. The other is the negative common logarithm of 50 μM (−1.7), because less potent compounds are considered to be non-active. If the QSAR value falls within two thresholds it is colored green and the compound is considered as active. Red indicates QSAR values outside the thresholds and non-active compounds. Blue relates to physicochemical properties.</p><p>MetaDrug™ estimates qualitative and quantitative models of several measures of toxicity. It includes a qualitative assessment of estrogen and androgen receptor toxicity in mammals. The MetaDrug™ tool for androgen binding (ADR-lig, prob) predicts the potential to bind to androgen receptor, range from 0 to 1. Cutoff is 0.5. Values higher than 0.5 indicate potential androgen receptor ligands. The compounds used in this QSAR model are from Fang et al. (2003).</p><p>The MetaDrug™ tool for estrogen binding (ESR-lig, prob) predicts the potential to bind to the estrogen receptor at 100 μM or less, range from 0 to 1. Cutoff is 0.5. Values higher than 0.5 indicate potential estrogen receptor ligands. The 219 compounds used in this QSAR model are based on U.S. EPA DSSTox KIERBL data. To establish the reliability of the QSAR to evaluate the chemical compound, the structure similarity is calculated for the most similar compound in the training set. This is done using the Tanimoto Prioritization score, which is calculated using the Accord Chemistry Cartridge™ (Accelrys, San Diego, CA, USA). Tanimoto prioritization scores are reported on a scale from 0 to 1.</p><!><p>OCHEM is a web-based platform that aims to automate and simplify the typical steps required for QSAR modeling. The platform consists of two major subsystems: the database of experimental measurements and the modeling framework (Sushko et al., 2011). OCHEM is widely used to perform QSPR/QSAR studies online and share those with other web users. The ultimate goal of OCHEM is to collect all possible chemoinformatics tools within one simple, reliable, and user-friendly resource. OCHEM is free to use and available online at http://www.ochem.eu.</p><p>For this work, consensus estrogen-binding class QSAR modeling was applied to the data set. This model was developed and applied under CERAPP. The descriptors were calculated using several program packages, which cover different representations of chemical structures from simple fingerprints and a count of chemical groups. Detailed information about chemicals in the data sets, descriptor calculations, model parameters, and statistics are provided on the OCHEM platform (Sushko et al., 2011).</p><!><p>VEGA is a freely available platform of QSAR models for regulatory purposes (Roncaglioni et al., 2008). It assesses the chemical space where the model predictions are reliable through the Applicability Domain Index (ADI), with scoring from 0 to 1 (VEGA v. 1.1.1. (http://www.vega-qsar.eu/)). This work uses the relative binding affinity (RBA) QSAR model for the qualitative prediction of RBA for screening estrogen mediated endocrine-disrupting activity. This is a classification model based on a classification and regression tree (CART) algorithm and data set of 806 chemicals from the Japanese Ministry of Economy, Trade, and Industry database, as reported by Roncaglioni et al. (2008).</p><!><p>Endocrine Disruptome is a freely available program package for estimating binding activities toward the following14 human nuclear receptors: androgen receptor (AR), estrogen receptors (ER α and ER β), glucocorticoid receptor (GR), liver X receptors (LXR α and LXR β), mineralocorticoid receptor (MR), peroxisome proliferator activated-receptors (PPAR α, PPAR β, and PPAR γ), progesterone receptor (PR), retinoid X receptor (RXR α), and thyroid receptors (TR α and TR β). (Kolsek et al., 2014) provide details of the docking protocol and development of the models, and data sets of compounds. Endocrine Disruptome's docking simulation, which is an assessment of a ligand's position within a target receptor, is performed by the Docking interface for Target Systems (DoTS) platform using the AutoDock Vina software package (Vedani et al., 2012; Kolsek et al., 2014).</p><p>The 18 proteins (14 agonists and 4 antagonists) in Endocrine Disruptome were carefully selected from a pool of 103 proteins extracted from Protein Data Bank, with consideration being given to the quality of data and docking results on reference compounds. The final results of the predictions are the binding affinities given numerically as binding free energies (in kcal mol−1) and, alternatively, presented as four colored classes at the tool output (Kolsek et al., 2014). For this classification, the program uses three threshold values, defined by calculations of sensitivity (SE) and the validation experiments. The threshold values are SE < 0.25 (red) for high binding affinity, 0.25 < SE < 0.5 (orange) and 0.5 < SE < 0.75 (yellow) for medium binding affinity (orange), and SE > 0.75 (green) for low binding affinity. The absolute values for the thresholds are also reported by (Plosnik et al., 2015). Thus, in this study, for classification of endocrine-disrupting chemicals as potential androgen or estrogen disruptors, if the receptor binding energy for agonists or antagonists is predicted as high or medium high, the chemical is classified as a potential disrupter (active).</p><!><p>A common way to evaluate the performance of classification models is to use a confusion matrix (Fjodorova et al., 2010). Such a matrix was used to show the number of correct and incorrect predictions made by each classification model compared with the actual (experimental value) in the data. The matrix is N × N, where N is the number of actual values (positive and negative). Performance of such models is commonly evaluated using the data in the matrix. The following illustration shows a 2 × 2 confusion matrix for two classes (positive and negative).</p><p>Several statistical parameters were used:</p><!><p>Sensitivity or recall: the proportion of actual positive cases correctly identified</p><p>Specificity: the proportion of actual negative cases correctly identified</p><p>Accuracy: the proportion of the total number of predictions that were correct</p><p>Balanced accuracy: (sensitivity + specificity)/2</p><p>Positive predictive value or precision (PPV): the proportion of positive cases correctly identified</p><p>Negative predictive value (NPV): the proportion of negative cases correctly identified</p><p>Matthews correlation coefficient (MCC): a coefficient of +1 represents a perfect prediction, 0 an average random prediction, and −1 an inverse prediction</p><!><p>The chemoinformatics tools in MetaDrug™ identify structurally exact or similar compounds annotated in the MetaDrug™ database and report information on proteins with which these exact or similar compounds interact. This allows users to predict biological effects using the workflow scheme shown in Fig.1. As a case study, a chemical similarity search was conducted for carbaryl, a carbamate insecticide used on a variety of crops, in a database of some 700,000 manually annotated compounds linked via physical compound-protein interactions with some 4500 proteins known as targets for at least one small molecule xenobiotic. Accord Chemistry Cartridge™ (Accelrys, San Diego, CA, USA) fragment-based fingerprints were applied to perform a similarity search. A list of compounds was generated using the chosen similarity of exact match or 100% similarity score (Tanimoto coefficient = 1), which will search the information related to this specific chemical to validate the system biology analysis. An exact match analysis was used to mine this database for curated information about carbaryl. In this way, only targets for carbaryl reported in the literature and those predicted from QSAR models would be used for further pathway analysis. From this list, protein targets were retrieved using the collection of protein–target interactions annotated and stored in the MetaDrug™ database.</p><p>A list of potential protein targets can be used for enrichment analysis in Canonical Pathway Maps, one of the proprietary Thomson Reuters ontologies, to identify and rank cellular pathways and processes most influenced by the uploaded chemical. Every map consists of molecular entities (genes, proteins, microRNAs, or compounds) participating in a pathway and linked by well-established interactions. The significance of enrichment was defined by p-values calculated using a hypergeometric distribution (applied Fisher exact test). As a result, the chemical was associated with a quantitatively ranked list of maps summarizing its possibly toxic effects at a systems-biology level.</p><p>In the next step, the protein target list for carbaryl was computationally expanded by the nearest neighbors in the process of network building (i.e., by first-step protein–protein interactions) and intersected to identify unique targets for carbaryl compounds. The network is based on valid protein–protein and compound–protein interactions identified through manual curation using full text articles within MetaDrug™. Individual chemicals and proteins are mapped to functional ontologies such as gene-disease associations, biological processes, and mechanisms of toxicity, and later used to reconstruct biological networks linking the component nodes into biologically meaningful clusters.</p><!><p>The performance and predictive ability of in silico methods, such as QSAR and docking, were investigated and compared with the U.S. EPA EDSP Tier 1 androgen assay results and with the CERAPP training set estrogen results.</p><p>The U.S. EPA Tier 1 androgen screening assay assessed the potential binding with the androgen receptor for 52 chemicals (Table 1S). Of these, 33 chemicals displayed no evidence of binding, 10 chemicals showed binding, and the remaining 9 chemicals showed an equivocal result. The CERAPP collected concentration-response results from 18 in vitro high-throughput screening (HTS) assays that explore several sites in the mammalian estrogen receptor pathway. (Judson et al., 2015) reported a mathematical model that integrates the in vitro data and calculates an area under the curve score, and proposed an activity consensus score across the estrogen receptor assays. The CERAPP training set data contain 1677 chemicals, of which 1440 were classified as inactive and 237 as active.</p><p>High priority issues in risk assessment include prioritizing chemicals for review, filling data gaps, and reducing the use of animals for experimental testing. HTS aids the investigation of thousands of chemical and provides a means to rapidly predict potential chemical toxicities or identify key receptors that regulate specific toxicity pathways (Judson et al., 2015).</p><p>In silico methods such as QSAR and docking approaches could support the identification of key characteristics in chemical structures responsible for a toxicity endpoint. These can be used to provide predictions about the possible activity of a chemical in virtual screening settings for regulatory purposes. The quality of in silico model development and assessment of commercial or open source models, based on chemical data sets from HTS experiments, varies depending on the efficiency of the assays, experimental design, data analysis, validation, access to the data, and quantity. However, the accuracy is usually high enough to support prioritizing chemicals identified as worth being subjected to experimental testing.</p><p>When using the U.S. EPA EDSP Tier 1 and CERAPP data set, differences were identified between the reported activities for an endpoint and a need for a balanced data set. CERAPP and Judson et al. (2015) reported the development of a classification method that integrates several HTS assays for estrogenic activity. This proposed classification for estrogen-binding activity was used to develop and evaluate QSAR models under CERAPP, led by U.S. EPA (Judson et al., 2015; Mansouri et al., 2016). Use of these data, which covers several different levels of the estrogen signaling pathway, is not equivalent to an individual-specific assay result.</p><!><p>For QSAR androgen receptor binding predictions, the predictive ability of the ADMET Predictor™ models were more reliable than those obtained with MetaDrug™. The Endocrine Disruptome docking results had similar reliability when compared with the ADMET Predictor QSAR model. The Endocrine Disruptome androgen receptor model was able to correctly predict the binding activity for more than 70% of the chemicals, with an overall prediction accuracy of 70% for the androgen receptor binding activity. This model showed good specificity (79%) and a low sensitivity (40%) (Table 1). Two molecules were outside of the applicability domain of the model: abamectin was predicted toxic and fenbutatin oxide was predicted negative.</p><p>For the androgen receptor MetaDrug™ model, overall prediction accuracy for binding activity was 35%, with low specificity (33%) and low sensitivity (40%) (Table 1). The applicability domain of the MetaDrug™ QSAR model for this particular data set of chemicals was poor; only 17 of the 52 chemicals showed a favorable Tanimoto score. Thus, it showed poor accuracy, sensitivity, and specificity, which could be because of the dissimilarity between the training (n = 162) and test set data (n = 32) used to develop and evaluate the QSAR model with a data set of pesticides. The model predicted the inactive chemicals better than the active chemicals. The ADMET Predictor docking model was able to correctly predict more than 70% of the chemicals, with an overall prediction accuracy of 78% for androgen receptor binding activity. The ADMET Predictor docking model also showed good specificity (94%), but low sensitivity (30%) (Table 1). The Endocrine Disruptome tool could not perform docking predictions for abamectin and fenbutatin oxide chemicals. The ADMET predictor QSAR model and the Endocrine Disruptome docking tools had a high chemical applicability domain for this data set; most chemicals were within the optimum prediction space, with high reliability of predictions. Both models show very good specificity, accuracy, and negative predictive value, which might be because these models were developed and evaluated on data sets that mainly consist of inactive chemicals.</p><!><p>Table 2 shows the results of the four QSAR models predictions (ADMET predictor, MetaDrug™, VEGA, and OCHEM) for the CERAPP estrogen-binding data set. The dataset collects 1677 chemicals, of which 1440 are classified as inactives, and 237 as active. Because these data contain an unbalanced ratio of actives (positive) and inactive (negative) chemicals, balanced accuracy and MCC (which makes use of all the four calculated measures) are best taken into consideration. For balanced accuracy, the higher the value the better (the values range from 0 to 1). MCC ranges from −1 to 1, where −1 indicates perfect negative correlation, 0 predicts random distribution, and 1 predicts perfect correlation. Sensitivity, also called true positive rate, and specificity (true negative rate) show the ratio of the active and inactive chemicals correctly identified by the QSAR models. Positive predictive value and negative predictive value are conditional probabilities. MetaDrug™ and ADMET predictor estrogen receptor QSAR models showed low MCC, compared with VEGA and OCHEM estrogen receptor QSAR models. The balanced accuracy values are higher for VEGA and OCHEM.</p><p>The best performance was observed with the OCHEM model, which was expected, because the model was developed using the CERAPP data set. The other models have a good balanced accuracy, but low MCC value compared with OCHEM. This could be because of the size and the distribution of the imbalanced data set used to develop the QSAR models. The models have good random distribution but less than perfect predictions. In general, all these models have high negative predictive value, most likely because they have a low number of active chemicals in their training sets, and positive predictive value is also poor for most of these models. These models provide relatively reliable predictions for inactive chemicals, but many incorrect active chemical predictions.</p><p>In terms of sensitivity (Sn), the ADMET predictor shows the lowest value (0.43) and also the highest number of false negatives (FN) (134), compared with the other models. Specificity (Sp) was very good for all models, with values ranging from 0.81 to 0.88. ADMET Predictor and OCHEM had higher chemical applicability domains for this data set compared with VEGA. The Tanimoto prioritization score calculated by the MetaDrug™ model for some chemical predictions was low, which brings into question the applicability of the QSAR results for those compounds.</p><!><p>The consensus approach combines the predictions of several QSAR models to maximize their strengths, minimize their weaknesses, and increase their global prediction accuracy. Estrogen receptor and androgen receptor QSAR and docking models were selected. These are based on different molecular descriptors and statistical and modeling techniques. They were merged in two types of consensus approaches. In the first (consensus I), all the model predictions agreed on the predicted class. In the second (consensus II), based on the majority agreement approach, the compound was classified according to the most frequently predicted class.</p><p>For androgen binding model predictions, a 2-model or 3-model consensus using both approaches did not increase prediction reliability, especially for the active compounds (Sn ranging from 0.3 to 0.4). When using a 2-model consensus, a combination of the ADMET Predictor or MetaDrug™ model with the docking model, results are strongly influenced by the docking predictions, showing good accuracy and specificity. However, the consensus model using all three models shows higher specificity, accuracy, and MCC (Table 3).</p><p>For estrogen-binding model predictions, the consensus II approach was used to obtain a 3- and 4-models consensus. A 4-model consensus, (164 chemicals have a tied call [2 and 2]), provided an increase in predictions reliability and a substantial decrease in false positives, with a resultant increase in specificity and accuracy. When using a 3-model consensus, the OCHEM model was excluded because the training set data used to develop the model were the CERAPP data, so the OCHEM model would have strongly influenced the predictions. The 4-model consensus performed better when compared with the 3-model consensus, as shown in Table 4.</p><p>The strengths and weaknesses of each model were considered, knowing that none of them was a perfect model. If the strength and advantage of a model is clearly known, the output of such a model can be used with greater certainty. The applicability domain of the estrogen or androgen binding QSAR models for the pesticide data set could be poor or strong (i.e., the chemical could be outside or inside the optimum prediction space). Estrogen and androgen receptor ADMET predictor models had higher chemical applicability domain for this data set when compared with the MetaDrug™ models. Both commercially available QSAR models were internally and externally validated. However, the ADMET predictor models had superior performance compared with the MetaDrug™ models, which might be because of the size and heterogeneity of the training data sets. A comparative QSAR analysis approach when using multiple models (commercially and open sources) could compensate for the limitations of the individual models that use different descriptors and statistical methods to model different aspects of the toxicological effects. Thus, the use of a consensus model approach could be more beneficial than using individual models. Such a consensus model approach could be used to confirm model predictions for the same endpoint or as a weight of evidence when supporting available data with related endpoints describing various modes of action.</p><p>When using multiple models with varying modeling techniques (molecular descriptors, statistical methods, and validation), judging the output from each can be particularly challenging when their performance is comparable but slightly different (Li and Gramatica, 2010a; Novic and Vracko, 2010; Ruiz et al., 2012, 2013; Plosnik et al., 2015). In such cases, conventional wisdom and expert knowledge could be used to examine how the QSAR models follow the development, validation, and application principles described by OECD.</p><!><p>In silico predictions for the estrogen-binding potential of carbaryl were negative across the QSAR models, and for androgen receptor binding, positive predictions were in agreement with the experimental data (U.S. EPA, 2010, 2011, 2015; Mansouri et al., 2016).</p><p>In this case study, Metadrug™ was used to predict the most likely metabolites of carbaryl, to allow a qualitative prediction for the metabolism of carbaryl, and to compare results with available published information. In silico metabolism approaches could be used as part of the in vitro testing as a means to improve metabolite identification strategies, particularly those mediated by CYP450 family enzymes. These approaches have been used to predict potential profile metabolites from a chemical structure, and to highlight the involvement of enzymatic reactions with the cytochrome P450 superfamily of enzymes.</p><p>Prediction of first pass and sequential metabolites, both phase I and phase II reactions, was selected to evaluate metabolite predictions for carbaryl, using MetaDrug™. Prediction of first pass and sequential metabolites, both phase I and phase II reactions, identified 59 possible phase I, and 2 phase II reactions. These include aromatic hydroxylation, N-dealkylation, epoxidation, epoxidation and hydrolysis, N-hydroxylation, N-/O-acetyl transfer reactions, N-/O-glucuronide transfer reactions, N-/O-methyl transfer reactions, N-/O-sulfate transfer reactions.</p><p>The prioritization of metabolites in MetaDrug™ is based on a score representing the occurrence rate. Occurrence rate (OC) is the ratio of the occurrence of a particular metabolic reaction to the total number of metabolic reactions in the MetaCore™/MetaDrug™ database. This occurrence frequency is then assigned to predicted metabolites as a negative log value (logOC). The larger the score, the higher the frequency of the metabolic reaction in the database. Metabolites are prioritized into major metabolites and minor metabolites on the basis of the logOC. Ranking the metabolites by the logOC values indicated three possible aromatic hydroxylation reactions, each with logOC = −1.08, as being the most likely to occur (Table 2S), and N-dealkylation was predicted to be only slightly less prevalent (logOC = −1.21). Nevertheless, numerous other metabolites were identified, confirming the diversity and relative ranking of metabolites predicted by MetaDrug™.</p><p>Carbaryl metabolite predictions using MetaDrug™ agreed very well with available published data (Tang et al., 2002). Carbaryl can be hydrolyzed by esterases and oxidized by cytochrome P450-mediated monooxygenases (CYP) to form hydrolysis and hydroxylation products, respectively. These are subject to further conjugation, such as sulfate and glucuronic acid conjugates of 1-naphthol and 4-hydroxycarbaryl. The major hydroxylation products include 5-hydroxycarbaryl (5-hydroxy 1-naphthyl N-methylcarbamate), 4-hydroxycarbaryl (4-hydroxy 1-naphthyl N-methylcarbamate), and carbaryl methylol (1-naphthyl N-[hydroxymethyl] carbamate). Although the contributions of hydrolysis and hydroxylation toward total metabolism of carbaryl have yet to be elucidated, hydroxylation by CYP is thought to be the more important route of carbaryl metabolism (Tang et al., 2002).</p><p>Assessment of the in silico prediction of potential carbaryl metabolites from a qualitative standpoint can substantially help in analyzing the available metabolism data. Fast and reliable in silico predictions could accelerate the in vitro/in vivo characterization of carbaryl metabolites. In silico tools could be continually used to explore or develop analysis of the potential endocrine disrupting effects of chemicals, the potential interferences with the metabolism of endogenous hormones, and the prediction of metabolism of chemicals by phase I and II enzymes. One of the most frequently cited limitations of in vitro assays without metabolic capacity concerns the qualitative and quantitative deficiencies in the metabolism of test chemicals. The case study presents the use of in silico metabolism prediction as a potential tool to integrate with endocrine-disrupting chemical HTS to be used for screening chemicals before endocrine disrupting chemical testing.</p><p>Based on QSAR predictions for carbaryl from the underlying database (MetaBase™), MetaDrug™ software generates a list of known and possible targets. Of these targets, three were cytochromes P450, including CYP1A2, CYP3A4, and CYP2D6, which were identified as possible targets by QSAR models (Tang et al., 2002; Myshkin et al., 2012). The androgen receptor was also identified as a potential target in this search, again supporting the results of the QSAR prediction (U.S. EPA, 2010, 2011, 2015). Another potential target included the fatty acid amide hydrolase (FAAH), with which carbaryl was previously reported to interact (Tarzia et al., 2003). This protein degrades bioactive fatty acid amides, such as oleamide, the endogenous cannabinoid, anandamide, and myristic amide to their corresponding acids, thereby serving to end the signaling functions of these molecules. This suggested that carbaryl has the potential to strongly affect metabolism and clearance.</p><p>For this study, the biological functions of the predicted protein targets for carbaryl were investigated in several ways, beginning with enrichment of the target list across three different ontological categories. The hypergeometric distribution p-value of Gene Ontology (GO) biological processes, GeneGo canonical pathway maps, and GeneGo toxicity networks, were calculated with respect to each category. GO processes and toxicity networks both confirmed the initial suggestion of a strong effect. Metabolic process and catabolic process were the two most affected GO processes. Metabolism_CYPs and Fanconi anemia group proteins and protein folding_HDAC, nucleophosmin were the two most affected GO toxicity networks (Fig. 1S).</p><p>Enrichment in canonical pathway maps gave a slightly different picture, with the highest enrichment in pathways representing metabolism of endogenous hormones, particularly estrogenic and androgenic steroid hormones. Fig. 2 shows the map for estradiol metabolism, the pathway with the highest enrichment in the dataset. Disruption of steroid hormone homeostasis, as is suggested by the enrichment in these pathways for the predicted targets of carbaryl, is a plausible mechanism for the developmental defects associated with exposure to this chemical and its metabolites. Indeed, disruption of steroid hormone homeostasis as a result of altered metabolism after activation of nuclear receptors such as pregnane X receptor (PXR), androgen receptor, estrogen receptor, and aryl hydrocarbon receptor (AhR) has recently been hypothesized as a possible mechanism of the observed developmental and reproductive effects.</p><p>This list of potential protein targets was used to build a biological network to further study pathway interactions and potential downstream effects. The carbaryl network was further refined using the shortest path algorithm network option on MetaDrug™. A Dijkstra's shortest paths algorithm calculates the shortest directed paths between selected objects. Thus, a complete network of interactions was generated for carbaryl targets (Fig. 3). The advanced search function, a Java application tool within the MetaDrug™ software, was also used to look for combined information. The advanced search made it possible to create Boolean queries and retrieve the results. A detailed methodological description of the systems biology procedures and protocols for using the software are available at http://lsresearch.thomsonreuters.com/. The advanced search tool was used to find all transcription targets of the androgen receptor that are part of the GO process "reproductive/development structure" and its sub-folders then used this list to build a network representation (Fig. 2S).</p><p>Computational systems biology was applied to carbaryl as a case study, for which few experimental developmental effects have been reported. Carbaryl is a carbamate pesticide still in use today. Animal studies on young and adult rodents have been the basis for risk assessment. Those studies have indicated that neonate animals are more sensitive to carbamates than are adult animals. The goal of the case study was to understand carbaryl's potential toxicity pathways and biological network interactions related to endocrine disruption. The results were compared with published findings (Klotz et al.,1997; Bjorling-Poulsen et al., 2008; U.S. EPA, 2010, 2011, 2015; Mansouri et al., 2016).</p><p>In toxicology, computational systems biology facilitates the identification of important pathways and molecules from large data sets. These tasks can be extremely laborious when performed through a classical literature search. Computational systems biology offers more advantages than just providing a high-throughput literature search engine. These tools might provide the basis for establishing hypotheses on potential links between environmental chemicals and human diseases. Comprehensive databases containing information describing networks of human protein–protein interactions and protein–disease associations make this possible. Experimentally determined targets of the specific chemical of interest can be fed into these networks to obtain additional information that can be used to establish hypothetical links between the chemical and human diseases. Such information can also be applied for designing animal and cell-based laboratory experiments that can test the established hypotheses more intelligently (Ruiz et al., 2016b).</p><p>By using the in silico metabolite prediction and the QSAR tools in MetaDrug™ for this study, modes of toxic action were inferred and a major mechanism of toxicity of carbaryl was identified. QSAR models and targets of structurally similar compounds in the software database, when analyzed using the systems biology molecular interaction data and network analysis tools in the product, pointed toward profound effects on metabolic enzymes, including cytochromes P450, conjugating phase II metabolic enzymes, and xenobiotic transporters. Pathway and network analysis suggested that the androgen receptor was a key regulator of these processes and, through its activation, carbaryl has the potential to affect xenobiotic metabolism and homeostasis of endogenous steroid hormones.</p><p>The pathway and network for carbaryl presents a novel computational systems biology and in silico models of different molecular mechanisms of endocrine disruption action that point to potential disease outcomes. Future experimental evaluation of this model might lead to the development of new predictive markers of endocrine-disrupting chemical effects that could translate into new disease prevention and clinical use strategies. A major objective of the present study is to stimulate experimental laboratory research by identifying good biological pathways and chemical candidates for investigation. Specific avenues of laboratory research might include in vitro and in vivo studies that should be conducted using exposure to selected endocrine-disrupting chemicals on an individual or mixture basis using a factorial design approach. Specific receptors or pathway nodes of interest identified using these combined in silico and computational systems biology model approaches could be technically evaluated by application of genomic, proteomic, or metabolomic methods.</p><!><p>This study integrated QSAR, docking, and computational system biology tools to develop a novel approach for evaluation of the estrogen and androgen binding activity of pesticides. The predictions of two commercially available (ADMET Predictor and MetaDrug™) and two open source QSAR tools (VEGA and OCHEM) were evaluated. The study used the publicly available Endocrine Disruptome docking tool using the U.S. EPA Tier 1 screening assay results for a set of pesticides in the Endocrine Disruptor screening program, and CERAPP data set. For estrogen receptor binding predictions, the predictive ability of the ADMET Predictor, VEGA, and OCHEM models (specificity: 0.88, 0.88, and 0.86, and accuracy: 0.81, 0.84, and 0.88, respectively) were more reliable than those obtained by MetaDrug™ (specificity 0.81 and accuracy 0.77). In general, all models have high negative predictive value, most likely because they have a low number of active chemicals in their training sets; thus, the positive predictive value is also poor for most of the models. The models have shown more accurate predictions for inactive chemicals than for active chemicals. On the other hand, for androgen receptor binding predictions, the predictive ability of the Endocrine Disruptome and ADMET Predictor models (specificity: 0.94 and 0.8, and accuracy: 0.78 and 0.71, respectively) were more reliable to those obtained by MetaDrug™ (specificity 0.33 and accuracy 0.4).</p><p>It is proposed that the consensus approach for the estrogen receptor prediction reduces the limitations related to an individual in silico prediction by reaching a general agreement among a collection of models (specificity 0.94 and accuracy 0.89). This systems biology pathways and network analysis shows insightful potential effects on metabolic enzymes (CYP450), androgen receptors, and xenobiotic transporters. The results provided here show that application or use of these tools as an integrated approach could support chemical safety assessment and guide further experimental testing. It could also be extended to evaluate other environmental chemicals classes because these tools allow screening of large libraries of molecules for potential endocrine-disrupting activity. This approach could also help circumvent the resource-intensive laboratory work currently used to evaluate the rapidly increasing number of chemicals detected in humans and the environment.</p><p>Key molecular networks, toxicity pathways, potential modes of action, and potential biomarkers of diseases can potentially be identified early during the process of chemical safety assessment. In silico tools such as QSAR and docking can be used for hazard identification and prioritization of chemicals for further experimental testing (Jensen et al., 2008; Mays et al., 2012; Nendza et al., 2013; Rybacka et al., 2015). Later, when empirical computational systems biology data have been generated for a chemical, (e.g., omics data, HTS, in vitro assays) or when specific toxicogenomic, metabolic, or toxicologic pathways are known, all these data can be analyzed and integrated. Using virtual screening to generate new hypotheses will expand and improve the comprehensive systems evaluation of a chemical's effects on biological systems.</p>

PubMed Author Manuscript

Carbene-based Difluoromethylation of Bisphenols: Application to the Instantaneous Tagging of Bisphenol A in Spiked Soil for Its Detection and Identification by Electron Ionization Gas Chromatography-Mass Spectrometry

The rapid and efficient difluoromethylation of a panel of eleven bisphenols (BPs) for their enhanced detection and identification by Electron-Ionization Gas Chromatography-Mass Spectrometry (EI-GC-MS) is presented. The derivatization employs the inexpensive, environmentally benign agent diethyl (bromodifluoromethyl) phosphonate (DBDFP) as a difluorocarbene-generating species that converts the BPs into bis-difluoromethylated ethers that can be detected and identified by GC-MS means. Key attributes of the protocol include its extreme rapidity (30 seconds) at ambient temperature, high specificity for BPs amidst other alcohol-containing analytes, and its biphasic nature that allows for its convenient adaptation to the analysis of BPs in organic as well as aqueous matrices. The protocol furnishes stable, novel BP ethers armed with a total of four fluorine atoms for their subsequent analysis by EI-GC-MS. Furthermore, each derivatized bisphenol exhibits unique retention times vastly different from their native counterparts leading to their unequivocal identification. The effectiveness and robustness of the developed methodology was applied to the tagging of the most famous member of this family of compounds, bisphenol-A (BPA), when spiked (at 1 μg.g −1 concentration) in the physically and compositionally complex Nebraska EPA standard soil. The method detection limit (MDL) for the bis-difluoromethylated BPA was determined to be 0.01 μg.mL −1 . The bis-difluoromethylated BPA was conveniently detected on the organic layers from the biphasic, derivatized mixtures, highlighting the protocol's practicality and utility in the rapid, qualitative detection of this endocrine disruptor during environmental analysis.Among the most persistent chemicals in the environment and whose emerging negative reputation is starting to garner considerable supporting data are the bisphenols. Bisphenols (BPs) are a class of aromatic hydrocarbons comprising two phenolic units linked together by a carbon atom (or a heteroatom such as sulfur) that may be, in the simplest member of the family, a methylene unit (CH 2 ) or a more elaborate structural motif (e.g. methylphenyl, bis-trifluoromethyl). While the two phenolic units may be linked symmetrically in three different ways,

carbene-based_difluoromethylation_of_bisphenols:_application_to_the_instantaneous_tagging_of_bisphen

Results and Discussion<!>Methods<!>EI-GC-MS experiments.<!>Derivatization protocol of individual bisphenols.<!>Conclusions

<p>Use of the fluorinated carbon units, mostly as the difluoromethyl (-CF 2 H) and the trifluoromethyl (-CF 3 ) moieties, as isosteric alternatives for the methyl (-CH 3 ) moiety in drug discovery is well documented 24 . Their use over the methyl group stems for the fact that once nitrogens, sulfurs and oxygen atoms present in a drug candidate are modified with this functionality, the net result is the blockade of metabolizing enzymes at that position that results in the increase of a drug's bioavailability and serum half-life. Equally important and within the realm of analytical chemistry, fluorine-bearing substituents play a crucial role as valuable chemical tags for analytes whose conventional detection by GC-MS is problematic as a result of their inherent low ionization potential 25 . Introduction of fluorine atoms into analytes, mostly accomplished as the universally-employed trifluoroacetyl or pentafluorobenzyl groups, act in synergistic and additive fashion to enhance their detection by augmenting their electron-capture ability. Introduction of the trifluoroacetyl moiety is accomplished using trifluoroacetic anhydride (TFAA) while the pentafluorobenzyl tag requires the use of pentafluorobenzyl bromide in the presence of a base 26,27 . As efficient and established as these reactions are in the analytical chemistry toolbox of derivatization methods, the cross-reactivity of the reagents with virtually any nucleophile is unavoidable leading to the fluorination of most, if not all, analytes in a mixture. Even though the areas of complex data analysis, processing and deconvolution have experienced an exponential evolution in the last decade, the avoidance of unnecessary, derivatized side-products that can interfere with the analysis has remained a prime and often welcomed requirement in analytical chemistry. Now, with regards to the difluoromethyl group, clever ways have been devised by chemists to introduce them into molecules, with the majority of these applicable for only modifying acidic hydroxyl groups (i.e. phenols) [28][29][30] . Most of these methods involved heating (>100 °C) for prolonged periods of time (often >16 h), until the Zafrani group introduced the stable and environmentally benign agent diethyl (bromodifluoromethyl) phosphonate (DBDFP) that expedited the difluoromethylation of acidic phenols under much milder conditions 31 . Due to its exclusive employment in the fields of medicinal chemistry and drug development, the difluoromethyl tag has experienced extremely limited use in the fields of analytical chemistry and forensic science. Earlier work by our group focused on the use of DBDFP for the derivatization of chlorinated phenols (pK a ~ 7-9) and demonstrated its efficiency, speed, and ability for detection of particularly toxic species by Proton and Fluorine Nuclear Magnetic Resonance ( 1 H and 19 F-NMR) spectroscopy 29 . Building on this work, we turned to expanding the methodology for the derivatization of bisphenols and their subsequent analysis by GC-MS means as shown diagrammatically in Fig. 1. Thus, under the basic conditions used in the protocol, DBDFP breaks down into diethylphosphonic acid and a fleeting bromodifluoromethyl carbanion that undergoes the loss of a bromide ion to generate the highly reactive difluorocarbene species that immediately reacts with the bisphenol to furnish the bis-difluorinated product (Fig. 1).</p><p>At the outset, we sought to determine the number of equivalents needed to successfully derivatize a given bisphenol without generating too many by-products from the degradation of excess reagent. After determining that 1.5 equivalents per hydroxyl group (thus a total of 3 for all bisphenols in this study) were sufficient for the derivatization of BPA without the creation of many by-products, we performed the derivatization on ten additional, structurally diverse bisphenols (Table 1). Interestingly, we were delighted to find that using large excess of DBDFP did not deleteriously affect the course of the reaction or the subsequent data analysis by providing additional by-products or interferences. Naturally, when encountering real world samples, one makes use of excess derivatization reagents to make sure that all of the analytes have been modified (e.g. BSTFA-mediated silylation), which is the reason why we employ such elevated concentrations when dealing with the soil samples (vide infra). Two interesting points that arise from the GC-MS studies of the reaction are, 1) the derivatization results in fluorinated bisphenol products with shorter retention times relative to the parent bisphenol, thus making them particularly useful in the analysis of late eluting analytes (e.g. B-FL with 38.7 min, SI, Page S16) with the difference in retention times between the derivatized and underivatized bisphenols ranging from 3.6-5.9 min., and 2) the protocol yields products that are m/z + 100 relative to the starting bisphenols, thus providing a strong, diagnostic, molecular ion peak when analyzing the data (Table 1).</p><p>In all the derivatized bisphenols studied in this work, the molecular ion peak is highly visible and in the case of bisphenol FL it is the base peak in the mass spectrum. In all other cases studied, the base peak was found to correspond to the species generated from the cleavage of one of the C-C bonds at the bridging carbon of the bis-difluoromethylated product (Fig. 2). This mode of ionization can be anticipated as the resulting radical carbocation generated at the bridging carbon is a highly stabilized doubly benzylic species while it is an even more stable trityl species where the remaining R group attached to the bridging carbon is a phenyl group. Thus, starting with bis-difluoromethyl bisphenol G (BP-G(CF 2 H) 2 ), its relatively simple mass spectrum shows the molecular ion peak at m/z = 412 [M] + as well as the spectra's base peak at m/z = 397 [M − CH 3 ] + (Figs. 2A and SI, Page S19). Similarly, in the case of BP-E(CF 2 H) 2 , one can observe its molecular ion peak at m/z = 314 [M] + while the base peak can be clearly noted at m/z = 299 [M − CH 3 ] + (Figs. 2B and SI, Page S13) arising from the loss of the bridging carbon's methyl group. Interestingly, it is the loss of a methyl rather than a proton to yield a more stable tertiary radical carbocation species that provides the base peak in the spectra of BP-E(CF 2 H) 2 , a dominant fragmentation pattern that is also exhibited by the native BPE (SI, Page S12). Other fragments that can be accounted for but yield very low abundance are the ones at m/z = 263 [M − CF 2 ] + and m/z = 247 [M − OCF 2 ] + . For the tetramethylated bisphenol analog, compound BP-C(CF 2 H) 2 , its mass spectrum seems to also offer very few fragment ions that include the molecular ion peak at m/z = 356 [M] + and its base peak at m/z = 341 [M − CH 3 ] + arising from the loss of one of the methyl groups at the bridging carbon (Figs. www.nature.com/scientificreports www.nature.com/scientificreports/ as one of the substituents at the bridging carbon, we begin our discussion with BP-AP(CF 2 H) 2 which provides us with one of the simplest mass spectra in our study (Figs. 2F and SI, Page S7). The spectrum is dominated by the presence of a low molecular ion peak at m/z = 390 [M] + and the base peak at m/z = 375 [M − CH 3 ] + arising from loss of the methyl group at the bridgehead, leaving a stabilized trityl radical cation. In the case of BP-BP(CF 2 H) 2 one can observe the molecular ion peak at m/z = 452 [M] + and the base peak at m/z = 375 [M − Ph] + arising from loss of one of the phenyl rings at the bridgehead for a species that resembles the one obtained with BP-AP(CF 2 H) 2 (vide supra) (Figs. 2G and SI, Page S9). Other peaks of interest that are significant in the spectrum are m/z = 309 [M − C 7 H 5 OF 2 ] + arising from the loss of one of the substituted phenyl groups to generate a trityl radical carbocation but not as strong as the one featured on the base peak.</p><p>One interesting product that provides a rich mass spectrum as its fluorinated derivative is BP-FL(CF 2 H) 2 . In the spectrum one can see several strong peaks arising from the breakdown of the product in and something quite interesting, the base ion peak is also the molecular ion peak at m/z = 450 [M] + . Other peaks in the spectra include one at m/z = 383 [M − OCF 2 H] + representing the loss of one of the difluoromethoxy groups. The peak at m/z = 307 [M − C 7 H 5 OF 2 ] + represents the loss of one of the substituted phenyl rings from the bridgehead junction to yield a stable trityl radical carbocation akin to the process observed for BP-BP(CF 2 H) 2 (Figs. 2H and SI, Page S17). In similar fashion, the spectrum of BP-Z(CF 2 H) 2 provides us ion fragments that seem to be analogous 1. Gas chromatographic properties for the native and bis-difluoromethylated bisphenols described in this work. The corresponding masses for the more abundant fragment ions along with their relative intensity to the base peak have been in included (in parentheses). In the MS peaks columns, the molecular ion peaks for each compound have been italicized. representing the loss of one of the difluoromethoxy groups. One of the most important bisphenols studied in this work is BPS due to its long time notoriety as a suitable substitute of BPA for manufacturing processes but still remaining a threat to human health as evidenced by numerous studies 32 . During the course of our studies, the native BPS possessed low response during the EI-GC-MS analysis (SI, Page S20) a characteristic that was diametrically obvious in its derivatized counterpart BP-S(CF 2 H) 2 . In the derivatized compound, BP-S(CF As the protocol involves a highly basic solution for the generation of the difluorocarbene species, we hypothesized that the acidic bisphenols would be amenable for selective derivatization over other less acidic alcohol species. Thus, in order to test the selectivity of the protocol for bisphenols over other alcohols, we carried out the difluoromethylation of BPA in the presence of six other alcohols. The alcohols were chosen so as to span a range of structural diversity and reactivity. Thus, incubation of BPA spiked in a mixture of 3-pentanol, 2-methyl-2-pentanol, 2,2-thiodiethanol, pinacolyl alcohol, 3,7-dimethyl-3-octanol and N,N-dimethylaminoethanol followed by treatment of the mixture with excess DBDFP resulted in the sole derivatization of BPA over the other alcohols (Fig. 3). This demonstrates that the protocol is a viable method for the specific labeling of other acidic alcohols (i.e. chlorophenols, resorcinols) under the basic conditions employed for its execution 23 . The observed selectivity lies in the magnitude of the pK a values for phenols that are orders of magnitude lower (i.e. ~7-8) than those exhibited by most alcohols (i.e. ~15-17). Consequently, under the basic conditions (pH ~ 12.8) at which the www.nature.com/scientificreports www.nature.com/scientificreports/ derivatization is conducted, the phenoxide not the alkoxide anion is the predominant species in the mixture and thus the most reactive towards difluoromethylation. Using the Henderson-Hasselbalch equation (Eq. 1) one can deduce that at pH = 12.8, the bisphenols practically exist as their phenoxide ions (with a pK a = 7, 99.9%) while the aliphatic alcohols employed in this study would only marginally exist as their alkoxide counterparts (with a pK a = 16, 0.001%).</p><p>a</p><p>In addition, one can invoke the much higher nucleophilicity of a phenoxide anion over that of an alkoxide anion as a reason for the selectivity observed in the process, however, this may not be a relevant contributor to the observed behavior as excess of the DBDFP is employed as eventually that could lead to the derivatization of the other alcohols.</p><p>In order to exploit the advantage that this protocol offers over other derivatization methods, we decided to test its derivatizing power in a Nebraska soil sample that was spiked with BPA initially at a 10 μg.g −1 concentration and then at a 1 μg.g −1 concentration. The protocol involves the direct conversion of the contaminated soil into a biphasic mixture (KOH/H 2 O//CH 3 CN) followed by the in situ derivatization of BPA with DBDFP. Upon vortexing and subsequently allowing the soil residue to settle to the bottom of the vial along with the aqueous layer, the top acetonitrile layer was then aliquoted, dried and analyzed by GC-MS. Nebraska soil was selected as a matrix due to its composition that encompasses several characteristics not normally present in other soils such as a total organic content (TOC, 1.9%) that provides several organic analytes that can act as interferences. The soil also possesses a large percentage of silt (58%) and clay matter (32%) that are recognized as challenging matrices in the field of soil extraction and analysis due to their inherent reactivity and its fine particulate texture making it difficult for derivatizing agents to access trapped analytes 33 . When carrying out the derivatization on the Nebraska soil, we were delighted to find it was successful at tagging the BPA providing its bis-difluoromethylated ether in enough concentration to be easily detected by GC-MS means (Fig. 4). After this initial assessment at 10 μg.g −1 BPA concentrations, we decided to explore the derivatization at an order of magnitude lower concentration of BPA (1 μg g −1 ) in the same soil. Again, the derivatization of BPA was successful in yielding the bis-difluoromethylated analog in the soil. Interestingly, when dealing with such a low concentration of the analyte, we made use of a selected-ion extraction mode (m/z = 313, strongest ion in the chromatograph) in order to unambiguously identify the fluorinated BPA demonstrating the ability of the protocol to be not only rapid (30 seconds) at tagging the BPA, but successful at detecting this endocrine disrupting compound at such low concentration (Fig. 5). The method detection limit (MDL) for the bis-difluoromethylated BPA in this soil was determined to be 0.01 μg.mL −1 (Supporting Information, Page S4).</p><!><p>Materials. All chemicals were purchased from commercial suppliers and used as received. Acetonitrile, methylene chloride, Bisphenol E, Bis-(4-hydroxyphenyl)methane (Bisphenol F), Bisphenol BP, Bisphenol G, 4,4′-(9-fluorenylidene)diphenol (Bisphenol FL), 3-pentanol, 2-methyl-2-pentanol, 2,2′-thiodiethanol (thiodiglycol), pinacolyl alcohol (3,3-dimethyl-2-butanol), 3,7-dimethyl-3-octanol, and 2-dimethylaminoethanol were Nebraska (EPA standard 54-135-4) soil was obtained from the soil matrix library at the Forensic Science Center in the Lawrence Livermore National Laboratory. All new bis-difluoromethylated bisphenols were purified using a Biotage Isolera flash column chromatography purification system using SNAP KP-Si gel column cartridges. HRMS analyses were obtained in the Forensic Science Center at the Lawrence Livermore National Laboratory using chemical ionization (CI). Combustion analyses were conducted at Galbraith Laboratories (Knoxville, TN).</p><!><p>EI-GC-MS analyses were carried out employing a 6890 Agilent GC equipped with a 5975 MS detector featuring a split/splitless injector 23,33 . The column used for our analyses was an Agilent DB-5MS capillary column (dimensions: 30 m × 0.25 mm i.d. × 0.25 µm i.f.). Ultra-high purity helium was employed as the carrier gas at a flow rate of 0.8 mL.min −1 . The injector temperature was set to 250 °C, and the injection volume was 1 µL. The oven temperature program used for the work was the following: 40 °C (held for t = 3 min), increased at a rate of 8 °C.min −1 to 300 °C and then held for t = 3 min. The MS ion source and quadrupole temperatures were set at 230 °C and 150 °C, respectively. The electron ionization energy used was 70 eV. The MS was operated to scan from m/z = 29 to m/z = 600 in t = 0.4 sec.</p><!><p>The initial protocol conditions included longer reaction times and stirring rather than vortexing aimed at the definitive derivatization of the panel of 11 BPs. Thus, the bisphenol (0.04 mmol) was placed in a glass autosampler vial equipped with a small stir bar and treated sequentially via pipette with 0.1 M KOH/H 2 O (800 μL), acetonitrile (CH 3 CN, 800 μL) and diethyl (bromodifluoromethyl) phosphonate (DBDFP, 21.4 μL, 0.12 mmol, 3.0 equiv. to bisphenol). The vial was capped and stirred vigorously at ambient temperature for 5 minutes. After the stirring was finalized, the mixture was allowed to stand to reveal a biphasic mixture and 500 μL of the clear, top layer (acetonitrile) was aliquoted into another glass autosampler vial containing anhydrous sodium sulfate (50 mg). The dried, organic fraction was passed through a syringe PTFE filter disc (0.45 μm) and 20 μL of the filtrate were aliquoted and diluted to 1.5 mL total volume with methylene chloride in an autosampler vial for GC-MS analysis.</p><p>Assessing the selectivity of the derivatization of bisphenol A in a mixture of aliphatic alcohols. In a 2 mL glass scintillation vial equipped with a stir bar, a 500 μL stock solution of seven alcohols, six of them aliphatic (3-pentanol, 2-methyl-2-pentanol, 2,2-thiodiethanol, pinacolyl alcohol, 3,7-dimethyl-3-octanol and N,N-dimethylaminoethanol) and BPA (each one at a 1000 μg mL −1 concentration) were treated sequentially with 0.1 M KOH/H 2 O (800 μL), acetonitrile (800 μL) and DBDFP (50 μL). The resulting biphasic mixture was vigorously stirred for 5 minutes at ambient temperature. After the stirring was stopped and the biphasic mixture revealed, 20 μL of the top layer (acetonitrile) was aliquoted into another autosampler vial and diluted to 1.5 mL with methylene chloride for GC-MS analysis. The extended period of time (5 min.) in contrast to the established Detection of bisphenol A in Nebraska soil sample at a 10 μg.g −1 concentration. Three sets of Nebraska soil samples (100 mg) in 4 mL vials were spiked with a BPA solution (1 μg mL −1 ) in methylene chloride and mixed, via tumbling, using a rotary evaporator at 40 °C for 15 minutes that after fully drying leads to a 10 μg g −1 BPA-contaminated soil. The contaminated soil was treated with a 0.1 M KOH aqueous solution (800 μL), followed by the sequential addition of acetonitrile (800 μL) and DBDFP (30 μL). The vials were capped, and the resulting biphasic suspensions were each mixed using a vortex for 30 seconds. After this time, 800 μL of the organic, top layer was aliquoted into an autosampler vial and dried with anhydrous sodium sulfate (30 mg). After drying, 100 μL of the organic phase was aliquoted into an autosampler vial equipped with a glass insert for GC-MS analysis.</p><p>Direct derivatization protocol for bisphenol A-treated Nebraska soil spiked at 1 μg.g −1 concentration. Nebraska soil (100 mg) was placed in a 4 mL vial and spiked with a BPA solution (0.1 μg mL −1 ) in methylene chloride. The suspension was mixed, via tumbling, using a rotary evaporator at 40 °C for 15 minutes that after fully drying leads to a 1 μg g −1 BPA-contaminated soil sample. The soil was treated using a pipette with 0.1 M KOH/H 2 O (1 mL), and then sequentially with acetonitrile (1 mL) and DBDFP (30 μL). The mixture was vortexed for 30 seconds at ambient temperature after which time an aliquot was extracted via pipette (800 μL), dried over anhydrous sodium sulfate (Na 2 SO 4 , 30 mg) and transferred (100 μL) to an autosampler vial equipped with a glass insert for GC-MS analysis. This set of experiments were conducted in triplicates.</p><p>General procedure for the synthesis of bis-difluoromethylated bisphenols. The bisphenol (0.84 mmol) is dissolved in CH 3 CN (3 mL) and then treated sequentially with DBDFP (500 μL, 2.8 mmol, 3.3 equiv. to bisphenol) followed by 0.1 M KOH/H 2 O (pH 12.8, 3 mL). The resulting mixture was stirred at ambient temperature for 30 minutes. The acetonitrile layer was dried over Na 2 SO 4 , evaporated in vacuo and purified by flash column chromatography (hexanes → 7:3 EtOAc/hexanes) to furnish the bis-difluoromethyl-bisphenol. All characterization data associated for each bis-difluoromethyl-bisphenol used in this study can be found in the SI section of this manuscript.</p><!><p>The methodology described herein represents the first expedient and practical derivatization of BPs via difluoromethylation for their enhanced and unequivocal detection by GC-MS. A remarkable aspect of the protocol is the fact that for the determination of BPs in soil, there are no prior extraction/concentration steps involved, thus greatly speeding up the process by removing the sample preparation step from the analysis. The fast and direct derivatization of BPA in Nebraska soil (at 1 μg.g −1 ) showcases the protocol's practicality and establishes it as a strong derivatization tool in the analytical chemist's toolbox for these types of compounds.</p>

Scientific Reports - Nature

Proton Transfer from C-6 of Uridine 5\xe2\x80\xb2-Monophosphate Catalyzed by Orotidine 5\xe2\x80\xb2-Monophosphate Decarboxylase: Formation and Stability of a Vinyl Carbanion Intermediate and the Effect of a 5-Fluoro Substituent

The exchange for deuterium of the C-6 protons of uridine 5\xe2\x80\xb2-monophosphate (UMP) and 5-fluorouridine 5\xe2\x80\xb2-monophosphate (F-UMP) catalyzed by yeast orotidine 5\xe2\x80\xb2-monophosphate decarboxylase (ScOMPDC) at pD 6.5 \xe2\x80\x93 9.3 and 25 \xc2\xb0C was monitored by 1H NMR spectroscopy. Deuterium exchange proceeds by proton transfer from C-6 of the bound nucleotide to the deprotonated side chain of Lys-93 to give the enzyme-bound vinyl carbanion. The pD-rate profiles for kcat give turnover numbers for deuterium exchange into enzyme-bound UMP and F-UMP of 1.2 \xc3\x97 10\xe2\x88\x925 and 0.041 s\xe2\x88\x921, respectively, so that the 5-fluoro substituent results in a 3400-fold increase in the first-order rate constant for deuterium exchange. The binding of UMP and F-UMP to ScOMPDC results in 0.5 and 1.4 unit decreases, respectively, in the pKa of the side chain of the catalytic base Lys-93, showing that these nucleotides bind preferentially to the deprotonated enzyme. We also report the first carbon acid pKas for proton transfer from C-6 of uridine (pKCH = 28.8) and 5-fluorouridine (pKCH = 25.1) in aqueous solution. The stabilizing effects of the 5-fluoro substituent on C-6 carbanion formation in solution (5 kcal/mol) and at ScOMPDC (6 kcal/mol) are similar. The binding of UMP and F-UMP to ScOMPDC results in a greater than 5 \xc3\x97 109-fold increase in the equilibrium constant for proton transfer from C-6 so that ScOMPDC stabilizes the bound vinyl carbanions, relative to the bound nucleotides, by at least 13 kcal/mol. The pD-rate profile for kcat/Km for deuterium exchange into F-UMP gives the intrinsic second-order rate constant for exchange catalyzed by the deprotonated enzyme as 2300 M\xe2\x88\x921 s\xe2\x88\x921. This was used to calculate a total rate acceleration for ScOMPDC-catalyzed deuterium exchange of 3 \xc3\x97 1010 M\xe2\x88\x921, which corresponds to a transition state stabilization for deuterium exchange of 14 kcal/mol. We conclude that a large portion of the total transition state stabilization for the decarboxylation of orotidine 5\xe2\x80\xb2-monophosphate can be accounted for by stabilization of the enzyme-bound vinyl carbanion intermediate of the stepwise reaction.

proton_transfer_from_c-6_of_uridine_5\xe2\x80\xb2-monophosphate_catalyzed_by_orotidine_5\xe2\x80\xb2

INTRODUCTION<!>Materials<!>5-Fluorouridine 5\xe2\x80\xb2-Monophosphate<!>Preparation of Solutions<!>Extinction Coefficient of 5-Fluorouridine 5\xe2\x80\xb2-Monophosphate<!>pKa of the N-3 Hydron of 5-Fluorouridine 5\xe2\x80\xb2-Monophosphate in H2O and D2O (pKNL)<!>Prepation of Yeast OMPDC in D2O<!>Enzyme Assays<!>NMR Analyses<!>Deuterium Exchange at C-6 of F-UMP Monitored by 1H NMR Spectroscopy<!>Deuterium Exchange at C-6 of UMP Monitored by 1H NMR Spectroscopy<!>Deuterium Exchange at C-6 of 5-Fluorouridine 5\xe2\x80\xb2-Monophosphate Catalyzed by Yeast OMPDC<!>Deuterium Exchange at C-6 of Uridine 5\xe2\x80\xb2-Monophosphate Catalyzed by Yeast OMPDC<!>Ionization of F-UMP and UMP at N-3<!>Nonenzymatic Deuterium Exchange at C-6 of Uridine, 5-Fluorouridine and 1,3-Dimethyl-5-Fluorouracil Catalyzed by Deuterioxide Ion in D2O<!>Kinetic Parameters for Enzyme-Catalyzed Deuterium Exchange into UMP and F-UMP<!>pD-Rate Profiles for Deuterium Exchange at C-6 of F-UMP and UMP<!>Identity of the Catalytic Base for Deuterium Exchange<!>Intrinsic Affinity of OMPDC for UMP and F-UMP<!>Mechanism of Deuterium Exchange: Formation of a Vinyl Carbanion at OMPDC<!>Carbon Acidity of Substituted Uracils in Aqueous Solution<!>Stabilization of the UMP and F-UMP Vinyl Carbanions by ScOMPDC<!>Enzymatic Rate Accelerations for Decarboxylation and Deuterium Exchange

<p>Orotidine 5′-monophosphate decarboxylase (OMPDC) is an essential enzyme that catalyzes the decarboxylation of orotidine 5′-monophosphate (OMP) to give uridine 5′-monophosphate (UMP), a critical step in the de novo biosynthesis of pyrimidine nucleotides (Scheme 1). In mammals, OMPDC is found as part of the bifunctional uridine 5′-monophosphate synthase complex, but in many lower organisms it exists as a discrete protein.1 The enzyme is an obligate dimer and residues from both subunits contribute to the active site architecture.2,3 Remarkably, OMPDC requires no metal ions or other cofactors but yet it effects an enormous 1017-fold acceleration of the decarboxylation of enzyme-bound OMP.3,4 A significant fraction of the total transition state stabilization for decarboxylation of ca. 31 kcal/mol3 results from utilization of the 12 kcal/mol intrinsic binding energy of the 5′-phosphodianion group of OMP.5–8</p><p>The first X-ray crystal structures for OMPDC appeared in 2000 and, notably, they revealed the absence of suitably placed catalytic residues required for several of the previously proposed reaction mechanisms.9–12 This precipitated a renewal of the scientific debate about the mechanism and the origin of the enzymatic rate acceleration for decarboxylation.13–19 The simplest possibility is the direct decarboxylation of OMP by a stepwise mechanism through the unstable UMP vinyl carbanion intermediate that undergoes subsequent proton transfer from the enzyme to generate the product (Scheme 1). We have argued that this mechanism is mandated by the observation of product isotope effects of unity for the decarboxylation of OMP and 5-fluoroorotidine 5′-monophosphate (F-OMP) in a mixture of 50:50 (v/v) H2O/D2O, which show that there is no discrimination between H and D at the proton donor in the transition state for the product-determining step (Scheme 2).20,21 These results eliminated a concerted reaction mechanism for decarboxylation in which proton transfer from the enzyme provides electrophilic push to the loss of CO2.9,13</p><p>We reported earlier that OMPDC from Saccharomyces cerevisiae (ScOMPDC) catalyzes the exchange for deuterium from solvent D2O of the C-6 proton of UMP to give UMP labeled with deuterium at C-6 (d-UMP, Scheme 2).22 These relatively slow reactions precluded extensive investigation of the kinetics and mechanism of OMPDC-catalyzed deuterium exchange into UMP. We reasoned that the addition of a 5-fluoro substituent to give 5-fluorouridine 5′-monophosphate (F-UMP) should result in stabilization of the negative charge at C-6 of the corresponding vinyl carbanion.23,24 Therefore, study of OMPDC-catalyzed deuterium exchange into F-UMP should be much more feasible than for UMP, and it would provide insight into the proton transfer reaction catalyzed by OMPDC. Our initial report of the OMPDC-catalyzed C-6 deuterium exchange reaction of F-UMP employed wildtype and the D70N and D70G mutants of OMPDC from Methanothermobacter thermautotrophicus.25</p><p>We present here a full report of our studies of the ScOMPDC-catalyzed C-6 deuterium exchange reaction of UMP,22 along with a complete set of data for the pD-dependence of the corresponding ScOMPDC-catalyzed C-6 deuterium exchange reaction of the more reactive substrate F-UMP. The results are consistent with a deuterium exchange reaction that is the formal reverse of the proton transfer "half reaction" that occurs during the decarboxylation of OMP, and they provide convincing support for our conclusion that the decarboxylation of OMP proceeds by a stepwise mechanism through the UMP vinyl carbanion intermediate (Scheme 2).20,21 We also report the first determination of reliable C-6 carbon acid pKas for a series of substituted uracils in aqueous solution, including the novel C-6 carbon acidities of the nucleosides uridine and 5-fluorouridine. A comparison of the data for the enzyme-catalyzed deuterium exchange reactions of UMP and F-UMP with those for proton transfer from uridine and 5-fluorouridine in aqueous solution shows that ScOMPDC stabilizes the bound UMP and F-UMP vinyl carbanions, relative to the bound nucleotides, by at least 13 kcal/mol. We conclude that a large portion of the rate acceleration for the decarboxylation of OMP can be accounted for by stabilization of the enzyme-bound vinyl carbanion intermediate of the stepwise reaction.</p><!><p>Orotidine 5′-monophosphate was purchased from Sigma or was prepared by chemical or enzymatic methods using modifications of literature procedures.24,26,27 Uridine, uridine 5′-monophosphate disodium salt (99%, from yeast), 5-fluorouridine and 2-(N-morpholino)ethanesulfonic acid sodium salt (MES) were purchased from Sigma. 5-Fluoro-1,3-dimethyluracil, phosphorus(V) oxychloride and imidazole were purchased from Aldrich. Trimethyl phosphate, glycylglycine (> 99%) and 3-(N-morpholino)propanesulfonic acid (MOPS, ≥ 99.5%) were purchased from Fluka. 6-Azauridine 5′-monophosphate (free acid) was purchased from R. I. Chemical. Deuterium oxide (99.9% D), sodium deuteroxide (30% in D2O, 99.5% D) and deuterium chloride (35% w/w, 99.9% D) were purchased from Cambridge Isotope Laboratories. Bovine serum albumin (BSA) was from Roche and was dialyzed against 0.1 M NaCl in D2O. Trimethyl phosphate was dried over phosphorus pentoxide and distilled under reduced pressure. Imidazole was recrystallized from benzene. Water was from a Milli-Q Academic purification system. All other chemicals were reagent grade or better and were used without further purification.</p><!><p>5-Fluorouridine 5′-monophosphate was prepared as its triethylammonium salt from 5-fluorouridine using modifications of a literature procedure.28 Phosphorus(V) oxychloride (0.13 mL, 1.4 mmol) and 5-fluorouridine (0.20 g, 0.76 mmol) were dissolved in dry trimethyl phosphate (2 mL) at 0 °C. After 16 h the mixture was poured into cold water (100 mL), the solution was adjusted to pH 7 using 1 M NaOH, and water was added to give a total volume of 180 mL. The mixture was applied to a column of DEAE Sephadex A25 equilibrated with 50 mM triethylammonium bicarbonate and an elution gradient of 50 – 600 mM triethylammonium bicarbonate was applied. Fractions containing the desired product were pooled and the buffer was removed by lyophilization to give a 50% yield of 5-fluorouridine 5′-monophosphate as its triethylammonium salt. This material was shown to be free of inorganic phosphate by 31P NMR. 1H NMR (500 MHz, D2O, pD ≈ 7) δ7.83 (1H, 3JHF = 6.0 Hz, C6-H), 5.87 (1H, 3JHH = 5.0 Hz, 5JHF = 1.5 Hz, C1′-H), 4.21 (1H, m, C2′-H), 4.19 (1H, m, C3′-H), 4.08 (1H, m, C4′-H), 3.84 (2H, m, C5′-H). Chemical shifts are reported relative to HOD at 4.67 ppm.</p><!><p>Solution pH or pD was determined at 25 °C using an Orion Model 720A pH meter equipped with a Radiometer pHC4006-9 combination electrode that was standardized at 25 °C. Values of pD were obtained by adding 0.40 to the observed reading of the pH meter.29</p><p>The acidic protons of glycylglycine were exchanged for deuterium by dissolution in D2O followed by evaporation and drying under vacuum at 55 °C. Buffered solutions of glycylglycine or MOPS in D2O were prepared by dissolving the acidic form and NaCl (if needed) in D2O followed by the addition of a measured amount of NaOD to give the desired acid/base ratio at I = 0.1. Buffered solutions of imidazole or MES in D2O were prepared by dissolving the basic form and NaCl (if needed) in D2O followed by the addition of a measured amount of DCl to give the desired acid/base ratio at I = 0.1. Solutions in D2O were stored in a dessicator.</p><p>The concentration of OMP in stock solutions was determined from its absorbance in 0.1 M HCl at 267 nm using ε = 9430 M−1 cm−1.30 Stock UMP (100 mM) was prepared by dissolving the disodium salt in D2O to give a solution at pD 8.2 (I = 0.3) and its concentration was determined from the absorbance in 0.1 M HCl at 262 nm using ε = 10,000 M−1 cm−1.31 Stock solutions of F-UMP were prepared by dissolving the triethylammonium salt in D2O to give a solution at pD ≈ 8 and its concentration was determined from the absorbance in 0.1 M HCl at 270 nm using ε = 9160 M−1 cm−1 (see below). Stock 6-azauridine 5′-monophosphate (6-aza-UMP, 12 mM) was prepared by dissolving the free acid in D2O and the pD was adjusted to 7.9 by the addition of 4.3 M NaOD. Solutions of OMP, UMP, F-UMP and 6-aza-UMP were stored at −20 °C.</p><!><p>A stock solution of F-UMP was prepared in D2O (pD ≈ 8) and was diluted 10-fold with 60 mM imidazole buffer (50% free base) in D2O at pD 7.6 and I = 0.1 (NaCl) to give a final concentration of 54 mM imidazole. The solution was analyzed by 1H NMR spectroscopy and the integrated areas of the signals due to the protons of F-UMP were compared with those due to the C-4 and C-5 protons of imidazole to give the concentration of F-UMP in the stock solution in D2O as 102 mM. The stock solution of F-UMP in D2O (102 mM) was then used to determine λmax = 270 nm and an extinction coefficient of ε = 9160 M−1 cm−1 for F-UMP in 0.1 M HCl.</p><!><p>The acidity of the N-3 hydron of F-UMP in H2O and D2O was determined spectrophotometrically by monitoring the decrease in absorbance at 269 nm, Aobsd, that accompanies ionization at N-3. The absorbance of 163 μM F-UMP was determined as a function of pL in H2O or D2O at 25 °C and I = 0.1 (NaCl) using the following buffers: 0.1 M HCl, 0.01 M LCl, acetate (pH 4.7), MES (pD 6.4), imidazole (pD 7.0), MOPS (pL 7.1 – 7.4), glycylglycine (pL 7.6 – 9.4), glycine (pL 9.7 – 10.3), 0.01 M NaOL. The data were fit to eq 1, where Aacid and Abase are the absorbances at the acidic and basic extremes, to give the acidity constants KNL for ionization of F-UMP at N-3 in H2O and D2O.</p><!><p>OMPDC from Saccharomyces cerevisiae (ScOMPDC) was prepared as described previously.32,33 The protein sequence differs from the published sequence for true wildtype yeast OMPDC by the following mutations: S2H,34 C155S,35 A160S and N267D.34 Except for the C155S mutation, the sequence is the same as that observed in the published crystal structure of wildtype yeast OMPDC.11 This C155S variant is more stable than, but kinetically and structurally essentially identical with, the Cys-155 enzyme.35 In this work, all experiments were conducted using this C155S variant.</p><p>Samples of overexpressed and purified ScOMPDC that had been stored at −80 °C were defrosted and extensively dialyzed at 4 °C against 10 mM MOPS (50% free base, pH 7.1) containing 100 mM NaCl in order to remove glycerol. This was followed by dialysis against several changes of buffer in D2O using a D-tube dialyzer (6 – 8 kDa MWCO, Novagen) placed inside a narrow vessel that was isolated from atmospheric moisture using parafilm. The buffers used for dialysis were: 62.5 or 125 mM glycylglycine, 80% free base (pD 9.3); 125 mM glycylglycine, 20% free base (pD 8.1); 100 mM MOPS, 50% free base or 60 mM imidazole, 50% free base (pD 7.6); 100 mM MOPS, 35% free base (pD 7.4); 60 or 125 mM imidazole, 20% free base (pD 7.0); 100 mM MES, 40% free base (pD 6.5). All buffers were in D2O at I = 0.1 (NaCl). The concentration of ScOMPDC in the stock solutions in D2O, determined by standard assay as described below, was 0.3 – 0.5 mM (9 – 15 mg/mL).</p><!><p>The activity of ScOMPDC in the enzyme stock solutions in D2O and in the deuterium exchange reaction mixtures was determined by monitoring the decrease in absorbance at 279 nm accompanying the enzyme-catalyzed decarboxylation of OMP (Δε = −2400 M−1 cm−1 at 25 °C).2,7,22,32 Prior to assay, the stock solutions of ScOMPDC were diluted with 10 mM MOPS, 50% free base (pH 7.1) containing 100 mM NaCl and 0.4 mg/mL BSA to give a final concentration of ca. 20 μM ScOMPDC. Assays in a total volume of 1 mL were conducted in 10 mM MOPS, 50% free base (pH 7.1), at 25 °C and I = 0.105 (NaCl), with 40 – 50 μM OMP (25 – 30Km). The reaction was initiated by the addition of 1 SL of the diluted stock solution of ScOMPDC or an aliquot (up to 100 μL) of the deuterium exchange reaction mixture to give a final enzyme concentration of 20 – 40 nM, and the initial velocity of the ensuing decarboxylation of OMP was determined within 1 min. The concentration of ScOMPDC in the stock solution or in the deuterium exchange reaction mixture was then calculated from the observed value of Vmax (M s−1) using the relationship Vmax = kcat[E], with kcat = 15 s−1.2,7</p><!><p>1H NMR spectra at 500 MHz (8 – 128 transients) were acquired at 25 °C using a Varian Unity Inova 500 spectrometer with a sweep width of 6000 Hz, a 90° pulse angle, an acquisition time of 6 s and a total relaxation delay between pulses of 90 s (≥ 10T1) for the reactions of UMP or 30 s (≥ 9T1) for the reactions of F-UMP. Chemical shifts were referenced to HOD at 4.67 ppm. Values of T1 = 9 s and 7 s were determined for the C-5 and C-1′ protons, respectively, of [6-2H]-uridine 5′-monophosphate (d-UMP) in glycylglycine buffer at pD 9.4. Values of T1 = 1.5 s and 3.3 s were determined for the C-6 and C-1′ protons, respectively, of [6-1H]-5-fluorouridine 5′-monophosphate (h-F-UMP) in imidazole buffer at pD 7.6. Before determination of the integrated peak areas, the signals of interest were greatly expanded, accurately phased and subjected to a first-order drift correction.</p><!><p>The C-6 deuterium exchange reactions of F-UMP catalyzed by ScOMPDC in D2O at pD 6.5 – 9.3 were followed directly by 1H NMR spectroscopy at 500 MHz. Reactions in a volume of 1 mL were initiated by the addition of an aliquot of F-UMP in D2O to a mixture of the appropriate buffer, NaCl, BSA and ScOMPDC in D2O to give final concentrations of 0.1 – 5 mM F-UMP, 0.4 mg/mL (0.04%) BSA (see below), and 0.2 – 40 μM ScOMPDC at I = 0.1 (NaCl). 700 μL of the reaction mixture was transferred to an NMR tube and the progress of deuterium exchange was monitored by 1H NMR spectroscopy at 25 °C for 2 – 8 h. The remaining portion of the reaction mixture was incubated at 25 °C and the activity of ScOMPDC was monitored by periodic standard assay as described above. After completion of the deuterium exchange reaction the enzyme was removed by ultrafiltration and the pD of the filtrate was recorded. There was no change in pD of the reaction mixture during these reactions at 25 °C which were followed for up to 8 h.</p><p>In all cases the concentration of enzyme used was chosen to accommodate the time needed to obtain at least 4 timepoints, with acquisition of an appropriate number of NMR transients, in one half-time for the deuterium exchange reaction. This required that lower concentrations of enzyme be used for the reactions at low concentrations of F-UMP for which more transients per timepoint were required for reasons of sensitivity. For the reactions of F-UMP in the presence of low concentrations of ScOMPDC (≥ 2 μM) in the absence of BSA there was up to a 25% decrease in the observed activity of ScOMPDC during the deuterium exchange reaction. However, the inclusion of 0.4 mg/mL (0.04%) BSA stabilized this loss of enzyme activity with time, which we attribute mainly to protein adsorption. There were also small decreases (up to 10%) in enzyme activity after extended reaction times in the presence of low concentrations of F-UMP (≤ Km) which we attribute to the instability of the dimeric form of unliganded ScOMPDC.2 A control experiment showed that the incubation of 10 μM ScOMPDC for 24 h at pD 9.4 (100 mM glycylglycine) and 25 °C resulted in a 15% loss in enzyme activity in the presence of 0.3 mM F-UMP (1.5Km) but no loss in enzyme activity in the presence of 2 mM F-UMP (10Km).</p><!><p>The C-6 deuterium exchange reactions of UMP catalyzed by ScOMPDC in D2O at pD 7.0 – 9.3 were followed by 1H NMR spectroscopy at 500 MHz. Reactions in a volume of 1 – 4 mL were initiated by the addition of an aliquot of UMP in D2O to a mixture of the appropriate buffer, NaCl and ScOMPDC in D2O to give final concentrations of 2.5 – 10 mM UMP and 100 – 300 μM ScOMPDC at I = 0.1 (NaCl). The reaction mixture was divided into 1 mL portions that were placed in screw-cap centrifuge tubes equipped with O-rings and incubated in a water bath at 25 °C. At various times a tube was withdrawn and lightly centrifuged to sediment any particulate matter. An aliquot of the reaction mixture was removed and the enzyme-catalyzed deuterium exchange reaction was stopped by the addition of a 2 to 3-fold excess over enzyme of the tight-binding competitive inhibitor 6-aza-UMP (12 mM stock, pD 7.9). The enzyme-inhibitor complex, which was formed quantitatively, was removed by ultrafiltration using an Amicon Microcon device (10K MWCO) and the filtrate was transferred to an NMR tube. The extent of deuterium exchange was then determined by 1H NMR spectroscopy. After NMR analysis the sample was removed from the NMR tube and its pD was determined. The activity of ScOMPDC in the reaction mixture was monitored by periodic standard assay as described above. There was no significant loss of enzyme activity or change in pD of the reaction mixture during these reactions at 25 °C which were followed for up to 12 days.</p><!><p>The exchange for deuterium of the C-6 proton of [6-1H]-5-fluorouridine 5′-monophosphate (h-F-UMP) ([S]o = 0.1 – 5 mM) to give [6-2H]-5-fluorouridine 5′-monophosphate (d-F-UMP) catalyzed by OMPDC from Saccharomyces cerevisiae (ScOMPDC, 0.2 – 40 μM) in D2O at pD 6.5 – 9.3, 25 °C and I = 0.1 (NaCl) was followed directly by monitoring the disappearance of the C-6 proton by 1H NMR spectroscopy (Scheme 3).</p><p>Figure 1 shows partial 1H NMR spectra at 500 MHz in the regions of the C-6 proton (left) and C-1′ (anomeric) proton (right) obtained at various times during the exchange for deuterium of the C-6 proton of h-F-UMP ([S]o = 2.5 mM) catalyzed by ScOMPDC (9.2 μM) in D2O at pD 9.3 (100 mM glycylglycine, 80% free base), 25 °C and I = 0.1 (NaCl). The spectra on the left of Figure 1 show that deuterium exchange at C-6 results in the clean disappearance of the doublet at 7.874 ppm (3JHF = 6.2 Hz) due to the C-6 proton. By contrast, after 43% exchange of the C-6 proton, there is a complex appearance of the signals in the region of the C-1′ (anomeric) proton (Figure 1B, right). This results from overlap of the signals of roughly equal size due to the anomeric proton of h-F-UMP at 5.885 ppm (double doublet, 3JHH = 5.2 Hz, 5JHF = 1.9 Hz, Figure 1A) and that of d-F-UMP, which appears as an upfield-shifted (Δδ = 0.002 ppm) double doublet at 5.883 ppm (3JHH = 5.2 Hz, 5JHF = 1.9 Hz, Figure 1C). At late reaction times the signal in the anomeric region is again simplified and is due mainly to the upfield-shifted anomeric proton of the product d-F-UMP. A similar upfield shift of the broad doublet due to the anomeric proton of UMP as a result of deuterium incorporation at C-6 was noted in our earlier communication.22</p><p>The progress of deuterium exchange into F-UMP was obtained from the integrated area of the doublet at 7.87 ppm due to the C-6 proton of h-F-UMP (AH), using the combined integrated areas of the signals due to the anomeric protons of both h-F-UMP and the product d-F-UMP at 5.88 ppm (AH+D) as an internal standard (Figure 1). Observed first-order rate constants for deuterium exchange were determined from the slopes of linear semi-logarithmic plots of reaction progress against time according to eq 2, covering 1 – 2 reaction half-times. Tables S1 – S6 of the Supporting Information give the values of kobsd (s−1) for exchange for deuterium of the C-6 proton of F-UMP in D2O at pD 6.45 – 9.33, 25 °C and I = 0.1 (NaCl) that were determined in these experiments.</p><!><p>The relatively slow exchange for deuterium of the C-6 proton of [6-1H]-uridine 5′-monophosphate (h-UMP) ([S]o = 2.5 – 10 mM) to give [6-2H]-uridine 5′-monophosphate (d-UMP) catalyzed by ScOMPDC (100 – 300 μM) in D2O at pD 7.0 – 9.3, 25 °C and I = 0.1 (NaCl) was followed by monitoring the appearance of the broad signal due to the C-5 proton of the product d-UMP by 1H NMR spectroscopy.22 The large coupling of the C-5 proton to the C-6 proton for h-UMP (3JHH = 8 Hz) is replaced by a small unresolved 3JHD coupling of the C-5 proton to deuterium at C-6 for d-UMP.22 The relatively high enzyme concentrations used in these experiments resulted in extremely broad 1H NMR signals for UMP so that the reactions could not be followed directly. At various times an aliquot of the reaction mixture was withdrawn and the reaction was stopped by the addition of a 2 to 3-fold excess over enzyme of the tight-binding competitive inhibitor 6-azauridine 5′-monophosphate (6-aza-UMP, Ki = 93 nM at pH 7.2).2,36 The enzyme•inhibitor complex, which was formed quantitatively, was removed by ultrafiltration and the filtrate was analyzed by 1H NMR spectroscopy.</p><p>The fractional extent of deuterium incorporation into UMP, f(d-UMP), was determined from analysis of the integrated areas of the signals due to the C-5 protons of h-UMP and the product d-UMP.22 Observed first-order rate constants for deuterium exchange were determined from the slopes of semi-logarithmic plots of the fraction of h-UMP remaining, f(h-UMP) = {1 - f(d-UMP)} against time, covering up to 30% reaction.22 The values of kobsd (s−1) for exchange for deuterium of the C-6 proton of UMP in D2O at pD 7.03 – 9.34, 25 °C and I = 0.1 (NaCl) were reported in the Supporting Information of the earlier communication.22 There was no detectable deuterium exchange into UMP (2 mM) after its incubation for 2 days in the presence of 260 μM ScOMPDC and 2 mM 6-aza-UMP at pD 9.3 (50 mM glycylglycine, 80% free base) at 25 °C. This shows that the observed enzyme-catalyzed deuterium exchange reactions occur at the active site of ScOMPDC.</p><p>The ScOMPDC-catalyzed deuterium exchange reactions of UMP at pD 7.6 – 9.3 were accompanied by up to 4% hydrolysis of the substrate 5′-phosphoryl group to give uridine. The very slow deuterium exchange reactions at pD 7.0 were accompanied by hydrolysis of both the 5′-phosphoryl group of UMP (up to 19%) and its glycosidic bond to give uracil (up to 24%). Control experiments conducted in the absence of ScOMPDC indicated that these side reactions are likely catalyzed by small contaminating enzyme activities present in our preparations of ScOMPDC. The rates of these hydrolysis reactions should be unaffected by the presence of deuterium at C-6 of UMP, so that they did not interfere with quantification of the extent of deuterium incorporation into the remaining UMP. However, at pD 7.0 in the presence of 2.5 mM UMP, the significant depletion of the pool of total UMP by these side reactions resulted in an apparent increase in the velocity of deuterium exchange with time (see kinetic analysis in the Discussion). In these cases, the observed rate constants for deuterium exchange were determined from the portion of the reaction that exhibited good first-order kinetics.22</p><!><p>The pKa of the N-3 hydron of F-UMP in H2O and in D2O at 25 °C and I = 0.1 (NaCl) was determined spectrophotometrically by monitoring the absorbance of F-UMP at 269 nm as a function of pL. The data gave pKNH = 7.84 ± 0.03 in H2O and pKND = 8.37 ± 0.05 in D2O, at 25 °C and I = 0.1 (NaCl). The pKa of 7.84 for the N-3 proton of F-UMP is very similar to the reported pKas of 7.90 for the N-3 proton of 5-fluoro-2′-deoxyuridine 5′-monophosphate,2 and 7.67 for the N-3 proton of 5-fluoro-2′-deoxyuridine,37 in H2O at 25 °C and I = 0.1. This shows that there is little or no effect of a 2′-OH or a 5′-phosphoryl group on the pKa of the N-3 proton of simple uridine derivatives. The pKa of 9.3 reported for the N-3 proton of 2′-deoxyuridine in H2O at 25 °C and I = 0.1 (NaCl)37 and the 0.5 unit difference in the values of pKNH and pKND for F-UMP determined here can be used to estimate pKND = 9.8 for ionization of the N-3 hydron of UMP in D2O. Therefore, there is little dissociation of UMP at N-3 in the range of pD examined in our enzyme-catalyzed deuterium exchange reactions.</p><!><p>The exchange for deuterium of the C-6 protons of uridine, 5-fluorouridine and 1,3-dimethyl-5-fluorouracil at various temperatures and values of pD in D2O was followed by monitoring the disappearance of the C-6 proton of the substrate by 1H NMR spectroscopy, as described in the Supporting Information. The reactions of 5-fluorouridine in the presence of 1 M NaOD and of 1,3-dimethyl-5-fluorouracil at pD 10.2 were conducted at 25 °C. The slower reactions of uridine in the presence of 1 M NaOD and of 5-fluorouridine at pD 7.1 – 9.4 (I = 0.1) were conducted at 60 – 90 °C and rate constants for the reactions at 25 °C were obtained by extrapolation of Arrhenius plots (see the Supporting Information). Table S7 of the Supporting Information summarizes the observed second-order rate constants (kDO)obsd (M−1 s−1) for the deuterioxide-ion-catalyzed exchange reactions that were determined in these experiments. The data gave the following second-order rate constants kDO for deuterioxide-ion-catalyzed deuterium exchange at 25 °C (see Table S8 of the Supporting Information): 3.8 × 10−4 M−1 s−1 for uridine, 1.5 × 10−8 M−1 s−1 for uridine N-3 anion; 1.7 M−1 s−1 for 5-fluorouridine, 6.8 × 10−5 M−1 s−1 for 5-fluorouridine N-3 anion; 0.043 M−1 s−1 for 1,3-dimethyl-5-fluorouracil.</p><!><p>The velocity of enzyme-catalyzed deuterium exchange into h-UMP or h-F-UMP to give d-UMP or d-F-UMP (Scheme 3) is given by eq 3 derived for Scheme 4 (written for the reaction of UMP), where kex (s−1) is the turnover number for the E•h-UMP or E•h-F-UMP complex, [S]o is the total concentration of UMP or F-UMP irrespective of its isotopic composition, Kd is the binding constant for dissociation of the E•UMP or E•F-UMP complex, and [E] is the concentration of ScOMPDC.</p><p>Eq 4 relates the observed second-order rate constants for deuterium exchange, kobsd/[E] (M−1 s−1), to the kinetic parameters kex and Kd (Scheme 4), where kobsd is the observed first-order rate constant for deuterium exchange that was determined from the slope of a semi-logarithmic plot of reaction progress against time (see Experimental Section). This equation has two limits: (1) At [S]o ≫ Kd, the observed second-order rate constant for deuterium exchange is inversely proportional to [S]o, kobsd/[E] = kex/[S]o, where kex is the turnover number or the first-order rate constant for deuterium exchange into saturating enzyme-bound substrate. (2) At [S]o ≪ Kd, the observed second-order rate constant for deuterium exchange tends to the true second-order rate constant kex/Kd.</p><p>Figure 2 shows the dependence of kobsd/[E] (M−1 s−1) for deuterium exchange into F-UMP catalyzed by ScOMPDC (0.22 – 14 μM) on the total concentration of F-UMP in D2O at pD 8.15, 25 °C and I = 0.10 (NaCl). This observed second-order rate constant decreases as the concentration of F-UMP is increased, which is the expected behavior for an enzyme-catalyzed isotope exchange reaction for which the substrate and the product bind to the enzyme with essentially equal affinities (Kd, Scheme 4). The data were fit to eq 4 to give values of kex = 0.048 s−1 and Kd = 0.14 mM for deuterium exchange into F-UMP at pD 8.15.</p><p>Eq 4 may be transformed to give eq 5 which describes the familiar hyperbolic dependence of enzymatic initial rate on total substrate concentration. Figure 3A shows the dependence of the quantity kobsd[F-UMP]o/[E] (s−1) for ScOMPDC-catalyzed deuterium exchange into F-UMP on the total concentration of F-UMP at pD 6.45 – 9.33, 25 °C and I = 0.1 (NaCl). The data were fit to eq 5 to give values of kex (s−1) for deuterium exchange into saturating enzyme-bound F-UMP, and Kd (M) for binding of F-UMP to ScOMPDC. Table 1 gives the values of kex (s−1), Kd (M) and the second-order rate constants kex/Kd (M−1 s−1) for enzyme-catalyzed deuterium exchange into F-UMP obtained from these fits.</p><p>The observed second-order rate constants kobsd/[E] (M−1 s−1) for the deuterium exchange reactions of h-UMP to give d-UMP at [S]o = 2.5 – 10 mM are inversely proportional to the total concentration of UMP which shows that [S]o ≫ Kd under these conditions (eq 4). Figure 3B shows the dependence of the quantity kobsd[UMP]o/[E] for ScOMPDC-catalyzed deuterium exchange into UMP on the total concentration of UMP at pD 7.03 – 9.34, 25 °C and I = 0.1 (NaCl). Table 2 gives the first-order rate constants kex (s−1) for enzyme-catalyzed deuterium exchange into saturating enzyme-bound UMP that were determined as the average of the values of kobsd[UMP]o/[E] at each pD ([UMP]o ≫ Kd, eq 5).</p><!><p>Figure 4 shows the pD-rate profile for the second-order rate constant kex/Kd (M−1 s−1) for ScOMPDC-catalyzed deuterium exchange at C-6 of F-UMP in D2O, constructed using the data in Table 1. This pD-rate profile is bell-shaped and the slope of the ascending limb at low pD approaches 2.0, which shows that the second-order rate constant for enzyme-catalyzed deuterium exchange into F-UMP is dependent on the deprotonation of two groups at the free enzyme/substrate. We assign the downward break at high pD to ionization of the phosphodianion form of F-UMP at N-3 to give the substrate trianion (KSD, Scheme 5), for which we have determined pKSD = 8.37 in D2O under the experimental conditions for deuterium exchange (see Results).</p><p>The pH-rate profile for the second-order rate constant kcat/Km (M−1 s−1) for decarboxylation of the natural substrate OMP by ScOMPDC in H2O depends on the deprotonation of a group with an apparent pKa of 6.1 which is very similar to the pKa of 6.38 for the phosphoryl group of OMP determined by titration under the same conditions (25 °C, I = 0.1).2 The X-ray crystal structure of ScOMPDC liganded with the potent inhibitor 6-hydroxyuridine 5′-monophosphate (BMP) shows hydrogen bonds between the ligand phosphoryl group and the side chains of Gln-215, Tyr-217 and Arg-235, and the backbone NHs of Arg-235 and Gly-234 (see Figure 6). This strongly implies that the enzyme binds and turns over the phosphodianion form of the substrate.8,11,32 Therefore, a very reasonable assignment for the apparent pKa of 6.1 in the pH-rate profile for kcat/Km for decarboxylation of OMP is the second ionization of the phosphoryl group of OMP to give the substrate phosphodianion.2 By analogy, we assign the two ionizations at the free substrate/enzyme evident in the ascending limb at low pD in the pD-rate profile for kex/Kd for deuterium exchange into F-UMP (Figure 4) as ionization of the substrate phosphoryl monoanion (SD2−, Scheme 5) to give the phosphoryl dianion (SD2−), and the deprotonation of a group at the free enzyme to give the basic form of the enzyme.</p><p>The pKas of 6.26 and 6.25 for the phosphoryl groups of UMP and ribose 5′-monophosphate, respectively, determined by titration at 25 °C and I = 0.1,2 are essentially identical, which shows that the ionization of this group is relatively insensitive to the presence of the pyrimidine ring. Therefore, we estimate pKSD2 = 6.8 for ionization of the phosphoryl group of F-UMP in D2O at I = 0.1 (Scheme 5, X = F), based on the pKa of 6.3 for the phosphoryl group of UMP in H2O at I = 0.1,2 and the 0.5 unit higher pKa of phosphate monoanion in D2O than in H2O.38,39</p><p>Scheme 6 shows our proposed kinetic model for the ScOMPDC-catalyzed deuterium exchange reactions of F-UMP and UMP that is consistent with the observed bell-shaped pD-rate profile for kex/Kd for F-UMP (Figure 4). In this model the ionization of the substrate phosphoryl monoanion to give the phosphodianion (KSD2) is required for binding and reaction of the substrate, and the deprotonation of an acidic residue at the enzyme (KE) is required to generate a Brønsted base for the proton abstraction from C-6 of the substrate, a necessary step for deuterium exchange. Further ionization of the substrate at N-3 (pKSD = 8.37) to give the F-UMP trianion results in a form of the substrate that is unable to bind and react, as evidenced by the downward break in the pD-rate profile at high pD (Figure 4). The values of kex/Kd for deuterium exchange into F-UMP were fit to eq 6, derived for Scheme 6, with the values of pKSD2 = 6.8 and pKSD = 8.37 to give pKE = 8.5 ± 0.2 for deprotonation of the essential residue at the free enzyme. The fit also gives (kex)max/Kd = 2300 ± 700 M−1 s−1 as the intrinsic second-order rate constant for the deuterium exchange reaction of the phosphodianion form of F-UMP (SD2−) catalyzed by the deprotonated enzyme (E). This intrinsic rate constant is ca. 6-fold larger than the largest experimental second-order rate constant for the deuterium exchange reaction, observed at pD 8.15 (Table 1).</p><p>Figure 5 shows the pD-rate profiles for the first-order rate constants kex (s−1) for the ScOMPDC-catalyzed deuterium exchange at C-6 of enzyme-bound F-UMP and UMP in D2O, constructed using the data in Tables 1 and 2. As required by the model in Scheme 6, these profiles exhibit a dependence on the ionization state of only the enzyme-substrate complex ED+•SD2−, and they are consistent with the deprotonation of an essential residue at this complex to generate a Brønsted base for proton abstraction from C-6 of the substrate (KES, Scheme 6). The values of kex for F-UMP and UMP were fit to eq 7, derived for Scheme 6, to give pKES = 7.1 ± 0.1 and 8.0 ± 0.1 for deprotonation of the essential residue at the ED+•F-UMP and ED+•UMP complexes, respectively. The data also give the intrinsic first-order rate constants (turnover numbers) for deuterium exchange into saturating F-UMP and UMP bound to the deprotonated enzyme as (kex)max = 0.041 s−1 and 1.2 × 10−5 s−1, respectively. Therefore, the addition of a 5-fluoro substituent results in a 3400-fold increase in the rate constant for deuterium exchange into saturating enzyme-bound UMP at the complex with the basic form of the enzyme (E•SD2−, Scheme 6).</p><!><p>The first X-ray crystal structures of OMPDC appeared in 2000 and they strongly implicated the protonated side chain of Lys-93 (numbering for the yeast enzyme) as the catalytic Brønsted acid needed to deliver the proton incorporated at C-6 during the decarboxylation of OMP to give UMP (Scheme 2).9–13 Figure 6, based on the X-ray crystal structure of ScOMPDC complexed with BMP,11 shows that Lys-93 at ScOMPDC is part of the strictly conserved DxKxxD "catalytic motif" that is found in all OMPDCs,40 the integrity of which is essential for robust decarboxylase activity.25,41 The hydrogen bond between the ammonium group of Lys-93 and the anionic oxygen at C-6 of BMP, an analog of the putative carbanion intermediate,36 is consistent with its proposed role as a Brønsted acid in the decarboxylation of OMP.9–12 This is in accord with the results of earlier mutagenesis studies of yeast OMPDC that showed that Lys-93 is essential for robust enzymatic activity. The inactive K93C mutant was rescued by covalent modification of the Cys-93 side chain by 2-bromoethylamine, which restores a primary amino group to the side chain at position 93.42 More recently, the OMPDCs from M. thermautotrophicus (MtOMPDC), Plasmodium falciparum (PfOMPDC) and the OMPDC domain of the bifunctional human UMP synthase (HsOMPDC)43 have been shown to undergo covalent modification by nucleophilic attack of the side chains of Lys-72, Lys-138 and Lys-314, respectively, at C-6 of bound 6-iodouridine 5′-monophosphate (6-I-UMP) and 6-azidouridine 5′-monophosphate (6-N3-UMP), resulting in displacement of the good leaving groups iodide and azide ions (Scheme 7).44–48 These lysine residues correspond to Lys-93 of the enzyme from yeast. The results of these structural and chemical studies strongly suggest that Lys-93 at the yeast enzyme functions not only to provide the proton incorporated at C-6 during the decarboxylation of OMP, but also as a Brønsted base in the reverse proton transfer "half reaction" that occurs as the first step of the C-6 deuterium exchange reactions of UMP and F-UMP (Scheme 2).</p><p>Figure 7 shows the X-ray crystal structures of wildtype HsOMPDC liganded by UMP and the D312N (D91N at ScOMPDC) mutant liganded by F-UMP. There is good superimposition of the side chains of the residues in the active sites of these enzymes and for simplicity Figure 7 depicts side chains of only the wildtype enzyme. Both UMP and F-UMP bind in the same manner observed for BMP at the yeast enzyme (Figure 6),11 with the pyrimidine ring in the syn conformation and C-6 directed towards Lys-314. The pyrimidine rings of UMP and F-UMP are "sandwiched" between the hydrophobic side chains of Pro-417 (Pro-202 at ScOMPDC) and Ile-318′ (Ile-97′ at ScOMPDC) of the second subunit. The fluorine of F-UMP projects into a hydrophobic pocket lined by the side chains of Ile-368, Met-371 and Ile-401 (Leu-150, Leu-153 and Ile-183 at ScOMPDC).46 The hydrophobic pocket thought to be occupied by the molecule of CO2 generated during the decarboxylation of OMP lies on the opposite side of the pyrimidine ring from the essential lysine, and is lined by the side chains of Phe-310, Ile-401 and Ile-448 (Phe-89, Ile-183 and Ile-232 at ScOMPDC).46 This pocket is occupied by a molecule of water in HsOMPDC liganded with UMP (Figure 7) and the same "conserved' water molecule is observed at OMPDCs liganded with BMP.49 Lys-314 is positioned close to C-6 of UMP and F-UMP with an Nζ-C6 distance of 3.06 Å in the UMP liganded structure. We therefore propose that the catalytic Brønsted base for the observed deuterium exchange reactions of F-UMP and UMP catalyzed by ScOMPDC is the deprotonated form of Lys-93.</p><p>Figure 8 shows the X-ray crystal structures of wildtype HsOMPDC liganded by UMP and the D312N mutant liganded by the substrate OMP. UMP and OMP bind in the same manner but the carboxylate group at C-6 of OMP is distorted out of the plane of the pyrimidine ring by 36 degrees,46,50 and there is no bound water molecule observed in the hydrophobic pocket. Similar out of plane distortions have been observed for the cyano group of 6-cyanouridine 5′-monophosphate and the functional groups at C-6 of other substituted pyrimidine nucleotides bound to OMPDC.47,48,51 Nζ of Lys-314 lies within hydrogen bonding distance of the carboxylate group of OMP with an Nζ-O7 distance of 2.54 Å. The close correspondence of these structures supports the proposal that the protonated side chain of Lys-93 at the yeast enzyme acts as a catalytic acid in the decarboxylation of OMP and that its conjugate base is the catalytic base in the C-6 deuterium exchange reactions of UMP and F-UMP.</p><p>The pH-rate profile for kcat/Km for decarboxylation of OMP by ScOMPDC in H2O is consistent with the requirement that a residue with a pKa of 7.7 be in the acidic form, and this pKa can be assigned to the side chain of Lys-93.2,26 The pKas of amine bases are ca. 0.6 unit higher in D2O than in H2O38 so that the catalytic acid is expected to have a pKa of around 8.3 in D2O, which is very similar to pKE = 8.5 for deprotonation of the essential group at the free enzyme determined from the pD-rate profile for kex/Kd for the deuterium exchange reaction of F-UMP in D2O (Figure 4). The close identity of the pKas at the free enzyme of the essential acid for decarboxylation of OMP and of the conjugate acid of the essential base for the deuterium exchange reaction of F-UMP supports our proposal that this base is the deprotonated form of Lys-93. Table 3 summarizes the pKas in D2O of the essential catalytic side chain at free ScOMPDC (pKE) and those at the enzyme liganded by OMP, UMP and F-UMP (pKES), determined from the pL-rate profiles for decarboxylation of OMP2 and for the C-6 deuterium exchange reactions of UMP and F-UMP. The value of pKE = 8.5 in D2O for the side chain of Lys-93 at free ScOMPDC is ca. 2.5 units lower than the expected intrinsic pKa of 11.0 for this residue in D2O.52 This depressed pKa is intriguing because the protonated side chain of Lys-93 appears to be involved in hydrogen bonds with the carboxylate side chains of both Asp-91 and Asp-96′ of the second subunit in the DxKxxD motif (Figure 6). The relatively low pKa of the ammonium group of Lys-93 has been noted previously,3,42 but the structural data offer no strong clues as to its origin. However, we note here that the 2.3 unit variation in the pKa of the ammonium group of Lys-93 at ScOMPDC complexed with OMP, UMP and F-UMP (Table 3) shows that the pKa of this residue is highly sensitive to its local environment. The downward perturbation in the pKa of Lys-93 may result in part from its proximity to the array of hydrophobic residues that surrounds the pyrimidine base (Figure 6), which should result in destabilization of the cationic protonated form.</p><!><p>The pD-rate profiles for kex for the deuterium exchange reactions of UMP and F-UMP (Figure 5) show that the pKa of the conjugate acid of the essential base is lowered from pKE = 8.5 for the free enzyme to pKES = 8.0 and 7.1 for the E•UMP and E•F-UMP complexes, respectively (Table 3). These results are consistent with a neutral rather than an anionic base because the presence of the hydrophobic pyrimidine ring at the liganded enzyme would be expected to destabilize a neighboring negative charge and result in an increase rather than a decrease in the basicity of the catalytic base. The relatively large 1.4 unit decrease in the pKa of the side chain of Lys-93 upon the binding of F-UMP (Table 3) is consistent with the destabilization of positive charge at Nζ of protonated Lys-93 by the combined effects of the electron-withdrawing 5-fluoro substituent and the hydrophobic pyrimidine ring of F-UMP (Figure 7). Scheme 6 includes the thermodynamic cycle that relates the binding of F-UMP and UMP to the protonated (Kd′) and deprotonated enzymes (Kd) to the pKas of the essential group at the free (pKE) and liganded (pKES) enzymes. The relationship Kd′/Kd = KES/KE and the difference (pKE - pKES) = 1.4 gives Kd′/Kd = 25 as the ratio of the dissociation constants for F-UMP from the protonated and the deprotonated enzyme, respectively. Therefore, F-UMP binds 25-fold more tightly to the deprotonated form of ScOMPDC, E, than it does to the form of the enzyme that dominates at neutral pD, ED+. Similarly, the difference (pKE - pKES) = 0.5 for UMP (Table 3) shows that UMP binds 3-fold more tightly to E than to ED+.</p><p>The intrinsic second-order rate constant for ScOMPDC-catalyzed deuterium exchange into F-UMP, (kex)max/Kd = 2300 M−1 s−1, can be combined with the intrinsic first-order rate constant (kex)max = 0.041 s−1 to give Kd = 20 μM as the intrinsic affinity of the deprotonated enzyme for the phosphodianion form of F-UMP in D2O. This intrinsic affinity is 6-fold greater than the largest experimental affinity observed at pD 7.6 – 8.2 (Table 1). The observed affinities, (Kd)obsd, of ScOMPDC for 2′-deoxyuridine 5′-monophosphate and 2′-deoxy-5-fluorouridine 5′-monophosphate of 170 μM and 250 μM, respectively, at pH 7.2 in H2O show that the addition of a 5-fluoro substituent results in only a small ca. 1.5-fold change in the observed affinity of ScOMPDC for the 2′-deoxypyyrimidine nucleotide in H2O at pH 7.2.2 These values of (Kd)obsd at pH 7.2 can be combined with eq 8, derived for Scheme 6 (in H2O), to give relative intrinsic affinities Kd of the deprotonated form of ScOMPDC for the phosphodianion forms of UMP and F-UMP of 1.5:1. This suggests that the intrinsic affinity of ScOMPDC for UMP in D2O is essentially the same as Kd = 20 μM determined for F-UMP. Therefore, we conclude that the larger decrease in the affinity of the enzyme for F-UMP than for UMP upon protonation of the catalytic base results largely from a specific destabilizing interaction of the electron-withdrawing fluorine at C-5 of bound F-UMP with the protonated enzyme.</p><p>By contrast with F-UMP and UMP, the binding of OMP to ScOMPDC results in a ca. 1.1 unit increase in the pKa of 8.3 in D2O for Lys-93 at the free enzyme, as evidenced by the downward break at pH 8.8 in the pH-rate profile for kcat for decarboxylation of OMP,2 which corresponds to a pKa of 9.4 in D2O (Table 3). This increase in basicity of the side chain of Lys-93 upon the binding of OMP is consistent with electrostatic stabilization of the positive charge at Nζ by its proximity to the negative charge of the carboxylate group at C-6 of bound OMP (Figure 8).</p><!><p>We reported earlier that the decarboxylation of OMP catalyzed by wildtype and several mutant ScOMPDCs in a mixed solvent of 50:50 (v/v) H2O/D2O results in the formation of an equimolar mixture of the products h-UMP and d-UMP labeled with H and D at C-6, respectively, so that the product deuterium isotope effect (PIE) on decarboxylation is unity (Scheme 2).20,21 Furthermore, the values of ΦNL3+ ≈ 1.0 for H/D fractionation between L2O and R-NL3+ show that the primary kinetic solvent isotope effect (KIE) is essentially equal to the PIE.21,53 Therefore, there is no discrimination between H and D at the proton donor in the transition state for the rate-determining step of the decarboxylation of OMP. The absence of any significant KIE shows that the rate-limiting step for enzyme-catalyzed decarboxylation of OMP is strongly uncoupled from the proton transfer step. It eliminates the possibility of a concerted mechanism for decarboxylation of OMP9,13 in which proton transfer from Lys-93 to C-6 provides electrophilic push to the loss of CO2. We concluded that the OMPDC-catalyzed decarboxylation of OMP proceeds by the stepwise mechanism through the UMP vinyl carbanion intermediate shown in Scheme 8, in which the transition state for the loss of CO2 from OMP to give the bound UMP carbanion is insensitive to the isotopic composition of the adjacent ζ-NL3+ group of Lys-93.20,21 The PIE of unity requires that hydron transfer to the bound vinyl carbanion be much faster than the possible rotation of the terminal CH2-NL3+ bond of the side chain of the proton donor Lys-93, which exchanges the positions of its NL3+ hydrons (kp ≫ krot). Therefore, the yields of h-UMP and d-UMP from the decarboxylation of OMP in an H2O/D2O mixture are determined by the initial isotopic enrichment of the ζ-NL3+ group of Lys-93 (governed by the fractionation factor ΦNL3+ ≈ 1) at the E•OMP Michaelis complex.20,21 The observation of a small inverse PIE of 0.93 for the decarboxylation of F-OMP catalyzed by wildtype ScOMPDC shows that the expected decrease in the chemical reactivity of the carbanion intermediate as a result of the 5-fluoro substituent does not result in any large change in the inequality kp ≫ krot.21 This small inverse PIE is consistent with a KIE of unity for hydron transfer to the enzyme-bound F-UMP vinyl carbanion intermediate, along with a small increase in the fractionation factor for the ζ-NL3+ group of Lys-93 at the E•F-OMP Michaelis complex to ΦNL3+ ≈ 1.1.21</p><p>The ScOMPDC-catalyzed deuterium exchange reactions of UMP and F-UMP examined here show that deuterium exchange into saturating enzyme-bound F-UMP [(kex)max = 0.041 s−1] is 3400-fold faster than exchange into enzyme-bound UMP [(kex)max = 1.2 × 10−5 s−1] so that there is a large 4.8 kcal/mol stabilization of the transition state for deuterium exchange by the electron-withdrawing 5-fluoro substituent at F-UMP. This is consistent with the formation of a carbanion intermediate at which the negative charge is stabilized by interaction with the electron-withdrawing 5-fluoro substituent, and a deuterium exchange reaction that is the reverse of the proton transfer "half reaction" involved in the physiological decarboxylation of OMP.</p><p>Scheme 9 shows our proposed mechanism for the OMPDC-catalyzed deuterium exchange reactions of UMP and F-UMP.22 Proton transfer from C-6 of UMP and F-UMP to the neutral side chain of Lys-93 (k-p) results in an enzyme-bound vinyl carbanion intermediate that can be rapidly reprotonated via "internal return" of the abstracted hydron to regenerate the unlabeled nucleotide (kp ≫ krot). This reversible chemical step is followed by the rare rate-limiting rotation of the terminal CH2-ND2H+ bond of the side chain of Lys-93 in to a "reactive position" for delivery of a deuteron to C-6 (krot).</p><p>The rate constant for CH2-NL3+ bond rotation in water is expected to be similar to the value of kreorg = 1011 s−1 estimated for reorganization that exchanges the relative positions of -O and -S at a thiobenzoate•carbocation ion pair.54 However, the side chain of protonated Lys-93 is "anchored" by hydrogen bonding to the carboxylate groups of Asp-91 and Asp-96′ in a "double salt bridge" (Figure 6), which is expected to result in a substantial barrier to rotation of the terminal CH2-NL3+ bond of this side chain at the ScOMPDC•carbanion complex. The strength of the stabilizing interactions in salt bridges buried in the interior of proteins is 3 –5 kcal/mol55,56 so that we estimate this barrier, which should involve the breaking of hydrogen bonds between protonated Lys-93 and both Asp-91 and Asp-96′, to be at least 8 kcal/mol. This gives an upper limit on the rate constant for bond rotation of krot ≤ 107 s−1.</p><!><p>Prior to the current work, a critical unknown was the effect of a 5′-phosphoribosyl group at N-1 and/or a 5-fluoro substituent on the C-6 carbon acidity of uracil derivatives in aqueous solution (Scheme 10). Therefore, we have determined second-order rate constants kDO (M−1 s−1) for the DO−-catalyzed deuterium exchange reactions of uridine, 5-fluorouridine and 1,3-dimethyl-5-fluorouracil in D2O at a variety of temperatures and pD (see Results and Supporting Information). The exchange reactions of uridine and 5-fluorouridine in the presence of 1 M NaOD provided kDO (M−1 s−1) for deuterium exchange into their N-3 anions, while the exchange reaction of 5-fluorouridine at pD 7.1 provided kDO for deuterium exchange into the neutral nucleoside (pKND = 8.4, this work). The value of kDO for deuterium exchange into neutral uridine was calculated from kDO determined for its N-3 anion, with the assumption that ionization of uridine at N-3 results in the same 2.5 × 104-fold decrease in reactivity of the C-6 proton that is observed upon ionization of 5-fluorouridine at N-3 (Table S8 of the Supporting Information).</p><p>The values of kDO for the substituted uracils (Table S8 of the Supporting Information) were used to calculate the second-order rate constants kHO (M−1 s−1) for deprotonation of these carbon acids by hydroxide ion in H2O given in Table 4, using a secondary solvent isotope effect of kDO/kHO = 2.4.57–59 The carbon acid pKas were then calculated using eq 9, derived for Scheme 11, with pKw = 14 for H2O and kHOH = 1011 s−1 for the reverse protonation of localized unstable carbanions by solvent water that is limited by the rotation of a molecule of water into a reactive position.57–60</p><p>Table 4 summarizes the C-6 carbon acidities, pKCH (Scheme 10), of the model nucleosides uridine and 5-fluorouridine in water, along with those of their N-3 anions and the corresponding 1,3-dimethyluracils. The value of pKCH = 30.7 for the C-6 proton of 1,3-dimethyluracil (Scheme 10, X = H, R1 = R2 = Me), obtained from rate constants for its competing exchange and hydrolysis reactions in 1 M NaOD at 25 °C,61 is lower than the earlier reported value of 34 ± 2 determined from the temperature dependence of deuterium exchange in acetate buffers at 175 – 217 °C.62 There is a good linear correlation (not shown) with a slope of ρI = −8.0 of the carbon acid pKas of 3-R-4-methylthiazolium cations (Scheme 12),63 "normal" carbon acids that undergo ionization to generate localized carbanions,59,64,65 with the inductive substituent constant σI for the 3-substituent, and, by analogy, a similar free energy relationship is expected for the carbon acid pKas of N-1 substituted uracils. The value of σI = −0.05 for the methyl group66 and the estimated value of σI = 0.14 for the ribosyl group67 can be combined with ρI = −8.0 to estimate a decrease in C-6 carbon acidity of 1.5 pK units upon the change from a methyl to a ribosyl group at N-1 of substituted uracils (R1, Scheme 12). This is consistent with the 1.6 unit difference in the values of pKCH for 1,3-dimethyl-5-fluorouracil and 5-fluorouridine determined here (Table 4).</p><p>The data in Table 4 show that the addition of a 5-fluoro substituent to substituted uracils results in an increase in the C-6 carbon acidity of 3.6 – 4.0 pK units. By contrast, the addition of a 5-fluoro substituent results in decreases of 1.5, 1.6 and 1.7 – 1.9 units in the pKa of the N-3 proton of uracil,37 2′-deoxyuridine,37 and UMP,68 respectively. These substituent effects are consistent with a fall-off factor of ca. 2 for the presence of the additional carbon atom between the 5-fluoro substituent and the center of negative charge at substituted uracil N-3 anions compared with their C-6 carbanions.57</p><!><p>Figure 9 shows free energy profiles for the enzyme-catalyzed deuterium exchange reactions of UMP and F-UMP through the respective C-6 vinyl carbanion intermediates. These profiles were constructed according to the mechanism shown in Scheme 9, with the observed 3400-fold faster deuterium exchange into enzyme-bound F-UMP [(kex)max = 0.041 s−1] than UMP [(kex)max = 1.2 × 10−5 s−1] at high pD (Figure 5), and a barrier to rotation of the terminal CH2-NL3+ bond of the side chain of Lys-93 of 8 kcal/mol (krot = 107 s−1, see above). These profiles refer to the exchange reactions of the substrate phosphodianions (SD2−) catalyzed by ScOMPDC at which Lys-93 is deprotonated (END2). The essentially identical values of Kd = 20 μM for binding of F-UMP and UMP to the deprotonated enzyme correspond to a free energy of binding to form the Michaelis complex of 6.4 kcal/mol. Proton abstraction from C-6 (k-p) leads to the formation of the initial enzyme-carbanion complex which can undergo rapid internal return by reprotonation (kp) or the occasional, rate-limiting, rotation of the side-chain of Lys-93 into a position to deliver a deuteron to C-6 (krot), with an estimated barrier to rotation of at least 8 kcal/mol (see above). Once this rotational event occurs there is rapid and irreversible transfer of a deuteron to C-6 to generate the deuterium exchange product, kp ≫ krot, as evidenced by the PIEs of unity determined for the decarboxylation of OMP in 50:50 (v/v) H2O/D2O.20,21</p><p>The formation of vinyl carbanions derived from UMP and F-UMP as intermediates of their C-6 deuterium exchange reactions (Scheme 9 and Figure 9) allows equilibrium constants (Keq)enz for proton transfer from C-6 of enzyme-bound UMP and F-UMP to the side chain of Lys-93 (Scheme 13A) to be obtained from the first-order rate constants (kex)max for deuterium exchange into the enzyme-bound nucleotides, according to eqs 10 and 11. The values of (kex)max = 1.2 × 10−5 and 0.041 s−1 for deuterium exchange into enzyme-bound UMP and F-UMP, respectively, and krot = 107 s−1 were substituted into eq 11 to give (dimensionless) equilibrium constants for proton transfer from C-6 of the bound nucleotide to ScOMPDC of (Keq)enz = 1.2 × 10−12 and 4.1 × 10−9, respectively, so that there is a large 4.8 kcal/mol stabilizing effect of the 5-F substituent on this proton transfer equilibrium (Figure 9). The values of pKES = 8.0 and 7.1 for the side chain of Lys-93 at ScOMPDC complexed with UMP and F-UMP, respectively (Table 3), show that there is an offsetting 1.2 kcal/mol destabilizing effect of the 5-fluoro substituent on equilibrium proton transfer to Nζ of Lys-93, as a result of interactions of the electron-withdrawing fluorine at C-5 of bound F-UMP with the positively charged ammonium group (vide infra). Therefore, the 5-fluoro substituent at F-UMP results in an overall decrease in the thermodynamic barrier for proton transfer from C-6 of 6.0 kcal/mol, which corresponds to a 4.4 unit increase in the carbon acidity of the bound nucleotide. This is similar to the 3.7 pK unit difference in the C-6 carbon acidities of the model nucleosides uridine (pKCH = 28.8 ± 1) and 5-fluorouridine (pKCH = 25.1 ± 0.5) in aqueous solution (Table 4).</p><p>Similarly, overall equilibrium constants for proton transfer from C-6 of UMP and F-UMP to a primary amine base of pKRNH3 = 7 in water (Scheme 13B) can be calculated using eq 12, with the values of pKCH for the model nucleosides uridine and 5-fluorouridine from Table 4, as (= 2 × 10−22 and 8 × 10−19, respectively. Therefore, the binding of UMP or F-UMP to Keq)aq ScOMPDC results in a ~5 × 109-fold increase in the equilibrium constant for proton transfer from C-6 to an amine base, which corresponds to a thermodynamic stabilization of the enzyme-bound UMP and F-UMP vinyl carbanions, relative to the enzyme-nucleotide complexes, of 13 kcal/mol. A critical unknown is the magnitude of the barrier to rotation of the terminal CH2-NL3+ bond of the side chain of Lys-93, which is hydrogen-bonded to the anionic side chains of Asp-91 and Asp-96′ at the enzyme•carbanion complex (Figure 6). We have estimated this barrier as ≥ 8 kcal/mol, so that 13 kcal/mol is a lower limit on the stabilization of the bound UMP and F-UMP vinyl carbanions resulting from their interactions with ScOMPDC.</p><!><p>The formation of vinyl carbanions as intermediates in the C-6 deuterium exchange reactions of UMP and F-UMP catalyzed by ScOMPDC provides convincing evidence for the decarboxylation of OMP and F-OMP by a stepwise mechanism through the same carbanions. The total rate acceleration for the decarboxylation of OMP by free ScOMPDC can be calculated as 4 × 1022 M−1, from the ratio of kcat/Km = 1.1 × 107 M−1 s−1 for decarboxylation at pH 7.1,7 which is the maximum in the pH-rate profile for decarboxylation,2 and ko = 2.8 × 10−16 s−1 for the nonenzymatic decarboxylation of OMP in aqueous solution.4 This corresponds to a total stabilization of the transition state for decarboxylation of OMP of 31 kcal/mol.3 It is more difficult to establish the rate acceleration for the enzyme-catalyzed deuterium exchange reactions of UMP and F-UMP, because the nonenzymatic proton transfer reactions of these normal carbon acids in solution proceed by proton transfer to lyoxide ion so that the observed rate constant depends strongly on pL. This notwithstanding, the observed rate constant for C-6 deprotonation of 5-fluorouridine (a model for F-UMP) by hydroxide ion in water at pH 7.1 can be calculated as ko = kHO[HO−] = 9 × 10−8 s−1 (Table 4). The intrinsic second-order rate constant for ScOMPDC-catalyzed deuterium exchange into F-UMP catalyzed by the deprotonated enzyme is (kex)max/Kd = 2300 M−1 s−1. The ratio of these rate constants gives the enzymatic rate acceleration for deuterium exchange as 3 × 1010 M−1, which corresponds to a transition state stabilization of 14 kcal/mol. Together with the ≥ 13 kcal/mol stabilization of the bound UMP and F-UMP vinyl carbanions by interactions with the enzyme estimated above, this shows that a large portion of the enzymatic rate acceleration for the decarboxylation of OMP results from stabilization of the bound carbanion intermediate by its interactions with ScOMPDC.</p><p>These data suggest that the total transition state stabilization for decarboxylation of OMP via the UMP vinyl carbanion intermediate exceeds that for its formation by proton transfer from C-6 of UMP to the enzyme by ca. 17 kcal/mol. We suggest several factors that may contribute to this apparent greater proficiency of ScOMPDC for decarboxylation than deuterium exchange.</p><p>(1) The difference in the transition state stabilization for decarboxylation compared with deuterium exchange would be lowered to 14 kcal/mol if there is uncertainty in the rate constant for the nonenzymatic decarboxylation of OMP such that this reaction is ca. 100-fold faster than the current best estimate.4</p><p>(2) The decarboxylation of OMP by ScOMPDC is limited in part by the barrier to the chemical step of the loss of CO2 to give the carbanion intermediate.2,26,33,69 By contrast, the rate-limiting step for enzyme-catalyzed deuterium exchange is the physical step of rotation of the terminal CH2-NHD2+ bond of the side chain of Lys-93 into a "reactive position", with an estimated barrier of ≥ 8 kcal/mol (krot, Figure 9). Therefore, the energy of the transition state for the chemical proton transfer step (k-p) will be lower than that for the overall deuterium exchange reaction by an amount RTln(kp/krot) so that the rate acceleration for deuterium exchange underestimates the rate acceleration for formation of the bound vinyl carbanion from UMP by this amount.</p><p>(3) The initial product of the decarboxylation of OMP is the ternary complex of OMPDC with UMP and a molecule of CO2 bound in the hydrophobic pocket below C-6 of UMP (Figure 7). If the presence of CO2 in the hydrophobic pocket activates the enzyme for catalysis of proton transfer, then the true reactivity of the enzyme towards proton transfer from C-6 of bound UMP and F-UMP could be higher than that observed in the deuterium exchange reactions described here.</p><p>(4) The transition state for decarboxylation of OMP may be stabilized by favorable "solvation" of the departing CO2 molecule by the development of stabilizing interactions with the hydrophobic side chains of Phe-89, Ile-183 and Ile-232 that line the CO2 binding pocket at yeast OMPDC (Phe-310, Ile-401 and Ile-448 at HsOMPDC, Figure 7). These interactions, which are not present in the ground state of the free enzyme and OMP, would result in selective stabilization of the transition state for formation of the UMP carbanion intermediate by decarboxylation that is not available for the transition state for its formation by proton transfer from C-6 of UMP.</p><p>Finally, we note that ground state destabilization by the introduction of "strain" at the protein and/or electrostatic "stress" at the substrate in the OMPDC•OMP Michaelis complex may contribute to an increase in kcat for the decarboxylation of OMP, and, possibly, proton transfer from C-6 of UMP and F-UMP.12,17,70 However, the overall rate accelerations for decarboxylation and exchange calculated here from the rate constants kcat/Km are unaffected by such effects because they refer to the barrier for conversion of the free enzyme and substrate in solution to the enzyme-bound transition state.71</p>

PubMed Author Manuscript

Bioinspired nonheme iron complex that triggers mitochondrial apoptotic signalling pathway specifically for colorectal cancer cells

The activation of dioxygen is the keystone of all forms of aerobic life. Many biological functions rely on the redox versatility of metal ions to perform reductive activation-mediated processes entailing dioxygen and its partially reduced species including superoxide, hydrogen peroxide, and hydroxyl radicals, also known as reactive oxygen species (ROS). In biomimetic chemistry, a number of synthetic approaches have sought to design, synthesize and characterize reactive intermediates such as the metal-superoxo, -peroxo, and -oxo species, which are commonly found as key intermediates in the enzymatic catalytic cycle.However, the use of these designed complexes and their corresponding intermediates as potential candidates for cancer therapeutics has scarcely been endeavored. In this context, a series of biomimetic first-row transition metal complexes bearing a picolylamine-based water-soluble ligand, [M(HN 3 O 2 )] 2+ (M ¼ Mn 2+ , Fe 2+ , Co 2+ , Cu 2+ ; HN 3 O 2 ¼ 2-(2-(bis(pyridin-2-ylmethyl)amino)ethoxy)ethanol) were synthesized and characterized by various spectroscopic methods including X-ray crystallography and their dioxygen and ROS activation reactivity were evaluated in situ and in vitro. It turned out that among these metal complexes, the iron complex, [Fe(HN 3 O 2 )(H 2 O)] 2+ , was capable of activating dioxygen and hydrogen peroxide and produced the ROS species (e.g., hydroxyl radical). Upon the incubation of these complexes with different cancer cells, such as cervical, breast, and colorectal cancer cells (MDA-MB-231, AU565, SK-BR-3, HeLa S3, HT-29, and HCT116 cells), only the iron complex triggered cellular apoptosis specifically for colorectal cancer cells; the other metal complexes show negligible anti-proliferative activity. More importantly, the biomimetic complexes were harmless to normal cells and produced less ROS therein. The use of immunocytochemistry combined with western blot analysis strongly supported that apoptosis occurred via the intrinsic mitochondrial pathway; in the intracellular network, [Fe(HN 3 O 2 )(H 2 O)] 2+ resulted in (i) the activation and/or production of ROS species, (ii) the induction of intracellular impaired redox balance, and (iii) the promotion of the mitochondrial apoptotic signaling pathway in colorectal cancer cells. The results have implications for developing novel biomimetic complexes in cancer treatments and for designing potent candidates with cancer-specific antitumor activity.

bioinspired_nonheme_iron_complex_that_triggers_mitochondrial_apoptotic_signalling_pathway_specifical

Introduction<!>Synthesis and characterization of rst-row transition metal complexes<!>Reactivity of 1 toward dioxygen and hydrogen peroxide<!>Cell viability and intracellular ROS detection<!>Immunocytochemistry, qRT-PCR, and western blot analysis<!>In vivo studies<!>Conclusions<!>Materials<!>Instrumentation<!>Cell culture<!>Cell viability<!>Statistical analysis<!>Western blot analysis<!>Immunocytochemistry<!>Quantitative reverse transcription-polymerase chain reaction (qRT-PCR)<!>In vivo experiments

<p>Cancer is one of the most harmful and serious heterogeneous diseases that represent abnormal cellular energy metabolism and remains one of the major causes of death in most developing and developed countries. 1 Although extensive labours have been devoted to discovering targeted therapies, it is still challenging to overcome poor prognoses and high mortality. Thus, the pursuit of new chemical tools dealing with biomedical functions of metal complexes has garnered tremendous interest due to their broad pharmaceutical properties as potent anticancer agents. For instance, cisplatin, one of the most clinically successful examples of platinum-based anticancer drugs, opened a new era for the development of anticancer transition metal compounds by covering about 50% of all cancer treatments to date. [2][3][4] Despite the outstanding applicability of Pt-based drugs, they suffer from low stability under physiological conditions resulting in copious toxic side effects such as necrosis, tissue injury, nausea, vomiting, and neurotoxicity. [5][6][7][8] Ever since the pioneering work of the Au(I)-NHC complex was reported by Bernes-Price and co-workers in 2004, 9 recent developments in the design of non-Pt-based metal Nheterocyclic carbene (NHC) complexes, such as those of Ru, Au, Ir, and Pd, have succeeded in the signicant improvement of their stability due to the strong donating ability of the NHC ligand. [10][11][12][13][14][15] Still, these metals are non-existing elements for the human body and may be the source of unexpected side effects along with the development of drug resistance. [16][17][18][19] Therefore, the use of rst-row transition metals such as Mn, Fe, Co, and Cu would be a credible alternative route for next generation anticancer agents with low general toxicity because they are biorelevant trace elements. In particular, iron, a redox-active essential element involved in oxygen transport in mammals and electron transport in iron-sulfur proteins, [20][21][22][23][24] would be expected to bypass the cytotoxicity concerns and function both as an electron donor and acceptor in order to disturb the redox homeostasis of the reactive oxygen species (ROS) level in tumors. 25 However, the anticancer activity of rst-row transition metal-based drugs have mostly been examined by introducing or vectorizing metal chelators as potential chemotherapeutics; [26][27][28][29][30][31] their in situ mechanism of action and reactivity were scarcely scrutinized due to the intrinsic instability of reactive metal-oxygen intermediates.</p><p>Recent investigations have highlighted the importance of redox balance and the deregulation of redox signalling in cancer cells due to the raised levels of ROS from multiple intracellular factors such as increased metabolic activity, mitochondrial dysfunction, and peroxisome activity. [32][33][34][35] On the contrary, ROS, which are by-products of normal cellular activity mainly generated in mitochondria and membrane-bound NADHP oxidase, 36,37 are persistently produced in a highly controlled manner in normal cells because the canonical production of ROS is required for the signalling processes of cell division, autophagy, and cellular proliferation. 38,39 Hence, a wealth of recent evidences have underlined ROS as a double-edged sword in cancer cells (e.g., a cancer-stimulating or a cancersuppressing agent). Such dichotomic role in cancer is also relevant to nitric oxide. 40,41 Nevertheless, aiming ROS regulation for the clinical treatment of cancer represents a valuable challenge to advance cancer therapeutic approaches. In this context, we reasoned that the use of biomimetic metal complexes would (i) advance our understanding of the molecular basis of their mechanism of action, (ii) offer an elegant way to control intracellular ROS dysregulation, and (iii) provide a wealth of repertoire of metallodrugs having a broader spectrum of cancer-specic antitumor activity with a prudent choice of ligands.</p><p>In the course of our investigation, we employed biomimetic metal complexes having picolylamine-based water-soluble ligand, which accommodates rst-row transition metals such as manganese, iron, cobalt, and copper. Their reactivity toward dioxygen and ROS was examined and then their anticancer activity was directly evaluated upon their incubation with cancer cells. It was found that the iron complex that was capable of activating both dioxygen and ROS triggered an efficient mitochondrial apoptotic signalling pathway of colorectal cancer cells by producing hydroxyl radicals and provoking the intrinsic mitochondrial apoptotic pathway (Scheme 1).</p><!><p>First-row transition metal complexes bearing a N,N-bis(2picolyl)amine backbone with a pendant ethoxyethanol side chain (HN 3 O 2 ), [M(HN 3 O 2 )] 2+ (M ¼ Mn, Fe, Co, and Cu) were synthesized according to the modied literature procedure (see ESI †). 42 The X-ray crystal structures of manganese and iron complexes, [Mn(HN 3 O 2 )(Cl) 2 ] and [Fe(HN 3 O 2 )(H 2 O)] 2+ (1), were obtained when recrystallization was carried out under nitrogen atmosphere (Fig. 1a, b and Table S1 †). As depicted in Fig. 1 were stable in both organic and aqueous solution; they did not undergo demetallation, aggregation, and precipitation. Interestingly, only when 1 was recrystallized in a solvent mixture of CH 3 CN : ether (v/v 1 : 2) under aerobic conditions for several days, deep yellow single crystals due the formation of the m-oxobridged diferric complex, [Fe 2 (m-O)(HN 3 O 2 ) 2 ] 4+ (2), was obtained (Fig. 1c and Table S2 †). This suggested that among [M(HN 3 O 2 )] 2+ , only 1 was susceptible to conduct the dioxygen activation reaction depending on the reaction conditions (vide infra). 47,48 From the charge consideration, the iron ions in 2 were both in the +3 oxidation state. Each Fe ion was at the center of a distorted N 3 O 3 octahedron with one bridging oxo ligand forming a {Fe 2 (m-O)} core with a Fe-O-Fe angle of 165.08(6) (Table S2 †). The Fe/Fe distance (3.530(1) Å) and the Fe-O distance (1.7799(4) Å) were consistent with a m-oxo bridged diiron complex (Table S2 †). [49][50][51][52][53]</p><!><p>To provide more insight into the reactivity of complex 1, we scrutinized the reactivity of complex 1 toward dioxygen and hydrogen peroxide (H 2 O 2 ). The UV-vis spectrum of isolated 2 clearly exhibited an intense absorption band at 355 nm (3 ¼ 6000 M À1 cm À1 ) in CH 3 CN at 20 C, which differed from that of 1 (Fig. 2a). Interestingly, when 1 was treated with iodosylbenzene (PhIO) or H 2 O 2 , 2 was generated instantaneously (Fig. S4 †). The ESI MS spectrum of 2 clearly exhibited two prominent peaks at m/z of 350.1 and 799.1, whose mass and isotopic distribution patterns are in a good agreement with [Fe 2 (O)(N 3 O 2 ) 2 ] 2+ and [Fe 2 (O)(N 3 O 2 ) 2 (ClO 4 )] + , respectively (Fig. 2, inset and Fig. S5a †). The use of 18 O-labeled PhIO resulted in two mass unit shi of the peak at m/z from 799.1 to 801.1, indicating that the oxygen atom is originated from PhI 18 O (Fig. 2, inset).</p><p>On the other hand, the addition of 1-benzyl-1,4dihydronicotinamide (BNAH), known as an NADH analog, to an CH 3 CN-solution of 1, also afforded the formation of 2 in the presence of acid (e.g., perchloric acid or hydrochloric acid) under aerobic conditions (Fig. S6 †). 54 The ESI MS spectrum of 2 generated upon the use of 18 O 2 evenly showed a peak at m/z 801.1, demonstrating that the source of the bridging oxygen atom was dioxygen (Fig. S5b †). Therefore, 1 is capable of activating dioxygen using biological ingredients (e.g., proton and cofactors). The electrochemical property of 1 was examined by cyclic voltammetry. Despite an irreversible redox wave of 1, the reduction peak potential (E red ) value of À0.15 V vs. Fc/Fc + in CH 3 CN was found; when an identical experiment was performed in deionized water, the E red value was positively shied (e.g., 0.08 V vs. Fc/Fc + ) (Fig. S7 †). According to the wellestablished results in the literature, the E red value below $À0.1 V vs. Fc/Fc + is dened to be a prerequisite value for dioxygen activation by nonheme iron(II) complexes. 47,55 Hence, 1 activates dioxygen with the help of a cofactor such as the NADH analog and proton under aerobic conditions or activates H 2 O 2 (Fig. S4 and S6 †); this results in the formation of the m-oxobridged dinuclear iron complex.</p><p>Furthermore, we attempted to detect the generation of ROS species such as hydroxyl radical ($OH) due to 1 using uorescence probe (e.g., terephthalic acid (TA) assay). 56 Upon incubation of 1 with non-uorescent TA under aerobic conditions for 24 h, a brilliant uorescence at 395 nm was detected by uorescence spectrophotometry (l ex ¼ 310 nm), indicating the hydroxylated product, 2-hydroxyterephthalic acid (TA-OH) was formed upon 24 h incubation (Fig. 2b). Indeed, over 50% generation of TA-OH in the reaction between TA and 1 can be achieved within 1 h under identical reaction conditions (Fig. S8 †). Since TA, 1, and 2 did not show any uorescence at 310 nm excitation wavelength, the appearance of the uorescence intensity at 395 nm due to TA-OH solely comes from the reaction between 1 and TA under aerobic conditions. Interestingly, the incubation of other metal complexes or ligands only with TA did not afford such strong uorescence (Fig. 2b). Therefore, the present result conrmed the efficacy of 1 for the in situ production of $OH under aerobic conditions.</p><p>By virtue of well-established dioxygen activation mechanism, 47,55, 57 1 could react with dioxygen to form a putative ironsuperoxo species (Scheme 2, pathway a), which might further be converted to a m-peroxo-bridged diferric species (Scheme 2, pathway b). Subsequent homolytic O-O bond cleavage results in the generation of iron-oxo species (Scheme 2, pathway c), followed by a comproportionation reaction with 1 to produce 2 (Scheme 2, pathway d). The addition of H 2 O 2 might facilitate the formation of the iron-oxo species and produce the cytotoxic hydroxyl radical species as seen in the Fenton-like reaction (Scheme 2, pathway e and f). 58</p><!><p>Encouraged by the abovementioned dioxygen and ROS activation by the iron complex, we surmised that 1 could promote the deregulation of the intracellular ROS level and could induce cancer cell cycle arrest and cell death. Indeed, it has also been shown that the m-oxo-bridged diiron(III) complexes could exhibit tumor-specic anticancer activity by generating the toxic hydroxyl radical. [59][60][61] The potential anti-proliferative activities of the above-mentioned biomimetic metal complexes against breast cancer cells (e.g., MDA-MB-231, AU565, and SK-BR-3), cervical cancer cell (e.g., HeLa S3), and colorectal cancer cells (e.g., HT-29 and HCT116), were examined by the WST-8 assay</p><p>monosodium salt), which relies on the mitochondrial activity (Fig. 3 and S9 †). Among biomimetic metal complexes, the iron complex, 1, only exhibits the anticancer activity toward the cancer cells, more specically, the colorectal cancer cell lines such as HT-29 and HCT116; 1 is signicantly effective to HCT116 than HT-29 (Fig. 3a and b). As shown in Fig. S10, † 1 clearly revealed enhanced cytotoxicity against HCT116 cells while other metal complexes did not. The half-maximal inhibitory concentration (IC 50 ) of 1 toward HCT116 cells is determined to be 24.5 mM (Fig. 3c). Such an IC 50 value is greater than that of other reported Pt-based anticancer agents; however, it is still largely below the millimolar range, thereby showing signicant cytotoxic activity against HCT116 cells. [62][63][64] This result led us to, notably, the time-dependent WST-8 viability, which showed that the anticancer activity of 1 was observed with a clear viability decay up to $50% within 24 h while other metal complexes did not exhibit any kinetics of cytotoxicity on the HCT116 cells (Fig. S11 †).</p><p>To show that 1 is not harmful to normal cells, we incubated 1 with CCD-986Sk cells, which are broblast cells for skin, under identical conditions. Very interestingly, all rst-row transition metal complexes including 1 did not show any perceivable cell death (Fig. 3d). Contrary to other abiological metal-based anticancer drugs including Pt, Pd, and Ru that are cytotoxic by nature, these biomimetic metal complexes are signicantly less toxic in the biological environment. These results strongly advocate that the use of the bio-relevant metal ions accommodating complexes might circumvent the side effects by better dealing with human physiological homeostasis. 65 The qualitative and quantitative detection of intracellularly generated ROS by 1 in HCT116 cells and CCD-986Sk cells were carried out under identical incubation conditions using the cell permeant reagent 2 0 ,7 0 -dichlorouorescin diacetate (DCFDA) (Fig. 4). Consistent with previously demonstrated results with Scheme 2 Proposed mechanism for the formation of the m-oxobridged diferric complex via the reductive activation of dioxygen or hydrogen peroxide. TA, only 1 effectively generated the ROS in HCT116 cells; the amount of intracellularly generated ROS was $1.8 fold greater than that observed in the negative control (NC) case (Fig. 4a). This result revealed that 1 efficiently penetrated into the HCT116 cells and produced intracellular ROS in the heterogeneous environment. Very importantly, the intracellular production of ROS by 1 in normal cells (i.e., CCD-986Sk) was determined to be $1.2 fold greater (Fig. 4b). These observations allowed us to speculate that the intracellular generation of ROS by 1 may be favored in heterogeneous environment (i.e., cancer cells) than in homogeneous environment (i.e., normal cells). It has been demonstrated that an increase in ROS was associated with cancer cell growth as compared to normal cells and there might exist a certain threshold of ROS concentration that is incompatible with cellular survival. 66 In the present case, we proposed that (i) 1 in HCT116 cells would activate intracellular ROS, (ii) provoke excessive levels of ROS that reaches a certain threshold, and (iii) nally exert cytotoxic effect. However, the less activation of ROS by 1 in normal cells would barely reach the threshold and the redox homeostasis would be maintained. Therefore, we concluded that once 1 was penetrated into the heterogeneous HCT116 cells, 1 increased the intracellular concentration of ROS in order to prompt the cell death by presumably affecting many regulatory signalling processes closely related to intracellular ROS homeostasis (vide infra).</p><!><p>To verify that 1 roused the signalling pathway of cell death in colorectal cancer cells, we rst performed immunouorescence staining with 4 0 ,6-diamidino-2-phenylindole (DAPI), which is frequently used to visualize the nuclear DNA in both living and xed cells, in HCT116 cells within 24 h aer treatment with 1 (Fig. 5a). The structural changes of the nuclei including nuclear condensation and shrinkage were noticed in HCT116 cells treated with 1; this result suggested the production of the apoptotic nuclei upon treatment with 1 (Fig. 5a, shown with a white arrow). We further examined the area and number of nuclei in HCT116 cells treated with 1 (Fig. 5b). A signicant decrease in the nuclei area as well as the average nuclei area per image in HCT116 cells incubated with 1 as compared to the control were conrmed. In addition, a decreased number of nuclei also suggested that 1 has a critical role in anticancer activity.</p><p>In order to understand whether the anticancer effect of 1 inuenced the apoptosis of HCT116 cells, we monitored the level of genes and proteins closely related to the apoptotic pathways. We rst checked the expression of the BCL-2 family members using qRT-PCR (Fig. 5c). It is well-documented that the BCL-2 family is considered to be divided into anti-apoptotic and pro-apoptotic members according to their functions. 67 Even though the expression of BCL-2 alpha and beta, known as antiapoptotic members, was slightly attenuated, the expression of BAX and BAK, known as pro-apoptotic members, was increased and prominent in HCT116 cells treated with 1. This clearly suggested that 1 induced apoptotic cell death.</p><p>Since up-regulated BAX and BAK commonly concerned the release of cytochrome c into the cytosol by increasing the permeability of the mitochondrial membrane, 68 we performed immunouorescence for COX IV to ascertain whether 1-triggered apoptosis was mediated by the mitochondria. COX IV is an inner membrane mitochondrial marker and their level is increased at the early stage of apoptosis and then decreased aer the release of cytochrome c into the cytosol during the apoptotic process. 69 As shown in Fig. 5d, COX IV was remarkably reduced in HCT116 cells treated with 1, whereas its level was abundant in control. These results supported that the apoptosis of HCT116 cells caused by 1 might be accompanied by a decrease in the mitochondrial function. Since it has been wellestablished that the decrease in the level of COX IV affected the activation of caspases, for instance, caspase 9 and 3, 69 we examined the level of the cleaved caspase 9 and 3 aer incubating HCT116 cells with 1 for 24 h. Western blot analysis undoubtedly showed that the level of the cleaved caspase 9 and 3 were remarkably increased in HCT116 cells (Fig. 5e). Moreover, an increased level of the cleaved poly(ADP-ribose) polymerase 1 (PARP1), which is a hallmark of apoptosis due to the caspases, 70 was noticeably perceived in HCT116 cells treated with 1 (Fig. 5e).</p><p>Next, to investigate whether 1-triggered apoptosis was mediated by the extrinsic pathway through FADD (fasassociated protein with death domain), immunouorescence for E-cadherin, which collaborates with death receptors, was conducted (Fig. S12 and S13 †). 71,72 We observed that there was no signicant change in the E-cadherin level aer treatment with 1. In addition, the expression of BID, which is proteolytically activated by FADD signalling, was not signicant between HCT116 cells treated with 1 and the controls (Fig. S14 †). These results supported that the apoptosis induced by 1 was presumably mediated by the intrinsic pathway including mitochondrial dysfunction but not by the extrinsic pathway. In addition, we performed the cell viability assay in the presence and absence of Z-VAD-FMK to evaluate the apoptotic effect of 1 on HCT116 cells. Z-VAD-FMK is a cell permeable caspase inhibitor that impedes caspase processing and apoptosis by irreversibly binding to the catalytic site. 73 1 induced cell death only in HCT116 and co-treatment with Z-VAD-FMK partly improved the viability level (Fig. 5f). CCD-986Sk is not affected by 1 and Z-VAD-FMK. These results showed that 1 was closely associated with the apoptosis-dependent pathway.</p><p>Taken together, we propose that aer cellular uptake, the intracellular generation of ROS by 1 prompted mitochondrial dysfunction and resulted in the apoptotic signalling pathway as follows: (i) the mitochondrial release of cytochrome c in cytosol, followed by (ii) the cleavage of caspase 9 and then that of caspase 3 occurred to nally (iii) stimulate the cleavage of PARP1. Thus, the intrinsic mitochondrial apoptosis pathway was activated to cause HCT116 cell cycle arrest upon treatment with 1.</p><!><p>Having established that 1 elevated the intracellular ROS concentration, we investigated the anticancer activity of 1 in vivo. To evaluate the anticancer effect of 1 in vivo, six mice models injected with HT29 cells were used; three were used as the vehicle with sterile distilled water and another three were treated with 1 intraperitoneally for 5 days (Fig. 6a). When the mice were treated with 5 mg kg À1 of 1, no adverse side effects were detected. Three cases of 1 resulted in approximately 50% to 80% tumor growth inhibition with respect to the control (Fig. 6b). Furthermore, the injection of 1 at 5 mg kg À1 signicantly inhibited the tumor growth of the xenogra relative to the control in the subcutaneous colorectal cancer model, indicating that 1 is a potential antitumor agent that operates in vivo (Fig. 6c). These results supported that 1 has the possibility of therapeutic target in vivo.</p><!><p>In conclusion, we successfully synthesized and characterized a series of biomimetic rst-row transition metal complexes with various spectroscopic methods including X-ray crystallography. Among these chemically designed complexes, only iron complex, 1, was capable of activating dioxygen and reactive oxygen species and producing hydroxyl radical species. This illustrated the importance of the choice of the redox active metal ion at the center of the biomimetic complex in order to judiciously attribute their chemical functionality and reactivity. This interesting chemical property of 1 was believed to trigger the cell cycle arrest of the cancer cells, more specically colorectal cancer cells, by efficiently producing intracellular ROS in a heterogeneous environment. Strikingly, when normal cells were treated with 1 under identical conditions, cell death did not occur due to the less production of ROS under homogeneous environment. The accumulated evidences presented in the cell viability assays clearly supported the effective intracellular generation of ROS by 1 as compared to other biomimetic metal complexes. More detailed investigation on the signaling pathway of cell cycle arrest revealed that the intrinsic mitochondrial apoptotic pathway was activated by 1 through (i) the release of cytochrome c, (ii) the cleavage of caspase 9 and caspase 3, and (iii) the cleavage of PARP1 located in the nucleus. Finally, we conrmed the anticancer effect of 1 both in vitro and in vivo as the inhibition of tumor growth by 1 in the subcutaneous colorectal cancer model was veried. Overall, the present strategy (e.g., the use of biomimetic metal complexes) might tremendously reduce the occurrence of off-target effects, thereby proving the possibility of a wide range of biological application and justify further investigation. Moreover, this would be an important asset-or, at least one of the central functions-that the designed biomimetic metal complexes merit more attention in the future direction of the cancer therapeutic research.</p><!><p>Commercially available chemicals were used without further purication unless otherwise indicated. Solvents were dried according to the published procedures and distilled under Ar prior to use. 74 was prepared by a literature method. 75 The nonheme manganese, iron, cobalt, copper(II) complexes were prepared according to the modied literature methods. 42</p><!><p>The UV-vis spectra were recorded on a Hewlett Packard Agilent Cary 8454 UV-visible spectrophotometer equipped with a T2/ sport temperature-controlled cuvette holder. Electrospray ionization mass spectra (ESI MS) were collected on a Thermo Finnigan (San Jose, CA, USA) LTQ™ XL ion trap instrument by infusing the samples directly into the source at 5.0 mL min À1 using a syringe pump. The spray voltage was set at 4. to each well and the absorbance at 450 nm was measured by Synergy™ HTX Multi-Mode Microplate Reader from Bio-Tek (Winooski, VT, USA) and VICTOR Nivo™ Multimode Plate Reader on a PerkinElmer (Waltham, MA, USA). Chemiluminescent signals were induced by chemiluminescent detection reagent from ATTO (Taito, Tokyo, Japan) and detected by Amersham Imager 600 from GE Healthcare (Chicago, IL, USA). The images observed with a LSM 700 laser scanning confocal microscope and Axio Vert. A1-inverted microscope from Zeiss (Oberkochen, Baden-Württemberg, Germany) and analyzed using ImageJ from NIH (Bethesda, Maryland, USA). qRT-PCR was measured by LightCycler 96 Instrument from Roche (Basel, Basel-Stadt, Switzerland).</p><p>Synthesis and characterization of mononuclear manganese(II), iron(II), cobalt(II), and copper(II) complexes 77 The CCD data were integrated and scaled using the Bruker-SAINT soware package. 78 An empirical absorption correction was applied using the SADABS program. 79 The structures were solved by direct methods, and all non-hydrogen atoms were subjected to anisotropic renement by full-matrix least squares on F 2 using SHELXTL Ver. 6.14. 80 The crystallographic data and selected bond distances and angles are listed in Tables S1 and S2, † respectively.</p><!><p>The human carcinoma-derived cells such as HCT116 and HT-</p><!><p>To determine cell viability, cells were seeded in 96-well plates and allowed to recover for 24 h. Then, the cells were exposed to the metal complexes for 24 h. The metal complexes were dissolved in distilled water and diluted in fresh culture medium (nal metal complex concentration is 50 mM). The added concentration of Z-VAD-FMK was 25 mM (Selleck Chem. #S7023). Aer 24 h treatment of the complexes, the cytotoxicity of the metal complexes was determined using the 2-(2-methoxy-4-nitrophenyl)-3-(4-nitrophenyl)-5-(2,4-disulfophenyl)-2H-tetrazolium monosodium salt (WST-8) assay as previously described. 81 Cellular reactive oxygen species (ROS) assay ROS were detected using 2 0 ,7 0 -dichlorouorescin diacetate (DCFDA) cellular ROS detection assay kit (Abcam # ab113851) following the manufacturers' protocol. Cells were plated overnight in 96-well plates in their medium with or without metal complex. The uorescence excitation/emission 485/535 was measured using the Synergy HTX Multi-Mode Microplate Reader (Bio-Tek).</p><!><p>Statistical analysis was performed using two-tailed paired t test, one way ANOVA-Dunnett's test, and nonlinear regression analysis in GraphPad Prism 5.01 Soware (San Diego, CA, USA).</p><!><p>Cells were seeded in a 6-well plate and cultured in a 37 C, 5.0% CO 2 incubator to adhere to the cells. The old culture medium was discarded and replaced with a fresh medium containing 50 mM of the iron complex. Incubation was then continued for 24 h. Aer treatment was completed, the cells were washed with PBS. The total protein was extracted using an RNA/protein extraction kit (MACHEREY-NAGEL) according to the manufacturer's instructions. Lysates 491A). The horseradish peroxidase-conjugated secondary antibodies in 2.0% skim milk in PBS with 1.0% Tween-20 were incubated at room temperature for 1 h. Signals were induced using the chemiluminescent detection reagent (ATTO) and detected using an Amersham Imager 600 (GE Healthcare).</p><!><p>Cells were xed with 4% paraformaldehyde and permeabilized with 0.20% Triton X-100 in phosphate-buffered saline (PBS). Following xation, cells were incubated at 4 C overnight with anti-COX IV and anti-E-cadherin primary antibody in PBS with 1.0% bovine serum albumin (BSA) and 0.20% Triton X-100. The stained proteins were visualized using Alexa Fluor 488 and Alexa Fluor 594-conjugated secondary antibodies. The nuclei were counterstained with DAPI. The stained cells were observed with a LSM 700 Confocal Laser Scanning Microscope (Zeiss). The nuclei area and number of nuclei were measured by the ImageJ soware (NIH).</p><!><p>The total RNA was extracted using an RNA/Protein extraction kit (MACHEREY-NAGEL). cDNA was synthesized from total RNA (2 mg) using M-MLV Reverse Transcriptase (Promega, M1705). qRT-PCR was performed on a LightCycler 96 System (Roche) using qPCRBIO SyGreen Blue Mix (PCR Biosystems, PB20.15), according to the manufacturer's instructions. The following targets were amplied using the indicated primer pairs (Table S3 †).</p><!><p>All animal experiments were performed according to the guidelines of the Chonnam National University Medical School</p><p>Research Institutional Animal Care Committee, and all experimental protocols were approved by the committee. Five-sixweek-old Balb/c-nu mice were obtained from OrientBio, Inc. (Seongnam, Korea) and housed in metal cages with free access to water and food. To generate the subcutaneous colon cancer models, HT29 human colon cancer cells were harvested during exponential growth and resuspended in a 1 : 1 mixture of saline and Matrigel (BD Biosciences). Next, 1 Â 10 7 HT29 cells per mouse were injected subcutaneously into the ank of Balb/c-nu mice. Mice were randomly assigned to three groups (N ¼ 3): sterile distilled water (control), 1 (concentration: 1 mg kg À1 ), and 1 (concentration: 5 mg kg À1 ). When the average tumor size reached a volume of approximately 100 mm 3 , 1 was administered via intraperitoneal injections ve times for 5 days. Tumor growth was monitored every 2 to 3 days; mice were monitored for signs of toxicity, and the size of the tumors were measured.</p>

Royal Society of Chemistry (RSC)

1D Polymeric Platinum Cyanoximate: A Strategy toward Luminescence in the Near-Infrared Region beyond 1000 nm

We report the synthesis and properties of the first representative of a new class of PtL2 complexes with ambidentate mixed-donor cyanoxime ligands [L = 2-cyano-2-oximino-N,N\xe2\x80\xb2-diethylaminoacetamide, DECO (1)]. Three differently colored polymorphs of \xe2\x80\x9cPt(DECO)2\xe2\x80\x9d (3\xe2\x80\x935) were isolated, with the first two being crystallographically characterized. The dark-green complex [Pt(DECO)2]n (5) spontaneously forms in aqueous solution via aggregation of yellow monomeric complex 3 into the red dimer [Pt(DECO)2]2 (4), followed by further oligomerization into coordination polymer 5. A spectroscopic and light-scattering study revealed a \xe2\x80\x9cpoker-chips\xe2\x80\x9d-type 1D polymeric structure of 5 in which units are held by noncovalent metallophilic interactions, forming a Pt---Pt wire. The polymer 5 shows a broad absorption at 400\xe2\x80\x93900 nm and emission at unusually long wavelengths in the range of 1000\xe2\x80\x931100 nm in the solid state. The near-infrared (NIR) emission of polymer 5 is due to the formation of a small amount of nonstoichiometric mixed-valence PtII/PtIV species during synthesis. A featureless electron paramagnetic resonance spectrum of solid sample 5 recorded at +23 and \xe2\x88\x92193 \xc2\xb0C evidences the absence of PtIII states, and the compound represents a \xe2\x80\x9csolid solution\xe2\x80\x9d containing mixed-valence PtII/PtIV centers. Exposure of KBr pellets with 5% 5 to Br2 vapors leads to an immediate \xe2\x88\xbc30% increase in the intensity of photoluminescence at 1024 nm, which confirms the role and importance of mixed-valence species for the NIR emission. Thus, the emission is further enhanced upon additional oxidation of PtII centers, which improves delocalization of electrons along the Pt---Pt vector. Other polymorph of the \xe2\x80\x9cPt(DECO)2\xe2\x80\x9c complex\xe2\x80\x94monomer\xe2\x80\x94did not demonstrate luminescent properties in solutions and the solid state. An excitation scan of 5 embedded in KBr tablets revealed an emission only weakly dependent on the wavelength of excitation. The NIR emission of quasi-1D complex 5 was studied in the range of \xe2\x88\x92193 to +67 \xc2\xb0C. Data showed a blue shift of \xce\xbbmax and a simultaneous increase in the emission line intensity with a temperature rise, which is explained by analogy with similar behavior of known quasi-1D K2[Pt(CN)4]-based solids, quantum dots, and quantum wells with delocalized carriers. The presented finding opens a route to a new class of platinum cyanoxime based NIR emissive complexes that could be used in the design of novel NIR emitters and imaging agents.

1d_polymeric_platinum_cyanoximate:_a_strategy_toward_luminescence_in_the_near-infrared_region_beyond

INTRODUCTION<!>General Considerations<!>pKa of the New Cyanoxime and Conductivity Measurements<!>NMR Spectroscopy<!>IR Spectroscopy<!>Absorption Spectroscopy<!>Emission Spectroscopy<!>Electron Paramagnetic Resonance (EPR) Spectroscopy<!>X-ray Crystallography<!>Electron Microscopy<!>Ligands<!>Metal Complexes<!>Intertransformation of Complexes in the Pt(DECO) System<!>Structures and NMR Spectra of the New Cyanoxime and Its Pd(DECO) and Pt(DECO) Complexes<!>Electronic Spectra of Platinum Cyanoximates<!>PL of Solid Complexes in the NIR Region<!>Transition Energy and Line Intensity<!>Line Width (Full Width at Half-Maximum)<!>Practicality of Studied Compounds and Future Plans<!>CONCLUSIONS

<p>In the past decade, near-infrared (NIR) technology opened up a variety of novel directions in industrial, scientific, military, and medical applications as new optical devices,1–3 sensors,4,5 contrast agents,6,7 and imaging techniques.8,9 As a result of the development of new technology, a diverse set of photoluminescent (PL) compounds emitting in NIR from small molecules to nanoparticles and aggregates have emerged. However, the great majority of these materials emits up to 800–850 nm, while the number of NIR emitters with emission above 1000 nm is still limited. Fundamental constraints, such as the energy gap law, stating that radiationless transitions at longer wavelengths increase because of vibrational overlaps between the ground and excited states, causing a decrease in the luminescence efficiency,10 significantly restrict the number of luminescent materials beyond 1000 nm. Reported compounds including some lanthanide complexes,11 cyanine dyes,12 certain types of quantum dots,13 small gold nanoparticles,14 single-wall carbon nanotubes,15,16 and, recently, supramolecular assemblies with noncovalent metal–metal (metallophilic) interactions have become an attractive way to synthesize NIR luminescent materials that feature palladium nanowires17 or platinum self-assembled aggregates.18 In these structures, a significant metal–metal interaction extended over multiple metal centers leads to lowering of the gap between the ground and excited states, resulting in longer wavelengths of emission. Such materials are especially promising because of their tunable optical properties, which can be adjusted by the type of ligand and controlled synthesis regulating the number of assembled complexes (polymorphs). Recently, some platinum complexes with their known propensity to metal–metal interactions and polymorphism have attracted attention because of their emissive properties up to 800 nm.19–22</p><p>The key factor in the development of supramolecular emitters with the desired optical properties is to control the aggregation process. Herein, we developed a new class of 1D platinum-based luminescent complexes with strong emission beyond 1000 nm that challenges the energy gap law. The choice of a suitable ligand system is critical for the successful preparation of such 1D complexes. Among ambidentate ligands, oximes used in this work take a special place because of their strong affinity for nickel triad metals that favor square-planar geometry.23–26 On the basis of our previous studies,27–30 we selected cyanoximes28,31–33 as ligands for binding PtII centers. Cyanoximes are particularly flexible mixed-donor ligands that allow modular design and are the platform of choice for developing metal complexes with controlled self-assembly in solutions and the solid state.34–36</p><p>In the design of long-wavelength PtII emitters, we rationalized that the environment facilitates metallophilic interactions, leading to a red shift of the emission21 and fulfilling the necessary geometric requirements for the formation of mixed-valence species. These interactions are kinetically controlled and lead to 1D aggregation along the M---M vector, resulting in a red-to-NIR luminescent 1D polymer. Recently, palladium-based mixed-valence 1D "metallic wire" complexes have shown a very intense NIR emission.17 We also rationalized that the anion's bulkiness from substituents would control the distance between the metal centers. Therefore, among known cyanoxime ligands,36 those with amide groups provide a greater variety of electronic and steric properties and solubility.37–39 Thus, novel bivalent palladium and platinum complexes were synthesized using a cyanoxime ligand well soluble in organic solvents, 2-cyano-2-oximino-N,N′-diethylaminoacetamide [HDECO, 1]. As we demonstrate below, the platinum complex self-assembles in solutions into a luminescent 1D polymeric "poker-chips" structure.</p><p>In this work, we report the synthesis and emission properties of a new PL platinum member of the cyanoxime family. We also report the synthesis and characterization of related bivalent palladium and platinum monomeric complexes of M(DECO)2 composition to understand the unique luminescence of the new material.</p><!><p>All of the necessary chemicals, such as cyanoacetic acid ester, K2CO3, K2[PdCl4], K2[PtCl4], and organic solvents, were obtained from commercial sources and used without further purification. The elemental composition of the starting cyanoxime and its palladium and platinum complexes on the C, H, and N content was determined using a combustion method at Atlantic Microlab (Norcross, GA). Melting points were measured on a Digimelt apparatus without correction.</p><!><p>The ionization constants for the new cyanoxime 1 were determined using a Sirius Analytical Instruments automated titration station (Sussex, U.K.) equipped with a temperature-controlled bath with details provided in the Supporting Information (SI), p S3. The aqueous pKa value was calculated with a Yasuda–Shedlovsky extrapolation in RefinementPro software. The pKa values for 1 and two other closely related amidocyanoximes are summarized in the SI, pp S4 and S5. Electrical conductivity measurements were carried out at 22 °C in dimethyl sulfoxide (DMSO) solutions at 1 mM concentrations of synthesized palladium and platinum complexes, using the Vernier LabQuest digital conductivity meter. An electrode was calibrated with 1 mM DMSO solutions of N(C4H9)4+Br−, P(C6H5)4+Br−, and K2PtCl4. The values of the conductivities are listed in Table S6 in the SI.</p><!><p>The synthesized organic compound 1 and its substituted acetonitrile precursor were characterized by 1H and 13C NMR spectroscopy [solutions in DMSO-d6; tetramethylsilane (TMS) was an internal standard; Varian INova 400 MHz spectrometer]. Variable-temperature experiments for 1, 3, and 6 were conducted in the 20–95 °C range.</p><!><p>IR spectra were recorded in KBr pellets for compounds 1 and 3–6 in the range of 400–4000 cm−1, using the Bruker Vertex S70 Fourier transform infrared (FTIR) spectrophotometer with 64 repetitions at 23 °C and 4 cm−1 resolution.</p><!><p>The UV–visible spectra of the cyanoxime 1 and its anion 2 (as NHEt3+ salt) in solutions were recorded at room temperature (293 K) using an HP 8354 spectrophotometer in the range of 200–1100 nm, in 1 and 10 mm quartz cuvettes (Starna, Inc., Atascadero, CA). The solid-state UV–visible spectra of compounds of interest [Magnus Green Salt (MGS), K2[Pt(CN)4] (KCP), and mixed valence partially oxidized with bromine K2[Pt(CN)4]·0.3Br·H2O (later POCP), as well as synthesized complexes 3–5, and Pd(DECO)2 complex 6; Scheme 1] were recorded as absorbance spectra from their fine suspensions in mineral oil using the above diode-array spectrophotometer. Absorption spectra of the tablets were recorded using a custom setup based on an integrating sphere fiber-optically connected to the silicon-based diode-array CCD camera Synapse (Horiba). Halogen light (HL-2000, Ocean Optics) was used as a light source. An integration time of 0.05 s and a slit of 2 nm were used. All spectra were collected 10 times and averaged. The spectra were collected for light intensity. Absorption spectra were referenced to the spectrum of a pure, neat KBr tablet.</p><!><p>The PL of solid metal complexes was investigated using the following experimental setup based on a Horiba spectrofluorimeter using the CCD liquid-nitrogen-cooled InGaAs diode-array camera Symphony (Horiba) sensitive in the 600–1600 nm range and a 3 s integration time with 20 times signal accumulation in summation mode. In both types of experiments, emission and excitation slits were kept constant at a width of 10 nm. Excitation was conducted with a xenon lamp using a double-grating (2 × 1200 g/mm at 500 mm) monochromator. A long-band-pass 830 nm Schott RG 830 filter was placed in front of the iHR-820 spectrograph with a grating of 100 g/mm at 800 mm. The system was calibrated with the laser (neodymium-doped) glass before every set of measurements. A custom-built anodized aluminum vacuum-pumped cryostat filled with liquid nitrogen/oil was used for variable-temperature experiments in the −195 to +70 °C range, where a J-Kem Scientific digital thermometer with a T-type thermocouple was used to monitor the temperature. 3D excitation–emission scans of the tablets (KBr matrix with 5% of metal complexes) were recorded with the simultaneous measurement of the light intensity (R) and correction of the emission by light intensity. The absolute fluorescence quantum yield of the tablet was measured using a large 150 mm (6 in.) integrating sphere with a fiber-optic bundle with high transmission in the NIR. The tablet was placed at the bottom-loading circular shallow drawer in the sphere. All tablets were pressed as 13 mm disks from a thoroughly homogenized mixture of the IR-spectroscopic-grade KBr and studied complexes 3–6 using a Carver hydraulic press at 23 °C and a pressure of 9 torr (∼6000 psi).</p><p>All glassware cuvettes and hardware parts that were in contact with either KBr pellets or powders of studied metal complexes 3–6 were washed using warm to ∼65 °C N,N-dimethylformamide (DMF) followed by deionized water to avoid contamination and exclude anything related to misinterpretation of the results of PL measurements. All synthesized complexes are well soluble in DMF.</p><!><p>Spectra were recorded on Bruker EMXplus X-band EPR spectrometer with dual-mode cavity and an Oxford cryostat system at +20 and −193 °C using a field sweep from 200 to 4000 G. The field was calibrated using DFPG, while the sensitivity of instrument was checked using a solid standard containing 1% of Cr3+ in Al2(SO4)3. Spectra were recorded as a sum of five repetitions with the time constant for each set at 160 ms.</p><!><p>Crystal structures were determined for the cyanoxime 1 and compounds 3, 4, and 6, suitable single crystals of which were grown by an ether vapor diffusion method in CH3CN solutions. All attempts to grow suitable crystals of 5 were unsuccessful (see the SI, p S24). Crystals of all studied compounds were mounted on a thin glass fiber or plastic MiTeGen holders attached to the copper pin positioned on the goniometer head of the Bruker APEX 2 diffractometer, equipped with a SMART CCD area detector. The intensity data were collected at 120 K in ω-scan mode using a molybdenum tube (Kα radiation; λ = 0.71073 Å) with a highly oriented graphite monochromator. Intensities were integrated from four series of 364 exposures, each covering 0.5° in ω at 20 s of acquisition time, with the total data set being a sphere. The space group determination was done with the aid of XPREP software. The absorption correction was performed by a crystals face-indexing procedure with a videomicroscope (see the SI, pp S8, S11, and S15) followed by a numerical input into the SADABS program that was included in the Bruker AXS software package. All structures were solved by direct methods and refined by least squares on weighted F2 values for all reflections using the SHELXTL program. In the structures of the free ligand HDECO, and in its Pd(DECO)2 complex 6, all hydrogen atoms were placed in calculated positions in accordance with the hybridization state of a hosting carbon atom and refined isotropically. However, in the structure of yellow Pt(DECO)2 (3), all hydrogen atoms were found in the electron difference map and refined anisotropically. No apparent problems or complications were encountered during structure solution and refinement. The crystal data for 1, 3, 4, and 6 are summarized in the SI, p S19, while selected bond lengths and valence angles are presented in the SI, p S21. Representative drawings of the crystal structures and packing diagrams were done using the ORTEP and Mercury software packages.</p><!><p>The scanning electron microscopy (SEM) images from a fine dark-green powder of polymeric complex 5 were obtained using four randomly selected areas at 200, 100, 50, 20, 10, and 5 μm resolution.</p><!><p>The cyanoxime ligand 1 (HDECO) was obtained in high yield according to known procedures from the respective diethylamide, as shown in the SI, p S32. The pure compound represents clear, ice-looking crystals soluble in most organic solvents except hydrocarbons. Yield: 68%. Mp: 87–90 °C. Rf = 0.32 in a 1:1 EtOAC/hexane mixture. Anal. Calcd (found) for C7H11N3O2: C, 49.70 (49.53); H, 6.55 (6.61); N, 24.84 (24.77). UV–visible [λnm, nm (ε, M−1·cm−1): CH3OH, 226 (10331; π → π*); deprotonated DECO− (as NHEt3+ salt), 394 (73; n → π*). 1H NMR (DMSO-d6): δ 1.16 (3H, t, methyl group), 1.12 (3H, t, methyl group), 3.40 (2H, quartet, methylene), 3.47 (2H, quart, methylene), 14.21 (1H, s, oxime); the value of a 3J(H–H) coupling constant in the ethyl group is 7.1 Hz. 13C NMR: two methyl groups at δ 12.74 (CH3), 14.63 (CH3), two ethyl groups at δ 40.95 (CH2), 43.99 (CH2), 127.83 (C=N–OH), 109.77 (CN), 158.82 (amide). IR (KBr pellet, cm−1): 3205 [ν(OH)], 2992 [νas(CH)], 2830 [νs(CH)]; 2250 [ν(C≡N)], 1630 [ν(CO, amide I)], 1478 [ν(C=N, oxime)], 1468 [ν(CO, amide II)], 1034 [ν(N–O)].</p><!><p>The preparation of MGS,40,41 KCP,42 and POCP43 (Krogman's salt) was carried out according to published procedures. Their microscopic photographs (at 40×) are shown in the SI, p S25. All of these complexes were used as standards of known 1D solids with well-defined Pt---Pt distances and acted as model compounds to compare with the newly obtained complexes 3–6. For the synthesis of 6, 0.330 g (1.95 mM) of HDECO were placed in a 100 mL Erlenmeyer flask and 5 mL of distilled water was added to the flask. Pure ligand 1 is not well soluble in water, but after being deprotonated, it becomes well soluble and reacts with transition-metal salts (Scheme 1). Therefore, the required stoichiometric amount of a 1.01 M KOH solution (1.93 mL) was added to the sludge of 1 in water. The cyanoxime dissolved, and the reaction mixture immediately changed to bright yellow. A solution of 0.318 g (0.98 mM) of K2[PdCl4] in 5 mL of water was added dropwise to a yellow solution of 2 under intense stirring. A thick yellow precipitate formed instantaneously, and after ∼5 min of stirring, it was filtered, washed three times with water, and then dried in a desiccator charged with concentrated H2SO4 (see the SI, p S32). Elem anal. Found (calcd) for PdC14H22N6O4: N, 18.24 (18.24); C, 37.05 (37.05); H, 4.50 (4.50). IR (KBr pellet, cm−1): 2983 [νas(CH)], 2940 [νs(CH)], 2211 [ν(C≡N)], 1631 [ν(CO, amide I)], 1440 [ν(CNO), oxime)], 1579 [ν(CO, amide II)], 1200 [ν(CNO)]. Complex 6 is soluble in CH3CN, well soluble in DMF, DMSO, Py, and its homologues, but not well soluble in acetone and alcohols. For the synthesis of Pt(DECO)2, 0.217 g (1.28 mM) of 1 in 8 mL of water was treated with 1.28 mL of a 1.01 M KOH solution, forming a yellow solution of deprotonated cyanoxime 2, to which 0.268 g (0.65 mM) of K2PtCl4 in 5 mL was added. When all components were mixed, the formation of the Pt(DECO)2 complex involving several stages was observed after an extended period of time. After ∼24 h, a very fine dark-green precipitate was filtered (see the SI, p S32), washed three times with water, and then dried in a vacuum desiccator under concentrated H2SO4. Elem anal. Found (calcd) for green-form PtC14H22N6O4 (5): N, 15.75 (15.78); C, 31.52 (31.46); H, 4.16 (4.03). IR (KBr pellet, cm−1): 2985 [νas(CH)], 2943 [νs(CH)], 2213 [ν(C≡N)], 1709 [ν(CO, carbonyl)], 1622; 1442 [ν(CNO), oxime)], 1585 [ν(CO, amide II)], 1227 [ν(CNO)]. The dark-green form 5 is not soluble in water but has some solubility in alcohols, slowly dissolves in DMSO (with the help of an ultrasound bath), CH2Cl2, and CHCl3, forming at first a dark-red solution of dimer 4. Also, complex 5 is well soluble in pyridine and its homologues with the formation of yellow solutions. The yellow form 3 of the same complex can be obtained after dissolution of the green form in CH3CN or DMF, which slowly changes color to red and yellow-orange. Crystals of 3 were grown from an overlap with a heptane/acetonitrile solution or by the vapor-diffusion method when ether slowly dilutes the solution. In both cases, nicely shaped yellow prism-type crystals can be isolated in moderate yield. Elem anal. Found (calcd) for the yellow form 3 (for 5): N, 15.75 (15.89); C, 31.52 (31.42); H, 4.16 (4.27). The data of UV–visible spectra for all synthesized compounds are summarized in the SI, p S43.</p><p>Safety Note! Complexes of bivalent palladium and platinum are known to be cytotoxic agents. Gloves at all times are required during work with these compounds.</p><!><p>One of the most remarkable observations during the Pt(DECO)2 synthesis (Scheme 1) is the color change. The ligand HDECO itself is colorless but turns into the bright-yellow anion DECO− (2) in the presence of a base. Shortly after the addition of the platinum salt, the color of the solution changes to red, then slowly the reaction mixture becomes turbid, and a very fine dark-green precipitate starts to form in the still red solution. The green precipitate makes up the bulk of the pure platinum complex 5 without any admixture of the red or yellow complexes 4 and 3, which are polymorphs. The spectra of the products in solutions are shown in Figure 1. The green form of 5 can be completely converted into the yellow monomeric form 3 by a simple dissolution of the complex in donor solvents such as pyridine and DMSO. Finally, all three forms of Pt(DECO)2 can be isolated (see the SI, pp S26 and S27). The UV–visible spectra of the monomer 3 were similar to those for the anionic ligand 2 in the same solvent (see the SI, Table S43). The red and yellow forms can be obtained from the green one using trituration of the solids with solvents such as CH3CN, CHCl3, and CH2Cl2, followed by either controlled crystallization or mechanical separation. The process of disaggregation appears to be reversible, showing spontaneous formation of the dark-green 5 from the red 4 or yellow complex 3 upon solvent evaporation (see the SI, p S26). The platinum complexes overall remain stable after dissolution: electrical conductivity in DMSO (as the donor solvent that disrupts metallophilic contacts and then dissolves all of the studied compounds) showed nonconductive solutions, which implies that the complexes keep their platinum ligand integrity (see the SI, p S6). For comparison, the bright-yellow 6 complex instantly forms upon mixing stoichiometric amounts of 2 and K2PdCl4 in water (see the SI, p S32). Compound 6 was prepared as a structurally similar, but nonaggregating, model complex used as a control (see the SI, pp S16–S18).</p><p>The dimension of aggregates formed is in the nanoscale, as is evident from their light scattering (Tindal effect) and particle size growth (see the SI, p S39). The linear trend in the increase of its size suggests the Oswald ripening mechanism of particle growth.</p><!><p>The crystal structure of the cyanoxime 1 (see the SI, p S7) reveals an adoption of trans–anti oxime geometry in the solid state. The molecule's core is not planar, with a significant dihedral angle of 53.06° between planes of the cyanoxime and carbonyl fragments due to the steric strain induced by two ethyl groups on the amide fragment. The ligand becomes significantly more planar when coordinated to the platinum metal as described below but adopts cis–anti geometry to adjust its chelate binding in the complex. Variable-temperature 1H NMR spectra of the free ligand 1 and monomeric complexes 3 (and its palladium counterpart 6) in DMSO-d6 revealed a significant increase of ∼6 kJ/mol in the rotation energy barrier of the N,N-diethylamino fragment around the C–N amide bond in the complex as opposed to the ligand. Such a barrier indicates an increased rigidity of the anion in the platinun complex (see the SI, p S4) and explains the overall propensity of the flattened complex to dimerization and further polymerization. Crystal structures revealed that different colors of complexes in the Pt(DECO) system relate to distinct polymorphs (see the SI, p S27). The crystal and refinement data are presented in the SI, pp S19 and S20. The yellow complex was found to correspond to monomeric 3 and adopts a shallow bowl-shapes structure (Figure 2).</p><p>The planar coordination [PtN2O2] environment of metal consists of two chelate ligands in a cis arrangement (Figure 2). Cyanoxime monoanions 2 (DECO−) in all metal complexes presented here (3, 4, and 6) are in the nitroso form, as is evident from the rather short N–O distance of ∼1.24 Å compared to the much longer ∼1.36 Å C–N bonds in the CNO fragment (Table S23 in the SI). The DECO− anion 2 in the monomeric complex 3 is considerably more planar than the structure of 1, with the values of the dihedral angles between the cyanoxime and amide fragments at ∼19–22° (see the SI, p S10). The structure represents "slipped stacks" of the cis-Pt(DECO)2 units, where the shortest intermetallic distance is 7.204 Å, suggesting minimum interaction between the platinum centers (see the SI, p S9). Yellow monomeric platinum complex 3 and palladium complex 6 are isostructural (see the SI, p S18).</p><p>The red complex was found to be the dimeric form of 3: [Pt(DECO)2]2 (4). The crystal and molecular structure of 4 is depicted in Figure 2, and the complex's actual appearance as clear red prisms can be seen in the SI, pp S11 and S27. This centrosymmetric dimer is formed by two bowl-shaped monomeric units of cis-Pt(DECO)2 that form an elegant, but shallow, double-bowl convex structure (Figure 2). The DECO− anion in the structure of 4 is even more planar than that in 3, with the values of the dihedral angles between the two cyanoxime and amide fragments decreasing to 4.4 and 14.7° (see the SI, p S12). Such increased planarity is apparently due to the demand of the metallophiliic Pt---Pt interactions to form the dimer. Two chelate rings in the cis-Pt(DECO)2 units form the dihedral angle of 23.1°, which is responsible for the adoption of the bowl shape by the individual units, but being combined in the dimer takes the convex structure (Figure 2). The Pt---Pt intermetallic distance of the dimer was found to be 3.1208 Å, which is shorter than the sum of the ionic and van der Waals platinum radii and significantly shorter than that for other published platinum dimers.44–47 The overall structure is best described as a column that consists of "slipped dimers" (see the SI, p S13), the geometrical details of which are shown in the SI, p S14. The formation of red and yellow polymorphs of PtLX2 [L = dipy, phen, bis(isoquinoline); X = Cl, Br, CN] has been well documented in the past48–55 and lately for cyclometalated platinum isonitriles.56 However, a study of their interconversion has not been carried out. It is important to mention that other platinum metals (rhodium, for example) in their complexes also exhibit yellow-to-red color changes when the intermetallic distance becomes shorter in the 1D chain.57 That peculiar color change upon the formation of aggregated structures with relatively short metallophilic contacts seems to have a rather general trend independent of the metal center. The color changes to a dark green (such as in MGS)41 or an intense blue like in platinum, rhodium, and iridium when a further decrease in the intermetallic separation occurs.58,59</p><p>Our efforts to grow crystals of the "green" complex 5 suitable for X-ray study were not successful. We were able to obtain very thin dichroic needles (green in reflected light and brown-yellow in passing through light; see the SI, p S28), but these were insufficient to determine the structure. This is, perhaps, due to significant problems of alignment and packing in the crystal dimeric units of cis-Pt(DECO)2 into polymer with the formation of a long-ordered lattice capable of diffracting X-rays, as suggested in the SI, p S24. According to SEM images, a very fine dark-green powder of the coordination polymer 5 represents fibrous microcrystalline materials without visible 5 μm resolution impurities or foreign phase (see the SI, p S29).</p><!><p>The yellow monomeric complex 3 shows typical π → π* transitions in UV–visible spectra around 250–400 nm (Figure 1; see the SI, p S43). The red dimeric complex 4 has three pronounced bands in the UV–visible spectrum at 258 nm (ε = 13670 M−1 cm−1), 374 nm (ε = 18500 M−1 cm−1), and 542 nm (ε = 4400 M−1 cm−1) with a weak shoulder at ∼420 nm (Figure 1). The uncommon dark-green color of the polymeric complex 5 deserves some careful consideration because the origin of such an intense transition is not obvious. The color of this complex is comprised of a "blue" band at ∼800 nm and a "yellow" band of tailing from the UV region of spectra of π → π* transitions at ∼380 nm and a weak n → π* transition at ∼420 nm, which is typical of all deprotonated cyanoximes.28,39,60 The presence of a broad band centered at ∼800 nm in the absorption spectra of 5 in solutions (Figure 1) and in the solid state (see the SI, pp S34, S44, and S45) immediately suggested the formation of a 1D coordination polymer observed in some other platinum complexes such as a well-known MGS, [Pt(NH3)4][PtCl4] with a Pt---Pt chain, and Millon's salt [Cu(NH3)4][PtCl4] with alternating Cu---Pt chains.61 The color of both green compounds is attributed to the charge-transfer (CT) band in "poker-chip stacks" formed by alternating donor [PtCl4]2− and acceptor [Pt(NH3)4]2+ or [Cu(NH3)4]2+ units.41 These exhibit metallophilic interactions and are separated by 3.23 Å in MGS, or 3.22 Å in Millon's salt compared to a typical covalent single Pt–Pt bond (2.6–2.8 Å in length). Similarly, in the absence of direct structural data, we propose the formation of 1D stacks for the "green" Pt(DECO)2 complex 5 in which short (∼3.1 A) metallophilic Pt---Pt contacts are present. We found that the green color of complex 5 quickly disappears upon its dissolution in donor solvents such as DMSO, DMF, CH3CN, pyridine, and 2-picoline, which evidenced a loss of the 1D polymeric structure in these solvents. Yellow solutions formed have UV–visible spectra identical with those for monomeric complex 3. However, the DMF and CH3CN solutions upon drying regenerate green solid 5 (often contaminated with red polymorph 4; see the SI, p S26), which suggests the transformation shown in Scheme 1 and Figure 3. Thus, the red dimer 4 is an intermediate product of such aggregation/disaggregation processes and acts as a building block with the observed short Pt---Pt interactions necessary for the formation of the green 1D polymer 5. It should be noted that the formation of dark-green, or blue-green, aggregates of other platinum-based alkynyl and terpyridyl62–65 complexes in solutions has been previously observed and correlated with the leading role of metallophilic interactions.</p><p>There is a plausible explanation for the origin of such an intense broad band in the NIR region of spectra for the polymeric complex 5 both in the solid state and in solutions in which the compound is stable: formation of a mixed-valence coordination polymer that contains a small amount of nonstoichiometric PtIV species. Thus, we suggest the formation of a "solid solution" of 5 containing other PtII metal centers. In this case, a small amount of higher-oxidation-state platinum would provide conditions for the appearance of a low-energy intravalence CT (IVCT) band in the PtII/PtIV or PtII/PtIII centers. This scenario was well documented during extensive investigations of "platinum blues" and related compounds in the past. Air oxygen or chemical oxidizers such as H2O2, FeIII,47 or AgI 46 cations were successfully employed in the past for the generation of mixed-valence complexes that had intense colors and also demonstrate anisotropic properties in the solid state, with electrical conductivity being one of them. In our case, we tried to address the issue of the partial air oxidation of the initial Pt(DECO)2 complex by carrying out its preparation at strictly anaerobic conditions using a specially designed 1 cm quartz cuvette that allows deaeration of starting solutions via repeated freeze/thaw cycles, accompanied by a system flash with argon (see the SI, p S33). The presence of air oxygen does not seem to be the sole source of the PtII oxidation because the "blue band" started to grow after ∼10 min, followed anaerobic mixing of the components. There is, however, a possibility for the chemical oxidation of PtII by the cyanoxime, which, perhaps, cannot be excluded from consideration (see the SI, p S58). Thus, a partial reduction of the ligand in an aqueous environment to hydroxylamine with the formation of a small amount of PtIV may lead to the formation of a mixed-valence complex, which self-aggregates using Pt---Pt interactions into 1D polymer 5. The latter shows intense bands in the visible region and emits in the NIR region of the spectrum (Figures 1 and 5). Interestingly enough, similar to complex 5 presented here, other solid dark-green Pt(PiPCO)2 and Pt(MCO)2 complexes66,67 conduct electricity in the upper range for semiconductors showing values of the solid-state conductivity at room temperature as 22 and 33 S/cm, respectively (see the SI, pp S30 and S31). As observed for the Pt(PiPCO)2 and Pt(MCO)2 complexes, the electrical conductivity at the high end of the semiconductors indicates the presence of delocalized carriers at room temperature, which is typical of mixed-valence compounds. Such carriers, for example, were introduced in the solid-state solution of the famous MGS upon partial oxidation of PtII and formed mixed-valence species, dramatically improving its conductivity.68 As mentioned above, dark-green polymeric Pt(PiPCO)2 and Pt(MCO)2 complexes being excited at 770 nm also emit in the NIR region at 1020 and 1038 nm, respectively (see the SI, p S53). These new cyanoxime ligands also represent N-acetamides formed with piperidine and morpholine, respectively.27,29,30,43 Details of syntheses, structures, and spectroscopic properties of these new polymeric platinum cyanoximates will be published elsewhere.</p><p>The formation of PtIII species as impurities in the dark-green complex 5 is excluded based on the EPR silence of this coordination polymer at 293 and 80 K (see the SI, p S46). The PdIII centers have readily observable low-spin d7 configurations in a square-planar environment.69–73</p><p>The formation of variable-length aggregates in 5, which were held together by metallophillic Pt---Pt interactions, can provide an explanation for some range of λmax of the "blue band" at ∼720–820 nm in visible/NIR spectra attributed to "poker-chip" stacks of variable size. Thus, a difference between the values of maxima in spectra of 5 in the solid state and different solvents, including micelles, can be as big as ∼100 nm. However, donor solvents such as DMSO and Py facilitate a fast and monotonic decrease of the "blue band" intensity and its eventual disappearance (see the SI, p S34), with the remaining spectrum identical with that of yellow complex 3. It is important to note the similar behavior of mixed-valence tetrameric platinum blue complex [Pt4(NH3)8(DMGI)4]·(NO3)5·2H2O (DMGI = 3,3-dimethylglutaimidate), where an intense band at ∼750 nm decreases upon complex disaggregation and reduction to bivalent species.74</p><p>The red dimer 4 has a pronounced band at 542 nm (Figure 1). The latter may be attributed to the metal-to-ligand CT band, as was observed for several red PtII dimers of [PtLX2] (L = dipy, phen; X = Cl, Br, CN) composition in the past.44 The phenomenon of a size-dependent change of λmax in the PL of quantum dots, wells, and nanoparticles is well established75 and also associated with extensive charge-carrier delocalization.</p><p>It is likely that both cases outlined above take place in our particular example of polymeric 1D platinum cyanoximates, where mixed-valence species with IVCT have the dominant effect. In that sense, the presence of small quantities of PtIV centers in complex 5 is similar to a doping effect well-known and widely used for semiconductors. In our case, however, it is manifested by a significant red shift of the emission into the extended NIR region beyond 1000 nm. Additional and specifically designed studies of complexes 4 and 5 and other similar platinum cyanoximates are necessary for obtaining a more definite answer regarding the origin of the observed transition. These studies were out of the scope of the current investigation but will be carried out for other recently discovered 1D platinum cyanoximates and reported elsewhere.</p><!><p>Crystal data for complexes 3, 4, and 6 revealed that palladium and platinum cyanoximates are typical Werner-type complexes of a nickel triad. Many complexes of this family are known to form dimers45–47 or extended 1D columnar structures that exhibit strong anisotropy of their physical and optical properties.3,42,76 For instance, they possess electrical conductivity and demonstrate unusual recently detected photoemission in the red part of the visible spectrum.77–80 The presence of an NIR absorption band associated with metallophilic interactions in the UV–visible spectra of 5 prompted us to investigate the PL of this compound in the NIR range. To avoid the observed interconversion of the species 3 ↔ 5 in solution, we investigated the emission in the solid state. The solids were embedded in the KBr matrix at 5% by weight concentration to minimize the uncertainties associated with the powder density and particle size and enhance the reproducibility of the results. Such a technique is commonly used in FTIR measurements but was originally also proposed for absorption spectra.22 Control compounds, such as model 1D platinum salts with known different Pt---Pt distances (Figure 4), as well as Pd(DECO)2 complex 6 (see the SI, p S16), were also investigated. The pellets shown in Figure 4 had a long shelf-life stability (more than 1 year) if stored in a desiccator at room temperature. Details of their preparation and experimental setup can be found in the SI, pp S35 and S37.</p><p>All pellets were excited at 770 nm with emission measured in the range of 800–1600 nm. Among the studied complexes, only the dark-green platinum complex 5 was found to emit in the NIR range with λmax ∼ 1060 nm (Figure 5). The effect of the KBr matrix was marginal; the powder also exhibited a similar NIR emission although at a slightly shorter emission maximum (∼1047 nm; see the SI, p S47). Also, it was found that in a vacuum the intensity of the NIR emission is much higher, which implies the role of air oxygen in the triplet state as the phosphorescence quencher (see the SI, p S48). The quantum yield of the emission from a solid pellet of 5 in KBr was measured absolutely and was found to be 0.51% (see the SI, p S49). Although this quantum yield appears to be low, it might be significantly higher in the tablets with a lower concentration of the active component. The graphs shown in the SI, p S45, demonstrate significant absorption from the tablet, thus pointing to potential reabsorption by the sample, which leads to a lowering of the measured quantum yield especially with the absolute method.81</p><p>The platinum and palladium monomeric complexes 3 and 6 were not emissive in the NIR region in accordance with the absence of the absorption band at the wavelength of excitation (770 nm). Similarly, known 1D platinum-coordination polymers such as MGS, POCP, and KCP (Figures 4 and 5), which were used as control compounds, did not show any measurable PL in the NIR region of the spectrum being excited in the 350–700 nm range (see the SI, pp S54 and S55). This range is important for potential biological applications of the NIR emitters. The absence of PL of the above control compounds and the presence of PL in the emissive complex 5 demonstrate the critical importance of the cyanoxime DECO− ligand 2 in the complex. This is perhaps due to its ability to alter the electronic structure of polymer 5 in the manner of dramatic lowering of the energy of key molecular orbitals corresponding to transitions. Also, the electron-withdrawing character of the cyanoxime can make the formation of mixed-valence species easier and thus facilitate electron "hopping" between two different oxidation states.</p><p>We suggest that the observed PL of 5 originated from low-energy transitions in an extended mixed-valence complex with metallophilic interactions between the centers. In order to prove this, a small drop of bromine was added to the chamber with the KBr pellet containing 5. The intensity of the emission in the first 1 min of such a treatment increased by ∼30%, with the pellet changing from dark green to copper-red, followed by a slow decrease in the NIR emission and the pellet turning yellow (see the SI, p S56). Thus, bromine partially oxidized PtII to PtIV and facilitated a significant increase in the degree of formation of the mixed-valence complex. The decrease in the intensity is associated with the complex decomposition in the presence of halogen (see the SI, p S57).</p><p>In the context of the role of an oxidizer and experimental conditions during the syntheses of quasi-1D platinum complexes, it is appropriate to mention inconsistencies of both emission intensities and its energies referred to in the past as a "sample history".82 No attempts to resolve and study this phenomenon were made. In our view, this is a clear manifestation of the formation or disappearance of mixed-valence species in that particular system.</p><p>Under the same synthetic conditions as those for Pt(DECO) complexes, Pd(DECO)2 forms exclusively a yellow monomer 6 (see the SI, pp S16 and S21), which has no NIR emission. Importantly, there are some significant differences in time between the platinum and palladium complexes' formation. For instance, the Pd(DECO)2 complex 6 formed within several minutes, as opposed to the much slower reaction of the formation of the green Pt(DECO)2 complex 5. This is in line with the much greater kinetic lability of bivalent palladium as opposed to platinum.83 Hence, kinetic lability explains why the PdII complexes did not form the extended 1D polymers necessary for emission in the NIR region.</p><p>The emission–excitation scan for the KBr pellet of 5 is shown in Figure 6A,B and evidenced that the position and intensity of the emission signal remain around 1000 nm and are weakly dependent on the excitation wavelength. This is typical of systems with delocalized charge carriers such as mixed-valence species, quantum dots, and wells. The emission spectrum of 5 was registered at the range of 825–1400 nm using a step of 5 nm and an integration time of 0.5 s with the help of a diode-array CCD detector. Moreover, the emission scan does not resemble the absorption spectra of the green [Pt(DECO)2]n complex 5 in KBr pellets (see the SI, p S45) because of severe distortion of the former due to strong chromophore coupling in the solid state.</p><!><p>Investigation of the temperature-dependent PL of the dark-green complex 5 in the KBr pellet shed some light on the mechanism of the emission. The experiments were carried out using a custom-built liquid-filled cryostat (see the SI, p S36) with the sample under vacuum at ∼6 × 10−2 torr to prevent moisture condensation on the quartz windows. Data displayed in Figure 7 indicate a pronounced bathochromic shift of λmax of the emission band and a significant intensity decrease upon lowering of the temperature from +50 to −170 °C. The observed signal intensity change for complex 5 is opposite to the common trend.78 However, a very similar red shift of the emission energy was found earlier for other quasi-1D solids such as numerous KCP-type compounds84 and columnar red polymorphs of PtLX2 (L = dipy, phen; X = Cl, Br, CN),49,52,55 or nonaggregated [Pt(ter)Y] (ter = terpyrinine; Y = Cl, solvent molecule).85 Nevertheless, in the latter series of complexes, the opposite to that observed for the complex 5 trend of a dramatic increase in the emission intensity with a temperature decrease was documented.</p><p>As we stated above, the yellow monomeric complex 3 is not emissive; only complex 5, which is assembled in a 1D polymer, shows pronounced luminescence beyond 1000 nm. We explain the simultaneous increase in the emission intensity with the temperature by analogy with the well-documented behavior of quantum dots and quantum wells with delocalized carriers. Thus, multicore and multishell CdSe/CdS/ZnS/ZnSe quantum dots, as well as single quantum wells GaAsSb/AlGaAs, demonstrate very similar temperature-induced changes in the emission energies and intensity profiles. Consequently, excitons with a temperature increase gain sufficient thermal energy to overcome small potential barriers in the local potential minimum, become mobile, and transfer to higher energy states of the band until the edge of the conduction band is reached.86,87 Hence, the emission energy undergoes a blue shift with a temperature increase. We suggest that at low temperatures excitons mostly localize in a mixed-valence aggregate, which is small in size, or even in the mixed-valence dimer. Elevation of the temperature creates delocalized excitons that can now reach higher energy states and, because of electron hopping, spread on a larger distance, increasing the probability of emission from a larger number of mixed-valence centers along the Pt---Pt wire. In general, increasing with temperature, electronic hopping interactions between sites lead to a greater degree of delocalization in both the ground and excited states.88 The energy that corresponds to an observed shift of 54 nm is 472 cm −1 (58 meV, or 5.65 kJ/mol) for complex 5 (Figure 7) and is consistent with delocalization of electrons along the 1D chain in compounds with metallophillic interactions.3,89</p><p>Ultimately, these results can be explained in terms of an increase of the distance traveled by electrons along the 1D Pt---Pt chain (e-delocalization) with rising temperature, which requires higher energy. Lower temperatures tend to localize this motion to a smaller distance. Similar blue-shift behavior was reported for a series of cadmiun halcogenide90,91 and In–As quantum dots92 and wells86 where even values of the PL temperature shifts are in the same range of 30–60 meV.93</p><p>The intermetallic Pt---Pt distances in KCP, MGS, and POCP were previously reported as 3.5, 3.25, and 2.89 Å, respectively. The crystal structure of the dimer 4 showed the Pt---Pt distance as 3.1208 Å, and apparently the emitting 1D polymeric complex 5 has the same or a similar intermetallic distance. Our data indicate rather clearly that the length of the metal–metal bond alone cannot explain the observed emission. Rather, the combination of several factors is important. The unique role of the cyanoxime ligand in the emission is suggested to be critical to (1) the formation of an appropriate low-energy electronic state including metal orbitals due to the asymmetric cis-PtN2O2 environment with significant covalent character in Pt–N bonding, (2) the electron-withdrawing character of the ligand, which allows depletion of the electron density from the metal centers and helps in the formation of mixed-valence centers due to an intramolecular redox process, and (3) providing a favorable geometrical configuration, leading to alignment of 1D polymeric "poker-chip stacks" of certain length and necessary for the NIR emission.</p><p>Other ligands with different substituents are currently being investigated to evaluate the role of the ligand in the assembly.</p><!><p>The almost symmetrical signal of the emission from polymeric complex 5 can best be described as two Gaussian-type lines (see the SI, pp S51 and S52). The minor slightly higher energy component contributes from 2.2 to 5.4% intensity to the overall fit at +50 and −165 °C. During variable-temperature studies, we also observed line narrowing upon a temperature decrease (see the SI, p S50), which is in agreement with the literature data for all the above-mentioned quasi-1D systems.</p><!><p>Similar to other platinum complexes such as cisplatin and oxaliplatin, platinum cyanoximates also demonstrate cytotoxicity27,28,30 and, in combination with their emissive properties, can be potentially utilized as theranostic agents. For these biomedical imaging and theranostics applications, the complexes have to be formulated for better bioavailability. Therefore, encapsulation of these NIR-emissive complexes in micelles, or water-soluble polymeric shells, is considered for future studies. Our preliminary data indicate just facile formation of monodispersed micelles based on sodium salts of long-chain carboxylic acids (see the SI, pp S40 and S41). This finding opens a new area of research for application of this and similar polymeric platinum cyanoximates. Future work also will be focused on a better understanding of the nature of the emission and tuning of the structures toward brighter complexes with a diverse range of emission. Such work will include, for example, characterization and measurement of the PL lifetime to characterize the type of emission. Given that the 1D aggregate might have a relatively large length distribution, it is also important to identify the relationship between the brightness of the emission and the size of the "poker-chip" stack. It is also critical to understand how the nature of the ligand affects the emissive properties of the complexes. This knowledge will expand the family of emitters to a variety of materials with different wavelengths of emission. Equally important is to demonstrate the utility of the complexes in real-life applications. Instead of embedding the complexes in KBr tablets, which are excellent models for the basic research, more realistic matrix materials should be explored.</p><!><p>In summary, we developed a unique supramolecular assembly based on the PtII complex of PtL2 composition (L = new cyanoxime ligand of the N-acetamides family, namely, 2-oximino-2-cyano-N,N′-diethylacetamide, HDECO). The complex has three polymorphic forms: monomeric yellow Pt(DECO)2, red dimeric [Pt(DECO)2]2, and dark-green [Pt(DECO)2]n. The latter complex represents a quasi-1D coordination polymer that strongly absorbs in the 400–900 nm range and luminesces at 1000–1200 nm. This compound also appears to be a solid solution containing nonstoichiometric amounts of PtIV centers. Thus, aggregation of monomeric Pt(DECO)2 units into a stacked polymer in aqueous solution is driven by metallophilic interactions and accompanied by the partial oxidation of PtII to PtIV. Contrary to other previously reported unstable mixed-valence palladium-17 or platinum-based72,74 "metal wires", presented in this work dark-green polymeric platinum cyanoximate is stable in the solid state at room temperature for many months. This and similar NIR emissive platinum cyanoximate complexes can be used in light-emitting devices and optical sensors to cover wavelengths that are currently unavailable. In addition, because of the transparency of the biological tissue at 1060–1150 nm (second optical window)8,9,94 and established cytotoxicity of platinum cyanoximates,27,30,95 the emissive platinum complexes can be utilized as theranostic agents for anticancer treatment and diagnostics.</p>

PubMed Author Manuscript

Activation of microglia with zymosan promotes excitatory amino acid release via volume-regulated anion channels: the role of NADPH oxidases

Microglia are the resident immune cells of the CNS, which are important for preserving neural tissue functions, but may also contribute to neurodegeneration. Activation of these cells in infection, inflammation, or trauma leads to the release of various toxic molecules, including reactive oxygen species (ROS) and the excitatory amino acid glutamate. In this study we used an electrophysiological approach and a D-[3H]aspartate (glutamate) release assay to explore the ROS-dependent regulation of glutamate-permeable volume-regulated anion channels (VRACs). Exposure of rat microglia to hypoosmotic media stimulated Cl\xe2\x88\x92 currents and D-[3H]aspartate release, both of which were inhibited by the selective VRAC blocker DCPIB. Exogenously applied H2O2 potently increased swelling-activated glutamate release. Stimulation of microglia with zymosan triggered production of endogenous ROS and strongly enhanced glutamate release via VRAC in swollen cells. The effects of zymosan were attenuated by the ROS scavenger MnTMPyP, and by two inhibitors of NADPH oxidase (NOX) diphenyliodonium and thioridazine. However, zymosan-stimulated glutamate release was insensitive to other NOX blockers, apocynin and AEBSF. This pharmacological profile pointed to the potential involvement of apocynin-insensitive NOX4. Using RT-PCR we confirmed that NOX4 is expressed in rat microglial cells, along with NOX1 and NOX2. To check for potential involvement of phagocytic NOX2 we stimulated this isoform using protein kinase C (PKC) activator PMA, or inhibited it with the broad spectrum PKC blocker G\xc3\xb66983. Both agents potently modulated endogenous ROS production by NOX2, but not VRAC activity. Taken together, these data suggest that the anion channel VRAC may contribute to microglial glutamate release, and that its activity is regulated by endogenous ROS originating from NOX4.

activation_of_microglia_with_zymosan_promotes_excitatory_amino_acid_release_via_volume-regulated_ani

Introduction<!>Materials<!>Cell culture preparation<!>Immunocytochemistry<!>Whole-cell patch clamp recordings<!>Excitatory amino acid efflux assay<!>Nitroblue tetrazolium (NBT) assay<!>Total RNA isolation and RT-PCR<!>Statistical analysis<!>Characterization of swelling-activated, glutamate-permeable pathway in primary rat microglia<!>Exogenous H2O2 enhances the release of glutamate from primary microglia via VRAC<!>Zymosan stimulates the endogenous production of O2- and enhances microglial VRAC activity in a ROS-dependent manner<!>Expression of different NOX isoforms in microglia<!>Activation and inhibition of NOX2 modulates O2\xe2\x88\x92 production without having any impact on microglial VRAC activity<!>Discussion<!>VRACs contribute to glutamate release from microglia<!>Microglial VRACs are potently regulated by exogenous and endogenous ROS<!>Differential impact of NOX isoforms on regulation of microglial VRAC by ROS<!>Pathological significance VRAC-mediated glutamate release from microglial<!>Cultured primary rat microglial cells express volume-regulated anion channels, which are permeable to excitatory amino acids<!>Exogenous reactive oxygen species H2O2 potentiate the excitatory amino acid release from primary microglia via VRAC<!>Zymosan stimulates the endogenous production of superoxide anion (O2\xe2\x88\x92) and enhances microglial excitatory amino acid release<!>Endogenous ROS production mediates zymosan effect on microglial excitatory amino acid release via VRAC<!>The NOX blocker apocynin potently suppresses the zymosan-induced ROS production, but not the zymosan effect on VRAC activity<!>The NOX inhibitor thioridazine blocks zymosan-enhanced ROS production and excitatory amino acid release<!>NOX1, NOX2, and NOX4 isoforms of NAD(P)H oxidase are expressed in primary microglial cells<!>The PKC inhibitor G\xc3\xb66983 completely prevents NOX2-medited O2\xe2\x88\x92 production but not the zymosan-enhanced release of excitatory amino acids via VRAC<!>PKC-dependent stimulation of NOX2 enhances O2\xe2\x88\x92 production without impacting the VRAC-mediated release of excitatory amino acids<!>Schematic representation of the hypothetical mechanisms contributing to the effects of zymosan on production of ROS and VRAC activity in rat microglial cells<!>

<p>Microglia are the resident immunocompetent cells of the central nervous system. In the adult brain, microglial cells are typically found in a resting state, in which they survey their local environment for any signals indicating traumatic injury, infection, or inflammation (Kreutzberg 1996; Farber and Kettenmann 2005). Upon exposure to invading pathogens, neuronal debris, elevated extracellular ATP levels, or pro-inflammatory cytokines and chemokines, microglia rapidly convert to an activated state. Once activated, they produce both cytotoxic and neuroprotective molecules (Kreutzberg 1996; Nakajima and Kohsaka 2001). The most prevalent microglial toxins are IL-1β, TNFα, reactive oxygen species (ROS), and reactive nitrogen species (RNS) (Nakajima and Kohsaka 2001; Block et al. 2007). Although ROS and RNS production is an intrinsic part of the CNS response to infections, trauma, neurodegeneration, or brain tumors, it frequently results in bystander neurotoxicity. ROS and RNS, including those produced by microglia, are thought to play an important role in numerous neurodegenerative processes (Nakajima and Kohsaka 2001; Halliwell 2006; Block et al. 2007). The primary source of ROS production in microglia are NADPH oxidases, multi-subunit heme-containing enzymes that catalyze the reduction of molecular oxygen to superoxide anion (O2−) using NADH or NADPH as a substrate (Bedard and Krause 2007).</p><p>A less recognized mediator of microglial neurotoxicity is the amino acid glutamate, which serves as the main excitatory neurotransmitter in the CNS. Several previous studies demonstrated that activated microglia release glutamate in quantities sufficient to induce neuronal death via overactivation of ionotropic glutamate receptors (Piani et al. 1991; Piani et al. 1992; Barger and Basile 2001; Takeuchi et al. 2006). Piani and co-workers suggested that glutamate, rather than ROS and/or inflammatory cytokines, is the key factor contributing to microglial toxicity (Piani et al. 1992). Two glutamate transporting pathways that have been proposed to mediate glutamate release from microglia are the cystine-glutamate antiporter and connexin hemichannels (Piani and Fontana 1994; Barger and Basile 2001; Takeuchi et al. 2006). The cystine-glutamate antiporter (system Xc-) exchanges one intracellular glutamate for extracellular cystine, a homodimer of cysteine, which is used for the synthesis of the antioxidant glutathione (McBean 2002). This transporter is upregulated in immune cells in response to oxidative stress in order to maintain sufficient levels of glutathione, and is abundantly expressed in activated microglia (Sato et al. 2001; Qin et al. 2006). However, Takeuchi et al. (2006) recently found that connexin hemichannels, rather than the cystine-glutamate antiporter, largely mediate microglial glutamate release and excitotoxic neuronal damage in a TNFα toxicity model. Additionally, in astrocyte cultures, it has been demonstrated that glutamate release in response to hypoosmotic swelling and chemical ischemia is mediated by two anion channels: volume-regulated anion channel (VRAC) and Gd3+-sensitive maxi-anion channel (Liu et al. 2006).</p><p>VRACs are ubiquitously expressed Cl− channels whose molecular identity remains unknown (Strange et al. 1996; Nilius and Droogmans 2003; Okada 2006). In addition to inorganic anions, VRACs are permeable to several amino acids, including the excitatory amino acids glutamate and aspartate (Kimelberg et al. 1990; Banderali and Roy 1992; Jackson et al. 1994; Abdullaev et al. 2006). Although it has not been demonstrated in microglia, VRACs may contribute to pathological glutamate release, as seen in ischemia and other neurological disorders (Kimelberg 1995; Kimelberg and Mongin 1998; Mongin and Kimelberg 2005). The presence of VRACs in mammalian cells is typically determined by measuring increases in Cl− conductance in response to cell swelling using electrophysiology. Besides sensitivity to cell volume changes, the key biophysical characteristics of VRACs include intermediate conductance, moderate outward rectification, Eisenman's type I anion permeability sequence (SCN− > I− >NO3− > Cl− > F− > gluconate), requirement for cytosolic ATP, and time-dependent inactivation at large positive potentials (Strange et al. 1996; Okada 1997; Nilius et al. 1997).</p><p>Several studies demonstrated that microglial cell functions are strongly regulated by changes in the expression and activity of K+, H+, and Cl− channels (reviewed in Eder 2005). Volume- and/or membrane stretch-sensitive Cl− channels, resembling VRACs, were found in cultured microglial cells and have been proposed to participate in morphological transitions, migration, proliferation, and phagocytosis in this cell type (Schlichter et al. 1996; Eder et al. 1998; Ducharme et al. 2007). Although the primary physiological role for VRACs is regulation of cellular volume, they are also activated during other cellular processes, such as apoptosis, cell motility, and proliferation (Okada et al. 2001; Lang et al. 2000; Stutzin and Hoffmann 2006). Several recent studies have found that VRACs may be activated or positively modulated by ROS, particularly H2O2 (Shimizu et al. 2004; Varela et al. 2004; Haskew-Layton et al. 2005). In ventricular myocytes, endogenous ROS have been demonstrated to regulate stretch-activated and swelling-activated Cl− currents, both likely mediated by VRAC (Browe and Baumgarten 2004; Ren et al. 2008). Since microglia robustly generate ROS as part of their immunological responses, they represent an ideal cell type for studying the impact of endogenously produced ROS on VRAC activity and VRAC-mediated release of organic osmolytes. In the present work we focused on the VRAC-mediated release of the excitatory amino acids, glutamate and aspartate, because of the significant physiological and pathological impact of these neurotransmitters on neuronal signaling and viability in the brain.</p><!><p>4-(2-Aaminoethyl) benzenesulfonyl fluoride hydrochloride (AEBSF), apocynin, dipehnyliodonium chloride (DPI), nitroblue tetrazolium chloride (NBT), thioridazine, zymosan A (S. cerevisiae), and H2O2 (30%), were purchased from Sigma-Aldrich. Gö6983, Mn(III)tetrakis(1-methyl-4-pyridyl)porphyrin pentachloride (MnTMPyP), and phorbol 12-myristate 13-acetate (PMA) were acquired from EMD-Calbiochem (La Jolla, CA, USA). 4-[(2-Butyl-6,7-dichloro-2-cyclopentyl-2,3-dihydro-1-oxo-1H-inden-5-yl)oxy]butanoic (DCPIB) acid was from Tocris Cookson (Ellisville, MO, USA).</p><!><p>Microglial cells were purified from mixed glial cultures, which were prepared from the cortices of neonatal Sprague-Dawley rat according to the procedure approved by the Albany Medical College Animal Care and Use Committee. After the isolation, the cortical tissue was enzymatically dissociated using three enzymatic extractions with protease Dispase II (Sigma-Aldrich, St. Louis, MO, USA). The first extraction was discarded and DNAse I (Sigma, St. Louis, MO, USA) was added. Dissociated cortical cells were plated in T-75 flasks and grown for two-three weeks in minimal essential medium (MEM) plus 10% fetal bovine serum (FBS) and additionally supplemented with 50 units/mL penicillin and streptomycin (all cell culture reagents were from Invitrogen, Carlsbad, CA, USA) in a humidified 5% CO2/95% air atmosphere at 37°C. Culture medium was replaced bi-weekly, and the penicillin and streptomycin were no longer added to the media after ten days of cultivation. This procedure yields a confluent mixed glial culture largely consisting of astrocytes and microglia. For purifying microglia, mixed glial cultures were shaken for approximately two to five minutes on a titer plate shaker. Floating microglial were collected, sedimented by a brief centrifugation, resuspended in Opti-MEM supplemented with 2% B-27 serum-free supplement minus antioxidants, and plated onto 18 mm2 glass coverslips (Caroline Biological Supply Co., Burlington, NC, USA) or cell culture dishes coated with poly-d-lysine (Sigma, St. Louis, MO, USA). Cells were maintained in Opti-MEM/B-27 for at least 12 hours before being used in subsequent experiments.</p><!><p>To verify the purity our microglia cultures, we stained the isolated microglial cells for OX-42 (CD11b), a complement receptor that within the CNS is expressed solely by microglia. Briefly, microglial cells plated onto 18 mm2 glass coverslips were fixed in 4% paraformaldehyde for 20 minutes at room temperature (22°C). The cells were then blocked in 10% normal goat serum (Invitrogen) for 30 minutes at room temperature. Blocked cells were incubated with an OX-42 mouse anti-rat monoclonal antibody (1:500; BD Pharmingen, San Diego, CA, USA) for two hours at room temperature, washed with physiological phosphate saline, and incubated with an Alexa488-conjugated goat anti-mouse secondary antibody (1:1,000 Molecular Probes, Carlsbad, CA, USA) for 45 minutes at room temperature. After completion of immunohisto-chemical staining microglia were additionally counterstained with DAPI (100 ng/mL; Sigma) for five minutes at room temperature. Images were analyzed with an Olympus Optical Provis AX70 microscope equipped with conventional fluorescence filters (Cy2/Alexa488: 460-500 nm excitation, 510-560 nm emission; DAPI/Hoechst: 375-400 nm excitation, 450-475 nm emission). Fluorescent images were captured with a high-resolution camera (Sony, DKC-ST5) interfaced with Northern Eclipse (Empix Imaging, Mississaugau, ON), Photoshop (Adobe, San Jose, CA), and Neurolucinda (MicroBrightField, Colchester, VT) software. This procedure revealed that >98% cells were positive for OX-42 (CD11b) and therefore are of microglial origin.</p><!><p>Microglial cells were plated on poly-d-lysine-treated glass coverslips at low density in MEM supplemented with heat-inactivated horse serum (HIHS, Invitrogen). After three hours they were transferred into B-27-supplemented Opti-MEM as described above. Adhered cells were moved into electrophysiological media described below. Whole-cell recordings were performed at room temperature as described previously (Kubo and Okada 1992; Abdullaev et al. 2006). Patch electrodes were fabricated from borosilicate glass capillaries using a micropipette puller (P-97, Sutter Instruments, Novato, CA), with a resistance of 3-3.5 MΩ when filled with pipette solution. Series resistance was ≤15 MΩ. Currents were recorded using an Axopatch 200B amplifier (Axon Instruments, Foster City, CA). pCLAMP software (version 9.2, Axon Instruments) was used for command pulse control, data acquisition, and analysis. Series resistance was compensated in all experiments. Current signals were filtered at 2 kHz using a four-pole Bessel filter and digitized at 4 kHz. The time course of current development was monitored by applying every 15 seconds alternating one-second step pulses from a holding potential of 0 to test pulses ±40 mV. After attaining steady-state activation of Cl− currents, their biophysical properties were determined by applying 2-s step pulses from 0 mV to test potentials of -100 to +100 mV in 20-mV increments. Short prepulse to -100 mV were applied before each test pulse to assure complete activation of VRAC channels. The isoosmotic external solution contained (in mM): 110 CsCl, 2 CaCl2, 1 MgSO4, 5 glucose, 10 HEPES, and 60 mannitol (pH 7.4, 290 mosM). The hypoosmotic solution was made by omitting mannitol from isotonic solution and had an osmolarity of 230 mosM. The pipette solution contained (in mM): 110 CsCl, 1 MgSO4, 1 Na2-ATP, 0.3 Na2-GTP, 15 Na-HEPES, 10 HEPES, and 1 EGTA (pH 7.3, 255 mosM). The osmolarity of the pipette solution was set lower than that of the isotonic bath solution in order to prevent spontaneous cell swelling after attaining the whole-cell mode (Worrell et al. 1989).</p><!><p>The release of excitatory amino acids from primary microglia was measured using a non-metabolized L-glutamate/L-aspartate analog, D-[3H]aspartate, as described elsewhere (Mongin and Kimelberg 2002). In brief, primary microglia plated on 18 mm2 glass coverslips were loaded for 3 hours with 8 μCi/mL D-[3H]aspartate in MEM plus 10% HIHS. Loaded microglial cells were washed of extracellular isotope and serum-containing medium with HEPES-buffered solution, and then placed in a Lucite perfusion chamber. Cells were suffused with isoosmotic or hypoosmotic media at 1.2 mL/min. The isoosmotic medium contained (in mM) 135 NaCl, 3.8 KCl, 1.2 MgSO4, 1.3 CaCl2, 1.2 KH2PO4, 10 d-glucose, and 10 HEPES (pH 7.4). The hypoosmotic medium was prepared by reducing NaCl concentration to 85 mM, which translates into a 30% reduction in the medium osmolarity. Suffused media fractions were collected at one minute intervals. Isotope (D-[3H]aspartate) content in each fraction was determined using a PerkinElmer Tri-Carb 1900TR liquid scintillation analyzer (PerkinElmer, Downers Grove, IL, USA) after addition of four mLs of Ecoscint A scintillation cocktail (National Diagnostics, Atlanta, GA, USA). The isotope remaining in the cells was extracted with a solution containing 2% sodium dodecylsulfate plus 8 mM EDTA. Percent fractional isotope release for each time point was determined by dividing radioactivity released in each 1-minute interval by the radioactivity remaining in the cells at this time point, as retroactively calculated using a custom computer program.</p><!><p>The quantitation of O2− production in primary microglia was performed using a nitroblue tetrazolium assay as described elsewhere (Choi et al. 2006). In brief, primary microglial cells grown on 24-well culture plates were incubated with nitroblue tetrazolium chloride and various compounds that promote and/or inhibit the production of O2− for 10 or 30 minutes at 37°C. Cells were then rinsed with warm PBS and fixed in 100% methanol. Reduced formazan particles, contained inside cells, were dissolved with 2M KOH in DMSO, and the absorbance was read at 620 nm on a BioTek ELx800 absorbance microplate reader (BioTek Winooski, VT, USA).</p><!><p>Total RNA was isolated from primary rat microglia using the RNAqueous-4PCR kit (Ambion, Austin, TX, USA) according to the manufacturer's protocol. The concentration and purity of the isolated microglial RNA was determined using a NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA). One μg of isolated RNA was then converted into cDNA using an iScript cDNA Synthesis Kit (Bio-Rad, Hercules, CA, USA), adhering to the manufacturer's instructions, in a Mastercycler thermocycler (Eppendorf, Westbury, NY, USA). One μl of the cDNA product was used for PCR amplification of the NOX1-4 and GAPDH transcripts, using rat specific primers included in Table 1.</p><!><p>The statistical analyses of the data were performed using either Student's t-test or one-way ANOVA followed by Tukey's post-hoc analysis for multiple comparisons. Results of the excitatory amino acid efflux assays were separately analyzed for the maximal release values under hypo-osmotic conditions, or integral release values during the whole duration of the hypoosmotic exposure. Because statistical significances of the data were very similar for both parameters, we presented only comparisons of maximal releases in the text and figures. Origin 6.0 (OriginLab, Northampton, MA) and Prism 5 (GraphPad, San Diego, CA) were used for statistical analysis.</p><!><p>In our initial experiments, we tested for the functional expression of a volume-sensitive Cl− permeability pathway in primary rat microglial cells using a whole-cell patch clamp electrophysiological approach. We hypothesized that microglia, as the majority of other cells, express VRACs. To activate VRACs, we induced cell swelling by exposing microglia to hypoosmotic medium (20.7% reduction in medium osmolarity), which elicited slowly developing, outwardly rectifying Cl− currents (Fig. 1a). The average current density in five successfully patched cells was 40.8 ±5.7 pA/pF at +40 mV. This was in the same order of magnitude as the VRAC current densities registered under identical experimental conditions in primary rat astrocytes (19.7 ±3.3 pA/pF, Abdullaev et al., 2006). The registered currents were similar to those defined for VRAC in other cell types in several respects. Similar to what is shown in Fig. 1b, in all five cells, swelling-activated Cl− currents were outwardly rectifying and showed time-dependent inactivation at +100 mV, although it was less pronounced than in other cell types. Importantly, these Cl− currents were completely and reversibly inhibited by DCPIB, a pharmacological compound that discriminates VRACs from other Cl− channels (Decher et al. 2001; Abdullaev et al. 2006).</p><p>To determine whether VRACs expressed in primary microglia are permeable to glutamate, we employed an excitatory amino acid efflux assay. Primary microglia were loaded with D-[3H]aspartate, a non-hydrolyzable analog of glutamate, and then suffused with isoosmotic or hypoosmotic media. The exposure of microglia to hypoosmotic medium (30% reduction in medium osmolarity) resulted in a significant release of D-[3H]aspartate, and this effect was potently suppressed by 20 μM DCPIB, suggesting that microglial VRACs are permeable to glutamate (Fig. 1c).</p><!><p>It has been demonstrated in various cell types, including astrocytes, that VRACs may be activated or modulated by the reactive oxygen species hydrogen peroxide (H2O2) (Shimizu et al. 2004; Varela et al. 2004; Haskew-Layton et al. 2005; Ren et al. 2008). To address the question of whether H2O2 impacts VRAC activity in microglia, we exposed microglia to exogenous H2O2 under isoosmotic and hypoosmotic conditions. In non-swollen cells, 500 μM H2O2 produced a small gradual increase in D-[3H]aspartate release levels (Fig. 2a, ∼2-fold increase above basal levels, p<0.05, repeated measures ANOVA). However, when applied in conjunction with hypoosmotic medium, H2O2 significantly potentiated the swelling-activated D-[3H]aspartate release from microglia by ∼70% (Fig. 2a). The effect of H2O2 was completely inhibited by DCPIB, suggesting that VRAC is the source (Fig. 2a). DCPIB on its own had no antioxidant properties, as verified in ROS assays in microglia and macrophages (data not shown). We further explored which concentrations of H2O2 are most effective in regulating VRAC activity. We found a biphasic effect of H2O2 with the maximal stimulation at 100 μM (Fig. 2b). At higher H2O2 concentrations, potentiation of excitatory amino acid release was diminished, which may suggest a redox-dependent inactivation of VRAC itself or its regulatory mechanism(s) (Fig. 2b). To assure that the effect of H2O2 is mediated by increases in VRAC activity, we additionally performed several electrophysiological experiments in which H2O2 was applied to microglia under isoosmotic or hypoosmotic conditions. Similar to glutamate release results, 100 μM H2O2 did not increase basal Cl− currents in non swollen cells (n=3, data not shown). However, when 100 μM H2O2 was applied in conjunction with hypoosmotic medium, after swelling-activated Cl− currents reached steady-state levels, there was additional 54% increase in current density (from 45.9 ±2.6 to 70.7±3.9 pA/pF, n=3, p=0.006).</p><!><p>To determine whether endogenous ROS can modulate VRAC activity, we stimulated microglia with zymosan (500 μg/mL). Zymosan is a desiccated preparation of yeast cell wall commonly employed to stimulate the activity of the ROS-producing enzyme NADPH oxidase (NOX) in immune cells (DeChatelet et al. 1975). We measured endogenous ROS production using a nitroblue tetrazolium (NBT) assay, which detects O2− via quantification of NBT reduction to formazan (Choi et al. 2006). Stimulation of microglia with zymosan strongly increased NBT reduction (Fig. 3a). The effect of zymosan was evident at 10 minutes, but roughly tripled after a 30-minute incubation (data not shown). In all the subsequent experiments we used a 30-minute incubation to increase assay sensitivity. Several previous studies found increased ROS production upon application of hypoosmotic media (Lambert 2003; Ortenblad et al. 2003; Varela et al. 2004). In our microglial experiments we failed to register changes in NBT reduction under hypoosmotic conditions; also, hypoosmotic medium did not increase the zymosan-induced ROS production (data not shown). The effect of zymosan was blocked by DPI (10 μM), a potent NOX inhibitor (Bedard and Krause 2007) (Fig. 3a). This suggests that zymosan stimulates the production of ROS in microglia via NOX activation.</p><p>We further tested zymosan in the excitatory amino acid release assay. Zymosan was applied for 10 minutes prior to and during the 10-min hypoosmotic stimulation. In non-swollen cells, zymosan caused a small and transient increase in D-[3H]aspartate release (Fig. 3b, 2.4-fold increase above basal levels, p=0.021, marked by arrow). In swollen microglia, zymosan stimulation potentiated hypoosmotic d-[3H] aspartate release by ∼two-fold (Fig. 3b). As in the case of exogenous H2O2 (see Fig. 2a), 20 μM DCPIB suppressed zymosan-induced excitatory amino acid efflux both under isoosmotic and hypoosmotic conditions (Fig. 3b), confirming VRAC contribution. The zymosan-enhanced release of D-[3H]aspartate was also blocked with 10 μM DPI, under both isoosmotic and hypoosmotic conditions (Fig. 3c), suggesting NOX involvement.</p><p>Because the effect of zymosan in non-swollen cells was very small, in all the subsequent experiments we applied zymosan during exposure to hypoosmotic medium only. Such application produced a consistently stronger stimulation of volume-sensitive excitatory amino acid release (increased from ∼2-fold to ∼4-fold, compare Fig. 3c to Fig. 4b).</p><p>To further confirm that the effects of zymosan on VRAC activity are mediated by ROS, rather than by activation of ROS-independent intracellular signaling pathway(s), we applied MnTMPyP, which is a cell permeable superoxide dismutase (SOD) and catalase mimetic (Day et al. 1997). One-to-30 μM MnTMPyP dose-dependently reduced NBT-detectable levels of O2− in microglia stimulated with zymosan (Fig. 4a). However, even at the highest MnTMPyP concentration used, inhibition of ROS production incomplete (Fig. 4a). We have chosen 20 μM MnTMPyP concentration for the subsequent experiments in order to reduce risk of toxicity and assure specificity of the effects. 20 μM MnTMPyP significantly attenuated the zymosan-enhanced release of D-[3H]aspartate under hypoosmotic conditions (Fig. 4b), strongly suggesting that the effects of zymosan are mediated by ROS.</p><p>To further implicate NOX as the source of ROS involved in the regulation of VRAC activity in microglia, we employed apocynin, another commonly used inhibitor of NOX, which is more selective than DPI (Stolk et al. 1994). We used the concentration of 2 mM, because at lower concentrations apocynin showed limited efficacy in NBT assays (data not shown). Two mM apocynin decreased the zymosan-stimulated production of O2- in microglia by 75% (Fig. 5a) but, surprisingly, was completely ineffective against the zymosan-enhanced release of d-[3H] aspartate (Fig. 5b). Similarly, AEBSF, a protease inhibitor that is frequently used to block the NOX2 isoform (Bedard and Krause 2007), prevented the zymosan-activated ROS production, but was ineffective against zymosan-stimulated glutamate release (Supplementary Fig. 1).</p><p>In an attempt to resolve the incongruity between results with DPI, apocynin and AEBSF, we tested thioridazine, another NOX inhibitor. Thioridazine is a phenothiazine derivative, which blocks dopaminergic receptors, as well as all NOX isoforms via the mechanism similar to DPI (Serrander et al. 2007). In our experiments, thioridazine completely blocked the zymosan-stimulated production of O2− at 30 μM (Fig. 6a), and effectively prevented the zymosan-enhanced release of D-[3H]aspartate from swollen microglia (Fig. 6b).</p><!><p>Because apocynin and AEBSF were unexpectedly ineffective in attenuating the zymosan-enhanced release of D-[3H]aspartate, we conducted an analysis of literature on NOX pharmacology. Five NOX isoforms (NOX1-5) are expressed in humans, while only four NOX isoforms (NOX1-4) have been identified in rat (Kawahara et al. 2007). Apocynin and AEBSF block NOX1-3, but are ineffective against NOX4, while DPI and thioridazine effectively inhibit all NOX isoforms (Bedard and Krause 2007; Serrander et al. 2007). Therefore, our data suggest that NOX4 is involved in the ROS-mediated effect of zymosan on VRAC.</p><p>To determine which NOX isoforms are expressed in primary rat microglia, we probed for the mRNAs expression of all four NOX rat isoforms (NOX1-4). Using RT-PCR, we found that primary rat microglia express high levels of NOX2 along with lower levels of NOX1 and NOX4, but not NOX3 (Fig. 7). The expression of NOX4 fits well with the pharmacological profile of zymosan-enhanced D-[3H]aspartate release, suggesting that NOX4, and not NOX2, may be the source of ROS that positively regulates VRAC activity.</p><!><p>NOX2 is considered to be the main NOX isoform expressed in microglia and other phagocytic cells; it serves as a prominent source of ROS in immunological reactions (Lambeth 2004), and, thus, is expected to be activated in response to zymosan. Our pharmacological data imply that NOX4, and not NOX2, is the source of ROS which impacts VRAC activity. NOX1-3, but not NOX4, require protein kinase activity for the phosphorylation-dependent assembly of an active NOX complex (Lambeth 2004; Bedard and Krause 2007). Therefore, we pharmacologically manipulated PKC activity in an attempt to discern whether ROS production by NOX2 (and perhaps NOX1, which is also expressed in rat microglia) regulates VRAC.</p><p>First, we interrupted the generation of ROS via NOX1/2 by inhibiting PKC with the broad-spectrum PKC blocker Gö6983 (Gschwendt et al. 1996). One μM Gö6983 completely blocked the zymosan-stimulated production of O2- (Fig. 8a) but had no effect on the zymosan-enhanced release of d-[3H] aspartate (Fig. 8b).</p><p>We then took an opposite approach by stimulating PKC directly using phorbol 12-myristate 13-acetate (PMA), which has been demonstrated to fully activate NOX2, as well as NOX1 and NOX3 (Bedard and Krause 2007). 500 nM PMA was more effective in stimulating microglial O2− production than zymosan (Fig. 9a). However, at the same concentration, PMA failed to enhance the VRAC-mediated release of glutamate in microglia (Fig. 9b).</p><p>Together, the experiments presented in Figs. 8 and 9 strongly suggest that NOX2, the major microglial NOX isoform, and possibly NOX1, are not a critical source of ROS that regulate VRAC activity in zymosan-treated cells.</p><!><p>In the present work, we found that in microglia, the immunocompetent cells of the CNS, both exogenous and endogenously produced reactive oxygen species strongly enhance excitatory amino acid release via volume-regulated anion channels (VRACs). To stimulate endogenous ROS production by NADPH oxidases, we challenged microglial cells with zymosan, thereby mimicking pathogenic infection. The surprising finding of this study is that the major phagocytic NOX isoform, NOX2, although activated by zymosan, does not contribute to VRAC regulation. Instead, such regulation is likely mediated by NOX4, which—until the present study—has not been identified in microglial cells.</p><!><p>Microglial activation in response to pathogenic infection or trauma is important for the preservation of neural tissue (Kreutzberg 1996; Nakajima and Kohsaka 2001; Block and Hong 2005). When activated, microglia produce and release various inflammatory cytokines, reactive oxygen and nitrogen species, and glutamate (Nakajima and Kohsaka 2001; Block and Hong 2005; Block et al. 2007). Glutamate is the primary excitatory neurotransmitter in the CNS, which may induce excitotoxic tissue damage if its extracellular levels are not properly regulated (Choi 1988; Choi 1992). One of several potential pathways for glutamate release are VRACs (Kimelberg and Mongin 1998; Mongin and Kimelberg 2005; Malarkey and Parpura 2008). In microglial cells, swelling-activated anion channels, which resemble VRAC, have been identified and suggested to play important roles in volume regulation, cell proliferation, and phagocytosis (Schlichter et al. 1996; Ducharme et al. 2007). Importantly, microglial swelling-activated Cl− channels are permeable to glutamate and aspartate with permeability values, relative to Cl−, of ∼0.18 and ∼0.22, respectively (Ducharme et al. 2007). However, Ducharme et al. (2007) have found that microglial swelling-activated anion channels are not inactivated at positive potentials and have much smaller unitary conductance of ∼1-3 pS in comparison to a prototypical VRAC. Conversely, our electrophysiological and pharmacological data suggest that rat microglia express classical VRACs. Several characteristics that support such a notion include: (i) outward rectification of the swelling induced Cl− currents, (ii) time-dependent current inactivation at positive potentials, and (iii) the sensitivity of swelling-activated Cl− currents to the selective VRAC blocker DCPIB. DCPIB has been found to discriminate VRACs from other cloned to-date Cl− channels (Decher et al. 2001). Additional studies are needed to determine if the method of cell preparation or differences in recording solutions explain variation between present electrophysiological findings and those in the work of Schlichter and co-workers (Ducharme et al. 2007). Along with our electrophysiological data, the inhibitory effect of DCPIB on amino acid release suggests that microglial VRACs serve as a major route for glutamate release, and in such a way may profoundly impact neural tissue function. This may especially be the case under pathological conditions associated with cell swelling or in the processes of microglial migration, proliferation, and phagocytosis, which likely involve enhanced VRAC activity (Schlichter et al. 1996; Eder 2005; Ducharme et al. 2007).</p><!><p>The major objective of this study was to determine if ROS may trigger glutamate release from microglial cells via glutamate-permeable VRACs. In several cell types, it has been demonstrated that ROS activate or positively modulate VRACs (Browe and Baumgarten 2004; Shimizu et al. 2004; Varela et al. 2004; Haskew-Layton et al. 2005; Ren et al. 2008), as well as similar volume-sensitive taurine permeability pathway (Ortenblad et al. 2003; Lambert 2003). Because activated microglia abundantly generate ROS, these cells represent an ideal subject for studying redox regulation of VRACs, and the functional impact of such regulation on cell physiology and function. In our experiments, application of exogenous H2O2 enhanced hypoosmotic medium-induced D-[3H]aspartate release and VRAC Cl− currents via a DCPIB-sensitive pathway, acting synergistically with cell swelling. Consistent with our previous findings in astrocytes (Haskew-Layton et al. 2005), H2O2 was essentially ineffective in non-swollen cells. Induction of endogenous ROS production with zymosan produced limited VRAC activation in non-swollen cells, and potently increased glutamate release by 2-5-fold from microglial cells exposed to hypoosmotic media. These data suggest that ROS preferentially modulate VRACs only when these channels are already active. The magnitude of D-[3H]aspartate release from non-swollen cells exposed to H2O2 or zymosan is too small to have physiological significance. Under isoosmotic conditions set in our experiments, VRACs are inactive. Therefore, we subject microglial cells to large hypotonic stimuli in order to activate the volume-sensitive Cl− currents and glutamate fluxes. Such hypoosmotic gradients do not exist in the brain, even in pathology. However, microglial VRACs may be activated under various physiological and pathological conditions without changes in extracellular osmolarity. Such a notion is supported by the observations that nonselective VRAC blockers inhibit proliferation, phagocytosis and shape transitions in microglia (Schlichter et al. 1996; Eder et al. 1998; Ducharme et al. 2007).</p><p>Zymosan is commonly used to stimulate ROS production in immune cells, including microglia, and this effect is mediated via activation of NADPH oxidases (Colton and Gilbert 1987; Sankarapandi et al. 1998). In our experiments, the effects of zymosan on D-[3H]aspartate release via VRACs were inhibited either by the antioxidant MnTMPyP or the NOX blockers DPI and thioridazine, which strongly suggests that they are mediated by ROS originating form NOX.</p><!><p>The paradoxical finding of the present work was insensitivity of zymosan-induced VRAC activation to the NOX inhibitors apocynin and AEBSF. Apocynin, DPI, AEBSF, and thioridazine all potently suppressed zymosan-induced superoxide production. However, only DPI and thioridazine inhibited VRAC-mediated glutamate release from zymosan-treated cells. This discrepancy may be explained by suggesting that the effect of zymosan on VRAC involves atypical NOX4. NOX1-3 isoforms require phosphorylation-dependent assembly of several cytosolic subunits in order to be catalytically active, while NOX4 activity does not require the same proteins and is regulated via alternative mechanisms (Bedard and Krause 2007, see Fig. 10). Apocynin and AEBSF suppress the activity of NOX2 (and related NOX1 and 3) by inhibiting the assembly of regulatory subunits (Stolk et al. 1994; Bedard and Krause 2007). However, these compounds are ineffective inhibitors of NOX4 (Serrander et al. 2007). DPI and thioridazine, on the other hand, inhibit all NOX isoforms, including NOX4, by interrupting electron flow in the catalytic subunit (Bedard and Krause 2007; Serrander et al. 2007). Although DPI and thioridazine are not specific for NOXs, and also block other electron-transferring enzymes, they are the only potent NOX4 inhibitors available to-date.</p><p>NOX4 is a recently discovered enzyme, whose expression in microglial cells has not been reported until the present study (Bedard and Krause 2007). Taken together, the presence of NOX4 messenger RNA and our pharmacological data strongly support the hypothesis of NOX4 involvement in regulation of VRAC. A similar idea has been previously put forth by Lambert and colleagues, who suggest that NOX4 is critical for activation of swelling-sensitive taurine release pathway in mouse NIH 3T3 fibroblasts (Friis et al. 2008).</p><p>It should be noted, however, that NOX4 is not a major NOX isoform in microglia, nor is it known to be activated by zymosan. In microglia, zymosan stimulates generation of ROS via activation of phagocytic NOX2 (Sankarapandi et al. 1998). It was not clear to us why NOX2-derived ROS were not contributing to VRAC activation by zymosan. In order to address this question, we stimulated NOX2 by treating microglial cells with PKC activator PMA, thereby promoting the assembly of catalytically active NOX2 complex (Bedard and Krause 2007). PMA potently enhanced generation of O2−, but it did not affect the VRAC activity. Conversely, the potent PKC inhibitor Gö6983 completely suppressed O2− generation in zymosan-treated cells, but did not affect VRAC activity. These data unequivocally prove that NOX2-derived ROS are not critical for VRAC regulation by zymosan under our experimental conditions.</p><p>The surprising lack of effect of NOX2-derived ROS may be explained by the location of ROS production and/or by the peculiarity of NOX4 biochemistry. As shown in Fig. 10, NOX2 is predominantly present in plasmalemmal and phagocytic membranes and secretes superoxide to the extracellular space or into the phagosome (Bedard and Krause 2007). In contrast, NOX4 is thought to be predominantly associated with the endoplasmic reticulum and/or endosomes and generates disproportionally higher amounts of intracellular H2O2 compared to other NOXs (Martyn et al. 2006; Bedard and Krause 2007; Serrander et al. 2007). Therefore, NOX4 may create sufficiently high local levels of H2O2 that stimulate VRACs. Our attempts to quantify intracellular H2O2 production in zymosan-treated cells using a fluorescent probe 5-chloromethyl-2′7′-dichlorofluorescein diacetate (CM-H2DCFDA) were unsuccessful because of the turbidity of zymosan solutions.</p><!><p>Microglia activation is a common feature of many neurological disorders. Although necessary for preservation of neural tissue, microglial responses may have deleterious effects, which are largely related to the excessive production of ROS (Kreutzberg 1996; Nakajima and Kohsaka 2001; Block and Hong 2005). Genetic deletion of the ROS-producing enzyme NOX2 or its regulatory subunits, or overexpression of ROS-scavenging enzymes, are protective in animal models of cerebral ischemia, Alzheimer's and Parkinson's diseases, and methamphetamine neurotoxicity (Yang et al. 1994; Walder et al. 1997; Deng and Cadet 2000; Zhang et al. 2004; Park et al. 2005). Tissue damage in the same animal models can also be prevented by antagonists of the NMDA subtype of glutamate receptors (Sonsalla et al. 1989; Zeevalk et al. 1994; Lipton 1999). Therefore, oxidative stress and glutamate toxicity are likely related processes. It is well known how glutamate promotes oxidative and nitrosative stress (Beckman and Koppenol 1996; Lipton 1999). Whether the opposite is true, i.e., if ROS production may lead to enhanced glutamate release, is less clear. Previous studies implicated the cystine/glutamate antiporter and connexin hemichannels in the ROS- and inflammation-induced glutamate release (Piani and Fontana 1994; Barger and Basile 2001; Takeuchi et al. 2006; Fogal et al. 2007). Our present and published work suggests that, in microglia, astrocytes and possibly other types of neural cells, oxidative stress may promote pathological glutamate release via the glutamate-permeable anion channel VRAC, and perhaps other permeability pathways (Haskew-Layton et al. 2005; Liu et al. 2006). Although our present in vitro data point to a key role for NOX4, in intact tissue NOX2 may be as important. A recent report by Z. Ren et al. (2008) implicated NOX2 in regulation of VRAC activity by angiotensin II in cardiac myocytes. NOX2-derived superoxide is converted extracellularly and intracellularly to H2O2, which in intact tissue is capable of accumulating in pathological quantities, regardless of the initial source (Halliwell 2006). In our experimental paradigm, NOX2-generated ROS may be consumed in phagosome or removed with perfusion, while NOX4-derived H2O2 may reach high intracellular levels (see Fig. 10). Overall, our study proposes one possible mechanism that links microglial ROS production to elevated glutamate release, which involves the anion channel VRAC. Such release may contribute to glutamate-associated neuronal damage seen in various models of neural pathologies.</p><!><p>(a) Representative electrophysiological recording of Cl− currents in a single cultured microglial cell exposed to hypoosmotic medium. Cells were held at 0 mV and alternative one-second pulses to ±40 mV were applied to measure activation of Cl− currents every 15 seconds. The VRAC blocker DCPIB (20 μM) was applied after Cl− currents reached steady-state level. (b) Cl− current responses to step pulses from -100 to +100 mV in 20 mV increments applied after swelling-activated Cl− currents reached steady-state level. Representative traces of electrophysiological recordings in 5 cells. (c) Effect of the VRAC blocker DCPIB (20 μM) on hypoosmotic medium-induced D-[3H]aspartate release from microglia. Cells were loaded for 3 hours with D-[3H]aspartate. Extracellular isotope was then washed and cells were placed in a Lucite perfusion chamber. Data represent the mean values ±SE of six experiments. *** p < 0.001, DCPIB vs. control hypoosmotic medium (Hypo).</p><!><p>(a) Primary microglia were exposed to H2O2 (500 μM) 10 minutes prior to and during the 10-minute suffusion with hypoosmotic medium in the presence or absence of the VRAC blocker DCPIB (20 μM). Data represent the mean values ±SE of six experiments. **p<0.01, H2O2 vs. Hypo, and p<0.001, H2O2 vs. H2O2+DCPIB. (b) The effect of varying concentrations of H2O2 on the release of D-[3H]aspartate from microglia under hypoosmotic conditions. These experiments were conducted in a similar fashion as in (a). For clarity only maximal D-[3H]aspartate release data are shown. Mean values ±SE of five to ten experiments in each group are presented. *p<0.05, **p<0.01, ***p<0.001, H2O2 vs. hypoosmotic control; one-way ANOVA with Tukey's post-hoc analysis.</p><!><p>(a) Effects of zymosan and the NOX inhibitor DPI on microglial production of O2−. Primary microglial cells were exposed to hypoosmotic media additionally containing zymosan (500 μg/mL) and/or DPI (10 μM) for 30 minutes at 37°C. O2- production was measured as the reduction of NBT (see Materials and methods for details). ***p<0.001 Zymosan vs. Control; ###p<0.001 Zymosan vs. Zymosan+DPI. (b) Effect of zymosan on microglial D-[3H]aspartate release. Microglia were exposed to zymosan (500 μg/mL) for 10 minutes prior to and during 10-minute application of hypoosmotic medium. Arrow indicates a transient increase in D-[3H]aspartate release upon zymosan application under isoosmotic conditions. In a separate set of experiments, zymosan was applied in combination with 20 μM DCPIB. The mean values ±SE of six experiments are presented. ***p<0.001 Hypo vs. Zymosan. ###p<0.001 Zymosan+DCPIB vs. Hypo or Zymosan alone. (c) Effects of zymosan (500 μg/mL) on hypoosmotic medium-induced D-[3H]aspartate release in the presence or absence of DPI (10 μM). The design of these experiments was similar to those presented in (b); DPI was present 30 min before and during application of hypoosmotic medium. For clarity only maximal hypoosmotic release values are presented. Data are the mean values ±SE of 5-6 experiments. ***p<0.001, Zymosan vs. Hypo; ###p<0.001, Zymosan vs. Zymosan + DPI.</p><!><p>(a) The effect of the ROS scavenger MnTMPyP on intracellular O2− levels in microglia. Primary microglial cells were exposed for 30 minutes to hypoosmotic media containing zymosan (500 μg/mL) and various concentrations of MnTMPyP. O2− production was measured as the reduction of NBT. Representative of two experiments. Data are the mean values ±SE of three independent measurements. *p<0.05, **p<0.01, p<0.001 vs. Hypo; #p<0.05, p<Zymosan+MnTMPyP vs. Zymosan control. (b) The effect of MnTMPyP (20 μM) on the zymosan-induced enhancement of hypoosmotic D-[3H]aspartate release. The mean values ± SE of five experiments are shown. **p<0.01, ***p<0.001 vs. Hypo; ##p<0.01, Zymo.+MnTMPyP vs. Zymosan. In these experiments one-way ANOVA with Tukey's post-hoc analysis was performed after logarithmic transformation of the data.</p><!><p>(a) The effects of apocynin on zymosan-induced O2- production in microglia. Microglial cells were exposed to hypoosmotic media additionally containing zymosan (500 μg/mL) and/or apocynin (2 mM). Apocynin was included in the media 30 minutes before and during 30-minute incubation with zymosan and NBT. Data are the mean values ±SE of four experiments. ***p<0.001, vs. Hypo; ###p<0.001 Zymosan vs. Zymo. +Apocynin. (b) The effect of zymosan on hypoosmotic medium-induced D-[3H]aspartate release in the presence or absence of 2 mM apocynin. The mean values ±SE of five experiments are shown. ***p<0.001 vs. Hypo. In these experiments one-way ANOVA with Tukey's post-hoc analysis was performed after logarithmic transformation of the data.</p><!><p>(a) The effect of various concentration of thioridazine on zymosan-induced production of O2− in microglia exposed to hypoosmotic medium. Thioridazine was applied for 30 minutes in conjunction with 500 μg/mL zymosan and NBT. The mean values ±SE of three independent measurements in a representative experiment are shown. *p<0.05, ***p<0.001, vs. Hypo, ###p<0.001 vs. Zymosan control. (b) The effect of 20 μM thioridazine on zymosan-induced enhancement of hypoosmotic D-[3H]aspartate release. The data are the mean values ±SE of five experiments. ***p<0.001 vs. Hypo. ##p<0.01 Zymozan vs. Zymo.+Thioridazine.</p><!><p>This figure shows a representative negative image of ethidium bromide staining of PCR products for transcripts of various NOX isoforms. PCR was performed after reverse transcriptase (RT) processing of the total RNA samples isolated from primary rat microglia. Total RNA was treated with DNAse in order to remove genomic DNA contamination. Lanes 2-6 show RT-PCR products for rat NOX1-4 and GAPDH in microglial samples, as indicated. NOX3 was not expressed in microglia. In the reaction presented in lane 7, PCR was performed using NOX4 primers, but the RT product was omitted. Lane 8 shows NOX4 RT-PCR product from primary rat endothelial cells (positive control). Lane 1 contains 50-bp MW ladder.</p><!><p>(a) Effect of Gö6983 on the zymosan-induced of O2- production (reduction of NBT). Microglial cells were exposed for 30 minutes to hypoosmotic medium additionally containing zymosan (500 μg/mL) and the broad-spectrum PKC inhibitor Gö6983 (1 μM). Data are the mean values ±SE of three independent measurements in a representative of three experiments. **p<0.01, ***p<0.001 vs. Hypo; ###p<0.001, Zymosan vs Zymosan+Gö6983. (b) The effect of zymosan on hypoosmotic D-[3H]aspartate release in the presence or absence of 1 μM Gö6983. Data are the mean values ±SE of six to nine experiments. ***p<0.001,. vs. Hypo.</p><!><p>(a) Effects of PMA (500 nM) or zymosan (500 μg/mL) on microglial O2− production measured as the reduction of NBT. Data are the mean values ±SE of four independent measurements in a representative of three experiments. ***p<0.001 vs. Hypo. (b) PMA (500 nM) had no effect on hypoosmotic D-[3H]aspartate release from primary microglia. Data are the mean values ±SE of six experiments</p><!><p>Zymosan activates TLR2 and perhaps another unidentified subclass of the TLR receptors. TLR2 activation stimulates cytosolic PKC via several additional steps (not shown). PKC phosphorylates regulatory cytosolic subunit p47phox (see NOX2 box on the right). Phosphorylation of p47phox prompts assembly of the enzymatically active NOX2 complex, which includes membrane proteins NOX2 (gp91phox) and p22phox, as well as cytosolic p47phox, p67phox, p40phox, and the small GTPase rac. An active NOX2 complex resides in plasmalemmal and phagocytic membranes and produces O2−, which is released to the extracellular space or into phagosome. An alternative signaling pathway causes activation of NOX4 via unidentified enzymatic steps. NOX4 is likely localized in endoplasmic reticulum and/or endosomes and does not require cytosolic regulatory subunits for its activity (see NOX4 box on the right). NOX4 produces predominantly H2O2, rather than O2−. The difference in the location of the ROS production and/or the nature of generated ROS likely determines differential contribution of NOX4 and NOX2 to the activation of VRAC by zymosan. Abbreviations: ER, endoplasmic reticulum; ES-SOD, extracellular superoxide dismutase; NOX, NAD(P)H oxidase, PKC, protein kinase C, TLR, Toll-like receptors. See text for additional details.</p><!><p>Species specific primers for RT-PCR analysis of NOX isoform expression in microglia</p>

PubMed Author Manuscript

Mapping of possible prion protein self-interaction domains using peptide arrays

BackgroundThe common event in transmissible spongiform encephalopathies (TSEs) or prion diseases is the conversion of host-encoded protease sensitive cellular prion protein (PrPC) into strain dependent isoforms of scrapie associated protease resistant isoform (PrPSc) of prion protein (PrP). These processes are determined by similarities as well as strain dependent variations in the PrP structure. Selective self-interaction between PrP molecules is the most probable basis for initiation of these processes, potentially influenced by chaperone molecules, however the mechanisms behind these processes are far from understood. We previously determined that polymorphisms do not affect initial PrPC to PrPSc binding but rather modulate a subsequent step in the conversion process. Determining possible sites of self-interaction could elucidate which amino acid(s) or amino acid sequences contribute to binding and further conversion into other isoforms. To this end, ovine – and bovine PrP peptide-arrays consisting of 15-mer overlapping peptides were probed with recombinant sheep PrPC fused to maltose binding protein (MBP-PrP).ResultsThe peptide-arrays revealed two distinct high binding areas as well as some regions of lower affinity in PrPC resulting in total in 7 distinct amino acid sequences (AAs). The first high binding area comprises sheep-PrP peptides 43–102 (AA 43–116), including the N-terminal octarepeats. The second high binding area of sheep-PrP peptides 134–177 (AA 134–191), encompasses most of the scrapie susceptibility-associated polymorphisms in sheep. This concurs with previous studies showing that scrapie associated-polymorphisms do not modulate the initial binding of PrPC to PrPSc. Comparison of ovine – and bovine peptide-array binding patterns revealed that amino acid specific differences can influence the MBP-PrP binding pattern. PrP-specific antibodies were capable to completely block interaction between the peptide-array and MBP-PrP. MBP-PrP was also capable to specifically bind to PrP in a Western blot approach. The octarepeat region of PrP seems primarily important for this interaction because proteinase K pre-treatment of PrPSc completely abolished binding.ConclusionBinding of MBP-PrP to PrP-specific sequences indicate that several specific self-interactions between individual PrP molecules can occur and suggest that an array of interactions between PrPC-PrPC as well as PrPC-PrPSc may be possible, which ultimately lead to variations in species barrier and strain differences.

mapping_of_possible_prion_protein_self-interaction_domains_using_peptide_arrays

Background<!>MBP-PrP expression and analysis<!><!>Binding domains of ovine PrP<!><!>Ovine versus bovine peptide-array<!>Peptide-array controls<!><!>Antibody blocking of peptide-array binding pattern<!><!>MBP-PrP mediated detection (reverse detection) of PrP in Western blot<!><!>Binding domains of ovine PrPC<!>Ovine versus bovine peptide-array<!>Antibody blocking of peptide-array binding pattern<!>MBP-PrP mediated detection (reverse detection) of PrP in Western blot<!>Conclusion<!>MBP-PrP construct<!>MBP-PrP expression and purification<!>Peptide-array analysis<!>Production of monoclonal antibody 6C2<!>Antibody aggregation test<!>PrPSc purification and analysis<!>Reverse detection assay<!>Abbreviations<!>Authors' contributions<!>Acknowledgements

<p>Transmissible spongiform encephalopathies (TSEs) are fatal neurodegenerative disorders characterized by formation and accumulation of partially protease resistant prion protein (PrPSc) mainly in tissues of the central nervous system. TSEs (or prion diseases) include (among others) familial, sporadic and variant Creutzfeldt-Jacob disease in humans, bovine spongiform encephalopathy (BSE) in cattle, and scrapie in sheep. Formation of PrPSc is a posttranslational process and involves refolding (conversion) of the host-encoded prion protein (PrPC) into partially protease resistant forms (PrPSc) [1]. Since no other proteins are known to be involved in this conversion, the existence of a specific and probably efficient self interaction between PrP molecules must be considered.</p><p>The molecular mechanism involved in PrP conversion is not well understood, but polymorphisms in PrP have been shown to be of importance in both interspecies and intraspecies transmissibilities [2] and cell-free conversion of PrPC provides an excellent in vitro model in which relative amounts of produced proteinase K (PK) resistant PrP reflect important biological aspects of TSEs at the molecular level [2-9]. Whereas differences in susceptibility of- and transmissibility in sheep can largely be explained at the molecular level by the effects of single polymorphisms in PrPC or PrPSc on PrP conversion [6,10-12], the exact molecular mechanism of disease development modulation by polymorphisms is still unknown, however we previously showed that disease associated polymorphisms do not affect the initial binding of PrPC to PrPSc [13]. Hölscher et al showed by deletion of residues 114–121 (mouse PrP) the necessity of the highly amyloidogenic AGAAAAGA motif in conversion of PrPC to PrPSc [14]. Many other studies have revealed the importance of the PrP regions encompassing amino acid sequence (AA) 90–120 (which confirms the importance of AGAAAAGA) [15-17] and 132–156 [8,15,18-27]. However, to our knowledge no attempts have been made to systematically map all possible AA involved in PrP interaction (During review of this manuscript a study with complementary results directed at the identification of regions of PrPC that tightly bind to PrPSc by using a limited set of sequential 24-mer polypeptides motif grafted onto an antibody was published [28]. Our study has its focus however, on systematical domain mapping at the single amino acid level by using a complete set of overlapping 15-mer PrP derived peptides). In order to elucidate which AA of PrP capable of interaction are involved in the primary interaction of PrPC to PrPSc, a peptide-array based on linear PrP sequences comprising the complete PrP sequence was utilized to determine which residues of PrP are capable of interacting with PrPC.</p><!><p>Expression of maltose binding protein N-terminally fused to PrP (MBP-PrP) revealed a mainly soluble recombinant MBP-PrP of approximately 70 kDa (Fig. 1, lanes 1 & 2) and is readily detected in Western blot using a PrP-specific antibody (9A2, Fig. 1B) or a MBP specific antibody (Fig. 1C). The MBP-PrP fusion-protein could be purified using the amylase-resin column and the naked PrP protein could be obtained by digestion with protease Factor Xa, indicating accessible folding (Fig. 1, lanes 3–6). After 24 hours approximately 45% of MBP-PrP was digested by factor Xa, however when aided by addition of 0.01% SDS factor Xa completely digested MBP-PrP within 24 hours (data not shown). Monoclonal (9A2 and 94B4, Table 1) and polyclonal (R521, epitope AA100–102 [29,30]) PrP-specific antibodies (Table 1) with specificity for epitopes dispersed throughout the PrP-protein detected MBP-PrP in Western blot. MBP expressed without additional fusion protein (PrP), which frequently served as negative control in this study, was also of homogeneous quality (Fig 1, lane 7) and of expected size (MBP-β-gal α fragment, 50.8 kDa) which is somewhat larger (as expected) than MBP cleaved from the fusion protein after factor Xa digestion (42.5 kDa). Though we did not study its physical state, the soluble MBP-PrP product used is most likely in a monomeric or low oligomeric state representative for PrPC, with a secondary structure that is high in alpha helix and random coil and low in beta-sheet [31,32].</p><!><p>Monoclonal antibody overview</p><p>a amino acid sequence of the reported antibody epitope using peptide mapping techniques</p><p>b position of the antibody epitope(s) in the ovine protein sequence</p><p>c concentration at which the antibody completely blocked binding on the peptide-array</p><p>d publication in which the epitope mapping results are reported</p><p>e epitope of octarepeat (partially); occurs as 5 respectively 6 successive sequences in ovine and bovine PrP</p><p>Analysis of MBP-PrP and MBP expression and MBP-PrP digestion by Factor Xa. Lane 1 contains untreated MBP-PrP, whereas lane 2 contains a mock digestion of MBP-PrP. MBP-PrP was digested with 1% w/w factor Xa and during digestion samples were taken at 2, 4, 7 and 24 hours (lanes 3,4,5 and 6 respectively). All samples were run on SDS-PAGE and the gel was stained with Sypro Orange (total protein stain, panel A) before western blotting and subsequent immunodetection using either a PrP-specific monoclonal antibody (9A2, panel B) or MBP-specific monoclonal antibody (α-MBP, panel C). Expression of MBP, expressed from the pMAL-c2X vector with no insert (MBP-β-gal α fragment), was analyzed by Western Blot using either 9A2 or α-MBP (lane 7, panel A & C respectively)</p><!><p>Using solid-phase arrays of 15-mer overlapping peptides systematically covering the whole mature part of PrPC, MBP-PrP was allowed to bind with the peptide-array, with the prospect that this would yield information on interaction sites between its PrP moiety and the linear peptides. Indeed interaction between the individual PrP sequences (peptides) and MBP-PrP was sufficient for immunodetection, resulting in a reproducible binding pattern (Fig. 2, line graph). This binding pattern, expressed in relative density values (Fig. 2, column graph), was characterized by two distinct high binding areas (peptides 43–102 and 134-177 respectively) as well as some lower binding areas. Analysis of the correlating peptide sequences revealed that these areas usually were characterized by consensus sequences which suggested the existence of the following interaction domains for the mature part of PrPC (Fig 3). Two consecutive binding peaks with peptides 22–28 + 29–33 (Fig. 3A) have [33-GWNTG-37] (ovine protein sequence position used throughout) as their consensus domain, followed by two consecutive minor binding peaks with peptides 35–38 + 39–42 (Fig. 3B) with [42-PGQGSPGG-49] as the common domain. The first high binding area is comprised of peptides 43–102, and encloses only two likely consensus domains: on the one hand peaks 43–52 (Fig. 3C), 53–60 (Fig. 3D), 61–68 (Fig. 3E), 69–77 (Fig. 3G) and 78–87 (Fig. 3H) each recognized an octarepeat with [PxGG, x = Q or H] as the consensus domain and on the other hand peaks 90–93, 94–97, 98–102 (Fig. 3I) with [102-WNK-104] as a common domain. The second high binding area is comprised of peptides 134–177 and encloses three likely consensus sequences. The shared sequence for peptides 134–136, 137–140 (Fig. 3J) is [140-PLIHFGNDY-148], for peptides 141–151 (Fig. 3K, AA152–154) and 153–155, 156–158, 59–164 (Fig. 3L, AA165–167) is [YYR] and for peptides 165–168, 170–173, 174–177 (Fig. 3M) is [177-NFV-179] respectively. The remaining lower binding areas also enclose three likely binding domain consensuses; [183-VNITVKQHTVT-193] for peptides 179–183 (Fig. 3N), [192-TTTTKGENFT-202] for peptides 188–193 (Fig. 3O) and [225-SQAY-228] for peptides 214–217, 219–221, 222–225 (Fig. 3P).</p><!><p>Peptide-array binding pattern of MBP-PrP. Dual plot of MBP-PrP binding to the ovine PrP peptide-array. The relative density value (r.d.v.) was calculated by dividing the optical density value (o.d.v.) by the background and binding was considered relevant when at least 3 consecutive peptides showed binding values of at least 3 times the background. The unprocessed optical density values (left X-axis) of each peptide (Y-axis, peptide number) are plotted in the graph, while relative density values (right X-axis) of each peptide are plotted in the column graph.</p><p>Correlation of relative binding pattern, peptide amino acid sequence and species variation between ovine and bovine PrP. Horizontal bars represent r.d.v of each peptide in a binding region with the position of each peptide indicated for ovine (left numbers) and bovine (right numbers) peptide array. Amino acid sequences are the ovine peptide sequences of all peptides in designated binding regions. The consensus domains found within each binding region for the ovine sequence are boxed and the consensus domains of the corresponding peaks found with the bovine peptide-array are in bold font. Blue bars represent r.d.v of ovine peptide-array, red bars r.d.v. of bovine peptide-array. Substitutions in bovine PrP are mentioned at right top side of each panel (panels H, I, J, K, N).</p><!><p>To further assess the extent of the specificity of the binding pattern found, MBP-PrP was also tested against a bovine PrP peptide-array. This yielded a rather similar binding pattern compared to the results with ovine PrP peptide array but with some differences. The binding pattern on the bovine peptide-array (Fig. 3, red bars) was compared to that on the ovine peptide-array (Fig. 3, blue bars). As expected an extra octarepeat was found (Fig. 3F), and of the six amino acid differences between the ovine – and bovine peptide arrays only two resulted in a difference in binding. Binding with the peptides containing the ovine to bovine substitutions S98A (ovine numbering used throughout, Fig. 3H–I), S146A (Fig. 3J) and Y158H (Fig. 3K) were comparable on both the ovine and bovine peptide-array, whereas the S100G (Fig. 3I) and Q189E (Fig. 3N) did result in altered binding patterns. No binding was found with peptides containing the I208M substitution (data not shown). Some differences in binding without a direct apparent reason were observed. Differences in the relative level of binding was observed with peptides 165–177 (Fig. 3M, bo# 173–785), 187–193 (Fig. 3O, bo# 195–201) and 220–222 (Fig. 3P, bo# 227–229), but these differences did not translate in differences in the determined consensus domains. However, binding with the array of bovine peptides 35–42 (Fig. 3B) remained below the cutoff value (3 times background), whereas low binding with peptides 103–105 (Fig. 3I) was observed with the bovine peptide-array but not with the ovine peptide-array.</p><!><p>Several control tests were carried out to determine the viability of the peptide-array to obtain PrP-specific binding patterns. Only minor non-significant differences in binding pattern were seen as a result of varying concentration of MBP-PrP (except for the expected difference in optical density value [o.d.v.]), storage buffers for MBP-PrP, peptide synthesis methods, or peptide-array batches (Fig. 4). Also, no significant binding was observed with each separate antibody or in the combination used for detection (thus in the absence of MBP-PrP). Furthermore no binding of MBP with the PrP peptide-array was observed. Interestingly the o.d.v. decreased after prolonged storage of MBP-PrP in PBS + 0.1% SB3–14. Further examination of the isolate showed that MBP-PrP had precipitated, indicating that interaction between the peptides and MBP-PrP only occurs when the latter is soluble. Furthermore, MBP-PrP was tested on an unrelated peptide-array containing overlapping peptides covering the sequence of VP2 of canine parvovirus yielding not any significant binding domains. All these controls confirm that binding of MBP-PrP to the PrP-peptides was as a result of the PrP moiety of MBP-PrP, and that this binding was PrP-specific.</p><!><p>Peptide array controls. Optical density value plot of MBP-PrP isolate 1 measured on minicard 1 (black line), MBP-PrP isolate 2 on minicard 1 (black dotted line), and MBP-PrP isolate 1 on minicard 2 (white line) and of MBP (background, grey dotted line) on peptide-array. For differences between isolates, compare black and white line. For differences between peptide synthesis batches compare black and black dotted line.</p><!><p>To find a correlation with structural properties, the relative binding pattern of MBP-PrP on the peptide array was compared to the Kyte-Doolittle hydrophilicity plot of mature PrPC revealing a high correlation between hydrophilic (exposed) regions of PrPC and binding pattern regions (Fig. 5). Even though the correlation was not absolute, it was necessary to determine if the binding pattern could be blocked with PrP-specific antibodies. Therefore several monoclonal antibodies with epitopes at different sites throughout PrPC (Table 1, Fig. 5) were tested. Pre-incubation (1 hour at room temperature) of MBP-PrP with all the tested PrP-specific monoclonal antibodies resulted in a concentration-dependant blocking of MBP-PrP binding over the whole set of PrP-peptides, albeit at different antibody concentrations (Table 1). No blocking of the MBP-PrP binding pattern occurred after pre-incubation with the unrelated Mycobacterium specific antibody M7. To ensure that blocking of the binding pattern is a result of immune-complex formation between antibody and MBP-PrP and not of incidental aspecific aggregation of MBP-PrP, a Mab-aggregation test for each PrP-specific antibody was carried out. Comparative amounts of MBP-PrP and antibody (necessary for blocking) were incubated. The soluble and insoluble fraction were separated by centrifugation at 20.000 × g and analyzed on SDS-PAGE, resulting in 75 ± 16% of MBP-PrP and 85 ± 7% of antibody detected in the supernatant fraction (data not shown), indicating that if an immune-complex is formed this complex remains soluble. Therefore formation of a soluble immune-complex must be responsible for loss of binding in the peptide-array instead of diminished binding as a result of aspecific aggregation. In addition preliminary results indicate that some selected peptides are also capable of blocking MBP-PrP binding to the peptide-array, confirming that binding of MBP-PrP to the PrP-peptides was as a result of the PrP moiety of MBP-PrP and PrP-specific.</p><!><p>Overview of PrPC secondary structures and antibody epitopes versus peptide-array binding pattern and Kyte-Doolittle hydrophilicity plot. Schematic representation of PrPC showing signal sequences, β-sheets (S1, S2), α-helices (H1, H2, H3), disulfide bridge site (S-S), glycosylation sites (CHO) and relative positions of the antibodies used in this study. The sequence of PrP is lined up with both the Kyte-Doolittle hydrophilicity plot (negative = hydrophobic and positive = hydrophilic) and the relative binding pattern found with the ovine peptide-array.</p><!><p>To further confirm the specificity of the observed PrP-PrP interaction, MBP-PrP was used as a detector in Western blot to further study its affinity towards intact PrPC in a brain homogenate. MBP-PrP could be used to detect recombinant His-tagged PrP (Fig. 6B, left panel lane marked HP) and intact PrP in both scrapie positive and negative brain homogenates (Fig. 6A, left panel lanes marked nt), albeit with a lower sensitivity under the standard Western blot conditions using monoclonal antibody 9A2 (Fig. 6A, compare left and right panel). MBP-PrP seems to preferably detect the un-glycosylated PrP in the scrapie negative brain homogenate (in contrast to the scrapie positive homogenate). Correspondingly PNGase F treatment of the brain homogenates did not alter the capability of MBP-PrP to detect PrP in brain homogenates (Fig. 6A, right panel lanes marked PF) even though detection with 9A2 showed decreased levels of glycosylated PrP (Fig. 6A, left panel lanes marked PF). MBP-PrP detection of PrP in the PNGase F treated scrapie positive homogenate still shows some detection of the different glycosylation forms (Fig. 6A, right lower panel lane marked PF). Comparison of MBP-PrP detection of PrP in the PNGase F treated – and non-treated scrapie positive brain homogenate samples shows that after PNGase F treatment the amount of di-glycosylated PrP has decreased while mono-glycosylated PrP has increased. Therefore the detection of PrP glycoforms after PNGase treatment is not due to aspecific binding of MBP-PrP, but rather a result of incomplete de-glycosylation of PrP in this particular sample. In contrast, PK treatment of the brain homogenate abrogated MBP-PrP detection (Fig. 6A, lanes marked PK) whereas immunodetection using monoclonal antibody 9A2 clearly shows the presence of PK-resistant PrPSc (Fig. 6A, right panel lane marked PK) in the scrapie positive brain homogenate. As a control the same samples were tested in Western blot using free MBP, resulting in no detectable signal with either His-PrP or any of the brain homogenate samples (Fig. 6, center panels). Only MBP-PrP (or MBP) was detected (Fig. 6B, center panel, positive detection control), thus proving that the detection with MBP-PrP was PrP-specific.</p><!><p>Interaction of recombinant MBP-PrP with various species of PrP in Western blot. Samples containing either brain derived PrP [A] or recombinant derived PrP [B] were analyzed by SDS-PAGE and subsequent Western blotting. PrP was detected using MBP-PrP (left panels), MBP alone (center panels) or PrP-specific antibody 9A2 (right panels). Refer to the Methods section on "reverse detection assay" for specific details concerning immunodetection of membrane bound PrP. Aliquots from PrPSc negative ([A] upper panels) and PrPSc positive brain ([A] bottom panels) homogenates were either treated with PNGase F (PF) or proteinase K (PK) and compared to non-treated (nt) aliquots of brain homogenate. His-PrP ([B], lanes marked HP) was used as a positive PrP-control and MBP-PrP ([B] lanes marked MP) was used as a positive detection control. Arrow heads indicate the approximate position of the un-glycosylated, mono- and di-glycosylated PrP isoforms.</p><!><p>Probing for possible PrP interaction domains using MBP-PrP and a solid-phase PrP peptide-array resulted in PrP specific interaction between specific PrP-sequences (peptides) and MBP-PrP. This probing revealed several likely interaction domains encompassed in two distinct high binding areas and some lower binding areas which will be discussed below in relation to structural features of PrPC. Suggested properties in conversion, the species barrier and self-interaction sites as hypothesized in structural models will be discussed.</p><p>The first distinct high binding area contains two different interaction domain consensuses. The domain [PHGG] is repeated five times in the ovine peptide-array and six times in the bovine peptide-array (Fig. 3C–H) and is part of the octarepeat sequence PHGGGWGQ, except for the first octarepeat (Fig. 3C) where H is replaced with a Q. This substitution is considered neutral [33], confirmed by the lack of effect on the binding pattern found, even though Q is a polar residue and the weak positive charge of H is neutralized. The octarepeats are an epitope for antibodies inhibiting PrPSc propagation in cell culture [26]. This study further reveals domain [102-WNK-104] (Fig. 3I) as a domain involved in PrP-PrP interaction. These three AAs are part of the epitope of the motif-grafted antibody containing mouse AA 89–112, which is capable of selective immuno-precipitation of infectivity [34].</p><p>The second high binding area contains 3 different interaction domain consensuses. This area also includes most of the polymorphisms found in sheep PrPC. The domain [140-PLIHFGNDY-148] (Fig. 3J) is situated between the first β-sheet (Fig. 5, S1) and first α-helix (Fig. 4, H1; the D and the Y are actually the first AA in α-helix 1) in PrPC and is part of the epitope of the motif-grafted antibody containing mouse AA 136–158 (together with [152-YYR-154]), capable of selective immuno-precipitation of prion infectivity [34]. This domain encompasses the two amino acid positions which appear to fully control species-specific kinetics of PrP23–144 [35] by affecting amyloid fibril conformation, thus limiting which PrPC molecule can adapt to the conversion seed. Therefore this domain is most likely involved in the species barrier and/or indirectly determines the susceptibility of sheep PrP to scrapie, maybe by influencing the accessibility of this domain and thus the adaptability of PrPC to the conversion seed. Involvement of this region in adaptability between species was also concluded for human, mouse and hamster prions [35]. The [YYR] sequence occurs twice in the PrP sequence. AA [152-YYR-154] (Fig. 3K) is situated within the first α-helix (Fig. 5, H1) and AA [165-YYR-167] (Fig. 3L) is situated within the second β-sheet (Fig. 5, S2) of PrPC. The charged residues in the first α-helix, especially at residues 151, 152 and 154 (D, Y and R respectively) have also been shown to be a determinant of conversion [36]; substitutions of neutral amino acids or oppositely charged residues impaired conversion. Furthermore, [152-YYR-154] (together with [140-PLIHFGNDY-148]) is part of the of the epitope of the motif-grafted antibody containing mouse AA 136–158, capable of selective immuno-precipitation of prion infectivity [34]. The [YYR] domain has also been described as the epitope of an antibody that selectively recognizes PrPSc [37], and is part of several antibody (fragment) epitopes that prevent scrapie infection in tissue culture [18,19,23,25,26] or in a mouse model [24]. The putative domain [177-NFV-179] is the common AA sequence of three consecutive peaks of which the fist two peaks encompass amino acids of a peptide (corresponding to AA 163–176) that inhibits in vitro conversion [15]. However, because of the third peak it is more likely that this domain is involved in self-interaction.</p><p>The remaining lower binding areas are more difficult to interpret. The domain consensus for the peptides 22–33 (Fig. 3A) seems to be [33-GWNTG-37], but may be a result of a cross-interaction with WN, instead of binding to domain [102-WNK-104] within the first high binding area. Binding with domain consensus [42-PGQGSPGG-49] (Fig. 3B) is relatively low and may likely be due to cross-interaction with the proline (P) and two consecutive glycines (G) of these peptides, instead of binding to the consensus octarepeat domain [PxGG]. Binding with the domains [183-VNITVKQHTVT-193] (Fig. 3N) and [192-TTTTKGENFT-202] (Fig. 3O) is also (very) low and these domains comprise the second helix in PrPC (Fig. 5, H2) which in turn may explain the relative low binding due to the structured nature of this part of the protein. It is not clear what the importance is of these domains, but part of the latter domain is contained within a peptide (corresponding to AA 197–220) that inhibits in vitro conversion [15] as well. The last low binding domain found is [225-SQAY-228], which (together with [YYR]) is part of the non-linear epitope of the PrPSc specific monoclonal antibody 15B3 [18].</p><p>Several studies identified antibodies [18,19,23-26] or peptides [15] able to inhibit prion propagation. The binding domains found with the peptide-array containing AA corresponding with these antibody epitopes or peptides may also be (in)directly involved in conversion. However, the inhibitory effects of antibodies seems simply due to steric hindrance preventing any PrP interaction at all (also confirmed by our peptide-array blocking results), making the binding domains corresponding to the conversion inhibiting peptides (Fig. 3M, peak1+2 and Fig. 3O) the more likely candidate domains influencing conversion.</p><p>Our data are in line with a theoretical two rung β-helical model described by Langedijk et al. [38], which tries to explain how AA sequence and secondary structure could explain strain properties and the species barrier. The binding domains found by peptide-array are all exposed in the periphery of the proposed fibril. The first distinct high binding domain makes up the incoming protein chain, whereas the second distinct high binding area forms the loop connecting the two-rung β-helical core with the α-helices (the outgoing protein chain). Furthermore, no binding is observed between the two distinct high binding areas and the AA sequence of these peptides correlate with the predicted two rung β-helical core.</p><p>Our data provide insight in the possible interaction domains of PrPC with itself or PrPSc, and most of the domains identified are likely to be involved in PrPC self-interaction. This may involve dimerisation of PrPC and/or formation of a trimer of three PrPC molecules as suggested in the theoretical two-rung β-helical model for PrP stacking during PrPSc induced fibrillization [38]. If one of the identified high binding domains is of influence on conversion, this possibly is exerted during pre-oligomerisation, which is an inefficient process. On the other hand, a direct effect on PrPC-PrPSc binding can not be excluded.</p><!><p>In addition to the ovine peptide-array, a similar array of overlapping 15-mer peptides bovine peptides was used. Since there are several sequential differences between bovine and ovine PrP this may be of influence on the overall binding pattern found with MBP-PrP (ovine PrP). Generally it seems that binding of MBP-PrP is somewhat less efficient (strong) with the bovine peptide-array compared to binding with the ovine peptide-array. However, this had no significant effects on the relative binding pattern. In the amino acid sequence of bovine PrP there is an extra octarepeat as well as six amino-acid substitutions. Only the largest differences will be discussed here. The non-discussed differences may well be the result of minor methodological variations when producing the peptide-arrays.</p><p>As expected, an extra octarepeat (Fig. 3F) was evident that confirmed the octarepeats consensus binding domain [PHGG]. Only two out of six amino acid substitutions were of influence on the binding pattern. At first glance the amino acid substitutions at AA 98 and 100 (sheep numbering used throughout) are seemingly both of influence on binding with peptides 93–97 (Fig. 3I, bo# 98–113), allowing detection of the consensus binding domain [102-WNK-104] only when these AA were no longer present in the peptides. The supposedly neutral substitution of glycine for serine at AA 100 is most likely responsible for the observed differences in binding and may be attributed to the greater conformational flexibility of G, affecting availability of other AA's for interaction with MBP-PrP. Taking this in account, binding with peptides 90–92 should be a result of cross-binding with two consecutive Glycines present in these peptides in stead of binding with the consensus octarepeat domain. Glutamatic acid only contains an oxygen in place of the amido group in glutamine and therefore these AA's are considered readily interchangeable [33]. However we observed that the substitution of glutamatic acid for glutamine at AA 189 did affect binding (Fig. 3N) on the bovine peptide-array with peptides 179–183; Both AA's are polar, but where glutamine interacts with other polar or charged atoms with its polar side chains, glutamatic acid is negatively charged and is frequently involved in salt-bridges and/or glutamatic acid interacts with positive charged AA's to form hydrogen bonds. These differences in AA reactivity are most probably responsible for the observed difference in binding between ovine and bovine PrP peptide sequences.</p><p>Analysis and comparison of the relative (consensus) ovine – and bovine peptide-array revealed that the detection of potential PrP-PrP interaction domains using this method is robust as well as sensitive to differences in structural flexibility and/or amino acid differences. Therefore, this peptide-array approach provides a possibly valuable tool to assess the influence of disease associated polymorphisms on available interaction domains and to test for PrP-PrP binding inhibitors potentially useful in therapy (i.e. antibodies or peptides).</p><!><p>All monoclonal antibodies (mab) recognizing PrP (Table 1, Fig. 5) are capable of blocking the complete binding pattern. Differences in the antibody concentration necessary for complete blocking are likely due to epitope availability and/or affinity for PrP. Complete blocking of the binding pattern can best be explained by steric hindrance of the antibodies preventing any interaction. It has been described that binding of a monoclonal antibody at the N-terminus of human PrP influences epitope availability at the C-terminus [39] and similar events may also attribute to completely abolishing the binding pattern. Furthermore, structural studies of PrPSc that resulted in prion propagation/fibrillization models [38,40,41], suggest that PrP-PrP interaction depends on the structure of the whole protein (not just the trimer or dimer core). These findings corroborate the notion that antibodies inhibiting prion propagation probably do so by preventing the interaction between PrPC and PrPSc or between PrPC molecules themselves in the pre-oligomerisation phase.</p><!><p>By using MBP-PrP as the detecting agent in Western blot we showed that binding of MBP-PrP to the peptide-arrays is PrP specific and indicative of a PrP-PrP interaction. MBP-PrP seems to preferably detect un-glycosylated PrPC in the scrapie negative brain homogenate. The exact reason for this preference is unclear, but does not seem to be due to the lack of glycosylated isoforms in the brain homogenate (except after de-glycosylation with PNGase F) as shown by immuno-detection with 9A2. It may be possible that the determined interaction domains do not (only) interact with the same amino acid motif (self-self), but that an intramolecular cross-interaction between different domains can also occur. The peptide-array data has revealed two high binding areas, one of which contains the N-terminal octarepeats and the other contains the first α-helix, second β-sheet up to the second α-helix directly adjacent to the glycosylation sites (Fig. 5). When PrPC in the homogenate is glycosylated this may sterically hinder binding of MBP-PrP to PrPC. Glycosylation is suggested to have a role in prion strain maintenance and the species barrier [42] by modulating the fidelity of interaction, which may explain the favorable binding of un-glycosylated PrP by MBP-PrP (which is also un-glycosylated, indicating that binding preferably occurs between compatible glycosylated molecules) in the scrapie negative homogenate. However, in the scrapie positive brain homogenate detection of PrP by MBP-PrP is more diffuse and might suggest that MBP-PrP detection of PrP in Western blot is due to interaction with both full-length PrPC as well as full length denatured PrPSc, which is comparable to PrPC. Alternatively, PrPSc might be partly endogenously truncated resulting in more heterogeneous binding of MBP-PrP to all glycosylation forms. In contrast, MBP-PrP detects un-glycosylated PrP as well as both mono – and di-glycosylated PrP in the scrapie positive homogenate, even though un-glycosylated PrP usually is the lesser component in the PrP-triplet of scrapie sheep brain samples, which may be indicative for preferable binding of un-glycosylated PrP in the scrapie positive homogenate. However interpretation of these results is difficult and in order to elucidate the precise effects of glycosylation on binding between PrP molecules, interaction should be studied under more native conditions. When proteinase K treatment was applied MBP-PrP did not detect PK-resistant PrPSc. This indicates that in order for MBP-PrP to detect PrP in brain homogenates (under the conditions used) full length (at least containing the high binding area with octarepeats) PrP molecules are required. It may be hypothesized that the first high binding area containing the octarepeats aids in stabilization of PrP self-interaction, perhaps by intramolecular interaction with other mapped interaction domains. This extra stabilization in turn allows further immunodetection in Western blot (under the conditions used). These results combined with the results obtained by peptide-array analysis support the concept of self-interactive domains of PrPC.</p><!><p>In summary, probing for possible interaction domains in PrP using a solid phase PrP peptide-array revealed that specific interactions take place between individual PrP molecules. Ten possible consensus binding domains were found, which includes one domain that likely is due to a cross-reaction with the octarepeat domain consensus – and for two domains it remains unclear what their importance is. The remaining seven domains are most likely involved in PrPC self-interaction. Furthermore, MBP-PrP was also capable to specifically bind to full length PrPC and PrPSc bound PrPC in Western blot confirming PrP-PrP specific interaction. Together these results indicate that in addition to direct PrPC-PrPSc interactions several other molecular interactions between PrPC molecules/sequences themselves may also be possible, facilitating initial steps in the oligomerisation process.</p><p>The PrP peptide-array may additionally facilitate in gaining insight into effects of disease associated polymorphisms in PrP on PrP-PrP binding, and the subsequent molecular conversion of PrPC into PrPSc. The (self-)interaction domains described here may ultimately prove useful in the design of therapeutics interfering in the PrP-PrP binding process.</p><!><p>In order to obtain the PrP gene suitable for cloning into the pMAL Protein Fusion and Purification System (New England Biolabs), the mature part of the sheep PrP (ARQ) open reading frame (ORF) was PCR amplified using primers ShBo-F-DraI (GGTGGTTTTAAAAAGCGACCAAAACCTGG) and Sh-R-STOP (GGTGGTCTATGCCCCCCTTTGGTAATAAGCC). The resulting PrP (AA25–233), without its N -and C-terminal signal sequences, was cloned into a general TA-cloning vector (Invitrogen) and sequenced to exclude PCR artifacts. The PrP fragment was subsequently sub cloned using DraI and EcoRI into the pMAL-c2X expression vector, resulting in the maltose binding protein (MBP) fusion to the N-terminus of PrP (MBP-PrP).</p><!><p>Expression and purification by affinity chromatography was performed as described in the manual of the pMAL Protein Fusion and Purifications System (method I; New England Biolabs) To improve binding of MBP-PrP to prevent formation of interchain disulfide upon lysis (as suggested in the protocol), β-mercaptoethanol was added. Quantity and quality of the eluted MBP-PrP was determined by SDS-PAGE (12% NuPAGE, Invitrogen). After separation the gel was either stained with Sypro Orange (total protein stain, Molecular probes) or analyzed by Western blotting and immunodetection of MBP-PrP with polyclonal antiserum R521-7 specific for PrP. To obtain MBP for cross-reaction aspecificity tests, the pMAL-c2x expression vector without insert was expressed and purified as described above.</p><!><p>Synthesis of complete sets of overlapping 15-mer peptides were carried out on grafted plastic surfaces, covering the ovine or bovine PrP amino acid sequence of mature PrP (residues 25–234 of ovine and 25–242 of bovine PrP) [43]. The plastic surface consisted of 455, 3 μl wells on a credit-card size plastic (minicard) carrier. Peptide-arrays covering the ovine or bovine PrP amino acid sequence of mature PrP were custom synthesized through two different synthesis techniques: either all peptides of the array were synthesized in situ to the grafted plastic surface by step-by-step amino acid coupling or the peptides were synthesized separately and coupled as complete 15-mer peptides to each well at their C-terminus [44-46]. In subsequent ELISA analyses on the minicards, MBP-PrP was incubated as an antigen followed by immuno-screening with mouse anti-MBP monoclonal antibody (Mab) obtained from New England Biolabs and rabbit anti-mouse-IgG-peroxidase, or a rabbit anti-MBP Mab and swine anti-rabbit-IgG-peroxidase (from DAKO, Denmark). Blocking studies were performed by pre-incubating the MBP-PrP with a PrP-specific Mab before incubating the mixture as the antigen on the minicard. The background was determined by calculating the mean value of 20 peptides with low density values of which at least 5 peptides were in consecutive order. The relative density value (r.d.v.) was calculated by dividing the optical density value (o.d.v.) by the background and binding was considered relevant when at least 3 consecutive peptides showed binding values of at least 3 times the background.</p><!><p>Monoclonal antibody 6C2 was newly prepared using PrP-knockout mice immunized with peptide KTNMKHVAGAAAAG (ovine PrP109–122), conjugated through a cysteine at its C-terminus to Keyhole limpet hemocyanine, using previously described procedures for synthesis and screening [47]. In ELISA and Western blot antibody 6C2 binds respectively to ovine – and bovine recombinant PrP and ovine – and bovine PrPres at the approximate residues HVAGAAA as determined by peptide mapping analysis using an ovine peptide-array.</p><!><p>Each reaction contained 500 ng MBP-PrP and monoclonal antibody in PBS containing 0.05% Tween80. For each antibody the inhibitory concentration as well as an excess concentration (max. 25 μg/ml) was tested. The reaction was incubated for 1 hour at room temperature and subsequently centrifuged for 30 minutes at 20,000 × g. Most of the supernatant was transferred to a new tube except approximately 3–5 μl to prevent disturbance of the pellet. The pellet fraction was dissolved in 0.1% SDS by sonification. Both fractions were subjected to methanol precipitation and analyzed by SDS-PAGE (12%, NuPAGE), Western blot and immunodetection using either R521-7 (rabbit anti-PrP serum [48]) and swine anti-rabbit-IgG-peroxidase (PrP detection) or rabbit anti-mouse-IgG-peroxidase (antibody detection). The relative amount of MBP-PrP or antibody band(s) detected as fluorescent signal (f.s.) in Western blot was determined by using the ECF substrate for detection and the Molecular Dynamics ImageQuant software for quantification. Subsequently the mean percentages of MBP-PrP or antibody in the soluble (supernatant) fraction (f.s.sup/(f.s.sup+f.s.pel) were calculated.</p><!><p>PrPSc was isolated from brain tissue of clinically ill scrapie sheep. PrP genotypes were determined by Sanger sequencing of the full PrP-ORF as described before [49]. PrPSc was purified by ultracentrifugational pelleting from sarcosyl-homogenated brains as described previously [3,50]. The final pellets were sonicated in phosphate-buffered saline containing 1.0% SB 3–14. Yields of PrPSc were quantified by SDS-PAGE (12% NuPAGE) and Western blotting using antiserum R521-7.</p><!><p>Confirmed scrapie positive and negative 10% sheep brain homogenates were digested with either proteinase K (PK) or PNGase F and compared to the non-treated samples. A separate aliquot of 10% brain homogenate was treated with 35 μg/ml PK for 1 hour at 37°C. Another aliquot was denatured by adding 1/10 volume denaturing buffer (5% sodium dodecyl sulphate and 10% β-mercaptoethanol in 20 mM Tris-HCl- 150 mM NaCl- 2 mM EDTA [pH 7.5]) for and subsequent heating 10 min. at 96°C. This aliquot was de-glycosylated in the presence of 1000 U PNGase F/ml for at least 36 hours at 37°. Untreated, PK treated – and PNGase F treated brain homogenates samples were analyzed by SDS-PAGE and Western blot. As positive controls His-PrP (positive PrP control) and MBP-PrP (positive detection control) were included. Reverse detection of PrP was accomplished by incubating the Western blot for 1 hour at room temperature with approximately 2 ng/μl MBP-PrP followed by immunodetection using mouse anti-MBP monoclonal antibody and rabbit anti-mouse-IgG-peroxidase (RAMPO). PrP was also detected on Western blot using the PrP-specific monoclonal antibody 9A2 and RAMPO. To determine if detection is PrP specific, a Western blot was carried out with MBP alone instead of MBP-PrP.</p><!><p>TSE: transmissible spongiform encephalopathy, PrP: general denotation for prion protein, PrPC: host-encoded cellular prion protein (protease sensitive), PrPSc: scrapie associated isoform of the prion protein (partially protease resistant), MBP-PrP: fusion protein of maltose binding protein linked to the N-terminus of the mature sheep prion protein. Non-Polar Amino Acids; I: isoleucine, V: valine, L: leucine, F: phenylalanine, M: methionine, A: alanine, G: glycine, W: tryptophane, P: proline (ordered from most hydrophobic to most hydrophilic). Neutral Polar Amino Acids; C: cysteine, T: threonine, S: serine, Y: tyrosine, N: asparagine, Q: glutamine (ordered from most hydrophobic to most hydrophilic). Charged Polar Amino Acids; H: histidine, D: aspartate, E: glutamatic acid, K: lysine, R: arginine (ordered in accordance to increasing hydrophilicity)</p><!><p>AR and AB conceived the study and together with JPML were responsible for study design and coordination. PLJMM was involved in later stage study design and responsible for development and testing of MBP-PrP. JGJ was responsible for the production and testing of monoclonal antibody 6C2. All peptide-arrays were performed at Pepscan B.V. by DPT and all other experiments were performed by AR. JPML and AR were responsible for choice of – and analyses of peptide-arrays. AB is project manager of the NWO-project under which this study was performed. AR and AB were responsible for data-analysis. AR drafted the manuscript and AB and JPML critically read the manuscript before submission. All authors have read and approved the final manuscript.</p><!><p>This work was supported by grant 903-51-177 from the Dutch Organization for Scientific Research (NWO), by a grant from the Dutch Ministry of Agriculture, Nature Management and Fisheries (LNV) and by EU NeuroPrion project STOPPRIONs FOOD-CT-2004-506579. We kindly thank Saskia Ruiter for her work on the MBP-PrP expression construct and the monoclonal mycobacterium antibody (M7) was courtesy of Dr. D. Bakker (CIDC-Lelystad, NL).</p>

PubMed Open Access

Differentiating antiproliferative and chemopreventive modes of activity for electron-deficient aryl isothiocyanates against human MCF-7 cells

Consumption of Brassica vegetables provides beneficial effects due to organic isothiocyanates (ITCs), a resultant product of the enzymatic hydrolysis of glucosinolate secondary metabolites. The ITC L-sulforaphane (L-SFN) is the principle agent in broccoli that demonstrates several modes of anticancer action. While the anticancer properties of ITCs like L-SFN have been extensively studied and L-SFN has been the subject of multiple human clinical trials, the scope of this work has largely been limited to those derivatives found in nature. Previous studies have demonstrated that structural changes in an ITC can lead to marked differences in a compound\xe2\x80\x99s potency to (1) inhibit growth of cancer cells, and (2) alter cellular transcriptional profiles. This study describes the preparation of a library of non-natural aryl ITCs and the development of a bifurcated screening approach to evaluate the dose- and time-dependence on antiproliferative and chemopreventive properties against human MCF-7 breast cancer cells. Antiproliferative effects were evaluated using a commercial MTS cell viability assay. Chemopreventive properties were evaluated using an antioxidant response element (ARE)-promoted luciferase reporter assay. The results of this study have led to the identification of (1) several key structure-activity relationships and (2) lead ITCs for continued development.

differentiating_antiproliferative_and_chemopreventive_modes_of_activity_for_electron-deficient_aryl_

Introduction<!>Preparation of aryl ITC analogues<!>Evaluation of antiproliferative activity for aryl ITC analogues<!>Evaluation of ARE-induction activity for aryl ITC analogues<!>Integrated analysis of antiproliferation activity and ARE- induction activity for aryl ITC analogues<!>Validation of ARE-induction byqPCR<!>Conclusions<!>Experimental Section<!>0,0-Di(pyridin-2-yl) carbonothioate (26):[31]<!>Isothiocyanatobenzene (2):<!>(Isothiocyanatomethyl)benzene (3):<!>1-lsothiocyanato-2-methylbenzene (27):[40]<!>1-lsothiocyanato-3-methylbenzene (28):[41]<!>1-lsothiocyanato-4-methylbenzene (29):[42]<!>1-(lsothiocyanatomethyl)-2-methylbenzene (30):<!>1-(lsothiocyanatomethyl)-3-methylbenzene (31):<!>1-(lsothiocyanatomethyl)-4-methylbenzene (32):[43]<!>1-(lsothiocyanato)-2-(trifluoromethyl)benzene (33):<!>1-(lsothiocyanato)-3-(trifluoromethyl)benzene (34):[40]<!>1-(lsothiocyanato)-4-(trifluoromethyl)benzene (35):[40]<!>1-(lsothiocyanatomethyl)-2-(trifluoromethyl)benzene (36):<!>1-(lsothiocyanatomethyl)-3-(trifluoromethyl)benzene (37):<!>1-(lsothiocyanatomethyl)-4-(trifluoromethyl)benzene (38):<!>1-isothiocyanato-2-nitrobenzene (39):<!>1-lsothiocyanato-3-nitrobenzene (40):[42]<!>1-lsothiocyanato-4-nitrobenzene (41):[44]<!>1-(lsothiocyanatomethyl)-2-nitrobenzene (42):<!>1-(lsothiocyanatomethyl)-3-nitrobenzene (43):<!>1-(lsothiocyanatomethyl)-4-nitrobenzene (44):<!>4-(Methylthio)phenylmethanamine hydrochloride (52):<!>1-lsothiocyanato-2-(methylthio)benzene (53):<!>1-lsothiocyanato-3-(methylthio)benzene (54):[45]<!>1-lsothiocyanato-4-(methylthio)benzene (55):[46]<!>1-(lsothiocyanatomethyl)-2-(methylthio)benzene (56):<!>1-(lsothiocyanatomethyl)-3-(methylthio)benzene (57):<!>1-(lsothiocyanatomethyl)-4-(methylthio)benzene (58):<!>1-lsothiocyanato-2-(methylsulfinyl)benzene (59):<!>1-lsothiocyanato-2-(methylsulfonyl)benzene (65):<!>1-lsothiocyanato-3-(methylsulfinyl)benzene (60):<!>1-lsothiocyanato-3-(methylsulfonyl)benzene (66):[45]<!>1-lsothiocyanato-4-(methylsulfinyl)benzene (61):<!>1-lsothiocyanato-4-(methylsulfonyl)benzene (67):[45]<!>1-(lsothiocyanatomethyl)-2-(methylsulfinyl)benzene (62):<!>1-(lsothiocyanatomethyl)-3-(methylsulfinyl)benzene (63):<!>1-(lsothiocyanatomethyl)-3-(methylsulfonyl)benzene (69):<!>1-(lsothiocyanatomethyl)-4-(methylsulfinyl)benzene (64):<!>1-(lsothiocyanatomethyl)-4-(methylsulfonyl)benzene (70):<!>1-(lsothiocyanatomethyl)-2-(methylsulfonyl)benzene (68):<!>MTS antiproliferation assay:<!>ARE induction assay:<!>Quantitative real-time PCR (qPCR) analysis:

<p>The Brassica vegetables, which include broccoli, cabbage, cauliflower, Brussels sprouts, kale, collard greens, pak choi and kohlrabi, are rich sources of glucosinolates (β-thioglucoside-A/-hydroxysulfates); tissue damage to the plant induces enzymatic hydrolysis of glucosinolates, resulting in evolution of various secondary metabolites.[1] At physiological pH, the principle products of glucosinolate hydrolysis are isothiocyanates (ITCs),[2] which are believed to be primarily responsible for the observed cancer chemoprevention that results from diets rich in these vegetables.[3] Although dietary glucosinolates have no known direct bioactivity, many of their resultant ITCs are well-studied anticancer agents; several of the most highly-studied ITC natural products include L-sulforaphane (L-SFN, 1, Figure 1), benzyl isothiocyanate (BITC, 3), phenethyl isothiocyanate (PEITC, 4), and allyl isothiocyanate (AITC, 5). L-SFN (1) is particularly abundant in broccoli and has been the subject of numerous clinical trials.[4] The unsubstituted aryl ITCs 3 and 4 have received significant attention for their ability to inhibit chemically-induced cancer in animal models at low micromolar concentrations,[5] well below the observed threshold for toxicity against noncancerous cells.[6] Enthusiasm for the aliphatic ITC 5 has decreased in part due to preclinical reports citing toxicity in the bladder and hematological tissues.[3i, 6] The literature is rich with many specific examples describing the effects of these (and other) dietary ITCs against cancer cell types and in specific model systems.[3i, 6] One of the most attractive features of using ITCs as anticancer agents is the ability of one agent to affect multiple mechanistic points involved in cancer pathogenesis, through a combination of cytostatic,[7] cytotoxic, [8] and chemopreventive mechanisms.[3c,8c,9]</p><p>Dietary ITCs demonstrate chemotherapeutic properties through their ability to damage and inhibit proliferation of cancerous cells. These agents are capable of directly modulating signaling pathways which promote cell proliferation[3i] and inhibiting features of cell division, leading to cell cycle arrest. The most common points of arrest are G0-G1 and G2-M, dependent on both the identity of the ITC and cellular mechanisms specific to a cancer cell type[3i, 6] ITCs can promote apoptosis and autophagic cell death through diverse mechanisms,[3i] modulate epigenetic marks through inhibition of histone deacetylase (HDAC),[7a, 10] promote anti-angiogenic effects through downregulation of vascular endothelial growth factor (VEGF), and elicit antimetastatic effects through suppression of ERK kinase.[3i] ITCs have also demonstrated anti-inflammatory and immunomodulatory activity[11] and can sensitize cancer cells to other anticancer agents.[12]</p><p>The second major mode of ITC activity relates to their ability to inhibit carcinogenesis through chemopreventive mechanisms. ITCs block tumorigenesis caused by various chemical carcinogens by reducing expression levels of phase I drug- metabolizing cytochrome P450s.[3i, 13] Arguably, the most significant chemopreventive activity of ITCs results from their potent induction of phase II enzymes that assist in clearing chemical carcinogens and reactive oxygen species from the body; noteworthy phase II enzymes include glutathione S-transferases (GST), NAD(P)FI:quinone oxidoreductase (NQO1), epoxide hydrolase, and UDP-glucuronosyl-transferases.[4c, 14] The most well-studied phase II chemopreventive mechanism involves the activation of nuclear factor E2 p45-related factor 2 (Nrf2).[3d,15] Under basal conditions, cytoplasmic Nrf2 is associated with Kelch-like ECFI-associated protein 1 (Keapl), a multidomain, cysteine-rich protein bound to the actin cytoskeleton that promotes Cul3-based ubiquitination of Nrf2 and its subsequent degradation by the proteasome.[3a,3d] ITCs contain an electrophilic carbon (R-N=C=S) capable of reaction with thiols, including specific cysteine residues on the surface of Keapl[16] Thiocarbamoylation of Keapl at Cys151 by ITCs disrupts the Keapl-Cul3 interaction, thus preventing Nrf2 ubiquitination, and resulting in Nrf2 release and its subsequent nuclear accumulation.[4a,16b,17] In the nucleus, heterodimerized Nrf2 binds to antioxidant response elements (ARE), regulatory DNA sequences upstream of chemopreventive genes, where it activates transcription.[18] The large number of genes impacted through ITC-mediated stimulation of Nrf2 transcriptional activity indicates the importance of this interaction in dietary cancer chemoprevention.[19]</p><p>Although the effects and mechanisms of naturally-occurring, dietary ITCs have been heavily investigated, fewer studies have described the anticancer properties of non-natural ITCs. A 2008 study described the preparation and evaluation of 35 non-natural ITCs and several of their non-natural glucosinolate precursors.[20] In this study, 5 of the 35 analogues screened demonstrated enhanced antiproliferative activity versus 1; each of these analogues were substituted aryl ITC variants. The percentage of hits produced by this panel (14%) was high compared to what is typically observed during a preliminary screen (0.1–5%),[21] suggesting that the structure of the parent compound 1 may be sub-optimal for antiproliferative activity. Several key structure– activity relationships (SARs) were identified, indicating that ITCs may be amenable to drug development strategies. Importantly, this study also demonstrated that synthetic, non-natural glucosinolates can serve as precursors for their analogous, improved non-natural ITCs.[3d, 22] This feature has been validated by more recent work[23] and may serve as an opportunity to circumvent the problem of aqueous instability of ITCs, which has plagued formulation strategies during clinical trials.[24] Together, this supports the premise that non-natural ITCs are viable targets for the design of novel anticancer agents, a field that has not been extensively explored.</p><p>The central motivation for this study was to test the hypothesis that unsubstituted, dietary aryl ITCs are not optimal as anticancer agents and that the structure of a substituted variant will impact it's observed anticancer properties. Despite the specific mechanistic differences between the action of 3 and 4 across various cancer cell types, both compounds demonstrate broad similarities in their ability to inhibit cancer cell growth, promote cell cycle arrest, stimulate apoptosis, generate ROS, inhibit phase I enzymes, and induce expression of ARE-dependent phase II genes. These trends are a stark contrast to the properties exhibited by their non-natural methylene homologue, phenyl isothiocyanate (PITC, 2), which has significantly reduced activity.[3h,20] The anticancer properties of substituted aryl ITCs have not been thoroughly investigated and would prove useful in developing a more comprehensive understanding of SARs for this class of compounds. Since natural ITCs are known to thiocarbamoylate cellular proteins,[16a]it is reasonable to expect that the steric and electronic properties of non-natural, substituted aryl ITCs may impact an analogue's observed anticancer properties.</p><p>To test this hypothesis, a panel of 36 substituted aryl ITCs were selected for preparation (Figure 2). Since the greatest difference in activity between natural aryl ITCs occurs between 2 and 3, this panel included derivatives of both PITC (n = 0) and BITC (n = 1). Substituted PEITC analogues (n = 2) remain an interest toward the larger hypothesis and were not included in this panel due to the increased synthetic complexity required for their preparation; these studies are ongoing and will be reported in due course. With the expectation of preparing ortho-, meta-, and para-regioisomers for each substituent variant, six R groups were selected for inclusion. Previous SARs noted improved antiproliferative properties for aryl ITCs bearing electron-deficient groups;[20,22] for this reason, trifluoromethyl (R = CF3) and nitro (R = NO2) groups were selected. Since fluorine is a bioisostere for hydrogen, methyl analogues (R = CH3) were included to directly contrast any electronic effects displayed by trifluoromethyl analogues.[25] Methylsulfanyl (R = SCH3), methylsulfinyl (R = S(O)CH3), and methylsulfonyl (R = S(O2)CH3) analogues were selected based on both previous SAR which noted the importance of the sulfinyl group in 1[20] and the goal of evaluating the effect of electron-deficient groups (e.g. sulfinyl, sulfonyl). Furthermore, these latter analogues served as an opportunity to hybridize the key structural features of 1 with aryl ITCs 3 and 4; even for such relatively simple compounds, most of these 18 sulfur-containing analogues have novel structures.</p><p>In an attempt to accommodate and differentiate the diverse anticancer modes exhibited by natural ITCs, it was desirable to develop a systematic and efficient screening process to broadly determine the properties of candidates against cancer cells. Preliminary screening efforts were conducted against MCF-7 human breast cancer cells; the selection of MCF-7 cells was based on a combination of factors, including: (1) its regard as one of the most highly-studied breast cancer cell lines,[26] (2) its established use as an in vitro model for proliferation and chemoprevention studies with |TCs,[12b,20'24,27] and (3) its reduced levels of cytosolic Nrf2 and subsequent chemosensitivity.[28] The dose-response for each ITC candidate would be determined following variable incubation time with MCF-7 cells (t = 24 h, 72 h). Antiproliferative properties would be evaluated using a commercial MTS cell viability assay which measures the reductive capacity of viable cells.[20,29] Chemopreventive properties would be evaluated using an in vitro reporter assay which correlates ARE activation to a spectrophotometric response against human MCF-7 cells stably transfected with an ARE-luciferase reporter plasmid vector (MCF-7-ARE cells).[30] Although in vitro assays exist for many biological endpoints, recent analysis has identified limited options to evaluate candidate ARE induction without use of an animal model;[3h] consequently, development and validation of this latter assay would greatly complement the body of available in vitro screening methods.</p><!><p>Approximately half of the target aryl ITCs were prepared directly from reaction of their corresponding, commercially-available primary amines (6–25, Scheme 1) with 0,0-di(pyridin-2-yl) carbonothioate (26).[23a,31] While many of the substituted PITC analogues (n = 0) have been described within the literature or are, in some cases, commercially-available, all compounds were synthetically prepared, purified, and characterized to ensure uniformity among the analogues to be evaluated. It was observed that the highest yields of ITC were obtained when using > 3.0 equivalents of 26. Although 26 is commercially-available, it was more cost-efficient to prepare 26 in-house using a modified version of the synthesis described by Kim, et. al.[31]</p><p>Sultane-, sulfoxide-, and sulfone-derived aryl ITCs were prepared using the divergent synthesis described in Scheme 2. Since two of the target ITC scaffolds lacked commercially-available primary amines, they were prepared from benzaldehydes 45–46 using reductive amination. Treatment of 45–46 with hydroxylamine hydrochloride provided oxime intermediates in high yield (data not shown), which were immediately reduced to amines with UBFI4.[32] It was observed that the reduction was most often responsible for the greatest loss in yield, due to a combination of sluggish reaction and the difficulty in isolating the free amine products. Passing gaseous FICI through the solution of the amines in dry MeOFI induced selective precipitation of the amine hydrochloride salts;[29b] intermediate 52 was obtained in 27% yield over three steps. Amines 47–52 were reacted with 26 to generate ITCs 53– 58 [23a, 33] Sulfoxide-(59–64) and sulfone-analogues (65–70) were prepared through treatment of 53–58 with mCPBA.[20,34] While sulfone-variants could be prepared in high yield using excess (> 3 equivalents) oxidant, it was challenging to obtain sulfoxide analogues in similar yield; even with careful, controlled addition of one equivalent mCPBA, some unreacted sulfane and sulfone were observed. Since both sulfoxide- and sulfone-analogues were desired for evaluation, reactions were intentionally performed with 1–2 equivalents mCPBA to provide a mixture of both sulfoxide and sulfone products which were obtained using flash chromatography separation. This approach provided high combined yields of both products (67–94%).</p><!><p>The antiproliferative properties of each non-natural aryl ITC candidate was evaluated against human MCF-7 breast cancer cells over the concentration range 200 to 0.78 pm. Although a candidate incubation time of 72 h was selected to maintain consistency with previous methods,[20,29] parallel evaluation of ARE-induction required shorter incubation periods; consequently, in order to maintain congruency between the datasets for both endpoint assays, candidates were evaluated after both 24 h and 72 h incubation. The commercial MTS assay was employed to assess the number of viable cells relative to control, similar to previous reports[35] Representative antiproliferation dose- response curves for aryl ITC analogues are depicted in Figures 3–5, with calculated GI50 values in Table 1. In this study, the graphical depiction of data provides the means to draw parallels between antiproliferation and ARE-induction modes of evaluation, as well as to illustrate unexpected features within the antiproliferation dataset.</p><p>The antiproliferative activity of 1 after 72 h was consistent with previous data (3.1 ± 0.9 pm), despite the differences in the method used to determine cell viability and the nature of the dosing regimen.[20] Flowever, antiproliferation data for 1 after 24 h unexpectedly demonstrated a significant non-sigmoidal relationship (NSR) in the dose-response curve, which was unable to be fit to the nonlinear log(inhibitor) vs. response (variable slope, 4 parameters) function in GraphPad 6.0. The deviation from an expected sigmoidal response was observed at concentrations between 50 to 200 pm and was consistent across individual trials obtained over a five year-span (see Figure SI-1 in the Supporting Information). Since the MTS assay was used to determine percent viability relies on mitochondrial function to provide a colorimetric measurement of living cells, these data alone do not directly provide insight to rationalize the presence of the NSR. While the antiproliferative properties of 1 against a variety of cancer cell lines have been previously examined,[36] few reports have even addressed a possible relationship between 1 and pro-viability effects.[37] Consequently, the literature is unclear whether a proviability NSR has been similarly observed and documented. It is plausible that this deficiency results from a combination of (1) the infrequency of reporting dose-response curves, (2) the increased dose-response resolution afforded by iterative, 2-fold serial dilutions used in this current study, and (3) the range of concentrations tested. Furthermore, an attenuated NSR for 1 also appears to be reproducibly present over the same concentration range after 72 h incubation; the slight deviation from sigmoidal response at [1] = 50 pm was initially believed to result from normal experimental variation and does not significantly impact the ability to fit the data to a nonlinear curve. Flowever, with the additional context of the 24 h incubation response, it seemed more likely that the presence of a NSR is time dependent, decreasing between 24 h and 72 h incubation.</p><p>Expanding the body of antiproliferation results to include the panel aryl ITCs draws attention to several comparisons, trends, and structure-activity relationships (SARs). Out of the 38 ITCs evaluated, 14 demonstrated non-sigmoidal dose-response curves after 24 h incubation; only three analogues demonstrated a NSR after 72 h incubation. All but one of the compounds with a NSR were substituted derivatives of BITC (3); the sole exception was 33, ortrto-CF3-substituted PITC. Similar to the trends for 1, the NSR effects were reproducible across trials (data not shown) and were largely attenuated at 72 h incubation. While it is remains unclear why BITC and its analogues exhibit a greater propensity to have a NSR, the alkyl linker between the isothiocyanate functional group and the aromatic ring (n = 0–1) appears to play an important role. Throughout the remainder of this account, compounds demonstrating NSR have been excluded from quantitative comparisons of GI50 trends; this designation most significantly affects analysis of certain analogue subsets after 24 h incubation. For most analogues, antiproliferative activity increased between 24 h and 72 h, consistent with previous observations.[36a] Of the 10 aryl analogues demonstrating the opposite qualitative trend, three were inactive across the dose range (59–61) and five had 24 h activity within error of 72 h activity (2, 27–29, 40); the anomalies (33 and 39) were both ortho- substituted PITC analogues bearing a strongly electron withdrawing group (CF3 and NO2, respectively). On average, PITC analogues (n = 0) demonstrated greater than 5-fold reduction in antiproliferative properties versus BITC analogues (n = 1) after both 24 h and 72 h incubation; this calculation was made using GI50 = 200 pm for inactive compounds (GI50 > 200 μm).</p><p>Comparison of substituted aryl ITCs to the unsubstituted 2 and 3 provided insight toward the impact of functional group on antiproliferation activity. All aryl ITCs bearing a methyl group or a trifluoromethyl group had lower GI50S relative to the unsubstituted aryl ITC analogue, with the trifluoromethyl group providing the larger reduction; the magnitude of this reduction ranged from a statistically-insignificant 0.1 pm (28, 72 h) to 24.4 pm (35, 72 h). The presence of a nitro substituent also generally lowered the GI50 (range of reduction = 3.0 pm to 20.1 pm); the sole exception to this trend was ortho-nitro PITC analogue 39, which had significantly reduced antiproliferative activity. Structure-activity trends for the three sulfur-containing substituents (methylsulfanyl, methylsulfinyl, and methylsulfonyl) differed between the PITC scaffold and the BITC scaffold. PITC analogues bearing methylsulfinyl, and methylsulfonyl groups were inactive (GI50 > 200 pm), as was the p-methylsulfanyl analogue 55 after 24 h incubation. The other methylsulfanyl-substituted PITC analogues (53, 54) exhibited no discernible pattern to their antiproliferation activity. In contrast, all of the sulfur-containing BITC analogues (56–58, 62–64, 68–70) demonstrated decreased GI50 after 72 h, relative to unsubstituted 3. Unfortunately, most comparisons between ortho-, meta-, and para-regioisomers were unable to provide conclusive, statistically-significant trends; the most noticeable difference between regioisomers was observed for the ortho-nitro PITC analogue 39, as previously noted.</p><!><p>A primary goal of this work was to develop an accurate and efficient method to screen the chemopreventive capacity of nonnatural ITC candidates. To facilitate rapid evaluation and identification of promising lead ITCs, a reporter assay which correlates ARE activation to a spectrophotometric response was utilized. The reporter vector contained eight ARE motifs upstream of a SV40 promoter and the gene for firefly luciferase.[38] Induction of Nrf2 would result in the production of luciferase which could be quantified using a commercial luciferase assay, then normalized to lysate protein concentration. To maintain consistency between antiproliferation studies and ARE-induction studies, human MCF- 7 cells stably transfected with a luciferase reporter vector (MCF- 7-ARE cells) were utilized.[30]</p><p>To evaluate the effectiveness of this reporter construct, MCF-7- ARE cells were treated with 1 over the concentration range 200– 0.1 pm and were evaluated between 3–48 h after treatment (Figure 6). The normalized response increased with the concentration of 1 and peaked between 5–10 pm; at higher concentrations, the normalized response sharply declined. This decline appears to be due to a reduction in the detected relative light units (RLU), the quantitative measure of luciferase activity that is proportional to the concentration of luciferase (see Figure SI-2B); the protein concentration remained constant over the range of concentrations evaluated (see Figure SI-2A). It is plausible that higher concentrations of 1 could promote a stress response in MCF-7-ARE cells which affects the stability, longevity, and activity of any luciferase produced by the reporter vector. Consequently, subsequent ARE studies were limited to ITC concentrations below 12.5 pm. Experiments conducted using MCF-7-ARE and parental MCF-7 cell in parallel demonstrated that an observed luciferase response was specific to MCF-7-ARE cells containing the vector and luciferase gene (see Figure SI-3). The data in Figure 6 also suggests that the maximal reporter response occurred 24 h after inoculation with 1; prior to 12 h, minimal ARE-activation was observed. These findings are consistent with previous reports describing the dose- and time- dependence of Nrf2-activation following treatment with 1,[39] further supporting the use of this reporter construct to screen the chemopreventive capacity of aryl ITC analogues.</p><p>Despite the semi-quantitative nature of data from the ARE reporter assay, a number of SARs for aryl ITC analogues have been identified. A clear trend was observed concerning the chain length between the isothiocyanate functional group and the aryl ring; after 24 h, none of the PITC analogues (n = 0) demonstrated significant capacity for ARE induction over the dose range, while BITC analogues (n = 1) all demonstrated some degree of ARE induction. While (1) the magnitude of ARE induction, (2) the concentration at which maximal ARE induction was observed, and (3) the presence of a sharp decline in ARE induction at higher concentrations differed between individual BITC analogues, it is unclear whether most of these differences are statistically significant. Qualitatively, it appears that trifluoromethyl- substituted BITC analogues (36–38) demonstrated consistently- reduced capacity for ARE induction. Although 1 demonstrated diminished (but measurable) ARE induction after 72 h, many of the BITC analogues demonstrated little to no capacity for ARE induction after 72 h. Those with noticeable ARE induction after 72 h included meta and para isomers of methylsulfanyl- (57–58), methylsulfinyl- (63–64), and methylsulfonyl-substituted BITC analogues (69–70); the activity of ortho isomers (56, 62, and 68) were diminished after 72 h. Interestingly, analogues 63, 64, and 70 were the only aryl ITC analogues that paralleled the ARE induction profile for 1, which lacked an observed decline in ARE induction over the concentrations tested. It is plausible that this similarity may be due to the presence of a methylsulfinyl or methylsulfonyl group, which is a common structural feature among these ITCs. Analogue 57 was unique among the ITCs evaluated, as its significant ARE induction did not noticeably change between 24 h and 72 h incubation. Given the prominent and well-characterized capacity of 1 to act as an ARE inducer, aryl ITCs 57, 63, 64, and 70 were identified as potential lead compounds due to their strong similarity in ARE-induction profile.</p><!><p>A secondary outcome of evaluating both the antiproliferative and chemopreventive modes of activity exhibited by aryl ITC analogues was to determine if there were any correlating factors between the two types of data. This underlying analysis served as a motivation to depict the time-and dose-dependence of each ITC analogue on a single plot (Figures 3–5). One of the most interesting features of the antiproliferation data for ITC analogues was the significant and unexpected presence of NSRs. With the exception of ITC 33, every analogue which demonstrated an antiproliferation NSR after 24 h or 72 h also demonstrated ARE induction through the luciferase reporter assay; the opposite correlation does not appear valid, as several compounds (38, 43, 44, 56, 58) elicited ARE induction without an evident non- sigmoidal antiproliferation response. Furthermore, the data suggest that these two effects appear to occur over a similar concentration range. This qualitative observation is complicated by the sharp reduction in ARE induction exhibited by many ITC analogues, which, as previously mentioned, may be an undesired artifact of the reporter assay. It is also possible that these two responses result from different cellular mechanisms, similar to conclusions drawn in a previous study.[37]</p><p>From a mechanistic perspective, cellular induction of the ARE by ITCs are known to result in increased transcription of several phase II genes and a chemopreventive response. It is plausible that ARE induction may be related to the activation of other, prosurvival responses that lead to increased relative proliferation, observed as a NSR. Based on the data, a NSR response only appears to occur over a relatively narrow concentration window (1–2 two-fold dilutions = approximately 2–4-fold change in [ITC]). Beyond this range, antiproliferation activity is reasserted, suggesting that an ITC-induced pro-survival response has limited capacity. Were these NSR effects only observed after treatment with substituted aryl ITC analogues, it would be easier to speculate that these non-natural agents were eliciting differentcellular responses than their natural analogues. However, both natural analogues 1 and 3 also demonstrate similar trends in antiproliferation dose-response, ARE induction, and a NSR. Additional studies to more fully understand the interplay between these types of cellular responses are ongoing and will be reported in due course.</p><!><p>In order to validate the screening effectiveness of the luciferase ARE reporter assay, a subset of five ITCs were selected for evaluation of their capability to induce expression of ARE promoted genes. L-SFN (1) was included as a well-described ARE inducer and positive control. Phenyl isothiocyanate (2) was a non-natural aryl ITC that demonstrated negligible capacity for ARE induction by the luciferase reporter assay. ITC 37 was selected as a representative substituted BITC analogue which elicited a non-sigmoidal antiproliferative response and a low capacity for ARE induction. ITC 57 was included for its strong and consistent ARE induction after both 24 h and 72 h incubation. ITC 64 was selected as a representative analogue which demonstrates ARE induction after both 24 h and 72 h and a non- sigmoidal antiproliferative response, criteria which are qualitatively similar to the activity profile of L-SFN (1). Human MCF-7 cells were treated with the selected ITCs over the same concentration range used in the luciferase reporter assay (12.5—0.4 pm) and gene expression was determined by quantitative polymerase chain reaction (qPCR). The levels of four well- described redox-sensitive, ARE-promoted genes were analyzed in parallel to provide a more-complete picture of the changes in ARE-promoted transcriptional response: NAD(P)H:quinone oxidoreductase (NQ01), heme oxygenase 1 (HMOX1),glutathione S-transferases a1 (GSTal), and thioredoxin reductase 1 (TXNRD1). Transcriptional responses were normalized to glyceradehyde-3-phosphate dehydrogenase (GAPDH).</p><p>Dose-response curves for the transcriptional expression of genes following treatment by the five selected ITCs are depicted in Figure 7. As a positive control, treatment with L-SFN (1) increased expression of all four genes in a dose-dependent fashion. Treatment with ITC 2 did not elicit significant transcription of any of the four genes, consistent with the ARE induction data from the luciferase reporter assay. Treatment with ITC analogues 37, 57, and 64 increased expression of all four analyzed genes. Unlike the data resulting from the evaluation of these analogues through the luciferase reporter assay, gene expression increased over the dose range that was tested; the sharp decline in ARE-induction from the reporter assay was not observed through the qPCR analysis, an observation which supports the premise that this decline may be an artifact of the reporter system. The increases in gene expression for 37 and 57 were similar in magnitude, both between the four gene targets and in comparison to the effects of 1. The major exception to this trend related to the significantly increased expression of HMOX1 following treatment with 37. The underlying rationale for this large and gene-specific difference is unclear and the subject of ongoing investigation. Of the ITC analogues evaluated, 64 was found to be an incredibly strong inducer of all four ARE-promoted genes. The magnitudes of the transcriptional responses elicited by 64 are especially compelling when normalized to the expression profile for L-SFN (1), a well- described ARE inducer. Expression of NQ01 (10–21 fold), HMOX1 (0.7–15 fold), GSTal (1.4–8.6 fold), and TXNRD1 (0.8– 3.8 fold) were all significantly larger following treatment with 64 relative to 1, especially at the higher ITC concentrations.</p><p>Taken together, the qPCR data provides support to validate the use of the luciferase reporter assay during initial evaluation of ITC candidates. It appears that the greatest strength of the luciferase reporter lies in discerning compounds which are capable of inducing expression of ARE-promoted genes (e.g. 1, 37, 57, 64) from those that are incapable (e.g. 2). However, there appears to be minimal correlation between both the magnitude and gene specificity of a response in the luciferase reporter assay and the magnitude of actual gene transcription assessed by qPCR</p><!><p>The central hypothesis of this study was that the anticancer properties of naturally-occurring, unsubstituted aryl ITCs are not optimized and that substituted aryl ITC variants will demonstrate differential (and hopefully improved) anticancer properties. To test this hypothesis, a panel of 36 substituted variants of PITC and BITC were identified and prepared. The time- and dose- dependence of ITC candidates were systematically evaluated against human MCF-7 breast cancer cells using two differential screening methods; one assay evaluated antiproliferative activity while the second utilized a luciferase reporter construct to indirectly measure induction of ARE-promoted genes. Together, the combined body of data resulting from these studies identified key several structure-activity relationships for both potential modes of aryl ITC bioactivity. In relation to the original hypothesis, these screening methods have provided the means to differentiate each candidate's anticancer activity profile and, in part, to identify structural features which confer improved modal selectivity. Several of the ITC analogues demonstrated antiproliferative properties with a non-sigmoidal dose-response, a unique feature that has not previously been documented for this class of compounds. Efforts to understand and rationalize this effect are ongoing and will be reported in due course. While the described luciferase reporter assay provided an efficient method to evaluate ARE induction, further efforts and revisions to this methodology will continue to address some of its limitations, including the need for improved dynamic range. Finally, p-methylsulfinyl BITC (64) was identified as a potent inducer of ARE-promoted genes, approximately an order of magnitude stronger than the natural product L-SFN (1). The anticancer properties of ITC 64, and other related analogues, will continue to be evaluated and explored as a new class of non-natural aryl ITC variants.</p><!><p>All reactions were carried out under nitrogen unless indicated otherwise. All reagents were obtained from available commercial sources and were used without further purification unless otherwise noted. Melting point analyses were conducted using an open capillary tube, unless otherwise noted. The silica gel used in flash chromatography was 60 A, 230–400 mesh. Analytical TLC was performed on Uniplate 250 pm silica gel plates with detection by UV light. NMR spectra were acquired on a JEOL ECS- 400 400 MHz NMR spectrometer with multinuclear capability and 24- sample autosampler, with solvent as internal reference; the chemical shifts are reported in ppm, in δ units. Infrared spectra were acquired on a Nicolet Avatar FTIR. High resolution mass spectroscopic data were obtained at the Mass Spectrometry & Analytical Proteomics Laboratory at the University of Kansas (Lawrence, KS). Cell viability absorption wasmeasured using a SpectraMax M2 Multi-Mode Microplate Reader. Firefly luciferase activity was measured using a Glomax 96-well Luminometer (Promega). RNA integrity was determined using an Agilent 2100 Bioanalyzer. Quantitative real-time PCR analysis was conducted using an Applied Biosystems 7500 Real Time PCR System.</p><!><p>To a solution of 2- hydroxypyridine (19.05 g, 200 mmol) and triethylamine (29.3 mL, 210 mmol) in dry CH2CI2 (300 mL) at 0 °C was slowly added thiophosgene (7.60 mL, 99 mmol) over 5 min. After warming to rt over 3 h, the reaction was diluted with saturated aqueous sodium bicarbonate (300 mL). The organic layer was collected, was washed with saturated aqueous sodium chloride (300 mL), dried (Na2S04), and concentrated. Recrystallization from CH2Cl2/hexanes afforded 26 as a colorless solid (17.80 g, 77%): m.p. 109.6–113.1 °C; 1H NMR (CDC3, 400 MHz) δ 8.46 (d, J = 4.6 Hz, 2H), 7.86 (tt, J = 7.3, 1.8 Hz, 2H), 7.30 (m, 2H), 7.20 (d, J = 8.2 Hz, 2H); 13C NMR (CDCb, 100 MHz) 6 192.3, 159.4 (2C), 148.9 (2C), 140.1 (2C), 123.2 (2C), 116.9 (2C); IR (filrn)Vmax 3075, 2993, 2883,2825, 1722, 1702, 1678, 1672, 1658, 1650, 1643, 1612, 1573, 1468, 1433, 1372, 1302, 1280, 1224, 1170, 1095, 1046, 994, 907, 853, 774, 728, 562 cm−1.</p><!><p>ITC 2 was prepared as previously described.[20]</p><!><p>ITC 3 was prepared as previously described.[20]</p><!><p>To a solution of o-toluidine (140 pL, 1.34 mmol) in dry CH2CI2 (19.5 mL) at rt was added 26 (413 mg, 1.72 mmol). The reaction was stirred for 20 h and the solvent was concentrated. Flash chromatography (silica gel, 25:1 hexanes:EtOAc) afforded 27 as a colorless oil (172 mg, 86%): 1H NMR (CDCb, 400 MHz) 6 7.24–7.16 (m, 4H), 2.40 (s, 3H); 13C NMR (CDCb, 100 MHz) 6 135.4, 135.2, 130.9, 130.5, 127.6, 127.1, 126.2, 18.6; IR (film) vmax 2923, 2854, 2175, 2086, 1598, 1580, 1501, 1485, 1460, 1381, 1290, 1115, 1037, 929 cm'1; HRMS (EI+) m/z [M]+ calcd for C8H7NS, 149.0299; found, 149.0293.</p><!><p>To a solution of m-toluidine (0.12 mL, 1.05 mmol) in dry CH2CI2 (11.0 mL) at rt was added 26 (268 mg, 1.16 mmol). The reaction was stirred for 24 h and the solvent was concentrated. Flash chromatography (silica gel, hexanes) afforded 28 as a colorless oil (93 mg, 59%): 1H NMR (CDCb, 400 MHz) 6 7.24 (t, J = 7.3 Hz, 1H), 7.12–7.01 (m, 3H), 2.35 (s, 3H); 13C NMR (CDCb, 100 MHz) 6 139.9, 135.0, 131.2, 129.5, 128.4, 126.5, 122.9, 21.4; IR (film) vmax 2921, 2229, 2136, 2099, 1604, 1583, 1484, 1484, 1453, 964, cm'1; HRMS (EI+) m/z [M]+ calcd for C8H7NS, 149.0299; found, 149.0277.</p><!><p>To a solution of p-toluidine (158 mg, 1.03 mmol) in dry CH2CI2 (15.0 mL) at rt was added 26 (292 mg, 1.26 mmol). The reaction was stirred for 12 h and the solvent was concentrated. Flash chromatography (silica gel, 25:1 hexanes:EtOAc) afforded 29 as a light yellow oil (126 mg, 82%): 1H NMR (CDCb, 400 Mz) 6 7.18–7.10 (m, 4H), 2.36 (s, 3H); 13C NMR (CDCb, 100 MHz) 6 137.7, 130.3 (2C), 128.5, 125.7 (2C), 21.4; IR(film)vmax 2923, 2848,2192, 2138, 2896, 1653, 1636, 1502,1456, 1324, 1113, 1101, 928, 815 cm4; HRMS (EI+) m/z. [M]+calcd for C8H7NS, 149.0299; found, 149.0286.</p><!><p>To a solution of 2- methylbenzylamine (100 μL, 0.81 mmol) in dry CH2CI2 (12.0 mL) at rt was added 26 (289 mg, 1.24 mmol). The reaction was stirred for 17 h and the solvent was concentrated. Flash chromatography (silica gel, 5:1 hexanes:EtOAc) afforded 30 as a yellow oil (130 mg, 99%): 1H NMR (CDCl3, 400 MHz) δ 7.32–7.23 (m, 4H), 4.70 (s, 2H), 2.36 (s, 3H); 13C NMR (CDCb, 100 MHz) δ 136.0, 132.5, 132.0, 130.9, 128.9, 128.1, 126.7, 47.3, 19.0; IR (film) Vmax 3067, 3023, 2924, 2858, 2168, 2087, 1695, 1652, 1606, 1493, 1461, 1437 1340 cm−1; HRMS (EI+) m/z [M]+ calcd for C9H9NS, 163.0456; found, 163.0476.</p><!><p>To a solution of 3- methylbenzylamine (98 μL, 0.81 mmol) in dry CH2CI2 (12.0 mL) at rt was added 26 (289 mg, 1.16 mmol). The reaction was stirred for 17 h and the solvent was concentrated. Flash chromatography (silica gel, 25:1 hexanes:EtOAc) afforded 31 as a colorless oil (129 mg, 98%): 1H NMR (CDCI3, 400 MHz) δ 7.29 (t, J = 7.8 Hz, 1H), 7.19–7.10 (m, 3H), 4.69 (s, 2H), 2.39 (s, 3H); 13C NMR (CDCb, 100 MHz) δ 139.0,134.3, 132.1, 129.3, 127.8, 124.1, 48.3, 21.6; IR (film) Vmax 3026, 2960, 2922, 2854, 2165, 2094, 1610.18, 1492, 1439, 1338, 1260, 1095, 1019 cm'1; HRMS (EI+) m/z [M]+ calcd forC9H9NS, 163.0456; found, 163.0450.</p><!><p>To a solution of 4- methylbenzylamine (100 pL, 0.70 mmol) in dry CH2CI2 (10.5 mL) at rt was added 26 (250 mg, 1.08 mmol). The reaction was stirred for 17 h and the solvent was concentrated. Flash chromatography (silica gel, hexanes) afforded 32 as a colorless oil (41 mg, 29%): 1H NMR (CDCIs): δ 7.22 (s, 4H), 4.67 (s, 2H), 2.39 (s, 3H); 13C NMR (CDCb): δ 138.3, 131.9, 131.3, 129.7 (2C), 127.0 (2C), 48.6, 21.3; IR (film) Vmax 3050, 3025, 2922, 2857, 2173, 2092, 1616, 1516,1438, 1414,1379, 1344, 1308, 1281, 1265, 1240, 1201, 1183, 1121, 1040, 1021 cm'1; HRMS (EI+) m/z [M]+ calcd for C9H9NS, 163.0456; found, 163.0470.</p><!><p>To a solution of 2- (trifluoromethyl)aniline (100 pL, 0.80 mmol) in dry CH2CI2 (12.0 mL) at rt was added 26 (273 mg, 1.17 mmol). The reaction was stirred for 72 h and the solvent was concentrated. Flash chromatography (silica gel, 25:1 hexanes:EtOAc) afforded 33 as a colorless oil (123 mg, 76%): 1H NMR (CDCb, 400 MHz) δ 7.67 (d, J = 7.8 Hz, 1H), 7.56 (td, J = 8.2, 0.9 Hz, 1H), 7.42–7.35 (m, 2H); 13C NMR (CDCb, 100 MHz) δ 138.0, 133.2, 129.5, 127.1 (q, J = 4.8 Hz), 127.0, 125.9 (q, J = 31.6 Hz), 122.9 (q, J = 273.2 Hz); IR (film) Vmax 2958, 2926, 2855, 2092, 1603, 1586, 1493, 1460, 1453, 1319, 1269, 1175, 1133, 1115, 1062, 1037, 942 cm-1; HRMS (EI+) m/z [M]+ calcd for C8H4F3NS, 203.0017; found, 203.0027.</p><!><p>To a solution of (trifluoromethyl)aniline (95 μL, 0.88 mmol) in dry CH2CI2 (11.1 mL) at rt was added 26 (264 mg, 1.14 mmol). The reaction was stirred for 40 h and the solvent was concentrated. Flash chromatography (silica gel, hexanes) afforded 34 as a colorless oil (70 mg, 39%): 1H NMR (CDCb, 400 MHz) δ 7.56–7.47 (m, 3H), 7.41 (m, 1H); 13C NMR (CDCb, 100 MHz) δ 138.3, 132.5 (q, J = 33.6 Hz), 130.5, 129.1, 124.0 (q, J = 3.8 Hz), 123.3 (q, J = 272.2 Hz), 122.9 (q, J = 3.8 Hz); IR (film) Vmax 3446, 2962, 2926, 2843, 2199, 2046, 1700, 1635,1614, 1590,1558, 1488, 1448, 1331, 1233, 1175, 1131, 1093, 1065 cm−1; HRMS (EI+) m/z [M]+ calcd for C8H4F3NS, 203.0017; found, 203.0006.</p><!><p>To a solution of (trifluoromethyl)aniline (100 μL, 0.80 mmol) in dry CH2CI2 (11.1 mL) at rt was added 26 (267 mg, 1.14 mmol). The reaction was stirred for 41 h and the solvent was concentrated. Flash chromatography (silica gel, 25:1 hexanes:EtOAc) afforded 35 as a colorless solid (149 mg, 91%): m.p. 40.1–42.6 °C; 1H NMR (CDCl3, 400 MHz) δ 7.63 (d, J = 8.2 Hz, 2H) , 7.33 (d, J = 8.2 Hz, 2H); 13C NMR (CDCb, 100 MHz) δ 138.5, 135.2, 129.3, (q, J = 32.6 Hz),127.0 (q, J = 3.8 Hz, 2C), 126.2 (2C), 123.8 (q, J = 272.2 Hz); IR (film) Vmax 2958, 2924, 2856, 2361,2196, 2121, 1611, 1505, 1412, 1327, 1154, 1123, 1105, 1064, 931 cm'1; HRMS (EI+) m/z [M]+ calcd for C8H4F3NS, 203.0017; found, 203.0015.</p><!><p>To a solution of 2-(trifluoromethyl)benzylamine (100 pL, 0.72 mmol) in dry CH2CI2 (11.2 mL) at rt was added 26 (267 mg, 1.14 mmol). The reaction was stirred for 17 h and the solvent was concentrated. Flash chromatography (silica gel, hexanes) afforded 36 as a colorless oil (153 mg, 99%): 1H NMR (CDCb, 400 MHz) δ 7.70 (d, J= 7.3 Hz, 1H), 7.67–7.61 (m, 2H), 7.51–7.45 (m, 1H), 4.97 (s, 2H); 13C NMR (CDCb, 100 MHz) δ 134.0, 132.8, 132.7, 129.2, 128.7, 127.6 (q, J = 30.7 Hz), 126.5 (q, J = 5.6 Hz), 124.1 (q, J = 274.1 Hz), 45.9; IR (film) Vmax 3074, 2926, 2855, 2184, 2090, 1609, 1587, 1498,1457, 1440,1353, 1314, 1171, 1121, 1060, 1039, 948 cm−1; HRMS (EI+) m/z [M]+ calcd for C9H6F3NS, 217.0173; found, 217.0177.</p><!><p>To a solution of 3-(trifluoromethyl)benzylamine (180 pL, 1.24 mmol) in dry CH2CI2 (12.0 mL) at rt was added 26 (289 mg, 1.24 mmol). The reaction was stirred for 16 h and the solvent was concentrated. Flash chromatography (silica gel, 5:1 hexanes: EtOAc) afforded 37 as a yellow oil (142 mg, 53%): 1H NMR (CDCl3, 400 MHz) δ 7.66–7.60 (m, 1H), 7.59– 7.53 (m, 3H), 4.81 (s, 2H); 13C NMR (CDCl3, 100 MHz) δ 135.5, 134.1, 131.6 (q, J = 32.6 Hz), 130.3, 129.8, 125.5 (q, J = 3.8 Hz), 123.9 (q, J = 272 Hz), 123.9 (q, J = 3.8 Hz), 48.3; IR (film) Vmax 3068, 2926, 2855, 2176, 2095, 1618, 1600, 1496, 1453, 1439, 1350, 1328, 1274, 1196, 1167, 1126, 1074 cm-1; HRMS (EI+) m/z [M]+ calcd for C9H6F3NS, 217.0173; found, 217.0174.</p><!><p>To a solution of 4-(trifluoromethyl)benzylamine (120 pL, 0.84 mmol) in dry CH2CI2 (12.0 mL) at rt was added 26 (288 mg, 1.24 mmol). The reaction was stirred for 17 h and the solvent was concentrated. Flash chromatography (silica gel, 25:1 hexanes:EtOAc) afforded 38 as a colorless oil (159 mg, 87%): 1H NMR (CDCb, 400 MHz) δ 7.67 (d, J = 8.2 Hz, 2H), 7.46 (d, J = 8.2 Hz, 2H), 4.81 (s, 2H); 13C NMR (CDCb, 100 MHz) δ 138.4, 133.9, 130.9 (q, J = 32.6 Hz), 127.3 (2C), 126.2 (q, J = 3.8 Hz, 2C), 124.0 (q, J = 272.2 Hz), 48.4; IR (film) Vmax 2927, 2855, 2187, 2096, 1622, 1437, 1420, 1326, 1239, 1167, 1126, 1067, 1019, 945, 819 cm'1; HRMS (EI+) m/z. [M]+ calcd forC9H6F3NS, 217.0173; found, 217.0170.</p><!><p>To a solution of 2-nitroaniline (298 mg, 2.16 mmol) in dry CH2CI2 (59.0 mL) at rt was added 26 (769 mg, 3.31 mmol). The reaction was stirred for 7 days and the solvent was concentrated. Flash chromatography (silica gel, 5:1 hexanes:EtOAc) afforded 39 as a near-colorless solid (152 mg, 39%): m.p. 70.8–71.6 °C; 1H NMR (CDCL3, 400MHz) δ 8.10 (dd, J = 8.2, 1.4 Hz, 1H), 7.64 (td, J = 7.8, 1.4 Hz, 1H), 7.47–7.39 (m, 2H); 13C NMR (CDCb, 100 MHz) δ 145.6, 139.3, 134.6, 129.4, 127.4, 126.1; IR (film) Vmax 3054, 2987, 2306, 1607, 1422, 1265, 896, 739, 705 cm−1; HRMS (EI+) m/z [M]+ calcd for C7H4N202S, 179.9993; found, 179.9995.</p><!><p>To a solution of 3-nitroaniline (321 mg, 2.32 mmol) in dry CH2CI2 (60.0 mL) at rt was added 26 (766 mg, 3.30 mmol). The reaction was stirred for 47 h and the solvent was concentrated. Flash chromatography (silica gel, 5:1 hexanes:EtOAc) afforded 40 as a pale yellow solid (411 mg, 98%): m.p. 57.2–58.0 °C; 1H NMR (CDCL3, 400 MHz) δ 8.14 (dt, J= 7.3, 2.3 Hz, 1H), 8.08 (m, 1H), 7.60– 7.53 (m, 2H); 13C NMR (CDCb, 100 MHz): δ 149.0, 139.9, 133.5, 131.7,130.7, 122.0, 120.9; IR (film) Vmax 3234, 3055, 1596, 1532, 1444, 1354, 1288, 1265, 1213, 1146, 1054, 897, 825,799, 736, 679 cm−1; HRMS (EI+) m/z [M]+ calcd for C7H4N202S, 179.9993; found, 180.0002.</p><!><p>To a solution of 4-nitroaniline (299 mg, 2.16 mmol) in dry CH2CI2 (60.0 mL) at rt was added 26 (761 mg, 3.27 mmol). The reaction was stirred for 65 h and the solvent was concentrated. Flash chromatography (silica gel, 5:1 hexanes:EtOAc) afforded 41 as a pale yellow solid (389 mg, 99%): m.p. 110.1–111.2 °C; 1H NMR (CDCb, 400 MHz) 8.26 (m, 2H), 7.36 (m, 2H); 13C NMR (CDCb, 100 MHz) δ 146.0, 140.5,138.1, 126.6(2C), 125.5 (2C); IR (film) Vmax 3233, 3054, 2361,2337, 1596, 1558, 1507, 1495, 1457, 1364, 1308, 1266, 1213, 1137, 1041,896, 843, 737, 706, 667 cm−1; HRMS (EI+) m/z [M]+ calcd for C7H4N202S, 179.9993; found, 179.9995.</p><!><p>To a solution of 2- nitrobenzylamine hydrochloride (304 mg, 1.61 mmol) in dry CH2CI2 (23.0 mL) at rt was added 26 (1.098 g, 4.36 mmol) and N,N- diisopropylethylamine (270 pL, 1.55 mmol). The reaction was stirred for 17 h and the solvent was concentrated. Flash chromatography (silica gel, 5:1 hexanes:EtOAc) afforded 42 as a yellow solid (212 mg, 64%): m.p. 69.3– 70.2 °C; 1H NMR (CDCl3, 400 MHz) δ 8.21 (d, J = 8.2 Hz, 1H), 7.81–7.73 (m, 2H), 7.57 (m, 1H), 5.29 (s, 2H); 13C NMR (CDCIs, 100 MHz) δ 146.7, 135.0, 134.7, 130.7, 129.5, 129.3, 125.7, 47.3; IR (film) Vmax 3054, 2987, 2305, 2094, 1529, 1421,1351, 1265, 896, 738, 705 cm−1; HRMS (EI+) m/z [M]+ calcd for C8H6N202S, 194.1050; found, 194.0144.</p><!><p>To a solution of 3- nitrobenzylamine hydrochloride (298 mg, 1.58 mmol) in dry CH2CI2 (23.0 mL) at rt was added 26 (1.097 g, 4.72 mmol) and N,N- diisopropylethylamine (280 pL, 1.61 mmol). The reaction was stirred for 16 h and the solvent was concentrated. Flash chromatography (silica gel, 5:1 hexanes:EtOAc) afforded 43 as a pale yellow solid (299 mg, 97%): m.p. 78.8–79.6 °C; 1H NMR (CDCIs, 400 MHz) δ 8.26–8.20 (m, 2H), 7.71 (d, J = 8.2 Hz, 1H), 7.62 (t, J= 7.8 Hz, 1H), 4.87 (s, 2H); 13C NMR (CDCIs, 100 MHz) δ 148.7, 136.6, 135.1, 132.9, 130.4, 123.6, 122.1,48.2; IR(film)vmax 3395, 3054, 2987, 2360, 2341, 1534,1516, 1422, 1393, 1352, 1265, 1204, 1147, 989, 896, 739, 705, 668, 446 cm−1; HRMS (EI+) m/z [M]+ calcd for CsHeNsOsS, 194.1050; found, 194.0141.</p><!><p>4-Nitrobenzylamine hydrochloride (301 mg, 1.596 mmol) was dissolved in dry CH2CI2 (23.0 mL) at rt. N,N-Diisopropylethylamine (0.270 mL, 1.633 mmol) was added, followed by 26 (1.0811 g, 4.654 mmol). The reaction mixture was stirred for 68 h, followed by solvent removal in vacuo. The residue was purified by flash chromatography (silica gel, 5:1 hexanes:EtOAc) to afford 44 as a yellow oil (202 mg, 65%): 1H NMR (CDCIs, 400MHz): δ 8.28 (m, 2H), 7.52 (m, 2H), 4.88 (s, 2H); 13C NMR (CDCIs, 100 MHz): δ 147.9, 141.4, 134.9, (2C), 124.3 (2C), 48.1; IR (film) Vmax 3403, 3054, 2987, 2361, 1653, 1521, 1421, 1350, 1265, 896, 739, 705, 449 cm−1; HRMS (EI+) m/z [M]+ calcd for CsHeNsOsS, 194.1050; found, 194.0147.</p><!><p>To a solution of hydroxylamine hydrochloride (944 mg, 13.58 mmol), pyridine (2.45 mL, 30.9 mmol) and EtOH (25.0 mL) was slowly added 4- (methylthio)benzaldehyde (0.87 mL, 6.54 mmol). The reaction was heated to reflux for 4 h and the solvents were concentrated. The residue was dissolved in a solution of L1BH4 in dry THF (2.0 M, 30 mL, 60.00 mmol) and was heated to reflux for 17 h. The reaction was diluted with EtOAc (100 mL) and aqueous HCI (6M, 100 mL). The aqueous phase was extracted with EtOAc (3 × 100 mL), adjusted to pH 13, and extracted with CHsCIs (3 × 100 mL). The combined organic layers were concentrated, the residue was dissolved in dry MeOH, and dry, gaseous HCI was passed through for 5 min. The solvent was concentrated and the residue was recrystallized from EtOAc:hexanes to afford 52 as a pale colorless solid. (608 mg, 49%): m.p. > 260 °C; 1H NMR (CD3OD, 400 Mz) δ 7.37 (m, 2H), 7.32 (m, 2H),4.06 (s, 2H), 2.49 (s, 3H); 13C NMR (CD3OD, 100 MHz) δ 142.0, 130.7, 130.5 (2C), 127.5 (2C), 43.9, 15.2; IR (film) Vmax 3424, 3005, 2581, 2037, 2037, 1594, 1496, 1479, 1463, 1434, 1409, 1384, 1094, 901, 821, 793, 525 cm−1; HRMS (ESI+) m/z: [M - Cl]+ calcd forC8Hi2NS, 154.0690; found, 154.0697.</p><!><p>To a solution of 2- (methylthio)aniline (0.42 mL, 3.35 mmol) in dry CH2CI2 (48.0 mL) at rt was added 26 (1.16 g, 4.99 mmol). The reaction was stirred at rt for 67 h and the solvent was concentrated. Flash chromatography (silica gel, 20:1 hexanes:EtOAc) afforded 53 as a light yellow oil (598 mg, 99%): 1H NMR (CDCI3, 400MHz): δ 7.28–7.19 (m, 3H), 7.13 (ddd, J = 7.8, 6.8, 1.8, 1H), 2.50 (s, 3H); 13C NMR (CDCIs, 100 MHz): δ 137.7, 136.9, 129.4, 127.8, 126.4, 125.9, 15.6; IR (film) Vmax 3061, 2986, 2920, 2850, 2165, 2071, 1579, 1567, 1464. 1319, 1283, 1235, 1163, 1130, 1076, 1036, 967, 955, 930, 851, 747, 731, 707, 668, 658, 529, 419, 412 cm−1; HRMS (EI+) m/z [M]+ calcd for C8H7NS2, 181.0020; found, 181.0004.</p><!><p>To a solution of 3- (methylthio)aniline (0.42 mL, 3.37 mmol) in dry CH2CI2 (48.0 mL) at rt was added 26 (1.17 g, 5.02 mmol). The reaction was stirred at rt for 67 h and the solvent was concentrated. Flash chromatography (silica gel, 10:1 hexanes:EtOAc) afforded 54 as a yellow oil (585 mg, 96%): 1H NMR (CDCIs, 400MHz): δ 7.25 (t, J= 7.8, 1H), 7.14 (ddd, J = 7.8, 1.8, 0.9, 1H), (t, J = 1.8, 1H), 6.98 (ddd, J = 7.8, 1.8, 0.9, 1H), 2.49 (s, 3H); 13C NMR (CDCIs, 100 MHz): δ 141.0, 136.0, 132.0, 129.8, 125.2, 123.0, 122.2, 15.6; IR (film) Vmax 3059, 2985, 2919, 2851,2195, 2106, 1934, 1582, 1554, 1472, 1434, 1423, 1415, 1272, 1096, 1078, 995, 957, 856, 775, 751, 719, 677, 657, 557, 540, 532, 523, 511,502, 480, 467, 446, 437, 434, 429, 423, 418, 403 cm−1; HRMS (EI+) m/z [M]+ calcd for C8H7NS2, 181.0020; found, 181.0006.</p><!><p>To a solution of 4- (methylthio)aniline (0.41 mL, 3.30 mmol) in dry CH2CI2 (48.0 mL) at rt was added 26 (1.15 g, 4.96 mmol). The reaction was stirred at rt for 41 h and the solvent was concentrated. Flash chromatography (silica gel, 20:1 hexanes:EtOAc) afforded 55 as a light yellow oil (598 mg, 99%): 1H NMR (CDCIs, 400 MHz) δ 7.20 (m, 2H), 7.16 (m, 2H), 2.49 (s, 3H); 13C NMR (CDCIs, 100 MHz) δ 138.6, 135.3, 128.0, 127.2 (2C), 126.3 (2C), 15.9; IR (film) Vmax 2918, 2179, 2089, 1734, 1699, 1684, 1653, 1635, 1558, 1540, 1506, 1486, 1465, 1457, 1436, 1420, 1404, 1094, 928, 816, 496, 484, 474, 466, 458, 446, 436, 421, 411 cm−1; HRMS (EI+) m/z [M]+ calcd for CsH/NSs, 181.0020 ; found, 181.0005.</p><!><p>Hydroxylamine hydrochloride (1.826 g, 27.6 mmol) was dissolved in ethanol (95%, 50.0 mL) and pyridine (4.70 mL, 58.1 mmol). 2-(Methylthio)benzaldehyde (1.748 mL, 13.5 mmol) was added and the reaction was heated to reflux for 16 h. After concentration, UBH4 (2.0 M in THF, 30 mL, 60.0 mmol) was added and the reaction was heated to reflux for an additional 5 h. The residue was partitioned between aqueous HCI (6M, 100 mL) and EtOAc (100 mL) and the aqueous phase was extracted and washed with EtOAc (3 × 100 mL). The aqueous phase was adjusted to pH 12 with aqueous NaOH (6M) and extracted with CH2CI2 (3 × 100 mL). The organic layers were concentrated, the residue was dissolved in dry CH2CI2 (85 mL) and 26 (746 mg, 3.25 mmol) was added. The reaction was stirred for 90 h and the solvent was concentrated. Flash chromatography (silica gel, 12:1 hexanes:EtOAc) afforded 56 as a green-yellow oil (236 mg, 27% over three steps): 1H NMR (CDCIs, 400 MHz): δ 7.40–7.29 (m, 3H), 7.23 (td, J = 7.3, 1.4 Hz, 1H), 4.82 (s, 2H), 2.51 (s, 3H); 13C NMR (CDCIs, 100 MHz): δ 137.1, 132.7, 132.4, 129.3, 128.1, 127.2, 125.9, 47.1, 16.6; IR (film) Vmax 3421,3053, 2986, 2926, 2361,2339, 2092, 1653, 1471, 1437, 1341, 1265, 1046, 896, 739, 705, 668 cm−1; HRMS (EI+) m/z [M]+ calcd for C9H9NS2, 195.0176; found, 195.0167.</p><!><p>To a solution of 3-(thiomethoxy)benzylamine hydrochloride (948 mg, 5.00 mmol) and N,N- diisopropylethylamine (955 pL, 5.05 mmol) in dry CH2CI2 (75 mL) was added 26 (2.45 g, 10.53 mmol). After 18 h, the reaction was concentrated. Flash chromatography (silica gel, 12:1 hexanes:EtOAC) afforded 57 as a light orange oil (913 mg, 94%): 1H NMR (CDCIs, 400 MHz) δ 7.30 (t, J = Hz, 1H), 7.23–7.16 (m, 2H), 7.06 (d, J= 7.3 Hz, 1H), 4.68 (s, 2H), 2.49 (s, 3H); 13C NMR (CDCIs, 100 MHz) δ 139.8, 135.1, 132.8, 129.4, 126.3,124.6, 123.4, 48.6, 15.7; IR (film) Vmax 2953, 2921,2852, 2170, 2085, 1576, 1559, 1473, 1437, 1335, 1207, 1088, 968, 862, 775, 669cm−1; HRMS (EI+) m/z [M]+ calcd for C9H9NS2, 195.0176; found, 195.0171.</p><!><p>To a solution of 52 (340 mg, 1.79 mmol) in dry CH2CI2 (57.0 mL) at rt was added 26 (662 mg, 2.85 mmol) and N,N-diisopropylethylamine (0.33 mL, 1.89 mmol). The reaction was stirred for 18 h and the solvent was concentrated. Flash chromatography (silica gel, 10:1 hexanes: EtOAc) afforded 58 as a light yellow oil (343 mg, 98%): 1H NMR (CDCIs, 400 MHz) δ 7.29–7.22 (m, 4H), 4.67 (s, 2H), 2.50 (s, 3H); 13C NMR (CDCIs, 100 MHz) δ 139.3, 132.6, 127.6(2C), 127.0(2C); IR (film) Vmax 3053, 2986, 2925, 2305,2175, 2094, 1602, 1496, 1438, 1406, 1346, 1265, 1095, 1016, 896, 801, 739, 705 cm−1; HRMS (EI+) m/z [M]+ calcd for C9H9NS2, 195.0176; found, 195.0167.</p><!><p>To a solution of 53 (257 mg, 1.42 mmol) in dry CH2CI2 (4.30 mL) was slowly added mCPBA (70%, 353 mg, 1.43 mmol). The reaction was stirred at rt for 4 h and was diluted with CH2CI2 (70 mL). The organic layer was washed with saturated aqueous sodium bicarbonate (70 mL), saturated aqueous sodium chloride (70 mL), dried (NaSCb), and was concentrated to afford the mixture of sulfoxide and sulfone. Flash chromatography (silica gel, 3:1 hexanes: EtOAc - 1:2 hexanes: EtOAc) afforded 59 as an off-colorless crystalline solid (208 mg, 75%): m.p. 76.0–77.8 °C; 1H NMR (CDCb, 400 MHz) δ 7.91 (m, 1H), 7.49 (m, 1H), 7.34 (m, 1H), 2.18 (s, 3H); 13C NMR (CDCb, 100 MHz) δ 142.0, 141.0, 132.2, 128.5, 128.2, 127.8, 124.9, 42.6; IR (film) Vmax 3072, 2996, 2922, 2853, 2167, 2060, 1582, 1572, 1466, 1439, 1414, 1293, 1127, 1076, 1047, 936, 760, 714, 532, 515, 496, 467, 457, 443, 434, 428, 418, 404 cm−1; HRMS (EI+) m/z: [M + H]+ calcd for C8H8NOS2, 198.0047; found, 198.0065.</p><!><p>The purification of 59 also provided 65 as a colorless solid (52 mg, 17%): m.p. 92.8–93.4 °C; 1H NMR (CDCb, 400 MHz) δ 8.05 (dd, J = 7.8, 1.4, 1H), 7.64 (td, J = 7.8,1.8, 1H), 7.45 (m, 2H), 3.22 (s, 3H); 13C NMR (CDCb, 100 MHz) δ 141.3,135.1, 134.5, 130.7, 129.9, 129.5, 127.5, 43.5; IR (film) Vmax 3090, 3069, 3024, 3008, 2955, 2925, 2853, 2177, 2099, 1874, 1737, 1727, 1584, 1571, 1469, 1443, 1409, 1316, 1271, 1244, 1212, 1152, 1127, 1070, 1036, 955, 937, 875, 777, 759, 716, 654, 545, 536, 502, 475, 465, 454, 445, 431,421, 413, 405 cm−1; HRMS (EI+) m/z: [M + H]+ calcd for C8H8N02S2, 213.9996; found, 213.9988.</p><!><p>To a solution of 54 (254 mg, 1.40 mmol) in dry CH2CI2 (4.20 mL) was slowly added mCPBA (70%, 351 mg, 1.42 mmol). The reaction was stirred at rt for 2 h and was diluted with CH2CI2 (70 mL). The organic layer was washed with saturated aqueous sodium bicarbonate (70 mL), saturated aqueous sodium chloride (70 mL), dried (NaSCb), and was concentrated to afford the mixture of sulfoxide and sulfone. Flash chromatography (silica gel, 3:1 hexanes: EtOAc 1:2 hexanes: EtOAc) afforded 60 as a yellow oil (194 mg, 70%): 1H NMR (CDCL3, 400 MHz) δ 7.57–7.48 (m, 3H), 7.33 (dt, J =6.9, 2.3, 1H), 2.75 (s, 3H); 13C NMR (CDCb, 100 MHz) δ 148.1, 138.3,133.2, 130.7, 128.1, 122.1, 121.1, 44.2; IR (film) Vmax 3055, 2996, 2923, 2196, 2104, 1586, 1574,1545, 1536,1532, 1474, 1421, 1415, 1299, 1279, 1248, 1085, 1079, 1050, 996, 960, 789, 680, 478, 470, 466, 448, 444, 436, 420, 415, 403 cm−1; HRMS (EI+) m/z: [M + H]+ calcd for C8H8NOS2, 198.0047; found, 198.0060.</p><!><p>The purification of 59 also provided 66 as a light yellow solid (37 mg, 13%): m.p. 75.2– 77.5 °C; 1H NMR (CDCb, 400 MHz) δ 7.84 (dt, J = 7.8, 1.8 Hz, 1H), 7.82 (t, J = 1.8 Hz, 1H), 7.59 (t, J = 8.2 Hz, 1H), 7.49 (ddd, J = 8.2, 2.3, 0.9 Hz, 1H), 3.09 (s, 3H); 13C NMR (CDCb, 100 MHz) δ 142.4,139.4, 133.4, 131.0,130.7, 125.8, 124.9, 44.6; IR (film) Vmax 2955, 2923, 2853, 2092, 2072, 2022, 2012, 1985, 1963 ,1314, 1297, 1143, 1088, 1075, 970, 774, 731, 676, 575, 531, 517, 508, 493, 481, 458, 445, 440, 421, 412, 406 cm−1; HRMS (EI+) m/z: [M + H]+ calcd for C8H8N02S2, 213.9996; found, 213.9989.</p><!><p>To a solution of 55 (256 mg, 1.41 mmol) in dry CH2CI2 (4.20 mL) was slowly added mCPBA (70%, 358 mg, 1.45 mmol). The reaction was stirred at rt for 2 h and was diluted with CH2CI2 (70 mL). The organic layer was washed with saturated aqueous sodium bicarbonate (70 mL), saturated aqueous sodium chloride (70 mL), dried (NaSCb), and was concentrated to afford the mixture of sulfoxide and sulfone. Flash chromatography (silica gel, 3:1 hexanes:EtOAc 1:2 hexanes:EtOAc) afforded 61 as a light yellow solid (171 mg, 62%): m.p. 74.7–77.5 °C; 1H NMR (CDCb, 400 MHz) δ 7.65 (m, 2H), 7.37 (m, 2H), 2.73 (s, 3H); 13C NMR (CDCb, 100 MHz) δ 144.5, 138.1, 134.4, 126.7 (2C), 125.1 (2C), 44.1; IR (film) Vmax 3080, 3058, 3024, 2994, 2957, 2923, 2853, 2361,2338, 2180, 2094, 1734, 1717, 1700, 1684, 1653, 1587, 1559, 1540, 1506,1488, 1466,1457, 1418, 1402, 1293, 1250, 1087, 1051, 1013, 955, 931, 831, 727, 676, 668, 516, 484, 472, 453, 440, 434, 417, 401 cm−1; HRMS (EI+) m/z: [M + H]+ calcd for C8H8NOS2, 198.0047; found, 198.0032.</p><!><p>The purification of 61 also provided 67 as a colorless solid (91 mg, 30%): m.p. 133.7– 134.4 °C; 1H NMR (CDCb, 400 MHz) δ 7.95 (m, 2H), 7.40 (m, 2H), 3.07 (s, 3H); 13C NMR (CDCb, 100 MHz) δ 139.8, 138.9, 137.2, 129.4 (2C), 126.7 (2C), 44.6; IR (film) Vmax 3092, 3067, 3008, 3021,2926, 2192, 2079, 1653, 1587, 1576, 1559, 1540, 1506, 1301, 1284, 1172, 1142, 1085, 931, 832, 777, 727, 668, 531, 450, 441,434, 430, 418, 414, 404 cm−1; HRMS (EI+) m/z: [M + H]+ calcd for C8H8N02S2, 213.9996; found, 213.9979.</p><!><p>To a solution of 56 (174 mg, 0.89 mmol) in dry CH2CI2 (2.80 mL) was slowly added mCPBA (70%, 230 mg, 0.93 mmol). The reaction was stirred at rt for 2.5 h and was diluted with CH2CI2 (35 mL). The organic layer was washed with saturated aqueous sodium bicarbonate (35 mL), saturated aqueous sodium chloride (35 mL), dried (NaSCb), and was concentrated to afford the mixture of sulfoxide and sulfone. Flash chromatography (silica gel, 1:3 hexanes:EtOAc) afforded 62 as a yellow green oil (145 mg, 77%): 1H NMR (CDCb, 400 MHz): δ 8.05 (dd, J= 7.8, 1.4 Hz, 1H), 7.64 (td, J = 7.3, 1.4 Hz, 1H), 7.58 (td, J = 7.3, 1.4 Hz, 1H), 7.47 (d, J = 7.8 Hz, 1H), 4.90 (s, 2H), 2.81 (s, 3H); 13C NMR (CDCb, 100 MHz): δ 144.5, 132.4,132.1, 131.6, 130.7, 129.4, 124.7, 45.6, 43.8; IR (film) Vmax 3054, 2987, 2360, 2340, 1653, 1559, 1540, 1507, 1420, 1265, 896, 740, 705, 668 cm- 1; HRMS (EI+) m/z: [M + H]+ calcd for C9HioNOS2, 212.0204; found, 212.0217.</p><!><p>To a solution of 57 (789 mg, 4.04 mmol) dissolved in dry CH2CI2 (15.0 mL) was slowly added mCPBA (70%, 1.64 g, 6.65 mmol). The reaction was stirred at rt for 2 h and was diluted with CH2CI2 (50 mL). The organic layer was washed with saturated aqueous sodium bicarbonate (2 × 70 mL), saturated aqueous sodium chloride (70 mL), dried (NaSCb), and was concentrated. Flash chromatography (silica gel, 5:1:1 hexanes:CH2CI2:EtOAc —► 1:1 CH2CI2:EtOAc) afforded 63 as a brown oil (156 mg, 18%): 1H NMR (CDCb, 400 MHz) δ 7.16 (s, 1H), 7.60–7.52 (m, 2H), 7.46 (m, 1H), 4.80 (s, 2H), 2.74 (s, 3H); 13C NMR (CDCb, 100 MHz) δ 147.0, 136.3, 134.0, 130.2, 123.7, 122.0, 48.5, 44.2; IR (film) Vmax 3455, 2997, 2923, 2853, 2178, 2095, 1599, 1476, 1429, 1342, 1084, 1049, 997, 957, 788, 710, 687 cm−1; HRMS (EI+) m/z: [M + H]+ calcd for C9HioNOS2, 212.0204; found, 212.0199.</p><!><p>The purification of 63 also provided 69 as a colorless solid (52 mg, 17%): m.p. 67.7–68.0 °C; 1H NMR (CDCb, 400 MHz) δ 7.92 (dt, J = 6.4, 1.8 Hz, 1H), 7.88 (s, 1H), 7.66–7.59 (m, 2H), 4.83 (s, 2H), 3.07 (s, 3H); 13C NMR (CDCb, 100 MHz) δ 141.6, 136.4, 134.7, 132.2, 130.4, 127.6, 126.0, 48.3, 44.7; IR (film) Vmax 3061, 3013, 2924, 2852, 2178, 2066, 1479, 1431, 1318, 1212, 1144, 1086, 962, 866, 759, 708 cm−1; HRMS (EI+) m/z: [M]+ calcd for C9H9N02S2, 227.0075; found, 227.0089.</p><!><p>To a solution of 58 (125 mg, 0.64 mmol) in dry CH2CI2 (1.70 mL) was slowly added mCPBA (70%, 158 mg, 0.64 mmol). The reaction was stirred at rt for2 h and was diluted with CH2CI2 (35 mL). The organic layer was washed with saturated aqueous sodium bicarbonate (70 mL), saturated aqueous sodium chloride (70 mL), dried (NaSCb), and was concentrated to afford the mixture of sulfoxide and sulfone. Flash chromatography (silica gel, 1:1 hexanes:EtOAc 1:1 CH2CI2) afforded 64 as a light brown oil (35.9 mg, 24.9%): 1H NMR (CDCb, 400 MHz) δ 7.69 (m, 2H), 7.50 (m, 2H), 4.81 (s, 2H), 2.74 (s, 3H); 13C NMR (CDCb, 100 MHz) δ 140.7, 140.5, 134.4,128.3, 48.3, 44.6; IR (film) Vmax 3420, 3055, 2927, 2854, 2361,2337, 2188, 2096, 1636, 1456, 1436, 1410, 1317, 1265, 1091, 1091, 1018, 958, 896, 811, 739, 705, 668, 651 cm−1; HRMS (EI+) m/z: [M + H]+ calcd for C9HioNOS2, 212.0204; found, 212.0206.</p><!><p>The purification of 64 also provided 70 as a green oil (92 mg, 68.6 %): 1H NMR (CDCb, 400 MHz) δ 8.00 (d, J = 8.2 Hz, 2H), 7.55 (d, J = 8.2 Hz, 2H), 4.86 (s, 2H), 3.08 (s, 3H); 13C NMR (CDCb, 100 MHz): δ 140.8, 140.6, 134.6, 128.4 (2C), 127.9 (2C); IR (film) Vmax 346, 3053, 2985, 2360, 2339, 2182, 2098, 1496, 1407, 1338, 1265, 1087, 1051, 1015, 956, 896, 807, 738, 704 cm−1; HRMS (EI+) m/z. [M + H]+ calcd for C9H10NO2S2, 228.0153; found, 228.0157.</p><!><p>To a solution of 56 (64.4 mg, 0.31 mmol) in dry CH2CI2 (6.10 mL) was slowly added mCPBA (70%, 150 mg, 0.61 mmol). The reaction was stirred at rt for 7 h and was diluted with CH2CI2 (30 mL). The organic layer was washed with saturated aqueous sodium bicarbonate (70 mL), saturated aqueous sodium chloride (70 mL), dried (NaSOi), and was concentrated. Flash chromatography (silica gel, 3:1 hexanes:EtOAc) to afford 68 as a nearcolorless solid (30 mg, 43 %): m.p. 95.5–95.7 °C; 1H NMR (CDCI3, 400 MHz): δ 8.08 (dd, J = 7.8, 1.4 Hz, 1H), 7.74 (td, J = 7.8, 1.4 Hz, 2H), 7.67 (m, 1H), 7.60 (td, J = 7.8, 1.4 Hz, 1H), 5.30 (s, 2H), 3.14 (s, 3H); 13C NMR (CDCI3, 100 MHz): δ 137.9, 134.8, 134.7, 133.8, 130.4, 130.3, 129.7, 46.5, 45.2; IR (film) Vmax 2927, 2855, 2254, 2090, 1599, 1468, 1380, 1346, 1316, 1156, 1096, 907, 733, 651, 449 cm−1; HRMS (EI+) m/z [M]+ calcd for C9H9NO2S2, 227.0075; found, 227.0090.</p><!><p>Human MCF-7 breast cancer cells were maintained in a 1:1 mixture of Advanced DMEM/F12 (Gibco) supplemented with L-glutamine (2 mM), streptomycin (500 pg/mL), penicillin (100 units/mL), and 10% FBS. Cells were grown to confluence in a humidified atmosphere (37 °C, 5% CO2), seeded (2000/well, 100 pL) in 96-well clear, flat-bottomed plates, and allowed to attach overnight. For each trial, nine, two-fold serial dilutions of ITC in DMSO (1 pLof 100x ssk, final concentration range = 200 to 0.78 pm) were added in triplicate, and cells were returned to the incubator for 24 h or 72 h. At the specified time, the number of viable cells was determined using an MTS/PMS cell proliferation kit (Promega #PR-G5430) following the manufacturer's instructions. The percent viability for each well was determined relative to cells incubated with vehicle (1% DMSO). Where applicable, data was fit with the nonlinear function describing log(inhibitor) vs. response (variable slope, 4 parameters) using GraphPad Prism 6.0, allowing determination of GI50 values.</p><!><p>MCF-7-ARE cells stably transfected with the pGL3-promoter vector (Promega) containing eight copies of the antixodiant response element (ARE, 5-GTGACAAAGCA-3') were utilized as previously described.1301 MCF-7-ARE cells were maintained in high glucose (25 mM) DMEM supplemented with 10% FBS, streptomycin (50 pg/mL), penicillin (50 units/mL), and G418 (400 pg/mL). Cells were grown to confluence in a humidified atmosphere (37 °C, 5% CO2), seeded (2000/well, 100 pL) in 96-well clear, flat-bottomed plates, and allowed to attach overnight. For each trial, six, two-fold serial dilutions of ITC in DMSO (1 pL of 100x stock, final concentration range = 12.5 to 0.39 pm) were added in triplicate, and cells were returned to the incubator for 24 h or 72 h. Cells were lysed in passive lysis buffer (Promega) containing protease inhibitors (1:1000, Sigma-Aldrich). Lysate was used as substrate in the Bright-Glo Luciferase Assay System (Promega, #E2610), following the manufacturer's instructions. RLU data was normalized for protein content, determined by the Pierce BCA method (Thermo Fisher Scientific, #PI23221), and reported relative to cells incubated with vehicle (1% DMSO).</p><!><p>MCF-7-ARE cells were grown to confluence in a humidified atmosphere (37 °C, 5% CO2), seeded (40,000/dish, 10 mL) in 10 cm Corning polystyrene dishes, and allowed to attach overnight. For each trial, six, two-fold serial dilutions of ITC in DMSO (100 pL of 100x stock, final concentration range = 12.5 to 0.39 pm) were added in triplicate, and cells were returned to the incubator for 48 h. Cells were trypsinized, washed with PBS, and lysed prior to RNA extraction using the Maxwell 16 Total RNA Purification Kit (Promega), following the manufacturer's instructions. cDNA was synthesized from samples with an RNA integrity number of 8.0 or greater. Primers and probes for qPCR were designed to only amplify genomic DNA with the following obtained from Thermo Fisher Scientific: NADPH dehydrogenase, quinone 1 (NQOI; Hs01045994_m1), glutathione S-transferase a1 (GSTal; Hs00275575_m1 ), and heme oxygenase 1 (HMOX1 ; Hs01110250_m1). qPCR primers and probes were designed using Beacon Designer 7.91 (Premier Biosoftware) forthioredoxin reductase 1 (TXNRD1, nm003330.2; forward 5'-GCTTCAGCATGTCATGTG-3', reverse 5'- CT CT GTTT CACAAACACAAC-3', probe [6~FAM]CCAATTCCGAGAGCGTTCCTTC[BHQ1a~6FAM]) and glyceradehyde-3-phosphate dehydrogenase (GAPDH, nm002046.4; forward 5'-CATCCATGACAACTTTGGTA-3', reverse 5'- CCAT CCACAGT CTT CTGG-3', probe [6~FAM]ACCACAGTCCATGCCATCACT[BHQ1a~6FAM]). Primers (900 nM) and probes (250 nM) were diluted in 2X Absolute Blue master mix (Thermo Fisher Scientific) and assayed as previously reported [47]. Standard MIQE guidelines were followed, including: internal primer validation through mass normalization, assessment of genomic DNA contamination, and assay efficiency.[48]</p>

PubMed Author Manuscript

NHCs and Visible Light Co-catalyzed 1,4-Sulfonylacylation of 1,3-Enynes for Tetrasubstituted Allenyl Ketones

The modulation of selectivity of highly reactive carbon radical crosscoupling for the construction of C-C bonds represents a challenging task in organic chemistry. N-Heterocyclic carbenes (NHCs) catalyzed radical transformations opened a new avenue for acyl radical cross-coupling chemistry. With this method, highly selective cross-coupling of acyl radical with alkyl radical for efficient construction of C-C bonds were succussfully realized. However, the cross-coupling reaction of acyl radical with vinyl radicals represents an uncharted domain. We herein describe NHCs and photocatalysis co-catalyzed radical 1,4-sulfonylacylation of 1,3-enynes, providing structurally diversified valuable tetrasubstituted allenyl ketones. Mechanistic studies indicated that ketyl radicals are formed from aroyl fluorides via oxidative quenching process of excited photocatalysis, allenyl radicals are generated from chemo specific sulfonyl radical addition to the 1,3-enynes, finally, unprecedented key allenyl and ketyl radical cross-coupling provides tetrasubstituted allenyl ketones.

nhcs_and_visible_light_co-catalyzed_1,4-sulfonylacylation_of_1,3-enynes_for_tetrasubstituted_allenyl

<!>Results and discussion<!>Mechanism investigations.<!>Methods

<p>Radical cross-coupling between two carbon radicals emerged as a powerful platform for constructing C-C bonds and received increasing attention. 1 Since the radical-radical coupling reactions proceeded via a diffusion-controlled manner, selectivity modulation is the critical challenge. 1b Via radial addition to unsaturated bond to form C-C bond, acyl radicals have been utilized in preparing diverse carbonyl compounds. 2 However, radical-coupling reaction between acyl and other carbon-centered radicals is rare. N-Heterocyclic carbenes catalysis (NHCs) has emerged as an attractive strategy in synthetic chemistry to access value-added organics via the formation of key Breslow intermediate (BI). 3 Recently, the single-electron-transfer (SET) of BI was found to provide ketyl-type radical species, which opens a new avenue for acyl radical chemistry. [4][5][6][7][8][9][10][11][12] As a result, NHCs catalyzed radical-couplings have attracted great attention after the pioneer work of Ohmiya in 2019. 7a Alkyl radical sources such as redox-active esters, 7 Katritzky pyridinium salts, 8 Hantzsch ester, 9 benzylic C-H bonds, 6e alkylborates, 10g olefins 6c,10 as well as cyclopropanes 6f could be used to perform cross-coupling reaction with acyl radicals to form C-C bond (Fig. 1a). However, to the best of our knowledge, radical-coupling reaction between vinyl radical and acyl radical has never been reported.</p><p>On the other hand, radical 1,4-difunctionalization [13][14] of 1,3-enynes provides an elegant and versatile strategy for tetrasubstituted allenes from easily available feedstocks. In this regard, in situ generated allene radicals undergo cyanation, 14a-d arylation, 14e-h halogenation, 14i alkynylation, 14j trifluoromethylation, 14k or intramolecular cyclization 14l to afford functionalized allenes. Radical acylation of 1,3enynes may provide straightforward access to value-added allenyl ketone units, which are crucial core in important nature products 15 and synthetic intermediates. 16 Recently, Studer et al. developed acylative difunctionalization of olefins 6c /cyclopropanes 6f and formal alkenyl 6d /benzylic 6e C−H acylation by employing aroyl fluorides as ketyl-type radical precursors. Inspired by those elegant approaches, we speculated that NHCs and visible light co-catalyzed system 6c-6f,9,11-12 enable generation of allenyl radicals and NHCs stabilized ketyl radicals under extremely mild conditions, which may offer an opportunity for radical acylation of 1,3-enynes. As our continuous interests in radical chemistry, 17 we now describe the development of NHCs and photocatalysis cocatalyzed three-component radical 1,4-sulfonylacylation of 1,3-enynes, providing direct access to structurally diversified tetrasubstituted allenyl ketones (Fig. 1b).</p><!><p>Reaction conditions development. We commenced our investigation by employing 1,3-enyne (1a), benzoyl fluoride (2a), TolSO2Na (3a) as the prototype substrates and PC-1 (1.5 mol %), NHC-1 (15 mol %) as catalysts. Pleasingly, in dichloromethane (DCM) under irradiation with Blue LED at room temperature for 4 h, the expected allenyl ketones 4 was obtained in 10% yield combination with competitive by-product 5 (Fig. 2, entry 1). Ir-based photocatalysis PC-2 and PC-3 improved reactivity and selectivity (entries 2 and 3), while PC-4 and PC-5 were inefficient for this reaction (entries 4 and 5). The employment of other solvents such as CH3CN, PhCF3, or THF provided 4 in relatively lower yields (entries 6-8). The structure of NHCs was crucial for chemo-selectivity control (entries 9-13). The NHC-2 and NHC-3 were unsatisfactory (entries 9 and 10). The N-2,6-diethyl phenyl substituted catalyst NHC-4 afforded 4 with a slightly diminished yield compared to NHC-1 (entry 11). For NHC-5 or NHC-6, decreased yield was observed (entries 12-13). To our delight, yield could be further improved upon running the reaction at lower concentration (entries 14−15), and affording 4 in 80% isolated yield with negligible 5 in 4 mL DCM (entry 15). The desired 1,4-sulfonylacylation product was isolated in 75% yield when the reaction was run at 0.2 mmol scale (entry 16), and these conditions were thus defined as the standard reaction conditions for subsequent investigations.</p><p>) (1:1 dr.) yield. Furthermore, the fluorides derived from natural products such as telmisartan and mefenamic acid were successfully converted into 54 and 55 in 85% and 61% yields, respectively. Synthetic applications. Large-scale synthesis and derivatization reactions were performed to showcase synthetic applications (Fig. 4a). Scale-up synthesis of 17 has been achieved at a 2.0 mmol scale, and a comparable yield was obtained (Fig. 4a1).</p><p>When employing PhLi as a base, the tetrasubstituted allenyl ketones 4 could isomerize to diene product 56 in 78% yield. 4 could undergo reduction of ketone unit with NaBH4.</p><p>The allenyl ketone 4 could easily be transformed into conjugated viny selenyl ether 58 in 50% yield with excellent Z/E selectivity. When treated with concentrated H2SO4, Nazarov cyclization product 59 was isolated in 86% yield.</p><!><p>A series of control experiments were performed to unravel the reaction mechanism (Fig. 4b). Light, NHCs, and photoredox catalysis were indispensable for this 1,4-sulfonylacylation reaction (Fig. 4b1). When the radical scavenger 2,2,6,6-tetramethylpiperidine 1-oxyl (TEMPO) was added, the reaction was suppressed, and TEMPO-trapping product 60 was separated in 55% yield (Fig. 4b2), thus suggesting the formation of ketyl radicals. Furthermore, a trace amount of 4,4'dimethyl-1,1'-biphenyl (62) was isolated under standard conditions, indicating the involvement of a sulfonyl radical. The intermediacy of acyl azoliums has been confirmed by coupling of acyl azolium ion 61 with 1,3-enynes 1a and sodium benzenesulfinate 3a in the absence of NHC (Fig. 4b3). The radical chain process could rule out based on light/dark experiments (Fig. S4, see Supplementary Information). Then</p><p>Stern-Volmer quenching studies were conducted to clarify the plausible photoredox mechanism (Fig. 4c). 1,3-Enynes 1a and sodium benzenesulfinate 3a do not show a significant luminescence quenching effect to the excited state of the Ir*(III). In contrast, Ir*-complex was effectively quenched by acyl azolium ion 61, pointing to the oxidative quenching process. In summary, we have realized an efficient 1,4-sulfonylacylation of 1,3-enynes by merging photocatalysis with NHCs. This transformation provided a facile and direct entry for tetrasubstituted allenyl ketones under mild conditions with broad functional group tolerance and excellent chemo-and regioselectivity. Mechanistic studies indicated that the key step of the transformation is unprecedented allenyl and ketyl radical cross-coupling, proving a new avenue for NHCs catalyzed radical chemistry.</p><p>Ketyl radical was formed from aroyl fluorides via the oxidative quenching process of excited photocatalysis. Further extension of this cross-coupling system to other destabilized transient radicals is ongoing in our laboratory.</p><!><p>General procedure for the synthesis of tetrasubstituted allenyl ketones. Into a nitrogenfilled glove box, a vial (15.0 mL) equipped with a magnetic stir bar was charged with NHC-1 (12.6 mg, 0.03 mmol), Cs2CO3 (130.3 mg, 0.4 mmol), PC-3 (2.7 mg, 0.003 mmol), sulfinate (71.3 mg, 0.4 mmol) and DCM (8.0 mL). Then 1,3-enynes (0.2 mmol) and acyl fluorides (0.4 mmol) were added. The vial was removed from the glove box, and then the reaction mixture was irradiated with Blue LED at room temperature for 4 hours. After the reaction finished that monitored by TLC, the reaction mixture was quenched by water. The mixture was extracted with EtOAc (3 x 5.0 mL). The combined organic phases were dried over anhydrous Na2SO4, and the solvent was evaporated under vacuum. The residue was purified by flash column chromatography (petroleum ether/ethyl acetate = 10 : 1) to get the desired product.</p>

ChemRxiv

Anionic Polymers Promote Mitochondrial Targeting of Delocalized Lipophilic Cations

Mitochondria are therapeutic targets in many diseases including cancer, metabolic disorders, and neurodegenerative diseases. Therefore, strategies to deliver therapeutics of interest to mitochondria are important for therapeutic development. As delocalized lipophilic cations (DLCs) preferentially accumulate into mitochondria, DLC-conjugation has been utilized facilitate therapeutic delivery systems with mitochondrial targeting capability. Here we report that upon DLC-conjugation, anionic polymers exhibited significantly improved mitochondrial targeting when compared to cationic polymers and charge-neutral polymers. Considering that cell membrane generally bears net negative charge, the observed phenomenon is unexpected. Notably, the DLC-conjugated anionic polymers circumvented the endosomal entrapment. The rapid mitochondrial accumulation of DLC-conjugated anionic polymers is likely a membrane-potential driven process, along with the involvement of the mitochondrial pyruvate carrier. Moreover, the structural variations on the side chain of DLC-conjugated anionic polymers did not compromise the overall mitochondrial targeting capability, widely extending the applicability of anionic macromolecules in therapeutic delivery systems.

anionic_polymers_promote_mitochondrial_targeting_of_delocalized_lipophilic_cations

Introduction<!>Cy3-conjugation on anionic polymers enhanced their cellular uptake<!>Cy3-conjugated anionic polymers localized on the mitochondria<!>Cy3-conjugation lead to the mitochondrial targeting of anionic polymers<!>The uptake of Cy3-conjugated anionic polymer is membrane potential-dependent<!>Side chain variations on Cy3-conjugated polymers did not affect the mitochondrial targeting<!>Cy3-conjugation on anionic polymers enhanced the small-molecule drug delivery efficiency of anionic polymers<!>Conclusions<!>General methods.<!>Synthesis and characterization of polymers.<!>Cellular uptake of polymers with different surface charge or different dye conjugation.<!>Subcellular localization of Cy3-tagged anionic polymers.<!>Effect of pharmacological inhibitors.<!>Cellular uptake of anionic polymers with varied side chains or different charge densities.<!>Delivery of small-molecule drugs using Cy3-conjugated anionic polymers.

<p>The mitochondrion is the central organelle for generating adenosine triphosphate (ATP), supplying requisite energy for cellular metabolism.1, 2 Mitochondrial dysfunction leads to cellular disorders and eventually causes severe diseases such as neurodegeneration.3 Beyond the role of cellular powerhouse, mitochondria are also involved in the regulation of the intrinsic apoptotic pathway.4 Briefly, mitochondria control the translocation of pro-apoptotic proteins to regulate the activation of apoptotic effectors (e.g. caspases).5 Defects in the apoptotic pathways have been associated with the resistance of tumor cells towards chemotherapy,6 further implying the fundamental importance of mitochondria as therapeutic targets. Besides, mitochondria are central mediators in regulated necrosis.7 Thus, strategies for precisely delivering drugs to mitochondria are necessary for the development of mitochondria-relevant therapeutics.</p><p>Chemical approaches to targeting mitochondria are mainly achieved through delocalized lipophilic cations (DLCs) or signal peptides.8 DLCs are positively charged small molecules, possessing delocalized electronic structures via resonance stabilization of the lipophilic molecule.9 The positive charge within DLCs is spread over a large hydrophobic molecular area, therefore requiring a lower enthalpy input to desolvate these charged species. As a result, moving DLCs through lipid bilayers requires a far lower activation energy than hydrophilic cations such as Na+.10 Driven by the cell and mitochondrial membrane potentials, DLCs readily transfer into the cytosol and efficiently accumulate to the mitochondria.8 Representative DLCs include triphenylphosphonium-based compounds,11 rhodamine-12 and cyanine (Cy)-based13 derivatives.</p><p>DLC-conjugation has been utilized to initiate the mitochondrial targeting capability of therapeutic delivery systems.9, 14-17 While the DLC-conjugation effectively targets small molecules to mitochondria, the DLC-conjugation with large and polar molecules has been far less effective in their mitochondrial targeting.18, 19 Herein, we report that conjugating DLCs onto anionic polymers resulted in a rapid mitochondrial targeting effect, largely exceeding the cationic and charge-neutral polymers with DLC-conjugation. Specifically, conjugation with anionic polymers did not compromise the mitochondrial targeting capability of delocalized lipophilic cations. To explain such unexpected phenomenon, we conducted structural variations on the DLC-conjugated anionic polymers, extending the fundamental understanding on the design principles of mitochondrial targeting materials. The study presents an effective strategy for enhancing the cellular uptake and mitochondrial targeting capability of therapeutic delivery systems.</p><!><p>The molecular design of polymers allows an azide group displayed as the end group of each polymer chain, providing the reaction site for further functionalization with click reaction.20 First, a library of methacrylate polymers with different surface charges were synthesized using reversible addition-fragmentation chain-transfer (RAFT) polymerization (Figure 1a). Next, dibenzocyclooctyne-functionalized fluorescent molecules were covalently conjugated to the azide-tagged polymers to allow their intracellular tracking (Figure S1). During the evaluation of cellular uptake, we fortuitously discovered that the Cyanine 3 (Cy3)-conjugated anionic polymers (NEG or SO3) exhibited a significantly higher uptake efficiency than the rest of the structural analog, including a cationic polymer (POS) and two charge-neutral polymers (PEG and MPC) (Figure S2). To further validate this phenomenon, we extended the screening to three more cell types, including human umbilical vein endothelial cells (HUVECs), mesenchymal stem cells immortalized with hTERT (hTERT-MSC), and mouse myoblast cells (C2C12). The results from these three cell types are all in agreement with the results from human cervical cancer cells (HeLa) (Figure 1b, S3), confirming that the Cy3-conjugated anionic polymers exhibit an enhanced cellular uptake compared to other polymers with different surface charge.</p><!><p>Next, we explored the intracellular localization of the Cy3-labeled anionic polymer (NEG). From the colocalization assessment with commercially available organelle-staining reagents, we found that NEG highly overlapped with the mitochondrial stain, rather than the lysosome and endoplasmic reticulum (ER) stains (Figure 2, S4). The colocalization analysis of ER stain and NEG exhibited the colocalization coefficient at ~0.59 in HeLa cells, and the coefficient decreased to ~0.35 in SK-MEL-2 cells (Figure S5), further proving that ER is less likely to be the localization of NEG. Since the ER-mitochondria contact sites closely correlate with the cellular metabolism,21, 22 the difference of ER stain-NEG colocalization coefficient can be potentially attributed to varied metabolic activities among different cell lines.23</p><p>To further validate the colocalization results with small molecule-based staining, we employed HeLa cells with green fluorescence proteins transiently expressed within the mitochondria (MitoGFP). After incubating Cy3-labeled anionic polymers with the MitoGFP-expressing HeLa cells, we observed similar colocalization results throughout the tubular morphology as with the small molecule-based mitochondrial stain (MitoTracker Green), confirming that Cy3-labeled anionic polymers are preferentially localized to mitochondria. Since the expression of MitoGFP is transient, it is reasonable to observe a lower colocalization coefficient (~0.57) than the MitoTracker Green group (~0.85). Note that NEG bypassed endosomal entrapment during its cellular uptake process. In contrast, Cy3-labeled cationic polymers (POS) were found to be mainly accumulated in the endosomes (Figure S6). Previous results have also demonstrated that DLC-conjugated cationic materials accumulate in endosomes and do not efficiently reach mitochondria.18</p><!><p>The enhanced cellular uptake and mitochondrial targeting capability of Cy3-labeled anionic polymer might be attributed to the conjugation of Cy3, as Cy3 itself is a delocalized lipophilic cation (DLC).24, 25 The azide-tagged end group allows a facile variation on the dye to be conjugated on the polymer, i.e. any azide-reactive dye derivatives can be tagged on the polymer. To test our hypothesis, we functionalized the azide-tagged anionic polymers (NEG) with five different dye molecules (Figure S7). The cellular uptake results from anionic polymers conjugated with different dyes altogether revealed that the conjugated dye molecule indeed determines the mitochondrial targeting capability of anionic polymers (Figure 3, S8). Polymers conjugated with DLC exhibited precise localization on mitochondria, whereas conjugation with non-DLC dye mostly resulted in the endosomal entrapment. Particularly regarding the sulfonated Cy3 (Sulfo-Cy3) analog, where the original DLC structure of Cy3 is altered to an overall charge-neutral dye with the sulfonate substitution, eventually resulting in the loss of overall mitochondrial targeting of anionic polymers. Note that the anionic polymer itself is based on sulfonate moieties, but only when installing this functionality in Cy3 itself that mutes the mitochondrial targeting capability. Also, based on the structural design of the polymer, introducing mitochondrial targeting capability to anionic polymers only requires one targeting moiety per polymer chain.</p><!><p>Efficient mitochondrial targeting of these Cy3-conjugated anionic polymers was observed within an hour through time-lapse microscopic imaging (Video S1). Considering the anionic feature of NEG, their cellular uptake and mitochondrial targeting may be contributed by the cell surface receptors that recognize anionic macromolecules.26 After treating HeLa cells with dextran sulfate or polyinosinic acid (Figure 4a,b), two inhibitors for scavenger receptors,27 the cellular uptake of NEG was not affected. Moreover, inhibition of mitochondrial voltage-dependent anion channels using erastin28 did not reduce the uptake of NEG (Figure 4c). The increase of NEG uptake after erastin-treatment is possibly due to the alteration of mitochondrial membrane permeability.28 Nonetheless, the role of DLC for the uptake of NEG is worthy of investigation. Since the uptake of DLC is known to be driven by mitochondrial membrane potential (ΔΨm), we treated cells with either FCCP or oligomycin to modulate the ΔΨm towards two different directions.29 As a result, FCCP-treatment decreased the uptake of NEG while oligomycin-treatment increased the uptake of NEG, confirming that the uptake is driven by membrane potential (Figure 4d,e). Meanwhile, we also evaluated the cellular uptake of NEG after treating cells with UK 5099, a potent inhibitor for the plasma membrane monocarboxylate transporters and the mitochondrial pyruvate carrier.30, 31 The role of mitochondrial pyruvate carrier during the uptake of Cy3-labeled anionic polymers was later evaluated by depleting sodium pyruvate within cell culture medium (Figure S13). The decrease of NEG uptake upon increased dosage of UK 5099-treatment indicates the participation of these transporters during the cellular uptake of NEG (Figure 4f). Based on our previous results, the cellular uptake of NEG also involves macropinocytosis.20</p><!><p>It is important to understand the effect of side chain structure on the mitochondrial targeting capability of anionic polymers. NEG is a random copolymer containing three side chain moieties, including sulfonate derivatives to ensure the anionic feature, poly(ethylene glycol) derivatives as the hydrophilic moieties, and pyridyl disulfide (PDS) derivatives as the hydrophobic moieties. We first varied the structure of the hydrophobic side chain and assessed if the mitochondrial targeting ability of these polymers are affected. It turns out that modulating either the disulfide bond (PDB) or the pyridine structure (Ph and C4S2) do not alter the mitochondrial targeting of Cy3-conjugated anionic polymers. Moreover, after replacing the PDS side chain with either hydrophobic hexyl (C6) or hydrophilic hydroxyl (OH) group, mitochondrial localization of polymers was still observed from 13 different cell types (Figure 5, S9).</p><p>The effect of charge density and negative charge type were also investigated regarding the Cy3-conjugated anionic polymers. Such Cy3-conjugated anionic polymers with varied density of sulfonate moieties again exhibited strong colocalization with mitochondrial staining, revealing that the mitochondrial targeting capability is independent of the negative charge density (Figure S10). We also synthesized an analog of anionic polymers with different negative charge type, containing either carboxylate (COOH) or phosphate (PO4) instead of sulfonate (SO3). After conjugating these polymers respectively with Cy3, similarly efficient mitochondrial targeting was observed from SO3 and PO4, while COOH did not present such efficiency (Figure 6, S12).</p><p>The results from pharmacological inhibitor screening indicates the role of plasma membrane monocarboxylate transporter and mitochondrial pyruvate carrier during the uptake of these Cy3-labeled anionic polymers (Figure 4f). Considering the pyruvate and carboxylate species (such as sodium pyruvate and linoleic acid) within the cell culture medium, the suppressed cellular uptake of COOH can be attributed to the competition with these small molecules (Figure S13). While we were preparing the current manuscript, a report in press demonstrated that conjugating carboxylate-decorated polymers with Cy5 targeted the polymers to the mitochondria of rat neural cells.32 In our systems, we found that COOH indeed can enhance mitochondrial targeting in two out of the nine cell lines tested (Myotube and SK-MEL-2; see Figure 6). We found however that sulfonates and phosphates have more consistent mitochondrial targeting capability. Interestingly, as nucleic acids similarly contain phosphate groups as PO4, conjugating DLC with nucleic acids is expected to efficiently localize nucleic acids to mitochondria. Such effect has been successfully demonstrated in a previous report,33 supporting our hypothesis.</p><!><p>As DLC-conjugated anionic polymers possess efficient mitochondrial targeting capability, DLC-conjugated anionic polymers are expected to have elevated delivery efficiency as drug carriers. We employed PDB as a drug carrier for small-molecule drugs (doxorubicin, lonidamine, and gossypol) to test our hypothesis. In detail, small-molecule drug of interest was encapsulated into the anionic polymer-based micelles respectively before and after Cy3-conjugation (PDB w/o Cy3 and PDB) (Figure S14). Resazurin-based assay was applied to evaluate the drug delivery efficiency.34 Resazurin is a redox-sensitive fluorogenic substrate. Once resazurin is reduced to resorufin (highly fluorescent) in mitochondria, relative level of the reduction reaction product reflects the mitochondrial metabolic activity.</p><p>PDB loaded with doxorubicin (IC50 = 5.0 μM), is more effective than doxorubicin alone (IC50 = 6.0 μM) to inhibit the mitochondrial metabolic activity (Figure 7a). Other than the well-known mechanism of action via the intercalation of DNA,35 doxorubicin has been associated with mitochondrial dysfunctions.36 In contrast, the therapeutic effect of doxorubicin was suppressed after being encapsulated inside PDB w/o Cy3. This is reasonable since anionic polymers without DLC-conjugation tends to be entrapped in the endosomes (Figure 3b), limiting the drug efficacy. Other than doxorubicin, similar trend was observed from two small-molecule drugs with mitochondrial proteins as the drug target, where lonidamine inhibits mitochondrial hexokinase37 and gossypol inhibits Bcl-2 family proteins as apoptosis regulator (Figure 7b,c).38 Note that before loading, both PDB w/o Cy3 and PDB carriers at each dosage demonstrated negligible effect on the metabolic activity of mitochondria in HeLa cells (Figure 7d). These datasets demonstrated that the DLC-conjugation enhances the drug delivery efficiency of anionic polymers. Meanwhile, such platform is a widely applicable for targeting small-molecule drugs to mitochondria, circumventing specific chemical modifications on the drug of interest.</p><!><p>In summary, we describe the conjugation of delocalized lipophilic cations (DLCs) on anionic polymers to facilitate macromolecules with efficient mitochondrial targeting capability. Unlike the endosomal entrapment of DLC-conjugated cationic or charge-neutral polymers, DLC-conjugated anionic polymers rapidly and efficiently localize to the mitochondria of live cells. Apart from membrane potential as the driving force, the mitochondrial targeting process of DLC-conjugated anionic polymers is also attributed to the mitochondrial pyruvate carrier proteins. Moreover, upon DLC-conjugation, the anionic polymer library with varied amphiphilicity and charge density successfully target at the mitochondria, providing fundamental understandings on the design principles of mitochondrial targeting materials. The efficient mitochondrial targeting capability clearly requires the co-existence of DLC and anionic polymeric chain, as replacing either component largely impairs such capability. While we attribute the membrane potential and mitochondrial pyruvate carrier proteins as key players in this targeting process, further understanding of the features that underlie this phenomenon still needs investigation, including elucidating the localization of these polymers in specific mitochondrial sub-compartments. Meanwhile, the pH gradient within mitochondria could also play a role in the mitochondrial localization of these polymers.39 From applications perspective, the versatility in the design of anionic macromolecules opens up potential avenues in biological imaging and therapeutic delivery applications.</p><!><p>Reagents were purchased from commercial sources without further purification. 1H NMR, 13C NMR, and 31P NMR spectra were recorded from a Bruker AdvanceIII 400 (or 500) NMR spectrometer. When using tetrahydrofuran (THF) as the eluent, gel permeation chromatography (GPC) was performed on an Agilent 1260 LC. Molecular weights are versus polystyrene standards. When using trifluoroethanol (TFE) as the eluent, GPC was performed on an Agilent 1200 series HPLC system, using polymethylmethacrylate standards for molecular weight calculation. Dynamic light scattering and zeta potential measurement were measured on a Malvern Zetasizer Nano ZS. Confocal microscopy images were acquired from a Nikon fluorescence microscope equipped with a spectral detector unit. Confocal microscopy video was obtained from a Nikon fluorescence microscope equipped with a Yokogawa spinning disk unit. Flow cytometry experiments were conducted on a ThermoFisher Attune NxT flow cytometer. The infrared spectra were obtained on a Bruker Alpha FT-IR Spectrometer with a spectral range from 3500 cm−1 to 400 cm−1. Thermogravimetric analysis was conducted under N2 flow from room temperature to 600 °C using a TA Instrument Q50 thermogravimetric analyzer.</p><!><p>Polymers in this study were all synthesized using reversible addition–fragmentation chain-transfer polymerization. Azido-derivatives of chain transfer agent and radical initiators were synthesized based on a previous report.20 Detailed procedures for the synthesis and characterization results of monomers and polymers are summarized in Section 2.1., 2.2., and Section 3. of the Supplementary Information. The conjugation between azide-tagged polymers and DBCO-derivatives was carried out based on our previous report.20 Details are available in Section 2.3. of the Supplementary Information. The stock solution of each polymer was prepared in deionized water (details in Section 2.4. of the Supplementary Information).</p><!><p>General cell procedures are summarized in Section 2.5 of the Supplementary Information. Cells of interest were seeded into a glass bottom dish (Cellvis, #D35C4-20-0-N) prior to the experiment. Next, the cells were incubated with Cy3-labelled polymers (30 μg·mL−1 for PEG, POS, NEG, and MPC in full growth medium) for 18 hours. After washing with phosphate buffer saline, the cells were incubated with Hoechst 33342 in FluoroBrite DMEM at 37 °C to stain the nucleus. The intracellular distribution was measured by confocal microscopy with excitation wavelengths of 405 nm (Hoechst 33342) and 561 nm (Cy3-labelled polymers). Details on each cell type are available in Section 2.6. and 2.8. of the Supplementary Information.</p><!><p>A total of 50 k HeLa cells were seeded into a glass bottom dish (Cellvis, #D35C4-20-0-N) for 24 hours prior to the experiment. Subsequently, the cells were incubated with Cy3-labelled anionic polymers (NEG, 60 μg·mL−1 in DMEM) for 2 hours. Before staining cells with organelle trackers, cells were washed with phosphate buffer saline. For organelle staining, the cells were incubated with LysoTracker Green, ER-Tracker Green, and MitoTracker Green were respectively used to stain the cellular lysosomes, endoplasmic reticulum (ER), and mitochondria. The mitochondria colocalization was further evaluated in HeLa cells with green fluorescence proteins transiently expressed within the mitochondria. The intracellular distribution was measured by confocal microscopy. Experimental details along with the colocalization analysis are available in Section 2.7. of the Supplementary Information.</p><!><p>A total of 10 k HeLa cells were cultured in a 96-well plate for 24 hours prior to the experiment. Next, cells were cultured for 1 hour in DMEM containing dextran sulfate polyinosinic acid, erastin, FCCP, oligomycin, UK 5099, respectively. Subsequently, in the presence of inhibitors, the cells were incubated with Cy3-labelled anionic polymers (NEG, 2 μg·mL−1 in DMEM) for additional 2 hours. After washing the cells with cold PBS, the fluorescence intensity of Cy3 within the cells was measured using flow cytometry with an excitation wavelength of 561 nm. For the positive control, HeLa cells were cultured in DMEM for 1 hour and incubated with Cy3-labelled polymers for another 2 hours. For the blank group, HeLa cells were cultured in DMEM for 3 hours. After subtracting the fluorescence signal from the blank group, Cy3 fluorescence of the positive control group was normalized as 100%. Details are available in Section 2.9. of the Supplementary Information.</p><!><p>A total of 50 k A549 cells were seeded into a glass bottom dish (Cellvis, #D35C4-20-0-N) for 24 hours prior to the experiment. Next, the cells were incubated with Cy3-labelled polymers in respective growth medium for 2 hours. Details are available in Section 2.10. and 2.11. of the Supplementary Information. After washing with phosphate buffer saline, the cells were maintained in FluoroBrite DMEM for confocal microscopic measurement. The intracellular distribution of Cy3-labelled polymers was measured by confocal microscopy with an excitation wavelength of 488 nm. Multispectral images were collected in each run to record with emission wavelengths from 492 nm to 682 nm, obtaining 32 images in total (5-nm wavelength each). Images with emission wavelengths from 560 nm to 630 nm were combined as the final intracellular distribution of Cy3-labelled polymers.</p><!><p>Polymer was dissolved in trifluoroethanol, followed by the dropwise addition of deionized water while stirring. The stock solution of small molecule-drug of interest in DMSO was added into the solution while stirring. The mixture was continuously stirred at room temperature overnight. Next, the mixture was purified with deionized water and concentrated using Amicon centrifugal filters with 3 k MWCO. With the assumption of no sample loss during the purification, the volume of the concentrated polymer solution was adjusted to result in the polymeric micelle stock solution at a concentration of 10 mg·mL−1, containing 0.5 mg·mL−1 small-molecule drug of interest. Empty polymeric micelles as the control groups were prepared similarly without the drug loading step. Mitochondrial metabolic activity of cells after delivery was evaluated using alamarBlue assay. Details are available in Section 2.12. of the Supplementary Information.</p>

PubMed Author Manuscript

Chemi-Ionization Reactions and Basic Stereodynamical Effects in Collisions of Atom-Molecule Reagents

A new theoretical method, developed by our laboratory to describe the microscopic dynamics of gas-phase elementary chemi-ionization reactions, has been applied recently to study prototype atom–atom processes involving reactions between electronically excited metastable Ne*(3P2,0) and heavier noble gas atoms. Important aspects of electronic rearrangement selectivity have been emphasized that suggested the existence of two fundamental microscopic reaction mechanisms. The distinct mechanisms, which are controlled by intermolecular forces of chemical and noncovalent nature respectively, emerge under different conditions, and their balance depends on the collision energy regime investigated. The present paper provides the first step for the extension of the method to cases involving molecules of increasing complexity, whose chemi-ionization reactions are of relevance in several fields of basic and applied researches. The focus is here on the reactions of Ne* with simple inorganic molecules as Cl2 and NH3, and the application of the method discloses relevant features of the reaction microscopic evolution. In particular, this study shows that the balance of two fundamental reaction mechanisms depends not only on the collision energy and on the relative orientation of reagents but also on the orbital angular momentum of each collision complex. The additional insights so emphasized are of general relevance to assess in detail the stereodynamics of many other elementary processes.

chemi-ionization_reactions_and_basic_stereodynamical_effects_in_collisions_of_atom-molecule_reagents

Special Issue<!>Introduction<!><!>General Trends<!><!>General Trends<!><!>The Ne*-Cl2 Case<!><!>The Ne*-Cl2 Case<!>The NH3 Reaction Stereodynamics<!><!>The NH3 Reaction Stereodynamics<!><!>The NH3 Reaction Stereodynamics<!><!>The NH3 Reaction Stereodynamics<!><!>The NH3 Reaction Stereodynamics<!><!>The NH3 Reaction Stereodynamics<!>Conclusions<!><!>Conclusions<!>

<p>Published as part of The Journal of Physical Chemistry virtual special issue "Cheuk-Yiu Ng Festschrift".</p><!><p>The weakly bound adducts formed by colliding reagents play an important role in the kinetics of elementary processes, serving as precursor states opening the passage to the transition state (TS) of several chemical–physical phenomena occurring in gaseous and condensed phases and at the gas–solid, gas–liquid interphases.1−4 The valence electrons quantum confinement of reagents in such adducts and the selectivity of electronic rearrangements, promoting the stereodynamical evolution of the systems toward the final products, is still today a topic of general interest.</p><p>Recently, we developed and applied a new/original method5−7 to the detailed investigation of chemi-ionization reactions (CHEMI), promoted by collisions of atoms electronically excited in metastable states with other atoms, that are prototypical of barrierless processes leading to the formation of ionic products plus electrons.8−10 The study of them remains of great interest for fundamental and applied research, since it allows a definition of the role of the barrierless reactions in cold chemistry11−13 and an opportunity to model energy-transfer phenomena occurring in flames, plasmas, and electric discharges.14,15</p><p>If CHEMI involving Ng atoms are important, especially from the point of view of basic research, those involving molecules are of a more general interest, especially in highlighting the role of electronic transfer, that is, the redox nature of this type of process.6,7 Indeed, they control the balance of phenomena occurring in interstellar environments, in combustion and flames, where CHEMI are considered as the primary initial step,14,15 in molecular plasmas and nuclear fusion. They also govern interstellar chemistry and planetary ionospheres16−18 affecting the transmission of radio and satellite signals.18 These reactions are also implicated in soft-ionization mass spectrometry techniques,19,20 since the controlled internal degrees excitation of the molecular ionic products limits the number of fragmentation channels.</p><p>In the present study we attempt to take the first step toward the extension/generalization of our approach to atom-molecule CHEMI, where the intermolecular interactions driving the dynamics are usually stronger and more anisotropic with respect to atom–atom CHEMI and often include further components. The combination of these interaction features can strongly vary the relative role of two basic (direct and indirect) mechanisms, initially demonstrated for atom–atom CHEMI reactions.5−7</p><p>In the next sections, after a comparison of the experimental total ionization cross sections, with their different magnitudes and collision energy dependence, the focus is on some prototypical CHEMI of molecules. The special case of Cl2 emphasizes how the typical harpooning effect can affect the reaction precursor state. The detailed study of NH3 reaction stereodynamics shows how the two basic mechanisms, first revealed for atom–atom CHEMI, are modulated by molecular orientation and by the orbital angular momentum of the collision complex, controlling the centrifugal component of the total interaction potential.</p><!><p>The optical potential model, formally introduced to describe nuclear reactions dynamics and applied also to CHEMI,9,10 is defined as a combination of a real and an imaginary part. We have demonstrated5−7 that the two parts—that control, respectively, the collision dynamics and the "opacity" or probability of CHEMI—must be interdependent, since they arise from the same interaction components.</p><p>The different balance of such components originates two competitive microscopic reaction mechanisms. They have been identified, respectively, as a direct mechanism, dominant at short separation distances of reagents being driven by chemical forces, and an indirect mechanism, prevalent at large separation distances and originating from noncovalent forces, such as dispersion, induction-polarization contributions, and those promoting spin–orbit and centrifugal-Coriolis effects.5−7 In particular, the direct mechanism is triggered by effective charge (electron) transfer (CT) effects between reagents favored by the overlap of valence orbitals. The indirect mechanism describes ionization that occurs also by a concerted emission-absorption of a "virtual" photon exchanged by reagents within the confines of the weakly bound collision complex. Therefore, while the direct mechanism controls the evolution of prototype elementary oxidation reactions, the indirect mechanism triggers typical radiative (photo)-ionization processes.7</p><p>The reactivity depends on the collision energy (Ecoll), separation distance R, and relative alignment of valence orbitals, important factors that affect the structure and stability of the adducts formed by collision of reagents and then of the reaction TS.</p><p>Twelve reaction channels, ascribed to specific passages from a quantum state of reagents to that of products, have been characterized, where each one is affected by a different relative role of the two basic mechanisms mentioned above.7</p><!><p>According to the pioneering work of Beijerinck and co-workers,21 molecular systems giving CHEMI, all experimentally investigated in detail in the gas phase under single collision conditions with the molecular beam technique, can be distinguished in two groups: CHEMI systems showing a pronounced increase in the total ionization cross section as Ecoll increases and CHEMI systems showing, contrariwise, cross sections with a decreasing trend. The energy dependence of the total ionization cross section has been measured in our laboratory in an internally consistent way for several systems involving Ne*;22−25 therefore, a direct and quantitative comparison of obtained results is straightforward. Some prototype examples are reported in Figure 1a,b.</p><!><p>Total ionization cross sections in some Ne*-molecule systems, as a function of collision energy. The curves are interpolating 3rd degree polynomials of experimental data.22,25 (a) The case of some inorganic molecules. (b) The case of the simplest saturated and unsaturated hydrocarbons.</p><!><p>The differing behaviors, exhibited by the various partners of the Ne* reagent, must selectively depend on their fundamental chemical physical properties, as depicted in Figure 2 for three cases of inorganic molecules. The present focus is on Ne*-Cl2, where the formation by harpooning of an effective ion pair is expected to increase the binding energy in the collision complex, at least 2 orders of magnitude with respect to that in Ne*-N2, favoring a closer approach of reagents. However, in addition to specific features of Ne*-Cl2, here we also analyze in detail the Ne*-NH3 system for which the intermolecular interaction is strongly anisotropic, exhibits an intermediate strength between Ne*-Cl2 and Ne*-N2, and has been recently provided in analytical form (see below).</p><!><p>Fundamental features of Cl2, NH3, and N2 molecules associated with different electronic charge distribution around their molecular axis. The chlorine molecule exhibits two σ-holes collinear with the bond axis. This justifies the large and positive quadrupole moment of Cl2. The ammonia molecule exhibits a large dipole moment. The nitrogen molecule exhibits a large and negative quadrupole moment. The positive charge density increase is approximately indicated by the increased extent of the red color; the corresponding change in negative charge density is likewise indicated in blue.</p><!><p>To cast light on the critical role of the interaction components that are expected to selectively modulate the relative weight of the two basic microscopic mechanisms indicated above as a function of collision energy, we make a preliminary attempt to rationalize the phenomenology observed for the Ne*-Cl2 system shown in Figure 1a. In particular, in the thermal range of Ecoll, Ne*-Cl2 is one of the systems showing the highest cross-section value. For the Cl2 reagent this behavior must relate to the electronic features of its structure. The potential curves for the ground and excited states of the chlorine molecule as well as of its positive and negative ions have been calculated by Peyerimhoff and Buenker using the MRD-CI method,26 to which the interested reader can refer. Specifically, this molecule exhibits a high permanent electric quadrupole moment (+3.8 au),27 a high electron affinity (2.44 eV),28 and is characterized by a σ-hole,29 with a positive electrostatic potential confined along the outer parts of the Cl–Cl bond (see Figure 2). An extended discussion on the σ-hole topic, in terms of electron density plots of Cl2 molecule in its ground electronic state, is presented in ref (29) (see also references therein).</p><p>These unique features of Cl2 are indeed responsible of the formation of the intermolecular halogen bond even with lighter Ng atoms in their ground electronic state.29 Therefore, during the approach to Cl2, the "floppy" outer electronic cloud of Ne* tends to be polarized by the long-range intermolecular interaction field. This electron transfer is primarily triggered in the collinear approach by the σ-hole presence. The newly formed Cl2– anion tends further to align with its axis along the interatomic Ne···Cl···Cl separation R, and the Coulomb attraction in the nascent ion-pair Ne+-Cl2– favors the trapping of reagents, shown schematically in Figure 3. Indeed, in Cl2– the added electron, populating the outer 3σu* antibonding orbital, confined in the external part of the Cl–Cl bond, completely fills the σ-hole and strongly reduces the molecular bond strength making Cl2– a highly unstable species, especially in the presence of Ne+.</p><!><p>A schematic diagram representing the microscopic dynamics for Ne*-Cl2 CHEMI reaction. Redox 1 (upper panel) At a large distance (∼6 Å) and with collinear Cl2 the Rydberg electron of Ne* can go to fill the σ-hole of Cl2 with subsequent Ne+-Cl2– ion pair formation. The extra electron in Cl2– is located in the antibonding 3σu* orbital. Redox 2 (middle panel) At shorter distances the ionization can take place involving a pair of electrons from the 3σ molecular orbitals. In this case, the Cl2+ ion is then formed in the dissociative B2Σg+. (lower panel) In case of a perpendicular approach, the Ne* atom is polarized, and the ionization can take place essentially through a radiative (physical-photoionization) mechanism at a large distance, or by an exchange (chemical-redox) mechanism at a short distance. The positive charge density increase is approximately indicated by the increased extent of the red color; the corresponding change in negative charge density is likewise indicated in blue.</p><!><p>The increased attraction arising from ion pair formation and the instability of Cl2– stimulates the formation of a highly excited NeCl* adduct that autoionizes leading to Ne + Cl+ + e– products. At a short R, an additional electronic rearrangement process become possible, illustrated in the middle panel in Figure 3, triggered by the overlap between the half-filled orbital of Ne+ and the populated 3σ molecular orbitals of Cl2. Specifically, this overlap promotes a single electron transfer from the outer 3σu* of Cl2– to the half-filled orbital of the Ne+ core, which is accompanied by an energy release sufficient to eject one of two electrons populating the 3σg bonding molecular orbital of Cl2. As a consequence, the product Cl2+ shows a propensity to be formed in the dissociative B2Σg+ state with a bond order of 0.5. Since electrons populating both 3σu* and 3σg molecular orbitals are mostly confined in the σ hole region, this peculiar feature of the chlorine molecule can be assimilated to a reaction catalyst. However, the formation of the fragment Cl+ requires a synchronization between the time required by an interacting complex to give an electronic rearrangement and the typical collision time. This synchronization is partially and totally relaxed with increasing Ecoll. Therefore, the Cl2+ production is predicted to increase with Ecoll, consistent with the experimental findings by Kischlat and Morgner30 and by our laboratory.23</p><p>It is intriguing to note that, under such conditions, the chemical (direct) mechanism is dominant and that it occurs through two basic steps: first, Cl2 undergoes a reduction to Cl2– by CT in which the resulting neon behaves as an alkali atom (i.e., Na) as reducing agent (Redox 1 in Figure 3); in the second step, the Coulomb attraction promotes the trapping of the Ne+-Cl2– ion pair at closer distances, where a concerted CT involving both internal 3σg and external 3σu* populated molecular orbitals of Cl2– (with the outer electron filling the p-orbital of the Ne and the other innermost electron being ejected) (Redox 2 in Figure 3), accompanied by molecular dissociation, determines the oxidation to the final state of Cl+. In this second case the Ne+ behaves like a halogen atom (i.e., F) as an oxidizing agent.</p><p>However, with increasing collision energy, the effectiveness of such a global mechanism, triggered by the Cl2 with the molecular axis aligned along R, decreases, since the collision time shortens, the passage through the crossing between neutral and ionic states assumes a less adiabatic character,31 and alignment effects are less probable. Under such conditions, collisions become statistically possible for all relative orientations of both partners, including the Cl2 molecule perpendicular approach to the Ne* atom and the global reactivity decreases. Here, in the perpendicular configuration of the formed adduct, rather unstable because of the absence of strong attractive components, both indirect (including possible radiative effects9,32−34) and direct (chemical or exchange) mechanisms become competitive, and above all an electron removal from the outer 3πu* molecular orbital becomes effective, leading to the single-step formation of Cl2+ in its ground electronic state X2Πg. This new channel increases the formation probability of Cl2+ with respect to Cl+.</p><!><p>A system useful for the generalization of our approach is Ne*-NH3, for which a multidimensional potential energy surface (PES) given in analytical form35 assists in the formulation of the real part of the optical potential. A previous study,25 adopting a radial dependent imaginary part, whose average strength is modulated only by the NH3 orientation defined by the polar coordinates R, θ, ϕ given in the lower panel of Figure 4, suggested that CHEMI occurs exclusively on the N-side of the molecular frame. Specifically, while the Ne* approach within an angular cone confined around the C3v ammonia axis promotes the formation of NH3+ in the X ground electronic state, the broadside approach in the vicinity of the perpendicular configuration leads to the formation of the electronically excited NH3+(A) ion that dissociates to NH2+ + H.25 The approach toward the hydrogen end of the molecule, along the C3v ammonia axis, is accompanied by the strong polarization of the "floppy" cloud of 3s1 valence electron of Ne* that enhances the propensity to give an intermolecular hydrogen bond with a consequent reduction of the redox reaction effectiveness.35</p><!><p>(lower panel) The polar coordinate system used to define the orientation of NH3 with respect to Ne*. (upper and intermediate panels) Two relevant configurations giving redox-reactive and nonreactive events. The metastable Ne* atom is differently polarized, accordingly to the NH3 dipole orientation. The positive charge density increase is approximately indicated by the increased extent of the red color; the corresponding change in negative charge density is likewise indicated in blue.</p><!><p>The optical potential formulation, including also an effective angular-dependent imaginary Γ component, permitted us to estimate the acceptance of two angular cones where the reactions mainly occur. Details on the acceptance angular cones have been discussed in details in ref (25). However, no information has been provided on the relative role of direct and indirect mechanisms and therefore on partial ionization cross sections associated with the different reaction paths.</p><p>The present study, exploiting the analytical PES,35 attempts to deconvolve the effective imaginary part,25 separating the contributions from chemical and physical components of intermolecular forces, in order to identify the relative role of the two direct and indirect (basic) mechanisms. This preliminary objective is fundamental for characterizing the dependence of the relative role of the two basic reaction mechanisms on Ecoll and therefore on the orbital angular momentum quantum number of the collision complex that, in a classical picture, relates to the impact parameter b. Here, we analyze in detail two geometries of reagents approach, the one close to the C3v molecular axis and the one in proximity of the perpendicular to this axis, that control the formation of NH3+ in the ground and first excited electronic states, respectively. Note that the two selected geometries are representative of the most probable configurations within the acceptance angular cones where the reaction occur, leading to the formation of a different type of ionic products. This allows the use of the same function for the imaginary Γ1 and Γ2 components, triggering direct and indirect mechanisms, respectively, whose general exponential formulation is borrowed from the atom–atom CHEMI reactions6,7 to describe direct and indirect mechanisms with their state-to-state dependence. In the present analysis, only the pre-exponential factors are adjusted to reproduce the magnitude and energy dependence of the total and partial ionization cross sections. Note also that any averaging over the angular acceptance cones is expected to change the pre-exponential factor values but not their ratios. The methodological choice of two selected configurations within the angular cones allows to highlight the different role of the centrifugal potential respect to the intermolecular interaction.</p><!><p>the quantities Γ1 and Γ2 must be related to intermolecular forces of a specific nature, whose strength scales in a different way with R. In particular, while chemical components, depending on the overlap integral between orbitals exchanging the electron, emerge at short separation distances and are strongly varying with R, those of physical origin show a much less radial dependence. Accordingly, completely different exponential functions have been adopted for the two imaginary terms, formulated as suggested from the detailed study of atom–atom reactions.6,7</p><p>Their relative and absolute strengths have been modeled in order to reproduce total and partial ionization cross sections in the right scale of experimental determinations. Considering the results provided by our25 and by another laboratory,36 cross sections represent a critical test of predicted values, since they cover 1–2 orders of magnitude and the probed Ecoll varies for subthermal (∼0.1 meV) up to hyper-thermal values (−103 meV), a changing of ∼4 orders of magnitude. More specifically, total ionization cross sections measured by the Losanna group36 vary from 300 to 400 Å2 (Ecoll = 01 meV) up to ≃100 Å2 (Ecoll = 10 meV), while those obtained in our laboratory cover a complementary range and are shown in Figure 1a.</p><p>Collision energy dependence of partial (σ1, σ2) and total (σtot) cross sections evaluated from individual (Γ1, Γ2) and total (Γtot) components of the imaginary part (see Table 1) and referred to the two selected geometries. The subscripts 1 and 2 indicate the direct and indirect reaction mechanism contributions separately, respectively.</p><!><p>Details of the selected geometries and on the formulation of the imaginary components are given in Table I.</p><!><p>The Γ1 and Γ2 components for the two selected geometries of the approach of reagents within the acceptance angular cones (see text) have the same analytical formulation in order to highlight the different role of the centrifugal potential with respect to the intermolecular interaction potential.</p><!><p>The total and partial (i.e., referred to each mechanism) ionization cross sections predicted by our method, and calculated within the semiclassical treatment, whose details can be found in refs (6) and (7), are shown in Figure 5. The results show that the role of two mechanisms and their energy dependence are completely different for the two geometries here considered, since the combination of the intermolecular potential and the centrifugal barrier selectively modulates the range of intermolecular distances probed so exalting the different role of direct and indirect mechanisms. In particular, for the geometry close to the C3v NH3 axis ("collinear"), producing the NH3+(X) ground state, the direct mechanism is dominant at all Ecoll values. However, for the "broadside" geometry, which gives rise to the formation of the NH3+(A) excited state with its subsequent dissociation into NH2+ + H, the direct mechanism becomes dominant only at hyper-thermal values of Ecoll. The contrasting behavior is ascribable to the different stability of the adduct formed in the two selected geometries for collision of reagents. Therefore, the detailed characterization of the dynamical evolution of the two types of collision complexes, leading to the turning points, which represent the most critical intermolecular distances where the reaction manifests the highest probability to occur, provides additional insight into critical features of the reaction stereodynamics. In particular, understanding the dependence of the turning points on b or on , that have been characterized, as emphasized in Figure 6, by a critical comparison between sum of real potential and centrifugal contribution with Ecoll, is of great interpretational value.</p><!><p>(upper panels) The dependence on the intermolecular distance R of the effective potential given as sum of the real component Vθφ(R) and of the centrifugal contribution . (lower panels) The turning point dependence on the impact parameter b, or in the quantum picture on , evaluated for the two geometries at selected Ecoll values that cover 4 orders of magnitude.</p><!><p>From the data reported in Figure 6 emerges the important selective role of the centrifugal barrier that generates, at collision energies lower than a critical Ecoll value, turning points confined in well-separated ranges of values, where the reaction probability is completely different, while for higher collision energies a unique extended range of turning points becomes effective. As depicted in Figure 6, for the "collinear" geometry the critical Ecoll value is ∼100 meV, while for the "broadside" geometry it amounts to ∼10 meV, and this variation arises from the different strength of the real potential that drives the collision. In particular, the centrifugal potential vanishes the trapping effect of the interaction more easily for the side approach because of the weaker attraction.</p><p>The striking selectivity feature is that, for Ecoll lower than the critical value, only a limited range of b or values controls the reactivity. Moreover, under such conditions the collision time is sufficiently long, the phase accumulation along each reaction path depends on a similar passage from long to short R values, and then the relative role of the two mechanisms is approximately constant, as shown in Figure 5. Along these trajectories, the chemical reactivity can be also enhanced by the possible orientation of the polar molecule within the strong and anisotropic intermolecular field probe, favored by the low Ecoll and by the long collision time. At higher Ecoll, in contrast, the range of turning points effective for reaction enlarges significantly, since the selective role of the centrifugal barrier (see upper panels in Figure 5), which separates short and large turning points, disappears. This is confirmed by the results in Figure 7, where, as a function of Ecoll, are plotted the total cross section (due to all b or values contributions) and the partial cross section, determined exclusively by b or lower than b or values determining the absolute maximum of centrifugal barrier associated at each Ecoll (bmax or , and for values see Table II). In particular, at Ecoll lower than the critical value, the centrifugal barrier completely separates the ranges of b or driving the collisions (see Figure 6), making ineffective turning points determined by b > bmax or , since they occur at too large R values.</p><!><p>Total, σtot, and partial, σpar, ionization cross sections determined by all values and by , respectively.</p><!><p>Obtained cross-section values and their ratios demonstrate that the contribution from highest b or values becomes appreciable only for Ecoll larger than the critical value. Accordingly, the selective role of the centrifugal barrier tends to disappear, and a unique interval of b or values promotes the reaction, making effective also those larger than bmax or , since determining turning points at intermediate and moderately large R.</p><!><p>It is important to stress that all stereodynamical effects emphasized for the CHEMI of Cl2 and NH3 must be considered averaged over all fine structure states accessible to the open-shell Ne*(3PJ) reagent, identified for atom–atom reactions by J and Ω quantum numbers. Note that Ω provides the absolute value of the projection of the total (sum of the orbital and spin components) electronic angular momentum J along R, and it indirectly defines also the alignment degree of the half-filled p-orbital of Ne*(3PJ) reagent respect to R. As demonstrated for atom–atom CHEMIs,6,7 both real and imaginary parts of the optical potential are depending on J and Ω, and this determines the opening of different state-to-state reaction channels. Figure 8 summarizes some basic differences between CHEMI reaction dynamics involving molecules in terms of a qualitative scheme of the potential energy curves that characterizes (see previous sections) two limiting cases for the direct mechanism. Figure 8a concerns molecules with positive electron affinity (e.g., Cl2 and O2), while Figure 8b concerns the other cases (e.g., NH3, N2, and CO).</p><!><p>Scheme of the potential energy curves characterizing the two limiting cases for the direct mechanism. (a) CHEMI involving molecules with positive electron affinity (e.g., Cl2 and O2); (b) other cases mentioned in the text (e.g., NH3, N2, and CO).</p><!><p>For CHEMIs of molecules, the characterization of state-to-state reaction channels, with their dependence on both atomic alignment and molecular orientation, will be the target of a future extension of our methodology. Particular attention must be further addressed to the N2 and O2 reagents for which differences in the collinear and perpendicular approach of the diatomic molecule to Ne* are expected to emphasize a new selectivity in the reaction dynamics. In particular, while N2, from a phenomenological point of view (see Figure 1), behaves similarly to CO and CH4, with a total ionization cross section that increases with Ecoll, under thermal collision energies the cross section of O2 is at least a factor 3 larger with respect to that of N2, and it decreases with Ecoll as for Cl2 and C2H2. The present paper suggests that the different behavior in the ionization cross sections of CHEMI involving molecules probably arises from a different balance of the intermolecular forces involved, which selectively depend on the fundamental physical/chemical properties of the molecules. In particular, while the electronic polarizability is comparable for N2 and O2, the electric quadrupole moment (that of N2, depicted in Figure 2, is approximately a factor 4 larger with respect to that of O2), energetics, and symmetry of highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO) molecular orbitals are completely different in the two cases.</p><p>Finally, it is interesting to note for Ne*-N2 important features of the isotropic optical potential were obtained from a multiproperty analysis of several experimental findings.37 In such a study the use of a combination of two imaginary components was necessary to reproduce simultaneously all analyzed experimental observables, which were probing complementary details of the interaction. According to the suggestions of this paper and our recent studies,6,7 this necessity probably relates to the occurrence of two competitive reaction mechanisms.</p><p>Therefore, the investigation from a phenomenological point of view of CHEMI reactions of prototype diatomic and polyatomic molecules emphasizes again the importance of experiments performed under single collision conditions and addressed to measure both ionization cross sections and PIES. The combined analysis of the experimental findings, that must be carried out adopting a proper formulation of the leading interaction components driving the collision dynamics, is then crucial to define the relative role of direct and indirect reaction mechanisms as a function of the geometry of the reagents approach and of the collision energy. The analysis of the reaction stereodynamics has allowed us to highlight important details on the microscopic redox mechanism of CHEMI, which is strongly dependent on the fundamental intrinsic characteristics of the target molecule and on the specific intermolecular interactions existing between the colliding partners, both crucial aspects in determining the formation of the transition state of the reaction. Our treatment is able to fully describe such reactions passing over the range from high temperature to ultracold collisions. This highlights the fact that that the "canonical" chemical oxidation process, dominant at a high collision energy, changes its nature in the subthermal regime to a pure direct photoionization process.7 It also points out differences between the cold chemistry of terrestrial and interstellar environments and the hot one of combustion and flames.11−13,38,39</p><!><p>The authors declare no competing financial interest.</p>

PubMed Open Access

Synthesis, Biological Evaluation, and Molecular Modeling of 11H-indeno[1,2-b]quinoxalin-11one Derivatives and Tryptanthrin-6-Oxime as c-Jun N-terminal Kinase Inhibitors

c-Jun N-terminal kinases (JNKs) play a central role in many physiologic and pathologic processes. We synthesized novel 11H-indeno[1,2-b]quinoxalin-11-one oxime analogs and tryptanthrin-6-oxime (indolo(2,1-b)quinazoline-6,12-dion-6-oxime) and evaluated their effects on JNK activity. Several compounds exhibited sub-micromolar JNK binding affinity and were selective for JNK1/JNK3 versus JNK2.The most potent compounds were 10c (11H-indeno[1,2b]quinoxalin-11-one O-(O-ethylcarboxymethyl) oxime) and tryptanthrin-6-oxime, which had dissociation constants (Kd) for JNK1 and JNK3 of 22 and 76 nM and 150 and 275 nM, respectively. Molecular modeling suggested a mode of binding interaction at the JNK catalytic site and that the selected oxime derivatives were potentially competitive JNK inhibitors. JNK binding activity of the compounds correlated with their ability to inhibit lipopolysaccharide (LPS)-induced nuclear factor-\xce\xbaB/activating protein 1 (NF-\xce\xbaB/AP-1) activation in human monocytic THP-1Blue cells and interleukin-6 (IL-6) production by human MonoMac-6 cells. Thus, oximes with indenoquinoxaline and tryptanthrin nuclei can serve as specific smallmolecule modulators for mechanistic studies of JNK, as well as potential leads for the development of anti-inflammatory drugs.

synthesis,_biological_evaluation,_and_molecular_modeling_of_11h-indeno[1,2-b]quinoxalin-11one_deriva

Introduction<!>Chemistry<!>Structure\xe2\x80\x93activity relationship (SAR) analysis for JNK1\xe2\x80\x933 binding affinity<!>Kinase inhibition profile of tryptanthrin-6-oxime<!>Molecular modelling<!>Evaluation of compound biological activity<!>Conclusion<!>Reagents and general procedures<!>6-Methyl-11H-indeno[2,3-b]quinoxalin-11-one oxime (6a) and general procedure for the synthesis of 6b-i.<!>6,8-Dichloro-11H-indeno[2,3-b]quinoxalin-11-one oxime (6b).<!>7,8-Dimethyl-11H-indeno[2,3-b]quinoxalin-11-one oxime (6c).<!>8-Carboxy-11H-indeno[2,3-b]quinoxalin-11-one oxime (6d).<!>7-Methyl-11H-indeno[2,3-b]quinoxaline-11-one oxime (6e).<!>7-Ethoxy-11H-indeno[2,3-b]quinoxalin-11-one oxime (6f).<!>8-Nitro-11H-indeno[2,3-b]quinoxalin-11-one oxime (6g).<!>8-Trifluoromethyl-11H-indeno[2,3-b]quinoxalin-11-one oxime (6h).<!>7-Chloro-11H-indeno[2,3-b]quinoxalin-11-one oxime (6i).<!>6H-Indeno[1,2-b]pyrido[3,2-e]pyrazin-6-one (3a).<!>3-Chloro-6H-indeno[1,2-b]pyrido[3,2-e]pyrazin-6-one (3b).<!>10H-Indeno[1,2-b]pyrido[3,4-e]pyrazin-10-one (5).<!>6H-Indeno[1,2-b]pyrido[3,2-e]pyrazin-6-one oxime (7a).<!>3-Chloro-6H-indeno[1,2-b]pyrido[3,2-e]pyrazin-6-one oxime (7b).<!>10H-Indeno[1,2-b]pyrido[3,4-e]pyrazin-10-one oxime (8).<!>6-(Hydroxyimino)indolo[2,1-b]quinazolin-12(6H)-one (tryptanthrin-6-oxime).<!>11H-indeno[1,2-b]quinoxalin-11-one O-methyl oxime (9a) and general procedure for synthesis of the 9b-d.<!>11H-Indeno[1,2-b]quinoxalin-11-one O-ethyl oxime (9b).<!>11H-indeno[1,2-b]quinoxalin-11-one O-benzyl oxime (9\xd1\x81).<!>11H-Indeno[1,2-b]quinoxalin-11-one O-allyl oxime (9d).<!>11H-Indeno[1,2-b]quinoxalin-11-one O-(O-ethylcarboxymethyl) oxime (10c) and general procedure for synthesis of 10a,b,d. Method A:<!>11H-Indeno[1,2-b]quinoxalin-11-one O-isobutyl oxime (10a).<!>11H-Indeno[1,2-b]quinoxalin-11-one O-carboxymethyl oxime (10b).<!>11H-Indeno[1,2-b]quinoxalin-11-one O-(2-hydroxyethyl) oxime (10d).<!>Kinase profiling and Kd\ndetermination<!>Cell culture<!>Analysis of AP-1/NF-\xce\xbaB activation<!>Cytokine analysis<!>Cytotoxicity assay<!>Western blotting<!>Molecular modeling

<p>c-Jun N-terminal kinases (JNKs) belong to the family of mitogen-activated protein kinases (MAPK) that are activated in response to various stress stimuli, such as ultraviolet radiation, oxidative stress, heat and osmotic shock, and ischemia-reperfusion injury of the brain and heart [1–4]. In mammals, 10 highly similar isoforms are expressed by alternative splicing of three different genes: JNK1 (four isoforms), JNK2 (four isoforms), and JNK3 (two isoforms) [5, 6]. JNK1 and JNK2 are found in all cells and tissues of the body, while JNK3 is expressed mainly in the brain, heart, and testicles [2].</p><p>The JNKs have been shown to play an important role in regulation of the signaling pathways involved in apoptosis, necrosis, inflammation, and ischemia/reperfusion injury [7–10]. They are involved in a wide range of diseases, including rheumatoid arthritis, osteoarthritis, multiple sclerosis, inflammatory bowel disease, insulin resistance, tumorigenesis, stroke, renal ischemia, Alzheimer's and Parkinson's diseases [6, 11–17].Upstream kinases of the MAPK cascade (MKK4 and MKK7) phosphorylate and activate JNK [18], whereas transcription factors such as c-Jun, specificity protein 1 (Sp1), activating transcription factor 2 (ATF2), and nuclear factors of activated T-cells (NFATc2 and NFATc3) are substrates for phosphorylation-activated JNKs [5, 7, 9, 19]. There are also numerous non-nuclear substrates of JNK that participate in the degradation of proteins, signal transduction, and regulation of apoptotic cell death [3, 20]. For example, JNK1 phosphorylates insulin receptor substrate 1 (IRS-1), a key molecule in the insulin-sensing pathway, which down-regulates insulin signaling [21]. Recently Tudor-SN, a multifunctional protein that is implicated in a variety of cellular processes, was identified as a novel JNK target [22].</p><p>A significant amount of pharmacological and genetic evidence suggests that inhibition of JNK signaling may represent a promising therapeutic strategy [23], and numerous efforts have focused on the development of selective and nontoxic JNK inhibitors. For example, selective JNK1/3 inhibitors may have clinical benefit in treating neurodegenerative disorders [24]. However, it has been difficult to design selective JNK inhibitors because of the high sequence identity among JNK isoforms (from 73 to 75%) and, specifically, sequence identity of their ATP binding pockets (close to 98%) [25]. Recently, JNK2/3 inhibitors with an aminopyrazole scaffold that have >30-fold selectivity over JNK1 were identified [25].</p><p>Although no JNK inhibitors have been approved for use in humans, a few small molecule JNK inhibitors have entered clinical trials for various indications, including tanzisertib for the treatment of lupus erythematous and idiopathic pulmonary fibrosis, bentamapimod for the treatment of inflammatory endometriosis, and D-JNKi1 for the treatment of inflammation and stroke (for review [26–28]). Previously, we identified a new class of JNK inhibitors based on the 11H-indeno[1,2-b]quinoxalin-11-one scaffold [29]. Specifically, compound IQ-1 (11-Hindeno[1,2-b]quinoxalin-11-one oxime) and its oxime analogs (Fig. 1) inhibited JNK activity and, consequently, proinflammatory cytokine production by murine and human leukocytes [29]. We also found that IQ-1 reduced inflammation and cartilage loss associated with collageninduced arthritis (CIA) [30] and protected against cerebral ischemia–reperfusion injury in mice [31]. These JNK inhibitors contain oxime (IQ-1) or O-acyl-oxime (IQ-2 through IQ-4 and IQ-6) groups (Fig. 1) and exhibited some (~3.6–5.3 fold) selectivity for JNK3 versus JNK1/2 [29, 30]. Thus, we propose that modification of IQ-1 by introduction of various substituents could increase potency and/or a selectivity of the resulting analogs toward the JNK isoforms.</p><p>In the present studies, novel 11H-indeno[1,2-b]quinoxalin-11-one oxime analogs, such as aza-analogues, O-substituted derivatives, and analogs with different substituents in the indenoquinoxaline tetracyclic moiety, were synthesized and evaluated against JNK1–3. We conducted molecular modeling for selected compounds and estimated their anti-inflammatory potential using in vitro cell-based assays. We also report for the first time that tryptanthrin-6oxime, a derivative of the natural alkaloid tryptanthrin, has a relatively high binding affinity for JNK1–3 and tropomyosin‐related kinases (TRK) A and B.</p><!><p>All new compounds were synthesized as reported in Scheme 1, and the structures were confirmed on the basis of analytical and spectral data. As reported previously, 11H-indeno[1,2b]quinoxalin-11-one (compound 1) was synthesized by the condensation of ninhydrin with ophenylenediamine [32, 33]. We synthesized oximes of known [34–36] and commercially available ketones 2a-i as described in Scheme 1A.</p><p>To synthesize indenoquinoxaline analogues 3a, 3b, and 5 containing an additional nitrogen atom in the tetracyclic nucleus, we used the reaction of 2,3-diaminopyridine, its 5chloro derivative, and 3,4-diaminopyridine, with ninhydrin in EtOH (Scheme 1B). It has been established that the use of H2O as a solvent instead of EtOH does not significantly affect yields of the products and selectivity of the process. The existence of two isomers is possible for each of the resulting aza-analogues. We determined the isomer ratios from integral intensities of the signals in 1H-NMR spectra of the products and found that 3,4-diaminopyridine reacts regiospecifically, with the sole formation of 5 (73% yield). To provide a rationale for regioselectivity of this reaction, we performed density functional theory (DFT) calculations (see the description and Supplementary Fig. S1). Previously, compound 5 was synthesized with comparable yield at higher temperature in boiling isobutyl alcohol [37]. From the reaction of 2,3diaminopyridine with ninhydrin, a mixture of isomers 3a and 4a (90:10%) was obtained, with a total yield of 82%, from which we isolated pure compound 3a. It should be noted that in boiling MeOH, an inseparable mixture of ketones 3a and 4a was obtained [36]. Reaction of 2,3-diamino6-methylpyridine leads to a mixture of compounds 3b and 4b (83:17%). From this mixture, we isolated pure isomer 3b by recrystallization from dimethylformamide.</p><p>Oximes were synthesized via ketone precursors through a reaction with hydroxylamine. Treatment of compounds 2a-i with hydroxylamine in hot EtOH in presence of NaOH led to the 11H-indeno[1,2-b]quinoxalin-11-one oximes (6a-i) (Scheme 1A). Likewise, oximes 7a, 7b, 8, and tryptanthrin-6-oxime were synthesized from ketones 3a, 3b, 5, and tryptanthrin, respectively, using hydroxylamine hydrochloride in EtOH or pyridine (Scheme 1B). For the synthesized oximes, the ratio of the Z- and E-isomers was determined from the integral intensities of the signals in 1H-NMR spectra. We found that compounds 6a-d, 6f-i, 8, and tryptanthrin-6-oxime were formed as individual isomers, while a mixture of Z- and E-isomers (90:10) was obtained for oxime 7a. The Z- and E-isomers of oximes 6e and 7a,b exist in dynamic equilibrium in solution and could not be isolated as individual forms. We speculate that the Z-isomer is predominant for the synthesized oximes, since it must be stabilized by an intramolecular H-bond between the OH group and the nitrogen atom of the pyrazine ring.</p><p>A convenient synthetic route to the O-substituted derivatives was synthesis from the corresponding ketone 1 by an oximation reaction with O-R-hydroxylamines (Scheme 1C). In the present work, we carried out the oximation of compound 1 using O-methyl, O-ethyl, O-benzyl, and O-allyl hydroxylamine hydrochlorides. According to the 1H-NMR spectra, products 9a-d synthesized according to Scheme 1C were isomerically pure individual compounds.</p><p>We also investigated the reactivity of IQ-1 towards alkylating reagents. IQ-1 has low solubility in most organic solvents, thus alkylation was evaluated in dimethylsulfoxide (DMSO); solubility of IQ-1 in this solvent is about 0.01 M at room temperature. KOH was used as a base. Being an aprotic solvent, DMSO easily solvates the potassium cations, while the OH-anions are solvated slightly, which leads to an extremely high basicity of the medium and activates the alkylation process in the DMSO-KOH system [38]. We also used Na2CO3 as a base. O-alkylation of IQ-1 in DMSO was carried out according to Scheme 1D at room temperature and vigorous stirring in the presence of a two-fold molar excess of a base (threefold on the synthesis of the carboxylic acid 10b).</p><p>Compound 10a in CDCl3 solution exists as a mixture of Z- and E-isomers with respect to the exocyclic C=N bond, as two sets of side-chain proton signals are observed in the 1H-NMR spectrum. The ratio of isomers is approximately 1:2, as determined from the integral intensities in each pair of signals. According to DFT calculations [B3LYP/6–31+G(d,p)] of the 10a isomers, the E-isomer is thermodynamically more stable. The effect of the solvent (chloroform) was taken into account within the polarizable continuum model (PCM). We determined that for the 10a(Z)10a(E) equilibrium, ∆Go298 is equal to 5.23 kJ/mol. The 1H-NMR results for compound 10c showed two sets of the side chain proton signals of the isomers with 1:3 integral intensity. Similarly to ⇌ 10a, indenoquinoxaline 10c in chloroform has a more stable E-isomer. Thus, for the 10c(Z)⇌10c(E) process, ∆Go298 evaluated by DFT is 10.38 kJ/mol. Obviously, the mixture of Z- and E-isomers is formed on synthesis under the reaction conditions.</p><p>It should be noted that the use of DMSO-Na2CO3 instead of DMSO-KOH (Scheme 1D) led to a longer reaction time: complete alkylation of IQ-1 was attained in 9–10 hours. However, on the alkylation by ethyl chloroacetate, isomerically pure 10c was formed with only traces of the minor isomer present, in contrast to the method using the superbasic medium DMSO-KOH. Comparison of the 1H-NMR spectra of the isomer mixture and the individual isomer 10c synthesized in DMSO-KOH and DMSO-Na2CO3 systems, respectively, showed that in the latter case, the product consisted of the isomer that was predominant when DMSO-KOH medium was used. According to results of our DFT calculations presented above, this isomer has an E-configuration, and this product was used for further biological evaluation.</p><!><p>All compounds were evaluated for their ability to bind to the three JNK isoforms in comparison with IQ-1, and the results presented in Table 1 demonstrated that the 11H-indeno[1,2-b]quinoxalin-11-one nucleus is an appropriate scaffold for JNK inhibitor development. Indenoquinoxalines 6e, 7a, and 10c exhibited Kd values in the nanomolar range for all three JNKs, with the most potent being 10c, which had even lower Kd values for JNK1 and JNK3 compared to IQ-1 [29]. Moreover, 10c had much higher specificity toward JNK1 and JNK3 (Kd values of 22 nM and 76 nM, respectively) versus JNK2 (Kd = 735 nM). To further evaluate the relative potency of 10c, we compared its binding affinity with that of a commercially available JNK inhibitor, SP600125. As shown in Table 1, the Kd of 10c toward JNK1 was even lower to that of SP600125.</p><p>As reported previously [29], we found that the side chain oxime R substituent was critical for JNK binding and biological activities. The observation that oxime derivatives 9a-d and 10a, which have hydrocarbon side chains, were inactive in the competition binding assay suggests that the R oxime substituent is involved in H-bond donor/acceptor interactions with JNK. These interactions occur possibly due to the presence of additional oxygen atoms in the carboxyl, ester, or OH groups of molecules 10b-d, which can be anchored in the binding site in a favorable conformation, as shown below for 10c. Although oxime groups may contribute important interactions in the JNK binding site, the tetracyclic nucleus seems to be responsible for proper ligand positioning. Indeed, substitution of an aromatic carbon atom at position 6 with nitrogen led to less active compound 8 with all JNK isoforms. On the other hand, substitution of a carbon atom at position 8 with nitrogen (7a) or introduction of a CH3 group as the R2 substituent (6f) had little effect on binding affinity with all three JNK isoforms. Other modifications of the tetracyclic nucleus, including introduction of CH3 at R3 (6a), OCH2CH5 or NO2 at R2 (6f and 6g, respectively), COOH at R1 (6d), and two CH3 groups at R1 and R2 (6d), led to compounds with relatively low JNK binding affinity. Furthermore, 6h containing a CF3 group at R2 and 6b with two Cl atoms at R1/R3 were completely inactive. The most interesting modification of the tetracyclic nucleus was the introduction of Cl at position R1 (6i), as it led to an increase in relative specificity toward JNK1/JNK3 versus JNK2.</p><p>The natural alkaloid tryptanthrin has an indolo(2,1-b)quinazoline-6,12-dion nucleus, which is analogous to the 11H-indeno[1,2-b]quinoxalin-11-one scaffold. Indeed, charge distributions in IQ-1 and tryptanthrin-6-oxime molecules are very similar (Fig. 2), although the latter has a very polar carbonyl group, which results in lower hydrophobicity (LogP values are 4.04 and 2.92 for IQ-1 and tryptanthrin oxime, respectively). Thus, we also evaluated JNK binding activity of this IQ-1 analog. Although tryptanthrin was inactive for JNK2/JNK3 and had a very low binding affinity for JNK1 (Kd ~23.0 µM), tryptanthrin-6-oxime exhibited high binding affinity for JNK1 and JNK3 (Table 1).</p><!><p>Since tryptanthrin-6-oxime demonstratedd high affinity for JNKs, we evaluated its specificity for various other kinases to evaluate its specificity compared to IQ-1. Specifically, it was profiled in a competition binding assay for its ability to compete with an active-site directed ligand for 97 different kinases (KINOMEscan, Eurofins Pharma Discovery, San Diego), representing all known kinase families. The panel included 10 kinases that were reported previously to be targets of SP600125 with similar or greater potency than the JNKs [39]. Tryptanthrin-6-oxime was screened at 10 µM, and the kinases for which >90% inhibition of ligand binding and kinase activity was observed were designated as "kinase targets of the compound." Five such kinase targets were identified, including casein kinase 1 σ (CK1σ, gene symbol CSNK1D), tropomyosin‐related kinase A (TRK-A, gene symbol NTRK1), JNK1, JNK2, and JNK3 (Fig. 3). Thus, similar to IQ-1 [30], tryptanthrin-6oxime had high specificity for inhibition of human JNK isoforms. Note however, that IQ-1 and IQ-3 (both potent JNK inhibitors with an indenoquinoxaline scaffold) did not bind TRK-A [29, 30]. Because TRKA-C are important targets for treatment of several tumors [40–42], the parent tryptanthrin and tryptanthrin-6-oxime were evaluated for their binding affinities (Kd) to these 3 kinases. We found that tryptanthrin-6-oxime had higher affinity toward TRKA-C in comparison with the parent alkaloid (Table 2).</p><p>Activity of TRK-family proteins (TRKA-C) is associated with poor survival in many types of cancer [43]. For example, TRK-A, a high affinity receptor for nerve growth factor (NGF) has been associated with the development of epithelial ovarian cancer [44]. Brain-derived neurotrophic factor (BDNF) is a potent neurotrophic factor that has been shown to stimulate breast cancer cell growth and metastasis via TRK-A and TRK-B [45]. Several compounds, including crizotinib and entrectinib, have been shown to inhibit the growth of tumor cells that express TRK-family fusion proteins and have demonstrated remarkable clinical response in patients with TRK-A fusion-positive tumors [46–48]. Sharma et al. [49] reported that some oxime derivatives of tryptanthrin exhibited anticancer activity in vitro against a panel of human cancer cell lines, but mechanisms of this activity are still non-identified. To our knowledge, this is first report demonstrating co-activity of a kinase inhibitor toward TRK and JNK isoforms. Using a selectivity score S(10), based on >90% inhibition of ligand binding at a single 10 µM screen concentration [50], we found that the S(10) for tryptanthrin-6-oxime was much lower (0.015 = 5/99) compared with the S(10) for SP600125 (0.328 = 39/119) [51], indicating much higher target kinase selectivity for tryptanthrin-6-oxime.</p><!><p>To further characterize our most active analogs, we performed docking studies of 10c and tryptanthrin-oxime into the binding sites of the three JNK isoforms. Since tryptanthrin was inactive, we were also able to directly compare binding of the inactive parent and active oxime derivative. According to our modelling, tryptanthrin formed a weak H-bond with Asn114 on binding with JNK1. At the same time, the highest partial interaction energy of this molecule was observed with Met111, which was due to van der Waals forces. The docking pose of tryptanthrin-6-oxime (Fig. 4) was characterized by strong H-bonding between the oxygen atom of the amide group and Met111. It should be noted that Met111 is considered as an important residue for small molecule interactions with JNK [52, 53]. The calculated docking score for tryptanthrin-6-oxime was about 15 kcal/mol more negative than for tryptanthrin, which may explain the higher binding affinity of the oxime derivative. Docking studies of these compounds to JNK2 showed that the parent alkaloid did not form H-bonds with any of the residues of this kinase, retaining in the binding center only by non-valent interactions. The highest attraction with a score of 14 kcal/mol was obtained for His149. In contrast, tryptanthrin-6-oxime was Hbonded with JNK2 through its oxime group with Gly171 (Fig. 4). Docking scores for tryptanthrin and its oxime derivative differed by 16 kcal/mol in favor of the oxime. According to the docking results obtained for JNK3, the low-energy pose for tryptanthrin formed a weak Hbond through its amide oxygen with Asn152 and was fixed in the binding site mainly by van der Waals interactions. On the other hand, tryptanthrin-6-oxime was anchored in the kinase cavity through H-bonding of the oxime group with Asp207 (Fig. 4). The docking score of the oxime in JNK3 was ~ 30 kcal/mol more negative than that of tryptanthrin. Thus, it can be assumed that, at least for JNK2 and JNK3, the introduction of an oxime moiety into the molecule of tryptanthrin caused the formation of a new H-bond with the kinase through participation of this moiety.</p><p>Docking of the highly active compound 10c in JNK1 gave a pose similar to tryptanthrin6-oxime, meaning that the molecule formed a strong H-bond with Met111 via the ester group of the ligand (Fig. 4). It is important that such an arrangement of the ester group is achieved for the Z-isomer of 10c. We also performed docking of the E-isomer, but another pose with a markedly worse docking score was obtained in this case. In its unbound form, the E-isomer of 10c is more stable; however, our DFT calculations show that the Z-isomer in solution is only slightly higher in energy than the E-isomer of the substituted oxime 10c. Obviously, when interacting with the kinase, 10c adopts the Z-configuration, which binds more effectively to the JNK1 active site, and in general, a gain in energy is achieved. When docking compound 10c in the JNK2 binding site, a more energy-efficient pose was obtained for the E-isomer (by 20.4 kcal/mol better according to the docking score) than for the Z-isomer. Compound 10c forms two strong H-bonds with Lys55 and Leu168 of JNK2 with participation of two nitrogen atoms in the heterocycle. It should be noted that the oxygen of the oxime group is located near one of these nitrogen atoms and forms an H-bond with Gly171 located in the vicinity of Lys55 for the pose of tryptanthrin-6-oxime (Fig. 4). We determined that 10c binds JNK3 in the form of the Z-isomer (by 60.2 kcal/mol lower in docking score than the corresponding E-isomer), forming two H-bonds to Lys93 with participation of the oxime and ethoxy oxygen atoms. In the pose of tryptanthrin-6-oxime, the oxygen atom, although located in the same region of space, forms an H-bond with Asp207 (Fig. 4). Consequently, there is a similarity in the location of the most active compounds (10c and tryptanthrin-6-oxime) by their location within the binding sites of the three JNK isoforms. Note that these molecules occupy the same region of space where co-crystallized ligand SP600125 is located. As shown in Supplementary Fig. 2S, the tetracyclic moieties of all three compounds are approximately parallel within a narrow binding site of JNK3.</p><p>We also performed docking studies of tryptanthrin and tryptanthrin-6-oxime into the TRK-A binding site. The major difference between their docking poses was the presence of Hbonding between the oxime moiety of tryptanthrin-6-oxime and the kinase (Fig. 5). Specifically, the oxygen atom of the oxime group is strongly H-bonded with Asp596 and Arg599. Additionally, a weaker H-bonding interaction is possible between a nitrogen atom in the tetracyclic alkaloid derivative and Arg599. In contrast, tryptanthrin interacted with the kinase via van der Waals forces only, although a strong attraction of the ligand to Asp596 exists according to our calculations. The dissimilarity in docking modes of tryptanthrin and tryptanthrin-6-oxime is likely responsible for the difference in their binding affinities to TRPA-C (Table 2).</p><!><p>All compounds were evaluated for their ability to inhibit LPS-induced NF-κB/AP-1 reporter activity and interleukin (IL)-6 production in human monocytic THP-1Blue and MonoMac-6 cells, respectively. As shown in Table 1, the 13 oxime compounds inhibited LPSinduced NF-κB/AP-1 activity and IL-6 production. As examples, the dose-dependent inhibitory effects of 10c and tryptanthrin-6-oxime on NF-κB/AP-1 activity and IL-6 production are shown in Fig. 6. As expected, these compounds also inhibited c-Jun phosphorylation in treated cells derivatives (6a, 6c, 6e, 6f, 6i, 7a, 7b, 8, 10c, and tryptanthrin-6-oxime) all had IC50 values close to that of SP600125 for inhibition of LPS-induced NF-κB/AP-1 activity and IL-6 secretion in biological assays (Table 1). Consistent with the JNK binding assay, 6b, 6h, 9a-d, and 10a-b did not inhibit NF-κB/AP-1 activity or IL-6 production (Table 1; examples are shown in Fig. 6), supporting the specificity of our assays. In contrast to the active oximes, ketone derivatives (2a2j, 3a, 5) (data not shown), as well as tryptanthrin (Fig. 6), did not inhibit LPS-induced NFκB/AP-1 activity or IL-6 production, even at concentrations up to 50 µM.</p><p>To verify that the results were not influenced by possible toxicity, cytotoxicity of the compounds was evaluated at concentrations up to 50 µM in MonoMac-6 and THP-1Blue cells during a 24-h incubation with the compounds. None of the compounds affected cell viability, even at the highest tested concentrations, thereby verifying that these compounds were not cytotoxic during the 24-h incubation period of our assays (data not shown).</p><p>It should be noted that many aryl oxime derivatives, including IQ-1, release nitric oxide (NO) during their oxidoreductive bioconversion to ketones [31, 54, 55]. Thus, biological activities of NO and these ketone precursors, including trypthantrin, should also be considered in biological experiments. For example, although compound 1 (ketone corresponding to IQ-1) has not been shown to be a DNA intercalator [56], the ketone precursors of compounds 9a, 6i, and 7a have cytotoxicity against some cancer cell lines, probably because of their topoisomerase I inhibitory activity [36]. Tryptanthrin is a natural alkaloid found in Polygonum tinctorium and Isatis tinctoria [57, 58] and has been reported to have various pharmacological effects, such as anti-inflammatory [59–61], antimicrobial [62], and anti-tumor activity [63, 64]. Tryptanthrin has also been reported to suppress NO and prostaglandin E synthesis in macrophages exposed to oxidative stress [65] and inhibit enzymatic activity of 5-lipoxygenase, cyclooxygenase-2, and indoleamine 2,3-dioxygenase [66–68]. Previously, several tryptanthrin derivatives with different substituents have been reported, including compounds with antiplasmodium and antitoxoplasma activities, inhibitors of indoleamine 2,3-dioxygenase, and DNA triplex stabilizing agents [68–73].</p><p>We synthesized novel 11H-indeno[1,2-b]quinoxalin-11-one oxime analogs and tryptanthrin-6-oxime (indolo(2,1-b)quinazoline-6,12-dion-6-oxime) and evaluated their effects on JNK activity. Several compounds exhibited sub-micromolar JNK binding affinity and were selective for JNK1/JNK3 versus JNK2. The most potent compounds were 10c (11H-indeno[1,2b]quinoxalin-11-one O-(O-ethylcarboxymethyl) oxime) and tryptanthrin-6-oxime, which had dissociation constants (Kd) for JNK1 and JNK3 of 22 and 76 nM and 150 and 275 nM, respectively. Molecular modeling suggested a mode of binding interaction at the JNK catalytic site and that the selected oxime derivatives were potentially competitive JNK inhibitors. JNK binding activity of the compounds correlated with their ability to inhibit lipopolysaccharide (LPS)-induced nuclear factor-κB/activating protein 1 (NF-κB/AP-1) activation in human monocytic THP-1Blue cells and interleukin-6 (IL-6) production by human MonoMac-6 cells. Thus, oximes with indenoquinoxaline and tryptanthrin nuclei can serve as specific smallmolecule modulators for mechanistic studies of JNK, as well as potential leads for the development of anti-inflammatory drugs.</p><!><p>Synthesis and analysis of novel 11H-indeno[1,2-b]quinoxalin-11-one oxime analogs and tryptanthrin-6-oxime demonstrated that several of these compounds had high affinity for JNK and were selective for JNK1/JNK3 versus JNK2. These analogs also inhibited LPS-induced nuclear NF-κB/AP-1 activation and IL-6 production in human monocytic cells. Our molecular modeling showed that oxygen atoms of the oxime or ester groups participated in the formation of strong H-bonds with residues in the JNK1/JNK3 binding sites. Thus, it is reasonable to suggest that further modification of the O-substituent in the oxime moiety could lead to more specific inhibitors with higher JNK selectivity. Our results also suggest that pan-JNK inhibition may be suitable for suppression of the production of proinflammatory cytokines by human monocytic cells. Finally, the identified oximes represent new chemical tools that may be useful in further development of JNK and/or TRK inhibitors and could find application in the treatment of inflammatory diseases, neurodegenerative pathologies, and cancer.</p><!><p>Indenoquinoxaline ketones 2b, 2d, 2f, and 2g were purchased from Vitas-M Laboratory (Moscow, Russia); 2e and 2h were from Maybridge (Cornwall, United Kingdom); and 2i was from Specs (Delft, The Netherlands). Tryptanthrin was purchased from Combi-Blocks (San Diego, CA). All other starting reagents were purchased from Sigma Aldrich. The chemicals were analytical grade and used without further purification. Compounds 1 (11H-indeno[1,2b]quinoxalin-11-one) and IQ-1 (11H-indeno[1,2-b]quinoxalin-11-one oxime) were synthesized, as described previously [32]. Ketone 2a was synthesized according to [34, 35], and compound 2c was synthesized according to [36]. Reaction progress was monitored by thin-layer chromatography (TLC) with UV detection using pre-coated silica gel F254 plates (Merck) or a Silufol UV-254. The synthesized structures were confirmed on the basis of analytical and spectral data. The melting points (m.p.) were determined using an electrothermal Mel-Temp capillary melting point apparatus. Elemental analysis was performed with a Carlo Erba instrument. GC-MS analysis was performed on an Agilent 7890A GC combined with an Agilent 5975C mass detector (Agilent Technologies, USA); carrier gas was helium. LC-MS analysis was performed on an Agilent 1260 Infinity combined with an Agilent 6530 Accurate Mass Q-TOF detector. Compounds dissolved in 3-nitrobenzyl alcohol were subjected to fast atom bombardment (FAB) ionization using a 10 kV argon beam, and the mass spectra were recorded with a VG 70–70 EQ spectrometer. IR spectra were recorded on a FT-IR spectrometer Nicolet 5700 with KBr pellets. 1H NMR spectra were recorded on Bruker 400 or 600 MHz spectrometers. For atom numbering detaails, see Supplementary Fig. 3S. Representative NMR spectra for compounds 8 and 10c are provided in supplementary material.</p><!><p>A mixture of 2a (2.17 g, 9.45 mmol), hydroxylamine hydrosulfate (3.06 g, 23.6 mmol), and NaOH (1.0 g, 25 mmol) in EtOH (100 mL) was heated for 8 h at 60 oC. After cooling, the mixture was poured into water (600 mL), the precipitate was filtered out, dried, and recrystallized from EtOH. Yield 2.26 g (92%). M.p. 303–304°. 1H NMR (600 MHz, DMSO-d6), δ, ppm: 2.85 (s, 3H, CH3), 7.38–7.64 (m, 4H, H-2, H-3, H-7, H-8), 7.95 (d, 1H, 3J = 8 Hz, H-9), 8.20 (d, 1H, 3J = 7.6 Hz, H-4), 8.85 (d, 1H, 3J = 7.6 Hz, H-1), 13.31 (s, 1H, OH). M.w. 261.29. C16H11N3O. LC-MS – m/z (I, %): 262.04377 (100) [MH]+; 244.03701 (83) [MH – H2O]+. FABMS – m/z (I, %): 262 (100) [MH]+. A similar procedure was used for synthesis of the following 11H-indeno[2,3-b]quinoxaline-11-one oximes. To isolate sodium oximates of 6b and 6h, a 2fold excess of NaOH was added (with respect to hydroxylamine salt) after completion of the reaction.</p><!><p>Yield of sodium salt 93%.M.p. 344о. 1H NMR (600 MHz, DMSO-d6), δ, ppm: 7.55–7.70 (m, 2H, H-2, H-3), 7.77 (s, 1H, H7), 7.92 (s, 1H, H-9), 8.2 (d, 1H, 3J = 8 Hz, H-4), 8.55 (d, 1H, 3J = 7.5 Hz, H-1). M.w. 338.13. C15H6Cl2N3NaO. LC-MS – m/z (I, %): 316.91 (100) [MH]+; 298.84 (30) [MH – H2O]+.</p><!><p>Yield 89%. M.p. 323–324°. 1H NMR (600 MHz, DMSO-d6), δ, ppm: 2.46 (s, 3H, CH3), 2.50 (s, 3H, CH3), 7.66–7.73 (m, 2H, H-2, H-3), 7.90 (s, 1H, H-7 or H-9), 7.91 (s, 1H, H-7 or H-9), 8.14 (d, 3J = 8 Hz, H-4), 8.53 (d, 1H, 3J = 8 Hz, H-1), 13.26 (s, 1H, OH). M.w. 275.31. C17H13N2O. LC-MS – m/z (I, %):276.0828 (100) [MH]+. FAB-MS – m/z (I, %): 276 (100) [MH]+; 258 (40) [MH – H2O]+.</p><!><p>Yield 90%. M.p. 323–324°.1H NMR (600 MHz, CDCl ), δ, ppm: 7.68–7.76 (m, 2H, H-2, H-3), 8.03 (d, 1H, 3J = 7 Hz, H-6),8.20 (d, 1H, 3J = 8 Hz, H-4), 8.32 (d, 1H, 3J = 7 Hz, H-7), 8.58 (s, 1H, H-9), 8.60 (d, 1H, 3J = 7 Hz, H-1), 13.65 (s, 1H, =N-OH). M.w. 291.27. C16H9N3O3. LC-MS – m/z (I, %): 292.03225 (100) [MH]+; 274.02414 (50) [MH – H2O]+.</p><!><p>Yield 80%. M.p. 297–298°.1H NMR (600 MHz, DMSO-d6), δ, ppm: 2.58 (s, CH3), 7.65–7.74 (m, 3H, H-2, H-3, H-8), 7.94 (s, 1H, H-6), 8.03 (d, 1H, 3J = 7.6 Hz, H-9), 8.17 (d, 1H, 3J = 7 Hz, H-4), 8.55 (1H, 3J = 7 Hz, H1), 13.30 (s, OH). M.w. 261.29. C16H11N3O. LC-MS – m/z (I, %): 262.27436 (100) [MH]+. FABMS – m/z (I, %): 262 (100) [MH]+.</p><!><p>Yield 87%. M.p. 303°−304°. 1H NMR (600 MHz, DMSO-d6), δ, ppm: 1.44 (t, 3H, 3J = 5 Hz, CH3), 4.26 (q, 2H, 3J = 5 Hz, CH2), 7.30–7.59 (m, 3H, H-2, H-3, H-8), 8.00 (d, 1H, 3J = 7.5 Hz, H-9), 8.06 (s, 1H, H-6), 8.15 (d, 1H, 3J = 7 Hz, H-4), 8.78 (d, 1H, J = 7 Hz, H-1), 13,17 (s, 1H, OH). M.w. 291.31. C17H13N3O2. LCMS – m/z (I, %): 292.07027 (100) [MH]+; 274.06221 (41) [MH – H2O]+. FAB-MS – m/z (I, %):292 (100) [MH]+.</p><!><p>Yield 84%. M.p. > 360°. 1H NMR (600 MHz, DMSO-d6), δ, ppm: 7.42–7.67 (m, 2H, H-2, H-3), 8.18 (d, 1H, 3J = 7 Hz, H-4),8.31 (d, 1H, 3J = 7 Hz, H-6), 8.87 (d, 3J = 8 Hz, H-1), 8.90 (d, 3J = 7 Hz, H-7), 8.99 (s, 1H, H-9), 13.47 (s, 1H, OH). M.w. 292.26. C15H8N4O3. LC-MS – m/z (I, %): 293.54 (100) [MH]+; 275.50 (41) [MH – H2O]+.</p><!><p>Yield of sodium salt 81%. M.p. 303 . H NMR (600 MHz, DMSO-d6), δ, ppm: 7.60–7.75 (m, 2H, H-2, H-3), 8.02 (d, 1H, 3J = 7 Hz, H-6), 8.26 (d, 1H, 3J = 8 Hz, H-4), 8.34 (d, 1H, 3J = 7 Hz, H-7), 8.48 (s, 1H, H-9), 8.74 (d, 1H, 3J = 8 Hz, H-1). M.w. 337.24. C16H7F3N3NaO. LC-MS – m/z (I, %): 316.04372 (100) [MH]+.</p><!><p>Yield 89%. M.p. 323–324°. 1H NMR (600 MHz, DMSO-d ), δ, ppm: 7.47–7.64 (m, 2H, H-2, H-3), 7.76 (d, 1H, 3J = 8 Hz, H-8), 8.13 (d, 1H, 3J = 8 Hz, H-9), 8.21 (d, 1H, 3J = 7 Hz, H-4), 8.85 (d, 1H, 3J = 7 Hz, H-1), 13.64 (s, 1H, OH). M.w. 281.70. C15H8ClN3O. LC-MS – m/z (I, %): 282.14929 (100) [MH]+.</p><!><p>A mixture of 2,2-dihydroxyindane1,3-dione (ninhydrin, 0.39 g, 2.2 mmol) and 2,3-diaminopyridine (0.22 g, 2.0 mmol) in EtOH (50 mL) was heated for 10 h (TLC monitoring) at 60 °C. The mixture was then cooled, and the resulting precipitate (mixture of isomers 3a and 4a, 90:10 %) was filtered and recrystallized from EtOH to give 3a (0.50 g, 81% yield) as a yellow solid. M.p. 266–268°С. 1H NMR (500 MHz, CDCl3), δ, ppm: δ 7.68 (td, 1H, 3J = 7.5 Hz, 4J = 1 Hz, H-8), 7.73 (dd, 1H, 3J = 8 Hz, 4J = 4.5 Hz, H-3), 7.83 (td, 1H, 3J =7.5 Hz, 4J = 1 Hz, H-9), 7.97 (d, 1H, 3J = 7.5 Hz, H-7), 8.25 (d, 1H, J = 7.5 Hz, H-10), 8.60 (dd, 1H, 3J = 8 Hz, 4J = 2 Hz, H-4), 9.17 (dd, 1H, 3J = 4.5 Hz, 4J = 1.5 Hz, H2). NMR 13C (125 MHz, CDCl3), δ, ppm: 123.8 (C-10), 125.1 (C-3), 125.6 (C-7), 133.6 (C-8), 137.4 (C-6a), 137.4 (C-9), 138.3 (C-4a), 140.4 (C-4), 141.2 (C-10a), 150.6 (C-5a), 152.0 (C11a), 155.6 (C-2), 160.0 (C-10b), 188.8 (C-6). IR bands, cm−1: 1720 (C=O), 1612, 1602, 1502, 1153, 787. Found, %: C 72.24, H 3.18, N 18.30. C14H7N3O. Calculated, %: C 72.10, H 3.03, N 18.02.</p><!><p>Compound 3b was synthesized, as described under 4.1.1.12 from ninhydrin and 3,4-diamino-5-chloropyridine. Yield 58 %, M.p. 293–295. 1H NMR (400 MHz, CDCl3), δ, ppm: δ 7.72 (t, 1H, 3J = 7.2 Hz, H-8),7.87 (t, 1H, 3J = 7.2 Hz, H-9), 8.00 (d, 1H, 3J = 7.6 Hz, H-7), 8.26 (d, 1H, 3J = 7.6 Hz, H-10),8.55 (s, 1H, 4J = 2.4 Hz, H-4), 9.08 (s, 1H, 4J = 2.4 Hz, H-2). IR bands, cm−1: 1725 (C=O), 1610, 1596, 1481, 1134, 803, 752. Found, %: C 62.93, H 2.04, N 15.85. C14H6ClN3O. Calculated, %: C 62.82, H 2.26, N 15.70.</p><!><p>Compound 5 was synthesized, as described under 4.1.1.12 from ninhydrin and 3,4-diaminopyridine. Yield 72%, M.p. 270–272°С. 1H NMR (600 MHz, CDCl ), δ, ppm: δ 7.72 (t, 1H, 3J = 7.8 Hz, H-8), 7.86 (t, 1H, 3J = 7.8 Hz, H37), 7.99 (d, 1H, 3J = 5 Hz, H-4), 8.00 (d, 1H, 3J = 7.2 Hz, H-9), 8.19 (d, 1H, 3J = 7.8 Hz, H-6), 8.89 (d, 1H, 3J = 5 Hz, H-3), 9.62 (s, 1H, H-1). NMR 13C (150 MHz, CDCl3), δ, ppm: 122.4 (C-4), 123.6 (C-6), 125.3 (C-9), 134.0 (C-8), 137.4 (C-7), 137.7 (C-11a), 137.8 (C-9a), 140.7 (C5b), 146.5 (C-4a), 149.8 (C-3), 151.2 (C-10a), 155.8 (C-1), 160.4 (C-5a), 188.5 (C-10). IR bands, cm−1: 1728 (C=O), 1559, 1572, 1430, 1123, 744. Found, %: C 72.37, H 2.89, N 18.24. C14H7N3O. Calculated, %: C 72.10, H 3.03, N 18.02.</p><!><p>A mixture of 3a (0.41 g, 1.74 mmol) and hydroxylamine hydrochloride (0.30 g, 4.33 mmol) in EtOH (50 mL) was heated for 10 h (TLC monitoring) at 60 °C. The mixture was then cooled and poured into H2O (500 mL).The resulting precipitate was filtered, washed with water, and recrystallized from EtOH to give 7a (0.35 g, 80% yield) as a colorless solid. M.p. 291–293°С. 1H NMR (400 MHz, pyridine-d5), δ, ppm: 7.65 (td, 1H, 3J = 7.6 Hz, 4J = 1.2 Hz, H-8), 7.71 (td, 1H, 3J = 7.6 Hz, 4J = 1.2 Hz, H-3),8.06 (d, 1H, J = 5.6 Hz, H-9), 8.35 (d, 1H, 3J = 7.2 Hz, H-7), 8.92 (d, 1H, 3J = 5.6 Hz, H-10), 8.96 (d, 1H, 3J = 7.6 Hz, H-4), 9.78 (1H, H-2). NMR 13C (100 MHz, pyridine-d5), δ, ppm: 122.6,123.5, 129.6, 132.1, 133.7, 135.4, 138.0, 145.9, 147.9, 148.2, 153.6, 155.9, 157.5. IR bands, cm-1:1631 (C=N), 1604, 1575, 1470, 1381 (O–H), 1096, 905 (N–O), 776. Found, %: C 67.47, H 3.02, N 22.69. C14H8N4O. Calculated, %: C 67.74, H 3.25, N 22.57.</p><!><p>Compound 7b was obtained from 3b and hydroxylamine hydrochloride, as described under 4.1.1.15. Yield 56%,M.p. 296–298°C. 1H NMR (400 MHz, DMSO-d6), δ, ppm: 7.77 (m, 2H, H-9, H-8), 8.25 (dd, 1H,3J = 8 Hz, J = 4 Hz, H-10), 8.57 (dd, 1H, J = 8 Hz, J = 4 Hz, H-9), 8.77 (s, 1H, J = 2.8 Hz, H4), 9.13 (s, 1H, 4J = 2.4 Hz, H-2), 13.58 (s, 1H, OH). IR bands, cm−1: 1639 (C=N), 1559, 1572, 1475, 1371 (O–H), 1091, 951 (N–O), 784, 739 (C–Cl). Found, %: C 59.18, H 2.47, N 19.93. C14H7ClN4O. Calculated, %: C 62.82, H 2.26, N 15.70.</p><!><p>Compound 8 was obtained from 5 and hydroxylamine hydrochloride, as described under 4.1.1.15. Yield 78%, M.p. > 300°С. 1H NMR (400 MHz, pyridine-d5), δ, ppm: 7.66 (t, 1H, 3J = 5 Hz, H-8), 7.71 (t, 1H, 3J = 5 Hz, H-7), 8.05 (d, 1H, 3J = 5 Hz, H-4), 8.34 (d, 1H, 3J = 5 Hz, H-9), 8.88 (d, 1H, 3J = 5 Hz, H-6),8.91 (d, 1H, 3J = 5 Hz, H-3), 9.71 (s, 1H, H-1). NMR 13C (100 MHz, pyridine-d5), δ, ppm: 121.2,122.1, 128.2, 130.8, 132.4, 133.9, 136.5, 144.4, 146.4, 146.8, 152.2, 153.5, 156.0. IR bands, cm−1: 1631 (C=N), 1575, 1540, 1486 (C–N), 1362 (O–H), 1005, 956 (N–O), 814 (see Supplementary Figs. 6S and 7S for 1H NMR and 13C NMR spectra, respectively). Found, %: C 68.02, H 3.47, N 22.21. C14H8N4O. Calculated, %: C 67.74, H 3.25, N 22.57.</p><!><p>A mixture of tryptanthrin (2.48 g, 10 mmol) and hydroxylamine hydrochloride (2.09 g, 30 mmol) in 30 mL of pyridine was stirred at 60 °C for 2 h (TLC monitoring). The reaction mixture was poured into 300 mL of water and the resulting precipitate was filtered, washed with water, and dried to give 2.50 g (95 %) of a slightly yellow solid, m.p. 280–282 °C. NMR 1H (500 MHz, DMSO-d6), δ, ppm: 7.44 (td, 1H, 3J = 7.5 Hz, 4J = 1 Hz, H-8), 7.44 (td, 1H, 3J = 7.5 Hz, 4J = 1 Hz, H-9), 7.64 (td, 1H, 3J = 7.5 Hz, 4J = 1 Hz, H-2), 7.80 (d, 1H, 3J = 7.5 Hz, H-4), 7.87 (td, 1H, 3J = 7 Hz, 4J =1.5 Hz, H-3), 8.27 (dd, 1H, 3J = 8 Hz, 4J = 1.5 Hz, H-7), 8.35 (dd, 1H, 3J = 7.5 Hz, 4J = 0.5 Hz, H-1), 8.35 (d, 1H, 3J = 8 Hz, H-10), 13.63 (s, 1H, C=N–OH). NMR 13C (125 MHz, DMSO-d6) δ, ppm: 116.2 (C-10), 118.8 (C-6a), 121.5 (C-12a), 126.5 (C-8), 126.6 (C-7), 127.4 (C-9), 127.5 (C-1), 128.1 (C-4), 132.0 (C-2), 134.7 (C-3), 139.3 (C-10a), 144.2 (C-5a), 147.0 (C-4a), 148.3 (C-6), 158.5 (C-12). IR bands, cm−1: 3114, 1690 (C=N), 1592, 1448, 1354, 1326, 1268, 1227, 1196, 1127, 1084, 1037, 924, 774, 688, 662. Found, %: C 68.70, H 3.31, N 15.65. C15H9N3O2. Calculated, %: C 68.44, H 3.45, N 15.96.</p><!><p>A mixture of 1 (0.120 g, 0.52 mmol) and O-methylhydroxylamine hydrochloride (0.216 g, 2.6 mmol) in EtOH (10 ml) was heated for 9 h (TLC monitoring) at 78 °C. The mixture was then cooled and poured into water (100 ml). The resulting precipitate was filtered, washed with water, and recrystallized from EtOH to give 9a (0.120 g, yield 88%) as a colorless solid. M.p. 172–174°. 1H NMR (400 MHz, CDCl3), δ, ppm: 4.32 (s, 3H, CH3), 7.457.57 (m, 2H, H-2, H-3), 7.70–7.61 (m, 2H, H-7, H-8), 8.03 (d, 1H, 3J = 7.6 Hz, H-9), 8.11 (d, 1H, 3J = 7.2 Hz, H-6), 8.17 (d, 1H, J = 7.6 Hz, H-4), 8.34 (d, 1H, J = 7.2 Hz, H-1). Found, %: C 73.82, H 4.14, N 15.86. C16H11N3O. Calculated, %: C 73.55, H 4.24, N 16.08. The same procedure was used for the synthesis of the following oximes from corresponding O-substituted hydroxylamine hydrochlorides (see Scheme 1) and 1.</p><!><p>Yield 90%, a colorless solid. M.p. 169–170°. 1H NMR (400 MHz, CDCl3), δ, ppm: 1.46 (t, 3H, 3J = 7.2 Hz, CH3), 4.60 (q, 2H, 3J = 7.2 Hz, CH2), 7.47–7.57 (m, 2H, H-2, H-3), 7.61–7.70 (m, 2H, H-7, H-8), 8.04 (d, 1H, J = 7.6 Hz, H-9), 8.12 (d, 1H, 3J = 7.2 Hz, H-6), 8.19 (d, 1H, 3J = 8.2 Hz, H-4), 8.38 (d, 1H, 3J = 6.8 Hz, H-1). Found, %: C 74.29, H 4.65, N 15.01. C17H13N3O. Calculated, %: C 74.17, H 4.76, N 15.26.</p><!><p>Yield 92%, a colorless solid. M.p. 191–193o. 1H NMR (400 MHz, CDCl3), δ, ppm: 5.59 (s, 2H, CH2), 7.28–7.57 (m, 7H, H-2, H-3, C6H5), 7.62–7.71 (m, 2H, H-7, H-8), 8.05 (d, 1H, 3J = 8.4 Hz, H-9), 8.12 (d, 1H, 3J = 7.6 Hz, H-6), 8.19 (d, 1H, 3J = 7.6 Hz, H-4), 8.33 (d, 1H, 3J = 7.6 Hz, H-1). Found, %: C 78.63, H 4.26, N 12.71. C22H15N3O. Calculated, %: C 78.32, H 4.48, N 12.46.</p><!><p>Yield 91%, a colorless solid in the form of needles. M.p. 123–124o. 1H NMR (400 MHz, CDCl3), δ, ppm: 5.02 (d, 2H, 3J = 5.6 Hz, OCH2), 5.27 (d, 1H, 3J = 10.4 Hz, =CH2, Htrans), 5.40 (d, 1H, 3J = 17.2 Hz, =CH2, Hcis), 6.066.18 (m, 1H, CH=CH2), 7.40–7.53 (m, 2H, H-2, H-3), 7.58–7.67 (m, 2H, H-7, H-8), 8.00 (d, 1H, 3J = 8.4 Hz, H-9), 8.06 (d, 1H, J = 6.8 Hz, H-6), 8.15 (d, 1H, J = 8 Hz, H-4), 8.33 (d, 1H, 3J = 7.2 Hz, H-1). Found, %: C 75.32, H 4.39, N 14.51. C18H13N3O. Calculated, %: C 75.25, H 4.56, N 14.63.</p><!><p>To a suspension of IQ-1 (0.247 g, 1.0 mmol) and KOH (0.112 g, 2.0 mmol) in 5 ml DMSO, a solution of ethyl chloroacetate (0.183 g, 1.50 mmol, in 5 ml DMSO) was added dropwise. The mixture was stirred for 1 h at room temperature and poured into 150 ml of water. The precipitate was filtered out and recrystallized from EtOH to give 10c (0.28 g, 84% yield) as colorless crystals. Method B: Similar to Method A, but Na2CO3 (1.2:1.0 molar ratio to IQ-1) was used instead of KOH, and stirring was continued for 10 h. Yield 56%. M.p. 193–195o. 1H NMR (400 MHz, CDCl3), δ, ppm: 1.34 (t, 3H, 3J = 7.2 Hz, CH3),4.31 (q, 2H, 3J = 7.2 Hz, CH2CH3), 5.18 (s, 2H, =N-O-CH2), 7.59–7.69 (m, 2H, H-2, H-3), 7.70–7.80 (m, 2H, H-7, H-8), 8.14 (d, 1H, 3J = 8 Hz, H-9), 8.22 (d, 1H, 3J = 7.6 Hz, H-6), 8.27 (d, 1H,3J = 8 Hz, H-4), 8.58 (d, 1H, J = 7.6 Hz, H-1) (see Supplementary Figs. 4S and 5S for NMR H and 13C NMR spectra, respectively of E-isomer obtained by Method B). A mixture of Z- and Eisomers obtained by Method A gave additional signals in the 1H NMR spectrum: 1.27 (t, 3J = 7.2 Hz, CH3), 3.74 (q, 3J = 7.2 Hz, CH2CH3), 5.31 (s, =N-O-CH2). 13C NMR (100 MHz, CDCl3), δ, ppm: 14.24, 61.34, 72.71, 122.30, 129.41, 129.65, 129.94, 130.35, 130.54, 132.16, 132.59, 133.07, 137.40, 141.88, 142.72, 149.32, 150.37, 153.60, 168.93. Found, %: C 68.78, H 4.32, N 12.34. C19H15N3O3. Calculated, %: C 68.46, H 4.54, N 12.61.</p><!><p>Compound 10a was synthesized similarly to 10c (Method A) by reaction of IQ-1 with isobutyl bromide (54% yield, M.p. 143–146°). 1H NMR (400 MHz, CDCl3), δ, ppm: 1.09 (d, 6H, 3J = 7 Hz, CH3), 2.29 (m, 1H, CH(CH3)2), 4.41 (d, 2H, 3J = 5.5 Hz, CH2), 7.5–8.5 (m, 8H, Har). The product contained Z-isomer, which gave additional signals in 1H NMR spectrum: 1.13 (d, 3J = 7 Hz, CH3), 2.59 (m, CH(CH3)2), 5.09 (d, 3J = 5.5 Hz, CH2). Found, %: C 74.95, H 5.38, N 13.54. C19H17N3O. Calculated, %: C 75.23, H 5.65, N 13.85. Compounds 10b, d were synthesized similarly to 10c (Method B) by reaction of IQ-1 with 2-chloroethanol or chloroacetic acid, and the following derivatives were obtained, respectively.</p><!><p>Yield 52%, M.p. 230–232°. 1H NMR (400 MHz, CDCl3), δ, ppm: 4.72 (s, 2H, CH2), 7.66–7.76 (m, 2H, H-2, H-3), 7.79–7.87 (m, 2H, H-7, H-8), 8.13 (d, 1H, 3J = 6.8 Hz, H-9), 8.19 (d, 1H, 3J = 8 Hz, H-6), 8.53 (d, 1H, 3J = 7.8 Hz, H-4), 8.61 (d, 1H, 3J = 7.6 Hz, H-1). Found, %: C 67.04, H 3.41, N 13.48. C17H11N3O3. Calculated, %: C 66.88, H 3.63, N 13.76.</p><!><p>Yield 83%, M.p.194°, decomp. 1H NMR (400 MHz, CDCl3), δ, ppm: 4.24 (t, 2H, 3J = 4.4 Hz, CH2OH), 5.43 (t, 2H, 3J = 4.4 Hz, =N-O-CH2), 7.54–7.59 (m, 2H, H-2, H-3), 7.64–7.73 (m, 2H, H-7, H-8), 8.01 (d, 1H, 3J = 8.4 Hz, H-9), 8.07 (d, 1H, 3J = 8 Hz, H-6), 8.12–8.18 (m, 2H, H-1, H-4). Found, %: C 70.38, H 4.27, N 14.13. C17H13N3O2. Calculated, %: C 70.09, H 4.50, N 14.42.</p><!><p>Kinase profiling was performed by KINOMEscan (Eurofins Pharma Discovery, San Diego, CA, USA) using a panel of 97 protein kinases, as described previously [50, 51]. In brief, the kinases were produced and displayed on T7 phage or expressed in HEK-293 cells. Binding reactions were performed at room temperature for 1 h, and the fraction of kinase not bound to a test compound was determined by capture with an immobilized affinity ligand and quantified by quantitative polymerase chain reaction. Primary screening at fixed concentrations of compounds was performed in duplicate. Selected compounds were submitted for dissociation constant (Kd) determination using the same platform. For dissociation constant Kd determination, a 12-point half-log dilution series (a maximum concentration of 33 µM) was used. Assays were performed in duplicate, and their average mean value is displayed.</p><!><p>All cells were cultured at 37°C in a humidified atmosphere containing 5% CO2. THP1Blue cells obtained from InvivoGen (San Diego, CA, USA) were cultured in RPMI 1640 medium (Mediatech Inc., Herndon, VA, USA) supplemented with 10% (v/v) fetal bovine serum (FBS), 100 µg/ml streptomycin, 100 U/ml penicillin, 100 µg/ml phleomycin (Zeocin), and 10 µg/ml blasticidin S. Human monocyte-macrophage MonoMac-6 cells (Deutsche Sammlung von Mikroorganismen und Zellkulturen GmbH, Braunschweig, Germany) were grown in RPMI 1640 medium supplemented with 10% (v/v) FBS, 10 µg/ml bovine insulin, 100 µg/ml streptomycin, and 100 U/ml penicillin.</p><!><p>Activation of AP-1/NF-κB was measured using an alkaline phosphatase reporter gene assay in human monocytic THP1-Blue cells, which are stably transfected with a secreted embryonic alkaline phosphatase gene that is under the control of a promoter inducible by NFκB/AP-1. THP-1Blue cells (2 × 105 cells/well) were pretreated with test compound or DMSO for 30 min, followed by addition of 250 ng/ml LPS for 24 h, and alkaline phosphatase activity was measured in cell supernatants using QUANTI-Blue mix (InvivoGen) as absorbance at 655 nm and compared with positive control samples (LPS). For selected compounds, the concentrations of inhibitor that caused 50% inhibition of the NF-κB reporter activity (IC50) were calculated.</p><!><p>A human IL-6 ELISA kit (BD Biosciences, San Jose, CA, USA) was used to assess the effect of selected compounds on IL-6 production. MonoMac-6 cells were plated in 96-well plates at a density of 2 × 105 cells/well in culture medium supplemented with 3% (v/v) endotoxin-free FBS. Cells were pretreated with test compound or DMSO for 30 min, followed by addition of 250 ng/ml LPS for 24 h. IC50 for IL-6 production was calculated by plotting percentage inhibition against the logarithm of inhibitor concentration (at least five points).</p><!><p>Cytotoxicity was analyzed with a CellTiter-Glo Luminescent Cell Viability Assay Kit from Promega (Madison, WI, USA), according to the manufacturer's protocol. Cells were treated with compound under investigation and cultivated for 24 h. After treatment, the cells were allowed to equilibrate to room temperature for 30 min, substrate was added, and the samples were analyzed with a Fluoroscan Ascent FL (Thermo Fisher Scientific, Waltham, MA, USA). The cell IC50 was calculated by plotting percentage inhibition against the logarithm of inhibitor concentration (at least five points).</p><!><p>MonoMac-6 monocytic cells were pretreated with different concentrations of the compounds under investigation for 30 min and treated with LPS (250 ng/ml) or vehicle for another 30 min. Cells were washed twice with Hanks' balanced salt solution, and cell lysates were prepared using lysis buffer from the JNK kinase assay kit (Cell Signaling Technology, Danvers, MA). Cell lysates (from 5×106 cells) were separated on ExpressPlus 4–20% PAGE Gels (GenScript, Piscataway, NJ, USA) using TRIS-MOPS running buffer (GenScript) and transferred to nitrocellulose membranes. The blots were probed with antibodies against c-Jun, phospho-c-Jun (Ser73), and total c-Jun (Cell Signaling Technology, Danvers, MA, USA), followed by horseradish peroxidase-conjugated secondary antibody (Cell Signaling Technology). The blots were developed using SuperSignal West Femto chemiluminescent substrate (Thermo Fisher Scientific) and visualized with a FluorChem FC2 imaging system (Alpha Innotech Corporation, San Leandro, CA, USA). Quantitation of the chemiluminescent signal was performed using AlphaView software (ver. 3.0; Alpha Innotech).</p><!><p>Geometries of JNK1–3 proteins were obtained by downloading crystal structures from the Protein Data Bank (PDB entry codes 1UKI, 3NPC, and 1PMV for JNK1, JNK2, and JNK3, respectively) into Molegro software (Molegro ApS, Aarhus, Denmark). All solvent molecules were removed. Additionally, tryptanthrin and tryptanthrin-6-oxime molecules were docked into TRK-A binding site (PDB code 4AOJ). A search space was chosen for each of the receptors as a sphere centered on co-crystallized ligand present in the corresponding PDB structure. Radii of the spheres were equal to 8, 11, 10, and 10 Å for JNK1, JNK2, JNK3, and TRK-A binding sites, respectively. Each sphere completely encompassed the co-crystallized ligand and the binding site. Side chains of all amino acid residues of a receptor within the corresponding sphere were regarded as flexible during docking. The number of such residues was equal to 21, 31, 39, and 17 for 1UKI, 3NPC, 1PMV, and 4AOJ structures, respectively. The flexible residues were treated with default settings of "Setup Sidechain Flexibility" tool in Molegro, and a softening parameter of 0.7 was applied during flexible docking, according to the standard protocol using the Molegro Virtual Docker (MVD) program (MVD 2010.4.2).</p><p>Before docking, structures of compounds were pre-optimized using HyperChem software (HyperCube, Gainesville, FL) with the MM+ force field and saved in Tripos MOL2 format (Tripos, St. Louis, MO). The ligand structures were imported into MVD. The options "Create explicit hydrogens," "Assign charges (calculated by MVD)," and "Detect flexible torsions in ligands" were enabled during importing. Appropriate protonation states of the ligands were also automatically generated at this step. Each ligand was subjected to 30 docking runs with respect to a given receptor structure using MVD software. The docking poses obtained were saved together with the corresponding optimal geometries for identified flexible residues. DFT calculations were performed with the use of Gaussian 09W (Revision D.01) software.</p>

PubMed Author Manuscript

Population pharmacokinetics of fedratinib in patients with myelofibrosis, polycythemia vera, and essential thrombocythemia

PurposeFedratinib (SAR302503, TG101348) is an orally administered Janus kinase (JAK) 2-selective inhibitor that is being developed for the treatment of patients with myelofibrosis (MF). The objectives of this analysis were to develop a population pharmacokinetic (PK) model to characterize fedratinib concentration-time profiles in patients with MF, polycythemia vera (PV) and essential thrombocythemia (ET) following oral fedratinib administration; and to investigate the effects of selected covariates on fedratinib PK parameters.MethodsNonlinear mixed effects modeling was employed in developing a population PK model for fedratinib. Intensive or sparse fedratinib concentration data collected in adult subjects with MF, PV or ET from six studies were pooled, and a total of 452 subjects and 3442 plasma concentration observations were included in the final model.ResultsFedratinib PK in patients with MF/PV/ET was adequately described by a two-compartment structural PK model with first-order absorption incorporating a lag time and first-order elimination. Following oral administration, fedratinib undergoes biphasic disposition and exhibits linear, time-invariant PK at doses of 200 mg and above. Compared to MF/ET patients, PV patients had higher apparent clearance (CL/F) and apparent central volume of distribution. Creatinine clearance was a statistically significant covariate on CL/F, and patients with mild and moderate renal impairment had 10% and 37% increases in fedratinib exposure as compared to patients with normal renal function. No clinically meaningful effect on fedratinib exposure was observed regarding age, body weight, sex, race and liver function.ConclusionsThese results should serve as the basis for dose adjustment of fedratinib for special populations.Electronic supplementary materialThe online version of this article (10.1007/s00280-019-03929-9) contains supplementary material, which is available to authorized users.

population_pharmacokinetics_of_fedratinib_in_patients_with_myelofibrosis,_polycythemia_vera,_and_ess

Introduction<!>Clinical study data<!>Bioanalytical methods<!>Population pharmacokinetic analyses<!><!>Structural pharmacokinetic model characterization<!><!>Structural pharmacokinetic model characterization<!><!>Discussion<!>

<p>Myeloproliferative neoplasms (MPNs) are clonal, BCR-ABL1 negative hematopoietic diseases of myeloid proliferation, and characterized by abnormal production of terminally differentiated functional blood cells [1, 2]. MPNs are classically categorized into three disease entities: primary myelofibrosis (primary MF or PMF), polycythemia vera (PV) and essential thrombocythemia (ET) [1, 2]. Patients with PV and ET are characterized by an abnormal increase in hemoglobin/hematocrit and platelet count, respectively, and PMF is more advanced subtype of MPNs, associated with bone marrow fibrosis, release of profibrotic and proinflammatory cytokines and splenomegaly due to extramedullary hematopoiesis [3]. PV and ET may lead to secondary myelofibrosis (post-PV MF and post-ET MF, respectively) [1], which are clinically indistinguishable from PMF [4].</p><p>Janus kinase (JAK)/signal transducer and activation of transcription (STAT) pathway is key to cytokine receptor signaling and plays a critical role in hematopoiesis and immune response [5]. In human, the JAK family comprises four members: JAK1, JAK2, JAK3 and tyrosine kinase (TYK) 2, each of which associates with different cytokine receptors [6, 7]. Dysregulation of JAK-STAT pathway has been found in hematological malignancies and autoimmune diseases [5, 8, 9]. JAK2 V617F, which induces constitutive activation of STATs, is identified in 95% of patients with PV and 50–60% of patients with PMF and ET, and this is the most prevalent mutation in MPNs [2].</p><p>Fedratinib (SAR302503, TG101348) is an oral kinase inhibitor with activity against wild type and mutationally activated JAK2 and FMS-like tyrosine kinase 3 (FLT3) that is being developed for the treatment of intermediate or high-risk primary or secondary (post-PV or post-ET) MF [10–13]. Fedratinib is selective for JAK2 over JAK1, JAK3 and TYK2, and inhibits wild-type JAK2, activated mutant JAK2 V617F, and FLT3, with IC50 values of 3 nM, 3 nM and 15 nM, respectively [14]. Fedratinib significantly inhibits JAK2 V617F-driven aberrant human PV progenitor erythroid differentiation [15]. Pharmacokinetics (PK) of fedratinib has been characterized in both healthy subjects [16, 17] and patients with MF [11, 13]. Fedratinib was rapidly absorbed following oral administration with peak plasma concentration attained within 0.5–4 h [11, 13]. Fedratinib exposure increased in an approximately dose-proportional manner over dose range of 300–500 mg at steady state [13]. Plasma fedratinib levels reached steady state within 15 days of once daily dosing, with mean accumulation ratio of 2.95–3.88 at 300–500 mg [13]. Mean terminal half-life of fedratinib was 62–78 h at the single dose of 300–680 mg in healthy subjects [17].</p><p>This article, to the best of our knowledge, for the first time describes a population PK model that was developed to characterize fedratinib concentration-time profiles in patients with MF, PV, or ET following oral fedratinib administration. In addition, the effects of covariates on fedratinib PK were investigated.</p><!><p>The population PK analysis utilized data from one phase 1 study (TED12037 [NCT00631462]), four phase 2 studies (ARD11936 [NCT01420770], ARD12042 [NCT01420783], ARD12181 [NCT01523171], ARD12888 [NCT01692366]) and one phase 3 study (EFC12153 [NCT01437787]). Study design, dosing regimen, and PK sampling information are presented in Supplementary Table 1. These studies were conducted in accordance with the Declaration of Helsinki and the International Council for Harmonisation Guideline for Good Clinical Practice (ICH E6). Written informed consent was obtained from all subjects.</p><!><p>Concentrations of fedratinib in plasma were determined using a validated high-performance liquid chromatography with tandem mass spectrometric detection, with good accuracy (− 6.75–8.8%) and precision (4.84–13.11%). The lower limit of quantification was 1 ng/mL, and the calibration range was 1–1000 ng/mL.</p><!><p>Population PK analyses were conducted using a nonlinear mixed-effect modeling approach, as implemented in the NONMEM software version 7.3.0 (ICON Development Solutions, Ellicott City, MD). Plotting of NONMEM outputs was conducted using the R software (version 3.4.1) and RStudio (version 1.1.456, Boston, MA).</p><p>In the development of the structural model, one- to three-compartment models with first-order elimination and different absorption models including first-order absorption with and without lag time, zero-order absorption and transit compartment model were tested to fit the plasma concentration-time data. First-order conditional estimation (FOCE) with interaction method was used for parameter estimation, with natural logarithm-transformed plasma fedratinib concentration data. The inter-subject variability was modeled assuming a log-normal distribution. Residual variability was modeled using an additive model. The model selection was based on the objective function value (OFV) using the log-likelihood ratio test, the goodness of fit criteria and visual predictive check (VPC). Covariate model building was carried out using a stepwise procedure, with significance levels set to 0.01 and 0.001 for the forward inclusion and backward elimination steps, respectively. Missing baseline covariates were imputed as the median value in the study population.</p><p>Stability of the final PK parameter estimates and the 95% confidence interval (CI) for the parameters were evaluated using the nonparametric bootstrap approach. With this approach, 500 datasets of size equal to the original dataset were generated by random resampling with replacement from the original dataset. The final model was fit to each of the 500 bootstrap datasets and all the model parameters were estimated for each dataset. The median and nonparametric 95% CI (2.5–97.5 percentiles) of the 500 estimates were calculated for each parameter. The ability of the final population PK model to describe the observed concentration data was evaluated by simulating 200 datasets having the same doses, dosing schedules and sampling times as the original dataset and by performing prediction-corrected VPC [18]. The 5th, 50th and 95th prediction percentiles of the fedratinib concentrations at each binned time point were computed for each simulated trial. Thereafter, the nonparametric 90% CI of the 5th, 50th and 95th prediction percentiles at each binned time point were calculated for the 200 simulated trials. The data were displayed graphically and overlaid with the corresponding percentiles of the observed data.</p><!><p>Demographic and baseline characteristics of 452 patients with myelofibrosis, polycythemia vera or essential thrombocythemia</p><p>ECOG Eastern Cooperative Oncology Group, ET essential thrombocythemia, MF myelofibrosis, N number of subjects, NCI-ODWG National Cancer Institute Organ Dysfunction Working Group, PV polycythemia vera</p><!><p>To identify the structural model, a one-compartment PK model was compared with a two compartment PK model. The two-compartment model with first-order oral input was preferred over the one-compartment model with first-order oral input (∆OFV: − 1038). The two-compartment model was selected over the three-compartment model as the additional compartment did not improve the model fitting (no change in OFV). In addition, incorporating a lag time improved the model fitting by significantly decreasing OFV (∆OFV: − 37). The zero-order absorption model worsened the model fitting compared with the first-order absorption model (∆OFV: + 29). The transit compartment model was not pursued because the improvement of the goodness-of-fit plot and VPC were not observed compared to the model with lag time in spite of the improvement of OFV.</p><!><p>Box plot of apparent clearance (CL/F, a and c) and volume of distribution for central compartment (V2/F, b and d) of fedratinib by dose in patients with myelofibrosis and essential thrombocythemia (a and b) or by disease status in patients receiving 400 mg dose once daily (c and d). Individual estimates of CL/F and V2/F from the base model were overlaid on top of box plot, and the number of subjects at each dose level or disease status were listed below the box and whiskers. The dashed lines show typical values of CL/F (13.6 L/h) and V2/F plot (340 L) from the base model, respectively. ET essential thrombocythemia, MF myelofibrosis, PV polycythemia vera</p><!><p>Aside from doses, a visual examination of the dose-normalized concentration versus time profiles showed lower plasma fedratinib concentrations in PV patients than that in MF or ET patients. Individual CL/F and V2/F values from the base model confirmed that both CL/F and V2/F values were higher in PV patients than those in MF or ET patients (Fig. 1c and d). Therefore, the disease status of PV was subsequently added to both CL/F and V2/F as a categorical covariate in the covariate analysis.</p><!><p>Population pharmacokinetic parameter estimates of fedratinib from the final model</p><p>ALAG1 absorption lag time, CI confidence interval, CL/F apparent clearance, COV covariance, CLcr creatinine clearance, Ka absorption rate constant, PV polycythemia vera, Q/F apparent intercompartmental clearance, TV typical value, V2/F apparent volume of distribution of central compartment, V3/F apparent volume of distribution of peripheral compartment</p><p>aBootstrap confidence interval values are taken from bootstrap calculation (484 successful out of a total of 500 bootstrap replicates)</p><p>bCL/F (L/h) = 13.0 * 1.54(if PV) * (CLcr/78.3)0.294</p><p>cV2/F (L) = 311 * 1.87(if PV) * (Weight/70.1)0.727 * (Dose/400)− 0.279</p><p>Forest plot of significant covariates on apparent clearance (CL/F, a) and volume of distribution for central compartment (V2/F, b) of fedratinib. Data are shown as median (90% confidence interval). References are myelofibrosis/essential thrombocythemia (diagnosis), normal renal function (creatinine clearance [CLcr] ≥ 90 mL/min), 400 mg (dose) and second tertile (weight). First, second and third tertile of weight at baseline are 39.5 to 64.6 kg, 65.0 to 76.8 kg and 77.0 to 135 kg, respectively. Mild 60 ≤ CLcr < 90 mL/min; moderate 30 ≤ CLcr < 60 mL/min; severe 15 ≤ CLcr < 30 mL/min</p><p>Goodness of fit plots of the final population pharmacokinetic model of fedratinib. The blue line represents the identity line or zero line. The red line represents the locally weighted scatterplot smoothing line. CWRES conditional weighted residuals, DV observed value, IPRED individual predicted values, PRED predicted values, TAD time after last dose (hour), TIME time after first dose (hour)</p><p>Prediction-corrected visual predictive check for plasma fedratinib concentration-time profiles. Circles represent observed data. Lines represent the 5th (dashed), 50th (solid), and 95th (dashed) percentiles of the observed data. Shaded areas represent nonparametric 90% confidence intervals about the 5th (blue), 50th (pink), and 95th (blue) percentiles for the corresponding model-predicted percentiles</p><!><p>The fedratinib population PK model provided an adequate description of plasma fedratinib concentration-time data from MF, PV, or ET patients receiving oral fedratinib doses of 100 mg and above. Fedratinib concentration-time data were well characterized by the structural PK model that consists of a two-compartment with first-order absorption incorporating a lag time and first-order elimination.</p><p>Patients with PV had 87% higher apparent central volume of distribution and 54% higher apparent clearance compared to those in patients with MF or ET. Apparent clearance of a typical PV patient (CLcr = 78.5 mL/min) was 20.0 L/h, which falls into between CL/F of a typical MF patient (13.0 L/h) and CL/F of healthy subjects at doses of 300–680 mg (20.7–46.0 L/h) [11]. While patients with PV appear to have higher CL/F compared with patients with MF or ET, after progression to MF (post-PV MF), the CL/F appear to reach the similar level to that in patients with MF or ET. There were no apparent differences in CL/F and V2/F among primary MF, post-PV MF and post-ET MF, which is consistent with the clinical finding that both post-PV MF and post-ET MF are clinically indistinguishable from primary MF [4]. These results indicate that some MPN-related factor is associated with the reduced CL/F of fedratinib.</p><p>Patients with MF, PV, or ET appeared to have a comorbid condition of mild renal impairment as indicated by the lower CLcr (median 78.5 mL/min), although no patients with latent renal impairment were enrolled in the clinical studies. The renal function marker (CLcr) appeared to be statistically and positively correlated with CL/F of fedratinib, and the typical steady-state area under the concentration-time curve (AUC, dose divided by CL/F) for MF patients with mild (60 ≤ CLcr < 90 mL/min) and moderate (30 ≤ CLcr < 60 mL/min) renal impairment were 10% and 37% higher than that in MF patients with normal renal function (90 mL/min ≤ CLcr). Typical AUC for MF patients with severe renal impairment was 59% higher than that in MF patient with normal renal function, however, this should be interpreted with caution due to the small sample size (N = 3).</p><p>Fedratinib exhibited a greater than dose-proportional increase in exposure across a wide dose range in a phase 1 dose-escalation study in patients with MF (30–800 mg) [11] and an ascending single-dose study in healthy subjects (10–680 mg) [17]. In the base population PK model, the CL/F and V2/F were decreased with dose from 30 to 120 mg and remained dose-invariant at doses above 200 mg (Fig. 1a and b). The finding of more than dose-proportional increase of fedratinib exposure across wide dose range could be explained by larger distribution and/or elimination clearance at lower doses below 120 mg. Since the doses lower than 100 mg were deemed to be less efficacious and not studied beyond phase 2 studies, the final population PK modeling focused on the clinically relevant doses at 100 mg and above. In the covariate analysis, dose was statistically significant covariate on V2/F, however, the magnitude of changes in V2/F by dose was less than 30% and was considered not to be clinically meaningful. These results are consistent with the finding of a dose-proportional increase in fedratinib exposure over dose range of 300–500 mg at steady state [13].</p><p>Body weight (a range from 39.5 to 135 kg) was found to be statistically and positively correlated with V2/F. Large volume of distribution of fedratinib suggests that fedratinib may be distributed by diffusion into the extracellular fluids, the volume of which increases with body weight; thus, the estimated increases in the central volume of distribution of fedratinib with increased body weight are consistent with the physiological effects of weight. Given that the magnitude of changes in V2/F by body weight were less than 30%, and it does not affect AUC, body weight was deemed not to be a clinically relevant covariate.</p><p>In addition, no clinically meaningful effect on the PK of fedratinib was observed with regard to age (20 to 95 years), race, sex, mild hepatic impairment (defined as total bilirubin ≤ upper limit of normal [ULN] and aspartate aminotransferase [AST] > ULN or total bilirubin 1 to 1.5 times ULN and any AST) or moderate hepatic impairment (defined as total bilirubin > 1.5 to 3 times ULN and any AST), in the population PK analysis.</p><p>In summary, PK of fedratinib in patients with MF, PV, or ET was adequately described by a two-compartment model with first-order absorption incorporating a lag time and first-order elimination, and the fedratinib exposure increased linearly for doses 200 mg and above. The PV patients had 1.87-fold higher V2/F and 1.54-fold higher CL/F compared to that in MF or ET patients. Creatinine clearance was a statistically significant covariate on CL/F, and patients with mild and moderate renal impairment had 10% and 37% increase in fedratinib exposure, respectively, compared to patients with normal renal function. No clinically meaningful effect on fedratinib PK was observed with regard to other covariates such as body weight, age, race, sex, and mild and moderate hepatic impairment.</p><!><p>Supplementary material 1 (DOCX 15 kb)</p><p>Publisher's Note</p><p>Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.</p>

PubMed Open Access

Strong inhibitory activities and action modes of lipopeptides on lipase

AbstractLipopeptides have been reported to exhibit anti-obesity effects. In this study, we obtained a Bacillus velezensis strain FJAT-52631 that could coproduce iturins, fengycins, and surfactins. Results showed that the FJAT-52631 crude lipopeptide, purified fengycin, iturin, and surfactin standards exhibited strong inhibition activities against lipase with dose-dependence manners (half maximal inhibitory concentration (IC50) = 0.011, 0.005, 0.056, and 0.005 mg/mL, respectively). Moreover, fengycin and surfactin had the comparable activities with orlistat, but iturin not. It was revealed that the inhibition mechanism and type of the lipopeptides were reversible and competitive. The quenching mechanism of lipase was static and only one binding site between lipase and lipopoeptide was inferred from the fluorescence analysis. The docking analysis displayed that fengycin and surfactin could directly interact with the active amino acid residues (Ser or Asp) of lipase, but not with iturin. Our work suggests that the B. velezensis lipopeptides would have great potential to act as lipase inhibitors.

strong_inhibitory_activities_and_action_modes_of_lipopeptides_on_lipase

Introduction<!>Chemicals and strains<!>Lipopeptide extraction and preparation<!>Lipopeptide identification and separation<!>Measurement of lipase activity<!>Spectrum analysis<!>In silico simulating molecular docking model<!>Identification of the strain FJAT-52631<!>Identification and separation of lipopeptides produced by the strain B. velezensis FJAT-52631<!><!>Identification and separation of lipopeptides produced by the strain B. velezensis FJAT-52631<!><!>Inhibitory activity of lipopeptide acting on lipase<!><!>Inhibitory activity of lipopeptide acting on lipase<!>Inhibition mechanism and type of lipopeptide on lipase activity<!><!>Inhibition mechanism and type of lipopeptide on lipase activity<!><!>Spectrum analysis of product in the presence and absence of lipopeptide<!><!>Spectrum analysis of product in the presence and absence of lipopeptide<!><!>Molecular docking analysis<!><!>Molecular docking analysis<!>Conclusion<!>Disclosure statement

<p>The incidence of obesity has increased with astounding rapidity worldwide, rendering obesity a serious public health concern of the 21st century. Obesity is a complex, multifactorial disease that arises from the interaction of excessive caloric intake, sedentary lifestyle, metabolic disorder, and genetic predisposition1. Indeed, obesity is the risk factor responsible for various chronic metabolic diseases or syndromes including diabetes mellitus, hypertension, hyperlipidaemia, osteoarthritis, hepatic steatosis, and so on2,3.</p><p>Dietary and lifestyle modifications such as calorie restriction and physical exercises are the common strategies adopted to control body weight; further, these methods have limited anti-obesity effects4. It has previously been reported that lipase inhibition is a potential strategy for counteracting obesity. Digestive lipase hydrolyses non-absorbable dietary triglycerides to smaller absorbable molecules of monoglycerides and free fatty acids, which are absorbed by the intestine. Inhibiting digestive lipase can reduce intestinal fat absorption5–7. Human pancreatic lipase is the main enzyme in intestinal digestion of dietary fats in the human digestive system. To date, a wide variety of natural products have been used as pancreatic lipase inhibitors, which originate from plants and metabolites of microorganisms. These include lipstatin, panclicins, saponins, polyphenols, flavonoids, caffeine, chitin, chitosan, etc5. The lipase inhibitor orlistat is the only one obesity-treatment drug currently available in the market, which reduces intestinal fat absorption via inhibition of pancreatic lipase; however, it has been reported to cause certain side effects, e.g. oily stools, oily spotting, and flatulence1,8. Some polyphenol compounds have been reported to have potential adverse effects on microorganisms and animal at high concentrations9,10. Thus, there is still a need to explore safe and effective anti-obesity drugs.</p><p>Interestingly, surfactants were found to produce inhibitory effects on the lipolytic efficiency of lipase by generating inactive aqueous enzyme-surfactant complexes or by blocking the congregation of enzymes at the lipid/water interface11. In fact, it has been reported that bacterial cyclic lipopeptides are the most popular amphiphilic molecules that are excellent surface active compounds12. The genus Bacillus is described as an efficient source of lipopeptide biosurfactants. The Bacillus lipopeptides are divided into three different families, including iturins, surfactins, and fengycins, consisting of a cyclic lipoheptapeptide or decapeptide with a long hydrophobic fatty acid moiety13. Surfactin is a well-known surfactant consisting of a peptide ring of seven amino acids with a β-hydroxy-fatty-acid chain that can lower the surface tension of water from 72 to 27 mN/m14. In contrast to surfactin, iturin contains a β-amino fatty acid linked to a peptide ring with seven amino acid residues, while fengycin is a cycle lipopeptide with 10 amino acid residues. It has been reported that the lipopeptide biosurfactants exhibit numerous bioactivities, such as antimicrobial, antiadhesive, antitumoral, antiviral, and hypoglycaemic activities15–17. Additionally, the Bacillus lipopeptides possess high biodegradability, biocompatibility, and high stability towards extreme environments. These remarkable properties make lipopeptides potent candidate drugs for therapeutic medical applications18.</p><p>It had been reported that lipopeptides of Bacillus subtilis SPB1 could significantly reduce the body weight of obese rats and relieve hyperlipidaemia without apparent side effects18,19. The anti-obesity effects are mediated by lipopeptides through inhibiting the serum pancreatic lipase activity to modulate dietary triglyceride digestion18,19. However, the molecular mechanism of lipopeptide interaction with lipase needs further exploration. The B. subtilis SPB1 lipopeptides consist of iturins, surfactins, fengycins, and other lipopeptide isoforms. Moreover, the surfactins were speculated to be the major contributor to the anti-obesity effects of B. subtilis SPB1 lipopeptide. However, it is still unknown whether all the types of lipopeptides display the inhibition effect on the lipase. Considering the structural differences between different lipopeptide families, the comparative studies of lipase inhibition activities of surfactin, iturin, and fengycin would be important for the application of lipopeptide as lipase inhibitor. The aim of this article is to report a new lipopeptide-produced Bacillus velezensis strain FJAT-52631 that could coproduce iturin, fengycin, and surfactin and to evaluate the inhibition activity of each type of lipopeptide. Furthermore, the action modes of lipopeptide on lipase catalysis was carried out.</p><!><p>Lyophilised powder of Mucor miehei lipase (EC3.1.1.3), 4-Nitrophenyl palmitate (4-NPP), iturin, and surfactin were purchased from Sigma-Aldrich (St. Louis, MO). Acetonitrile, hydrochloric acid (HCl) and Tris were purchased from Sinopharm (Shanghai, China).</p><p>The strain FJAT-52631 (CCTCC No. M 2019760) was isolated from a soil sample from Wuyi Mountain, Fujian Province, China and it was identified through whole genome sequence analyses.</p><!><p>A single clone of the strain FJAT-52631 was inoculated in a 25-ml sterile tube with 5 ml liquid culture media (beef Extract 3 g/L, peptone 5 g/L, and glucose 10 g/L) and incubated for 25 h at 30 °C and 170 rpm. The pre-culture was inoculated (1%) into 250-mL flasks with a 50 mL potato dextrose broth culture medium and then cultivated for 48 h in a rotary shaker at 30 °C, 170 rpm. After fermentation, the cells were removed by centrifugation (6000 g for 5 min) and the lipopeptide in the culture supernatant was precipitated by adding 3 N HCl to achieve a final pH of 2. The precipitates were dissolved in a phosphate buffer and then lyophilised for anti-lipase activity tests and liquid chromatography quadrupole time-of-flight tandem mass spectrometry (LC-QTOF-MS/MS) analyses.</p><!><p>The qualitative and quantitative analyses of lipopeptides produced from the FJAT-52631 were carried out using the LC-QTOF-MS/MS method described in our previous studies20. Then, the lipopeptides were purified using the C18 solid phase extraction method with methanol/water (v/v) as an elution solvent. Each elution fraction was evaporated at a reduced pressure (−50 psig, 50 °C), dissolved in water and then lyophilised.</p><!><p>The lipase inhibition was determined according to the method described by Liu et al.21 10 µg/mL lipase in water and 7.5 mmol/L 4-NPP in acetonitrile solutions were prepared. The crude lipopeptide and purified fengycin were dissolved in water, while the iturin and surfactin standards were dissolved in methanol, and then all were diluted to their appropriate concentrations. The 1 mL reaction mixture contained 0.75 mM 4-NPP, 0.4 µg/mL lipase, and different concentrations of inhibitor in Tris-HCl buffer (pH 7.8). The reaction was carried out at 37 °C and the detection wavelength was set at 405 nm.</p><p>The inhibition mechanisms were studied by fixing the concentration of substrate and changing the lipopeptides and enzymes to monitor enzymatic reaction. The inhibition types were determined based on the Lineweaver-Burk plot22; this reaction system contained different concentrations of substrate and lipopeptide and 100 µL of lipase in Tris-HCl buffer. Then, the inhibition constant was calculated from a secondary plot of 1/Vm versus the inhibitor concentration.</p><!><p>Ultraviolet (UV) wavelength scanning spectra of the 4-NPP hydrolysis product was measured in the absence and in the presence of lipopeptide using a model 1510 spectrophotometer (Thermo Fischer Scientific, Waltham, MA). The 1 mL reaction mixture contained 0.75 mmol/L 4-NPP and 0.015 mg/mL of lipopeptide in Tris-HCl buffer (pH 7.8). The final concentration of the lipase was 0.4 µg/mL.</p><p>Fluorescence quenching spectra of the lipase and the lipopeptide were carried out according to the method described by Liu et al.21, using a Cary Eclipse spectrophotometer (Agilent, Santa Clara, CA). The Stern-Volmer quenching constant (Ksv) was calculated according to Equation (1): (1)F0=1+Ksv[l], where F and F0 are the fluorescence intensities with and without lipopeptide, respectively; further, [I] is the concentration of the lipopeptide. The binding constant (KA) and the binding affinity (n) were calculated according to the Equation (2): (2)lg[(F0−F)/F]=IgKA+nlg[l].</p><!><p>The interactions between lipopeptide and lipase were modelled using molecular operation environment software (MOE). The energy between the lipase and the ligand was minimised before docking. The docking parameters were set according to a previous study21.</p><!><p>A lipopeptide-produced strain FJAT-52631 was isolated from the soil sample, which was further accurately identified through the whole genome sequence analyses. The genome sequence of strain FJAT-52631 contains 3929791 bp (GenBank accession number: CP045186). The G + C content of the chromosomal DNA for strains FJAT-52631 was 46.5 mol %. The genome-based similarity calculated based on the OrthoANIu between strain FJAT-52631 and the type strain Bacillus velezensis CBMB205T was 99.95%. This value was above the threshold ANI value of 95–96% used for delineating prokaryotic species, suggesting that strain FJAT-52631 is a strain of the species B. velezensis23.</p><!><p>The lipopeptides produced from the strain FJAT-52631 were identified using the LC-QTOF-MS/MS method. Results revealed that three sets of homologue molecules with retention times in the range of 12.2–21.5, 27.0–38.7, and 45.2–54.0 min could be categorised as iturins, fengycins, and surfactins, respectively. Peptide sequences of each lipopeptide group were determined based on previous literature reports24–26. The retention times, MS and MS2 spectral data, and identification results were summarised, as shown in Table 1; the iturins consisted of C14–C16 iturin A; the surfactins consisted of C12–C16 surfactin A, and C16 surfactin A derivative; further, the fengycins consisted of C16/C18 fengycin A, C16 fengycin A2/B2, C16–C17 fengycin B, and C15 fengycin A/B derivatives. The iturin, fengycin, and surfactin content in the supernatant were calculated as 2.66 ± 1.50, 86.95 ± 4.08, and 35.93 ± 2.28 mg/L, respectively. The results demonstrated that fengycin was the most abundant lipopeptide family that was produced by the strain FJAT-52631.The lipopeptide production from Bacillus spp. firstly depends on itself. For example, the B. subtilis SPB1 strain have the ability to coproduce iturins, fengycins, and surfactins and highest content of surfactin in lipopeptide mixture was observed15.</p><!><p>Identification of lipopeptides produced by B. velezensis FJAT-52631 using LC-QTOF-MS/MS.</p><p>aMonounsaturated β-hydroxy fatty acid.</p><!><p>The crude lipopeptide from strain FJAT-52631 was further purified by solid-phase extraction method, and characterised by LC-TOF-MS analysis. Results show that the purified fengycin in the crude lipopeptide was obtained by elution with 80% methanol (Figure 1), while the iturin and surfactin in the crude lipopeptide were not successfully purified using the above method. LC-QTOF-MS analysis revealed that the composition of iturin and surfactin in the crude lipopeptide yield from FJAT-52631 were identical to the iturin and surfactin standards (Figure 2). Thus, iturin and surfactin standards were used to substitute the corresponding substances in further study.</p><!><p>The full scan LC-ESI-MS chromatogram of fraction eluted using 80% methanol.</p><p>The full scan LC-ESI-MS chromatogram of (a) iturin and (b) surfactin standards.</p><!><p>As reported in literatures, surfactants could inhibit lipase activity at certain concentrations due to their special property14. Lipopeptides from the Bacillus group have been proved to be excellent surfactants and could reduce lipase activity in plasma from alloxan-induced diabetic rats18. In spite of this, the manner of lipopeptide interaction with lipase and the mechanism of enzyme inhibition are still unknown. In the present study, we report the effects of crude lipopeptide (Figure 3(a)), purified fengycin (Figure 3(b)), iturin (Figure 3(c)), and surfactin (Figure 3(d)) standards on the lipase. Both the lipase from M. miehei and human pancreatic lipase are serine proteases, which possess the same active site and catalytic mode27. Moreover, the lipase from M. miehei exhibits much higher enzyme activity, better purity, and more mature processes than the human pancreatic lipase28. Thus, the M. miehei lipase was selected for use in the present study.</p><!><p>Lipase inhibitory activity of (a) crude lipopeptide, (b) purified fengycin, (c) iturin, and (d) surfactin.</p><!><p>Results showed that lipase inhibition was dose dependent. After adding these four inhibitors, the relative enzyme activity decreased significantly with increasing concentration of effectors. The half maximal inhibitory concentration (IC50) of crude lipopeptide, purified fengycin, iturin, and surfactin standards were 0.011, 0.005, 0.056, and 0.005 mg/mL, respectively. We have previously reported that furoic acid and oxalic acid inhibited lipase with IC50 of 0.242 and 1.425 mg/mL, respectively21. Moreover, we found that the inhibitory effects of the lipopeptides were within one order of magnitude of orlistat (IC50 = 0.004 mg/mL)7. These results suggest that lipopeptides have strong lipase inhibition activity and could act as lipase inhibitors to prevent obesity.</p><p>The anti-lipase activities of natural compounds depend on their structure characteristics, including number and position of hydroxyl groups, degree of polymerisation, elimination of glycosylation, and the size of the molecules6. Buchholz et al.6have reported that a high number of phenolic hydroxyl groups in active flavonoids increases their inhibitory effect. In present study, it was found that the fengycin and surfactin exhibit much stronger inhibition activities on lipase (about 10-fold) than that of iturin, which possibly be attributed to the fengycins and surfactins composing of β-hydroxy fatty acids, whereas the iturins carrying a β-amino fatty acid modification14,15. The crude lipopeptide produced by FJAT-52631 was composed of 2.7% iturin, 73.9% fengycin, and 23.4% surfactin, which indicated that the inhibition effect of crude lipopeptide produced by FJAT-52631 was mainly attribute to the fengycin. To our knowledge, this is the first report of the strong inhibitory activity of lipopeptide from B. velezensis against lipase in vitro.</p><!><p>The mechanism of enzyme inhibition by drugs can be either reversible or irreversible29. The inhibition mechanisms of crude lipopeptide, fengycin, and surfactin on lipase catalysis were studied. As shown in Figure 4, all the plots of the remaining enzyme activity versus the enzyme concentrations generated a series of lines intersecting at the origin, and whose slopes decreased with increasing lipopeptide content. These results demonstrated that the above catalytic action is reversible, which means that the lipopeptide declined the catalytic activity of lipase but did not lead to enzyme deactivation. Studies showed that most common mechanism of the inhibitor observed is reversible, such as furoic acid and oxalic acid, others like orlistat display irreversible effects21,29.</p><!><p>Inhibition mechanisms of (a) crude lipopeptide, (b) fengycin, and (c) surfactin on lipase.</p><!><p>Four inhibition methods have been proposed for enzyme inhibitors, including competitive, non-competitive, uncompetitive, and mixed type30. The inhibition type could be related to the properties of inhibitors, such as structure, molecular weight, etc.21. To determine the inhibition type of crude lipopeptide, fengycin, and surfactin on lipase catalysis, plots of 1/v versus 1/[S] of lipopeptide on lipase shown in Figure 5 were drawn, where v is the reaction rate and S is the substrate concentration. The inhibition constants KI of different lipopeptides were calculated and summarised in Table 2. Our results demonstrated that all the tested lipopeptides (mixture, fengycin, or surfactin) inhibit lipase in a competitive manner with respect to substrate concentration, which is subvert the explanation reported by Zouari et al.18, who deduced that B. subtilis SPB1 lipopeptide biosurfactant exhibits an uncompetitive inhibition mode action on lipase activity. Our above finding suggested that the lipopeptide binds to the free enzyme at the substrate-binding site.</p><!><p>Inhibition types of (a), crude lipopeptide, line 5-1: mean 0, 0.0013, 0.002, 0.0025, 0.003 mg/mL, respectively; (b) fengycin, line 5-1: mean 0, 0.0013, 0.002, 0.0025, 0.003 mg/mL, respectively; and (c) surfactin, line 5-1: mean 0, 0.0008, 0.0009, 0.0011, 0.0013 mg/mL, respectively on lipase.</p><p>Inhibition effects of lipopeptides for lipase.</p><!><p>The UV spectra of lipase catalysis in the absence (Figure 6(a)) and presence (Figure 6(b)) of the crude lipopeptide were measured. Results show that the typical peak absorbent intensity of the product at 405 nm increased along with an extension of the catalytic reaction time. After 10 min, in the presence of 0.015 mg/mL lipopeptide, the peak absorbance was reduced to 54.5%, which indicates that the crude lipopeptide produced from FJAT-52631 was a stronger inhibitor. This result was consistent with the enzyme activity assay.</p><!><p>Consecutive spectra obtained during the lipase catalysis in the (a) absence and (b) presence of crude lipopeptide. Curves 1–10 depict the addition of the enzyme in 10 min.</p><!><p>The fluorescence emission spectra of crude lipopeptides on lipase are shown in Figure 7. Results showed that the fluorescence intensities of the emission peak at 338 nm gradually declined with increasing lipopeptide content (Figure 7(a,b)); this indicated that the lipopeptide and the lipase form a complex21,31. The effects of different concentrations of lipopeptides on lipase did not lead the endogenous fluorescence peak to red shift or blue shift; this suggests that there has not been a complete change of the composition of the enzyme molecule and that the enzyme is still active32. The plot of F0/F versus [I] was drawn and a linear regression equation of F0/F = 1 + 0.525[I] was obtained (Figure 7(c)). The value of Stern-Volmer quenching constant Ksv was found to be 0.525 L/g according to the Stern-Volmer equation. This value is larger than the maximum dynamic quenching constant of biomacromolecule (100 L/mol); this indicated that the fluorescence quenching mechanism of lipase is static when the lipopeptide is present33. Further, a linear regression equation lg[(F0−F)/F] = 1.429lg[I] − 0.108 (R2=0.9255) was obtained from Figure 7(d). The values of the binding constant KA and binding site n from the intercept of the above equation and the slope were calculated as 1.429 L/g and 0.78, respectively. This result indicates that the interaction between lipase and lipopeptides is quite intensive and there is only one binding site between them33.</p><!><p>Effect of lipopeptides on the emission spectrum of lipase. (a) Emission spectra of lipase with crude lipopeptide concentration of 0, 0.194, 0.265, 0.324, 0.375, 0.419, and 0.457 mg/mL, respectively (1–7, respectively); (b) Florescence intensity changes with lipopeptide; (c) Plot of F0−F against [I] for lipopeptide; (d) Plot of lg[(F0−F)/F] versus lg[I] for lipopeptide.</p><!><p>The M. miehei lipase have three active site residues Ser, His, and Asp that forming an oxygen negative ion hole, which are covered by an "alpha-screw lid"34,35. The lipase was composed of hydrophilic groups surrounded by exposed hydrophobic and its activity depends upon the conformation of lid (in open or in closed)35. The conformation of lipase would be altered when the active site of lipase interacted with an inhibitor, leading to the reduced catalytic activity of lipase36. Previous studies have shown that the ring structure and the carbonyl groups of the inhibitor is important for their correct binding in the active site of lipase37. The molecular docking analysis was carried out to explore the possible binding action mode between lipopeptide and lipase. Results demonstrated that the carbonyl group on the Tyr residue in iturin A might directly act on Gly 5 of lipase (Figure 8(a)); the carbonyl group on the Leu residue in surfactin A might directly act on Asp 238 and Ser 237 of lipase (Figure 8(c)); and the oxygen atoms of the hydroxyl group on the Thr and Glu residues in fengycin might directly act on Ser 259 and Asn 100 of lipase (Figure 8(b)), respectively. The hydroxyl and carbonyl groups on the lipopeptides formed hydrogen bonds with the amino residues of lipase to stabilise the interactions between the lipopeptide and the lipase. This manner could block substrate access to the catalytic active site of the lipase enzyme. Our results displayed that the fengycin and surfactin could be in direct contact with the active amino acids of lipase, and not those of iturin. The interactions of oxygen with the catalytic site might enhance the lipase inhibition activity. Moreover, the active site residue Ser is essential for the hydrolytic activity of the enzyme6. This could explain in part the weaker potency of iturin. The above modelling results support the data derived from the enzymology studies.</p><!><p>Interactions between key amino acids in (a) lipase and iturin, (b) fengycin, or (c) surfactin were investigated by in silico modelling.</p><!><p>Studies showed that the small molecules are easily to enter the gap in the oxygen negative ion to interact with the catalytic site of the enzyme. For example, the furoic acid and oxalic acid can directly act on active amino acid of Ser, while the big molecular orlistat does not contact with the active amino acids of, Ser, His or Asp21. Lipopeptides are the famous surfactants that composed of both hydrophilic and hydrophobic groups. The structures of surfactants play an important role in protein–surfactant interactions35. This suggested that lipopeptides lead to strong inhibition of M. miehei lipase activity by direct interact with the active site, which are closely related to the amphiphilic molecular structures of lipopeptides.</p><!><p>We report here on the strong inhibitory activity and action modes of lipopeptide from B. velezensis against lipase in vitro. The lipopeptides produced from B. velezensis FJAT-52631 were categorised as iturin, fengycin, and surfactin. The crude lipopeptides, iturin, fengycin, and surfactin exhibited strong inhibition activities against lipase with dose-dependence. The lipopeptide inhibition catalytic reaction was reversible and the inhibition type was competitive. The fengycins and surfactins directly interact with the residues of the enzyme's active centre, while iturins do not. Fengycins were the most abundant compounds among the crude lipopeptides, which were considered to contribute significantly to the inhibited activities against lipase. Our results suggest that these lipopeptides could be used directly as lipase inhibitors and the in vivo anti-obesity effects of lipopeptides will be explored in further studies.</p><!><p>No potential conflict of interest was reported by the author(s).</p>

PubMed Open Access

Biomimetic CO oxidation below -100 ⁰C by a nitrate-containing metal-free microporous system

CO oxidation is of importance both for organic and inorganic systems. Transition and precious metals on various supports can oxidize CO to CO2. Among them, few systems, like Au/TiO2, can perform CO oxidation at the low temperature of -70 ⁰C. Living (an)aerobic organisms perform CO oxidation with nitrate using complex enzymes under ambient temperatures which is an important pathway of their living cycle that enables them to "breathe"/produce energy in the absence of oxygen and leads to the carbonate mineral formation. Herein, we report that CO can be oxidized to CO2 by nitrate at -140 ⁰C in completely inorganic system (zeolite) without metals. The transformation of NOx and CO species in zeolite as well as the origin of this unique activity (catalyzed by Bronsted acid sites) are clarified using spectroscopic and computational approach.

biomimetic_co_oxidation_below_-100_⁰c_by_a_nitrate-containing_metal-free_microporous_system

<!>Supplementary Materials:<!>Computational Details and Models<!>Table S1

<p>CO oxidation is important both for automotive emissions control and in living microorganisms [1][2][3][4][5][6][7][8]. More specifically, inorganic materials such as transition/noble metals on solid supports are capable of oxidizing CO at elevated temperatures [1,2]. Among such systems, Au nanoparticles supported on titania, discovered by Haruta, represents a material active for CO oxidation at the lowest known temperature of -70 ⁰C [1]. In anaerobic and aerobic microorganisms, enzymes evolved to oxidize CO to CO2 by nitrate as shown in the pioneering studies of King and co-workers [3][4][5]8]. The energy produced is used to sustain life while emitted CO2 leads to formation of carbonate minerals, for example [3][4][5][6][7]. Moreover, search for carbonate minerals/CO2 on other planets (Mars) is ongoing to possibly confirm whether anaerobic life was ever present on such planets [5,8].</p><p>We discovered a pathway of CO oxidation by nitrate in a completely inorganic, non-metal crystalline system at -140 ⁰ C, more specifically zeolite SSZ-13. First, we had to clarify the chemistry of NOx species in zeolite under different conditions. In the 1980s, it was first discovered that NO + species can form in Na-zeolites upon interaction with (NO+O2) or NO2 using Raman and infra-red spectroscopy [9]. Later on, K. Hadjiivanov and co-workers' pioneering studies [10][11][12] allowed to establish, using FTIR spectroscopy and isotopic methods on NxOy molecules, that NO + can indeed form in ZSM-5 upon interaction with (NO+O2) mixture. NO2 was shown to disproportionate to NO + and NO3in the zeolites in Li/K/Na forms as well [8]. Although it was clear that NO + can be produced in zeolites from NO+O2 mixtures and specific mechanisms were proposed, we investigated its production and chemical properties using spectroscopy and density functional theory calculations.</p><p>We started by introducing NO2 onto small pore H-SSZ-13 zeolites with Si/Al ratios 12 and 6. SSZ-13 zeolite was chosen because it is the one often encountered in nature [13], it is a robust, hydrothermally stable framework used extensively in catalytic aftertreatment systems to decrease automotive pollution [14][15][16][17]. Additionally, it is the framework with only one equivalent framework T site [15] which is more straightforward to model than frameworks with broad distribution and location of T-sites. cm -1 (belonging to NO + ) and one with a maximum at ~1650 cm -1 (belonging to free HNO3). This corresponds to disproportionation reaction of NO2 to NO and HNO3: 2NO2 + H-zeolite  Zeolite-NO + + HNO3 Indeed, NO2 is known to be easily dimerizable to N2O4 which, in highly polar solvents (such as ethylacetate, for example) or in the highly polar zeolitic micropores favors disproportionation to</p><p>, which can react with zeolite protons (acid denoted as H-zeolite) in a scheme depicted above. The band representing NO + is not symmetrical and it corresponds to two NO + stretches at ~2195 and ~2171 cm -1 . Concomitant to the development of these two absorption bands is the gradual intensity decrease of the two OH stretching features at 3612 and 3585 cm -1 (Fig. 2B). Thus, NO + replaces protons near at least two distinct Al atom environments. Concurrently, a new OH stretching band appears at 3667 cm -1 , corresponding the OH stretch of nitric acid. Thus, NO2 reacts with zeolite to form NO + near the framework oxygens (in place of acidic protons) and HNO3.</p><p>Similar chemistry is observed for H-SSZ-13 with a different Si/Al ratio ~ 6 with slightly different distribution with NO + species formed (see difference spectra in SI Fig. S1, HAADF-STEM images of SSZ-13 crystals are shown in Fig. S2). This is fully consistent with the DFT calculations, which</p><p>show that the reaction: Zeo/H + + 2×NO2(g) → Zeo/NO + + HNO3(g) is exothermic by -46 kJ/mol.</p><p>Conceptually, NO + can bind with an Xanion in two fundamentally different ways to form NOX complexes. 1).</p><p>[N≡O] + complexes with weakly coordinating anions, like BF4 -. In these complexes NO + behaves as a free (or semi-free) cation exhibiting an N-O stretching vibrational frequency of ≥2,300 cm -1 , for example 2340 cm -1 in NO[BF4], 2326 cm -1 in NO[AuF6], and 2298 cm -1 in concentrated sulfuric acid solutions for free [NO] + [20]. 2). NO + forms complexes with X-but a covalent bond between N and X (X is most often a halogen) is preserved: in this case a bent O=N-X molecule forms with an O-N-X angle <180⁰. Complexes like these exhibit N-O vibrational frequencies between 1,950 and 1,800 cm -1 [21]. Furthermore, even for non-halogens the bent nonionic M-N=O moieties with fully covalent N-M bonds and bent Metal-N-O angle ( 120<angle<180) show νNO<1,900 cm -1 in the IR spectra [22]. DFT calculations for free NO + and NOX complexes are in good agreement with the experimental trend (Table S1). However, for the NO + in zeolite we observe νNO in the intermediate region between 2,300 and 1,950 (but closer to the former). This does indeed suggest that NO + and -O-Zeolite interaction has a significant covalent character (i.e., NO + is not a free-floating ion in zeolite). Further confirmation of this is the observation of the same FWHM of NO + band upon warming NO + /SSZ-13 system from 77 to 298 K (the NO + band FWHM does not change, Fig. S3) suggesting that NO + structure is attached to the zeolite and remains stable with temperature change (no shifting or broadening upon significant temperature changes). Unlike the 2133 cm -1 NO + band in ZSM-5 [11], the NO + bands in SSZ-13 are located at higher wavenumbers (at ~2170 and 2200 cm -1 ) and are characterized by higher thermal stability up to 200 ⁰C under high vacuum (Fig. S4) in contrast to NO + in H/ZSM-5 located at 2133 cm-1 and which starts desorbing above room temperture.</p><p>Adsorption of excess NO2 on NO + /zeolite with Si/Al ~ 6 (Fig. S5) and 12 (Fig. 1C) leads to the decrease of the intensity of the NO + stretching band in the expense of a new NO band at ~2230 cm -1 (lying higher than the original NO + bands by ~35 and ~58 cm -1 correspondingly).</p><p>Simultaneously, two new vibrational features appear,: a sharp band at ~1740 cm -1 and a shoulder at ~2080 cm -1 (clearly seen in the difference spectra and spectra upon evacuation, Fig. S5, S8) as the excess NO2 produces an NO + -NO2 complex. The formation of such a complex has been suggested, on the basis of Rietveld refinement of synchrotron XRD data, for NO + interacting with excess NO2 in the supercages of Ba-FAU zeolite [19]. The onset of the formation of this complex coincides with the appearance of the band at 1746 cm -1 [18]. DFT calculations (Fig. S10) show the N-O stretches of NO + -NO2 complex lie at ~2042 and 1722 cm -1 . Considering the systematic shift of calculated DFT NO stretches relative to experimental values by 20-30 cm -1 (due to the non-innocent nature of NO adsorption, unlike in the CO case) this agrees well with ~2080 and ~1746 cm -1 bands for the NO + -NO2 complex. The nature of the 2,230 cm -1 band is discussed below.</p><p>The chemistry we observe when reacting a mixture of NO+O2 over H-zeolites is somewhat different from the case discussed above for NO2. It was shown in the 1980s that NO+O2 in microporous materials can easily produce NO2 [25]. The thus formed NO2, in turn, can either dimerize to form N2O4 or react with NO to form N2O3. Due to the presence of excess NO in the system the primary reaction is N2O3 formation. The thus formed N2O3 can disproportionate to the ion pair NO + NO2that, in turn, can react with a zeolitic proton. Indeed, when we mix NO with sub-stoichiometric amounts on oxygen, we see the immediate appearance of the IR bands characteristic of both NO + (2174 cm -1 ) and N2O3 (1570 and 1970 cm -1 ) (Fig. S6). The intensities of the IR bands of N2O3 initially grow rapidly, and then, after reaching their maxima, they begin to lose their intensities. The IR feature of adsorbed NO + continuously grows during the entire experiment (Fig. S6) while Bronsted acidic protons are consumed:</p><p>HNO2 then quickly reacts with H + to form NO + and H2O. The major difference between the two NO + formation processes (NO2 vs. NO+O2) is the production of nitrates in the NO2 only process, and the formation of H2O in the NO+O2 reaction. NO + , formed in either processes, is stable under vacuum at 77 K and only desorbs above 150 ⁰C under high vacuum. Furthermore, the chemistry of NO + confined in zeolite is peculiar. Adsorption of NO2 at room temperature leads to the formation of NO + -NO2 complex as well as a shift of the NO + band to 2230 cm -1 . Evacuation restores the IR signature of the original NO + as the [NO + -NO2] complex decomposes (Figs. S7 and S8). Adsorption of CO at low temperature (77 K) on the NO + /H/SSZ-13 shows the production of an NO + -CO complex evidenced by the blue shift of the NO + vibrational signature to ~2220 cm -1 (vide infra) (Fig. 3A). In contrast, upon the adsorption of NO on the NO + /H/SSZ-13 sample at 77 K the IR band of NO + redshifts to 2013 cm -1 evidencing the selective production of the NO + -NO complex.</p><p>Thus, this novel chemistry provides a hitherto unknown insight how adsorption of an adsorbate (NO, CO, NO2) changes the properties of the cation with which it interacts. Normally, such cations are metal cations that have no corresponding IR active vibrations, and the information regarding the changes incurred during adsorption is hidden. In our case, however, because NO + itself has an active N-O vibration, we see that adsorption of CO, NO and NO2 shifts its electronic signature, our DFT calculations showed that the interaction between NO + and CO or NO2 is relatively weak with a binding energy of the adsorbed molecule below -20 kJ/mol in absolute value (Table S1). Hence, we conclude that since the concentration of CO or NO2 is high in the zeolite pores, these gas phase molecules slightly shift NO + further from its equilibrium position in the zeolite. This, in turn, weakens the interaction between NO + and zeolite, leading to a shortening in the N-O distance and a blue shift of N-O vibrational frequency to ~2230 cm -1 (in the case of NO2) and 2220 cm -1 (in the case of CO). This provides a unique insight into the interaction of extra-framework cations with adsorbates, not routinely available even from the most sophisticated synchrotron XRD and Rietveld refinement methods.</p><p>However, in the case of NO, the shift is to a significantly lower frequency. This finding can be rationalized by our DFT results, since in this case a stable Zeo/NO + -NO complex is formed (Fig. S10) in the zeolite with a binding energy of NO to Zeo/NO + of -52 kJ/mol (Table S1). This structure has two frequencies at 2009 and 1911 cm -1 . They can rationalize the experimental bands at 2013 and 1871 cm -1 . In addition, calculated ONNO species in gas phase has frequencies at 1879 and 1727 cm -1 . Thus, they can rationalize the experimental bands at 1870 and 1685 cm -1 .</p><p>In-situ heating of the NO-CO complex produced from N2O3 reaction with H-SSZ-13 leads to no CO2 formation (Fig. S9). However, when the NO + /SSZ-13 sample prepared by the disproportionation of NO2 (i.e., contained large amounts of nitrates) was heated from -170 to -140</p><p>⁰C in the presence of CO, the immediate formation of CO2 inside the zeolite micropores was observed, evidenced by the appearance of a sharp band at 2345 cm -1 characteristic signature of adsorbed CO2 [23]. CO was oxidized by nitrates: as the intensity of the characteristic vibrational feature of adsorbed CO2 gradually increased, the intensity of the 1645 cm -1 nitrate band simultaneously decreased. These results unambiguously show that CO can be oxidized by nitrates in this zeolite at the very low temperature of -140 ⁰C ( Fig. 3). This biomimetic chemistry by a completely inorganic non-metal system occurs at temperatures previously unseen for such a conversion. Moreover, when we react CO with NO3at room temperature (in the same system), no reaction takes place at all! How can this seemingly puzzling fact be rationalized? It is well-known the catalysis can occur when the reacting molecule is adsorbed/chemisorbed. In the case of CO, at room temperature CO is not adsorbed by the Bronsted acid protons of -Si-OH-Al groups, as evidenced by the lack of CO stretches other than gas-phase CO. However, CO interacts with Bronsted acid protons of the zeolite at lower temperatures forming -H + ---CO complex [25]. IR CO stretching feature in this complex is centered at ~2175 cm -1 [25]. The major consequence of the binding of CO to H + is the polarization of the C-O bond as C (δ+) -O (δ-) because in the H + -CO complex no backdonation from the proton to CO is present, and only electrostatic interaction and formation of a sigma bond takes place with charge transfer from C to H + . This polarization (i.e., formation of C (δ+) -O (δ-) ) makes CO susceptible to nucleophilic attack by NO3to form CO2 and reduced NOx species. DFT calculations further support the proposed route. We considered four mechanisms for CO oxidation by HNO3 in H-CHA. They are presented schematically in Fig. 4, as well as the corresponding energetic diagrams. Shorter names are used for the structures in the energy diagrams with respect to the schemes above, as: IS -Zeo1/CO/HNO3; FS -Zeo1/CO2/HNO2; IC -Zeo2/NO2 + /HCO2 -; ID -Zeo2/HCO2 -/NO2 + , FS2 -Zeo2/CO2/HNO2. Color coding: Si -blue, O -red, Al -green, Nlight blue, C -brown, H -white.</p><p>All mechanisms start with the adsorption of the reactants, CO and HNO3, via hydrogen bonds to the bridging OH group and basic zeolite oxygen center, respectively. In this initial state (IS) structure (Zeo1/CO/HNO3) the C-O vibrational frequency was calculated to be 2175 cm -1 , in line with the corresponding experimental band. Mechanisms A and B are one-step mechanisms. In the former one the CO molecule is oxidized by the HNO3 via transition state structure TS1 with no direct participation of the zeolite. In mechanism B, the CO molecule is oxidized to CO2 via transition state structure TS2, where zeolite participates actively in the process via movement of the proton from the bridging OH group towards C atom from the CO molecule and via the basic zeolite oxygen, which attracts the nitric acid's proton. TS2 structure seems to correspond to a bifurcation point, since it may be decomposed in two ways (i) directly to CO2 and HNO2 via the electron transfer shown by the arrows in the mechanism B; and (ii) to formation of a complex HCOO -NO2 + (IC) as the zeolite proton interacts with an O center from NO2 + , as shown in mechanism C. From the latter intermediate can be formed the final products (CO2 and HNO2) via the same TS2 structure (Mechanism C). Alternatively, if the intermediate is coordinated to the zeolite (see Zeo2/HCO2 -/NO2 + , ID) CO2 and HNO2 can be formed via TS3, which includes a Htransfer from HCOOto NO2 + (Mechanism D). In all mechanisms considered the CO oxidation was calculated to be exothermic by 187 to 217 kJ/mol, depending on the coordination of the products (Figure 4), thus we investigated only kinetic aspects in depth. Energy diagrams show that the most plausible are Mechanisms B and C with a barrier of the rate limiting steps (which are the same) of 81 kJ/mol. In addition, we calculated the Gibbs free energy barriers of this rate limiting step of both mechanisms at 140 K to be 58 kJ/mol. Since these steps are the first ones, formation of an intermediate is not expected in line with our experimental results. Both mechanisms involve the direct participation of the zeolite via H + transfers and as well as the formation of initial complex of CO to the zeolite H + is crucial for the oxidation process. The calculated energy barriers for the rate limiting steps of mechanisms A and D are notably higher (Figure 4).</p><p>We also considered CO oxidation on Zeo/NO2 + (Fig. S11), in which NO2 + are the chargecompensating species of the negatively charged zeolite framework, forming kind of nitrate with the O center from the zeolite. The calculated IR frequencies of such Zeo/NO2 + species are 2046 and 1358 cm -1 and move to 2055 and 1353 cm -1 after CO adsorption. However, the barrier for the reaction: Zeo/NO2 + + CO → Zeo/NO + + CO2 is very high, 143 kJ/mol, thus the role of NO2 + species as a CO oxidizing agent can be discarded.</p><p>In summary, we discovered that CO can be oxidized to CO2 by nitrate in zeolite micropores at temperatures as low as -140 ⁰C. This reaction was previously known to be catalyzed by complex enzyme molecules in living (an)aerobic organisms. However, no noble (or other) metal systems have been known that can perform this biologically important reaction under mild conditions.</p><p>Remarkably, fully inorganic system with no (noble) metals performs such a reaction at -140 ⁰C.</p><p>Interaction of CO with Bronsted acid protons in confined nanopore produces -H + -CO complex making the carbon atoms susceptible for nucleophilic attack by a nitrate, revealing a hitherto unknown pathway for CO conversion chemistry in inorganic systems at low temperatures.</p><!><p>Materials and Methods H-SSZ-13 was synthesized using previously described methods [14,16 in the main text]. More specifically, a Na-form of SSZ-13 with Si/Al ratio ~6 and 10-12 were prepared, and then exchanged three times with 2 M ammonium nitrate solution at 80 ⁰C. The powder was purified by consecutive centrifugation cycles after washing with DI. The wet powder was dried at 100 ⁰ C, and then calcined at 550 ⁰C for 5 hours in the flow of dry air.</p><p>All the chemicals used were the highest-grade purity available.</p><p>The in situ static transmission IR experiments were conducted in a home-built cell housed in the sample compartment of a Bruker Vertex 80 spectrometer, equipped with an MCT detector and operated at 4 cm -1 resolution. The powder sample was pressed onto a tungsten mesh which, in turn, was mounted onto a copper heating assembly attached to a ceramic feedthrough. The sample could be resistively heated, and the sample temperature was monitored by a thermocouple spot welded onto the top center of the W grid. The cold finger on the glass bulb containing CO was cooled with liquid nitrogen to eliminate any contamination originating from metal carbonyls, while NO was cleaned with multiple freeze-pump-thaw cycles. Prior to spectrum collection, a background with the activated (annealed, reduced or oxidized) sample in the IR beam was collected. Each spectrum reported is obtained by averaging 64 scans.</p><p>HAADF-STEM was performed with a FEI Titan 80-300 microscope operated at 300 kV. The instrument is equipped with a CEOS GmbH double-hexapole aberration corrector for the probeforming lens, which allows for imaging with 0.1 nm resolution in scanning transmission electron microscopy mode (STEM). The images were acquired with a high angle annular dark field (HAADF) detector with inner collection angle set to 52 mrad.</p><!><p>Periodic DFT calculations were performed using the Perdew-Burke-Ernzerhof (PBE), [1] exchange-correlation functional with the additional empirical dispersion correction as proposed by Grimme (DFT-D2), [2] as implemented in Vienna ab initio simulation package (VASP). [3,4]. We also employed PAW pseudopotentials [5,6] and the valence wave functions were expanded in a plane-wave basis with a cutoff energy of 415 eV. The Brillouin zone was sampled using only the Γ point. [7] We used a monoclinic unit cell of the CHA framework, which consists of 36 T atoms. It was optimized for the pure silicate structure with dimensions: a = b = 13.675 Å, c = 14.767 Å; α = β = 90 o , γ = 120 o [8]. One Si center in the unit cell located in one six-member ring was replaced with Al, as the negative charge around this Al was compensated by an H + cation. During the geometry optimization, atoms were allowed to relax until the force on each atom became less than 5×10 −2 eV/Å. The vibrational frequencies were calculated using a normal mode analysis where the elements of the Hessian were approximated as finite differences of gradients, displacing each atomic center by 1.5×10 −2 Å either way along each Cartesian direction.</p><p>The reported binding energies (BE) of the various adsorbates (CO, CO2, NO, HNO3, HNO2) were calculated as BE = -Ead -Esub + Ead/sub, where Ead is the total energy of the adsorbate in the gas phase (ground state), Esub is the total energy of the pristine zeolite system, where the framework negative charges are compansated by some of the modeled cationic species (H + , NO + , or NO2 + ) considered, and Ead/sub is the total energy of the zeolite, together with the adsorbate in the optimized geometry. With the above definition, negative values of BE imply a favorable interaction.</p><p>When Gibbs free energies were obtained the enthalpy values were calculated from the total energy values (Eel) corrected for the internal vibrational energy (Ev) [9] and zero point vibrations (ZPE) derived from frequency calculations of the optimized structures: H140 = Eel + Ev + ZPE.</p><p>The entropy values of the initial and transition states (TS) include only the vibrational degrees of freedom (Sv), since the adsorbates are bound to the zeolite and the rotational and translational degrees of freedom are converted into vibrations [10,11]. The expressions of all enthalpy and entropy contributions can be found elsewhere [9].</p><!><p>Binding energy (in kJ/mol) of NO2 and NO to Zeo/NO + structure, vibrational frequencies (in cm -1 ), ν(N-O), as well as selected interatomic distances, R(A-B), in pm.</p>

ChemRxiv

J o u r n a l Na me Local energy decomposition analysis and molecular properties of encapsulated methane in fullerene (CH 4 @C 60 ) †

Methane has been successfully encapsulated within cages of C 60 fullerene, and it is an appropriate model system to study confinement effects. Its chemistry and physics is also relevant for theoretical model descriptions. Here we provided insights into intermolecular interactions and predicted spectroscopic responses of the CH 4 @C 60 complex and compared with results from other methods and with literature data. Local energy decomposition analysis (LED) within the domain-based local pair natural orbital coupled cluster singles, doubles, and perturbative triples (DLPNO-CCSD(T)) framework was used, and an efficient protocol for studies of endohedral complexes of fullerenes is proposed. This approach allowed us to assess energies in relation to electronic and geometric preparation, electrostatics, exchange, and London dispersion for the CH 4 @C 60 endohedral complex. The calculated stabilization energy of CH 4 inside the C 60 fullerene was −13.5 kcal/mol and its magnitude was significantly larger than the latent heat of evaporation of CH 4 . Evaluation of vibrational frequencies and polarizabilities of the CH 4 @C 60 complex revealed that the infrared (IR) and Raman bands of the endohedral CH 4 were essentially "silent" due to dielectric screening effect of the C 60 , which acted as a molecular Faraday cage. Absorption spectra in the UV-Vis domain and ionization potentials of the C 60 and CH 4 @C 60 were predicted. They were almost identical. The calculated 1 H/ 13 C NMR shifts and spin-spin coupling constants were in very good agreement with experimental data. In addition, reference DLPNO-CCSD(T) interaction energies for complexes with noble gases (Ng@C 60 ; Ng = He, Ne, Ar, Kr) were calculated. The values were compared with those derived from supermolecular MP2/SCS-MP2 calculations and estimates with London-type formulas by Pyykkö and coworkers [Phys. Chem. Chem. Phys., 2010, 12, 6187−6203], and with values derived from DFT-based symmetry-adapted perturbation theory (DFT-SAPT) by Hesselmann & Korona [Phys. Chem. Chem. Phys., 2011, 13, 732−743]. Selected points at the potential energy surface of the endohedral He 2 @C 60 trimer were considered. In contrast to previous theoretical attempts with the DFT/MP2/SCS-MP2/DFT-SAPT methods, our calculations at the DLPNO-CCSD(T) level of theory predicted the He 2 @C 60 trimer to be thermodynamically stable, which is in agreement with experimental observations.

j_o_u_r_n_a_l_na_me_local_energy_decomposition_analysis_and_molecular_properties_of_encapsulated_met

Introduction<!>Methods<!>∆E<!>C−(T) int<!>Computational details<!>The choice of PNO truncation thresholds and basis sets<!>DLPNO-CCSD(T)-LED analysis of the CH 4 @C 60<!>DLPNO-CCSD(T)-LED Repulsion Stabilization<!>Spectroscopic properties of the CH 4 @C 60 in IR and UV-Vis<!>NMR properties of the CH 4 @C 60<!>C−(T) int<!>Conclusions<!>Conflicts of interest

<p>Carbon displays a rich chemistry and physics with a variety of molecular allotropes, including common graphite and diamond, but also fullerenes, carbon nanotubes, and graphene. [1][2][3][4] The most investigated fullerene is the C 60 "Buckminsterfullerene" composed of 20 hexagons and 12 pentagons of sp 2 -hybridized carbon atoms fused into a pseudosphere with a ∼7 Å diameter, as displayed in Figure 1. C 60 occurs in trace amounts on Earth in carbon-rich rocks and soot. 5,6 It has also been observed in the micrometeorite impact residue on the NASA Long Duration Exposure a b</p><p>Fig. 1 The C 60 fullerene (a) and its endohedral complex CH 4 @C 60 (b).</p><p>Facility orbiter, 7 which indicates that it either survived impact at nominal encounter velocity of orbital debris (∼11 km/s), 8 or was created in situ in space. Fullerenes isolated from meteorites re-vealed encapsulated atoms of noble gases with a 3 He/ 4 He isotope ratio of clearly extraterrestial origin. 9 Moreover, analyses of the 2019 data collected by the NASA/ESA Hubble Space Telescope confirmed spectral features of the ionized C + 60 species in diffuse interstellar bands making it the largest molecule observed in space and indicating that fullerenes might play an important role in interstellar chemistry. 10,11 The properties and chemistry of C 60 have been studied; for example, the wave-particle duality was experimentally observed for C 60 . 12 Methods for preparation and separation have been established, 13,14 and the possibilities to encapsulate atoms and molecules inside the fullerene cage was recognized soon after its discovery. 15 C 60 endohedral complexes with metal ions, noble gases, H 2 , N 2 , and H 2 O have been prepared by high-energy collisions of ionized fullerene species, harsh conditions of high temperature and pressure, electric arc, or by organic synthesis methods called as molecular surgery. [16][17][18][19] The successful synthesis of the endohedral complex with CH 4 was reported in 2019 by Whitby and coworkers. 20 Methane is the largest, and the first organic molecule encapsulated in the C 60 fullerene, and such a complex denoted as CH 4 @C 60 is the main object of this study.</p><p>There is no obvious direct route to measure the stabilization energy in fullerene endohedral complexes and obtain insights into the interaction mechanisms. Experimental observations in this respect have been limited to assessing the efficiency/probability of the given complex to be formed, and the main focus has been on spectroscopic and diffraction studies in relation to unusual physical properties of the encapsulated species. 21 Substantial theoretical efforts have directed to studies of C 60 endohedral complexes and associated intermolecular interactions. Pioneering ab initio studies by Jerzy Cioslowski [22][23][24] at the Hartree-Fock (HF) level of theory were expanded by studies of Bühl et al. 25,26 , Patchkovskii et al. 27 , Darzynkiewicz et al., 28 and Autschbach et al. 29 among others, [30][31][32][33][34][35] where density functional theory (DFT) and secondorder Møller-Plesset perturbation theory (MP2) were used. However, within the supermolecular approach with the interaction energy being the arithmetic relation of related energies (E int = E AB − (E A + E B )), DFT is essentially blind to long-range dispersion. This limitation has typically been addressed by using empirical correction schemes for the dispersion contributions. [36][37][38][39][40] MP2 is the lowest ab initio method that accounts for "real" dispersion effects but it is unbalanced and its performance for weak, noncovalent interactions are modest and system dependent. 41 Symmetry-adapted perturbation theory (SAPT) with monomers description at the DFT level (DFT-SAPT) developed by Krzysztof Szalewicz and coworkers is an alternative approach to account for dispersion contributions. [42][43][44] Within SAPT, the interaction energy is obtained as a sum of physical contributions, free from basis set superposition error (BSSE). 45 Hence, the most reliable stabilization energies for C 60 endohedral complexes so far have been obtained with DFT-SAPT by Hesselmann and Korona. 46,47 In parallel developments, approximate London-type formulas have been derived by Pyykkö and coworkers for the estimation of dispersion interaction in endohedral systems. 48,49 Attractive dispersion interactions between nonpolar species such as C 60 and CH 4 (or noble gases) are purely quantum me-chanical and originate from instantaneous effects of dynamical electron correlation. 50 For systems of chemical interest that can be correctly described by a single reference wave function, the most robust (and still tractable) way of introducing electron correlation is the coupled cluster singles, doubles, and perturbative triples CCSD(T) method. 51 It is the "gold standard" of quantum chemistry. However, a canonical implementation of the CCSD(T) model exhibits a seventh-order scaling with the system size, which results in tremendous computational expenses when considering systems larger than 15−25 atoms. Frank Neese and coworkers have recently developed an efficient implementation of the domain-based local pair natural orbital coupled cluster method (DLPNO-CCSD(T)). [52][53][54] Briefly, in the DLPNO-CCSD(T) approach the correlation energy is expressed as a sum of electron pair correlation energies, which enables the distinction between the "weak pairs" with negligible contribution to the total correlation energy and the "strong pairs" that constitute the dominant, desired part. In this way the "weak pairs' can be treated with a computationally more efficient second-order perturbation theory, whereas only the essential "strong pairs" are subjected to an accurate coupled cluster treatment, which greatly reduces the computational complexity. With appropriately selected pairselection thresholds, this model is capable to recover 99.9 % of the correlation energy of its canonical counterpart. It reproduces the CCSD(T) results within a chemical accuracy at substantially reduced computational efforts. 55,56 This approach extends the possibility of obtaining accurate ab initio energies to systems for which only DFT has been applicable so far. 57,58 Moreover, using a local energy decomposition (LED) protocol allows for a physical meaningful decomposition of the interaction energy within the DLPNO-CCSD(T) framework. 50,59,60 In this study, the goal was to providing an accurate interaction energy decomposition for the CH 4 @C 60 complex and the encapsulation energy barrier using the DLPNO-CCSD(T) method. This approach is not biased by the parametrization inherent to the DFT models, including type of exchange-correlation approximation and dispersion correction scheme. The reference interaction energies for endohedral complexes with noble gases were provided and compared with results by Pyykkö et al. and Hesselmann and Korona. [47][48][49]</p><!><p>The counterpoise-corrected interaction energy of molecular fragments X and Y can be expressed as: 59</p><!><p>where the E B A (C) notation denotes energy of fragment A calculated at the energy optimized coordinates of B and using a basis set of system C. The ∆E int term is the "electronic interaction", whereas ∆E geo−prep is the geometric preparation contribution that accounts for the differences between equilibrium molecular ge-ometries of isolated fragments and those in a complex ("deformation energy").</p><p>The electronic interaction energy ∆E int is decomposed within the following DLPNO-CCSD(T)-LED decomposition scheme:</p><p>The interaction energy ∆E int is decomposed into that from the Hartree-Fock level of theory ∆E HF int and the corrections due to inclusion of electron correlation ∆E C int . The latter is decomposed further into the interaction energy contribution at the CCSD level of theory ∆E C−CCSD int and that resulting from the perturbative triple excitations ∆E</p><!><p>. The Hartree-Fock interaction energy ∆E HF int is decomposed into the electronic preparation contribution ∆E HF el−prep , which correspond to the energy needed to bring the electronic structures of the isolated fragments into the one optimal for the interaction ("energy investment") and into attractive electrostatic E elstat and exchange E exch contributions. The CCSD correlation interaction energy is partitioned further into the genuine London dispersion interaction energy E C disp and the nondispersive correlation contribution ∆E C non−disp . The latter provides (dynamical) corrections to the Hartree-Fock polarization effects, "dynamic charge polarization". We refer the reader to the original articles for a detailed description of the method and implementation. 50,59,60</p><!><p>All calculations were performed with the ORCA code 61,62 using a very tight convergence tolerance of 1 × 10 −9 E h . The evaluation of Coulomb and exchange integrals was accelerated with the RIJCOSX approximation 63 with def2/J Coulomb-fitting basis set 64 and tightened grid (GridX5; further increase was verified to have negligible effect). Geometry optimizations were converged to very tight thresholds (VeryTightOpt setting) using the revised PBE "revPBE" exchange-correlation DFT approximation 65,66 together with atom-pairwise dispersion correction based on tight binding partial charges (D4). 40 The polarization-consistent segmented pcseg-1 basis set, 67 and the increased DFT integration grid (Grid5 NoFinalGrid) were used. The choice of the revPBE model was based on its performance in a recent and thorough benchmark study (best among the computationally efficient gradient-corrected GGA functionals). 68 To confirm the global energy minima at the potential energy surfaces, and to evaluate vibrational IR and Raman spectra, Hessians and dipole polarizabilities were calculated. The transition state search involved many consecutive computations of the Hessians towards the firstorder saddle point. Thereby computational efforts were reduced by using the smaller pcseg-0 basis set for atoms of aryl groups for the open-cage models. Cartesian coordinates of the models are provided in the Electronic Supplementary Information (ESI) † . DLPNO-CCSD(T) calculations were performed with a Foster-Boys localization scheme, a full MP2 guess, and employed correlationconsistent cc-pVXZ (X = D, T, Q, 5) orbital basis sets [69][70][71] together with the corresponding cc-pVXZ/C auxiliary basis sets. 72 The chosen PNO truncation thresholds are discussed in the results and discussion section. Computations were performed on a cluster node equipped with two Intel R Xeon R Gold 6126 CPUs (2.6 GHz; 12-core) and 256 GB of RAM.</p><!><p>To facilitate accuracy control in a user friendly manner, the authors of the DLPNO-CCSD(T) method have implemented three levels of predefined PNO truncation thresholds. These levels converge towards the method limit at increasing computational cost: LoosePNO, NormalPNO, and TightPNO. 55 The first two offer sufficient accuracy for most applications (<0.5 kcal/mol deviation for the evaluations of total energy with respect to the CCSD(T) reference), 55 but for analysis of weak intermolecular interactions, the TightPNO setting should be used. It ensures that the electron pairs that dominate the interaction are being treated at the coupled cluster level. 59 However, fullerenes are challenging for local coupled cluster methods. The large number of (long-range) π − π interactions in the highly delocalized π-system of fullerenes render calculations with the TightPNO setting and accurate basis sets very demanding. 56,73,74 The DLPNO-CCSD(T) in its current implementation is formally a linear-scaling method when considering the iterative part, but the RI-PNO integral transformations on large systems add substantial prefactors to the total computation times, which in turn limit the feasibility, also due to the memory and disk space requirements. Therefore, based on test calculations for the CH 4 • • • C 6 H 6 dimer (Figure 2), which is expected to exhibit similar nature of noncovalent interactions to those in the CH 4 @C 60 complex, we used a multilevel approach as proposed by Sparta et al. 57 Within the multilevel DLPNO approach the test CH 4 • • • C 6 H 6 system was divided into CH 4 and C 6 H 6 fragments. By this division, the intrafragment electron pairs with their orbitals entirely localized on one molecular fragment could be separated from those that gave rise to intermolecular interaction. 57 In Table 1, dispersion interaction and time used for the DLPNO-CCSD(T)-LED calculations for the CH 4 • • • C 6 H 6 dimer are presented. We monitored the convergence of the energy component for the dispersion interaction (E C disp ) since it depends solely on the treatment of the electron correlation, and therefore is critically sensitive to the truncation thresholds of the PNO and the (in)completeness of the basis set. By using a TightPNO setting for both intrafragment and interfragment pairs, smooth convergence towards the limit of a complete basis set (CBS) was observed (see Fig. 2). The dispersion interaction energy essentially converged at the TightPNO/cc-pVQZ level. Unfortunately, this setup would involve prohibitive computational costs when applied to endohedral fullerene complexes. Therefore, for routine applications we propose a more tractable multilevel DLPNO scheme. In this scheme, the truncation thresholds for the intrafragment PNO Open circles correspond to the calculation with the multilevel scheme proposed in the last row of Table 1.</p><p>for the guest (in this case CH 4 ) and for a troublesome delocalized π-system (C 6 H 6 , C 60 ) are reduced to the NormalPNO and LoosePNO, respectively, whereas the critical interfragment pairs are subjected to an accurate TightPNO evaluation. Together with the combination of the cc-pVQZ/cc-pVTZ basis sets, the multilevel scheme proposed in the last row in Table 1 offers massive computational savings without compromising accuracy to any significant extent. For a test conducted for the CH 4 • • • C 6 H 6 dimer, this approach recovered >95 % of the dispersion interaction energy when compared to the TightPNO/CBS reference, while being only twice as expensive as the TightPNO/cc-pVDZ calculation. The contribution from weak pairs to the E C disp was less than 4 % throughout the calculations presented in Table 1. Moreover, the related interaction energies compared well with previously reported accurate ab initio calculated energies for the CH 4 • • • C 6 H 6 dimer. When using the proposed multilevel DLPNO-CCSD(T) setup, a total interaction energy ∆E int =−1.32 kcal/mol was calculated, which agreed very well with the CCSD(T)/aug-cc-pVTZ result of −1.39 kcal/mol by Ringer et al. 77 The estimated dispersion contribution of −2.03 kcal/mol at the SAPT2/aug-cc-pVDZ level of theory by Ringer et al. 77 was close to our value of −2.22 kcal/mol. Above indicated that the multilevel DLPNO-CCSD(T) setup tailored for substantially more demanding calculations on endohedral complexes of fullerenes is robust.</p><!><p>In Table 2 the corresponding bond lengths of energy optimized geometries of the CH 4 , C 60 , and CH 4 @C 60 endohedral complex are shown. As a consequence of I h symmetry the C 60 fullerene molecular structure is defined by the two distinct C−C distances r 1 and r 2 that originate from the bonds between fused pentagons and hexagons (r 1 ) and the shorter ones between two hexagons (r 2 ). The energy optimized model of C 60 exhibited excellent agreement with experimental bond lengths estimates. For r 1 the deviation was <0.005 Å, and r 2 coincided with the empirical C−C distance. This indicated that the revPBE-D4/pcseg-1 level of theory was capable to deliver robust models of fullerene systems. For comparison, reported geometries of C 60 optimized at the (ab initio) HF and MP2 levels of theory have revealed considerable deviations of C−C bond lengths, 33,34 whereas previous tests of different DFT approximations have not included dispersion corrections. 33,34 Noteworthy, at the revPBE-D4/pcseg-1 level of theory the geometries of both CH 4 and C 60 remained essentially unchanged upon formation of the CH 4 @C 60 endohedral complex.</p><!><p>Fig. 3 Results of the DLPNO-CCSD(T)-LED/cc-pVQZ(CH 4 )/cc-pVTZ(C 60 ) local energy decomposition analysis for the CH 4 @C 60 endohedral complex, energies are given in kcal/mol. Inset shows total interaction energies (∆E int ) corresponding to the Hartree-Fock, CCSD, and CCSD(T) levels of theory.</p><p>respond to an energy minimum at coupled cluster level of theory, for which the equilibrium C−H distance was shorter, and closer to experimental estimate. The encapsulation of CH 4 in C 60 is associated with a tiny shortening (<0.001 Å) of the C−H bond lengths. Therefore, the total DLPNO-CCSD(T) energy of the CH 4 molecule calculated at molecular geometry corresponding to that in the CH 4 @C 60 complex was lower compared to that for the isolated CH 4 . Although this effect was very small and had no implications on the evaluation of the electronic interaction energy ∆E int in the CH 4 @C 60 complex, it would lead to an unphysical lowering of the "deformation energy", ∆E geo−prep term in Equation 1. Therefore, to provide the most realistic values the evaluation of ∆E geo−prep included only the contribution from the deformation of the C 60 cage.</p><p>In Figure 3, interaction energy contributions are presented from the DLPNO-CCSD(T)-LED analysis of the CH 4 @C 60 endohedral complex. The total interaction energy for the complex is represented as a sum of seven physical contributions:</p><p>, and ∆E geo−prep (according to Equations 1 and 2). The large and positive electronic preparation term ∆E HF el−prep = +81.22 kcal/mol is counteracted by attractive contributions due to electrostatics and exchange (E elstat = −39.73 kcal/mol, E exch = −21.00 kcal/mol). However, the summed components of the interaction energy at the Hartree-Fock level (∆E HF el−prep + E elstat + E exch ) resulted in substantially repulsive interaction of ∆E HF int = +20.49 kcal/mol. This value was basically identical to the value calculated by Pyykkö and coworkers at the HF/def2-QZVPP level of theory (the same as the +20.50 kcal/mol). 49 This summation can be regarded as an estimate of the extent of steric repulsion. 48 Noteworthy is that the CH 4 @C 60 is predicted to be unstable also by DFT if empirical dispersion corrections are not used. 30,31 The non-dispersive correction due to electron correlation was small and repulsive (∆E C non−disp = +1.00 kcal/mol). As expected, London dispersion is the dominant intermolecular interaction mechanism. The magnitude of the E C disp term of −29.96 kcal/mol was larger than the substantially repulsive Hartree-Fock interaction and the ∆E C non−disp correction, resulting in an endohedral complex stabilization by −8.47 kcal/mol. Yet, further attractive correction came from the contribution from perturbative triple excitations ∆E C−(T) int that stabilized the complex by an additional estimated contribution of −5.03 kcal/mol. The correction from perturbative triples was important, given that it increased the net binding energy in the complex by nearly 60 % (from −8.47 to −13.50 kcal/mol; see inset in Fig. 3). Therefore, the final electronic interaction energy at the DLPNO-CCSD(T)/cc-pVQZ(CH 4 )/cc-pVTZ(C 60 ) level of theory for the CH 4 @C 60 endohedral complex was −13.50 kcal/mol. The geometry preparation ("deformation energy") term was very small (∆E geo−prep = +0.03 kcal/mol). C 60 is a rigid molecule and encapsulation of the CH 4 guest had a negligible effect on its geometry (see Table 2). Our reference stabilization energy of ∆E = −13.47 kcal/mol was compared with best reported estimates. For the CH 4 @C 60 complex, the most robust results have been reported by Pyykkö and coworkers. 49 In that study interaction energies were obtained with supermolecular MP2 and its spin component scaled counterpart (SCS-MP2). Calculations were performed with the def2-TZVPP and def2-QZVPP basis sets, interaction energies were corrected for basis set superposition error and extrapolated to the complete basis set limit. The obtained MP2 interaction energy of −21.37 kcal/mol was clearly overestimated. The value obtained with SCS-MP2 (−11.97 kcal/mol) was closer to the coupled cluster reference, but underestimated. These calculated interaction energies followed the pattern observed in previously reported benchmark calculations. For the CH 4 • • • C 6 H 6 dimer, MP2 overesti-mates the CCSD(T) interaction energy, as was shown by Ringer et al., 77 and for the endohedral CH 4 @C 60 complex, this overestimation seems to be even more pronounced. The same trend of deviation was observed by Pyykkö and coworkers for dimers composed of atoms of noble gases and benzene (Ng• • • C 6 H 6 ), where MP2 overestimated the CCSD(T) reference interaction energies significantly, whereas SCS-MP2 was generally much closer to coupled cluster results, but consistently underestimated the interaction. 49 Pyykkö and coworkers have also developed London-type formulas to estimate dispersion interaction energies in endohedral systems. 48,49 The input parameters to these formulas such as ionization potentials and polarizabilities can be readily computed at the DFT level. Using the data from the study of Pyykkö and coworkers 49 (Table 17, equations 69+86) the dispersion energy estimate of −17.33 kcal/mol was obtained for the CH 4 @C 60 complex. This energy was much smaller than for the DLPNO-CCSD(T)-LED (−29.96 kcal/mol), and would not overcome the steric repulsion estimate of +20.49 kcal/mol, and the complex would not be estimated to be stabilized in that description. vations that high pressure and elevated temperature conditions (1645 atm, 190 • C for 22 h) are necessary to achieve a high degree of CH 4 insertion. 20 Noteworthy is that the electronic repulsive interaction at the orifice ∆E int = + 9.19 kcal/mol amounted to only less than half of the insertion energy barrier. The remaining geometry preparation term corresponded to the energy needed to deform the open-cage fullerene from its equilibrium geometry to the one optimal for CH 4 insertion (∆E geo−prep = +10.62 kcal/mol). After insertion, the CH 4 molecule is predicted to be stabilized inside the open-fullerene cage by ∆E = −10.16 kcal/mol.</p><!><p>Harmonic vibrational frequencies of the endohedral CH 4 @C 60 complex have been computed at the level of GGA and hybrid DFT approximations without using dispersion corrections, 30,31 and at the Hartree-Fock level of theory. 32 Hence, those frequencies were evaluated on structures corresponding to energy minima in situation where London dispersion interactions had not been accounted for. In those studies the intensities in the resulting calculated IR and Raman spectra were not discussed as well. Therefore, we calculated the harmonic vibrational frequencies together with the respective IR absorption coefficients and Raman scattering factors for the CH 4 , C 60 , and the CH 4 @C 60 complex at the revPBE-D4/pcseg-1 level of theory. Related frequencies, absorption coefficients and scattering factors are presented in Table 3.</p><p>The calculated vibrational frequencies of the C 60 were in very good agreement with experimental data and virtually not changed upon CH 4 encapsulation. This situation was in agreement with the experimental IR spectrum of the H 2 O@C 60 , where the vibrational frequencies of the fullerene cage were the same as those of the free C 60 . 19 The calculated IR absorption coefficients and Raman scattering factors (for the fullerene cage) were predicted to be slightly affected by CH 4 encapsulation and resulted on average in a <5 % loss in spectral intensity. Frequencies of the encapsulated CH 4 on the other hand were blue shifted with respect to the free CH 4 molecule, and were in agreement with previous theoretical predictions. 31,32 Our results suggested that the triple degeneracy of the asymmetric bending and stretching IR modes (1287 and 3107 cm −1 ) of CH 4 was partially removed due to the interaction with the cage. However, what was the most important, for both IR and Raman a substantial loss in spectral intensities for the encapsulated CH 4 was revealed. This intensity loss was in line with experimental IR spectra of the CH 4 @openfullerene complex, where vibrations of the CH 4 could not be observed. 35 In addition, vibrational features of the H 2 O were very weak in the experimental IR spectrum of H 2 O@C 60 , and the potential screening effect of the fullerene cage was indicated. 19,83 Dielectric measurements conducted at low temperature and IR spectra of H 2 O@C 60 collected at liquid helium temperature have revealed that the dipole moment of the encapsulated H 2 O was 0.5±0.1 D, 84,85 which is about 25 % of free the H 2 O molecule (in agreement with theoretical predictions 86 ). A very similar extent of dipole moment reduction has been observed for HF in HF@C 60 as well. 87 Table 3 Harmonic vibrational frequencies (ν; cm −1 ), IR absorption coefficients (A; 10 5 cm/mol) a and Raman scattering factors (S; Å 4 /amu) a calculated at the revPBE-D4/pcseg-1 level of theory b ; experimental values for the free CH The fullerene cage protects encapsulated species from influence of the external electric field, and acts as a molecular Faraday cage. 88,89 Such a screening effect can be assessed by calculating a difference between the dipole polarizability (α) of an endohedral complex, and the sum of polarizabilities of an isolated guest and the empty fullerene: 90</p><p>Negative ∆α corresponds to the polarizability depression that results from the decrease of polarizability of the encapsulated guest. Therefore, the dielectric screening coefficient can be evaluated: 90</p><p>To inspect these effects for the CH 4 @C 60 endohedral complex, the dipole polarizabilities of CH 4 , C 60 , and CH 4 @C 60 were calcu-lated at the CAM-B3LYP/Sadlej-pVTZ level of theory. The rangeseparated and Coulomb-attenuating method called as the CAM-B3LYP DFT approximation 91 was shown to deliver accurate polarizabilities, 92 and a balanced description of electronic excited states. 93 The Sadlej-pVTZ basis set was specifically developed for calculations of polarizabilities and other electric molecular properties. [94][95][96] The calculated polarizabilities and the respective values obtained from Equations 3 and 4 are shown in Table 4. The values for the polarizabilities for CH 4 and C 60 were in good agreement with experimental data. The polarizability of the CH 4 @C 60 complex was predicted to be almost the same as that of the empty C 60 , which in turn was reflected in a substantial polarizability depression of ∆α=−2.48. The value for the dielectric screening coefficient (c = 0.97) indicated a particularly strong effect for the CH 4 @C 60 endohedral complex. The polarizability of the encapsulated CH 4 molecule was essentially quenched. a From ref. 97,98 The electronic excited states energies and absorption in the UV-Vis domain were calculated at the CAM-B3LYP/cc-pVTZ level of theory with a time-dependent DFT (TD-DFT) within the efficient sTD-DFT implementation of Bannwarth and Grimme. 99 Transition energies, wavelengths, and oscillator strengths for C 60 and CH 4 @C 60 are shown in Table 5. The predicted spectroscopic characteristics of C 60 and CH 4 @C 60 were almost identical in the UV-Vis range. Both entities exhibited the same transition energies. The oscillator strengths were marginally lower for the complex compared to the empty fullerene. These results were in agreement with experimental observations for C 60 and H 2 O@C 60 . They display close to identical UV-Vis spectra, despite revealing a slightly lower absorption for the complex. 19 The data in Table 5 compare favorably with the experimental UV-Vis spectrum of C 60 in the gas phase. 100 The set of three most intense bands at 187/231/300 nm that reveal oscillator strengths of 1.907/1.922/0.384 correspond to the experimentally observed transitions at 205/257/330 nm that exhibit extinction coefficients of 4.8/3.5/0.9 (10 5 L mole −1 cm −1 ). 100 Hence, the pattern of UV-Vis bands correlated well between theoretical predictions and the experimentally observed spectra, however, transition energies were overestimated (too short wavelengths) at the TD-DFT level of theory.</p><p>Because of the rich chemistry of the ionized fullerene species (C n+ 60 ; n = 1, 2, 3), it is interesting to compare the ionization potentials of the C 60 and its endohedral complex. Ionized fullerenes exhibit a diversity of ionization mechanisms and a variety of reactions with potential implications to chemistry in the interstellar medium. 101 The vertical ionization energies (V IE) for C 60 and CH 4 @C 60 were calculated at the CAM-B3LYP/Sadlej-pVTZ level of theory according to:</p><p>where E n+ cation and E 0 neutral denote the total energies of the ionized and neutral species, respectively. They were calculated at the equilibrium geometry of the ground state. The calculations for ionized species involved an unrestricted (UDFT) formalism due to higher than singlet multiplicities. The obtained results indicated that the ionization potentials of the CH 4 @C 60 complex were almost identical to those of empty C 60 . The calculated V IE for C 60 agreed very well with experimental data, as can be seen from values in Table 6.</p><!><p>Fullerenes constitute the only known allotrope of carbon that can be dissolved in organic solvents at room temperature. 103 Therefore, high resolution liquid-state NMR spectra of C 60 and its endohedral complexes can be measured in common NMR solvents. 104,105 Whitby and coworkers collected and analyzed 1 H and 13 C NMR spectra of the CH 4 @C 60 complex dissolved in 1,2dichlorobenzene. 20 For prediction of such spectra with quantum chemistry methods, a robust model including subtle interactions of the fullerene cage with solvent molecules is important. Hence, we constructed systems composed of the C 60 and CH 4 @C 60 explicitly solvated by 25 molecules of 1,2-dichlorobenzene, whereas solvent effects at the outer sphere were accounted implicitly by a polarizable continuum model (PCM) assuming a dielectric con-stant ε = 9.93. The initial configuration was obtained with the Packmol software, 106 and the coordinates were energy optimized at the revPBE-D4/pcseg-1(CH 4 , C 60 )/pcseg-0(C 6 H 4 Cl 2 ) level of theory up to the energy change of < 5 × 10 −6 E h ; see Figure 5. The 1 H and 13 C nuclear magnetic shielding tensors 107,108 (σ σ σ ; assessed with the GIAO approach 109 ) and indirect nuclear spinspin coupling constants (J) were calculated at the level of PBE0 DFT approximation. 110 The calculations were performed with segmented pcS-1 and pcJ-1 basis sets, which have been specifically developed to provide fast convergence towards the Kohn-Sham limit for NMR shieldings and spin-spin couplings. 111,112 The chosen setup represents a reasonable compromise for the calculations of 1 H/ 13 C NMR parameters when compared to more accurate methods given the size of the system. 113,114 For solvent molecules, the pcseg-0 basis set was used. The calculations of NMR observables involved all electrons (no frozen core) and very tight grids (GridX8 Grid7). The calculated isotropic 1 H/ 13 C NMR shielding:</p><p>was converted into isotropic NMR chemical shift δ according to:</p><p>where δ j and σ j correspond to the chemical shift and shielding of an atom of interest j, whereas σ re f , calc CH 4 (gas) and δ re f , exp CH 4 (gas) represent the calculated shielding and experimental shift of the reference, respectively. The CH 4 molecule was used as a reference, since its proton and carbon chemical shifts measured in the gas phase are available. 115 Shifts of chemically equivalent atoms were averaged. Spin-spin coupling constants were represented as a sum of four physical contributions: the Fermi contact (FC), spin-dipole (SD), paramagnetic spin-orbit (PSO), and diamagnetic spin-orbit (DSO) terms.</p><p>a From ref. 20,115,116 The calculated 1 H and 13 C NMR chemical shifts as well as the 1 H− 13 C spin-spin couplings presented in Table 7 revealed a very good agreement with experimental data. For protons in the CH 4 molecule, the change in chemical shift (∆δ ) upon encapsulation in C 60 was predicted to be −7.54 ppm, which compared very well to −7.88 ppm observed in an experiment. 20 This change on encapsulation is associated with the NMR shielding inside the fullerene cage, where the locally induced magnetic fields counteract the applied external field of the NMR instrument. The corresponding effect for the 13 C in CH 4 was slightly smaller, as revealed by the calculated and experimental ∆δ values of −5.87 and −4.98 ppm, respectively. For carbon atoms of the fullerene cage, the presence of endohedral CH 4 results in a deshielding of the 13 C NMR signal. Therefore, this effect is opposite to that observed for the 13 CH 4 inside the cage. The calculated 13 C deshielding of the cage of +0.51 ppm was in an excellent agreement with the experimental value of +0.52 ppm. 20 The spin-spin coupling constant 1 J HC in the CH 4 was dominated by the Fermi-contact (FC) mechanism. The calculated coupling strength of 1 J HC = 126.4 Hz for the free CH 4 was close to the experimental value of 125.3 Hz. 116 Encapsulation of CH 4 in C 60 had little effect on the 1 J HC , and the calculated value of 124.1 Hz was very close to the experimentally determined 124.3 Hz. 20 Therefore, our theoretical model predicts correctly the sign and the small magnitude of the ∆J. The small change of the coupling was consistent with negligible deformation of the CH 4 geometry upon encapsulation.</p><p>Complexes with noble gases Ng@C 60 (Ng = He, Ne, Ar, Kr) The efficient DLPNO-CCSD(T) setup designed for probing intermolecular interactions in the CH 4 @C 60 complex was also used to obtain reference interaction energies for endohedral complexes with atoms of noble gases (Ng@C 60 ; He, Ne, Ar, Kr). Hence, in this section we compared our coupled cluster results to previously reported estimates at lower levels of theory. a Two-point extrapolation to the CBS limit based on the def2-TZVPP/def2-QZVPP basis sets; results from ref. 49 b DFT-SAPT calculations with PBEac functional, aug-cc-pVDZ(He, Ne), aug-cc-pVTZ(Ar, Kr), and TZVP(C) basis sets; results from ref. 47 In Table 8, interaction energies (∆E int ) calculated at the DLPNO-CCSD(T)/cc-pVQZ(Ng)/cc-pVTZ(C 60 ) level of theory are presented together with values from the MP2/SCS-MP2 calculations by Pyykkö and coworkers, 49 and from the DFT-SAPT calculations by Hesselmann and Korona. 47 The MP2 method exhibited the most unbalanced performance. The associated interaction energies for He was too low with this method, for Ne they were quite reasonable (accidentally), but those for Ar and Kr were severely overestimated. These trends resembled those observed for values derived with the MP2 method for the CH 4 @C 60 complex. The spin component scaled variant (SCS-MP2) displayed an improved description. Although interaction energies for He and Ne were too low, the result for Ar was close to the coupled cluster reference, and the overestimation for Kr was not as severe as with the MP2. The situation with the DFT-SAPT results was more complex and somewhat difficult to judge. On the one hand, all interaction energies for these systems obtained with the DFT-SAPT were consistently underestimated when compared to the DLPNO-CCSD(T) values. On the other hand, the performance of the DFT-SAPT was consistent and well balanced across the He, Ne, Ar, and Kr complexes. This complexity became apparent when the results of DLPNO-CCSD(T) were plotted against the DFT-SAPT counterpart in Figure 6a. The correlation among the two approaches was very good, despite the consistently underestimated interaction energies by the latter. For comparison, the correlation with the SCS-MP2 data shown in panel b was much worse and significantly less convincing.</p><p>In Table 9, estimations of dispersion contributions to the interaction energy for the Ng@C 60 complexes are presented for different methods of calculations. The dispersion contributions (E C disp ) from the DLPNO-CCSD(T)-LED were in good agreements with the dispersion components from the DFT-SAPT for all four considered complexes. Hence, it was concluded that the DFT-SAPT compared favorably to the DLPNO-CCSD(T)-LED for prediction of the dispersion interaction in fullerene complexes. This favorable comparison for DFT-SAPT is illustrated in Figure 7a where results of DLPNO-CCSD(T) and DFT-SAPT are plotted against each other. Therefore, the too low interaction energies obtained with the DFT-SAPT for the complexes with noble gases did not originate from inappropriate description of the dispersive part of the in- Fig. 6 Interaction energies (∆E int ) calculated at the DLPNO-CCSD(T)/cc-pVQZ(Ng)/cc-pVTZ(C 60 ) level of theory plotted against the results from DFT-SAPT (a), 47 and supermolecular SCS-MP2 (b). 49 Grey line corresponds to the ideal correlation y = x, whereas red line to the linear regression fit.</p><p>teraction, but resulted from deficiencies in the remaining components of the interaction energy. Analysis with DLPNO-CCSD(T)-LED revealed significant contributions from perturbative triples (∆E</p><!><p>), and yet, for small guests (He, Ne) for which repulsive interaction at the Hartree-Fock level (∆E HF el−prep + E elstat + E exch ) is small, the non-dispersive corrections due to electron correlation (∆E C non−disp ) were attractive. The dispersion interaction as predicted with London-type formulas by Pyykkö and coworkers were substantially underestimated when compared to the DLPNO-CCSD(T)-LED results, although a linear trend with the latter was revealed; see Table 9 and Figure 7b. This underestimation indicates that the sum of the dipole-dipole and quadrupole-quadrupole terms in the "Pyykkö model 2010" (equations 69+72 from ref. 49 ) was not sufficient, and that the higher order multipole-multipole contributions are necessary to include to obtain better agreement with high-level quantum chemistry methods. 47,49 The He 2 @C 60 trimer</p><p>The existence of the He 2 @C 60 trimer, where two helium atoms are encapsulated inside the C 60 was discovered with 3 He NMR by Rabinovitz and coworkers. 117 The observed He 2 @C 60 :He@C 60 ratio of 1:200 was 10 times smaller than that for the He 2 @C 70 :He@C 70 (1:20). This reduction suggested that the smaller cavity of C 60 was significantly less suited for the encapsulation of two He atoms as compared to the C 70 fullerene.</p><p>The stability of the He 2 @C 60 trimer was studied with quantum chemistry methods by Darzynkiewicz & Scuseria, 28 Krapp & Franking, 118 and Hesselmann & Korona. 47 However, all methods applied, including DFT, MP2, SCS-MP2, and DFT-SAPT predicted repulsive interaction in the range from +1.13 to +10.23 kcal/mol, depending on the method and the basis sets used. Hence, theoretical investigations reported so far suggest that the He 2 @C 60 is thermodynamically unstable towards loss of the noble gas atom, in stark contrast to the experimental observation.</p><p>To gain insight into this challenging system and confront discrepancy between theory and experiment with the DLPNO-CCSD(T) approach, a relaxed potential energy surface scan at the revPBE-D4/pcseg-1 level of theory was performed for the He 2 @C 60 . The He−He distance was sampled with a 0.02 Å increment, and for each step all other coordinates were subjected to unconstrained optimization. Subsequently, single-point DLPNO-CCSD(T) calculations were performed on the resulting geometries to locate the "true" energy minimum; see Figure 8.</p><p>At the DLPNO-CCSD(T)/cc-pVQZ(He)/cc-pVTZ(C 60 ) level of theory, the equilibrium He−He distance inside the C 60 was 1.94 Å. This distance was not only substantially shorter than that of 3.00 Å calculated for the free He 2 dimer, but it corresponded to a clearly repulsive regime for the latter. comparably smaller. Note that the interaction well for the isolated He 2 dimer was relatively shallow, whereas the relation of potential energy change upon He−He distance variation was very steep for the He 2 @C 60 trimer.</p><p>For the equilibrium distance r He−He = 1.94 Å inside C 60 the stabilization energy ∆E int of the He 2 @C 60 trimer:</p><p>− (E XY Z X (XY Z)</p><p>evaluated at the DLPNO-CCSD(T)/cc-pVQZ(He 2 )/cc-pVTZ(C 60 ) level of theory was −1.43 kcal/mol. Noteworthy was that the stabilization energy for the He 2 @C 60 trimer at the DLPNO-CCSD(T) level of theory was almost as high as for the complex with a single He atom with the DFT-SAPT. The ab initio calculations predicted the He 2 @C 60 trimer to be stable, which is in agreement with experimental observations.</p><!><p>The reference interaction energies for endohedral complexes of the C 60 fullerene with He, Ne, Ar, Kr, and CH 4 were calculated at the DLPNO-CCSD(T) level of theory and decomposed into physical contributions with the LED scheme. An accurate and efficient multilevel DLPNO-CCSD(T) setup was proposed, which was applicable to routine studies of endohedral complexes of C 60 and larger fullerenes. Calculated molecular properties of the CH 4 @C 60 complex revealed that the IR and Raman bands of the endohedral CH 4 were essentially "silent" due to the dielectric screening effect of the C 60 , which acted as a molecular Faraday cage. Absorption spectra in the UV-Vis and ionization potentials of C 60 and CH 4 @C 60 were predicted to be almost the same. Calculated 1 H/ 13 C NMR shifts and spin-spin coupling constants were in very good agreement with experimental data. Lastly, selected points at the potential energy surface of the endohedral He 2 @C 60 trimer were calculated at the DLPNO-CCSD(T) level of theory. In contrast to previous theoretical studies with DFT, MP2, SCS-MP2 and DFT-SAPT, where all these methods predicted the He 2 @C 60 to be thermodynamically unstable towards the loss of the noble gas atom, our calculations predicted the He 2 @C 60 to be stable, which is in agreement with experimental observations. Therefore, the case of the He 2 @C 60 trimer clearly indicated that the DLPNO-CCSD(T) level of theory is indispensable in studies of weakly interacting systems, and should be used whenever applicable.</p><!><p>There are no conflicts to declare.</p>

ChemRxiv

The Unnatural Substrate Repertoire of A, B, and X Family DNA Polymerases

As part of an effort to develop unnatural base pairs that are stable and replicable in DNA, we have examined the ability of five different polymerases to replicate DNA containing four different unnatural nucleotides bearing predominantly hydrophobic nucleobase analogs. The unnatural pairs were developed based on intensive studies using the Klenow fragment of DNA polymerase I from E. coli (Kf) and were found to be recognized to varying degrees. The five additional polymerases characterized here include family A polymerases from bacteriophage T7 and Thermus aquaticus, family B polymerases from Thermococcus litoralis and Thermococcus 9\xc2\xb0N-7 and the family X polymerase, human polymerase \xce\xb2. While we find that some aspects of unnatural base pair recognition are conserved among the polymerases, for example the pair formed between two d3FB nucleotides is typically well recognized, the detailed recognition of most of the unnatural base pairs is generally polymerase dependent. In contrast, we find that the pair formed between d5SICS and dMMO2 is generally well recognized by all of the polymerases examined, suggesting that the determinants of efficient and general recognition are contained within the geometric and electronic structure of these unnatural nucleobases themselves. The data suggest that while the d3FB:d3FB pair is sufficiently well recognized by several of the polymerases for in vitro applications, the d5SICS:dMMO2 heteropair is likely uniquely promising for in vivo use. T7-mediated replication is especially noteworthy due to strong mispair discrimination.

the_unnatural_substrate_repertoire_of_a,_b,_and_x_family_dna_polymerases

1. Introduction<!>2. Results<!>2.1.1 Kf-mediated synthesis and extension of natural base pairs<!>2.1.2 Recognition of unnatural base pairs by Kf<!>2.2.1 T7-mediated synthesis and extension of natural base pairs<!>2.2.2 Recognition of unnatural base pairs by T7<!>2.3.1 Taq-mediated synthesis and extension of natural base pairs<!>2.3.2 Recognition of unnatural base pairs by Taq<!>2.4.1 Vent-mediated synthesis and extension of natural base pairs<!>2.4.2 Recognition of unnatural base pairs by Vent<!>2.5.1 Therminator-mediated synthesis and extension of natural base pairs<!>2.5.2 Recognition of unnatural base pairs by Therminator<!>2.6.1 Polymerase \xce\xb2-mediated synthesis and extension of natural base pairs<!>2.6.2 Recognition of unnatural base pairs by polymerase \xce\xb2<!>3. Discussion<!>3.1.1 Self pairs<!>3.1.2 The d5SICS:dMMO2 heteropair<!>3.2 Structure-activity relationships<!>3.3 Progress toward the expansion of the genetic alphabet<!>Materials and Methods<!>Steady-State Kinetics

<p>The ability to replicate and amplify DNA in vitro has revolutionized biotechnology by enabling a wide range of techniques, including PCR, cloning, and DNA sequencing. Much interest is also focused on the use of oligonucleotides for the development of novel materials,1,2 diagnostics,3 and therapeutics.4-7 All of these efforts require, or are at least are facilitated by, sequence specific amplification of DNA by DNA polymerases, which is currently limited by the natural substrate repertoires of the polymerases. The identification of novel nucleotides that selectively pair in duplex DNA as well as during replication, i.e. an unnatural base pair, would allow for the expansion of these in vitro techniques to include oligonucleotides that are site specifically modified with interesting functionalities (catalytic groups, fluorophores, etc.) thereby expanding their scope and potential applications. Additionally, an unnatural base pair would lay the foundation for the construction of a synthetic organism with increased potential for information storage that might be used, for example, to site-specifically label RNA in vivo or to encode proteins with unnatural amino acids.8-10</p><p>Toward these goals, we have synthesized and evaluated a large number of nucleotides bearing predominantly hydrophobic nucleobase analogs,11-15 and identified several that are recognized with at least moderate efficiency and selectivity by the exonuclease deficient Klenow fragment of DNA polymerase I from E. coli (Kf). These studies have elucidated many of the determinants of efficient unnatural base pair replication and were punctuated by the discovery of four base pairs of particular significance (Figure 1). First, while examining a variety of isocarbostiryl-based nucleoside analogs, we showed that Kf synthesizes the dPICS self pair (i.e. by insertion of the unnatural nucleotide opposite itself in the template) with moderate efficiency and fidelity; however, the resulting primer is not extended (i.e. by continued primer elongation resulting from the insertion of the next correct dNTP).13 Structural studies revealed that in duplex DNA the dPICS nucleobases interact by intercalation,16 which we speculate facilitates self pair synthesis but results in a primer terminus structure that is refractory to extension, at least with Kf. Second, after examining a variety of indole-based scaffolds, we found that Kf synthesizes the d7AI self pair with moderate efficiency, but cannot extend it,14 while rat pol β cannot synthesize the self pair, but extends it with a rate comparable to that of a natural base pair.15 Thus, full length synthesis of DNA containing the d7AI self pair is possible when both Kf and pol β are present.</p><p>The results of these early studies prompted us to examine nucleotides bearing small phenyl-ring nucleobase analogs. These analogs are expected to be incapable of interstrand intercalation, and we hypothesized that they might be derivatized to facilitate synthesis and extension. After extensive investigations, we found that the inclusion of a single fluorine substituent at the position meta to the C-glycosidic bond significantly facilitates Kf-mediated self pair synthesis and extension (d3FB, Figure 1).11 Structural and biochemical studies have shown that the d3FB self pair forms a more natural-like 'in plane' primer terminus, and have also suggested that the specific dipole alignment within the self pair may be important for both its efficient synthesis and extension.11,16</p><p>Finally, we recently conducted two separate screens of 3600 possible base pair candidates for self pairs or heteropairs that are efficiently synthesized and extended by Kf.12 After optimization of the initial hits, these screening efforts yielded the d5SICS:dMMO2 heteropair (Figure 1). Steady-state kinetic analysis revealed that this heteropair is synthesized and extended by Kf with remarkable efficiency and fidelity in either strand context. Further characterization suggested that, at least with Kf, efficient replication relies on the substituents at the position ortho to the glycosidic linkage, which are likely disposed in the developing minor groove at the primer terminus.</p><p>Our efforts to design predominantly hydrophobic unnatural base pairs that are efficiently replicated have been based mainly on their recognition by Kf. However, Kf appears to recognize nucleoside analogs differently than other polymerases, such as the thermostable polymerases Vent and 9°N-7,17-19 or the mesophilic polymerase pol α.20,21 Moreover, work from the Benner lab with nucleobase analogs that pair via orthogonal H-bonding topologies indicates that recognition can be polymerase dependent.22-26 In addition, Kool et al. have shown that efficient extension of H-bonding deficient nucleobase isosteres by different polymerases relies to different extents on polymerase-nucleobase H-bonds,27 and that different polymerases accommodate changes in nucleobase size differently.28-31 Finally, our own data showing the significant difference in recognition of the d7AI self pair by Kf and pol β15 suggests that there may be significant variability in how different polymerases recognize the predominantly hydrophobic base pairs.</p><p>No comprehensive comparisons of substrate recognition have been reported for DNA-dependent DNA polymerases; however, they have been grouped into six families A, B, C, D, X, and Y, based on sequence homology.32-34 Of these, the two largest and most thoroughly studied are the A family (or type I), exemplified by DNA polymerase I from E. coli, and the B (or pol α) family, exemplified by eukaryotic DNA polymerase α. In addition, the family X polymerases, especially pol β, have also been well characterized, and are structurally and functionally diverged from the A and B family polymerases. To begin to explore the potential substrate repertoires of natural DNA polymerases, we have analyzed the replication of the dPICS, d7AI, and d3FB self pairs and the d5SICS:dMMO2 heteropair by two family A polymerases (T7 and Taq), and two family B polymerases (Vent and Therminator). These polymerases are among the most useful for biotechnology applications, and they have been characterized extensively.35-41 In addition, each is a replicative polymerase that is responsible for both leading and lagging strand replication in vivo. Due to its structural and functional divergence, we also analyzed unnatural DNA replication by the human family X polymerase pol β, which is a repair polymerase involved in gap filling.42-47 The results, along with available structural data, allow us to begin to define the general and polymerase specific determinants of natural and unnatural base pair recognition. In addition, these studies should help in the design of unnatural base pairs, not only for in vitro applications, but also for in vivo applications, which will require general recognition by the multiple polymerases involved in genome replication and maintenance.</p><!><p>Two A family polymerases were selected for characterization to augment our data with Kf, the mesophilic polymerase from bacteriophage T7 (T7 polymerase) and the thermophilic polymerase from Thermus aquaticus (Taq polymerase), both of which have high homology to Kf and have been particularly well characterized. For our initial characterization of B family polymerases we selected two thermophilic polymerases, Vent, from the archaea Thermococcus litoralis,48 and Therminator, a mutant of the native DNA polymerase from the Thermococcus species 9°N-7.18,19 In addition, based on our previous characterization of pol β with d7AI (vide supra), and its diverged role as a DNA repair enzyme, we included the human variant of this X family polymerase. All of the polymerases used in this study are either naturally exonuclease deficient or rendered so by point mutation. The mesophilic polymerases were characterized at 25 °C, with the exception of pol β, which due to generally reduced activity with both natural and unnatural substrates, was characterized at its optimal temperature of 37 °C. The thermophilic polymerases were characterized at 50 °C, to allow the use of the same primer-template substrates employed in our previous studies (significant melting of the duplex would occur at higher temperatures). While this is approximately 20 °C below their optimal temperatures, each polymerase retains the majority of its activity under these conditions.49-51</p><p>The recognition of the different unnatural base pairs was evaluated using steady-state kinetics, which provides a convenient means to assay the overall rate at which product is produced. The most relevant and interpretable data available from this approach is the ratio kcat/KM (or efficiency), which is a second order rate constant relating the duplex-bound polymerase and free dNTP to the rate limiting transition state for multiple turnover dNTP insertion. The lower limit of detection of the assay is kcat/KM ∼ 1.0 × 103 M-1min-1. The individual values of kcat and KM are reported in the Supporting Information but are generally not discussed as their interpretation depends on the rate limiting step of the reaction, which may vary among the different polymerases and unnatural base pairs. The absolute efficiencies of synthesis (correct pair and all possible mispairs) and extension (correct pair and select mispairs) are listed in Tables 1 and 2, respectively for the A family polymerases, and Tables 4 and 5, for the B and X family polymerases. While the synthesis of all possible mispairs was characterized, we only characterized the extension of the most efficiently synthesized mispairs in order to approximate the overall fidelity of replication. In cases where no mispair is efficiently synthesized, the mispair with dA was characterized as it is commonly one of the most problematic.11,16,52-54 The overall fidelity is calculated as the product of the synthesis and extension fidelities. It is important to note that unless otherwise noted, the calculated fidelities are minimum fidelities, meaning the rate of correct unnatural base pair synthesis and/or extension, relative to the most competitively synthesized and/or extended mispair. Tables 3 and 6 summarize the most relevant information, including the efficiencies of synthesis and extension, and the overall fidelity for the replication of each unnatural base pair by the A family polymerases (Table 3) or the B and X family polymerases (Table 6). To facilitate comparison of the mesophilic and thermophilic polymerases in the following sections, it is convenient to compare the second order rate constants after normalization by the efficiencies for natural synthesis under identical conditions. Throughout, heteropairs (natural or unnatural) are denoted as dX:dY, with dX indicating the nucleotide in the primer strand and dY indicating the nucleotide in the template strand.</p><!><p>To provide a reference for the unnatural base pairs, we first measured the steady-state rates at which dATP or dGTP is inserted opposite a template dT. Kf inserts the correct and incorrect triphosphate with second order rate constants of 3.2 × 108 M-1min-1 55 and 5.7 × 104 M-1min-1, respectively, resulting in a fidelity under these conditions of 5.6 × 103. Preferential extension of a correct primer terminus is also critical for the high fidelity replication of DNA, thus we determined the steady-state rates of dA:dT pair and dG:dT mispair extension by the insertion of dCTP opposite a template dG. Kf extends the dA:dT pair and dG:dT mispair with second order rate constants of 1.7 × 108 M-1min-1 55 and 4.8 × 105 M-1min-1, respectively, resulting in a fidelity in this context of 350. The overall fidelity, which is the selectivity for the synthesis and extension of dA:dT, relative to the dG:dT mispair in this strand context is 2 × 106.</p><!><p>The synthesis of the unnatural base pairs, and to some extent their extension by Kf have been reported previously11-14 and are included here to facilitate comparison. At 25 °C, Kf synthesizes the dPICS self pair with a second order rate constant of 2.4 × 105 M-1min-1, but no extension of the nascent self pair is detected. The only mispair synthesized with a significant rate results from the insertion of dTTP, which results in a minimal synthesis fidelity of 20.13 Similarly, as mentioned above, Kf synthesizes the d7AI self pair with reasonable efficiency (kcat/KM = 2.2 × 105 M-1min-1), while mispairs are synthesized only inefficiently, resulting in a synthesis fidelity of 37; however, neither the dPICS nor the d7AI self pair is extended at a detectable rate. In contrast, Kf not only synthesizes the d3FB self pair with a high efficiency (2.1 × 106 M-1min-1) that is only 150-fold reduced relative to the efficiency for natural base pair synthesis, but it also extends the self pair with a reasonable efficiency (3.3 × 105 M-1min-1).11 Mispair synthesis with dGTP or dCTP is inefficient, although dATP and dTTP are inserted opposite d3FB slightly more efficiently, reducing the fidelity of Kf-mediated d3FB self pair synthesis to only ∼4. The mispair with dA is extended with a kcat/KM of 6.3 × 104 M-1min-1, resulting in an extension fidelity of ∼5. Combined, these single step fidelities result in an overall fidelity of 20 for Kf-mediated d3FB self pair replication.</p><p>The d5SICS:dMMO2 pair is the only predominantly hydrophobic heteropair we have identified to date that is efficiently synthesized and extended by a polymerase in both possible strand contexts.12 With d5SICS in the template, Kf inserts dMMO2TP with a second order rate constant of 3.6 × 105 M-1min-1, and it inserts dGTP, d5SICSTP, dATP, and dTTP, with low to moderate efficiency, resulting in a synthesis fidelity of ∼3. In the opposite context, Kf inserts d5SICSTP opposite dMMO2 with the remarkable efficiency of 4.7 × 107 M-1min-1, which is only 7-fold slower than natural synthesis under identical conditions. The only mispairs synthesized with any efficiency are those with dA and dMMO2, but the correct heteropair is still synthesized with a fidelity of 390.</p><p>Extension of dMMO2:d5SICS and d5SICS:dMMO2 (referring to primer:template) by Kf is also efficient, with a kcat/KM of 1.9 × 106 M-1min-1 and 6.7 × 105 M-1min-1, respectively. The extension of the most efficiently synthesized mispair, dG:d5SICS, is not efficient; however, dTTP is inserted with reasonable efficiency, and the resulting dT:d5SICS mispair is also extended with moderate efficiency, which limits the extension fidelity to only 4.8.12 For dMMO2 in the template, dATP and dMMO2TP insertion results in the most efficiently synthesized mispairs, but extension of the self pair is barely detectable and extension of the dA:dMMO2 mispair proceeds inefficiently, resulting in an extension fidelity of 15. Taken together the single step fidelities result in overall minimum fidelities of 130 (for d5SICS in the template) to 7100 (for dMMO2 in the template).</p><!><p>T7 inserts dATP and dGTP opposite dT in the template with second order rate constants of 1.5 × 107 M-1min-1 and 9.8 × 102 M-1min-1, respectively, resulting in a fidelity under these conditions of 1.5 × 104, which is approximately 3-fold higher than that observed for Kf. T7 extends the dA:dT and dG:dT pairs with second order rate constants of 1.3 × 107 M-1min-1 and 1.8 × 104 M-1min-1, resulting in a fidelity in this context of 720, 2-fold higher than that observed for Kf. Overall, the selectivity for the synthesis and extension of the dA:dT pair, relative to the dG:dT mispair in this strand context is 1.1 × 107.</p><!><p>Despite its homology with Kf, T7 does not synthesize or extend the dPICS or d7AI self pairs; however, it does synthesize and extend the d3FB self pair with second order rate constants that are only 40- and 2-fold reduced relative to Kf. Also, similar to Kf, T7 competitively inserts dATP and dTTP opposite d3FB, with the rates of both actually exceeding that of d3FBTP. Despite their reasonably efficient rates of synthesis, neither the dA:d3FB or dT:d3FB mispairs are efficiently extended, resulting in an overall fidelity of 29.</p><p>Generally, T7 recognizes the d5SICS:dMMO2 heteropair better than it recognizes any of the self pairs. T7 inserts dMMO2TP opposite d5SICS with a second order rate constant of 3.7 × 104 M-1min-1, which is virtually identical to that for the insertion of d3FBTP opposite d3FB. Insertion of the natural dNTPs or d5SICSTP is barely detectable, resulting in a fidelity in this context of 22. With dMMO2 in the template, T7 inserts d5SICSTP with an efficiency of 1.1 × 106 M-1min-1, which remarkably, is only 14-fold less efficient than that for a natural base pair. T7 does not synthesize any mispairs except those with dA and dMMO2, and these are only synthesized inefficiently, resulting in a synthesis fidelity of 19.</p><p>Extension of d5SICS:dMMO2 and dMMO2:d5SICS by T7 proceeds with high efficiency (kcat/KM = 1.1 × 106 M-1min-1 and 5.4 × 105 M-1min-1, respectively). The most efficiently synthesized mispair with either d5SICS or dMMO2 in template is that with dA; however, neither is extended at a detectable level. Although the dMMO2 self pair is synthesized less efficiently than the dA mispair, it is extended more efficiently. These rates result in a selectivity for correct pair extension of at least 540 with d5SICS in template and 190 with dMMO2 in template. Remarkably, the overall fidelities for T7-mediated unnatural heteropair synthesis are at least 1.2 × 104 and 1.4 × 104 in its two strand contexts.</p><!><p>At 50 °C, Taq polymerase inserts dATP and dGTP opposite dT in the template with second order rate constants of 4.0 × 107 M-1min-1 and 2.5 × 104 M-1min-1, resulting in a fidelity of 1.6 × 103. Taq extends the dA:dT and dG:dT pairs with second order rate constants of 1.9 × 107 M-1min-1 and 7.4 × 103 M-1min-1, resulting in a fidelity in this context of 2.6 × 103. Overall, the selectivity for the synthesis and extension of dA:dT, relative to the dG:dT mispair in this strand context is 4.2 × 106. (It is not appropriate to compare the fidelities of this thermophilic polymerase at 50 °C with those of the mesophilic polymerases at 25 °C, due to the temperature differences and because the temperature at which Taq is assayed is further from its optimal temperature of ∼75 °C50).</p><!><p>Despite the temperature differences, self pair recognition by Taq is similar to that by Kf, with second order rate constants that only vary between 1.2 × 104 M-1min-1 and 2.6 × 104 M-1min-1, and extension observed only in the case of the d3FB self pair (kcat/KM = 1.6 × 104 M-1min-1). Surprisingly, Taq extends the d3FB self pair with the same efficiency that it synthesizes it. Taq does not efficiently synthesize any mispairs with either dPICS or d7AI in the template, resulting in fidelities of 22 and 2.9, respectively, but with d3FB in the template, Taq actually inserts dATP and dTTP opposite d3FB faster than it inserts d3FBTP. Nonetheless, as also observed with T7, Taq does not extend either the dA:d3FB or dT:d3FB mispair, resulting in a minimum overall selectivity for d3FB self pair synthesis of 2.6.</p><p>Taq inserts dMMO2TP opposite d5SICS with a second order rate constant of only 8.6 × 104 M-1min-1, but it does not efficiently insert d5SICSTP or any natural dNTP, resulting in a heteropair synthesis fidelity in this context of 2.3. As with the other polymerases, Taq synthesizes the heteropair more efficiently in the opposite strand context, inserting d5SICSTP opposite dMMO2 with an efficiency of 3.5 × 106 M-1min-1, which is only 11-fold lower than that for a natural base pair. While Taq does not insert dCTP or dGTP opposite dMMO2, it does insert dATP, dTTP, and MMO2TP, but it does so at least 32-fold slower than it inserts the correct unnatural triphosphate.</p><p>Taq extends dMMO2:d5SICS and d5SICS:dMMO2 with moderate efficiencies (kcat/KM = 2.2 × 104 M-1min-1 and 6.4 × 104 M-1min-1). The most efficiently synthesized mispair with d5SICS in the template, dG:d5SICS, is not extended at a detectable level, nor is the self pair or the mispairs with dA, while the mispair with dT is extended only very inefficiently, resulting in an extension fidelity in this context of 4.7. The most efficiently synthesized mispair with dMMO2 in the template, dA:dMMO2, is extended at a barely detectable rate, resulting in an extension fidelity of 38. These efficiencies of correct pair and mispair synthesis and extension result in a minimum overall fidelity for dMMO2:d5SICS and d5SICS:dMMO2 replication of at least 51 and 1200, respectively.</p><!><p>At 50 °C, the family B polymerase Vent inserts dATP and dGTP opposite dT in the template with second order rate constants of 1.2 × 107 M-1min-1 and 1.7 × 104 M-1min-1, resulting in a fidelity of 710. Vent extends the dA:dT and dG:dT pairs with second order rate constants of 2.9 × 106 M-1min-1 and 1.5 × 104 M-1min-1, resulting in a fidelity in this context of 190. Overall, the selectivity for the synthesis and extension of the dA:dT, relative to the dG:dT mispair in this strand context is 1.3 × 105.</p><!><p>Vent synthesizes each self pair with reasonable efficiency (between 2.2 × 104 M-1min-1 and 6.4 × 104 M-1min-1). While the absolute efficiencies are similar to those of the A family polymerases, relative to the synthesis of a natural base pair, synthesis of the dPICS and d7AI self pairs is significantly better with Vent than with the A family enzymes. Vent does not insert dCTP opposite dPICS, but does insert the other natural dNTPs with moderate efficiencies, resulting in a synthesis fidelity of 4. The insertion of dGTP and dCTP opposite d7AI is barely detectable, but dATP and dTTP insertion is slightly more efficient. Opposite d3FB, Vent does not insert dGTP or dCTP efficiently, but does insert dTTP and dATP with moderate efficiencies. Nonetheless, none of the synthesized mispairs are efficiently extended, resulting in overall fidelity of 3.7 for Vent-mediated d3FB self pair synthesis.</p><p>Vent synthesizes the unnatural heteropair much more efficiently than it synthesizes the self pairs. Insertion of dMMO2TP opposite d5SICS proceeds with a second order rate constant of 1.3 × 106 M-1min-1, an efficiency only 9-fold lower than that of a natural base pair. dATP, dGTP, and dTTP are inserted with only moderate efficiencies, but the self pair of d5SICS is synthesized with a slightly greater efficiency which limits fidelity in this context to 2.6. As with the other polymerases, d5SICSTP insertion opposite dMMO2 is even more efficient, in fact, it is as efficient as the insertion of dATP opposite dT in the same sequence context. Unlike the A family polymerases, which preferentially insert dATP opposite dMMO2, Vent most efficiently inserts dTTP and dMMO2 opposite dMMO2 in the template, while dCTP and dATP are inserted less competitively, and dGTP the least competitively. This results in a fidelity of 83. Also unlike the other polymerases, Vent cannot efficiently extend the heteropair in either strand context (kcat/KM ∼ 7 × 103 M-1min-1), nor can it extend any of mispairs of these two nucleotide analogs with the natural nucleotides.</p><!><p>At 50 °C, the family B polymerase Therminator inserts dATP and dGTP opposite dT in the template with second order rate constants of 1.5 × 108 M-1min-1 and 1.9 × 106 M-1min-1, resulting in a fidelity of 79, which is significantly lower than that of the other polymerases. Therminator extends both the dA:dT and dG:dT pairs efficiently with second order rate constants of 3.4 × 106 M-1min-1 and 7.1 × 105 M-1min-1, resulting in a fidelity of only 4.8. The low fidelities at each step result in an overall fidelity for natural DNA synthesis of only 380.</p><!><p>At 50 °C, Therminator synthesizes each of the self pairs with remarkable a efficiency, varying only between 2.3 × 106 M-1min-1 and 4.2 × 106 M-1min-1. The efficient synthesis results in large part from a small apparent KM for triphosphate binding (Supporting Information). However, Therminator also efficiently synthesizes a wide range of mispairs. Opposite both dPICS and d7AI in the template, Therminator inserts the natural dNTPs with reasonable efficiency, limiting fidelity to 2.5 and 4.6, respectively. Opposite d3FB, Therminator inserts dCTP, dGTP, and dTTP with moderate efficiencies, but as with all of the other polymerases except Kf, it inserts dATP more efficiently than it inserts d3FBTP.</p><p>The d5SICS:dMMO2 heteropair is recognized by Therminator at 50 °C in both strand contexts with remarkable efficiencies that are only 3 to 4-fold reduced relative to natural synthesis. This is again due to a remarkably small apparent KM value (Supporting Information), as was also observed for self pair synthesis by Therminator. However, as with the self pairs, Therminator also efficiently synthesizes mispairs with both d5SICS and dMMO2. With d5SICS in the template, dCTP is inserted with moderate efficiency, while the remaining dNTPs are inserted with slightly higher efficiencies. The most efficiently synthesized mispair with d5SICS in the template is its self pair, which limits heteropair synthesis fidelity to 3.2. Opposite dMMO2, Therminator inserts dMMO2, dATP, and dTTP very efficiently, and dCTP and dGTP only slightly less efficiently. This results in a fidelity of heteropair synthesis of 5.5.</p><p>Like Vent, Therminator at 50 °C catalyzes the extension of d5SICS:dMMO2 and dMMO2:d5SICS less efficiently than it catalyzes their synthesis. However, with second order rate constants between 2 and 5 × 105 M-1min-1, the absolute efficiency of Therminator-mediated extension is significantly greater than that of Vent-mediated extension. With d5SICS in the template, the most efficiently synthesized mispairs are those with d5SICS and dG (although the mispairs with dA and dT are synthesized with nearly identical efficiencies). While extension of the self pair is inefficient, extension of the mispairs with dT and dG is reasonably efficient, and extension of the mispair with dA is only slightly less efficient. This results in a minimum selectivity for correct pair extension of 2. Mispairs with dMMO2 in the template are extended only inefficiently, resulting in an extension fidelity of 3.7. Synthesis and extension fidelities combine to yield an overall fidelity for the heteropair of 20 and 170 in its two strand contexts. Remarkably, this is only 2- to 19-fold less than the fidelity observed for natural synthesis.</p><!><p>Due to low activity, the family X polymerase pol β was assayed at its optimal temperature of 37 °C. At this temperature, pol β inserts dATP and dGTP opposite dT in the template with second order rate constants of 9.8 × 104 M-1min-1 and less than 1.0 × 103 M-1min-1, resulting in a fidelity of at least 98. Pol β extends the dA:dT and dG:dT pairs with second order rate constants of 2.1 × 105 M-1min-1 and 2.8 × 103 M-1min-1, resulting in a fidelity in this context of 75. Overall, the selectivity for the synthesis and extension of the dA:dT pair, relative to the dG:dT mispair in this strand context is at least 7.4 × 103. Relatively low fidelity pol β-mediated extension of a simple primer template (i.e. not a gapped substrate) has been reported previously,47,56,57 as has the high frequency of the dG:dT mispair.58</p><!><p>Pol β does not synthesize any of the unnatural pairs examined, except for the heteropair, and in this case, only by insertion of d5SICSTP opposite dMMO2, and even then, only at a barely detectable rate. Likewise, pol β also does not synthesize any mispairs, with the exception of both mispairs between dC and the nucleotides of the heteropair. In dramatic contrast, pol β extends all of the unnatural pairs examined with second order rate constants that vary from as efficient as that for natural base pair extension to only 20-fold less efficient. However, mispair extension is also efficient, resulting in fidelities of 5.8 to 12 for the self pairs and 2 and 18 for the heteropair (in its two strand contexts).</p><!><p>We continue to explore the potential of nucleotides bearing predominantly hydrophobic nucleobase analogs as part of an effort to expand the genetic alphabet. The replication of these analogs has been extensively examined with the family A polymerase Kf. However, our work with rat pol β and the d7AI self pair suggests that the recognition of predominantly hydrophobic base pairs may be polymerase specific. Indeed, this has been observed with other candidate unnatural base pairs,22-26 hydrophobic nucleobase isosters,27-31 and nucleotides with modified sugars or nucleobase substituents.17 However, there has been no detailed and systematic report of the kinetic rates with which different polymerases synthesize and extend unnatural base pairs. To explore the potential substrate repertoire of natural DNA polymerases with the predominantly hydrophobic unnatural base pair candidates, and to determine whether some aspects of replication are base pair-specific or polymerase-specific, we have characterized the efficiency and selectivity of five different DNA polymerases with four different unnatural base pairs (Figure 1), including the dPICS, d7AI, and d3FB self pairs and the d5SICS:dMMO2 heteropair. Two family A polymerases, one mesophilic (T7) and one thermophilic (Taq); two thermophilic family B polymerases (Therminator and Vent); and one family X polymerase (pol β) were selected for characterization. Along with our previous studies of Kf, the results of this study provide a rather broad survey of natural polymerase diversity. A direct comparison of kinetic data acquired at different temperatures is not straightforward; thus, for the purposes of this discussion, we focus on relative rates determined for each polymerase by normalizing the efficiencies for unnatural base pair synthesis and extension to the corresponding values for a natural base pair under identical conditions.</p><!><p>From a practical perspective, self pairs are particularly attractive for expanding the genetic alphabet, as they require DNA polymerases that efficiently and selectively catalyze the insertion of only a single unnatural dNTP and the subsequent extension of only a single unnatural primer terminus. Previous structural studies16 and kinetic studies with Kf11,13,14 suggest that the three self pairs examined represent two structurally distinct classes. The dPICS nucleotides self pair within duplex DNA via intercalation, which with Kf appears to facilitate synthesis but inhibit extension. Other large aromatic nucleobase analogs have been shown to pair by intercalation,59 and d7AI is also likely to pair in this manner. In contrast, and as would be expected based on the reduced surface area of the simple phenyl ring, structural studies have demonstrated that the d3FB nucleobases self pair within duplex DNA in an in-plane manner, which with Kf appears to facilitate extension. Interestingly, these differences in structure appear to be accommodated very differently by the different polymerases, as we find that self pair recognition is polymerase dependent.</p><p>The intercalative self pairs dPICS and d7AI are not synthesized or extended by T7 and are not synthesized by pol β. The behavior of T7 is consistent with that observed in previous studies using nucleobases of various sizes which suggested that T7 has greater structural discrimination than Kf.28 Likewise the data with pol β is consistent with previous reports that this enzyme does not synthesize base pairs between analogs with modified H-bonding functionality.22,60 Despite not being able to synthesize either self pair, pol β can extend both with rates that are only 5 to 20-fold less efficient than extension of a natural base pair. However, due to the generally low rates with which this enzyme replicates natural DNA, the absolute values of the rates remain modest. Somewhat higher efficiencies for the synthesis of the dPICS and d7AI self pairs are observed with Taq, although not as high as with Kf, and like the other family A polymerases, Taq does not extend either self pair with a rate sufficient to detect. Generally, relative to natural synthesis, the family B polymerases synthesize the self pairs with the greatest efficiency. Therminator extends them with high efficiency, as well, which is consistent with its ability to synthesize and extend pairs between orthogonal H-bonding nucleobases.24 However, Therminator is also the only enzyme that synthesizes mispairs of either dPICS or d7AI with rates that are competitive with those for unnatural base pair synthesis. In contrast, while Vent synthesizes the self pairs only 200- to 500-fold slower than natural synthesis, it is virtually unable to extend any of the self pairs once synthesized. This behavior also is consistent with previous reports that Vent is unable to extend primers past an orthogonal H-bonding nucleotide analog in the template.24</p><p>While the d3FB self pair is neither synthesized nor extended with even moderate efficiency by Taq, and extended, but not synthesized by pol β (the same behavior pol β shows with all the unnatural base pairs examined), it is synthesized and extended by all of the other polymerases with efficiencies only 4- to 500-fold reduced relative to a natural base pair. Nonetheless, when efficiencies and fidelities of both synthesis and extension are considered, it is apparent that Kf is uniquely efficient at recognizing the d3FB self pair. However, even though Therminator inserts dATP opposite d3FB more efficiently than its cognate unnatural triphosphate, the self pair is replicated with high efficiency and reasonable fidelity, as the mispair is not efficiently extended. Moreover, while T7 synthesizes the self pair with only marginal efficiency, it does so with reasonable efficiency due to its very low rate of mispair synthesis and extension. Thus, both Therminator and T7 may also prove useful for replicating DNA containing the d3FB self pair, perhaps in the presence of a 3′ to 5′ exonuclease to remove any mispairs formed at the primer terminus.</p><!><p>The d5SICS:dMMO2 pair is the only unnatural heteropair we have identified to date that is efficiently synthesized and extended in both possible strand contexts. While no structural information is yet available, it is presumably packed in duplex DNA with the nucleobases interacting edge-on. In general, both synthesis and extension of this heteropair are efficient, but several interesting similarities and differences are apparent among the different polymerases. While Kf synthesizes the heteropair relatively efficiently in either strand context, the insertion of d5SICSTP opposite dMMO2 is significantly more efficient than synthesis of the heteropair by insertion of dMMO2TP opposite d5SICS. Polar recognition of the heteropair is also observed with the other family A polymerases. The family A polymerases differ regarding the insertion of dGTP opposite d5SICS, which is not catalyzed by T7, but which proceeds only 2- to 3-fold slower than insertion of dMMO2TP by Kf and Taq. Interestingly, the polar recognition of the heteropair is less pronounced with Vent and not observed with Therminator; each synthesizes the heteropair in both strand contexts with rates that are within 10-fold of those for a natural base pair. As with the self pairs, virtually no synthesis of the heteropair in either strand context is observed with pol β.</p><p>Kf extends the heteropair with at least reasonable efficiency in both strand contexts, only 100- to 200-fold slower than it extends a natural base pair. The other A family enzymes show divergent behavior: T7 extends the heteropair in either strand context better than Kf, only 10- and 20-fold slower than extension of a natural base pair, while Taq extends it worse, 300- to 900-fold slower than natural synthesis. As with extension of the self pairs, the two family B polymerases showed very different behavior; Therminator efficiently extends the heteropair in both strand contexts, only 10-fold slower than natural synthesis, while Vent extends it in both contexts 400- to 500-fold slower than natural synthesis. Although the absolute rates are modest, pol β extends the heteropair in both strand contexts virtually as efficiently as it extends a natural base pair. Interestingly, and unlike synthesis of the heteropair, extension of the heteropair is not strongly dependent on strand context with Kf, T7, Taq, or Therminator. However, strand context is more important with pol β, which extends the heteropair with dMMO2 in the template ∼10-fold more efficiently than with d5SICS in the template. Thus, while differences are apparent, many aspects of heteropair recognition appear to be conserved among the polymerases, suggesting that, unlike self pair synthesis, much of the recognition of the heteropair is polymerase-independent and is instead encoded within the nucleobases themselves. This is likely to include a well packed, edge-on interface between the nucleobases as well as sulfur and methoxy moieties appropriately positioned in the developing minor groove to accept H-bonds from polymerase H-bond donors (see below).</p><!><p>Structures of ternary complexes with primer/template duplex and triphosphate have been reported for the catalytic domain of Taq (KlenTaq),37-39 T7,36 and pol β.42-44,61 As a model of the Kf active site, we use the reported ternary complex of DNA polymerase I large fragment from a thermostable strain of Bacillus stearothermophilus (Bf), which has high sequence and structural homology to Kf.35,62 As a model for Therminator and Vent we use the reported apo-enzyme structure of Therminator with substrates modeled in,41 as well as the ternary complex of the B family polymerase from RB69.40 It is apparent from these structures that each polymerase adopts the standard DNA polymerase fold, which consists of 'thumb', 'palm', and 'fingers' subdomains.</p><p>The dPICS and d7AI self pairs are unique in that their intercalative mode of pairing is likely to induce structural distortions at the primer terminus. Thus, the different recognition of the dPICS and d7AI self pairs, relative to other unnatural pairs may be related to the different ability of the polymerases to accommodate these distortions as they develop at the primer terminus during dNTP incorporation or after they are fully manifest during extension. The polymerase structures reveal that the binding site for the base pair being synthesized is provided by a narrow slot formed by the flanking nucleobases on one side and the polymerase on the other. With the family A polymerases, the polymerase side of this binding slot is provided by an aromatic ring (F710, F667, and Y562, in BF, Taq, and T7, respectively) that tightly packs on the nucleobase of the incoming dNTP, and the methylene moiety of a glycine residue (G711, G668, and G527, in BF, Taq, and T7, respectively) that packs on the nucleobase of the templating base. (The template strand bends to expose the templating base to the glycine backbone.) We refer to this packing arrangement as aromatic/aliphatic. The data suggests that this packing motif contributes to efficient catalysis of dNTP insertion when the pairing nucleobases are planar and interact in an edge-on manner, resulting in the most efficient recognition of the d3FB self pair and the unnatural heteropair in either strand context. In contrast, the polymerase side of the binding slot in pol β, when bound to a simple primer-template, is provided by side chain methylene groups (D276 and K280), and in RB69 and Therminator, by side chain methylene and methyl groups (N564 and L561, and L491 and I488, respectively in RB69 and Therminator), arrangements that we refer to as aliphatic/aliphatic. The data suggest that relative to the aromatic/aliphatic motif, the aliphatic/aliphatic motifs may be better able to accommodate distortions associated with nucleobase intercalation.</p><p>While this potential structure activity relationship is interesting, there are clearly exceptions to these generalizations. For example, while Taq, Kf, and T7 all employ aromatic/aliphatic packing, T7 appears to be more sensitive to nucleobase distortions (as it does not synthesize the intercalative self pairs at all), and Taq appears to be less sensitive (as it synthesizes the intercalative self pairs as efficiently as it synthesizes the d3FB self pair). Furthermore, Vent (which by homology to the other B family polymerases is likely to employ aliphatic/aliphatic packing interactions via N494 and I498) is consistently the least proficient at extending the unnatural pairs, and pol β does not synthesize any of the pairs. Thus, packing at the primer terminus cannot fully account for the differences in recognition. The 'tightness' of the T7 active site might contribute to the unique behavior observed here as well as to the hypersensitivity of T7 to increased nucleobase size that was reported previously.28 It is unclear what other factors might contribute to the unique behaviors observed with Taq and Vent, but they may be related to the generally high fidelity observed with these polymerases. With pol β, the substrate specificity during dNTP incorporation is likely to be dominated by its unique H-bonding interactions with both the dNTP and template nucleobase (see below).</p><p>While these exceptions currently render the relationship between primer-template packing motif and unnatural base pair recognition speculative, it is interesting that pol β appears to be able to utilize both packing motifs. As mentioned above, pol β appears to employ the aliphatic/aliphatic motif when bound to a simple primer-template.61 However, when bound to a gapped substrate, a conformational change results in the packing of the primer-template with a aromatic(His34)/aliphatic motif.42 Interestingly, the fidelity of pol β is different with the simple primer-template and gapped substrates, suggesting that the different packing arrangements may indeed contribute importantly to substrate recognition.47</p><p>In addition to differences in packing interactions, the polymerases examined also show differences in the number of H-bonds that they make with their bound substrates. These H-bonds, between donors in the polymerase and acceptors within the developing minor groove of the substrates, are conserved in all natural nucleobases but appropriately positioned only within the context of a correct Watson-Crick base pair.63-66 Interestingly, the A family polymerases form more of these H-bonds than the B or X family polymerases. T7, Klentaq, and BF make seven, five, and four contacts, respectively, including H-bonds with both the primer terminus nucleobase and its partner in the template, while pol β and RB69 (and by inference, Therminator41) make only two, including only a single H-bond with the primer terminus nucleobase or its partner in the template strand. Fewer H-bonds in Therminator and pol β may impart the primer terminus with increased freedom to adopt structures required for efficient unnatural base pair synthesis or extension. As with the packing interactions described above, if these H-bonding interactions do contribute to efficient extension, the data suggests that they may have diverged in Vent, or that other interactions between Vent and its substrates may be more important.</p><p>Interestingly, pol β is the only one of the polymerases examined that also forms analogous H-bonds with the incoming dNTP and the templating nucleobase. We hypothesize that these unique interactions with the forming base pair make pol β hypersensitive to its structure and H-bonding topology, possibly explaining the enzyme's general inability to synthesize unnatural base pairs. This is consistent with the inability of pol β to synthesize pairs between nucleobases with orthogonal H-bonds22 or pairs where one or both nucleobases are nonpolar isosters.60</p><p>A primer terminus formed by a natural base pair is efficiently recognized by all DNA polymerases because its adopts an appropriate geometry, which is ensured by a combination of Watson-Crick interbase H-bonding and shape complementarity,67-69 and because it presents appropriately positioned H-bond acceptors.63-65 However, the structural analysis presented above suggests that the different polymerases evolved somewhat different mechanisms to recognize these termini, and that each polymerase may have a rather different potential unnatural substrate repertoire. This is likely the origin of the strongly polymerase-dependent self pair recognition. It is also possible that in the absence of suitably positioned H-bonds or shape complementarity, self pair incorporation and extension depend not on recognition per se, but rather, on the avoidance of the polymerase-dependent mechanisms that evolved to discriminate against natural mispairs, as suggested previously by Kuchta and co-workers.20,21 In contrast, while generally orthogonal to the natural nucleobases, the nucleobases of the d5SICS:dMMO2 heteropair appear to interact in a manner that better mimics a natural base pair, and/or avoids the structural and/or electrostatic checkpoints employed by polymerases to discriminate against mispairs between the natural nucleobases.</p><!><p>An unnatural base pair that is selectively and efficiently replicated by a DNA polymerase would significantly increase the scope of biotechnologies that either employ selectively modified oligonucleotides (i.e. for novel materials development1,2) or that rely on the selective amplification of unnatural oligonucleotides (i.e. for SELEX-based selections of aptamers, ribozymes, or DNA enzymes with novel functionalities70). Such in vitro applications are somewhat less demanding than in vivo applications in terms of efficiency and fidelity; for example, the unnatural dNTPs may be present in excess, or more time may be allowed for amplification. Moreover, for in vitro applications the unnatural base pair need only be recognized by the polymerase present. For these applications, the most promising base pair-polymerase combinations appear to be the d3FB self pair with Kf or T7, and the d5SICS:dMMO2 heteropair with Kf, T7, or Therminator (Table 3 and 6). The remarkably high fidelity with which T7 replicates DNA containing the self pair and the heteropair are particularly noteworthy. Thus, at least to help facilitate the development of such systems, these unnatural base pairs already appear to be sufficient, although continued optimization is expected to even further increase their utility.</p><p>The demands of efficiency and fidelity are much more exacting for any in vivo application. To ensure that the unnatural base pairs are stable in a genome, they should be efficiently recognized by any polymerase involved in the replication and or maintenance of the genome. It is thus critical to characterize the unnatural base pairs according to those whose recognition is polymerase-specific and those recognized more generally by different polymerases. Clearly, the self pairs may be placed in the former category and the heteropair, in the latter category. In fact, the d5SICS:dMMO2 heteropair is generally well recognized by all of the polymerases examined, suggesting that it might be well recognized by any replicative or repair polymerase. Our results indicate that the determinants of efficient replication are generally contained within the unnatural base pair itself. It is particularly noteworthy that the N-glycosidic sulfur and the C-glycosidic ortho methoxy substituents, which are likely to engage the polymerase from within the developing minor groove and which are so central to Kf recognition,12 also appear to be well recognized by different polymerases. The slow step with each polymerase examined is the insertion of dMMO2TP opposite d5SICS in the template. Not only is this expected to result in polymerase stalling, which might be expected to be problematic in vivo, but it also results in somewhat reduced fidelity relative to the other strand context. Current efforts are focused on the optimization of this step of unnatural base pair replication, and particularly on its optimization during T7-mediated replication.</p><!><p>The unnatural triphosphates and phosphoramidites of dPICS, d7AI, d3FB, dMMO2, and d5SICS were prepared as described,11-15 and the phosphoramidites were incorporated in to oligonucleotides using standard procedures. Kf and T7 polymerases were obtained from GE Healthcare (Piscataway, NJ); Taq, Therminator, and Vent polymerases were obtained from New England Biolabs (Ipswich. MA); and pol β was obtained from CHIMERx (Madison, WI).</p><!><p>The unnatural nucleotides were evaluated as substrates for DNA polymerases by measuring the initial rates at which a [γ—33P]-labeled primer-template, 5′—d(TAATACGACTCACTATAGGGAGAX), annealed to the 45-mer template, 5′—d(CGCTAGGACGGCATTGGATCGYTCTCCCTATAGTGAGTCGTATTA), was extended with varying concentrations of natural or unnatural nucleoside triphosphates. Each reaction included 40 nM—primer template, 0.3 to 1.2 nM enzyme, and Klenow reaction buffer (50 mM Tris-HCl, pH 7.5, 10 mM MgCl2, 1 mM DTT, and 50 μg/mL acetylated BSA) for Kf and T7 polymerases, ThermoPol reaction buffer (20 mM Tris-HCl, 10 mM (NH4)2SO4, 10 mM KCl, 2 mM MgSO4, and 0.1 % Triton X-100, pH 8.8) for Taq, Therminator, and Vent polymerases, or pol β reaction buffer (50 mM Tris-HCl, pH 8.7, 10 mM MgCl2, 100 mM KCl, 1 mM DTT, and 0.4 mg/mL acetylated BSA) for pol β. The reactions were initiated by adding a 5 μL 2 × dNTP solution to a 5 μL solution containing the polymerase and primer-template and incubating at 25 °C (Kf and T7), 50 °C (Taq, Therminator, and Vent), or 37 °C (pol β) for 2 to 10 min, and were then quenched with the addition of 20 μL of loading dye (95% formamide, 20 mM EDTA, and sufficient amounts of bromophenol blue and xylene cyanole). The reactions were analyzed by polyacrylamide gel electrophoresis and a Phosphorimager (Molecular Dynamics) was used to quantify gel band intensities corresponding to the extended and unextended primer. The measured velocities were plotted against the concentration of dNTP and fit to the Michaelis-Menten equation to determine Vmax and KM. kcat was determined from Vmax by normalizing by the total enzyme concentration. Each reaction was run in triplicate and standard deviations determined.</p>

PubMed Author Manuscript

Structural arrangement within a peptide fibril derived from the glaucoma-associated myocilin olfactomedin domain

Myocilin-associated glaucoma is a new addition to the list of diseases linked to protein misfolding and amyloid formation. Single point variants of the ~275-residue myocilin olfactomedin domain (mOLF) lead to mutant myocilin aggregation. Here, we analyze the 12-residue peptide P1 (GAVVYSGSLYFQ), corresponding to residues 326\xe2\x80\x93337 of mOLF, previously shown to form amyloid fibrils in vitro and in silico. We applied solid-state NMR structural measurements to test the hypothesis that P1 fibrils adopt one of three predicted structures. Our data are consistent with a U-shaped fibril arrangement for P1, one that is related to the U-shape predicted previously in silico. Our data are also consistent with an antiparallel fibril arrangement, likely driven by terminal electrostatics. Our proposed structural model is reminiscent of fibrils formed by the A\xce\xb2(1\xe2\x80\x9340) Iowa mutant peptide, but with different arrangement of molecular turn regions. Taken together, our results strengthen the connection between mOLF fibrils and the broader amylome and contribute to our understanding of the fundamental molecular interactions governing fibril architecture and stability.

structural_arrangement_within_a_peptide_fibril_derived_from_the_glaucoma-associated_myocilin_olfacto

Introduction<!>Peptide Fibrillization Experiments<!>Solid state NMR<!>NMR-related spin simulations<!>Molecular dynamics modeling<!>Isotope labeling strategy and fibril formation<!>2D finite-pulse radio-frequency-driven recoupling (2D fpRFDR) experiments reveal a single uniform structure<!>2D dipolar assisted rotational resonance (2D DARR) experiments reveal inter-residue proximities and exclude the S-shaped model<!>Dipolar recoupling NMR and 2D CHHC experiments are consistent with antiparallel \xce\xb2-sheets<!>Structural modeling<!>Discussion

<p>Amyloid formation is associated with over 50 human diseases, creating chemically stable, protease resistant nanostructures that form proteinaceous plaques in various tissues.1–4 Amyloids are unified by the cross-β structural motif, in which β-strands are oriented perpendicular to the fibrillar axis. Still, there is considerable structural diversity within this motif.5–8</p><p>Missense mutations in the myocilin olfactomedin (mOLF, Fig. 1A) domain comprise the strongest genetic link to early-onset open angle glaucoma9, 10 and disease-causing myocilin variants are associated with toxic misfolding.11–17 Data demonstrate amyloid formation by full-length myocilin variants in cells and by the isolated mOLF domain in vitro.18, 19 Analysis of the mOLF sequence by amyloid prediction software identified the 12-residue P1 peptide (G326AVVYSGSLYFQ337) within mOLF (Fig. 1A) as one with high amyloid propensity. Experiments including atomic force microscopy (Fig. 1B) and Thioflavin-T (ThT) fluorescence measurements demonstrated that P1 assembles into amyloid fibrils under conditions similar to those that promote mOLF fibrils (37 °C, pH 6.8, gentle rocking) and with similar morphologies.19 The similarities in peptide and protein assembly behavior led us to hypothesize that the P1 peptide is part of the core amyloid-forming region full-length mOLF.</p><p>The structure of P1 peptide amyloid fibrils is also interesting for understanding what molecular structures are possible in amyloid formation and how structure could be predicted. Just like P1 was discovered through analysis of the mOLF protein, fragment peptides of similar size to P1 derived from larger known amyloid-forming proteins have been identified. Examples include residues 16–22 of Alzheimer's amyloid-β (Aβ(1–40) and Aβ(1–42))20, 21, residues 20–41 of β2-microglobulin (K3)22, and residues 105–115 of transthyretin (TTR)23, 24. Aβ(16–22) adopts a linear β-strand conformation within the Aβ amyloid fibril structure, and K3 and TTR(105–115) both adopt linear β-strands within natively folded β2-microglobulin and transthyretin, respectively. These precedents support the interpretation that fragment peptides derived from β-strand regions of larger aggregated or folded molecules are prone to adopt β-strand conformations when they self-assemble on their own. The P1 peptide is different from these examples, however, in that residues 326–337 do not adopt a linear β-strand conformation within mOLF. These residues instead form a β-hairpin within a well-folded β-propeller (see Figure 1) with a β-turn at residues S331 and G332. Although β-hairpin configurations (Figure 2A) are compatible with amyloid formation, as has been demonstrated for the MAX1 family of designer peptides and proposed for Aβ protofibrils (pre-fibrillar aggregates)25–27, P1 would be (to our knowledge) the smallest peptide shown to form this structure. Coarse-grained molecular dynamics (cgMD) simulations of P1 assembly predicted that the P1 peptide could assemble into amyloid fibrils without adopting linear β-strand or β-hairpin conformations28. Simulations predicted U-shaped (Figure 2B) or S-shaped (Figure 2C) molecular conformations, with each β-strand oriented parallel to neighbors, reminiscent of fibril structures of amyloid-β peptides29, 30. Our study of P1 amyloid structure was motivated by the desire to differentiate between multiple structural predictions.</p><p>In this contribution, we gain insight into the arrangement of P1 by testing hypothesized structures of P1 amyloid fibrils by solid-state NMR. The data presented herein are most consistent with the cgMD-predicted U-shaped molecular conformation, but with molecules arranged into antiparallel rather than parallel β-sheets. Our proposed structural model is reminiscent of Aβ(1–40) Iowa mutant peptide fibrils, but with a difference in the orientations of molecular turn regions. Electrostatic attraction between the termini likely drives the antiparallel arrangement in the P1 fibril. Our results strengthen the connection between mOLF fibrils and the broader amylome, and contribute to our understanding of the fundamental molecular interactions governing fibril architecture and stability.31–37</p><!><p>Isotope-labeled and unlabeled P1 short peptides (>95% final purity) were synthesized by CPC Scientific (Sunnyvale, CA). The amyloid sample was prepared as previously reported.19, 28 Briefly, P1 peptide was first stored at room temperature as 5 mg/mL solution in DMSO. To generate amyloid fibrils, 500 μM peptide was dissolved into 10 mM Na2HPO4/KH2PO4 buffer at pH 7.2, which contained 200 mM NaCl plus 10 μM ThT, and the solution was incubated at 36 °C for 24−48 h. The products of three 4 mL reactions were combined for NMR experiments. Insoluble aggregates were directly packed into 3.2 mm solid-state NMR rotors through ultracentrifugation (280,000 × g, 30 minutes at 4 °C). For isotope-diluted experiments, 30% labeled and 70% unlabeled peptides were mixed in DMSO solution and then subjected to the same aggregation procedure.</p><!><p>All of the solid-state NMR experiments were performed on a Bruker narrow-bore 11.7 Tesla magnet (1H frequency of 500 MHz), equipped with a 3.2-mm HCN magic angle spinning (MAS) probe. The 2D fpRFDR,38, 39 2D DARR,40, 41 and 2D CHHC42 experiments use different mechanisms to reintroduce the dipolar coupling between 13C atoms and thus provide different structural information. For 2D fpRFDR experiments, the power of the π pulse on the 13C channel was adjusted to 33 kHz to match the duration (15.2 μs) of one-third of rotor period at 22 kHz MAS. For the 2D DARR experiments, continuous irradiation with power corresponding to 11-kHz nutation frequencies (same as the MAS spinning rate) in the 1H channel was applied during the exchange periods, which were set to 500 ms to detect inter-residue contacts. In the 2D CHHC experiments, 150 μs 13C-1H and 1H-13C cross-polarization periods and a 182 μs 1H-1H spin diffusion period (2 rotor cycles in 11 kHz MAS spinning) was used. Proton decoupling with two-pulse-phase modulation43 and a 1H radiofrequency field of 100 kHz was used in all of the 2D experiments. The signal averaging of 2D fpRFDR, 2D DARR, and 2D CHHC required 36 to 48 h to produce the spectra in Figures 4 and 5, and it was increased to 72 h for the 2D DARR experiment on the isotope-diluted sample. After getting the 2D spectra, non-linear fitting with a 3D Gaussian function was performed for every distinguished crosspeak to determine the chemical shifts, the linewidths and the peak heights. The intensities of peaks were calculated by integrating the 3D Gaussian functions of the crosspeaks or the diagonal peaks. For the weakest crosspeaks (V3 Cα/F11 Cα, V3 Cα/F11 Cδ, A2 Cβ/F11 Cϵ, and V3 Cγ/S8 Cβ) in the isotope-diluted 2D DARR spectrum, the signal-to-noise was too low to directly measure the peak intensity. For these crosspeaks, we estimated the intensities using the corresponding linewidths we measured in the non-diluted spectrum and assumed that the peak heights that were half the height of the noise in the isotope-diluted spectrum.</p><p>To quantify the effects of isotopic dilution on 2D DARR NMR crosspeaks (dilution ratios), the relative crosspeak intensity for pairs of atoms was measured in the spectrum of Sample A (Figure 4B). The intensity of each crosspeak was scaled to corresponding diagonal peaks using (IAB + IBA) / (IAA+ IBB). IAB and IBA are the intensities of crosspeaks on different sides of the diagonal and IAA and IBB are intensities of the corresponding diagonal peaks. Then, the relative intensities from the 2D DARR spectrum of Sample C (Figure 7C) and those from Sample A were calculated to get the dilution ratio.</p><p>PITHIRDS-CT44 and R2W45 experiments measured the distances between specific 13C atoms in the amyloid sample. PITHIRDS-CT experiments were performed with a MAS spinning rate of 12.5 kHz. The dipolar recoupling time was adjusted by the number of blocks of pulses (k1, k2, and k3 defined by Tycko44), and it was fixed to be between 0 and 61.4 ms in our measurements. Proton decoupling with 100 kHz 1H radiofrequency field using continuous wave was applied during PIRHIRDS recoupling and acquisition. PITHIRDS curves in Figures 5A required signal averaging of about 24 h. R2W experiments were performed according to Costa et. al.45 We set a fixed evolution time of 50 ms and variable MAS spinning rates, which vary from on-resonance to ±0.4 kHz off-resonance. The spectra with 0 ms evolution time at each spinning rate were also collected as reference. Each 1D spectrum in R2W required about 2 h signal averaging to provide sufficient signal-to-noise.</p><!><p>Spin simulations were performed using SPINEVOLUTION46 with the same conditions as in the experiments. The chemical shift anisotropy parameters of Phe11 CO and Ala2 Cβ were measured in static CP 1D spectra and were used in all of the simulations. For Phe11 CO, the δaniso is −75 ppm, and ηΩ is 0.75. For Ala2 Cβ, the δaniso is −18 ppm, and ηΩ is 0.89.</p><p>In the PITHIRDS-CT simulations, a linear eight-spin system was used to mimic the atom coordinates in the parallel β-sheet structure, which is a linear array of eight 13C spins separated by constant distance. For the R2W simulations, a two-spin system was chosen to better describe the relative atom positions in the antiparallel β-sheet model (Figure 5D). R2W simulations with different combinations of T2, zq and distances were performed, and the results are compared with experimental data to find the best fit (Figure S3).</p><!><p>The all-atom models were built by constraining β-strands with NAMD scripts.47 First, a single peptide was generated using standard β-strand backbone torsion angles (antiparallel β-sheets: ϕ =−139°, ψ = 135°; parallel β-sheets: ϕ = −119°, ψ = 113°). Second, the torsion angles of the turn region (S6, G7, S8) were manually adjusted to make the desirable conformation. The SGS region for the U-shaped model was similar to previous DMD simulations,28 while the SGS for the β-hairpin model was similar to 3-residue β-turns.48 Third, the U-shaped P1 peptides were placed in antiparallel alignment to form β-sheets for the U-shaped antiparallel model, and the β-hairpin P1 peptides were placed into syn alignment to form β-sheet layers for the non-native β-hairpin model. These β-sheet layers were optimized in NAMD to get the energy-minimized conformation. Finally, the U-shaped antiparallel layers (or the syn β-hairpin layers) were brought to close positions parallel to the fibril axes, and the inter-molecular sidechain constraints (Val3-Ser8) were used to guide their stacking. The multi-layer structural models were generated from the molecular dynamic simulation in NAMD.</p><p>To calculate the expected dilution ratio for atom pairs in Table 4, we counted the corresponding 13C atom contacts in the structural model. We counted the fraction of intra-molecular contacts in total contacts for every atom pairs, and then calculate the expected signal attenuation in the isotope-diluted sample (dilution ratio).</p><!><p>Peptide P1 was synthesized with or without 13C, 15N labels on specific residues. Labeling schemes were designed to differentiate between the hypothesized structures in Figure 2. For the P1 fibrils in Sample A (see Table 1), we applied uniform 13C labeling of residues near the N-terminus (Ala2 and Val3), the center of the primary structure (Ser8), and the C-terminus (Phe11 and Gln12) with the intention of employing 2D 13C-13C NMR to probe inter-residue proximities in the assembled structure. In Figure 2, black ovals with white letters are used to illustrate model-predicted relative positions of the sidechains that are 13C-labeled in Sample A. Figure 3 indicates model-predicted patterns of inter-residue "contacts" (inter-atomic proximities that could be observed by 2D NMR experiments) detectable through off-diagonal peaks between 13C-labeled sites on different residues (inter-residue crosspeaks). Specifically, the orange-colored squares indicate nearest-neighbor residues separated by the distance of 0.6 nm or less (minimum distance between one C atom on one residue and one C atom on the other) as would be necessary for detection of inter-residue contacts. The lightness of orange in Figure 3 indicates the fraction of each of these residue pairs predicted to contribute to observable 2D DARR crosspeaks. The native β-hairpin structure (Figures 2A and 3A) predicts contacts for Ala2/Phe11 and Val3/Ser8, the U-shaped parallel β-sheet predicts a contact for Val3/Phe11, and the S-shaped parallel β-sheet predicts a contact for Ser8/Gln12. Additional inter-residue contacts to those shown in Figure 3 would be expected if the structures modeled in Figure 2 correspond to subunits (protofilaments) that associate into larger structures. To obtain more specific constraints on the organization of molecular backbones, P1 fibrils were produced (Sample B) with 13C labels on the carbonyl (CO) of Phe11 and the Cβ methyl of Ala2. These labels were designed for compatibility with 13C-13C dipolar recoupling experiments.44 Sample C was created through fibrilization of P1 solutions of 30% peptide labeled as with Sample A and 70% unlabeled peptide. Sample C was designed to reveal the effects of isotopic dilution (co-assembly of labeled and unlabeled peptide) on 2D NMR spectra, which can be interpreted in terms of model-dependent intra- and inter-molecular 13C-13C couplings. Samples A, B, and C were generated de novo as described previously19, 28 and used directly in experiments.</p><!><p>Consistent with prior biophysical data, the 2D fpRFDR spectral signatures in Figure 4A of P1 fibril Sample A are consistent with amyloid. The 2D fpRFDR 13C-13C spectrum was collected with a time for 13C-13C dipolar coupling (1.45 ms) chosen to reveal crosspeaks between directly bonded 13C atoms. The colored horizontal and vertical lines in Figure 4A indicate spectral assignments (correspondence between NMR peaks and 13C-labeled sites) based on bonding patterns that are unique to each uniformly 13C-labeled amino acid. Each directly bonded pair of 13C atoms exhibits only one crosspeak, indicating a homogeneous sample and a structure that produces a single chemical environment (uniform positions of atomic neighbors and local torsion angles) per labeled site. The spectral crosspeaks have linewidths (0.6 – 1.7 ppm) consistent with those of amyloid-β fibrils49–52 and mostly larger than linewidths of peptide microcrystals.53, 54 We also calculated secondary chemical shifts for CO, Cα, and Cβ by comparing the measured chemical shifts to the values from the same amino acid in random-coil model peptides55. The negative secondary chemical shifts for Cα and CO, and positive secondary chemical shifts for Cβ in all the labeled residues except the terminal Gln12 (Table 2) indicate that most of the residues in P1 adopt β-strand conformation56 as expected in amyloids formed by short peptides.</p><!><p>To interrogate inter-atomic contacts up to distances of ~0.6 nm, we performed 2D DARR experiments with 500 ms mixing time40, 49–51 on Sample A. Due to the longer mixing time, the 2D DARR spectrum (Figure 4B) reveals more crosspeaks than the 2D fpRFDR spectrum (Figure 4A), including inter-residue crosspeaks corresponding to atoms on distinct 13C-labeled residues (see Figures 2 and 3). Extensive crosspeaks from the labeled residues (Ala2, Val3, Ser8, Phe11, Gln12) detailed in Table 2 lead to the inter-residue contacts in Figure 4C. These results indicate that the N- and C-termini of the peptide are close in space (Ala2/Phe11, Ala2/Gln12, Val3/Phr11, and Val3/Gln12 contacts) and that Val3 is close to Ser8. These results do not support the S-shaped parallel model (Figure 2C and Figure 3C). We did not detect contacts between Ser8 and Phe11 or Gln12, as the model predicts, and none of the observed inter-residue contacts were predicted by the model. For the native β-hairpin (Figures 2A and 3A) and the U-shaped parallel β-sheet (Figures 2B and 3B) models, some model-predicted contacts corresponding to 13C-labeled sites were observed because their NMR spectral frequencies were resolved. However, the contact charts in Figure 3A and 3B each predict fewer inter-residue contacts than were observed experimentally, necessitating the further experiments described subsequently.</p><!><p>To further define the inter-molecular arrangement of P1 backbones, the 13C-13C dipolar recoupling techniques PITHIRDS-CT44 and R2W45 were applied to Sample B and the 2D CHHC technique57, 58 was applied to Sample A. The PITHIRDS-CT experiment produces a decay of NMR peak intensity that is sensitive to spatial proximity in the range of 0 to ~0.6 nm between equivalent labeled sites on adjacent molecules. This distance range encompasses 0.5 nm, the distance between backbones of neighboring β-strands in a β-sheet. The R2W technique can probe distances between 13C atoms with different peak frequencies (distinct sites), such as the Ala2 Cβ and Phe11 CO labeled cites in Sample B.45 Finally, the 2D CHHC experiment exhibits crosspeaks between labeled Cα sites within uniformly 13C-labeled residues when corresponding Hα atoms are separated by 0.3 nm or less. Such proximity between Hα atoms is a signature of an antiparallel β-sheet.57</p><p>The PITHIRDS-CT experiment on Sample B was performed to test for the arrangement of β-strands into in-register parallel β-sheets, as predicted by the U-shaped and S-shaped models (Figures 2B and 2C, respectively). The dashed curves in Figure 5A bracket expected PITHIRDS-CT decays for in-register parallel β-sheets, as indicated by previous measurements on amyloid fibrils labeled with 13C at backbone CO or Ala Cβ sites.44, 49, 59 These curves correspond to simulated data for 8 13C atoms positioned as predicted for an-in register parallel β-sheet: in linear arrangements with constant nearest-neighbor 13C-13C distances of 0.5 nm or 0.6 nm (see Figure S1). The measured decays in Figure 5A indicate weaker 13C-13C dipolar couplings between equivalent labeled sites and a structure that is not composed of in-register parallel β-sheets. As explained in Figure S2, the non-negligible measured PITHIRDS-CT decays in Figure 5A appear to be consistent with a 13C-13C distance of approximately 0.7 nm with precise distance depending on the 3D arrangement of atoms (see Figure S2A and S2B). However, when 13C-13C dipolar couplings between labeled sites are this weak, the apparent inter-atomic distance can be significantly affected by couplings to natural-abundance 13C atoms at nearby unlabeled sites (Figure S2C)60. In addition, PITHIRDS-CT decays can be affected by imperfect compensation of the pulse sequence for transverse spin relaxation44, 59. Hence, we do not believe that weak PITHIRDS-CT decays can be reliably interpreted in terms of specific structural features beyond indicating that labeled sites are separated by more than ~0.6 nm.</p><p>We conducted R2W measurements to constrain the distance between the Cβ atom of Ala2 and the CO atom of Phe11, which are both 13C-labeled in Sample B. This measurement assesses the dephasing (loss) of 13C NMR signal intensity when the magic angle spinning rate is near the difference in NMR peak frequency between the two labeled sites (ωMAS ~ ωCO − ωCβ). The dependence of signal loss (due to 13C-13C dipolar coupling) on the resonance mismatch (ωMAS − ωCO + ωCβ) indicates an inter-atomic distance of 0.57 ± 0.01 nm (Figure 5B). Because two parameters affect the theoretical dependence of signal loss on resonance mismatch (the zero-quantum relaxation time and the inter-nuclear distance) a two-parameter optimization was conducted accordingly using spin simulation results to get the best fit of data (Figure S3). This result supports the interpretation that P1 β-strands are organized into antiparallel β-sheets.</p><p>To further constrain the antiparallel arrangement of β-strands within the P1 fibril, we performed 2D CHHC measurements. The 2D CHHC spectrum in Figure 5C was collected with 182 μs 1H-1H dipolar coupling, and the observation of a crosspeak between the Cα sites of Val3 and Phe11 indicates that the Hα atoms of these two residues are separated by 0.3 nm or less. This result confirms that P1 β-strands are arranged into antiparallel β-sheets or β-hairpins (Figure 5D). However, the result is not consistent with the native β-hairpin conformation (Figure 2A). Figures S4A–C illustrate that the 0.3 nm distance between the Val3 Hα and Phe11 Hα atoms is not predicted within the native β-hairpin and cannot be rationalized through reasonable arrangements of native β-hairpins into β-sheets. In addition, Figure S5, a 2D CHHC spectrum collected on Sample C, shows that the Val3 Cα/Phe11 Cα crosspeak is attenuated with isotopic dilution, indicating that the atoms that contribute to this crosspeak are on neighboring molecules. Curiously, the 2D CHHC spectrum also shows a crosspeak between F11 Cα and Q12 Cα though the corresponding Hα atoms should be separated by about 0.4 nm in a β-strand. This crosspeak suggests that terminal Q12 backbone torsion angles may depart from those of β-strands. Alternatively, the unexpected sharpness and strength of the Q12 Cα signal could have made this crosspeaks more easily detected despite an F11 Hα – Q12 Hα distance that is longer than 0.3 nm. We did not detect a crosspeak between A2 Cα and V3 Cα in the 2D CHHC spectrum despite these residues also being adjacent in the primary structure.</p><!><p>Although the NMR data in Figures 4 and 5 reveal inconsistencies with all of the hypothesized models in Figure 2, two of the models can be modified for harmony with the data. To summarize, we observed the following main constraints regarding the P1 fibril: (a) a single chemical environment for each 13C-labeled site indicating a homogeneous fibril, (b) an antiparallel arrangement β-strands within β-sheets, (c) close inter-molecular proximity between Val3 and Phe11 backbone atoms, and (d) inter-molecular contact between sidechains of Val3 and Ser8. Based on AFM characterization (Figure 1B),19 where the height of fibril was measured to be ~1 nm, while the widths of the fibrils were much larger (varying between 10 and 20 nm), we infer that the proposed fibril structures would contain stacking of multiple layers to form mature fibrils.</p><p>We considered a modification of the U-shape model we predicted by cgMD (Figure 2B). Figure 6A shows an antiparallel β-sheet fibril that retains the U-shaped secondary structure. In this model, the Val3 backbone is within hydrogen bonding distance of that to Phe11, the Val3 sidechain is at the opening of the "U" and the Ser8 sidechain is at the bottom turn of the "U". To rationalize the inter-molecular Val3-Ser8 contact, the U-shape β-sheets were stacked top-to-bottom (Figure 6B). Analysis of the multi-layer stacked structural model (Figure 6B) for predicted DARR contacts (Figure 6C), in which the orange color represents observable contact between residue pairs, is consistent with our experimental 2D DARR data (Figure 4B). Although we modeled only 4 U-shaped layers, additional layers would be necessary for agreement with the fibril dimensions measured by AFM (10–20 nm).</p><p>Next, we considered a model with P1 molecules in a β-hairpin conformation, one that largely preserves the native structure of P1 within the mOLF crystal structure (Figure 1 and S4A). We changed the 2-residue β-turn (Ser6-Gly7) in the native β-hairpin into a 3-residue β-turn (Ser6-Gly7-Ser8), with torsion angles of the 3-residue β-turn estimated by comparing with similar structures48 (see Figure S6A). Thus, the two β-strands in the hairpin model are formed by Val3-Tyr5 and Leu9-Phe11, and the β-strands are arranged into an antiparallel β-sheet by bringing Val3 and Phe11 backbones within hydrogen bonding distance (Figure S6B, C). Following the terminology of Leonard et. al.26 and Nagy-Smith et. al.,25 which defines the terms "syn" and "anti" to describe whether turns of β-hairpins are oriented in the same or opposite directions, respectively, the formation of an extended P1 β-sheet fibril requires syn arrangement of the β-hairpins within each β-sheet (Figure S6C); β-hairpins aligned in an anti configuration (Figure S6B) cannot be stabilized by sufficient inter-molecular hydrogen bonds nor satisfy the short-distance contact between the Hα of Val3 and Phe11 (Figure S4B, C). The β-hairpin model resolves the inter-molecular Val3 - Ser8 sidechain contact through stacking of β-sheets with anti arrangements of adjacent β-sheets (Figure 6E). When we put the β-sheets formed by β-hairpins in multiple stacked layers (Figure 6E), the Val3 and Ser8 sidechain proximity happens between every layer. Like the U-shaped model, the stacked β-hairpin model predicts a DARR contact pattern that is consistent with the experimental data (Figure 6F). Again, additional layers would be required for agreement with the fibril dimensions detected by AFM.</p><p>Though the models in Figures 6B and 6E both satisfy the experimental constraints in Figures 4 and 5, they differ in their predictions of relative orientations of amino acids and the relative positions of amino acids within the same or adjacent molecules. These differences make it possible for us to differentiate between the models through 2D NMR spectra collected on our isotopically diluted sample (Figure 7). Intensities of crosspeaks in this spectrum are attenuated relative to diagonal peaks when they correspond to atoms on different molecules. Figures 7A and 7B illustrate model-dependent differences in the relative positions and orientations of the Val3 and Phe11 residues. The U-shaped model (Figure 6A–6C) predicts that each Val3 has two adjacent Phe11 residues on different molecules and one adjacent Phe11 residue on the same molecule. The β-hairpin model (Figure 6D–6F) predicts one adjacent Phe11 residue within the same molecule and one adjacent Phe11 residue on a different molecule, and no Val3/Phe11 contact from stacking between layers (Figure S7). Thus, the models predict different degrees of crosspeak attenuation between labeled sites on Val3 and Phe11. Close examination of residue orientations in the U-shaped model motivates greater degrees of isotopic dilution for crosspeaks between Val3 and Phe11 backbone atoms when compared to crosspeaks between Val3 and Phe11 sidechain atoms (Figure S8). Noting that 2D DARR crosspeaks are sensitive to 13C-13C dipolar couplings corresponding to distances of up to ~0.6 nm, the U-shaped model predicts that crosspeaks between backbone atoms (e.g., Phe11 Cα – Val3 Cα, Phe11 CO – Val3 Cα, Phe11 Cα – Val3 CO) would be exclusively between atoms on different molecules, whereas Phe11/Val3 sidechain crosspeaks (Val3 Cγ1/Cγ2 – Phe11 Cγ, Val3 Cγ1/Cγ2 – Phe11 Cδ1/Cδ2, Val3 Cγ1/Cγ2 – Phe11 Cϵ1/Cϵ2) would include contributions from inter- and intra-molecular dipolar couplings. In contrast, the β-hairpin model would predict uniform isotopic dilution effects for all crosspeaks between atoms on Val3 and atoms on Phe11. Thus, to differentiate between the two models, we considered the predicted dilution ratio from the two models compared to the calculated ratios from the experimental 2D DARR data (Figure 7C, Figure 8, and Table 4). The calculated effect of isotopic dilution on 2D NMR peak intensities was based on peak integrals tabulated in Figure S1. These results favor the U-shaped model. Taken together, although the U-shaped fibril model and the β-hairpin model both largely satisfy the available data, the U-shaped parallel fibril model is favored (Figure 9).</p><!><p>In this study, we characterized the unanticipated amyloid structure of a recently identified peptide fibril, that of the 12-residue P1 of the mOLF β-propeller (Figure 1) whose misfolding is associated with glaucoma. Prior to solving the mOLF structure, amyloid prediction servers (Waltz, AmylPred, and TANGO)61–63 converged on three peptide sequences, P1, P2 and P3 within mOLF with high propensity to form amyloid. Consistent with predictions, fibril formation by P1 and P3 (but not P2) were confirmed experimentally.19 Unlike with P3, which forms heterogeneous fibrils,28 incubation of dissolved P1 peptide at 37 °C and pH 7.2, without seeding, reproducibly produces fibrils with consistent morphology and ThT aggregation kinetic curves. Fibril morphology was akin to that observed for full-length mOLF aggregated under similar conditions.19 While effective computational tools exist that enabled discovery of P1 within the mOLF sequence, fibril structure predictions are still lacking. Despite having three seemingly reasonable structural predictions to guide our experiments, the NMR data in this paper describe a unique amyloid molecular structure.</p><p>Our data are most consistent with stacked U-shaped P1 protofibrils. Although the structure does not contain the canonical steric zipper with interdigitated side chains forming a dry core, this structure maximizes H-bonding interactions like other peptide amyloid structures30, 64, 65 and also has a buried hydrophobic region within each protofibril, suggesting it is a stable arrangement. Specifically, the opportunities for intermolecular main chain H-bonding span the residues within each peptide of the protofibril, including the S-G-S turn in which H-bonding was not constrained during our modeling process (Figure S9A). In forming the U shape, each peptide shields hydrophobic residues Val3 and Leu9 creating a hydrophobic pocket, which is then topped with π-stacked Tyr5 and Phe11 that form a hydrophobic patch on each protofibril (Figure 9A and Figure S9B). The protofibril surface containing the hydrophobic patch then nestles with a neighboring protofibril wherein the two Ser residues of the S-G-S turn form polar contacts with the peptide termini from the neighboring protofibril (Figure S9C). Comparatively, the non-native hairpin arrangement would seem less thermodynamically stable even though this model predicts stacking in a direction perpendicular to the β-strand backbone (see Figure 6E) commonly seen in amyloid structures51, 64–66. First, it has fewer options for H-bonding: the S-G-S turn within each hairpin is not in proximity of S-G-S turns in other peptides. Second, the hydrophobic residues are on the surface where they stabilize interactions among the protofibril layers and may be less effective at protecting the hydrophobic core from water compared to the U-shaped structure.</p><p>One reason for studying new amyloid-forming peptides is to explore the diversity of sequences capable of forming amyloid and their structural arrangements. Amyloid assembly can produce a variety of possible structures within a single sequence; amyloid structures are not uniquely determined by amino acid sequence.49, 67 The observation of new structural motifs for any peptide reveals structures that may be accessible to other peptides of similar size. The U-shaped antiparallel amyloid structure of P1 has not been observed previously, even if assemblies of other larger peptides are considered. To our knowledge, all other amyloid-forming peptides of this small length (e.g., the 11-mer peptide of TTR)24, 68 have been reported to form extended β-strands when they assemble. Our P1 amyloid model is reminiscent of a model reported by Qiang et. al. for an assembly of the Iowa mutant (D23N) of the Aβ(1–40) peptide.69 However, in this structure, the C-terminal 25 residues form a U shape wherein 14 residues form 2 antiparallel β-sheets associated along a sidechain-interdigitated steric zipper and the remaining residues form a connecting loop. Unlike our P1 model, the model of Qiang et. al. predicts that the turn regions of Aβ alternate in orientation along the long axis of the β-sheet. Since the P1 peptide is only 12 residues, each β-strand consists of only 4 or 5 residues, the loop is shorter, and the extent of sidechain interdigitation is less between β-sheets. These features result in a widened U shape that invites other such U shape fibrils to bind and satisfy the solvent-exposed hydrophobic surface formed by the Val3 and Phe11 sidechains (explaining our observed proximity between Val3 and Ser8). Even though the β-hairpin models we considered in this study were not favored by data on isotopically diluted samples, we could formulate a β-hairpin model with sidechain proximities that almost matched those for the final U-shaped antiparallel model. A similar β-hairpin model has been reported for the 20-amino-acid designed peptide MAX125, which form β-hairpins by taking advantage of a Pro-DPro hinge, suggesting that the β-hairpin structure may be more accessible for larger peptides. Previously, we predicted a parallel U-shaped fibril arrangement, using cgMD simulations (Figure 2B).28 These simulations did not consider the charges of the peptide termini,70 which would favor antiparallel over parallel β-sheets. Indeed, in the antiparallel model presented here, the electrostatic compatibility of adjacent termini adds stability to the antiparallel arrangement, forming H-bonding interactions with polar residues (see Figure S9). As suggested by Qiang et. al.,71 who also observed both antiparallel and parallel β-sheet structures for D23N-Aβ(1–40) fibrils, a parallel fibril arrangement might be more thermodynamically stable. Even though our de novo-formed P1 fibrils are antiparallel β-sheets, future studies could test whether this β-strand alignment persists upon seeding P1 fibrils. Importantly, even though short peptide fragments derived from amyloidogenic proteins (e.g., Aβ(16–22)21) often adopt antiparallel β-sheets, experimental results from numerous larger peptide and full-length protein fibrils (Aβ(1–40)30, 49, 51, Aβ(1–42)29, 64, 65, α-syn72, amylin73, HET-s74, β2m36, Ure2p75, SUP3576, PrP77) indicate hinged parallel fibril structures78. It is plausible that our cgMD simulations are relevant to the arrangement of full-length mOLF fibrils where terminal charges would be remote from the core fibril-forming segment. Further, the full-length mOLF fibril core may include discontinuous segments, e.g., from P3. More generally, as mOLF is a member of the large protein family eponymously named olfactomedin, the propeller domain may serve as a rich new source of amyloid-forming proteins and peptides from which to test generalities deduced from well-studied model systems associated with human disease and for the development of new peptide-based nanomaterials.</p>

PubMed Author Manuscript

Dithiocarbamates strongly inhibit carbonic anhydrases and show antiglaucoma action in vivo#

A series of dithiocarbamates was prepared by reaction of primary/secondary amines with carbon disulfide in the presence of bases. These compounds were tested for the inhibition of 4 human (h) isoforms of the zinc enzyme carbonic anhydrase, CA (EC 4.2.1.1), hCA I, II, IX and XII, involved in pathologies such as glaucoma (CA II and XII) or cancer (CA IX). Several low nanomolar inhibitors targeting these CAs were detected. X-ray crystal structure of hCA II adduct with morpholine dithiocarbamate evidenced the inhibition mechanism of these compounds, which coordinate to the metal ion through a sulfur atom from the dithiocarbamate zinc-binding function. Some dithiocarbamates showed effective intraocular pressure lowering activity in an animal model of glucoma.

dithiocarbamates_strongly_inhibit_carbonic_anhydrases_and_show_antiglaucoma_action_in_vivo#

Introduction<!>Chemistry<!>Carbonic anhydrase inhibition<!><!>X-ray crystallography<!>Intraocular pressure lowering in hypertensive rabbits<!>Conclusions<!>Chemistry<!>General procedure for the synthesis of compounds 1a-27a.30<!>CA inhibition<!>Co-crystallization and X-ray data collection of CA II complex<!>Structure determination of CA II drug complex<!>Animals and glaucoma induction

<p>Carbonic anhydrases (CAs, EC 4.2.1.1) are widespread zinc metalloenzymes found in higher vertebrates including humans.1–3 16 isozymes have been characterized to date, many of which are involved in critical physiological processes. They catalyze the following reaction: CO2+H2O ↔H++HCO3−.1–3 In humans, CAs are present in a large variety of tissues including the gastrointestinal tract, the reproductive tract, the nervous system, kidneys, lungs, skin and eyes.2,3 The different isozymes are localized in different parts of the cell with CA I and CA II, important isozymes in normal cells, being localized in the cytosol. 1–4</p><p>Many of the CA isozymes are important therapeutic targets with the potential to be inhibited to treat a range of disorders.1–7 CA II plays a role in bicarbonate production in the eye and is therefore a target for therapy of eye disease such as glaucoma.7–9 Indeed, CA inhibitors (CAIs) of the sulfonamide type such as dorzolamide DZA or brinzolamide BRZ are topically used antiglaucoma agents,7–10 whereas the older drugs, such as acetazolamide AAZ or dichlorophenamide DCP show the same action through systemic administration, which however leads to a wide range of side effects due to inhibition of the enzyme from other organs than the target one, i.e., the eye.11 CA XII, a transmembrane isoform with an extracellular active site, was shown to be overexpressed in glaucomatous patients eyes.12</p><p> </p><p>As some solid tumors grow in cancer patients, hypoxic regions are formed, particularly in the interior of the tumor.13 The gene expression profile of a hypoxic cancer cell is different from that of other cancer cells in a normally-oxygenated environment, i.e., in normoxic conditions.13–15 Under hypoxia, the distribution of CA isoforms is altered as compared with normoxic cells.13,14 As a result, CA isozymes IX and XII are overexpressed in hypoxic tumor cells, in a variety of solid tumors.13–15 Unlike other CAs, CA IX and CA XII are both extracellularly localized on hypoxic tumor cells. 13–15 These enzymes play various roles in tumorigenesis, by regulating pH inside and outside the tumor cell,15 interfering with phosphorylation of various proteins16 or by playing a role in the cell-cell adhesion.13–15 They therefore provide a target for cancer therapy because they are relatively specific to the hypoxic tumor cells and appear to be important in their survival and proliferation.15 Indeed, several antibodies targeting CA IX are in Phase III clinical development for the treatment of solid tumors (or for their imaging)17 whereas some small molecule inhibitors are also in advanced preclinical evaluation.15,18,19</p><p>The classical carbonic anhydrase inhibitors (CAIs) are the sulfonamides and their isosteres (sulfamates, sulfamides, etc).1–4 However, most of these compounds indiscriminately inhibit many of the 16 CA isoforms known to date in mammals.1–3 Thus, efforts have been made to find different CAIs, from the sulfonamide, sulfamate and sulfamide ones. Indeed, recently the coumarins were discovered as mechanism-based inhibitors which act as prodrugs and bind in a very different mode compared to sulfonamides and their isosteres,20 whereas some polyamines (such as spermine),21 as well as a range of phenols22 were also investigated and showed interesting such properties and novel mechanisms of inhibition. Among CAIs investigated to date there are also the inorganic anions, which coordinate to the zinc ion from the enzyme active site.1,23,24 Indeed, trithiocarbonate (CS32−) an anion similar to carbonate, has recently been investigated and shown to constitute a "lead" for novel CAIs.23 The X-ray crystal structure for the adduct of trithiocarbonate (CS32−), bound to hCA II has recently been reported (Fig. 1).23 The inhibitor binds to the Zn2+ in the hCA II active site in a slightly distorted tetrahedral geometry of the metal ion, occupying a position similar to that observed in the case of hCA II-bicarbonate complex.23 Trithiocarbonate was monocoordinated to the Zn(II) ion by means of one of the sulfur atoms. The same sulfur made a hydrogen bond to the OH of Thr199 whereas a second sulfur atom participated to another hydrogen bond with the NH group of the same amino acid residues, Thr199. This binding mode explains the low micromolar affinity of this inhibitor to many of the CA isoforms investigated to date.23 Based on this binding mode of a millimolar inhibitor, trithiocarbonate (CS32−), we hypothesized that compounds incorporating this new zinc-binding function, CS2−, may act as even stronger CAIs. Indeed, we have recently demonstrated in a preliminary communication that dithiocarbamates (DTCs), compounds possessing the general formula R1R2N-CS2−M+ act as highly efficient CAIs.25 Here we report the first detailed study of the DTCs as a class of potent CAIs, with a mechanism of action different of that of the sulfonamides. Furthermore, we prove that some of these highly water soluble compounds possess excellent intraocular pressure (IOP) lowering properties in an animal model of glaucoma, making them interesting candidates for developing antiglaucoma drugs.</p><!><p>DTCs are well known metal complexing agents and they also possess interesting biomedical and agricultural applications.26–29 Although this class of compounds (and their metal complexes) started to be used as fungicides more than 50 years ago,29 few studies investigated their interactions with metalloenzymes.27 Apart studies of DTCs as inhibitors of tyrosinase, a copper enzyme,27 only one such work investigated the inhibition of N,N-diethyl-DTC with bovine CA (bCA).28 By using Co(II)-substituted CA, Morpurgo et al.28 showed that the inhibitor does not extrude the metal ion from the enzyme active site (as it does with the copper ion from the tyrosinase active site)27 and that it binds to it, probably in a trigonal-bipyramidal geometry of Co(II). However, no other DTC were subsequently investigated for their interaction with CAs till our group reported that trithiocarbonate and related compounds containing the new zinc binding group (ZBG) found in it, i.e., CS2−, inhibit several CA isoforms in the low micromolar or submicromolar range.23 Here we extend those findings, showing that a wide range of DTCs incorporating various aliphatic and/or aromatic moieties at the nitrogen atom, acts as low nanomolar and even subnanomolar CAIs.</p><p>We prepared a series of 27 DTCs, of types 1a-27a, by the classical reaction27,30,31 of dithiocarbamoylation between primary/secondary amines 1-27, and CS2, in the presence of a base, which most of the time was NaOH, KOH but in the case of more basic amines, the amine itself can act as base (Scheme 1). As shown in Table 1, a large variety of R1 and R2 moieties are present in DTCs reported here, i.e., hydrogen, alkyl, aryl, aralkyl, hetaryl and cyclic such moieties, which lead to the generation of a wide chemical diversity in these compounds. Presumably, this should be reflected also in varied physico-chemical and biological properties of these DTCs. Compounds 1a-27a were characterized by physico-chemical standard procedures (IR, 1H- an 13C-NMR spectroscopy, MS) and were > 99 % pure, as determined by HPLC (see Experimental protocols for details)</p><!><p>Compounds 1a-27a were assayed32 for the inhibition of four physiologically relevant CA isoforms, hCA I, II, IX and XII. All of them are drug targets: hCA I, II and XII for ophthalmologic diseases, mainly glaucoma,1,10 whereas CA IX and XII for antitumor drugs/tumor imaging agents.1,15,17–19 Inhibition data with the sulfonamide, clinically used agent acetazolamide (AAZ) are also reported in Table 1, for comparison reasons.</p><p>The following structure-activity relationship (SAR) can be observed for the CA inhibition data with DTCs 1a-27a investigated here:</p><!><p>The cytosolic isoform hCA I was strongly inhibited by DTCs 1a-27a investigated here, with KIs in the range of 0.88 – 1838 nM. It may be observed that irrespective of the nature of R2, the DTCs prepared from primary amines (R1 = H) 1a-12a were highly effective hCA I inhibitors, with inhibition constants in the low nanomolar range (3.5 – 33.5 nM). On the contrary, the compounds prepared from secondary amines showed a more varied biological activity. Thus, the simple dimethyl- and diethyl-DTCs 13a and 14a were weak hCA I inhibitors, with KIs in the range of 699–790 nM. The same is true for the di-n-propyl derivative 17a (KI of 1838 nM). However, the cyclic derivative 15a, which differs from 14a by the cyclic structure and an extra carbon atom present in its molecule, is a subnanomolar hCA I inhibitor, showing a dramatic increase of potency of 823 times compared to 14a. It is also interesting to compare the potencies of 16a and 18a, which incorporate iso-butyl and n-butyl moieties (and are isomers), and differ by a factor of 44, unexpectedly, in favor of the compound with a branched scaffold. Increasing the length of the aliphatic chains to C6, as in 19a leads to a slight loss of potency compared to the most active compounds in the aliphatic series, which are 15a and 16a. Furthermore, the presence of hydroxyethyl moieties (instead of the ethyl ones) as in 21a, lead to a steady increase of potency, 21a being 85.9 times a better hCA I inhibitor compared to 14a. The compounds incorporating one alkyl and one aryl moiety at the nitrogen atom of the DTC function, such as 22a and 23a were also effective hCA I inhibitors (KIs of 39.6–69.9 nM), as were also the heterocyclic derivatives 24a-27a, some of which showed subnanomolar activity (the morpholine DTC 24a had a KI of 0.88 nM and was the best DTC hCA I inhibitor and also the best hCA I inhibitor ever described, as far as we know). Thus, many of these chemotypes explored here show excellent hCA I inhibitory activity, which range from the subnanomolar to the micromolar. Furthemore, many of the DTCs are much more effective as hCA I inhibitors compared to the sulfonamide AAZ, which has a KI of 250 nM against this isoform (Table 1).</p><p>The physiologically dominant cytosolic isoform hCA II also showed an interesting inhibition profile with DTCs 1a-27a. Thus, several primary DTCS (1a, 2a, 7a, and 10a) and several secondary ones (16a, 21a, 24a and 25a) were excellent hCA II inhibitors, with KIs in the range of 0.70 – 4.6 nM, being more effective (even one order of magnitude) than the clinically used sulfonamide AAZ (Table 1). It may be observed that these compounds incorporate aromatic, arylakyl, hetaryl, alkyl and hydroxyalkyl moieties substituting the nitrogen atom from the DTC moiety. Another rather large group of derivatives, such as 3a-6a, 8a, 9a, 12a, 15a, 17a-20a, 22a, 23a, 26a and 27a, were slightly less effective hCA II inhibitors, but still possessed a high efficacy, with KIs in the range of 13.5 - 55.5 nM. Again both primary and secondary DTCs are present in this subgroup. They incorporate various types of substituents, such as alkyl, aryl, aralkyl, and hetaryl ones. It is obvious that small structural changes in the DTC scaffold influence dramatically the biological activity. For example, for the aliphatic secondary DTCs, the isomeric pair 16a – 18a, which differ only by the nature of the aliphatic chain (iso-Bu moieties in the first compound and n-Bu in the second derivative) have KIs which differ by a factor of 53.6. The length of the alkyl chain also strongly influence activity, with compounds possessing a medium chain (e.g., 15a-20a) being more effective than the ones with shorter chains, such as 13a and 14a, which are rather ineffective as hCA II inhibitors (KIs of 3.1 – 6.9 μM). Also the glycine DTC 11a was a medium potency hCA II inhibitor, with a KI of 325 nM.</p><p>The tumor-associated isoform hCA IX was highly inhibited by the DTCs investigated here, with KIs in the range of 3.6 – 1413 nM. The simple aliphatic secondary DTCs 13a/14a and the bulky cyclic derivative 26a, were the least effective inhibitors (KIs of 0.714–1.413 μM), and four other compounds (11a, 15a, 17a and 18a) were effective, medium potency inhibitors, with KIs in the range of 50.3 – 70.4 nM. They incorporate the carboxyalkyl moiety present in glycine (11a), the five-membered aliphatic ring (from 15a) and 3- or 4-carbon atom n-alkyl chains (17a and 18a). All the remaining derivatives showed highly effective hCA IX inhibitory properties, with KIs < 30 nM. Thus, a rather high structural diversity (aliphatic, aromatic, aralkyl, hetaryl moieties) present in primary/secondary DTCs lead to highly effective hCA IX inhibitors, with minor structural changes drastically affecting enzyme inhibition. Many DTCs were more effective hCA IX inhibitors compared to acetazolamide (Table 1).</p><p>A rather similar SAR as the one discussed above for hCA IX, was observed for the inhibition of the second transmembrane isoform, hCA XII, with DTCs 1a-27a. Thus, 13a/14a and 26a, were the least effective inhibitors (KIs in the range of 169 – 1105 nM), whereas the remaining DTCs were highly effective hCA XII inhibitors, with KIs in the range of 0.78 – 31.7 nM (Table 1). Among the best hCA XII inhibitors (subnanomolar inhibition constants) were the di-isobutyl-DTC 16a and the piparazine-bis-DTC 25a. Again the main conclusion is that a large number of substitution patterns, incorporating varied moieties, lead to highly effective hCA XII inhibitors.</p><p>The DTCs investigated here showed a rather promiscuous inhibitory activity against all four CA isoforms described here, although each of these CAs had a different inhibition profile with all these compounds. For example, 15a was a subnanomolar inhibitor of hCA I and inhibited the remaining 3 isoforms with KIs in the range of 27.5–70.4 nM, having thus an acceptable selectivity ratio for the inhibition of hCA I over the remaining three CAs. 25a was a subnanomolar inhibitor of hCA II and XII, and inhibited hCA I and IX with higher KIs, of 12.6–37.5 nM. Compound 23a showed a rather good selectivity for inhibiting hCA XII over hCA I, II and IX (Table 1). It should be also mentioned that being negatively charged, these compounds show membrane impermeability33 which may be a favorable pharmacological property in vivo.</p><!><p>In order to explain the potent CA inhibitory properties of the DTCs, which are a new class of CAIs, we resolved the X-ray crystal structure of hCA II in complex with the very potent inhibitor 24a (KI of 0.95 nM). Compound 24a was well ordered and refined with an occupancy of 1.0 with B-factors that were comparable to the solvent within the active site (Table 2). The compounds is buried deep into the active site, displacing the catalytic zinc-bound solvent, such that one of the sulfur atoms coordinates directly to the zinc ion of the enzyme. The overall zinc coordination (3N from the coordinating histidine residues His94, 96 and 119, and 1S from the inhibitor ligand) can be described as a distorted tetrahedron (Fig. 2). The details of this tetrahedral geometry are provided in Table 3. The zinc bound sulfur also interacts with the O atom of Thr199 in a similar manner to that observed in the more classical clinically used sulfonamides and sulfamates CAIs.34–36 Compound 24a possesses a puckered ring and binds with a slightly higher B-factor of 23.9 Å2, compared to the two other DTCs for which the X-ray structure in complex with hCA II were reported earlier,25 i.e., 23a and 26a. For a structural comparison the three compounds were superposed onto each other (Fig. 3). For unbound hCA II the side chain conformation of His64 has been shown to be dependent upon the buffer pH, which effects the protonation state of the imidazole ring. It is widely believed that this side chain flips from an "in" to "out" conformation as part of the proton transfer mechanism in hCA II, hence two conformations of the residue are often observed in crystal structures.34–36 His64 in the hCA II structure in complex with compound 26a (PDB ID 3P5L) and 24a (PDB ID 3P5A) has a dual conformtion. In the case of 26a, the terminal six-membered hydrophobic ring sites close to F131, V135 and P202 at the rim of the active site in a hydrophobic pocket of hCA II. Whereas, for 24a, the six-membered ring does not extend far enough out of the active site to either reach this hydrophobic pocket or close enough to the in-conformation of His64. Hence, compounds 24 and 26a are 5.6 and 5.0 Å respectively from His64 and therefore do not affect its conformation. Whereas in the hCA II – 23a structure (PDB ID 3P58) the six-membered planar ring forms a T-shaped π-stacking with the imidzole ring of His64 and stabilizes this amino acid in the "in" conformation.25 In addition 24a is positioned 3.2 Å from Thr200 but does not form hydrogen bonds with either the protein main- or side-chain. However, the endocyclic oxygen atom in the tail ring of 24a was within hydrogen bond distance of 3.4 and 3.2 Å from water393 and water540, respectively (Fig. 2) which probably also contribute to its high affinity to hCA II.</p><!><p>Two of the new CAIs investigated here, compounds 24a and 25a, which show excellent hCA II and XII inhibitory properties (Table 1) were investigated in vivo, for their ability to lower intraocular pressure (IOP) in carbomer-induced glaucoma in rabbits.10c Normal IOP in rabbits, like in humans is of around 15–20 mm Hg. In this model, the IOP is quite elevated, mimicking thus the pathologic situation observed in the human disease.10 Clinically used drugs such as dorzolamide, induce a maximal IOP lowering of 4–5 mm Hg, as reported recently by us.10 The two dithiocarbamates investigated here in detail were chosen both due to their excellent enzyme inhibitory activity in vitro and also because they show very good water solubility, being formulated at 2 % eye drop solution at neutral pH (due to their salt character, whereas DZA is formulated at pH 5.5, as hydrochloride salt, and induces eye irritation).10 In fact, water solubility of eye drugs is a significant problem,8–10 with many classes of drugs achieving an acceptable solubility only as salts with strong acids, such as HCl, which leads to acidic pH values causing eye irritation (dorzolamide DZA is a well known such case).10</p><p>Rabbits were treated with 2 % solutions of DTCs 24a and 25a and their IOP was monitored for 48 h (Fig. 4a). The contralateral eye was treated with vehicle and was used as control (Fig. 4B). As observed from Figure 4, both compounds were effective in reducing elevated IOP time dependently, for a rather long period. The maximal effect (of –6–10 mm Hg) has been observed after 2 hours post administration, and it lasted for up to 4–8 hours, being almost double that reported for DZA (of 4–5 mm Hg, and lasting only for about 3–4 hours).10 DTC 24a was slightly more effective than 25a as an IOP lowering agent. The IOP in vehicle-treated eyes was rather constant during the entire duration of the experiments, varying between 37–39 mm Hg for animals treated with 25a and between 37–40 mm Hg for animals treated with 24a (the data of the Figures are the mean for 3 different animals and the error range is shown in Fig. 4).</p><!><p>We report here that DTCs represent a novel class of highly effective CAIs. DTCs are easy to prepare from simple starting materials, they can incorporate a very high chemical diversity, and act as inhibitors of several physiologically relevant CA isoforms, with potencies from the subnanomolar to the micromolar. SAR for the inhibition of isoforms hCA I, II, IX and XII were straightforward and slightly different, with small modifications in the backbone of the compound leading to dramatic changes of biological activity. The inhibition mechanism of the DTCs was also explained, by resolving the X-ray crystal structure for hCA II complexed with a heterocyclic DTC. The CS2− moiety present in DTCs represents a new zinc-binding function. It is directly coordinated to the Zn(II) ion from the enzyme active site and also participates in an interaction (hydrogen bond) with the OH moiety of Thr199, an amino acid essential for the binding of many classes of CAIs (and of the substrates). The organic scaffold of the DTC is deeply buried within the enzyme active site and also participates in favorable interactions with it, which leads to a high stabilization of the enzyme-inhibitor adduct. Some of the most potent CAIs detected here showed favorable IOP lowering effects in an animal model of glaucoma. Being water soluble, with pH of the solution in the neutral range, and with duration of action lasting up to 4–8 h, this new class of CAIs may constitute interesting candidates for developing novel antiglaucoma therapies, a field in which no new drug emerged in the last 15 years.</p><!><p>1H, 13C, DEPT, COSY, HMQC and HMBC spectra were recorded using a Bruker Advance III 400 MHz spectrometer. The chemical shifts are reported in parts per million (ppm) and the coupling constants (J) are expressed in Hertz (Hz). For all new compounds DEPT, COSY, HMQC and HMBC were routinely used to definitely assign the signals of 1H and 13C. Infrared spectra were recorded on a Perkin Elmer Spectrum R XI spectrometer as solids on KBr plates. Melting points (m.p.) were measured in open capillary tubes, unless otherwise stated, using a Büchi Melting Point B-540 melting point apparatus and are uncorrected. Thin layer chromatography (TLC) was carried out on Merck silica gel 60 F254 aluminium backed plates. Elution of the plates was carried out using ethyl acetate/n-hexane or MeOH/DCM systems. Visualization was achieved with UV light at 254 nm, by dipping into a 0.5 % aqueous potassium permanganate solution, by Hanessian's Stain solution and heating with a hot air gun or by exposure to iodine.</p><p>All other solvents and chemicals were used as supplied from Aldrich Chemical Co., Acros, Fisher, Alfa Aesar or Lancaster Synthesis. Aniline 1, morpholin-4-amine 2, 4-methylpiperazin-1-amine 3, (±) sec-Butylamine 4, 2-Morpholinoethanamine 5, N1,N1-bis(2-Aminoethyl)ethane-1,2-diamine 6 (Tris), Benzylamine 7 (CAS 100-46-9), Pyridin-4-ylmethanamine 8 (CAS 3731-53-1), 2′-(Piperidin-1-yl)ethanamine 9, 2-Aminothiazole 10, Glycine 11, 3-(1H-Imidazol-1-yl)propan-1-amine 12, sodium dimethyldithiocarbamate 13a, sodium diethyldithiocarbamate 14a, Pyrrolidine 15, Diisobutylamine 16, Dipropylamine 17, Dibutylamine 18, Dihexylamine 19, Ethylbutyamine 20, Diethanolamine 21, N-methylbenzenamine 22, N,N-Benzylmethylamine 23, Morpholine 24, Piperazine 25, 4-Cyano-4-phenylpiperidine hydrochloride 26, L-Proline 27 were purchased from Sigma-Aldrich (Milan, Italy) and were of the highest available purity. Purity of the prepared DTCS has been determined by HPLC and was > 99 %.</p><!><p>Secondary/primary amines 1-27 (1.0 g, 1.0 eq) were treated with a NaOH, KOH or Et3N (1.0 - 2.2 eq), 4.0 ml of MeOH as co-solvent was used, and the solutions were stirred at 0°C for 20 min (Scheme 1). Then carbon disulfide (1.2 – 2.4 eq) was added dropwise and the mixture was stirred at r.t. until starting material was consumed (TLC monitoring). The solvents were removed under vacuo at r.t. and the residues obtained were dissolved in MeOH, filtered off trough Celite and the filtrate was concentrated in vacuo not exceeding 20 °C.</p><p>Synthesis of triethylammonium phenylcarbamodithioate 1a</p><p> </p><p>Aniline 1 (0.5 g, 1.0 eq) was treated with triethylamine (1.0 eq) in benzene (0.5 ml) followed by addition of carbon disulfide (1.0 eq) at 0°C. The mixture was warmed to r.t. and stirred O.N. at r.t.. The solid formed was washed with diethyl ether and dried under vacuo to afford the titled compound as a light yellow solid in 51% yield.</p><p>Triethylammonium phenylcarbamodithioate 1a: νmax (KBr) cm−1, 2960, 2886, 1648, 1599, 1520, 1451; δH (400 MHz, DMSO-d6) 1.13 (9H, t, J 6.8, 3x CH2CH3), 2.98 (6H, brs, 3x CH2CH3), 6.97 (1H, t, J 8.0, 4-H), 7.22 (2H, dd, J 8.3, 8.0, 2x 3-H), 7.93 (2H, d, J 8.3, 2x 2-H), 9.00 (1H, brs, exchange with D2O, (CH3CH2)3N+-H), 10.10 (1H, brs, exchange with D2O, N-H); δC (100 MHz, DMSO-d6) 10.0, 46.6, 114.8, 122.9, 128.4, 143.2, 215.5 (C=S); m/z (ESI), 168 [M-Na]−.</p><p>Diisobutylcarbodithioic acid sodium salt 16a: m.p. 220 °C with dec; νmax (KBr) cm−1, 2961, 2933, 2867, 1640, 1601, 1520, 1480, 1090; δH (400 MHz, DMSO-d6) 0.84 (12H, d, J 6.8, 4 x CH3), 2.43 (4H, m, 2 x CH), 3.86 (4H, d, J 7.2, 2 x CH2); δC (100 MHz, DMSO-d6) 21.2, 27.4, 61.5, 215.4 (C=S); m/z (ESI), 204 [M-Na]−. Data are in agreement with reported data.30</p><p>Synthesis of morpholinecarbamodithioate sodium salt 24a.</p><p> </p><p>Morpholine 24 (1.0 g, 1.0 eq) was treated according to the general procedure with 1.0 M aqueous solution of NaOH (1.0 eq) followed by addition of carbon disulfide (1.2 eq). The title compound was obtained as a white solid in quantitative yield.</p><p>Morpholinecarbamodithioate sodium salt 24a: m.p. 320 °C with dec; νmax (KBr) cm−1, 2966, 2901, 2854, 1625, 1520, 1416, 1215; δH (400 MHz, DMSO-d6) 3.52 (4H, t, J 8.0 CH2), 4.33 (2H, t, J 8.0, CH2); δC (100 MHz, DMSO-d6) 50.6, 67.1, 215.4 (C=S); m/z (ESI), 162 [M-Na]−.</p><!><p>An Applied Photophysics stopped-flow instrument has been used for assaying the CA catalysed CO2 hydration activity. Phenol red (at a concentration of 0.2 mM) has been used as indicator, working at the absorbance maximum of 557 nm, with 20 mM Hepes (pH 7.5) as buffer, and 20 mM Na2SO4 (for maintaining constant the ionic strength), following the initial rates of the CA-catalyzed CO2 hydration reaction for a period of 10–100 s.32 The CO2 concentrations ranged from 1.7 to 17 mM for the determination of the kinetic parameters and inhibition constants. For each inhibitor at least six traces of the initial 5–10% of the reaction have been used for determining the initial velocity. The uncatalyzed rates were determined in the same manner and subtracted from the total observed rates. Stock solutions of inhibitor (0.1 mM) were prepared in distilled-deionized water and dilutions up to 0.01 nM were done thereafter with distilled-deionized water. Inhibitor and enzyme solutions were preincubated together for 15 min at room temperature prior to assay, in order to allow for the formation of the E-I complex. The inhibition constants were obtained by non-linear least-squares methods using PRISM 3, as reported earlier,33 and represent the mean from at least three different determinations. All CA isofoms were recombinant ones obtained in house as reported earlier.9,10</p><!><p>Co-crystals for the hCA II – 24a complex were obtained using the hanging drop vapor diffusion method.37 Drops of 10 μL (0.3 mM hCA II; 0.7 mM DTC 24a; 0.1 % dimethyl sulfoxide; 0.8 M sodium citrate; 50 mM Tris-HCl; pH 8.0) were equilibrated against the precipitant solution (1.6 M sodium citrate; 50 mM Tris-HCl; pH 8.0) at room temperature (~20 °C). Crystals were observed after 5 days. Based on visual selection a crystal of the complex was cryoprotected by quick immersion into 20% glycerol precipitant solution and flash-cooled by exposing to a gaseous stream of nitrogen at 100 K. The X-ray diffraction data was collected using an R-AXIS IV++ image plate system on a Rigaku RU-H3R Cu rotating anode operating at 50 kV and 22 mA, using Osmic Varimax HR optics. The detector-crystal distance was set to 80 mm. The oscillation steps were 1° with a 5 min exposure per image. Indexing, integration, and scaling were performed using HKL2000.38</p><!><p>Starting phases were calculated from Protein Data Bank (PDB) entry 3KS339 with waters removed. Refinement using Phenix package,40 with 5% of the unique reflections selected randomly and excluded from the refinement data set for the purpose of Rfree calculations,41 was alternated with manual refitting of the model in Coot.42 The validity of the final model was assessed by PROCHECK.43 Complete refinement statistics and model quality are included in Table 2.</p><!><p>Adult male New Zealand Albino rabbits weighing 2–2.5 Kg were employed in this study. The animals were utilized in groups of eight for each of the chosen specific treatments. The experimental procedures were conform to those of the Declaration of Helsinki and with the Guideline for the Care and Use of Laboratory Animals as adopted and promulgated by the U.S. National Institute of Health, and were conducted upon authorization of the Italian Regulations on Protection of Animals used for experimental and other scientific purpose (DM 116/1992) as well as with the European Union Regulations (OJ of ECL 358/1, 12/12/1986), and the experimental protocol was approved by the local animal care committee of the University of Florence (Florence, Italy). The rabbits were kept in individual cages; food and water were provided ad libitum. The animals were identified with a tattoo on the ear, numbered consecutively and maintained on a 12–12h light/dark cycle in a temperature controlled room (22°–23°C). All selected animals were examined before the beginning of the study and were determined to be normal on ophthalmic and general examinations. Glaucoma was induced by injection of 0.1 ml 0.25% carbomer (Siccafluid, FarMila - THEA Pharmaceuticals) into anterior eye-chamber bilaterally in New Zeeland albino rabbits] anesthetized with tiletamine and zolazepam (Zoletil 100, 0.05 mg/Kg b.w.) plus xilazine (Xilor 2%, 0.05 ml/Kg b.w.) i.m., by the procedure previously reported.10c IOP was measured before carbomer injection and after 1, 2 and 4 hours the first day and three times a day until stabilization and then every 24 hours. All rabbits treated with carbomer presented a net increase in IOP. One drop of 0.2% oxybuprocaine hydrochloride (Novesine, Sandoz) diluted 1:1 with sterile saline was instilled in each eye immediately before each set of pressure measurements. IOP was measured using a Tono-Pen XL tonometer (Medtronic Solan, USA) as reported by earlier.10 The pressure readings were matched with two-point standard pressure measurements at 1, 2, 4 and 8 hours after the instillation of the drug and once a day for the following days using a Digilab calibration verifier. All IOP measurements were done by the same investigators using the same tonometer. As soon as a stable IOP increase was obtained, the animals were treated with the drugs in study. The efficacy of the different drugs in lowering IOP was evaluated after drug administration over four hours, with the following schedule: before and after 30, 60, 90, 120 and 240 minutes after drug administration. The treatment was performed in three animals per drug in one eye and compared to the contralateral eye treated with vehicle. A group of 4 non glaucomatous albino rabbits was treated with the drugs of this study and used as control. At the end of the experiments the animals were killed with a lethal dose of Pentothal (Abbott S.p.A., Campoverde di Aprilia, LT).</p>

PubMed Author Manuscript

Integration of exonuclease III-powered three-dimensional DNA walker with single-molecule detection for multiple initiator caspases assay

Initiator caspases are important components of cellular apoptotic signaling and they can activate effector caspases in extrinsic and intrinsic apoptotic pathways. The simultaneous detection of multiple initiator caspases is essential for apoptosis mechanism studies and disease therapy. Herein, we develop a sensitive nanosensor based on the integration of exonuclease III (Exo III)-powered three-dimensional (3D) DNA walker with single-molecule detection for the simultaneous measurement of initiator caspase-8 and caspase-9. This assay involves two peptide-DNA detection probe-conjugated magnetic beads and two signal probe-conjugated gold nanoparticles (signal probes@AuNPs). The presence of caspase-8 and caspase-9 can induce the cleavage of peptides in two peptide-DNA detection probes, releasing two trigger DNAs from the magnetic beads, respectively. The two trigger DNAs can serve as the walker DNA to walk on the surface of the signal probes@AuNPs powered by Exo III digestion, liberating numerous Cy5 and Texas Red fluorophores which can be quantified by single-molecule detection, with Cy5 indicating caspase-8 and Texas Red indicating caspase-9. Notably, the introduction of the AuNP-based 3D DNA walker greatly reduces the background signal and amplifies the output signals, and the introduction of single-molecule detection further improves the detection sensitivity. This nanosensor is very sensitive with a detection limit of 2.08 Â 10 À6 U mL À1 for caspase-8 and 1.71 Â 10 À6 U mL À1 for caspase-9, and it can be used for the simultaneous screening of caspase inhibitors and the measurement of endogenous caspase activity in various cell lines at the single-cell level. Moreover, this nanosensor can be extended to detect various proteases by simply changing the peptide sequences of the detection probes.

integration_of_exonuclease_iii-powered_three-dimensional_dna_walker_with_single-molecule_detection_f

Introduction<!>Results and discussion<!>Conclusions<!>Chemicals and materials<!>Preparation of the detection probe-conjugated MBs<!>Detection of caspase activities<!>Ensemble uorescence measurement<!>Single-molecule detection and data analysis<!>Gel electrophoresis<!>Kinetic analysis<!>Inhibition assay<!>Cell culture and preparation of cell extracts

<p>Apoptosis is programmed cell death which can result in the disappearance of cells without any inammatory phenomena. 1 Caspases are a family of cysteinyl aspartate-directed proteases, and they are the central executioners of apoptosis. 2 Once activated by a specic stimulus (e.g., ultraviolet (UV) radiation, girradiation, heat, DNA damage, viral virulence factors), 3 caspases execute the part-proteolysis of downstream substrates to trigger a cascade of events that culminates in the desired biological response: disruption of cellular membranes, breaking down of cytoplasmic and nuclear skeletons, extruding of cytosol, degradation of chromosomes, and fragmentation of the nucleus. 4 Inappropriate apoptosis and deregulation of caspase activity are implicated in various human diseases, including Alzheimer's disease, 5 ischemic damage, 6 autoimmune diseases, 7 and cancers. 8 The apoptotic-associated caspases are classied into initiators and effectors by their location in the apoptosis cascade signalling pathways. The initiator caspases (e.g., caspase-8, caspase-9, and caspase-10) are responsible for initiating the apoptosis cascade, and the effector caspases (e.g., caspase-3, caspase-6, and caspase-7) can be activated by initiator caspases to destroy cellular proteins and execute apoptosis. 9 Extrinsic and intrinsic apoptotic pathways are the only two pathways of apoptosis, and caspase-8 and caspase-9 are critical participants in extrinsic and intrinsic apoptosis, respectively. In the extrinsic apoptotic pathway, caspase-8 is specically activated by death receptors such as Fas/CD95. 10 In the intrinsic apoptotic pathway, caspase-9 is activated by the accumulation of cytochrome C in the cytosol due to dissipation of the mitochondrial membrane potential. 11 Therefore, the simultaneous measurement of caspase-8 and caspase-9 are of great importance for the study of the apoptosis mechanism and disease therapy.</p><p>Conventional methods for caspase assays include enzymelinked immune sorbent assay (ELISA), 12 western blot, 13 ow cytometry, 14 and mass spectrometry. 15 They usually involve multi-step and time-consuming processes, 16 and extensive pretreatment of samples. 17 Recently, a variety of new methods have been developed for in vitro and in vivo detection of caspases, including electrochemical, 18 colorimetric 19 and uorescence measurements. 20,21 Most of these methods rely on either antibodies 18 or peptide substrates. [19][20][21] The main limitation of antibody-based assays is the requirement for high quality antibodies. 22 Alternatively, peptide substrates are a more attractive choice for caspase assay with the distinct advantages of accessibility, simplicity, cost-effectiveness, and chemical denition. 23 In the peptide substrate-based assays, peptide substrates are oen labelled with uorophores and the uorescence signals are directly measured aer caspase cleavage, resulting in limited sensitivity. 24 The development of new methods for sensitive caspase assay is highly desirable.</p><p>Due to the highly selective and programmable Watson-Crick complementarity, DNA is becoming an increasingly attractive building material for the construction of a series of functional nanodevices (e.g., walker, [25][26][27] gear, 28 and tweezer 29 ). In particular, a DNA walker that is able to convert chemical energy to mechanical motion is the most widely studied DNA nanodevice and has been extensively applied for molecular transport, 30 DNA computing, 31 chemical synthesis, 32 and biosensing. 33,34 DNA walkers can be classied into one-dimensional (1D), twodimensional (2D), and three-dimensional (3D) DNA walkers in terms of dimensionality. The 1D DNA walkers move along 1D tracks such as double-stranded DNA (dsDNA) and single walled carbon nanotubes (SWCNTs). 35,36 2D DNA walkers can move along the 2D surfaces of gold electrodes and DNA tile. 37,38 3D DNA walkers can move along 3D scaffolds on nanoparticles such as gold nanoparticles (AuNPs), 39 magnetic beads (MBs), 40 SiO 2 @CdTe, 41 Au@Fe 3 O 4 , 42 hollow carbon nanospheres (HCS), 43 and DNA origami. 44 Such 3D DNA walkers are usually powered by nicking endonucleases, 45,46 exonuclease III (Exo III), 47 or the DNA catalytic hairpin assembly reaction, 48 and possess large specic surface area and high DNA loading capability, 49 and they can greatly enhance the local concentration of track DNA for improved walking efficiency and higher processivity, 50 facilitating high signal gain in a given period of time. 51 A rolling circle amplication (RCA)-based method has been developed for caspase assay, 52 but it is only suitable for single target detection and is prone to a high background due to nonspecic amplication. In addition, it requires laborious procedures for the preparation of the circular template, and its accuracy is challenged by the inner lter effect and the collisional quenching of ensemble uorescence measurements. Herein, we develop a simple and sensitive caspase nanosensor with the capability of multiplexed assay based on the integration of an Exo III-powered 3D DNA walker with singlemolecule detection. The Exo III mediates highly efficient signal amplication through catalyzing the cyclic digestion of one strand of dsDNA from the 3 0 end, 53 which can be performed under isothermal conditions with the elimination of complex thermal cycling and nonspecic articial amplication involving in polymerase chain reaction (PCR). 54 In comparison with other isothermal amplication methods such as strand displacement amplication (SDA), 55 exponential amplication reaction (EXPAR), 56 rolling circle amplication, 57 and loop mediated isothermal amplication (LAMP), 58 the Exo IIImediated amplication takes advantage of a much simpler reaction scheme without the requirement of a specic recognition sequence for nicking endonuclease, complicated procedures for preparation of the circular template, and multiple primers for completing the amplication reaction. Moreover, in comparison with the ensemble measurement, single-molecule detection has the distinct advantages of high sensitivity, simplicity, and low sample consumption. [59][60][61][62][63][64][65] This assay involves two peptide-DNA detection probe-conjugated magnetic beads and two signal probe-conjugated gold nanoparticles (signal probes@AuNPs). The integration of Exo IIImediated amplication with the AuNP-based DNA walker facilitates facile and efficient amplication of the caspase signal to acquire high sensitivity with reduced background, and the introduction of single-molecule detection further improves the detection sensitivity. The presence of caspase-8 and caspase-9 can induce the cleavage of peptides in two peptide-DNA detection probes, releasing two trigger DNAs from the magnetic beads. The two trigger DNAs can serve as the walker DNA to walk on the surface of signal probes@AuNPs powered by Exo III digestion, liberating numerous Cy5 and Texas Red uorophores which can be quantied by single-molecule detection, with Cy5 indicating caspase-8 and Texas Red indicating caspase-9. This nanosensor is very sensitive with a detection limit of 2.08 Â 10 À6 U mL À1 for caspase-8 and 1.71 Â 10 À6 U mL À1 for caspase-9, and it can be used for simultaneous screening of caspase inhibitors and measurement of endogenous caspase activity in various cell lines at the single-cell level. To the best of our knowledge, the integration of Exo III-powered 3D DNA walker with single-molecule detection for the simultaneous measurement of multiple caspases has not been explored so far.</p><!><p>The schematic illustration of the nanosensor based on the integration of the Exo III-powered 3D DNA walker with singlemolecule detection for the simultaneous measurement of initiator caspase-8 and caspase-9 is shown in Scheme 1. This assay involves two peptide-DNA detection probe-conjugated magnetic beads and two signal probe-conjugated gold nanoparticles (signal probes@AuNPs). The biotinylated peptide-DNA detection probes (Fig. S1, ESI †) can be self-assembled on the streptavidin-coated MBs through the specic streptavidinbiotin interaction. The peptide domain of detection probe 1 (Scheme 1, green color) contains a tetrapeptide sequence Ile-Glu-Thr-Asp (IETD) for caspase-8 recognition, and the peptide domain of detection probe 2 (Scheme 1, magenta color) contains a tetrapeptide sequence Leu-Glu-His-Asp (LEHD) for caspase-9 recognition. The DNA domain of detection probe 1 (Scheme 1, yellow color) is complementary to signal probe 1 (Scheme 1, blue color), and the DNA domain of detection probe 2 (Scheme 1, cyan color) is complementary to signal probe 2 (Scheme 1, red color). Signal probes are attached to the surface of AuNPs via a Au-S bond to form the 3D scaffold of the DNA walker which is powered by Exo III. In the presence of caspase-8, it specically recognizes the sequence IETD and hydrolyzes the peptide bond adjacent to the carboxylic group of the aspartic acid residue (i.e., the peptide bond between aspartic acid and glycine), releasing the cleaved detection probe 1 from the magnetic beads (Fig. S1A, ESI †). In the presence of caspase-9, it specically recognizes the sequence LEHD and hydrolyzes the peptide bond between aspartic acid and glycine, releasing the cleaved detection probe 2 from the magnetic beads (Fig. S1B, ESI †). The intact detection probes are separated by magnetic separation. The trigger DNA in the cleaved detection probes (Scheme 1, yellow and cyan color) can serve as the walker DNA to initiate the recyclable cleavage of signal probes on the surface of the AuNPs, releasing numerous Cy5 and Texas Red uorophores from the signal probes@AuNPs. Specically, the hybridization of trigger DNA 1 of the cleaved detection probe 1 (Scheme 1, yellow color) and signal probe 1 (Scheme 1, blue color) form a dsDNA with 5 0 overhangs, which can be recognized by Exo III. The stepwise catalytic removal of mononucleotides from the 3 0 -terminus of signal probe 1 (Scheme 1, blue color) leads to the liberation of the Cy5 uorophore from the signal probe 1@AuNPs, while the digestion of trigger DNA 1 (Scheme 1, yellow color) is blocked by the oligo-T tail. As the digestion of signal probe 1 (Scheme 1, blue color) proceeds, the trigger DNA 1 (Scheme 1, yellow color) is partially released and hybridizes with another signal probe 1 (Scheme 1, blue color) nearby. In this way, the trigger DNA 1 (Scheme 1, yellow color) walks along the AuNP surface, and abundant Cy5 uorophores are liberated from the signal probe 1@AuNPs, resulting in the restoration of Cy5 uorescence which can be quantied by single-molecule detection, with Cy5 indicating the presence of caspase-8. Similarly, the hybridization of trigger DNA 2 of the cleaved detection probe 2 (Scheme 1, cyan color) with signal probe 2 (Scheme 1, red color) forms a dsDNA with 5 0 overhangs, which can be recognized by Exo III. The Exo III-mediated stepwise catalytic removal of mononucleotides from the 3 0 -terminus of signal probe 2 (Scheme 1, red color) leads to the liberation of abundant Texas Red uorophores from the signal probe 2@AuNPs, resulting in the restoration of Texas Red uorescence which can be used for the quantication of caspase-9. Because there are 120 signal probes per AuNP (ESI †), the high density of the scaffold DNA enhances the movement of trigger DNA along the 3D tracks, greatly amplifying the uorescence signals. In contrast, in the absence of caspase-8 and caspase-9, the detection probes remain intact, and no trigger DNA is released. Consequently, no Exo III-mediated digestion of signal probes occurs, and neither the Cy5 nor Texas Red signal can be detected.</p><p>This assay mainly relies on the caspase-mediated cleavage of peptide to initiate the recycling liberation of Cy5/Texas Red uorescent molecules from the signal probes@AuNPs. We used 12% nondenaturing polyacrylamide gel electrophoresis (PAGE) to verify the cleavage of the detection probes by caspase-8 and caspase-9. For the caspase-8 assay, when caspase-8 is absent, only one distinct band is observed (Fig. 1A, lane 2), which is identical to that of the intact detection probe 1, indicating that the cleavage of detection probe 1 cannot occur in the absence of caspase-8. When caspase-8 is present, a new smaller-sized band is observed (Fig. 1A, lane 3), which corresponds to the cleaved peptide-DNA 1 (Fig. 1A, lane 1), indicating the cleavage of detection probe 1 induced by caspase-8. When only caspase-9 is present, only one identical band to that of the intact detection probe 1 (Fig. 1A, lane 2) is observed (Fig. 1A, lane 4), indicating that the cleavage of detection probe 1 cannot occur in the presence of caspase-9. For the caspase-9 assay, when caspase-9 is absent, only one distinct band is observed (Fig. 1B, lane 2), which is identical to that of intact detection probe 2, indicating that the cleavage of detection probe 2 cannot occur in the absence of caspase-9. When only caspase-8 is present, only one distinct band identical to lane 2 (Fig. 1B) is observed (Fig. 1B, lane 3), indicating that the cleavage of detection probe 2 cannot occur in the presence of caspase-8. When caspase-9 is present, a new smaller-sized band is observed (Fig. 1B, lane 4), which corresponds to the cleaved peptide-DNA 2 (Fig. 1B, lane 1), indicating the cleavage of detection probe 2 induced by caspase-9. To verify the feasibility of the proposed nanosensor, we performed uorescence measurements (Fig. 1C and D). As shown in Fig. 1C, a distinct Cy5 uorescence signal with a characteristic emission peak of 670 nm is observed in the presence of 0.025 U mL À1 caspase-8 (Fig. 1C, red line). In contrast, in the control group without caspase-8, a low Cy5 signal (Fig. 1C, black line) is detected. As shown in Fig. 1D, a distinct Texas Red uorescence signal with a characteristic emission peak of 620 nm is observed in the presence of 0.025 U mL À1 caspase-9 (Fig. 1D, green line). In contrast, in the control group without caspase-9, a low Texas Red signal (Fig. 1D, black line) is detected.</p><p>We further employed single-molecule detection technology to detect the uorescence signals (Fig. 2). In the absence of caspase-8 and caspase-9, neither Cy5 (Fig. 2A) nor the Texas Red uorescence signal (Fig. 2E) is observed, indicating that the Exo III-mediated recycling liberation of uorophores cannot occur in the absence of caspase-8 and caspase-9. In contrast, in the presence of 0.025 U mL À1 caspase-8, distinct Cy5 uorescence signals are observed (Fig. 2B, red color), but no Texas Red uorescence signal is detected (Fig. 2F), indicating that caspase-8 can induce the recycling liberation of Cy5 uorophores from the signal probe 1@AuNPs. In the presence of 0.025 U mL À1 caspase-9, distinct Texas Red uorescence signals are observed (Fig. 2G, green color), but no Cy5 uorescence signal is detected (Fig. 2C), indicating that caspase-9 can induce the recycling liberation of Texas Red uorophores from the signal probe 2@AuNPs. In the presence of both caspase-8 and caspase-9, distinct Cy5 (Fig. 2D, red color) and Texas Red uorescence signals (Fig. 2H, green color) are simultaneously observed. Notably, both the Cy5 and Texas Red uorescence spots exhibit single-step photobleaching (Fig. S3, ESI †), suggesting that the observed individual uorescent spot originates from a single dye molecule. These results clearly demonstrate that the proposed nanosensor can be applied for the simultaneous detection of caspase-8 and caspase-9 at the single-molecule level.</p><p>Under the optimized experimental conditions (Fig. S4 and S5, ESI †), we evaluated the sensitivity of the proposed nanosensor by measuring the variance of uorescent counts with the caspase concentration. As shown in Fig. 3A, the Cy5 counts improve with increasing concentration of caspase-8 from 2.50 Â 10 À6 to 0.05 U mL À1 . In the logarithmic scale, the Cy5 counts show a linear correlation with the concentration of caspase-8 over the range from 2.50 Â 10 À6 to 2.50 Â 10 À3 U mL À1 . The regression equation is N ¼ 510.06 + 89.12 log 10 C with a correlation coefficient (R 2 ) of 0.9990 (inset of Fig. 3A), where N is the measured Cy5 count and C is the concentration of caspase-8 (U mL À1 ). By evaluating the average response of the control group plus three times the standard deviation, the limit of detection (LOD) was calculated to be 2.08 Â 10 À6 U mL À1 (7.51 pM). The sensitivity of the proposed nanosensor is 96-fold higher than the conjugated polymer-based uorescence assay (0.2 U mL À1 ), 66 and 439-fold higher than the reported uorescence assay (3.3 nM). 67 As shown in Fig. 3B, the Texas Red counts improve with increasing concentration of caspase-9 from 2.50 Â 10 À6 to 0.05 U mL À1 . In the logarithmic scale, the Texas Red counts show a linear correlation with the concentration of caspase-9 over a range from 2.50 Â 10 À6 to 2.50 Â 10 À3 U mL À1 . The regression equation is N ¼ 451.37 + 71.46 log 10 C with a correlation coef-cient (R 2 ) of 0.9997 (inset of Fig. 3B), where N is the measured Texas Red count and C is the concentration of caspase-9 (U mL À1 ). The LOD was calculated to be 1.71 Â 10 À6 U mL À1 (92.37 pM). The sensitivity of the proposed nanosensor is 40-fold higher than the up-conversion nanoparticles-based FRET assay (0.068 U mL À1 ), 68 and 650-fold higher than the gold nanoparticle-polydopamine-based electrochemical immunosensor (0.06 mM). 18 The improved sensitivity of the proposed nanosensor can be ascribed to (1) the efficient conversion of the caspase activity signal into the DNA signal through the cleavage of the peptide-DNA detection probe, (2) recyclable liberation of uorescent molecules from the signal probes@AuNPs induced by the Exo III-driven 3D DNA walker, and (3) the high signal-tonoise ratio of single-molecule detection.</p><p>To investigate the selectivity of the proposed nanosensor for caspase assay, we used caspase-3, DNA (cytosine-5)methyltransferase 1 (Dnmt1) and human alkyladenine DNA glycosylase (hAAG) as the negative controls. Caspase-3 is a core effector caspase that can be activated by caspase-8 in the extrinsic apoptotic pathway and can be activated by caspase-9 in the intrinsic pathway, but it cannot cleave the peptide substrates of caspase-8 and caspase-9. 16 Dnmt1 can specically recognize the hemimethylated sequence 5 0 -CG-3 0 , and catalyze the transfer of a methyl group from S-adenosyl-L-methionine (SAM) to the cytosine in genomic DNA. 69 hAAG can recognize and excise a diverse group of alkylated purine bases and cleave the N-glycosidic bond between the sugar and the damaged base. 70 As shown in Fig. 4, in the presence of caspase-3, Dnmt1 and hAAG, neither Cy5 nor Texas Red uorescence signal is observed, consistent with the control with only the reaction buffer. In contrast, in the presence of caspase-8, an enhanced Cy5 uorescence signal is observed, but no Texas Red uorescence signal is detected. In the presence of caspase-9, an enhanced Texas Red uorescence signal is detected, but no Cy5 uorescence signal is detected. Moreover, in the presence of both caspase-8 and caspase-9, both Cy5 and Texas Red uorescence signals can be simultaneously detected. These results demonstrate that the proposed nanosensor exhibits excellent selectivity toward caspase-8 and caspase-9.</p><p>To investigate the feasibility of the proposed nanosensor for kinetic analysis, we measured the initial velocity (V) in response to various concentrations of the detection probes. To evaluate the enzyme kinetic parameters of caspase-8, we measured the V in the presence of 0.025 U mL À1 caspase-8 and different concentrations of detection probe 1 for 5 min at 37 C. As shown in Fig. 5A, the initial velocity of caspase-8 enhances with the increasing concentration of detection probe 1. V max is calculated to be 146.55 min À1 and K m is calculated to be 1.39 mM. The K m value is consistent with that obtained by the rolling circle amplication (RCA)-based uorescence assay (1 mM). 52 To evaluate the enzyme kinetic parameters of caspase-9, we measured the V in the presence of 0.025 U mL À1 caspase-9 and different concentrations of detection probe 2 for 5 min at 37 C. As shown in Fig. 5B, the initial velocity of caspase-9 enhances with the increasing concentration of detection probe 2. V max is calculated to be 109.12 min À1 and K m is calculated to be 1.21 mM. The K m value is consistent with that obtained by the multicolor gold-selenium bonding nanoprobe-based uorescence assay (8.53 mM). 71 These results suggest that the proposed nanosensor can be used to accurately evaluate the kinetic parameters of caspases. Fluoromethylketone (FMK), 72 chloromethylketone (CMK) 73 and diuorophenoxymethyl (OPh) 74 are competitive caspase inhibitors, and they can irreversibly inactivate caspases by forming a covalent thioether adduct with the cysteine of the active site in caspases. We used the caspase-8 inhibitor Z-IETD-FMK, caspase-9 inhibitor Ac-LEHD-CMK and broad-spectrum caspase inhibitor Q-VD-OPh as the model inhibitors to investigate the feasibility of the proposed nanosensor for the caspase inhibition assay. As shown in Fig. 6A, the relative activity of caspase-8 decreases with the increasing concentration of Z-IETD-FMK from 0 to 20 mM. The concentration of inhibitor required to reduce the activity of caspase-8 by 50% (IC 50 ) is determined to be 1.95 mM, consistent with that obtained by the RCA-based uorescence assay (0.9656 mM). 52 As shown in Fig. 6B, the relative activity of caspase-9 decreases with increasing concentration of Ac-LEHD-CMK from 0 to 300 nM. The IC 50 is determined to be 72.80 nM, consistent with the data from the manufacturer (70 nM). As shown in Fig. 6C and D, the relative activities of caspase-8 and caspase-9 decrease with increasing concentration of Q-VD-OPh, respectively. The IC 50 is determined to be 0.28 mM for caspase-8 and 0.31 mM for caspase-9. These results suggest that the proposed method can be applied for the screening of caspase inhibitors.</p><p>To demonstrate the feasibility of the proposed method for cellular caspase assays, we simultaneously measured endogenous caspase-8 and caspase-9 activity in a human cervical cancer cell line (HeLa cells), a human breast cancer cell line (MCF-7 cells), and a human acute T-lymphocytic leukemia cell line (Jurkat cells). Staurosporine (STS) is a broad-spectrum inhibitor of protein kinases with the capability of inducing in vitro apoptosis. 75 To avoid the interference from nonspecic caspases, we employed anti-caspase-8 and anti-caspase-9 antibodies to completely inhibit target caspase-8/9 activity (Fig. S6, ESI †), and calculated the specic caspase-8/9 signal DC according to eqn (1).</p><p>where C is the uorescent count in the absence of antibody, and C a is the uorescent count in the presence of antibody. We rst veried its capability of inducing apoptosis. In contrast to the low background signal in the control group without any cell extracts (Fig. 7A and B DC Texas Red in HeLa cells (Fig. 7A and B, green columns), MCF-7 cells (Fig. 7A and B, orange columns) and Jurkat cells (Fig. 7A and B, cyan columns) compared to the cells without STS treatment (Student's t-test, P < 0.001), suggesting that the activation of caspases is involved in STS-induced apoptosis. These results demonstrate that the proposed nanosensor can be used for the sensitive detection of endogenous caspases activity.</p><!><p>In summary, we develop a sensitive nanosensor based on the integration of exonuclease III-powered 3D DNA walker with single-molecule detection for simultaneous measurement of initiator caspase-8 and caspase-9. The presence of caspase-8 and caspase-9 can induce the cleavage of peptides in two peptide-DNA detection probes, releasing two trigger DNAs from the magnetic beads. The two trigger DNAs can serve as the walker DNA to walk on the surface of signal probes@AuNPs powered by Exo III digestion, liberating numerous Cy5 and Texas Red uorophores which can be quantied by singlemolecule detection, with Cy5 indicating caspase-8 and Texas Red indicating caspase-9. This nanosensor possesses the following distinct advantages: (1) the introduction of the AuNPbased 3D DNA walker greatly reduces the background signal and amplies the output signals, and the introduction of singlemolecule detection further improves the detection sensitivity, endowing this nanosensor with higher sensitivity and less assay time than the reported caspase assays (Table S1, ESI †) and facilitating the accurate detection of low-abundant caspase with low sample consumption for biomedical research and disease diagnosis;</p><p>(2) the amplication reaction can be conducted at constant temperature without the requirement of complicated thermal cycling, and the signal can be easily quantied by single-molecule counting; (3) the use of peptide-DNA detection probe-conjugated magnetic beads and signal probes@AuNPs greatly simplies the experimental procedures, facilitating the simultaneous detection of multiple caspases. This nanosensor is very sensitive with a detection limit of 2.08 Â 10 À6 U mL À1 for caspase-8 and 1.71 Â 10 À6 U mL À1 for caspase-9, and it can be used for the simultaneous screening of caspase inhibitors and measurement of endogenous caspase activity in various cell lines at the single-cell level. Moreover, this nanosensor can be extended to detect various proteases by simply changing the peptide sequences of the detection probes.</p><!><p>All oligonucleotides (Table 1)</p><!><p>The assembly of detection probes onto the MBs was carried out according to the protocol of the manufacturer. 50 mL of the 10 mg mL À1 streptavidin-coated MBs solution was washed three times using 1Â PBS. Aer resuspending with 50 mL of 1Â PBS, 10 mL of 10 mM detection probes were added to form the detection probes-MB nanostructure through biotin-streptavidin interaction at room temperature for 30 min. The detection probe-MBs were then washed ve times using 1Â PBS to remove the uncoupled probes by magnetic separation, followed by resuspending in 50 mL of 1Â PBS.</p><p>Construction of signal probes@AuNPs AuNPs (10 nm) were functionalized with thiolated signal probes using the freeze-directed methods. 76 The method is reagentless without the involvement of extra salts, acids, and surfactants. Specically, 20 mL of 100 mM signal probes were mixed with 1 mL of AuNPs. The mixture was subsequently placed in a laboratory freezer at À20 C for 2 h, followed by thawing at room temperature. At last, the signal probe-coated AuNPs were centrifuged and washed three times with ultrapure water to remove excess signal probes, and resuspended in 40 mL of sterile water, and stored at 4 C. In the signal probes@AuNPs solution, the concentration of DNA was measured using a NanoDrop 2000c Spectrophotometer (Thermo Scientic, Wilmington, Delaware, USA) and the number of signal probes per AuNP is estimated to be 120 AE 1 for both Cy5-labeled signal probe 1 and Texas Red-labeled signal probe 2 (Fig. S2, ESI †).</p><!><p>The caspase activity assay includes two steps: (1) caspasemediated cleavage of detection probes, and (2) recycling liberation of uorophores induced by the Exo III-powered 3D DNA walker. The caspase-mediated cleavage of the detection probe was performed in 20 mL of solution containing 3.5 mL of 1Â PBS, 1 mM DTT, 8 mL of detection probe 1-conjugated MBs or/and detection probe 2-conjugated MBs, and 0.5 mL of caspase at different concentrations at 37 C for 1 h. Aer magnetic separation, the supernatant with cleavage products was collected for the next-step use. The Exo III-powered 3D DNA walker-based recycling liberation of uorophores was performed in 20 mL of reaction solution containing 1Â CutSmart buffer (50 mM potassium acetate, 20 mM tris-acetate, 10 mM magnesium acetate, 100 mg mL À1 BSA, pH 7.9), 0.71 mL of signal probe 1@AuNPs or/and 0.36 mL of signal probe 2@AuNPs, and 0.5 U of Exo III at 37 C for 1 h.</p><!><p>The 20 mL of reaction products was diluted to a nal volume of 50 mL with ultrapure water for the measurement of uorescence emission spectra using a 1 cm path length quartz cuvette on a Hitachi F-7000 uorescence spectrophotometer (Tokyo, Japan). The excitation wavelength was 645 nm for Cy5 and 560 nm for Texas Red, and the emission spectra were recorded in the wavelength range of 660-750 nm for Cy5 and 600-680 nm for Texas Red with a slit width of 5 nm for both excitation and emission.</p><!><p>The reaction products of caspase-8 were diluted 200-fold in the imaging buffer (1 mg mL À1 glucose oxidase, 0.4% (w/v) Dglucose, 0.04% mg mL À1 catalase, 50 mg mL À1 BSA, 67 mM glycine-potassium hydroxide, 1 mg mL À1 Trolox, 2.5 mM magnesium chloride, pH 9.4), and the reaction products of caspase-9 were diluted 500-fold in the imaging buffer. For total internal reection uorescence (TIRF) imaging, 10 mL of the sample was directly pipetted onto the coverslips. The Cy5 and Texas Red uorescent molecules were excited by the sapphire 640 and 561 nm lasers (Coherent, USA), respectively. The resulting photons were collected by an oil immersion objective (CFI Apochromat TIRF 100Â). The Cy5 and Texas Red uorescence signals were imaged on an Andor ixon Ultra 897 EMCCD camera (Andor, Belfast, UK) with an exposure time of 500 ms. For data analysis, the ImageJ soware was used for counting the Cy5 and Texas Red uorescent molecules from an imaging region of 600 Â 600 pixels.</p><!><p>The reaction products were analyzed by 12% nondenaturing polyacrylamide gel electrophoresis (PAGE) in 1Â TBE buffer (9 mM Tris-HCl, 9 mM boric acid, 0.2 mM ethylenediaminetetraacetic acid, EDTA, pH 7.9) at a 110 V constant voltage at room temperature for 50 min. The gels were stained by SYBR gold and analyzed by a Bio-Rad ChemiDoc MP Imaging System (Hercules, CA, USA).</p><!><p>To evaluate the enzyme kinetic parameters of caspases, we measured the initial velocity in the presence of 0.025 U mL À1 caspase (caspase-8 or caspase-9) and different concentrations of the detection probes at 37 C for 5 min. The kinetic parameter is tted to the Michaelis-Menten equation.</p><p>where V max is the maximum initial velocity, [S] is the concentration of the detection probe, and K m is the Michaelis-Menten constant.</p><!><p>To evaluate the effect of inhibitor upon the caspase activity, different concentrations of the inhibitor (Z-IETD-FMK for caspase-8, Ac-LEHD-CMK for caspase-9, and broad-spectrum caspase inhibitor Q-VD-OPh for both caspase-8 and caspase-9) were preincubated with 0.025 U mL À1 caspase (caspase-8 or caspase-9) at room temperature for 15 min, respectively. Then 8 mL of detection probe 1-conjugated MBs or detection probe 2conjugated MBs were added into the mixture, and the reaction volume was adjusted to 20 mL with 1Â PBS, followed by the same detection procedure as described above. The relative activity (RA) of caspase was measured according to eqn (3).</p><p>where N 0 represents the Cy5 counts in the absence of caspase-8 or the Texas Red counts in the absence of caspase-9; N t represents the Cy5 counts in the presence of caspase-8 (0.025 U mL À1 ) or the Texas Red counts in the presence of caspase-9 (0.025 U mL À1 ); and N i represents the Cy5 counts in the presence of caspase-8 (0.025 U mL À1 ) + inhibitor or the Texas Red counts in the presence of caspase-9 (0.025 U mL À1 ) + inhibitor. The IC 50 value was calculated from the curve of RA versus the inhibitor concentration.</p><!><p>HeLa cells and MCF-7 cells were cultured in Dulbecco's Modied Eagle's Medium (DMEM, Life Technologies, USA) with 10% FBS (Life Technologies, USA) and 1% penicillin-streptomycin (Gibco, USA). Jurkat cell was cultured in 1640 cell medium (Life Technologies, USA) with 10% FBS (Life Technologies, USA) and 1% penicillin-streptomycin (Gibco, USA). The cells were cultured at 37 C in a humidied atmosphere containing 5% CO 2 . For real sample analysis, cells in the exponential phase of growth were collected and counted using Countstar BioTech Automated Cell Counter IC1000 (Shanghai, China), washed twice with ice-cold 1Â PBS, and centrifuged at 800 rpm for 5 min. To isolate cytoplasmic components from nuclear ones, the cells were treated with a nuclear protein extraction kit (Beyotime Biotechnology, Wuhan, China) and centrifuged at 3400 rpm for 15 min at 4 C. For STSinduced apoptosis analysis, cells were incubated in 5 mL of cell medium containing 0.4 mM STS for 4 h prior to the cell lysis procedure. The protein concentration was measured using a NanoDrop 2000c Spectrophotometer (Thermo Scientic, Wilmington, Delaware, USA).</p>

Royal Society of Chemistry (RSC)

Truly Random Degradable Vinyl Copolymers via Photocontrolled Radical Ring-Opening Cascade Copolymerization

Degradable vinyl polymers by radical ring-opening polymerization have become a promising solution to the challenges caused by the widespread use of non-degradable vinyl plastics. However, achieving even distribution of labile functional groups in the backbone of degradable vinyl polymers remains challenging. Herein, we report a photocatalytic approach to truly random degradable vinyl copolymers with tunable main-chain composition via radical ringopening cascade copolymerization (rROCCP). The rROCCP of the macrocyclic allylic sulfone and acrylates or acrylamides mediated by visible light at ambient temperature achieved near-unity reactivity ratios of both comonomers over the entire range of the comonomer compositions and afforded truly random vinyl copolymers with degradable units evenly distributed in the polymer backbone. Experimental and computational evidence revealed an unusual reversible inhibition of chain propagation by in situ generated sulfur dioxide, which was successfully overcome by reducing the solubility of sulfur dioxide in the reaction mixture. This study provided a powerful approach to truly random degradable vinyl copolymers with tunable main-chain labile functionalities and comparable thermal and mechanical properties to traditional non-degradable vinyl polymers.

truly_random_degradable_vinyl_copolymers_via_photocontrolled_radical_ring-opening_cascade_copolymeri

Introduction<!>PET-RAFT Polymerization of Macrocyclic Allylic Sulfones<!>Copolymerization of Macrocyclic Allylic Sulfones and Acrylates or Acrylamides<!>Understanding the Unusual Kinetic Behavior<!>Conclusion

<p>Vinyl polymers have been widely used in an array of applications including packaging, structural materials, synthetic fibers, coating, absorbent, and many others. While the all-carbon backbone makes vinyl polymers highly robust materials, it has also created significant challenges in their degradation, leading to critical environmental issues caused by plastic accumulation in landfills and the ocean. [1][2] Therefore, significant efforts have been made in recent years to develop innovative synthetic polymers that possess thermal and mechanical properties comparable to the original, nondegradable vinyl polymers and can undergo facile degradation at the end of their life cycle. [3][4] Among various approaches to degradable vinyl polymers, radical ring-opening polymerization (rROP) is of great interest. Attractive features of rROP include its abilities to incorporate labile functional groups (e.g. esters, thioesters, disulfide, etc.) into the polymer main chain [5][6][7] and interface with a plethora of reversible deactivation radical polymerization (RDRP) techniques for the synthesis of polymers with complex and defined macromolecular architectures. 8 Since the advent of rROP, various cyclic monomers have been successfully developed for the synthesis of degradable vinyl (co)polymers. 6 As a representative class of rROP monomer, cyclic ketene acetals (CKAs) have been extensively investigated since 1980s. 9 Despite recent progress made by Dove, [10][11][12][13][14] Nicolas, [15][16][17][18][19][20] and Sumerlin, [21][22] unfavorable reactivity ratios in the copolymerization of CKA with other vinyl monomers often lead to gradient or tapered compositions of the resultant copolymer. 23 The gradient composition in turn resulted in highly dispersed degradation products and large non-degradable fragments, as part of the copolymer lacked main-chain degradable units. Although new cyclic monomer classes including macrocyclic allylic sulfide (MAS) [24][25][26][27] and dibenzo[c,e]oxepane-5-thione (DOT) [28][29][30] (Figure 1A) have demonstrated promising properties, truly random copolymerization of these cyclic monomers with acrylates or acrylamides remains challenging. In 2018, we reported an approach to the radical ring-opening cascade polymerization of allylic sulfone macrocyclic monomers. 31 The radical cascade reaction of macrocyclic allylic sulfone could extrude sulfur dioxide (SO2) and generate a secondary alkyl radical capable of controlled chain propagation. 32 However, copolymerization of the macrocyclic allylic sulfone and acrylates exhibited unfavorable reactivity ratios at high temperatures. Therefore, it is essential to develop a method that provides access to truly random copolymers with tunable compositions and evenly distributed main-chain functional groups.</p><p>Recent studies suggest that temperature has a strong influence on the reactivity ratios in the radical copolymerization of cyclic and acyclic vinyl comonomers. 6 We reasoned that performing the copolymerization at lower temperatures would provide a key opportunity to modulate the reactivity ratios of vinyl comonomers. Therefore, we turned our attention to light-mediated polymerization techniques, as recent works have demonstrated that they are versatile tools to mediate controlled polymerization following radical, [33][34][35][36][37][38][39][40][41] cationic, [42][43][44][45][46] and metathesis pathways [47][48][49] at ambient temperature (Figure 1B). 50 In particular, we envisioned that the photoinduced electron/energy transfer-reversible addition/fragmentation chain transfer (PET-RAFT) polymerization developed by Boyer and coworkers [51][52][53][54][55][56] could be employed to mediate radical ring-opening cascade copolymerization (rROCCP) [57][58] of the macrocyclic allylic sulfone and acrylates or acrylamides (Figure 1C). Unlike the polymerization initiated by azobisisobutyronitrile (AIBN) that required high temperatures (80-100 °C) to maintain a sufficiently high rate of propagation, PET-RAFT can be performed at mild temperatures, thereby enabling favorable comonomer reactivity ratios in copolymerization. To the best of our knowledge, the photocontrolled rROCCP represents the first method that achieved truly random radical copolymerization of cyclic monomers and acrylic monomers over the entire range of comonomer compositions.</p><!><p>Our investigation began by screening various wellestablished photocatalysts to mediate the photocontrolled homopolymerization of allylic sulfone macrocyclic monomer 1 under visible light irradiation (Table S1). [51][52][53][54][55] We screened an array of photocatalysts, including fac-[Ir(ppy)3], Ru(bpy)3Cl2, ZnTPP, and Eosin Y, and identified fac-[Ir(ppy)3] as a promising photocatalyst for the reaction due to the excellent control over the polymerization when combined with CTA1. At a monomer/CTA ratio of 50:1, our initial attempt of the polymerization of macrocyclic allylic sulfone 1 mediated by fac-[Ir(ppy)3] and CTA1 under 450 nm light irradiation yielded P-1 with Mn (SEC) of 9.8 kg/mol and Ð of 1.11 (Table S2). Further examination of the reaction conditions found that optimal polymerization was achieved when the monomer concentration was at 0.2 M in DMF and the catalyst loading reached 200 ppm (Table S3-S5). Polymerization of 1 at other monomer/CTA ratios of 25:1, 100:1, and 200:1 successfully yielded polymers with predictable Mn and low Ð, demonstrating excellent control over the polymerization (Table S6). Similarly, macrocyclic allylic sulfone 2 with a smaller ring size was also polymerized with good control under the same conditions (Table S7). It is noteworthy that no ring-retaining propagation of both allylic sulfone macrocyclic monomers 1 and 2 has been observed.</p><p>Following the exploration of reaction conditions, we examined the living characteristics of the polymerization. First, the kinetic analysis revealed that the polymerization of 1 deviated from first-order kinetics in the late stage (Figure S1). This observation was consistent with our previous results when the cascade polymerization of macrocyclic allylic sulfone was thermally initiated. 31,59 Despite the kinetic anomaly, the polymerization of 1 still exhibited a linear increase of Mn with respect to the monomer conversion and remained low Ð throughout the reaction, suggesting that control over the polymerization was well maintained even after the rate decreased in the late stage (Figure 2A). 1 H-NMR analysis of P-1-6k (M n (SEC) = 6.4 kg/mol, Ð = 1.07) confirmed the fidelity of the chain end groups (2.46 and 1.21 ppm for a-chain end and 4.81 and 3.36 ppm for w-chain end, Figure S2), an important indicator of controlled polymerization. Besides, the discrete oligomers of P-1-5k (M n (SEC) = 5.5 kg/mol, Ð = 1.10) observed by matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometry showed masses consistent with the predicted values of these oligomers with intact chain ends (Figure 2B). Furthermore, chain extension of the macroinitiator P-2-4k (M n (SEC) = 3.9 kg/mol, Ð = 1.16) by 1 exhibited a clear shift to the higher molecular weight region on the SEC chromatogram, suggesting the formation of a diblock copolymer P-2-b-P-1 (M n (SEC) = 13.0 kg/mol, Ð = 1.20, Figure 2C). Finally, the reaction exhibited excellent temporal control: chain propagation completely halted when the light was switched "off"; polymerization resumed efficiently after the light was switched back "on" (Figure 2D). Taken together, these results unambiguously supported that the PET-RAFT polymerization of macrocyclic allylic sulfones maintained an excellent control throughout the reaction despite the deviation from first-order kinetics at the late stage.</p><!><p>Building upon the results of photocontrolled homopolymerization of macrocyclic allylic sulfones, we then investigated copolymerization of 1 and various acrylates or acrylamides (denoted hereafter as comonomer B). First, 1 was copolymerized with methyl acrylate (MA) at the feed composition of 𝑓 𝟏 " = 0.05, where 𝑓 𝟏 " is the molar fraction of 1 in the initial comonomer mixture, yielding copolymer P-1-co-MA with Mn (SEC) of 44.0 kg/mol and Ð of 1.28 (Table 1, entry 1). The propagation of both comonomers demonstrated first-order kinetics throughout the copolymerization (Figure 3A). The molecular weight also increased linearly with respect to the overall monomer conversion, which is defined by Eq 1: where [1(t)] and [B(t)] are the respective instantaneous concentrations of 1 and comonomer B at time t, and [1(0)] and [B(0)] are the respective initial concentrations of 1 and comonomer B (Figure 3B). Importantly, the instantaneous molar fraction of 1 incorporated in the copolymer (denoted hereafter as 𝐹 𝟏 ) remained identical to 𝑓 𝟏 " throughout the copolymerization (Figure S3). Correspondingly, the final copolymer composition, 𝐹 𝟏 (*+,) , when the reaction reached the end point, was also identical to 𝑓 𝟏 " (Table 1, entry 1). These results suggested that the reactivities of the two comonomers are highly similar in chain propagation. To determine the reactivity ratios of the copolymerization, the compositional data of 1 and B throughout the copolymerization was fitted to the Beckingham−Sanoja−Lynd (BSL) integrated model reported by Lynd et al. 60 𝑐𝑜𝑛𝑣. = 1 − 𝑓 𝟏 " ,</p><p>𝑐𝑜𝑛𝑣. = 1 − 𝑓 𝟏 " ,</p><p>where 𝑟 𝟏 and 𝑟 ) are reactivity ratios of 1 and comonomer B. It is noteworthy that although the BSL model is derived for ideal copolymerization where 𝑟 . × 𝑟 ) = 1, such as ionic or metalcatalyzed copolymerization systems, we reasoned that the copolymerization of the macrocyclic allylic sulfone and acrylic monomers is a close approximation of the ideal copolymerization, because the allylic sulfone motif was designed such that the propagating secondary alkyl radical formed after the radical cascade process is structurally similar to the propagating radical of polyacrylates. 61 Independent fitting of the polymer compositional data to Eq 2 and Eq 3 supported this rationale, as the derived reactivity ratios of the comonomers were 𝑟 𝟏 = 1.07 and 𝑟 ) = 0.94, with 𝑟 𝟏 ´ 𝑟 ) = 1.006 (Figure 3C). These results suggest that the copolymerization is truly random and that it is indeed highly analogous to an ideal copolymerization in which the product of the two reactivity ratios equals 1. The reactivity ratios of 1 and MA in the entire range of monomer feed compositions (𝑓 𝟏 " = 0-1) remained close to unity ( To investigate how degradability was influenced by the composition and distribution of degradable building blocks in copolymers, copolymers were treated with sodium methoxide to cleave the main-chain esters. SEC analysis of the degradation of the copolymer P-(1-co-MA) prepared by the photocontrolled rROCCP ( 𝐹 𝟏 (*+,) = 0.10, Mn (SEC) = 19.2 kg/mol, and Ð =1.27) exhibited a dramatic molecular weight reduction after degradation, resulting in oligomers with Mn (SEC) of 1.3 kg/mol and Ð of 1.33 (Table 2, entry 2 & Figure 4A). In contrast, degradation of the copolymer with a similar overall composition ( 𝐹 𝟏 (*+,) = 0.08, Mn (SEC) = 16.4 kg/mol, and Ð =1.52) generated by the thermallyinitiated copolymerization produced frag ments with higher Mn and Ð (Mn (SEC) = 7.3 kg/mol, Ð = 1.98) (Figure 4A). Furthermore, the degradation of copolymers with different comonomer compositions generated by the photocontrolled rROCCP consistently produced fragments with low Mn and narrow molecular weight distributions (Table 2). These results indicated that while the thermally initiated copolymerization yielded a gradient copolymer that could only be partially degraded, copolymers generated by the photocontrolled rROCCP possessed even and tunable distributions of main-chain degradable functionalities and could be degraded efficiently into low molecular weight fragments.</p><p>The thermal properties of copolymers were further evaluated by thermogravimetry (TGA) and differential scanning calorimetry (DSC) analyses. P-(1-co-MA) with main-chain degradable functionalities at different copolymer compositions established similar thermal stability comparable to polymethylacrylate with 5% weight loss decomposition temperature (Td) between 363-368 °C (Figure 4B). Furthermore, glass transition temperature (Tg) of P-(1-co-MA) can be fine-tuned by the initial comonomer feed composition in copolymerization, highlighting the potential utility of this method in generating degradable vinyl polymers with tailormade material properties (Figure 4C).</p><!><p>Our studies have shown that while the PET-RAFT homopolymerization and copolymerization (Figure S3 & S20-26) involving macrocyclic allylic sulfones deviated from the first-order kinetics, the polymerization remained well-controlled. This phenomenon was in stark contrast to traditional controlled polymerization in which deviation of first-order kinetics is usually a sign of loss of control, suggesting an unusual kinetic behavior that warranted further investigation. We suspected that the in situ generated SO2 in the radical cascade polymerization affected the reaction kinetics. To investigate this hypothesis, Density Functional Theory (DFT) calculations were carried out using the M06-2X/6-311++G(d,p)//B3LYP/6-31G(d) method in conjunction with the Solvation Model based on Density (SMD) simulating the effect from DMF to compute a plausible potential energy surface of the cascade process in the polymerization of macrocyclic allylic sulfones (Figure 5B). 65 Our calculation showed that the ascission/SO2 extrusion step (G2 to G3) has a low energy barrier of 5.9 kcal/mol, and that this transformation is exergonic by 2.8 kcal/mol. The low activation energy and relatively small change in Gibbs free energy indicates that this step is likely reversible. The DFT calculations also suggest that G3, with the lowest energy in the whole cascade process, exists at a high enough concentration during steady-state conditions, making it a plausible intermediate for chain propagation (Figure S27). Compared to chain propagation (G3-TS4-G4, with an energy barrier of 20.7 kcal/mol), two alternative reaction pathways of G3 with lower energy barriers are the reversible addition by the CTA (G3-TS5-G5, with an energy barrier of 12.0 kcal/mol) or SO2 (G3-TS3-G2, with an energy barrier of 8.7 kcal/mol). While the former serves as the reversible deactivation of the chain propagation to achieve controlled polymerization, the latter is a reverse reaction of the ascission/SO2 extrusion step and regenerates the sulfonyl radical G2. Because of a high energy barrier of 19.7 kcal/mol and being endergonic by 9.8 kcal/mol, chain propagation of G2 by the monomer (G2-TS6-G6) is prohibited thermodynamically and kinetically. These results indicate that excess SO2 in the reaction could indeed recombine with the propagating alkyl radical to regenerate the sulfonyl radical and inhibit chain propagation.</p><p>To provide further evidence of the presence and accumulation of sulfonyl radical over the course of reaction, we employed electron paramagnetic resonance (EPR) to monitor the evolution of radical species in the reaction in situ (Figure 5C). In the early stage (initial two hours) of the reaction, the EPR spectrum only consisted of signals corresponding to the alkyl radical a (g0 = 2.004) and the degenerative intermediate b (g0 = 2.009) (Spectrum I, Figure 5C). The g-values of the peaks and patterns of the spectrum are consistent with the radical species generated in the radical polymerization of MA. The EPR spectrum gradually evolved as the polymerization proceeded. In the late stage (after five hours) of the reaction, a new peak c with a gvalue of 2.014 appeared in the EPR spectrum (Spectrum II, Figure 5C), which is consistent with the g-value of the sulfonyl radical reported in literature. 66 Furthermore, simulated EPR spectra (dotted lines) based on the absence and presence of the sulfonyl radical in the reaction perfectly fit the experimental data as shown in Spectrum I and II, respectively, confirming the proposed assignments. Notably, Spectrum II is also consistent with Spectrum III obtained after the exogenous SO2 gas was introduced to the system at the early stage of the reaction (Figure 5C). Collectively, the DFT calculations and EPR analyses are consistent with the observed kinetic results, confirming that G2 (peak c in Figure 5C), G3 (peak a in Figure 5C), and G5 (peak b in Figure 5C) are long-lived radical intermediates in the polymerization of macrocyclic allylic sulfones, and that the concentration of SO2 could have a significant effect on the direction of the reaction. Overcoming the Propagation Inhibition by SO2.</p><p>Based on DFT calculations, we reason that the propagation inhibition by the in situ generated SO2 may be reversible, given the low energy barrier of the process. This reversibility implies that the extrusion of SO2 and the formation of the alkyl radical are favored at low SO2 concentrations, whereas the recombination of SO2 and the formation of the sulfonyl radical are favored at high SO2 concentrations. Therefore, the propagation inhibition could be alleviated by removing SO2 from the reaction. Indeed, we found that sparging the reaction mixture with argon steadily increased the rate of the PET-RAFT homopolymerization of 1 in the late stage of the reaction at 25 °C (Figure S28). In fact, both the SO2 inhibition and reactivation of chain propagation by argon sparging were reversible and the polymerization could be switched "on"/"off" by alternating the exogenous SO2 and argon introduced into the reaction vessel (Figure S29-S30). Similarly, the propagation inhibition was also alleviated in the copolymerization of 1 and BnA (𝑓 𝟏 " = 0.09) by argon sparging at 25 °C (Figure S31). Additionally, increasing the reaction temperature to 50 °C was also found to improve the rate of the homopolymerization of 1 in the late stage (Figure S32-S33).</p><p>Combining the argon sparging and the temperature elevation to 50 °C proved to further improve the reaction kinetics of the homopolymerization of 1, allowing it to remain pseudo first-order throughout the reaction (Figure 6A). The rate of copolymerization of 1 and MA was also improved when the reaction temperature was elevated to 50 °C (Figure S34), but a modest deviation of the comonomer reactivity ratios from unity was observed (Figure S35). We reasoned that an alternative strategy to reduce the propagation inhibition by SO2 was to switch the solvent from DMF to dioxane, in which SO2 has lower solubility (Figure S36). Encouragingly, we found that the kinetics of copolymerization of 1 and BnA at 25 °C remained pseudo first-order throughout the reaction when dioxane was used as the solvent (Figure 6B).</p><!><p>A novel approach to the truly random degradable vinyl polymers with tunable main-chain composition via photocontrolled radical ring-opening cascade copolymerization (rROCCP) is presented in this article. Compared to existing rROP systems, the photocontrolled rROCCP enabled the synthesis of truly random degradable vinyl copolymers with evenly distributed, tunable composition of the main-chain labile groups at ambient temperature. Computational and EPR analyses revealed that the reversible inhibition of the chain propagation by in situ generated SO2 caused an unusual kinetic behavior that showed a deviation from first-order kinetics in the late stage of the reaction. Removal of SO2 was found to reverse the inhibition of the chain propagation and improve the reaction kinetics in both the homopolymerization and the copolymerization involving macrocyclic allylic sulfones. Taken together, excellent control and favorable comonomer reactivity ratios make photocontrolled rROCCP a powerful strategy for the preparation of truly random degradable vinyl copolymers with tunable main-chain compositions for a wide range of applications. In addition, the mechanistic insights into the reversible inhibition of chain propagation by SO2 shed light on using chemical cues to control radical chain-growth cascade polymerization systems.</p>

ChemRxiv

RSU-1 interaction with prohibitin-2 links cell–extracellular matrix detachment to downregulation of ERK signaling

Cell–extracellular matrix (ECM) detachment is known to decrease extracellular signal–regulated kinase (ERK) signaling, an intracellular pathway that is central for control of cell behavior. How cell–ECM detachment is linked to downregulation of ERK signaling, however, is incompletely understood. We show here that focal adhesion protein Ras Suppressor 1 (RSU1) plays a critical role in cell–ECM detachment induced suppression of ERK signaling. We have identified prohibitin 2 (PHB2), a component of membrane lipid rafts, as a novel binding protein of RSU1, and mapped a major RSU1-binding site to PHB2 amino acids 150 to 206 in the C-terminal region of the PHB/SPFH (stomatin/prohibitin/flotillin/HflKC) domain. The PHB2 binding is mediated by multiple sites located in the N-terminal leucine-rich repeat region of RSU1. Depletion of PHB2 suppressed cell–ECM adhesion–induced ERK activation. Furthermore, cell–ECM detachment increased RSU1 association with membrane lipid rafts and interaction with PHB2. Finally, knockout of RSU1 or inhibition of RSU1 interaction with PHB2 by overexpression of the major RSU1-binding PHB2 fragment (amino acids 150–206) effectively suppressed the cell–ECM detachment induced downregulation of ERK signaling. Additionally, expression of venus-tagged wild-type RSU1 restored ERK signaling, while expression of venus-tagged PHB2-binding defective RSU1 mutant in which the N-terminal leucine-rich repeat region is deleted did not. Taken together, Our findings identify a novel RSU1-PHB2 signaling axis that senses cell–ECM detachment and links it to decreased ERK signaling.

rsu-1_interaction_with_prohibitin-2_links_cell–extracellular_matrix_detachment_to_downregulation_of_

<!>Identification of PHB2 as an RSU1-binding protein<!><!>Cell–ECM detachment promotes RSU1 association with membrane lipid rafts<!><!>Cell–ECM detachment promotes RSU1 interaction with PHB2<!><!>RSU1 is a multifunctional protein that is positively involved in cell spreading, directional migration, and invasion<!>Discussion<!><!>Discussion<!>Cell culture<!>Antibodies, siRNAs, and other reagents<!>Generation of RSU1 KO cell lines<!>Generation of RSU1-Clover knock-in cell lines<!>Generation of mouse monoclonal anti-RSU1 antibody<!>DNA cloning<!>Yeast two-hybrid screen<!>Lipid raft isolation<!>Lipid raft staining<!>Generation of GST- or MBP-tagged RSU1 and PHB2 fusion proteins<!>GST-fusion protein binding assays<!>Western blotting<!>Coimmunoprecipitation assay<!>Immunofluorescent staining<!>Time-correlated single-photon fluorescence lifetime microscopy<!>Quantification of Western blotting and statistical analysis<!>Cell spreading assay using IncuCyte ZOOM apparatus<!>Cell spreading area analysis using actin staining<!>Cell migration<!>Cell invasion<!>Data availability<!>Conflict of interest<!>Supporting information

<p>Edited by Alex Toker</p><p>It has been well established that cell–extracellular matrix (ECM) adhesion is crucial for regulation of the extracellular signal–regulated kinase (ERK) signaling pathway, which is central for control of cell behavior including cell proliferation, migration, and survival (1, 2, 3, 4, 5, 6, 7, 8, 9). Detachment of cells from the ECM often results in downregulation of ERK activation and consequently alteration of cell behavior (1, 10, 11, 12). Aberrant ERK signaling and anchorage-independent growth are intimately associated with cancer development and progression (13, 14). Thus, elucidation of the molecular mechanisms underlying cell–ECM adhesion–mediated regulation of ERK signaling is important for understanding the pathogenesis of cancer and identifying therapeutic targets for control of cancer progression.</p><p>Cell–ECM adhesion triggers recruitment of a selective number of intracellular proteins to focal adhesions, where these proteins or protein complexes act as signaling intermediators regulating cell proliferation, migration, and survival (15, 16, 17, 18, 19, 20, 21). One of the key focal adhesion components is a tertiary protein complex composed of integrin-linked kinase (ILK), particularly interesting new cysteine-histidine-rich protein (PINCH), and parvin (hereinafter referred to as the ILK-PINCH-parvin complex) (19, 22, 23). The ILK-PINCH-parvin complex binds to other adaptor proteins as well as actin cytoskeleton, thereby regulating various signaling pathways critical for cell-cycle progression, cell migration, and survival (22, 23, 24, 25, 26). The formation of the ILK-PINCH-parvin complex is mediated by the interactions of the ILK amino- and carboxyl-terminal domains to PINCH and parvin, respectively (22). PINCH also interacts with RSU1, a leucine-rich repeat (LRR)–containing protein that was originally identified and named after its ability to suppress v-Ki-Ras–induced oncogenic transformation (27, 28, 29). The interaction with PINCH recruits RSU1 to focal adhesions (19, 30, 31, 32). Functionally, both clinical evidence and experimental studies in model organisms and cells have indicated that ILK, PINCH, and parvin play important roles in promoting tumor development and progression (22, 23, 25, 33, 34, 35). While RSU1 has been found to play a positive role in cancer cell migration and invasion (36), substantial evidence suggests that RSU1 may also function as a tumor suppressor. For example, loss or reduced expression or mutations of RSU1 was often found in human cancers (e.g., hepatocellular carcinoma and gliomas) (37, 38, 39). Furthermore, overexpression of RSU1 significantly reduced human breast cancer and glioblastoma cell growth and tumorigenic potential (39, 40). How RSU1 suppresses oncogenic transformation and tumor growth, however, is not well understood.</p><p>Lipid rafts are membrane microdomains found in eukaryotic cells, which are enriched in cholesterol, sphingolipid, and many lipid-linked proteins including caveolin (41, 42). It has been shown that membrane lipid rafts are discrete platforms for anchoring important signaling molecules such as ERK and its upstream kinases such as mitogen-activated protein kinase kinase (MEK) to transduce signals critical for cell proliferation, growth, and migration (43, 44, 45, 46). Of note, inhibition of caveolin-mediated internalization of lipid rafts effectively prevented cell–ECM detachment induced downregulation of ERK signaling (47, 48). PHB2 is a component of the ubiquitously expressed PHB complex that is present in various subcellular locations including membrane lipid rafts (49, 50). Structurally, PHB2 is composed of an amino-terminal hydrophobic sequence that facilities its localization to the membrane, a large PHB/SPFH domain known to have affinity for binding to lipid rafts, and a carboxyl-terminal coiled-coil domain required for the assembly of the PHB complex (49). Previous studies have shown that the PHB complex is essential for Ras-induced ERK activation (51, 52, 53). Furthermore, rocaglamides, natural compounds that potently inhibit proliferation of various cancer cells, have been shown to directly bind the PHB complex and thereby inhibit ERK signaling (54, 55).</p><p>In the current study, we show that RSU1 plays a critical role in cell–ECM detachment induced downregulation of ERK signaling. We have identified PHB2 as a novel binding protein of RSU1 and mapped a major RSU1-binding site to PHB2 amino acids 150 to 206 in the C-terminal region of the PHB/SPFH domain. In addition, we have found that multiple sites in the N-terminal LRR region of RSU-1 mediate the PHB2 binding. Functionally, depletion of PHB2 suppressed cell–ECM adhesion–induced ERK activation. Furthermore, cell–ECM detachment promoted RSU1 association with membrane lipid rafts and interaction with PHB2. Finally, knockout (KO) of RSU1 or inhibition of RSU1 interaction with PHB2 by overexpression of the RSU1-binding PHB2 fragment (amino acids 150–206) effectively suppressed the cell–ECM detachment–induced downregulation of ERK signaling. Expression of venus-tagged wild-type RSU1, but not that of venus-tagged PHB2-binding defective RSU1 mutant in which the N-terminal LRR region is deleted, restored cell–ECM detachment–induced downregulation of ERK signaling. Our results identify a novel RSU1-PHB2 signaling axis that senses cell–ECM detachment and links it to downregulation of ERK signaling.</p><p>RSU1 deficiency impairs cell–ECM detachment–induced downregulation of MEK and ERK signaling.A, verification of RSU1 deficiency in RSU1 KO cells by Western blotting. HT1080 (lane 1) and RSU1 KO (lane 2) cells were analyzed by Western blotting with anti-RSU1, anti-PINCH-1, and GAPDH antibodies as described in Experimental procedures. B, verification of RSU1 disruption in RSU1 KO cells by DNA sequencing. DNA sequencing revealed that a 351-nucleotide (nt) insertion was detected at the expected cleavage site (shown in red) in RSU1 KO cells resulting in frameshift and early termination of RSU1. The protospacer adjacent motif (PAM) is shown in red, and the gRNA-targeting sites are shown in green. C, RSU1-deficient cells failed to downregulate MEK and ERK signaling in response to cell detachment from the ECM. Wild-type and RSU1 KO HT1080 cells were either allowed to adhere to fibronectin (10 μg/ml) (Adh) or maintained in suspension in HEMA-coated cell culture dishes (Sus) for 5 h. The levels of total MEK and ERK and phosphorylated MEKSer 221 and ERKThr202/Tyr204 were determined by Western blotting. The densiometric ratios of phosphorylated MEKSer 221 to the total MEK and that of phosphorylated ERKThr202/Tyr204 to the total ERK were analyzed as described in Experimental procedures. In each data set, data were normalized to those observed in adherent cells. Differences between the attached and suspended cells were examined for statistical significance as described in Experimental procedures. n = 4 experiments, ∗p < 0.05. D, re-expression of RSU1 in RSU1-deficient cells restored downregulation of MEK–ERK signaling in response to loss of cell–ECM adhesion. RSU1 KO cells stably expressing venus-tagged RSU1 or venus alone were cultured in adhesion and suspension conditions, respectively, and the levels of total MEK and ERK and phosphorylated MEKSer 221 and ERKThr202/Tyr204 were assessed as described in (C). n = 4 experiments, ∗p < 0.05. ECM, extracellular matrix; ERK, extracellular signal–regulated kinase; HEMA, hydroxyethyl methacrylate; KO, knockout; MEK, mitogen-activated protein kinase kinase; NS, not significant.</p><p>RSU1 interacts with PHB2.A–B, coimmunoprecipitation of PHB2 (A) and PINCH-1 (B) with RSU1. HT1080 cells were harvested and lysed with octyl glucoside (OG)–containing lysis buffer. Endogenous RSU1 was immunoprecipitated with anti-RSU1 antibody and mouse IgG as a negative control, and the immunoprecipitants were analyzed by Western blotting with anti-RSU1 (A and B), anti-PHB2 (A), and anti-PINCH-1 (B) antibodies. The heavy chain of IgG is marked by asterisk. C, coimmunoprecipitation of RSU1 with PHB2. HT1080 cells were harvested and lysed with OG-containing lysis buffer. Endogenous PHB2 was immunoprecipitated with anti-PHB2 antibody and mouse IgG as a negative control, and the immunoprecipitants were analyzed by Western blotting with anti-RSU1 and anti-PHB2 antibodies. The heavy chain of IgG is marked by asterisk. D, Pull down of RSU1 by GST-PHB2. GST pull-down assay was performed as described in Experimental procedures. The samples (as indicated) were analyzed by Western blotting with anti-RSU1 antibody (left) and Coomassie blue staining (right), respectively. GST was served as a negative control. E, recombinant MBP-fusion proteins containing full-length (FL) PHB2, PHB2 N-terminal region (aa 1–152), middle region (aa 150–206), or C-terminal region (aa 201–299) or PHB2 mutant in which aa 150 to 206 are deleted (Δ150–206) were expressed and purified, and their interactions with GST-tagged RSU1 were analyzed using a pull-down assay as described in the Experimental procedures. MBP was used as a negative control. The samples were analyzed by Western blotting with anti-RSU1 antibody (top) or Coomassie blue staining (middle). In the Western blot, a protein band with molecular mass identical to MBP-PHB2 C-terminal region (aa 201–299), which is smaller than that of the GST-RSU1 band, was indicated with asterisk (lane 6). This band was likely caused by a cross-reactivity of the anti-RSU1 antibody used for the Western blotting. Bottom panel, a schematic drawing summarizing the RSU1-binding activities of the FL and deletion mutants of PHB2 used in the pull-down assay. The PHB/SPFH domain (amino acid residues 39–201) known to have affinity for binding to lipid rafts is marked. F, recombinant GST-fusion proteins containing different regions of RSU1 (as indicated) were used to pull down venus-PHB2 expressed in HT1080 cells. The samples were analyzed by Western blotting with anti-PHB2 antibody (top) and Coomassie blue staining (middle). Bottom panel, a schematic drawing summarizing the PHB2-binding activities of the FL and deletion mutants of RSU1 used in the pull-down assay. MBP, maltose-binding protein; PINCH, particularly interesting new cysteine-histidine-rich protein; PHB2, prohibitin 2; SPFH, stomatin/prohibitin/flotillin/HflKC.</p><!><p>To test whether RSU1 directly interacts with PHB2, we expressed recombinant GST-tagged RSU1 and maltose-binding protein (MBP)–tagged PHB2 in Escherichia coli, respectively, and analyzed the ability of purified GST-RSU1 to interact with MBP–PHB2 in pull-down assays. The results showed that GST-RSU1 was readily pulled down by MBP–PHB2 (Fig. 2E, lane 3) but not by MBP (Fig. 2E, lane 2), suggesting that RSU1 binds to PHB2 directly. Next, we generated MBP-tagged PHB2 fragments containing the N-terminal region (amino acids 1–152), the middle region (amino acids 150–206), or the C-terminal region (amino acids 201–299) and tested their RSU1-binding activity using the pull-down assays. The results showed that GST-RSU1 was pulled down by MBP-tagged PHB2 fragment containing the middle region (amino acids 150–206) (Fig. 2E, lane 3) but not by that containing the C-terminal region (amino acids 201–299) (Fig. 2E, lane 4) or the N-terminal region (amino acids 1–152) (Fig. 2E, lane 6), suggesting that PHB2 middle region (amino acids 150–206) contains a major RSU1-binding site. However, GST-RSU1 was pulled down by MBP-tagged PHB2 mutant in which amino acids 150 to 206 were deleted (Δ(150–206)) (Fig. 2E, lane 5), albeit in a smaller amount compared with that of GST-RSU1 pulled down by the PHB2 middle region (amino acids 150–206) (Fig. 2E, compare lane 3 with lane 5), suggesting that deletion of amino acids 150 to 206 does not completely eliminate RSU1 binding.</p><p>Next, we sought to identify the RSU1 region that mediates PHB2 binding. To do this, we generated GST-tagged proteins containing various RSU1 regions (Fig. 2F) and tested their PHB2-binding activity in a pull-down assay. The results showed that GST-tagged RSU1 N-terminal LRR region (amino acids 1–208) (Fig. 2F, lane 4), but not GST-tagged RSU1 C-terminal region (amino acids 209–277) (Fig. 2F, lane 3), pulled down venus-tagged PHB2. In control experiments, venus-tagged PHB2 was pulled down by GST-RSU1 (Fig. 2F, lane 2) but not by GST (Fig. 2F, lane 1), confirming the specificity of the pull-down assay. Venus-tagged PHB2 was also pulled down by GST-tagged LRR1-LRR2 (amino acids 1–66) (Fig. 2F, lane 6) or LRR3-LRR8 (amino acids 67–208) (Fig. 2F, lane 7) but not LRR1 (amino acids 1–43) (Fig. 2F, lane 5) or LRR2 (amino acids 44–66) (Fig. 2F, lane 8). Collectively, these results suggest that the PHB2 binding is mediated by multiple LRRs located in the N-terminal region (amino acids 1–208) of RSU1.</p><!><p>PHB2 silencing impairs cell–ECM adhesion–induced MEK and ERK activation.A, silencing PHB2 with siRNA results in downregulation of MEK and ERK activation under basal condition. HT1080 cells transfected with control siRNA (siControl) or two different PHB2 siRNAs (siPHB2-1 and siPHB2-2) for 48 h were harvested and examined for the levels of total MEK and ERK and phosphorylated MEKSer 221 and ERKThr202/Tyr204 by Western blotting. The densiometric ratio of PHB2 to GAPDH was used to indicate the knockdown efficiency of PHB2. The densiometric ratio of phosphorylated MEKSer 221 to the total MEK and that of phosphorylated ERKThr202/Tyr204 to the total ERK were analyzed as described in Experimental procedures. In each data set, data were normalized to those observed in cells transfected with siControl. Differences between the adherent and suspended cells were examined for statistical significance as described in Experimental procedures. n = 5 experiments, ∗p < 0.05, ∗∗p < 0.005, ∗∗∗p < 0.001. B, PHB2 silencing impairs cell–ECM adhesion–induced MEK and ERK activation. HT1080 cells transfected with control siRNA (siControl) or two different PHB2 siRNAs (siPHB2-1 and siPHB2-2) for 48 h were trypsinized and then were either maintained in suspension in HEMA-coated cell culture dishes (Sus) or allowed to adhere to fibronectin (10 μg/ml) (Adh) for 5 h. The cells were analyzed by Western blotting with antibodies as indicated. The densiometric ratio of phosphorylated MEKSer 221 to the total MEK and that of phosphorylated ERKThr202/Tyr204 to the total ERK were analyzed as described in Experimental procedures. n = 3 experiments, ∗p < 0.05. ∗∗∗∗p < 0.0001. ECM, extracellular matrix; ERK, extracellular signal–regulated kinase; HEMA, hydroxyethyl methacrylate.</p><p>Cell–ECM detachment promotes RSU1 association with membrane lipid rafts.A, distribution of caveolin, PHB2, and RSU1 in lipid rafts and cytosol prepared from HT1080 cells that were either maintained in suspension or allowed to adhere to fibronectin (10 μg/ml). Lipid rafts were isolated from the total membrane fraction as described in Experimental procedures. The blots were probed with antibodies against caveolin-1, GAPDH, PHB2, and RSU1. Noted that the amount of caveolin-1 associated with lipid rafts was similar in both adherent and suspended cells. The densiometric ratio of PHB2 or RSU1 protein levels relative to caveolin-1 was analyzed as described in Experimental procedures. Noted that while PHB2 present in lipid rafts remained essentially the same in suspended and attached cells, the amount of RSU1 detected in lipid rafts was significantly increased in suspended cells compared with that in attached cells. n = 5 experiments, ∗p < 0.05, B, Colocalization of RSU1-Clover with lipid rafts. The RSU1-Clover HT1080 cells in which the DNA sequence encoding Clover was inserted immediately to the 3ʹof RSU1 loci were maintained in suspension or allowed to adhere to fibronectin (10 μg/ml) for 5 h before incubation with cholera toxin B subunit (CTxB), a marker for lipid rafts, as described in Experimental procedures. Cells were analyzed by confocal microscopy, and representative images were shown. Note that an increased fraction of RSU1-Clover was colocalized with Alexa555-CTxB in suspended cells. Bars = 20 μm, 5 μm, or 10 μm as indicated in the figure. NS, not significant.</p><!><p>To further analyze this, we engineered cells in which Clover, a green fluorescence tag, was inserted immediately to the 3ʹof RSU1 loci to allow tracing subcellular localization of endogenous RSU1 (hereinafter RSU1-Clover). As expected, in cells that were adhered to fibronectin, abundant RSU1-Clover was detected in focal adhesions (Fig. S2). To detect lipid rafts by florescence confocal microscopy, we stained the cells with cholera toxin B subunit (CTxB), a marker for lipid rafts. Consistent with the biochemical analyses of lipid rafts (Fig. 4A, lanes 1 and 3), confocal microscopic analyses of RSU1-Clover and CTxB-positive lipid rafts showed that RSU1-Clover was largely not colocalized with the lipid rafts in adhered cells (Fig. 4B, left panels). However, significantly more RSU1-Clover was found to colocalize with CTxB-positive membrane lipid rafts in response to loss of cell–ECM adhesion (Fig. 4B, right panels). These results are highly consistent with those of the biochemical assays (Fig. 4A), and collectively, they suggest that loss of cell–ECM adhesion promotes RSU1 association with membrane lipid rafts.</p><!><p>Cell–ECM detachment promotes RSU1 interaction with PHB2.A, FRET analysis of RSU1 interaction with PHB2. HT1080 cells were cotransfected with mClover-N1 vector carrying RSU1 (RSU1-Clover) and Ruby2-N1 vector carrying PHB2 (PHB2-Ruby2) or Ruby2-N1 vector alone as a negative control. The transfected cells were allowed to adhere to fibronectin (10 μg/ml) or maintained in suspension for 5 h before fixation with 4% PFA as described in Experimental procedures (A). Bars = 20 μm or 10 μm as indicated in the figure. The fluorescence lifetime (τ) of RSU1-Clover was measured as described in Experimental procedures. The mean ± SEM of τ is plotted. In adherent cells, τRSU1-Clover/Ruby2 = 2.226 ± 0.02 ns (n = 33), τRSU1-Clover/PHB2-Ruby2 = 2.234 ± 0.018 ns (n = 35). In suspended cells, τRSU1-Clover/Ruby2 = 2.315 ± 0.019 ns (n = 44), τRSU1-Clover/PHB2-Ruby2 = 2.143 ± 0.015 ns (n = 46). Noted that in suspended cells, τRSU1-Clover in the presence of PHB2-Ruby2 was significantly reduced compared with that in the presence of Ruby2 control. ∗∗∗∗p < 0.0001, FRET efficiency (E) was calculated as described in Experimental procedures. n = 3, ∗∗p < 0.005. B, coimmunoprecipitation of PHB2 with endogenous RSU1 in attached and suspended HT1080 cells, respectively. HT1080 cells were trypsinized and seeded on fibronectin (10 μg/ml) or maintained in suspension for 5 h. The cells were then harvested and lysed with octyl glucoside (OG)–containing lysis buffer for immunoprecipitation experiments with anti-RSU1 antibodies as described in Experimental procedures. The anti-RSU1 IgG and control IgG immunoprecipitants were analyzed by Western blotting. Note that the amount of PHB2 coprecipitated with RSU1 in suspended cells was more than that in attached cells. C and D, coimmunoprecipitation of PHB2 (C) or PINCH-1 (D) with Flag-RSU1 in attached and suspended HT1080 cells that express Flag-RSU1. HT1080 cells were transfected with Flag-RSU1 or control Flag vector and analyzed by immunoprecipitation with anti-Flag antibodies. The immunoprecipitants were analyzed by Western blotting with antibodies as indicated. Note that the amount of PHB2 (compare lanes 7 and 8 in panel C) but not that of PINCH-1 (compare lanes 4 and 8 in panel D) coprecipitated with Flag-RSU1 was increased in response to cell–ECM detachment. ECM, extracellular matrix; FRET, Förster resonance energy transfer; PFA, polyformaldehyde; PINCH, particularly interesting new cysteine-histidine-rich protein.</p><!><p>To further test this, we analyzed the interaction between RSU1 and PHB2 in cells that were adhered to the ECM or maintained in suspension by coimmunoprecipitation. The RSU1 immunoprecipitants were analyzed by Western blotting with antibodies for RSU1 and PHB2, respectively. The results showed that the amount of PHB2 coimmunoprecipitated with RSU1 was increased in suspended cells compared with that in adherent cells (Fig. 5B, compare lanes 3 and 4). As an additional test, we transfected HT1080 cells with expression vectors encoding FLAG-RSU1 or FLAG only as a control, allowed them to adhere to fibronectin or maintained them in suspension, and analyzed the interaction between PHB2 and FLAG-RSU1 by coimmunoprecipitation with anti-FLAG antibodies (Fig. 5C). Again, significantly more PHB2 was coimmunoprecipitated with FLAG-RSU1 in the sample derived from the suspended cells than that derived from the cells adhered to fibronectin (Fig. 5C, compare lane 8 with lane 7). These results are highly consistent with those of the FRET experiments, and collectively, they demonstrate that loss of cell–ECM adhesion promotes RSU1 interaction with PHB2. In parallel experiments, the amount of PINCH-1 coimmunoprecipitated with FLAG-RSU1 was not increased in response to cell–ECM detachment (Fig. 5D, compare lane 8 with lane 4), suggesting that cell–ECM detachment selectively enhances the interaction of RSU1 with PHB2.</p><!><p>PHB2-binding defective RSU1 deletion mutant is unable to suppress MEK/ERK activation in response to loss of cell–ECM adhesion. RSU1 KO cells re-expressing venus-tagged PHB2-binding defective RSU1 mutant (209–277) or venus were allowed to adhere to fibronectin (10 μg/ml) (Adh) or maintained in suspension in HEMA-coated cell culture dishes (Sus), and the levels of total MEK and ERK and phosphorylated MEKSer 221 and ERKThr202/Tyr204 were analyzed by Western Blotting and quantified as described in Figure 1D. n = 3 experiments. ECM, extracellular matrix; ERK, extracellular signal–regulated kinase; KO, knockout; HEMA, hydroxyethyl methacrylate; NS, not significant.</p><p>Overexpression of GFP-PHB2 fragment (150–206) reduces the RSU1–PHB2 interaction and suppresses cell–ECM detachment–induced downregulation of ERK signaling.A, coimmunoprecipitation of RSU1 with PHB2 fragment (150–206). HT1080 cells were transfected with vectors encoding GFP or GFP-PHB2 fragment (150–206) for 24 h and then analyzed by immunoprecipitation with anti-GFP antibodies. The immunoprecipitants were analyzed by Western blotting with anti-RSU1 and anti-GFP antibodies. B, overexpression of PHB2 fragment (150–206) disrupts the RSU1–PHB2 interaction. HT1080 cells were transfected with vectors encoding GFP or GFP-PHB2 fragment (150–206) for 24 h and then analyzed by immunoprecipitation with anti-RSU1 antibodies. The immunoprecipitants were analyzed by Western blotting with anti-PHB2 and RSU1 antibodies. Note that the amount of PHB2 coimmunoprecipitated with RSU1 was reduced in the presence of the GFP-PHB2 fragment (150–206) (compare lanes 5 and 6). C and D, overexpression of the GFP-PHB2 fragment (150–206) (C) but not that of the GFP-PHB2 fragment (201–299) (D) attenuates cell–ECM detachment–induced suppression of MEK-ERK activation. HT1080 cells transfected with vectors encoding GFP, GFP-PHB2 (150–206), or GFP-PHB2 (201–299) for 24 h were trypsinized and replated on fibronectin (10 μg/ml) (Adh) or maintained in suspension in HEMA-coated cell culture dishes (Sus) for 5 h. The cells were analyzed by Western blotting as indicated. The levels of total MEK and ERK and phosphorylated MEKSer 221 and ERKThr202/Tyr204 were assessed and quantified as in Figure 1. n = 4 to 9 experiments. ∗∗∗∗p < 0.0001, ∗∗p < 0.005, ∗p < 0.01. ECM, extracellular matrix; ERK, extracellular signal–regulated kinase; NS, not significant.</p><p>RSU1 and its interaction with PHB2 interaction suppresses ERK activation in cells grown in 3D culture.A, wild-type or RSU1 KO HT1080 cells were grown in 3D culture as described in Experimental procedures. The levels of total MEK and ERK and phosphorylated MEKSer 221 and ERKThr202/Tyr204 were determined by Western Blotting. The densiometric ratio of phosphorylated MEKSer 221 to the total MEK and that of phosphorylated ERKThr202/Tyr204 to the total ERK were analyzed as in Figure 1. n = 3 experiments, ∗p < 0.05. B, re-expression of RSU1 in RSU1 KO cells suppresses ERK activation in cells grown in 3D culture. RSU1 KO cells stably expressing venus-RSU1 or venus alone were cultured in 3D, and the levels of total MEK and ERK and phosphorylated MEKSer 221 and ERKThr202/Tyr204 were assessed as in (A). n = 4 experiments, ∗∗∗p < 0.001, ∗p < 0.05. C, overexpression of the GFP-PHB2 fragment (150–206) that reduces the RSU1–PHB2 interaction increases ERK activation in cells grown in 3D culture. HT1080 cells transfected with GFP or GFP-PHB2 (150–206) for 24 h were trypsinized and then cultured in 3D as described in Experimental procedures. The levels of total MEK and ERK and phosphorylated MEKSer 221 and ERKThr202/Tyr204 were assessed and quantified as in (A). n = 6 experiments. ∗∗∗p < 0.001, ∗p < 0.05. ECM, extracellular matrix; ERK, extracellular signal–regulated kinase; KO, knockout; NS, not significant.</p><p>RSU1 promotes cell spreading, directional migration, and invasion.A, the spreading of wild-type and RSU1 KO HT1080 cells on culture dishes coated with fibronectin (30 μg/ml) were analyzed with IncuCyte ZOOM apparatus over the course of 5 h. Cell surface areas (mean ± SE) of wild-type HT1080 cells and RSU1 KO cells were quantified as described in Experimental procedures. B, cells (as indicated in the figure) were seeded on fibronectin (30 μg/ml)-coated culture dishes overnight and then stained with Alexa 32 Fluor555–conjugated phalloidin. Cell area was quantified using ImageJ software (NIH). Data are shown as mean ± SEM of cell areas measured in 4 to 6 randomly selected fields. ∗∗∗∗p < 0.0001. C, cell tracks of migrating wild-type HT1080 (n = 65) and RSU1 KO (n = 44) cells. The directional persistence (directionality) and mean migration speed during cell migration were measured. The mean ± SEM is plotted in red. ∗p < 0.05. D, wild-type HT1080, RSU1 KO, and RSU1 KO cells expressing venus or venus-RSU1 were analyzed in an invasion assay as described in Experimental procedures. ∗p < 0.05, ∗∗p < 0.01. KO, knockout; NS, not significant.</p><!><p>Next, we assessed the effect of RSU1 deficiency on cell migration by live-cell imaging. The migration speed and directionality (i.e., directional persistence, which can be determined by the ratio of Euclidian distance to accumulated migration distance for each track (58)) of individual cells were tracked. The results showed that while the mean migration speed was not reduced in RSU1 KO cells compared with that of wild-type HT1080 cells, the directional persistence was significantly impaired in response to loss of RSU1 (Fig. 9C). Consistent with the impaired directionality of cell migration, the invasion of RSU1 KO cells through matrigel was inhibited in response to loss of RSU1, which was restored when venus-tagged RSU1 was re-expressed in RSU1 KO cells (Fig. 9D). Taken all together, these results suggest that RSU1 is positively involved in cancer cell spreading, directional migration, and invasion through ECM.</p><!><p>Aberrant activation of ERK signaling, which is critical for control of cell behavior including cell growth, migration, and survival, plays a pivotal role in the pathogenesis and progression of cancer (9, 59, 60). It has been well established that cell–ECM adhesion exerts strong influence on ERK activation and detachment of cells from the ECM often results in downregulation of ERK activation (1, 2, 3, 4, 5, 6, 7, 8, 9). Indeed, aberrant ERK signaling contributes to anchorage-independent cell growth, a hallmark of malignant transformation. However, despite the importance, our understanding of the molecular mechanism by which cell–ECM detachment regulates ERK activation remains rather incomplete. The studies presented in this paper have shed important new insights into this process. Specifically, we have identified RSU1, an evolutionally conserved protein localized to cell–ECM adhesion, as a key mediator of cell–ECM detachment–induced downregulation of ERK activation. KO of RSU1 abolished cell–ECM detachment–induced downregulation of ERK activation. Thus, cells lacking RSU1 exhibited a constitutively high level of ERK activation despite the absence of cell–ECM adhesion (Fig. 1). Additionally, loss of RSU1 results in increases of ERK activation in cells cultured in 3D (Fig. 8). Together, these findings suggest a suppressive role of RSU1 in ERK activation, which provides an explanation as to why loss or reduced expression or mutations of RSU1 was often found in human cancers (e.g., hepatocellular carcinoma and gliomas) (37, 38, 39) and why overexpression of RSU1 can inhibit anchorage-independent growth of human cancer cells (39, 40).</p><!><p>A model of RSU1-mediated regulation of ERK signaling in response to changes of cell–ECM adhesion. The figure depicts a model in which RSU1 regulates ERK signaling in response to changes of cell–ECM adhesion. In cells with abundant cell–ECM adhesions, RSU1 is concentrated in cell–ECM adhesions where it interacts with the ILK–PINCH–parvin complex and facilitates cell spreading, directional migration, and invasion (A). Loss or reduction of cell–ECM adhesion releases RSU1 from cell–ECM adhesions, resulting in increased RSU1 association with membrane lipid rafts and interaction with PHB2 and consequently suppression of ERK activation (B). This negative regulatory mechanism is compromised in cells lacking RSU1 (C) or overexpressing dominant negative PHB2 fragment (DN) that inhibits the RSU1–PHB2 interaction (D), resulting in aberrant activation of ERK despite the absence or reduction of cell–ECM adhesion. ECM, extracellular matrix; ERK, extracellular signal–regulated kinase; PINCH, particularly interesting new cysteine-histidine-rich protein.</p><!><p>Our studies have demonstrated that the interaction of RSU1 with PHB2 is influenced to a great extent by cell–ECM adhesion (Fig. 5). How does the interaction of RSU1 with PHB2 sense the status of cell–ECM adhesion? Previous studies have shown that cell–ECM adhesion exerts a strong effect on membrane order through a process that depends on both membrane lipid rafts and clustering of focal adhesion proteins (62). PHB2 is a component of membrane lipid rafts (49, 50). Furthermore, it has been shown that membrane lipid rafts are involved in cell–ECM detachment–induced downregulation of ERK activation (47, 63). In the studies presented here, we showed that a fraction of RSU1 was associated with membrane lipid rafts. Importantly, although the amount of PHB2 in membrane lipid rafts was not significantly altered in response to cell–ECM detachment, the amount of RSU1 associated with membrane lipid rafts was increased in this process (Fig. 4). Concomitantly, the interaction of RSU1 with PHB2 was enhanced (Fig. 5). Thus, it is attractive to propose that the interaction of RSU1 with PHB2 senses the status of cell–ECM adhesion through, at least in part, alteration of membrane lipid rafts. Clearly, future studies are required to further test this possibility.</p><p>Given the central role of aberrant ERK signaling in the development and progression of cancer (9, 59, 60), our findings that RSU1, whose expression is frequently lost or reduced in human cancers (37, 38, 39), binds PHB2 and consequently mediates cell–ECM detachment–induced suppression of ERK signaling have important implications for development of therapeutic agents that control cancer progression. Indeed, recent studies have shown that rocaglamides, a class of natural anticancer compounds, directly bind PHB1 and PHB2 and inhibit ERK activation (54). Thus, the RSU1-PHB signaling axis identified in the current study may serve as an effective target for therapeutic control of aberrant ERK signaling, anchorage-independent growth, and cancer progression.</p><!><p>Human HT1080 fibrosarcoma cells were from ATCC and cultured in cell culture dishes with minimum essential medium supplemented with 10% fetal bovine serum (FBS), 2 mM L-glutamine, 0.1 mM nonessential amino acids, 1.0 mM sodium pyruvate, and penicillin and streptomycin with 50 U/ml each at 37 °C in 5% CO2 and 3% O2. Transfection was performed using LipofectAMINE 3000 (Invitrogen) according to the manufacturer's instructions. For cells cultured in adhesion and suspension, respectively, trypsinized cells were either allowed to adhere to fibronectin (10 μg/ml) or maintained in suspension in hydroxyethyl methacrylate-coated (12 mg/ml, Sigma) non-adhesive cell culture dishes for 5 h. For 3D cell culture, type I collagen (corning 354249) was diluted with sterile deionized water to a final concentration of 3.0 mg/ml. Cells (as specified in each experiment) were harvested with trypsin, washed, and mixed with collagen I (the final cell density = 5.0 × 105 cells/ml and the final concentration of collagen I = 1.0 mg/ml), and pH was adjusted to 7.4 with 1 M NaOH. The cells were cultured in 3D collagen I gels for 24 h and then transferred into tubes and treated with 1 mg/ml collagenase D (Roche, 37334226) for 0.5 h at 37 °C. The cells were collected by centrifugation at 300g for 3 min and lysed with the radio-immunoprecipitation assay buffer for further analyses.</p><!><p>Rabbit polyclonal anti-RSU1 antibody used for immunoprecipitation was from BETHYL (Montgomery, AL). Mouse monoclonal anti-PINCH-1 antibody was from BD. Mouse monoclonal anti-GAPDH antibody was from Abmart (Berkeley, NJ). Rabbit monoclonal anti-MEK, anti-phosphorylated MEKSer 221, anti-ERK, anti-phosphorylated ERKThr202/Tyr204, anti-caveolin-1, and anti-PHB2 antibodies used for Western blotting were purchased from Cell Signaling. Alexa fluor647-conjugated goat anti-mouse IgG antibody was from ThemoFisher Scientific. Horseradish peroxidase–conjugated secondary antibodies were from Jackson Immuno Research Laboratories (West Grove, PA). Alexa555-CTxB used to label lipid rafts was purchased from Invitrogen (Carlsbad, CA).</p><p>Small interfering RNA (siRNA) specifically targeting RSU1 and PHB2, respectively, and their corresponding scramble control siRNA were purchased from Genepharma (Shanghai). The sequences of synthetic siRNAs directed against PHB2 are siPHB2-1: 5′-gccucaucaaggguaagaatt-3′ and siPHB2-2: 5′-gugauuuccuacaguguuguucccu-3′, respectively. siRNA was transfected using RNAiMax (Invitrogen) following the manufacturer's protocol. Experiments were carried out 48 h after the transfection.</p><p>Restriction endonucleases were obtained from New England Biolabs, Inc (Beverly, MA). Cell culture media and corresponding items were from Sigma-Aldrich (St Louis, MO) or Invitrogen (Carlsbad, CA). All other chemicals were from Fisher Scientific (Fairlawn, NJ) or Sigma-Aldrich (St Louis, MO).</p><!><p>HT1080 cells in which RSU1 was knocked out were generated by CRISPR/cas9-mediated gene disruption. Briefly, a guide RNA (gRNA) oligo designed to target the sequence of 5′-ccggggcatctccaacatgctgg-3′ located at the exon 1 of RSU1 was cloned into pSpCas9 (BB) -2a-GFP (PX458 containing cas9, was a gift from Dr Feng Zhang, Addgene plasmid # 48138) via BbsI sites and transfected into HT1080 cells. Single GFP-positive cells were sorted into each wells of 96-well plates by FACS sorter (BD FACS Aria III). Once the single colonies are propagated, PCR-based analyses as well as Western blotting were employed to assess targeted gene disruption of RSU1. Genomic DNA of individual colonies was extracted and amplified with a pair of DNA oligos flanking the gRNA-targeting site, i.e., P1: 5′-ccaaccctggggaagcctcaga-3′ and P2: 5′-tactgcaaaccctctgcgcg-3′ for RSU1. The PCR products were subcloned into pTA vectors. For each single colony, 6 to 8 clones were selected for DNA sequencing, and Western blotting using anti-RSU1 antibody was used to further verify the RSU1 KO cell line.</p><!><p>The DNA sequence encoding Clover was inserted into exon 8 of RSU1 in HT1080 cells using CRISPR/cas9-mediated targeted gene editing and homology-mediated recombination. Briefly, a guide RNA (gRNA) oligo (5′-caccgaaagatcagccggaaacccc-3′) designed to target the sequence (5′-aaagatcagccggaaacccctgg-3′) located at the exon 8 of RSU1 was cloned into pSpCas9 (BB) -2a-GFP (PX458 containing cas9, Addgene plasmid # 48138) via BbsI sites. The engineered pUC19 vector carried Clover sequence–flanked respectively, by the 600 to 700 bp of DNA sequences immediately upstream and downstream of the CRISPR/cas9 targeting site was served as donor for homology-mediated recombination. The guide RNA targeting site (i.e., 5′-aaagatcagccggaaacccctgg-3′) was also inserted to the 5′ and 3′ of the homolog sequences, respectively, to allow the linearization of the vector by Cas9 and the same guide RNA. Single GFP-positive cells were sorted into each well of 96-well plates by FACS sorter (BD FACS Aria III). Once the single colonies are propagated, PCR-based analyses as well as Western blotting were employed to assess targeted gene knock-in.</p><!><p>Mouse monoclonal antibodies recognizing RSU1 were prepared using GST-fusion proteins containing RSU1 residues 1 to 299 (full-length) as an antigen based on a previously described method (24, 65). Hybridoma supernatants were initially screened for anti-RSU1 activities by ELISA using MBP-RSU1 protein. Positive clones were selected and further tested by Western blotting using GFP- and FLAG-tagged RSU1.</p><!><p>The ORFs of RSU1 and PHB2, respectively, were amplified by PCR and subloned into mClover-N1 (Addgene, no. 54538) and mRuby2-N1 (Addgene, no. 54614). DNA fragments encoding PHB2 deletion mutants were generated by PCR and were cloned into pEGFP-C1. All DNA constructs were confirmed by DNA sequencing.</p><!><p>A full-length RSU1 cDNA was cloned from a human Kidney cDNA library (Clontech) by PCR using the following primers: 5ʹ-gcgaattcatgtccaagtctctgaagaagttggtg-3ʹ and 5ʹ-cggtcgacttatctgttcttggctgccaggggtttcc-3ʹ. The sequence of the RSU1 cDNA was confirmed by automated DNA sequencing. The RSU1 cDNA was inserted into the pGBKT7 vector (Clontech). The pGBKT7/RSU1 construct was used as bait to screen a human keratinocyte MATCHMAKER cDNA library following a previously described protocol (24, 64, 65). Six positive plasmids containing cDNA inserts with an identical size (1.1 kb) were selected and sequenced. They contain an identical cDNA fragment encoding the C-terminal fragment of PHB2 (residues 101–299). The full-length PHB2 cDNA was isolated from the human lung cDNA library by PCR using the following primers: 5ʹ-gctgaattcatggcccagaacttgaaggacttgg-3ʹ and 5ʹ-gagctcgagtcatttcttacccttgatgaggctg-3ʹ.</p><!><p>Lipid rafts were isolated from HT1080 fibrosarcoma cells using MinuteTM Lipid Raft Isolation Kit for Mammalian Cells/Tissues (LR-039, Invent Biotechnologies, MN) following the manufacturer's instructions. Briefly, 30 × 106 cells were collected and spun down at 500g for 5 min. The cell pellet was resuspended in 500 μl of buffer A and vortexed for 20 s. The cell suspension was centrifuged in the filter cartridge at 16,000g for 30 s. The pellet was then resuspended and centrifuged at 1000g for 5 min. The pellet containing nuclei, large cell debris, and some unruptured cells was discarded, and the supernatant was centrifuged further at 16,000g for 30 min. The resulting supernatant was saved as the cytosolic fraction, and the total membrane fraction in the pellet was resuspended in buffer B and incubated on ice for 30 min before centrifugation at 16,000g for 10 min. The supernatant was collected and mixed with buffer C to be subjected to centrifugation at 10,000g for 10 min, resulting in lipid rafts floating on the top of the tube. The aqueous phase was removed, and the lipid rafts was resuspended in buffer A and was centrifuged at 16,000g for 5 min. Finally, the lipid rafts in the pellet were resuspended with the radio-immunoprecipitation assay buffer (10 mM Tris-HCl pH 8.0, 0.5 mM EGTA, 1 mM EDTA pH 8.0, 1% TritonX-100, 0.1% sodium deoxycholate, 0.1% SDS, 140 mM NaCl).</p><!><p>Lipid rafts were labeled by Vybrant Alexa Fluor 555 Lipid Raft Labeling Kits (Invitrogen, CA) following the manufacturer's instructions. In brief, cells were incubated with fluorescent CTxB conjugate working solution after washing with complete growth medium at 4 °C. After the incubation, cells were spun down and then were resuspended with prechilled anti-CTxB antibody working solution at 4 °C for 15 min. Cells were fixed with ice-cold PBS containing 4% formaldehyde, and lipid rafts were visualized by fluorescence microscopy.</p><!><p>The ORF of RSU1 and PHB2 was inserted into the pGEX-5x-1 vector (Pharmacia), respectively. The full-length or deletion mutant forms of PHB2 were inserted into the pMAL-C2 vector (New England BioLabs), respectively. The recombinant vectors were used to transform E. coli cells. The expression of the GST- or MBP-fusion proteins was induced with IPTG. GST- and MBP-tagged fusion proteins were isolated using glutathione-Sepharose 4B and amylose-agarose beads, respectively, as we previously described (24, 65).</p><!><p>The association of PHB2 with endogenous RSU1 derived from mammalian cells was analyzed using a GST–PHB2 fusion protein in the pull-down assay as we previously described (65). Briefly, HeLa cells from three100-mm plates were lysed with 6 ml of lysis buffer (1% TX-100 in 50 mM Hepes (pH 7.1) containing 150 mM NaCl, 10 mM Na4P2O7, 2 mM Na3VO4, 100 mM NaF, 10 mM EDTA, and protease inhibitors (final concentrations of 1 μg/ml aprotinin, 10 μM leupeptin, 1 μM pepstatin, and 1 mM PMSF). The lysates were precleared with glutathione-Sepharose 4B beads. The precleared lysates (1.4 mg total amount of protein) were mixed with glutathione-Sepharose 4B beads containing GST-PHB2 or GST (25 μl) and incubated at 4 °C overnight. After washing, proteins coprecipitated with GST-PHB2 were detected by Western blotting with anti-RSU1 antibody.</p><p>To analyze direct interaction between RSU1 and various mutant forms of PHB2, GST-RSU1 (10 μg/ml) were incubated with 25 μl amylose-Sepharose 4B beads containing MBP-tagged full-length or deletion mutant forms of PHB2, or MBP as a negative control in binding buffer (1% Triton X-100, 10 mM Tris, and 100 mM NaH2PO4, pH 8.0). The mixtures were incubated at 4 °C for 3 h. At the end of incubation, the beads were washed six times with the binding buffer. GST-RSU1 bound to MBP-tagged full-length or deletion mutant forms of PHB2 was detected by Western blotting with an anti-RSU1 antibody.</p><!><p>Western blotting was performed as previously described (64). In brief, whole-cell proteins were extracted using the radio-immunoprecipitation assay buffer (10 mM Tris-HCl pH 8.0, 0.5 mM EGTA, 1 mM EDTA, 1% Triton X-100, 0.1% sodium deoxycholate, 0.1% SDS, 140 mM NaCl). The concentration of total proteins was determined using BCA protein assay kit (Pierce). Ten to 30 μg of proteins was loaded per lane. The proteins were separated by SDS-PAGE and transferred to a polyvinylidene fluoride membrane (Millipore). The membranes were blocked with 5% BSA in Tris-buffered saline (50 mM Tris-HCl and 150 mM NaCl, pH 7.4) containing 0.1% Tween 20 for 1 h at room temperature, followed by overnight incubation at 4 °C with a specific primary antibody. After washing and incubating with appropriate horseradish peroxidase–conjugated secondary antibodies (anti-mouse, 711-005-152, or anti-rabbit, 715-005-151; Jackson ImmunoResearch), blots were developed using an ECL kit (Bio-Rad) and then exposed to an x-ray film (super RX-N-C, Fuji Film).</p><!><p>Cells were lysed with the lysis buffer containing 10 mM Tris/HCl, pH 7.6, 500 mM NaCl, 1.75% n-octyl-β-D-glucopyranoside, and protease inhibitors cocktails (Roche) as specified. The cell lysates were mixed with agarose beads conjugated with anti-RSU1 (4 μl) or anti-PHB2 antibody (5 μl). The beads were washed four times, and the immunoprecipitates were analyzed by Western blotting with antibodies as specified.</p><!><p>Human HT1080 cells were fixed with 4% paraformaldehyde in PBS, permeabilized with 0.1% Triton X-100 in PBS, and stained with the primary rabbit monoclonal anti-PHB antibody and Alexa fluor647–conjugated anti-mouse IgG antibody. The cells were observed under a Nikon A1R confocal microscope equipped with a 100× oil objective.</p><!><p>Time-correlated single-photon fluorescence lifetime microscopy (TCSPC-FLIM) experiments were performed as follows. HT1080 cells were transiently transfected with mClover-N1-RSU1 and mRuby-N1-PHB2, respectively, and incubated in minimum essential medium in the absence of FBS for 24 h. The cells were fixed with 4% polyformaldehyde, and fluorescence images were recorded with Nikon A1R confocal microscope equipped with a 485-nm pulsed diode laser (PDL 800-D, PicoQuant), a photodetector (PMA Hybrid 40, PicoQuant), and a 100× oil objective. The detection covered a time window of 44 ns after the excitation pulse with the TCSPC resolution of 25.0 ps.</p><p>For each group of cells, 33 to 46 cells were analyzed for each experiment. Data analysis was performed by the software "SymPhoTime 64" (PicoQuant). In brief, "n-Exponential Tailfit" was selected as the fitting model, the signals on focal adhesions was selected as ROIs, and "calculated IRF" was selected as initial fit to get the lifetime value (τamp). FRET efficiency (E) was calculated according to the equation E = 1 − τDA/τD, where τD is the fluorescence lifetime of mClover-N1-RSU1 in the presence of mRuby2-N1-PHB2 and τDA is the fluorescence lifetime of mClover-N1-RSU1 in the presence of mRuby2-N1.</p><!><p>The chemiluminescent blots were imaged with the Tanon 6100B imager (Tanon, Shanghai, China). ImageJ software (NIH) was used to select and determine the density of the bands showing total levels of MEK and ERK and phosphorylated MEK (Ser 221) and ERK (Thr202/Tyr204) in all the blots. The densiometric ratio of phosphorylated MEK to total MEK and that of phosphorylated ERK to total ERK were calculated and normalized to the control in each experiment. Statistical analysis of the quantification data was expressed as means ± SEM. Differences between the groups were examined for statistical significance using one- or two-tailed paired t test (GraphPad Prism, version 5.00, for Windows, GraphPad Software, La Jolla, CA). A value of p < 0.05 was considered to be statistically significant.</p><!><p>To monitor cell spreading, 2 × 104 cells were seeded in 6-well plates that were precoated with fibronectin (30 μg/ml) and placed into IncuCyte ZOOM apparatus (Essen BioScience, MI). Images of the collective cell spreading at 16 different locations were captured every 15 min for a total duration of 8–10 h using the IncuCyte ZOOM live-cell imaging system (Essen BioScience, MI). After 2 h of plating, the cell surface area does not show significant variation over time. Therefore, the mean value of near-maximal cell surface area was calculated by averaging the values that show no significant variation over the time (up to 5 h after plating).</p><!><p>Cell spreading was quantified by measuring the cell area (μm2) as described previously (66). A two-tailed paired Student's t-test was performed to test for significance in all experiments.</p><!><p>To analyze cell migration, cells were seeded in a culture dish coated with fibronectin (30 μg/ml) and placed in a heated and air-humidified chamber built in a Nikon inverted microscope TE2000E. Phase-contrast time-lapse imaging of a field containing 7 to 10 cells at a 2-min interval for 3 h was captured on the microscpe with a 10× Ph1 objective, perfect focus system (PFS), and Hamamatsu C-11440-22CU camera controlled by NIS-elements software (Nikon). Image stacks were processed using ImageJ software. Cell motility was tracked by using ImageJ (NIH) with plug-in "manual Tracking" (Fabrice Cordelières, Institut Curie, Orsay, France). Data were analyzed by the Chemotaxis-and-Migration-Tool) (ibidi, http://ibidi.com/xtproducts/en/Software-and-Image-Analysis/Manual-Image-Analysis/Chemotaxis-and-Migration-Tool).</p><!><p>Cell invasion assay was performed in Corning biocoat matrigel invasion chambers (Corning, NY) with uncoated porous inserts (pore size: 8 μm) according to the manufacturer's protocol. Briefly, cells were plated at a density of 3 × 104 in each well with minimum essential medium free of FBS, and 700 μl of culture medium containing 10% FBS was added to the bottom of the 24-well plate. Following incubation for 24 h, noninvading cells were removed from the upper surface of the membrane using a cotton-tipped swab. The invading cells were subsequently fixed in 4% formaldehyde for 10 min and stained with Hochest 33342 for 5 min. The stained cells were counted as cells per field using Nikon ECLIPSE Ti at 10× magnification in 5 fields.</p><!><p>All data discussed are contained within the manuscript or the supporting material.</p><!><p>The authors declare that they have no conflicts of interest with the contents of this article.</p><!><p>Figures S1 to S3</p>

PubMed Open Access

Quantification of Sulforaphane Mercapturic Acid Pathway Conjugates in Human Urine by High-Performance Liquid Chromatography and Isotope-Dilution Tandem Mass Spectrometry

We report validation of the first high-pressure liquid chromatography isotope-dilution mass spectrometry method to measure sulforaphane (SFN) and its glutathione-derived conjugates in human urine. As epidemiological evidence continues to mount that the consumption of a diet rich in cruciferous vegetables may reduce the risk of certain cancers, the development of analytical methodologies to accurately measure isothiocyanates (ITCs) and their subsequent metabolic products becomes paramount. SFN, the principal ITC produced by broccoli, is an effective chemopreventive agent with multiple modes of action. SFN and SFN conjugates have often been measured collectively utilizing a cyclocondensation assay with 1,2-benzenedithiol. More recently, some of the major SFN conjugates have been determined using mass spectrometry. Here, triple-quadrupole mass spectrometry has been coupled with the use of stable isotope-labeled internal standards of D8-SFN and all four D8-SFN mercapturic acid pathway conjugates to provide an accurate, precise, sensitive, and specific method for analysis of these compounds. Using urine samples collected during an earlier intervention with broccoli sprouts, the concentrations of SFN, SFN-cysteine, and the mercapturic acid SFN-N-acetylcysteine were sufficiently high such that as little as 50 nL of urine was required for analysis. Although each study participant received an equivalent dose of broccoli sprout preparation, the interindividual conversion of the precursor glucosinolate to SFN varied over 100-fold. These 98 urines provided an ideal sample set for examining the robustness of the assay. The mean urinary concentrations \xc2\xb1 standard deviations in overnight voids following ingestion of the first dose were 4.7 \xc2\xb1 5.1, 0.03 \xc2\xb1 0.05, 0.06 \xc2\xb1 0.06, 18 \xc2\xb1 15, and 42 \xc2\xb1 23 nmol/mg creatinine for SFN, SFN-glutathione, SFN-cysteine-glycine, SFN-cysteine, and SFN-N-acetylcysteine, respectively. This method determines SFN and all four SFN glutathione-derived metabolites with minimal sample preparation and will be extremely useful in understanding the role of SFN-rich foods in preventing cancer and other chronic diseases.

quantification_of_sulforaphane_mercapturic_acid_pathway_conjugates_in_human_urine_by_high-performanc

Introduction<!>Study Population<!>Chemicals<!>Preparation of Standards and Internal Standards<!>Isolation of SFN and Its GSH-Derived Conjugates from Urine<!>HPLC-MS/MS Analysis of SFN and Its GSH-Derived Conjugates<!>Results<!>Discussion

<p>Epidemiological studies have reported that the consumption of a diet rich in cruciferous vegetables reduces the risk of cancer (1) as well as many chronic degenerative diseases (2). Cruciferous vegetables such as broccoli, cauliflower, cabbage, watercress, and Brussels sprouts contain a number of biologically active constituents including glucosinolates that can be degraded to isothiocyanates (ITCs)1 and indoles (3). Glucosinolates are hydrolyzed to ITCs through the action of the plant enzyme myrosinase or to a lesser degree by microbes in the human gut (4). Several ITCs, including sulforaphane (SFN; 1-isothiocy-anato-4-methylsulfinylbutane), the principal isothiocyanate contained in broccoli (5), are effective chemopreventive agents in animals (6, 7). Efficacy is attributed to the activation of cytoprotective enzymes through Nrf2 signaling, induction of apoptosis, and inhibition of cytochrome P450s and histone deacetylases (8–10). Although only a few human clinical trials have begun to examine the role of SFN as a preventive agent, an inverse association was found for the excretion of SFN metabolites and both aflatoxin-N7-guanine DNA adducts and trans,anti-phenanthrene tetraol, a metabolite of the combustion product phenanthrene (11).</p><p>Studies in humans and experimental animals have shown that ITCs are metabolized primarily to mercapturic acids (12, 13). ITCs are initially conjugated to glutathione (GSH) by glutathione transferases and metabolized sequentially to form cysteinyl-glycine (CysGly), cysteine (Cys), and, finally, N-acetylcysteine conjugates (NAC). ITCs and conjugates can be collectively measured by a cyclocondensation reaction of deconjugated ITCs with 1,2-benzenedithiol yielding 1,3-benzenedithiol-2-thione, which is readily quantifiable by high-pressure liquid chromatography (HPLC) coupled to UV detection (14). This method has proven to be a valuable tool for examining the total amount of ITCs in plant extracts (14) or the collective levels of ITCs and metabolites in plasma, urine, and tissue (15, 16); however, it does not allow a distinction to be made between different ITCs or ITC conjugates. In addition, analyte specificity of the cyclocondensation reaction can be problematic because 1,2-benzenedithiol also reacts with dithiocarbamates, carbon disulfide, substituted thioureas, disulfiram, and chemicals used in the rubber industry (14). More recent methods for measuring SFN and SFN metabolites directly have relied on HPLC coupled with UV (17) or mass spectrometric (MS) detection (12, 18–20). Although these reports have provided a platform for further examination, they may require cumbersome sample preparation (12), may only measure SFN-NAC (18), use other biological matrices for validation (19), or use N-acetyl-S-(N-butylthiocar-bamoyl)-L-Cys as an internal standard (20). Here, we report a sensitive and specific HPLC/isotope dilution mass spectrometric method to quantify SFN and the four glutathione-derived conjugates in human urine. Stable isotope-labeled internal standards for each of the four conjugates were synthesized from D8-SFN, purified by HPLC, and characterized by MS. Utilizing human urine samples collected in a previously reported broccoli sprout intervention in China (11), we demonstrate the specificity, sensitivity, and versatility of this method. As this method quantitatively measures the individual SFN -conjugates and SFN, it will be extremely useful in clinical trials seeking to elucidate the role of SFN-rich foods in the prevention of disease as well as for examining determinants of interindividual variation in SFN pharmacokinetics.</p><!><p>A clinical trial evaluating the chemopreventive activity of broccoli sprouts was conducted in 2003 as a collaborative study between the Johns Hopkins Bloomberg School of Public Health and the Qidong Liver Cancer Institute. Study design and consent forms were approved by the Institute Review Boards monitoring human studies at both institutions: Details of the study are reported in Kensler et al. (11). Briefly, individuals participating in a randomized clinical trial drank a hot water infusion prepared from 3 day old broccoli sprouts. This preparation contained 400 μmol of glucoraphanin, the primary glucosinolate found in broccoli sprouts and the precursor for SFN. Because the clinical trial utilized a hot water infusion that inactivated myrosinase found in the plant, the bioavailability of SFN was solely dependent on the hydrolysis of glucoraphanin by microbes present in the gut (16, 21). Overnight urine voids collected approximately 12 h after drinking the first broccoli sprout infusion were used to determine the levels of SFN and its conjugates by the isotope dilution tandem MS methodology described herein. In the earlier report (11), total ITC concentrations (as assessed using the cyclocondensation assay) were determined from urines collected at later time points in the study. For our current evaluation, we chose the first overnight urine collection for analysis because these samples had not been previously thawed and refrozen.</p><!><p>SFN and custom-synthesized D8-SFN [1-isothio-cyanato-4-methylsulfinyl(1,1,2,2,3,3,4,4-2H8)butane] were purchased from LKT Laboratories (St. Paul, MN). Glutathione (GSH), cysteineglycine (CysGly), Cys, and NAC were purchased from Sigma Chemical Co. (St. Louis, MO). All other chemicals and solvents were of analytical/reagent grade or higher. Creatinine levels were determined by Hagerstown Medical Laboratory, Inc. (Hagerstown, MD) using a kinetic alkaline picrate (Jaffee reaction) method measured on a Dade-Behring Dimension analyzer.</p><!><p>SFN-GSH, SFN-CysGly, SFN-Cys, and SFN-NAC were synthesized independently using both SFN and D8-SFN according to the method of Kassahun et al. (12) with a slight modification. At the completion of the 3 h reaction, the aqueous layer was acidified with 1 M HCl to pH 3, and a large excess of ice-cold acetonitrile was added, thereby precipitating the SFN conjugates. Following centrifugation at 13000g for 5 min, the solvent layer containing unconjugated SFN was removed. The aqueous fraction containing the SFN conjugate was subsequently purified by reverse-phase HPLC using a Luna C18 (2) 4.6 mm × 250 mm column (Phenomenex, Torrance, CA) and a Waters 996 photodiode array detector monitoring 200–300 nm. Chromatographic separation of the SFN conjugates and SFN was achieved using a linear gradient of 1% acetic acid/water and 1% acetic acid/acetonitrile over 40 min. Purity and structural confirmation of the conjugates were confirmed by HPLC and direct electro-spray MS prior to combining aliquots. Collected HPLC fractions were dried using a Savant Speed Vac (Savant Instruments, Inc., Farmington NJ) over several days until a constant weight was obtained. Each purified, dried conjugate was dissolved in 5% acetonitrile/1% acetic acid/water (initial MS mobile phase) and scanned using a Beckman DU 800 spectrophotometer (Beckman Coulter, Inc., Somerset, NJ). Millimolar absorption coefficients (mM−1 cm−1) for SFN-GSH, SFN-CysGly, SFN-Cys, and SFN-NAC were determined to be 7.46, 8.2, 7.02, and 5.9, respectively. An absorption coefficient of 8.0 mM−1 cm−1 was previously reported for SFN-GSH (13). The molar extinction of 0.90 mM−1 cm−1 for SFN and D8-SFN was obtained from the Merck Index (22). Solutions were stored at −80 °C at pH 3 and protected from direct light.</p><!><p>Urine samples were thawed and centrifuged, and aliquots were removed for SFN conjugate and creatinine analyses. Urine samples designated for MS analysis were immediately acidified to pH ~ 3 in either the initial HPLC/MS mobile phase [water/acetonitrile/acetic acid (95/5/0.1, v/v/v)] or in 10 mM ammonium acetate buffer to prevent degradation.</p><p>For the determination of SFN-Cys, SFN-NAC, and SFN, 10 μL of urine was diluted 1–50 in the initial HPLC/MS mobile phase (Scheme 1a). Ten microliters of this solution was then transferred to a silanized glass injection insert, such that an equivalent of 0.2 μL of urine was used for each analysis. This sample was then spiked with 50 μL of initial HPLC/MS mobile phase containing 1 ng each of the three stable isotope-labeled internal standards: D8-SFN, D8-SFN-Cys, and D8-SFN-NAC. This solution was then brought to 100 μL by the addition of 40 μL of initial HPLC/MS mobile phase. For two urine samples with SFN-Cys concentrations below 15 pg/0.2 μL urine, the analysis was repeated with an equivalent of 1 μL of urine. For 14 urine samples with SFN-NAC concentrations greater than 4000 pg/0.2 μL, the analysis was repeated with an equivalent of 0.05 μL of urine.</p><p>For the determination of SFN-GSH and SFN-CysGly (Scheme 1b), which were present in concentrations more than 100-fold lower than SFN-NAC, 50 μL of urine was diluted with 950 μL of 10 mM ammonium acetate buffer (pH 3). This sample was then spiked with 50 μL of initial HPLC/MS mobile phase containing 1 ng each of the two stable isotope-labeled internal standards: D8-SFN-GSH and D8-SFN-CysGly. This solution was applied to a 3 mg Varian Bond-Elut LRC C-18 SPE column (Varian, Inc., Walnut Creek, CA) that had been preconditioned with 3 mL of methanol followed by 3 mL of water. SFN-GSH and SFN-CysGly were eluted with 4 mL of water/methanol/ acetic acid (50/50/0.1, v/v/v). Eluants were taken nearly to dryness in amber vials under nitrogen and redissolved in 100 μL of initial HPLC/MS mobile phase prior to injection of 10 μL.</p><!><p>Analysis of the SFN and SFN conjugates was carried out on a Thermo-Finnigan TSQ Quantum Ultra triple quadrupole mass spectrometer coupled to a Thermo-Finnigan Surveyor Plus HPLC and autoinjector (ThermoElectron Corporation, San Jose, CA). Samples were maintained at 10 °C before injection of 10 μL aliquots. Chromatographic separation was carried out at 35 °C at a flow rate of 100 μL/min on a 1 mm × 150 mm × 5 μm Luna C18(2) microbore HPLC column (Phenomenex, Torrance, CA). Mobile phase compositions were as follows (v/v/v): (A) water/acetonitrile/acetic acid (95/5/0.1) and (B) acetonitrile/acetic acid (100/0.1). The initial composition was held at 100% A for 3 min and then ramped to 60% B at 14 min. The column was then re-equilibrated for 9 min at initial conditions. The column flow was diverted away from the electrospray ionization (ESI) ion source except for the time period from 3 to 15 min.</p><p>Positive ESI-MS/MS was conducted with the capillary temperature set at 220 °C, the sheath gas at 49 arbitrary units, and the spray voltage at 5 kV. The same HPLC conditions were used for analysis of the two groups of metabolites. MS/MS transitions and collision energies for analysis of higher concentration analytes (SFN-Cys, SFN-NAC, and SFN) are given in Scheme 1a. In all transitions, a resolution of ±0.01 amu was chosen. The corresponding selected reaction monitoring chromatogram for these analytes in a typical urine sample is given in Figure 1a. MS/MS transitions and collision energies for analysis of lower concentration analytes (SFN-CysGly and SFN-GSH) are given in Scheme 1b. All transitions were resolved at ±0.01 amu. For SFN-GSH and D8-SFN-GSH, the sum of the two transitions monitored was used for quantitation. Identification of SFN-GSH was dependent on the presence of the two peaks monitored with the expected area ratio. The corresponding selected reaction monitoring chromatogram for these analytes for the same urine sample is shown in Figure 1b.</p><!><p>Sensitivity and range of analysis for the HPLC-MS/MS analysis of SFN and four SFN conjugates in urine are given in Table 1. Limits of detection ranged from 1 to 6 pg, depending upon the analyte. Linear 11-point isotopic dilution standard curves for each analyte (r2 > 0.98) were generated by duplicate injection of 10 μL mixtures of D8-SFN-GSH, D8-SFN-CysGly, D8-SFN-Cys, D8-SFN-NAC, and D8-SFN (each at 1000 pg/ 100 μL) with varying concentrations of each of the five nonlabeled standards (4000, 2000, 1000, 500, 250, 125, 63, 31, 16, 8, and 0 pg/100 μL). Peak areas were fitted using the method of least-squares with a 1/X weighting factor. Limits of determination and range of analysis for the method in urine (Table 1) depended on the analyte and the amount of urine used for analysis.</p><p>Validation of the method for the analysis of SFN and SFN conjugates is shown in Tables 2 and 3. Accuracy was determined by using urine from a study participant who had received the placebo beverage in the clinical trial. The absence of any endogenous SFN or SFN metabolites in the urine sample was confirmed prior to use. Accuracy was evaluated by spiking the blank urine with the five SFN analytes at known levels and comparing the concentrations measured in the samples. All four concentration levels were examined in replicate (n = 5). The accuracy ranged from 70 to 101% and was worst for lower concentrations of SFN-GSH (Table 2).</p><p>Intraday reproducibility was evaluated in urine samples from three study participants who excreted high, medium, or low amounts of SFN and SFN conjugates. Concentrations were determined in five independent analyses, and the % coefficient of variation was less than 12% for all compounds at the three concentrations tested (Table 3).</p><p>The LC-MS/MS method described above was used to analyze 98 urines collected over a 12 h period after drinking a broccoli sprout infusion containing 400 μmol of the SFN precursor glucoraphanin. Selected reaction monitoring chromatograms for a typical urine sample are given in Figure 1a,b. The urinary concentrations in these samples of SFN and SFN conjugates corrected for creatinine excretion are summarized in Table 4. Sensitivity was very dependent on the particular SFN conjugate quantified. Levels of SFN, SFN-NAC, and SFN-Cys were orders of magnitude higher in urine than levels of SFN-GSH and SFN-CysGly, such that different size aliquots of each urine sample were analyzed separately. The wide ranges reported for each analyte reflect interindividual differences in SFN bioavailability and/or factors affecting biotransformation and excretion.</p><!><p>Evidence promoting the health benefits of vegetable consumption continues to mount with an emphasis on the role of crucifers. This family of vegetables is of particular interest because it is exceptionally rich in glucosinolates, precursors to ITCs. Numerous epidemiological studies have shown that dietary intake of ITCs is inversely associated with cancer risk (1, 2). Therefore, analytical methodologies examining specific ITCs and their metabolic products become increasingly important in understanding the multiple anticarcinogenic mechanisms through which cruciferous vegetables may act (10, 24). For example, SFN-Cys is thought to be the most potent inhibitor of histone deacetylase among SFN and its metabolites (9). Although urinary levels of total ITC equivalents may be an excellent biomarker of exposure to ITCs in general, cancer preventative potency varies widely for individual ITCs (7, 10, 24, 25). As a consequence, we focused on developing a method to measure SFN and SFN conjugates because SFN is the most potent anticarcinogenic ITC reported and is easily consumed in cruciferous vegetables such as broccoli.</p><p>In this report, we present an isotope dilution MS/MS method utilizing D8-SFN and D8-SFN conjugates as internal standards to quantify the GSH conjugation products of SFN in urine. Urines previously collected in a broccoli sprout intervention study (11) were used to demonstrate the validity, accuracy, and sensitivity of the method. In the broccoli sprouts intervention, urines of study participants collected after 5, 9, 10, and 12 days of intervention were previously analyzed by the cyclocondensation assay for the total ITC equivalents excreted (11). Although ITC excretion rates were observed to vary widely between individuals, excretion rates were relatively constant over time for each individual. As the urine samples examined in the previous study had been frozen and rethawed several times, we chose to analyze a set of heretofore unthawed urines that were collected for roughly 12 h overnight after participants drank the first dose of broccoli sprout infusion. In our analyses, we attempted to diminish any degradation of SFN conjugates by quick-thawing these urines and keeping them on ice, maintaining all sample processing in an acidic environment and minimizing light exposure throughout the analyses.</p><p>As shown in Table 2, the SFN mercapturic acid pathway conjugates isotope dilution MS/MS assay is very sensitive and has a wide dynamic range for detection of SFN and its metabolites. For SFN-GSH and SFN-CysGly, 50 μL of urine was used for analysis. However, as SFN-Cys, SFN-NAC, and SFN urine concentrations were much higher, only 0.2 μL and in a few cases 0.05 μL of urine was required for determination of these three compounds, and solid-phase column cleanup was not required (Scheme 1a,b). In particular, the SFN-Cys and SFN-NAC concentrations in urine are spread over a large dynamic range, recapitulating the high interindividual variability in excretion rates as previously reported (11).</p><p>With the exception of SFN-GSH, accuracy and reproducibility were excellent and well within desired limits (Tables 2 and 3). The determination of the SFN-GSH conjugate proved somewhat problematic due to the very low concentration in the urine, but reanalysis of a larger aliquot of urine enabled accurate measurement. We also examined the impact of GSTM1 or GSTT1 polymorphisms on the rate of SFN metabolite excretion but found no significant effect (data not shown), as also reported earlier based upon analysis of samples collected at later time points and analyzed by the cyclocondensation assay (11).</p><p>The average urinary concentrations of each SFN conjugate and SFN are summarized in Table 4. In agreement with previous reports, SFN-NAC was the primary metabolite of SFN excreted in urine comprising 64.9% of the total products; SFN-Cys and SFN represented 27.6 and 7.2% of the total, respectively. The proportional distribution of metabolites compares well with SFN-urinary profiles as published by Al Janobi et al. (20). In the current analysis, the average rate of SFN and SFN-GSH-derived conjugates excreted was 41.0 μmol/12 h. Using the cyclocondensation method, Kensler et al. reported a total dithiocarbamate excretion rate of 42.1 μmol/12 h for all participants receiving the broccoli sprouts preparation averaged over the urine samples analyzed at later time points (days 5, 9, 10, and 12) in the same clinical trial (11). The concordance of the total SFN metabolite levels for the two methods indicates that the LC-MS/MS method utilizing D8-SFN internal standards gives reliable results. In addition, it provides sensitive, reliable, and accurate measurements for individual metabolites while greatly minimizing or eliminating sample preparation. Application of this method will now be expanded to examine the multifaceted actions of SFN as a chemopreventive agent.</p>

PubMed Author Manuscript

Immobilization of \xce\xb2-Galactosidases from Lactobacillus on Chitin Using a Chitin-Binding Domain

Two \xce\xb2-galactosidases from Lactobacillus, including a heterodimeric LacLM type enzyme from Lactobacillus reuteri L103 and a homodimeric LacZ type \xce\xb2-galactosidase from Lactobacillus bulgaricus DSM 20081, were studied for immobilization on chitin using a carbohydrate-binding domain (chitin-binding domain, ChBD) from a chitinolytic enzyme. Three recombinant enzymes, namely, LacLM-ChBD, ChBD-LacLM, and LacZ-ChBD, were constructed and successfully expressed in Lactobacillus plantarum WCFS1. Depending on the structure of the enzymes, either homodimeric or heterodimeric, as well as the positioning of the chitin-binding domain in relation to the catalytic domains, that is, upstream or downstream of the main protein, the expression in the host strain and the immobilization on chitin beads were different. Most constructs showed a high specificity for the chitin in immobilization studies; thus, a one-step immobilizing procedure could be performed to achieve up to 100% yield of immobilization without the requirement of prior purification of the enzyme. The immobilized-on-chitin enzymes were shown to be more stable than the corresponding native enzymes; especially the immobilized LacZ from L. bulgaricus DSM20081 could retain 50% of its activity when incubated at 37 \xc2\xb0C for 48 days. Furthermore, the immobilized enzymes could be recycled for conversion up to eight times with the converting ability maintained at 80%. These results show the high potential for application of these immobilized enzymes in lactose conversion on an industrial scale.

immobilization_of_\xce\xb2-galactosidases_from_lactobacillus_on_chitin_using_a_chitin-binding_domain

Introduction<!>Chemicals and Reagents<!>Bacterial Strains and Media<!>Plasmid Construction and Transformation<!>Evaluation of the Expression of Recombinant \xce\xb2-Galactosidases<!>Fermentation<!>Immobilization on Chitin Beads<!>Influence of pH and Temperature on the Activity and Stability of Immobilized Enzymes<!>Steady-State Kinetic Measurements<!>Reusability of the Immobilized Enzymes<!>Lactose Hydrolysis and Transgalactosylation Experiments<!>\xce\xb2-Galactosidase Assay<!>Protein Measurement<!>Sodium Dodecyl Sulfate\xe2\x80\x93Polyacrylamide Gel Electrophoresis (SDS-PAGE) Analysis<!>Statistical Analysis<!>Plasmid Construction and Expressions of Recombinant \xce\xb2-Galactosidases in L. plantarum WCFS1<!>Immobilization of \xce\xb2-Galactosidases on Chitin Beads<!>Partial Characterization of Immobilized \xce\xb2-Galactosidases<!>Steady-State Kinetic Constants<!>Effect of Temperature and pH on Enzyme Activity<!>Reuse of Immobilized Enzymes<!>Lactose Transformation

<p>β-Galactosidases (β-gals) (lactases, EC 3.2.1.23) are known as important enzymes for applications in the dairy industry,1 where they are used for lactose hydrolysis to produce low-lactose or lactose-free products as a response to lactose intolerance of consumers, which affects approximately 70% of the world population.2 Another useful property of β-gals is their transgalactosylation activity, by which health-promoting prebiotic galacto-oligosaccharides (GOS) can be formed from lactose.1,3 Many studies have demonstrated the indirect health benefits of GOS, where GOS promote growth and activity of beneficial intestinal microbes in the host.4–6</p><p>It is conceivable that β-gals from probiotic strains produce GOS that are also specific for these probiotics,7 and therefore β-gals from several probiotic Bifidobacterium spp.8,9 and prominently Lactobacillus spp. such as Lactobacillus reuteri, Lactobacillus plantarum, Lactobacillus sakei, Lactobacillus pentosus, or Lactobacillus bulgaricus have been studied in relation to GOS production during the past decade.10–14 β-Galactosidases from Lactobacillus species are reported to be of two main types, consisting of either two different subunits (heterodimeric or LacLM type) or two identical subunits (homodimeric or LacZ type).15,16 Two well-studied examples are the LacLM β-galactosidase from L. reuteri L10314 and the LacZ β-galactosidase from Lactobacillus delbrueckii subsp. bulgaricus (L. bulgaricus) DSM 20081,16 both belonging to glycoside hydrolase family GH2.</p><p>The recombinant β-gals from L. reuteri L103 and L. bulgaricus DSM20081 were successfully developed and overexpressed in food grade Lactobacillus hosts,16–18 yielding remarkably high levels of activity, for example, ca. 23−53 kU per liter of fermentation broth, respectively.16,17 The lactose conversion toward GOS of both native and recombinant β-gals from L. reuteri L103 and L. bulgaricus DSM 20081 were studied in detail in both batch and continuous bioreactors.14,16,19–21 Maximum GOS yields achieved were approximately 38% of total sugars (at a lactose conversion rate of 80% and an initial lactose concentration of ∼200 g/L) using LacLM from L. reuteri L103,20 or 50% GOS for recombinant LacZ from L. bulgaricus DSM 20081 (at 90–95% of lactose conversion and an initial lactose concentration of ∼200 g/L).16 These GOS mixtures contained mainly (nonlactose) disaccharides, trisaccharides, and tetrasaccharides, in which the transferred galactosyl moiety is attached via β-1,3- and β-1,6-linkages. These structures are important for potential prebiotics.20</p><p>To study the application of these enzymes in more detail, we aimed at immobilization of these two β-gals. To date, there have been no efforts on immobilization of β-gal from L. reuteri L103, even though β-gals are well-studied enzymes in terms of immobilization.22 Immobilization will enable the reuse of these biocatalysts, and it might also contribute to stabilization of the two β-gals, because their thermostability could be a limiting factor for their application.23</p><p>The chitin-binding domain (ChBD) is part of some chitin- or chitosan-degrading enzymes, chitinases or chitosanases, where it is responsible for the tight binding of the enzyme onto the substrate, thus increasing the hydrolytic activity.24,25 The ChBD of Bacillus circulans WL-12 chitinase A1 is one of the most well studied carbohydrate-binding domains. It belongs to the carbohydrate-binding module family 12 of the CAZY (carbohydrate-active enzyme) database. The three-dimensional structure of this ChBD has been determined by NMR.26 It consists of 45 amino acids, Ala655–Gln699, including several hydrophobic and aromatic residues with low solvent accessibility, thus forming a rigid and compact structure.26 Two β-sheets are composed of five strands including Thr660–Tyr662, Gln666–Tyr670, Lys673–Cys677, His681–Ser683, and Trp696–Leu698 residues.26 This ChBD has been reported to bind to only insoluble or crystalline chitin but not to the chitooligosaccharide, soluble derivatives of chitin, or other polysaccharides.27 On the basis of the high affinity to chitin, ChBD from WL-12 has been successfully applied for immobilization of several enzymes such as d-hydantoinase28 or levansucrase29 on chitin material. The procedure for the simultaneous purification and immobilization was a simple mixing of the enzyme solution with the insoluble chitin without the requirement of purified enzyme.28,29 Moreover, this method can overcome the disadvantages of other immobilization methods. For example, the covalent binding method immobilization required the harsh condition to create the bond between the enzyme molecule to the support material.23 The absorption method is based on weak forces such as van der Waals, hydrophobic interaction, or hydrogen interaction; thus, the immobilized enzymes are loosely bound to support materials.30</p><p>This study focuses on the immobilization of heterodimeric L. reuteri L103 LacLM and homodimeric L. bulgaricus DSM20081 LacZ β-gal on chitin using ChBD of B. circulans WL-12 chitinase A1. The recombinant enzymes were fused to the ChBD via DNA-based molecular methods. The biochemical characteristics of these fusion enzymes in comparison to the native enzymes are shown; thus, the effects of the ChBD and the immobilization on the properties of these enzymes are elucidated.</p><!><p>All chemicals were purchased from Sigma (St. Louis, MO, USA) unless otherwise stated and were of the highest quality available. The endonucleases were purchased from New England BioLabs (Ipswich, MA, USA) and used as recommended by the supplier. T4 DNA ligase was from Fermentas (Vilnius, Lithuania).</p><!><p>The bacterial strains, plasmids, and primers used in this study are listed in Tables 1 and 2. L. plantarum WCFS131 was grown in MRS medium (Oxoid, Basingstoke, UK) at 37 °C without agitation. Escherichia coli NEB5 α (New England BioLabs, Ipswich, MA, USA) as cloning host was cultivated in Luria-Bertani (LB) medium at 37 °C with shaking at 200 rpm. Agar media were prepared by adding 1.5% agar to the respective media. Unless otherwise stated, the antibiotic concentrations were 5 or 200 μg/mL of erythromycin (Erm) for Lactobacillus or E. coli, respectively, and 100 μg/mL of ampicillin for E. coli.</p><p>Oligonucleotide primers for PCR amplification were supplied by VBC-Biotech Service GmbH (Vienna, Austria). The appropriate endonuclease restriction sites were introduced in the forward and reverse primers (Table 2).</p><!><p>The DNA amplification was performed with Phusion High-Fidelity DNA polymerase (Finnzymes, Espoo, Finland). Plasmid DNA from E. coli was purified using the Gene Elute plasmid Miniprep kit (Sigma, St. Louis, MO, USA). DNA was purified with the WizardRSV Gel and PCR Clean-up system kit (Promega, Madison, WI, USA). The pJET1.2 plasmid (CloneJET PCR cloning kit, Fermentas) was used for subcloning when necessary.</p><p>Three recombinant fusion proteins were constructed. Two are based on LacLM from L. reuteri L103, and the chitin-binding domain ChBD was attached upstream of LacL (termed ChBD-LacLM) or downstream of LacM (termed LacLM-ChBD). The third fusion protein was based on LacZ of L. bulgaricus DSM20081 with ChBD linked downstream of LacZ (termed LacZ-ChBD).</p><p>For construction of the expression plasmids of the fusion protein LacLM-ChBD or LacZ-ChBD, the fragment of chbd of B. circulans WL-12 chitinase A1 was amplified from the plasmid pTxB1 (New England BioLabs) with the primer pair F1 and R1 (Table 2). After double digestion with XhoI and Acc65I, this fragment was ligated to the fragment of pSIP403,32 which contained gusA as the original reporter gene and was treated by the same endonucleases, resulting in plasmid pSCBD-gusA (Table 1; Figure 1). The gusA gene was then excised by double digestion with NcoI and XhoI to obtain the empty vector pSCBD. This empty vector was ligated to the NcoI–XhoI fragment of lacLM (encoding LacLM from L. reuteri L103) from pHA103.1,15 resulting in pSCBDlac1 (Table 1; Figure 1). Similarly, lacZ was amplified from pJETlacZ16 by PCR with the two primers F4 and R4 (Table 2) and then digested by BsmBI and XhoI prior to ligation into empty pSCBD, resulting in pSCBDlac3 (Table 1; Figure 1).</p><p>For construction of the expression plasmid for ChBD-LacLM, chbd from pTxB1 was amplified with the two primers F2 and R2, resulting in a fragment of F2-chbd-R2. The lacLM gene was PCR-amplified from the lacLM template in pTH103.1 using the F3 and R3 primers, resulting in the fragment F3-lacLM-R3. Because of the 18 nt complementary sequence between F3 and R2 (Table 2), the fragment F2-chbd-R2 could be used as a mega forward primer in combination with R3 as reverse primer to amplify the F3-lacLM-R3 template, resulting in the fragment F2-chbd-lacLM-R3. The PCR product from this amplification was digested by BsaI and XhoI and ligated to NcoI–XhoI-digested pSIP403, resulting in pSCBDlac2 (Table 1; Figure 1).</p><p>This strategy resulted in expression vectors, in which the target gene, lacLM or lacZ linked to the chbd sequence, is controlled by the inducible promoter PsppA, similar to gusA in the original pSIP403 vector.32 These plasmids were then electroporated into competent cells of L. plantarum WCFS1 as previously described.33 The transformants carrying the plasmids were determined and screened by using a colony PCR amplification of the target fusion genes.</p><!><p>This experiment was performed following a previous method.18</p><!><p>To obtain sufficient amounts of the enzymes for characterization, immobilization, and application, L. plantarum WCFS1 harboring different plasmids was cultivated in 1 L of medium to obtain larger amounts of the recombinant enzymes. The bacterial cells were induced at OD600 nm ∼ 0.3 and harvested at OD600 nm ∼ 6. The cell pellets were disrupted by homogenization with a French press (Aminco, Silver Spring, MD, USA). The cell-free extract, clarified by ultracentrifugation (Beckman, USA) at 30000g and 4 °C for 20 min, was used either directly for immobilization on chitin beads (New England BioLabs) or for a single-step affinity purification using the p-aminobenzyl 1-thio-β-d-galactopyranoside (ABTG) resin (Sigma) as described in a previous study.14</p><!><p>Before use, the chitin beads were washed three times with sodium phosphate buffer (50 mM, pH 6.5) and resuspended in the same buffer. Five hundred microliters of chitin bead suspension was mixed with the same volume of diluted cell-free crude extracts of 1000 U/mL of oNPG activity. The immobilization experiments were carried out at 4 °C with gentle agitation for 1–18 h. Chitin beads were separated from the supernatant by filtration and rinsed with sodium phosphate buffer. The supernatant and wash solutions were collected and pooled for SDS-PAGE analysis and protein measurement. Chitin beads were resuspended in buffer for further studies.</p><p>The immobilization yield (IY) and the activity retention (AR) were calculated according to the methods of Klein et al.34 IY (%) is defined as the ratio of immobilized activity in relation to total applied activity, in which the immobilized activity was determined by subtraction of the residual activity in the supernatant after immobilization from the total applied activity. AR (%) is the percentage of activity measured on chitin beads in relation to the theoretical immobilized activity.</p><p>The amount of chitin-bound protein was indirectly estimated by subtraction of the protein concentration in the supernatant after immobilization from that in the crude extract (prior to the immobilization).</p><!><p>The pH optimum of β-gal activity was evaluated in Britton–Robinson buffer17 with the pH ranging from 4 and 9. To determine the pH stability, the immobilized enzymes were incubated at 30 or 37 °C in Britton-Robinson buffer of different pH values. At various time intervals, the residual activities were measured with the oNPG assay. The inactivation constants kin were obtained by linear regression of ln(residual activity) versus time. The half-life values τ1/2 were calculated using τ1/2 = ln(2)/kin.35</p><p>The optimum temperature for hydrolysis activity of β-gals with both substrates lactose and oNPG was determined in the range of 20–80 °C. For estimation of the kinetic thermostability, enzymes were incubated in 50 mM sodium phosphate buffer, pH 6.5, at different temperatures ranging from 20 to 80 °C. At various time intervals, residual activities were measured with oNPG as substrate and plotted versus the incubation time. The half-life values of thermal inactivation (τ1/2) were similarly calculated as above.</p><!><p>To estimate the kinetic parameters of recombinant and fusion β-gals for lactose and oNPG, substrate concentrations were varied from 1 to 25 mM for oNPG and from 10 to 600 mM for lactose.14 The enzyme assays were performed for 10 min at 30 °C in 50 mM sodium phosphate buffer, pH 6.5. The kinetic parameters were calculated by using the Henri–Michaelis–Menten model and nonlinear least-squares regression.</p><!><p>To assess the possibility of recycling the immobilized enzyme preparations for conversion, immobilized β-gal LacLM L. reuteri L103 was added to sodium phosphate buffer (50 mM, pH 6.5) containing 600 mM lactose and maintained at 30 °C and 600 rpm agitation for the conversion of lactose. Every 24 h, the chitin beads carrying the enzyme were filtered off, rinsed with buffer, and reused in another batch conversion experiment under identical conditions. The filtrates from every conversion cycle were collected to determine the glucose concentration by using a commercial d-glucose assay kit (GOPOD Format, Megazyme, Wicklow, Ireland). The glucose concentration in the filtrate from the first conversion was taken as 100% of converting ability of the immobilized enzyme preparation.</p><p>Immobilized LacZ from L. bulgaricus DSM 20081 was also tested for its reusability but using the substrate oNPG. After 5 min of incubation with o-nitrophenyl-β-d-galactopyranoside (oNPG) at 30 °C, chitin beads carrying LacZ were separated from the liquid phase and reused in the next conversion experiments. The concentration of oNPG in the liquid phases after the conversion was measured with a spectrophotometer at 420 nm and used to calculate the residual converting ability of enzyme in each repeating conversion.</p><!><p>The transformation of lactose was carried out in batch mode. Both chitin-immobilized I-LacLM-ChBD and purified soluble LacLM-ChBD were added at equal concentrations of 1.5 Ulactose/mL to sodium phosphate buffer (50 mM, pH 6.5) containing 10 mM MgCl2 and 205 g/L lactose. Lactose conversion experiments were performed at 30 °C for 24 h at 600 rpm of agitation.</p><p>The chitin-immobilized β-gal of L. bulgaricus was applied in different amounts ranging from 1.7 to 9.7 Ulactose/mL to sodium phosphate buffer (50 mM, pH 6.5) containing 10 mM MgCl2 and 50 or 205 g/L lactose. This enzyme preparation was also used for conversion of lactose in ultrahigh-temperature-treated whole cow's milk. The incubation temperature was varied from 37 to 60 °C.</p><p>In all conversion experiments, samples were periodically withdrawn, heated at 99 °C for 5 min, and further analyzed for lactose, galactose, glucose, and GOS present in the samples. Qualitative analysis of sugar was performed by thin-layer chromatography (TLC) as described previously by Nguyen and co-workers.14 Furthermore, samples from conversion experiments were quantitatively analyzed by high-performance liquid chromatography (HPLC) (Dionex, Germany) equipped with a refractive index detector and an Aminex HPX-87K (300 mm × 7.8 mm) carbohydrate analysis column (Bio-Rad, Hercules, CA, USA). The chromatographic separation was performed at 80 °C with ultrapure water used as eluting solvent at a flow rate of 0.5 mL/min. The concentration of saccharides was calculated by interpolation from external standards. Total GOS concentration was calculated by subtraction of the quantified saccharides (lactose, glucose, galactose) from the initial lactose concentration. The GOS yield (%) was defined as the percentage of GOS produced in the samples compared to initial lactose.</p><!><p>β-Galactosidase activity was determined using oNPG or lactose as substrate following the method of Nguyen et al.14 In brief, the activity assays were performed in 50 mM sodium phosphate buffer of pH 6.5 at 30 °C, and the final substrate concentrations in the 10 min assays were 22 mM for oNPG and 600 mM for lactose. One unit of oNPG activity was defined as the amount of enzyme releasing 1 μmol of oNP per minute, whereas 1 unit of lactase activity was defined as the amount of enzyme releasing 1 μmol of d-glucose per minute under the given conditions.</p><!><p>Protein concentration was determined by using the method of Bradford36 with bovine serum albumin (BSA; Sigma) as standard.</p><!><p>For visual observation of the expression level of the recombinant β-gals in L. plantarum WCFS1 and the effectiveness of the immobilization, cell-free extracts, liquid phases, and chitin beads (after immobilization) were analyzed by SDS-PAGE using 3-(N-morpholino)propanesulfonic acid (MOPS) as running buffer.16</p><!><p>All experiments and measurements were performed at least in duplicate, and the data are given as the mean ± standard deviation when appropriate. Student's t test was used for the comparison of data with significance value α = 0.05.</p><!><p>In this work, the three expression plasmids pSCBDlac1, pSCBDlac2, and pSCBDlac3, containing the sequences of the three recombinant fusion proteins LacLM-ChBD, ChBD-LacLM, and LacZ-ChBD, respectively, were constructed and successfully electroporated into L. plantarum WCFS1. The transcription of the encoding sequences lacLM-chbd, chbd-lacLM, and lacZ-chbd is regulated by the inducible promoter PsppA in these plasmids (Figure 1). The expression levels of the different constructs and the recombinant β-galactosidases in L. plantarum WCFS1 were investigated.</p><p>Induced cells formed intracellular LacLM-ChBD at around 32 U/mL of fermentation broth with a specific activity of ca. 179 U/mg protein (Table 3). ChBD-LacLM was expressed in 3-fold lower yields (p < 0.05), giving ca. 11 U of β-galactosidase activities per milliliter of fermentation broth with a specific activity of 54 U/mg. The basal expression from the expression plasmids in noninduced cells was unexpectedly high (7.35 and 4.06 U/mg for LacLM-ChBD and ChBD-LacLM, respectively (Table 3)). The induction ratios (ratio of the β-galactosidase activities obtained under induced conditions divided by the activity under noninduced conditions in cells harvested at similar OD600 values of 1.8) were 24 and 14, respectively, which is lower than previously reported for different target genes expressed with the same expression vector pSIP403.16,17 It should be noted that the activity level produced by the wildtype host strain under identical growth conditions is very low (0.07 U/mg),15 and hence the activities measured can be attributed solely to the recombinant enzymes.</p><p>The expression of LacZ-ChBD in L. plantarum WCFS1 gave a higher volumetric activity (around 65.83 U/mL) but lower specific activity (40.27 U/mg) when compared to the expression of LacLM-ChBD (p < 0.05). Again, a rather high basal activity was observed in noninduced cultivation (16.77 U/mL and 9.43 U/mg), even though PsppA is known as a tightly controlled promoter.32,37 As a consequence, a low induction factor of around 4 was obtained with this system (Table 3).</p><p>The fusion of ChBD had significant effects on expression levels of the recombinant enzymes. This is obvious from, for example, LacZ, as expression of the lacZ gene without fusion to the chbd fragment gave much higher expression (i.e., 180-190 U/mg) with the same host, expression system, and induction conditions.16 Moreover, the expression levels of the recombinant proteins are remarkably different depending on the position of ChBD when fused to LacLM. The lower expression levels were observed when chbd was fused upstream of lacL (ChBD binding N-terminally to the large subunit LacL) (Table 3). It is shown that the nucleotide sequences at the beginning of a gene can change the stability of the secondary structure of the transcribed mRNA and thus affect the translation process.38 The secondary structures of the chbd-lacLM mRNA and lacLM-chbd mRNA are predicted by the Mfold tool (http://unafold.rna.albany.edu/?q=mfold) (data not shown). With chbd-lacLM, a hairpin loop including the ribosome-binding site RBS (AGGAG) is predicted in the region of the 5'UTR and the initial nucleotides of the chbd. This hairpin loop might prevent the binding of ribosome to the mRNA to initiate the translation process. Meanwhile, with lacLM-chbd, the RBS is not involved in the hairpin loop; hence, it might be an explanation for the higher expression level of the LacLM-ChBD fusion protein mentioned above.</p><p>SDS-PAGE analysis of cell-free extracts of L. plantarum WCFS1 harboring the different plasmids clearly indicates strong bands for the recombinant proteins at approximately 72 and 41 kDa (lane 2, Figure 2A) for LacLM-ChBD, 78 and 35 kDa (lane 4, Figure 2A) for ChBD-LacLM, and ∼120 kDa (lane 2, Figure 2C) for LacZ-ChBD. These sizes are in agreement with those of LacL (72 kDa) and LacM (35 kDa) from L. reuteri L10314,15 and lacZ from L. bulgaricus DSM20081 (115 kDa),16 and ChBD (∼6 kDa).</p><p>The deduced amino acid sequences of the three fusion proteins were used to predict their 3D structures using the RaptorX tool (http://raptorx.uchicago.edu/). The prediction shows that the ChBD domain arranges well separated from the main protein LacM (in the case of LacLM-ChBD) and LacZ (in the case of LacZ-ChBD), but not with LacL (for ChBD-LacLM), where it interacts more tightly with the LacL subunit (see Figure 10). It should be noted that with LacZ-ChBD, the ChBD fold was not predicted accurately, as it was not in agreement with the structure of ChBD alone (see Figure 10D) or in LacM-ChBD (see Figure 10A); nevertheless, it was predicted to be separately positioned from the main protein (Figure 10C). It should be noted that the chbd from pTxB1 encodes for ChBD from chitinase A1 of B. circulans WL-12 with 45 amino acids, from Ala655 to Gln699, as previously reported26 and 7 extra residues at the N-terminus. Moreover, 5 residues upstream of Thr660 are not involved in the essential structural region of ChBD.26 The sequence of 12 residues might be a flexible linkage between LacM or LacZ to ChBD. Meanwhile, for ChBD-LacLM, the linkage is only the last Gln699 of ChBD, which is not involved in forming the β-sheet.26 These predicted structures indicate that in both LacLM-ChBD and LacZ-ChBD, the interaction of ChBD with chitin beads will affect the main protein to a lesser extent than in ChBD-LacLM. Even though these structures need to be confirmed by crystallization/structure elucidation, they are in agreement with our further results in this study (see below).</p><!><p>The three different fusion proteins were used at equal activities of 500 UoNPG to study their immobilization on chitin beads (11 mg of dry weight), and the residual activities in the liquid phase after the immobilization reaction as well as the bound activities were subsequently analyzed for all three constructs (Table 4). An important parameter in immobilization is the immobilization yield (IY) as it is an indication of how much of the applied protein is actually bound to the carrier. Immobilization yields for both LacLM-ChBD and LacZ-ChBD were very high, at >91%, and especially the IY of LacLM- ChBD, which was close to 100% (0.67% of residual activity detected in the supernatant after immobilization), is very promising. In contrast, the IY of ChBD-LacLM was much lower, at ca. 52% (Table 4). These values are further corroborated by SDS-PAGE analysis using samples of the liquid phases and chitin beads after immobilization (Figures 2B and 2C). The bands of ChBD-LacL and lacM or LacZ-ChBD can be seen in the lanes of liquid phase samples (lane 5 in Figure 2B; lane 3, 4, and 5 in Figure 2C). The patterns of chitin bead samples, with only two bands corresponding to the two subunits of immobilized LacLM (LacL and LacM-ChBD in lane 3 and ChBD-LacL and LacM in lane 6, Figure 2B) or the single strong band of LacZ-ChBD (Figure 2C, lane 6) with only a minor very faint band indicating an impurity, indicate the high specific affinity of ChBD to chitin.25 This observation is in accordance with many other studies using ChBD for enzyme immobilization.28,29,39 LacLM from L. reuteri L103 is a heterodimer consisting of two subunits. Fusing the ChBD fragment N-terminally to LacL or C-terminally to LacM clearly had different effects on the immobilization of the recombinant fusion enzymes on chitin beads. A possible explanation for these results could be that in the ChBD-LacLM fusion the ChBD is positioned in such a way that it cannot interact with the chitin beads unperturbed. This is in agreement with the predicted structures of these fusion proteins (see Figure 10). LacZ from L. bulgaricus DSM20081 is a homodimer,16 and therefore each of the identical subunits will carry its own ChBD. These two chitin-binding domains per fusion protein might interfere with each other when binding on chitin, which could be the reason for the slightly lower efficiency in immobilization compared to LacLM-ChBD (Table 4).</p><p>Even though the immobilization yields were very high for two of the constructs (Table 4), the activity retention (AR) on chitin beads was much lower than expected. The residual activities on chitin beads for ChBD-LacLM and LacLM-ChBD were ∼9.19 and 25.89% in comparison with the initially applied activity, respectively. This amounted to AR values of 19–25%. A low AR value of ca. 13.7% was also observed for LacZ-ChBD. However, AR depends on the applied activity of our enzymes, because applying lower activities resulted in higher AR values (data not shown). To date, various methods have been used for immobilization of fungal or bacterial β-gals, yielding different ARs.22,40 For instance, a range of 2.5–9.5% of ARs of β-galactosidase from E. coli, which was immobilized on poly(2-hydroxyethyl methacrylate) membrane using entrapment, was reported by Baran and co-workers.41 However, most of these immobilization methods are based on purified enzymes, whereas in this study a single-step procedure was performed for the immobilization.</p><!><p>The effect of ChBD and of the immobilization on the characteristics of the β-gals was further assessed. Three chitin-immobilized enzymes of LacLM-ChBD, ChBD-LacLM, and LacZ-ChBD are termed I-LacLM-ChBD, I-ChBD-LacLM, and I-LacZ-ChBD, respectively, hereafter. Free LacLM-ChBD was also purified and partially characterized to compare with the immobilized I-LacLM-ChBD.</p><!><p>Table 5 presents the kinetic constants of the recombinant fusion β-gals for their two substrates, lactose and oNPG. For I-LacLM-ChBD, the vmax values for both substrates are higher than those of I-ChBD-LacLM (p < 0.05). With 1.74 and 0.72 μmol min−1 mg−1 these vmax values for the natural substrate lactose are much lower than the corresponding value of the free fusion protein (17.8 μmol min−1 mg−1), as well as LacLM isolated from its natural source L. reuteri L103,14 or expressed recombinantly either in E. coli15 or in L. plantarum WCFS17 with vmax values ranging from 34 to 43 μmol min−1 mg−1 (Table 5). The Michaelis constant Km was also negatively affected by the immobilization, but judging from the Km values determined, this effect is much less dramatic than for vmax. This obviously indicates that the hydrolysis activity of the chitin-bound fusion enzymes is negatively influenced by immobilization. The effect of immobilization on kinetic parameters of I-LacZ-ChBD was also negative, but here the decrease in vmax and hence kcat was less drastic (Table 5).</p><!><p>The immobilized fusion enzymes typically showed their highest activity at 50–55 °C with both oNPG and lactose as substrate for the 10 min assays (Figure 3); this is also the optimal range of temperature observed with one of the counterparts, the soluble enzyme LacLM-ChBD (data not shown). At higher temperatures (>70 °C) the enzyme activity was rapidly lost. The range of optimal temperature for the activity of the immobilized enzymes is comparable to that of the free enzyme without ChBD from the native strain L. reuteri L103 or recombinant LacLM expressed in different hosts.14,15,17 Interestingly, the immobilized preparation I-LacZ-ChBD showed a slightly increased temperature optimum of 65 °C (Figure 3) compared to the free enzyme LacZ without the ChBD with an optimum of 60 °C.16 We did not observe a similar increase for I-LacLM-ChBD and I-ChBD-LacLM, the reason for which might be that only one subunit is involved in binding to chitin when LacLM is used.</p><p>The thermostability of the different immobilized β-galactosidase preparations was tested at various temperatures in 50 mM sodium phosphate buffer, pH 6.5, by determination of the half-life times τ1/2 (Table 6). I-LacZ-ChBD proved to be the most stable preparation at all tested temperatures. For example, τ1/2 of this enzyme was 203 h at 37 °C, whereas it was <170 h for I-LacLM-ChBD and I-ChBD-LacLM at 30 °C. Free β-gal of L. bulgaricus was previously shown to be relatively thermostable (τ1/2 for 145 and 345 h without and with, respectively, the presence of 10 mM Mg2+ at 37 °C).16 Furthermore, the stability of the immobilized preparation I-LacZ-ChBD was further increased in the presence of 10 mM Mg2+ to a half-life time τ1/2 of almost 48 days (1155 h) at 37 °C (Table 7). The soluble fusion protein LacLM-ChBD was considerably less thermostable than its immobilized counterpart I-LacLM-ChBD (τ1/2 of 19 h versus 169 h at 30 °C; Table 6); thus, immobilization onto chitin beads via the chitin-binding domain can improve the kinetic stability of an enzyme significantly.</p><p>The immobilized enzyme preparations showed their optimum activity in the pH range of 6–6.5 for both substrates lactose and oNPG, with the exception of I-LacZ-ChBD and lactose having an optimal pH of 8 without clear reason (Figure 4). It is likely related to the isoelectric point (pH 9) of ChBD.27 When stored at different pH values, the enzymes were most stable in the pH range that also coincides with the pH optima, that is, 6–6.5 (Table 7). The binding of ChBD from B. circulans WL-12 chitinase to chitin was shown to be best at pH 9 (binding capacity of 90%) and less at pH 4 (binding capacity of 80%).27 However, LacLM from L. reuteri L103 and LacZ from L. bulgaricus were reported to have very low activity at pH 4,14,16 which was also observed with the immobilized enzymes in this study (Figure 4). This suggests that the effect of pH on the fusion enzymes is mainly on the catalytic activity rather than the binding of ChBD to chitin.</p><!><p>A study on the reusability of the immobilized β-gals, using subsequent hydrolysis steps with oNPG or lactose as substrate, was carried out. The residual converting ability of immobilized enzyme preparations was plotted versus the number of cycles of application (Figure 5). In this study I-LacLM-ChBD and I-LacZ-ChBD exhibited better stability, because these immobilized enzymes retained approximately 80% of their initial ability of substrate conversion after four to eight cycles. I-ChBD-LacLM showed a more rapid decrease of relative activity (Figure 5).</p><p>These results are comparable with previously reported data. Chiang et al. immobilized levansucrase on chitin using the B. circulans WL-12 ChBD, which retained approximately half of its initial activity at the end of the seventh cycle.29</p><!><p>Figures 6 and 7 show a qualitative analysis by TLC of products formed during lactose conversion using different immobilized β-gal preparations under various conditions. In general, the profile of products shown is comparable to the profile of the corresponding free enzymes previously reported14,16 as well as to the profile of Vivinal GOS (Amersfoort, The Netherlands) used as "reference" (Figure 6). This indicates that similar GOS mixtures are formed by the free and immobilized enzymes.</p><p>As expected, a higher enzyme activity or higher temperatures led to faster conversion (Figure 6). The β-gal LacZ from L. bulgaricus was shown to be a thermostable enzyme,16 and therefore it catalyzed lactose conversion efficiently at the higher temperature of 50 °C. In this study, the chitin-immobilized LacZ again showed good thermostability, thus resulting in a comparable profile of products at 60 °C compared to 37 °C. Regardless of the conditions of the conversion, the maximal GOS yield was around 23–24% (Table 8; Figure 8), which is lower than the yield obtained with the free enzyme LacZ (approximately 50%) as reported by Nguyen and co-workers.16 A possible explanation could be that the immobilization reduces the contact between enzyme and lactose as well as the monosaccharide sugars for transgalactosylation, thus resulting in reduced GOS products. This may be also the reason for the rather high residual lactose concentration at the end of the conversion run (Table 8; Figure 6). This effect of immobilization was also observed by Sheu et al. when using chitosan-immobilized β-gal from Aspergillus oryzae for lactose conversion.42 The conversion of lactose in the samples of ∼5% w/v, which is the concentration of lactose in milk, using immobilized recombinant β-gal I-LacZ-ChBD with a relatively high activity (9.7 Ulactose/mL) at a high temperature of 60 °C was fast, as 81% of lactose was converted after only 1 h (Table 8).</p><p>When used for batch conversion experiments of lactose, I-LacLM-ChBD and I-ChBD-LacLM also resulted in a profile of GOS products comparable to that of the native β-gal L103 (Figure 7). Lactose conversion was faster with the free enzyme of LacLM-ChBD compared to I-LacLM-ChBD at equal activity loading (Figure 9), which might be explained by limitations in diffusion of the substrate to the active site. A maximum GOS level of ca. 39–40% of total sugars was reached by both enzyme preparations, soluble and immobilized (Figure 9), yet the time for reaching this maximum was significantly longer for I-LacLM-ChBD at 12 h (Figure 9B) compared to 8 h for LacLM-ChBD (Figure 9A). These maximum yields were obtained at a similar lactose conversion of ca. 85%. This indicates that the conversion of lactose and the yield of GOS obtained by the recombinant enzymes are comparable to the conversion catalyzed by their counterpart, the native enzyme LacLM L. reuteri L103.14</p><p>In conclusion, this work describes the immobilization of two lactobacillal β-galactosidases, a homodimeric β-galactosidase of the LacZ type and a heterodimeric β-galactosidase of the LacLM type, onto chitin via a chitin-binding domain. This could provide a promising and efficient approach for lactose hydrolysis and production of prebiotic galacto-oligosaccharides because the enzyme can be purified from the crude cell extract and immobilized in one simple step. The immobilized fusion enzyme can be stably reused for several cycles for lactose hydrolysis and transformation. Preliminary results from an ongoing investigation of lactose conversion in continuous mode using these immobilization enzymes also show a high potential for an industrial application of these immobilized enzymes.</p>

PubMed Author Manuscript

Cost-effective sol-gel synthesis of porous CuO nanoparticle aggregates with tunable specific surface area

CuO nanoparticles (NPs) are applied in various key technologies, such as catalysis, energy conversion, printable electronics and nanojoining. In this study, an economic, green and easy-scalable sol-gel synthesis method was adopted to produce submicron-sized nanoporous CuO NP aggregates with a specific surface area > 18 m²/g. To this end, a copper-carbonate containing precursor was precipitated from a mixed solution of copper acetate and ammonia carbonate and subsequently calcinated at T ≥ 250 °C. The thus obtained CuO nanopowder is composed of weakly-bounded agglomerates, which are constituted of aggregated CuO NPs with a tunable size in the range of 100-140 nm. The CuO aggregates, in turn, are composed of equi-axed primary crystallites with a tunable crystallite size in the range of 20-40 nm. The size and shape of the primary CuO crystallites, as well as the nanoporosity of their fused CuO aggregates, can be tuned by controlled variation of the degree of supersaturation of the solution via the pH and the carbonate concentration. The synthesized submicron-sized CuO aggregates can be more easily and safely processed in the form of a solution, dispersion or paste than individual NPs, while still offering the same enhanced reactivity due to their nanoporous architecture.Advanced nanotechnologies in the fields of catalysis, energy conversion, storage and sensing devices rely on the accurate control of the shape and the size of materials from the nano-up to the micrometer scale. This requirement has boosted the development of a wide range of synthesis techniques for producing metallic, insulating and semiconducting nanoparticles (NP) of controllable sizes and shapes (cf. reviews 1 and 2 ). Reported studies on engineered and environmental NP-based systems generally focus on the size and shape of the smallest undividable entity (or building block), commonly referred to as the primary particle (or crystallite). However, successive manufacturing and processing steps (or environmental exposure) generally induce aggregation and/or agglomeration of primary NPs into larger entities with sizes of up to several microns. NP aggregates (or secondary particles) consist of strongly bonded or fused primary NPs, which cannot be separated by subsequent handling and processing steps (i.e. the aggregation process is irreversible). NP agglomerates comprise assemblies of more weakly bound primary NPs and NP aggregates (and/or a mixture thereof), which can be separated into their individual constituents by providing sufficient external energy and stabilized through the addition of suitable dispersion agents 3 . For practical applications, the actual properties of the nanomaterial (e.g. specific surface area, chemical reactivity, dispersibility and toxicity) will be governed by the size, shape and density (i.e. nanoporosity) of the NP aggregates and/or agglomerates and not solely by the primary particles 4,5 . For example, the extent to which the internal surface of a loosely clustered (and thus nanoporous) NP aggregate is accessible to interpenetrating gaseous or liquid species will critically influence its dispersibility, chemical reactivity and sintering behavior, which are of key importance for the development of e.g. catalysis, printed electronics and joining technologies 6,7 . Notably, sub-micron sized nanoporous NP aggregates also have the advantage that they can be more easily (i.e.

cost-effective_sol-gel_synthesis_of_porous_cuo_nanoparticle_aggregates_with_tunable_specific_surface

<!>Experimental<!>2−<!>Material characterization.<!>Results and Discussion<!>2−<!>Effect of pH and [CO<!>2−<!>Effect of pH, [CO 3 2−<!>Conclusion<!>2−<!>Data Availability

<p>; in addition of OH − ) and the pH 24,25 . Inorganic or organic additives, such as urea, polyethylene glycol or polyvinylpyrrolidone can be added to steer the precipitation and aggregation of the precursor phase, thus enabling the formation of specific hierarchical CuO nanostructures 22,26 . For example, Cu(NO 3 )-salt solutions in the presence of urea were aged at 90 °C to yield spherically-shaped malachite (CuCO 3 •Cu(OH) 2 ) precursor particles, which after calcination at 700-800 °C transform into micro-sized spherical CuO particles 27 . Reproducible results can only be obtained if key factors such as the pH, type and concentrations of the reactants are well controlled 22 .</p><p>In the present study, a simple, green and low-cost sol-gel synthesis method based on a copper-carbonate species containing precursor phase, using copper acetate and ammonia carbonate salts without the addition of any additives, was selected and optimized to produce loosely clustered (and thus nanoporous) CuO NP aggregates with an average size in the range of 100-140 nm and a resulting SSA that is significantly higher than typical SSA values in the range of 10-15 m²/g, as reported for commercially available high-purity (i.e. ≥99.99%) CuO nanopowders. Sol-Gel processes are usually used to synthesize nonmetallic inorganic materials from particle dispersions. They are easily upscalable and can be conducted with cheap and non-toxic chemicals 28 . Importantly, the synthesis was performed without organic additives (which contaminate the product phase and complicates the complex forming process of precursor phases). The evolution of the crystal structure, size and shape of the primary crystallites, which develop from the precursor phase during the calcination treatment, were monitored by measuring the peak broadening through high temperature in-situ XRD. The size, shape and cluster density of the resulting CuO aggregates, as constituted of individual clusters of firmly bonded primary CuO crystallites, was investigated by complementary analytical techniques, including Scanning Electron Microscopy (SEM), Transmission Electron Microscopy (TEM), Dynamic Light Scattering (DLS) and Brunauer-Emmett-Teller (BET) analysis. Key parameters in the sol-gel synthesis procedure were identified (e.g. the supersaturation level, carbonate concentration and the resulting pH of the mixed solution) and tuned to obtain an enhanced SSA of the synthesized CuO nanopowder for the targeted applications.</p><!><p>Material synthesis. Synthetic malachite can be prepared by reacting a solution of copper(II) nitrate with a solution containing the equivalent amount of sodium carbonate at room temperature; i.e. the copper-to-carbonate concentration ratio is fixed at [CO 3 2− ]/[Cu 2+ ] = 1.0 29 . Accordingly, solutions were prepared by mixing stoichiometric amounts of fresh aqueous 15 mM copper acetate (Sigma Aldrich) and 15 mM ammonia carbonate (Alfa Aesar) solutions at room temperature, while continuously stirring the mixed solution 30 . This resulted in a pH of 5.8 for the mixed solution, which remained stable during a stirring duration of 2 h, as monitored using a pH-meter (Metrohm 914 pH/Conductometer). Upon combining the starting solutions at room temperature (further referred to as mixed solution), precipitation of particles (the gel) from the blue-greenish solution was immediately observed. In an acidic environment, copper(II) ions would remain in solution without forming precipitates over the described time period. For more neutral environments of a pH above 5.6 (as is the case for the present study; see above), complexation of Cu 2+ leads to the formation of precipitates. Due to the supersaturation of the solution with respect to the formation of malachite, successive precipitation of the copper(II) complexes and gel formation occurs, and the composition of this precursor phase moves towards the composition of malachite. The chemical reaction to malachite can be written as www.nature.com/scientificreports www.nature.com/scientificreports/</p><p>When the pH exceeds 8.5, the copper-carbonate-hydroxide precursor phase complexes would gradually dissolve over time by ammonia leaching (resp. they would not even be formed), eventually leading to the formation of primarily copper(II) tetra ammine complexes 31 , i.e. The stability window for the formation of malachite in the mixed solution is therefore between pH 5.6 and 8.5. To evaluate the effect of the pH of the mixed solution on the size and shape of the copper-containing precursor phase (and thus on the final CuO end product), additional synthesis routes were performed at a fixed pH value of either 5.6 or 7.0 by adjusting the added volume of the ammonia carbonate solution, resulting in [CO 3 2− ]/[Cu 2+ ] ratios of 0.4 and 2.5, respectively. Depending on the regulated pH, either an ammonium carbonate deficiency (i.e. [CO 3</p><!><p>]/[Cu 2+ ] < 1 for pH < 5.83) or surplus [CO 3</p><p>]/[Cu 2+ ] > 1 for pH > 5.83) is established, which affects the supersaturation level of the mixed solution with respect to the nucleation of the targeted malachite precursor phase. Unless stated otherwise, the mixed solution was continuously stirred over the defined duration time of t stirring = 2 h. Afterwards, the gel precipitate was collected by centrifugation, washed with ethanol and distilled water. The freshly washed precipitate was then dried in a muffle furnace in air at a temperature of 60 °C for a fixed duration of t drying = 6 h. The resulting (largely) dehydrated precursor phase was decomposed into CuO by annealing in a muffle furnace in air at 400 °C for a fixed duration of t calcination = 4 h (further referred to as calcination), i.e.</p><p>A schematic of the above-described synthesis procedure is depicted in Fig. 1.</p><!><p>The crystallinity of the gel precursor, as well as the phase purity and average primary crystallite size of the synthesized CuO nanopowder, were investigated by X-ray diffraction (XRD) using a PANanalytical X'Pert Pro Multi-Purpose (MPD) X-ray diffractometer. The XRD data were collected in the 2Θ range of 10-100° 2Θ with a step size of 0.026° using Cu-Kα 1-2 radiation (λ average = 0.15418 nm, 45 kV and 40 mA).</p><p>A time-resolved XRD study of the precursor-to-CuO transformation during calcination was conducted using a PANalytical X'Pert PRO MPD X-ray diffractometer (same measurement parameters) with a gas-tight Anton Paar XRK-900 heating chamber equipped for heating and gas feeding (5850 TR, Brooks instrument) in static air. Thermo-gravimetric analysis (TGA) combined with differential scanning calorimetry (DSC) was conducted using a Netzsch STA449 F3 Jupiter with a low-temperature silver oven. The sample (mass: ~31.6 mg) was handled in an Al sample pan (85 μl) and heated from room temperature to 400 °C at a rate of 5 K/min under flow of synthetic air (60 ml/min). The data were corrected by an empty pan scan measured under identical conditions. The morphology and elemental composition of the calcinated nanopowder was investigated using a Hitachi S-3700N scanning electron microscope (SEM), equipped with an energy-dispersive X-ray detector (EDX) (Ametek EDAX, www.nature.com/scientificreports www.nature.com/scientificreports/ Octane Pro). In addition, high-resolution SEM investigations were conducted using a FEI Nova NanoSEM 230. For the analysis by transmission electron microscopy (TEM; JEOL JEM-2200FS operated at accelerating voltage of 200 kV), the synthesized particles were dispersed in ethanol and transferred onto a gold grid (300 mesh, TED PELLA, INC). ATR-FTIR analysis of the different precursor samples were performed with a Cary 640 FTIR spectrometer (Agilent). The diamond ATR accessory with a type IIa synthetic diamond crystal has a penetration depth of ~2 µm. The spectra were recorded in a frequency range of 4000-600 cm −1 with a spectral resolution of 2 cm −1 . A total of 128 scans were co-added for every spectrum. The background was measured with the same settings against air. The spectrometer was controlled by Agilent Resolutions Pro software 5.2.0. DLS experiments were conducted to determine the size distribution of the CuO NP aggregates. To this end, the synthesized nanopowder was dispersed in a water-based solution with 0.5% sodium dodecylsulfat (SDS) as a detergent and ultra-sonicated (UP200ST with VialTweeter, hielscher) to disrupt the weakly linked agglomerates. The DLS measurements were performed with a Zetasizer Nano ZS (Malvern Instruments Ltd., Malvern, UK) equipped with a max 4 mW He-Ne laser (emitting at 633 nm). Each measurement was performed at the non-invasive backscatter angle (NIBS) of 173° at a temperature of 25 °C and was preceded by a 30 s equilibration time. The ultrasonic treatment procedure was optimized to yield a stable minimal particle size after successive DLS measurements of the dispersion (thus ensuring near-complete disruption of the agglomerates). The specific SSA of the synthesized CuO nanopowders after calcination, as well as after ultrasonic treatment and subsequent drying in air, was determined from a 5-point N 2 adsorption BET isotherm, as measured with a Beckman-Coulter SA3100 instrument (Switzerland). Before the BET analysis, the powder samples were dried for 2 h at 180 °C in synthetic air.</p><!><p>Transformation of the precursor phase into CuO by calcination. The nature of the freshly-washed blueish-greenish precipitate collected for [CO 3</p><!><p>]/[Cu 2+ ] = 1.0 at pH = 5.83 (step 5 → 6 in Fig. 1) to CuO was first investigated by XRD (Fig. 2). The measurement indicates that the precipitated gel is XRD-amorphous, see Fig. 2a. Only a very broad intensity hump characteristic of an amorphous phase was detected in the 2-theta region of the expected principal (10-2) reflection of crystalline malachite, but no indications of crystalline CuO were found. To confirm the presence of a malachite-like precursor phase, the obtained blue-greenish precipitate was slowly dried to completion over 48 h at room temperature (RT) in air (note: for the standard synthesis route, the precipitate is dried at an elevated temperature for much shorter times, as described above). The XRD pattern after drying indeed matched the crystalline structure of malachite, which crystallizes in a monoclinic space group P21/a1 with lattice parameter a = 9.502 Å, b = 11.974 Å, c = 3.24 Å, alpha = 90.00°, beta = 98.75° and gamma = 90.00°3 2 : see Fig. 2b. These findings suggest that the amorphous precursor phase is chemically and structurally similar to crystalline malachite. Subsequent calcination of the thus-obtained crystalline malachite phase (as obtained by slow drying at RT in air), as well as of the amorphous precursor phase (obtained using the default synthesis route), both lead to the formation of CuO: see Fig. 2c and Eq. ( 3). CuO crystalizes in the monoclinic space group C12/c1 with lattice parameters a = 4.6837(5) Å, b = 3.4226(5) Å, c = 5.1288(6) Å, alpha = 90.000°, beta = 99.54(1)° and gamma = 90.000°3 3 .</p><p>In a next step, the calcination of the amorphous precursor gel, as obtained from the standard synthesis procedure was studied by in-situ HT-XRD: see Figs 3 and S1. To this end, a drying step was performed first for 12 h at 75 °C (i.e. 15 K higher than the default drying step), after which the temperature was step-wise increased with a rate of 1 °C/min to 250 °C, followed by an isothermal annealing step of 8 h at 250 °C (see Fig. 3a). In parallel, the reflections of the 2Θ region from 20-40°, which contains the principle reflections of crystalline malachite, Cu(OH) 2 , 4) and the Williamson-Hall formula (W-H) (5), i.e.</p><p>where β is the full width at half maximum of the respective reflection at the Bragg angle 2Θ, K is a numerical factor (here: K = 0.9, which is a good approximation for spherical particles in the absence of detailed shape information [34][35][36] ), ε is a factor for the strain induced peak broadening and λ is the incident X-ray wavelength (here Cu-Kα) 37,38 .</p><p>As evidenced by Fig. 3c, the average size of the CuO nanocrystallites synthesized at pH = 5.8 initially increases rapidly with increasing annealing time at T = 250 °C, reaching an average final size of roughly 17 ± 2 nm (by D-S) and 17.4 ± 0.4 nm (by W-H) within a few hours. Here it is emphasized that the XRD analysis probes the average size of the coherent scattering domains and therefore presents a measure of the smallest CuO crystalline building block only (and not of the CuO NP aggregate size). Notably, the Debye-Scherrer analysis on the basis of the CuO(111) and CuO(11-1) reflections gives similar crystallite sizes, which indicates that roughly equiaxed CuO crystallites are formed (i.e. single crystallites with a platelet or needle morphology are of minor importance).</p><p>After complete calcination, only reflections that match the reference pattern of CuO remain (see Fig. 2c), which suggests that the amorphous precursor phase is fully transformed into single-phase CuO within several hours of calcination at T ≥ 250 °C, in accordance with ref. 30 .</p><p>Differential scanning calorimetry (DSC) coupled with thermal gravimetric analysis (TGA) was carried out in synthetic air (60 mL/min) and at a heating rate of 5 K/min to 400 °C as shown in Fig. S2. An exothermic effect was detected at ~190 °C with a subsequent endothermic effect at ~260 °C, associated with a pronounced weight loss. The exothermic effect can be attributed to the crystallization of the amorphous precursor phase, as formed for the default synthesis route at pH = 5.8. The following endothermic effect is caused by the subsequent thermal decomposition of the crystallized precursor phase to CuO. The relative weight loss of 29%, as detected with TGA during the thermal decomposition, is in good agreement with the theoretical value of malachite decomposition: i.e. 1 − (2 × M CuO )/M malachite , where M CuO and M malachite correspond to the molar masses of CuO and malachite, www.nature.com/scientificreports www.nature.com/scientificreports/ respectively. This corroborates the findings from XRD that the thermal decomposition of the precursor gel into CuO (in static air) becomes thermally activated at around 250 °C. A somewhat lower transformation temperature of 180 °C was reported in ref. 39 , which might be due to a slightly different nature of the precipitate phase and/or the much longer annealing time of 48 h.</p><p>Noteworthy, DSC-TGA shows a clear crystallization peak upon heating whereas no distinct crystalline malachite reflections could be detected during heating by in-situ XRD (see Figs 3 and S1). This can be attributed to the overlap of the crystalline malachite peaks in XRD with the broad peak of the amorphous as-prepared gel (cf. Fig. 2), which hinders unique identification of the malachite phase as long as the amorphous gel is still present. ]/[Cu 2+ ] < 1 for pH < 5.83) or surplus (i.e. [CO 3</p><!><p>2− ]/[Cu 2+ ] > 1 for pH > 5.83), respectively. To provide a better insight into the precursor composition, the chemical speciation of Cu in aqueous solution, as well as the thermodynamic equilibrium with possible solid precipitate species (i.e. malachite, azurite, copper carbonate, copper hydroxide and copper oxide), were assessed as a function of pH for different carbonate concentrations using the MEDUSA software and the respective equilibrium and solubility constants from the HYDRA database 40 . The predictions were performed for various ammonia carbonate and copper acetate starting concentrations, as used in the experiments. The calculation results are plotted in Fig. 4a-c 4a-c,). Therefore, a second set of calculations was performed, which only considered the species remaining in solution (orange colored bars in Fig. 4a-c). The respective XRD patterns which were experimentally collected for the precipitate gels after 2 h of stirring are plotted for comparison in Fig. 4d-f.</p><p>For a carbonate deficiency of the mixed solution (Fig. 4a, pH = 5.6), [Cu 2+ •(OAc) − ] + , [Cu 2+ •2(OAc) − ] (with OAc = CH 3 COO) and [Cu 2+ •(HCO 3 ) − ] + are the dominant Cu complexes in solution. Due to the low solubility of malachite, the equilibrium is largely shifted towards the solid malachite phase (compare calculation with and without precipitation species in Fig. 4). According to the thermodynamic equilibrium assessment, Cu can to some extend also precipitate as CuO. Indeed, as predicted both malachite and CuO are being detected by XRD in the (freshly washed) precipitated gel, as collected after the default stirring duration of 2 h (see step 4 → 5 in Fig. 1). The CuO signal appears to be much more dominant than that of malachite in the recorded diffractograms, ]/[Cu 2+ ] concentration ratios and pH's of (a-c) are shown in (d-f) respectively including reference pattern for CuO (red) 33 and malachite (blue) 32 . (2019) 9:11758 | https://doi.org/10.1038/s41598-019-48020-8 www.nature.com/scientificreports www.nature.com/scientificreports/ which would implicate that the formation of malachite is suppressed compared to CuO under carbonate-deficient conditions, in contrasts to the thermodynamic assessment. However, the crystalline reflections of malachite and CuO are superimposed on a very broad intensity hump indicating the presence of an amorphous (malachite-like) precipitate phase. In this case, the summed-up signal intensity from crystalline malachite plus the amorphous malachite-like precursor clearly dominates over that of CuO, in accordance with the model predictions. For the [CO 3</p><!><p>]/[Cu 2+ ] ratio equal to 1.0, which corresponds to a slightly higher pH of 5.8, similar calculation results are obtained, although the CuCO 3 -complexes get slightly more dominant and the equilibrium further shifts to malachite, being by far the most dominant precipitating species (note: the predicted fraction of CuO becomes negligible, Fig. 4b). The freshly washed precipitate collected for the 1:1 ratio mixture is fully XRD-amorphous (Fig. 4e), but transforms into crystalline upon complete drying in air (see Fig. 2b). Finally, the calculations for a carbonate surplus of [CO 3</p><p>]/[Cu 2+ ] = 2.5 (pH = 7.0) predict CuCO 3 as the most dominant solution species and malachite as the only precipitation species; the acetate-containing copper complexes have only a minor contribution and the [Cu 2+ •(HCO 3 ) − ] + contribution is increased (Fig. 4c). In this case, the XRD analysis of the freshly-washed precipitate indeed only shows crystalline malachite, as well as the amorphous copper-carbonate-hydroxide (malachite-like) precursor phase (Fig. 4f).</p><p>The three different wet precursors with [CO 3 2− ]/[Cu 2+ ] = 0.4, 1.0 and 2.5 were also analyzed by ATR-FT-IR to disclose the differences in their molecular compositions. The recorded ATR-FT-IR spectra are shown Fig. S3 in the Supplementary Information. In all cases, vibration bands at 3311 cm −1 , 2400-1900 cm −1 and 1637 cm −1 , corresponding to vibrations of the H 2 O solvent were identified. The FTIR spectra do not show any characteristic vibration bands for copper acetate 41 or acetate ions 42 , which indicates that acetate is not contributing to the precursor phase and is removed during the washing process. The vibration detected at 880 cm −1 for [CO 3</p><p>2− ]/ [Cu 2+ ] = 1.0 and 2.5 corresponds to the non-planar rocking of bonded CO 3 2− , which does not appear for the sample with carbonate deficiency (i.e. [CO 3</p><p>]/[Cu 2+ ] = 0.4) 43 . Also the absorption bands for the symmetric C-O stretching at 1045 cm −1 and 1085 cm −1 appear less intense for the sample with carbonate deficiency. The largest differences in FTIR spectra for the three synthesis conditions are detected in the region of 1300-1530 cm −1 , in which asymmetric C-O stretching and CO 2 stretching vibrations of carbonate species appear 43 . The precursor obtained from the carbonate surplus reaction shows an absorption at ~1510 cm −1 , corresponding to asymmetric C-O stretching in basic conditions of complexated carbonates (fully deprotonated CO 3 2− ) 43 and at ~1400 cm −1 (CO 2 stretching in carbonates). Its small shoulder at 1420 cm −1 coincides with the broad absorption of the [CO 3</p><p>2− ]/[Cu 2+ ] = 0.4 sample. In refs 43 and 44 it was shown that the carbonate vibrations in acidic environments and in bicarbonate compounds shift as compared to their carbonato counterparts. The asymmetric C-O stretching vibrations in acidic environments shift towards 1620-1660 cm −1 and are thus covered by the absorption band of the solvent. The CO 2 stretching vibrations in HCO 3</p><p>− compounds appear at ~1410 cm −1 and ~1475 cm −1 . Hence the FTIR analysis reflects a dominant contribution of HCO 3 − species (1410 cm −1 and 1475 cm −1 ) and only a minor contribution of CO 3 2− species (visible at 1045 cm −1 ) at a carbonate deficiency, in agreement with the theoretical assessment of the solution species. Accordingly, the CO , as expected for [CO 3</p><p>2− ]/[Cu 2+ ] = 2.5 and 0.4, can be identified. The FTIR analysis thus indicates that the amorphous precursor phase, as formed for the default synthesis route at pH = 5.8, is predominantly constituted of copper carbonate and bicarbonate species and can thus be designated as an "amorphous malachite-like precursor phase". In conclusion, the experimental findings in combination with the thermodynamic calculations indicate that the observed amorphous phase is composed of randomly-packed clusters and chains of the predicted copper-complexes with HCO 3 − , CO 3 2− and H 2 O ligands (see Fig. 4). For the stoichiometric ratio of [CO 3</p><p>]/[Cu 2+ ] = 1.0, this amorphous precursor phase is the dominant product phase of the complex forming reaction and preferably convert into crystalline malachite upon slow drying and/or subsequent calcination. For deviations from the ideal ratio (and its corresponding pH-value), the crystallization of malachite from the amorphous precursor phase seems to be accelerated, but also the crystallization of CuO can then be observed.</p><!><p>]/[Cu 2+ ] -ratio and precipitation time on the morphology and size of the calcined CuO nanopowder. In order to tailor the crystallite and agglomerate sizes of the CuO nanopowder, the influence of the solution concentrations and resulting pH on the size and shape of the synthesized CuO NP aggregates as function of precipitation time was investigated in more detail. Since the hierarchical structures of the Cu precursor phase are generally largely preserved upon thermal decomposition into CuO by calcination 22 , the effect of the calcination treatment on the synthesized CuO product was not specially considered in the present study. Therefore, all calcinations were performed for 4 h at 400 °C.</p><p>Calcination of the precipitate collected after different stirring times of 1 min, 1 h, 2 h and 72 h leads to a clear difference in shape and compactness of the primary CuO crystallites and aggregates for the two [CO 3</p><p>2− ]/ [Cu 2+ ]-ratios considered, as becomes apparent from the SEM analysis of the CuO nanopowder (Fig. 5). For a very short stirring time of 1 min, the aggregated NPs and agglomerates produced at both pH values are predominantly constituted of large crystallites with needle-and platelet-like morphologies. This anisotropic crystallite shape found at the onset of mixing hints at the presence of an intermediate precursor phase which differs from both the above discussed copper-containing amorphous malachite-like precursor phase and crystalline malachite, which can be found after 2 h of stirring. A likely explanation is the formation of a precipitating phase similar to the amorphous malachite-like precursor phase, but with an increased hydroxide content, which is typical for the formation of needle-and platelet-shaped Cu-hydroxides 30,45 . Indeed, it may be assumed that the complexation of Cu 2+ by (larger and less mobile) carbonate CO 3 2− and bicarbonate HCO 3 − ions, as associated with the nucleation and growth of a malachite precursor phase, is relatively sluggish as compared to complexation of Cu 2+ by OH − . (2019) 9:11758 | https://doi.org/10.1038/s41598-019-48020-8 www.nature.com/scientificreports www.nature.com/scientificreports/ Upon further stirring, this intermediate precursor phase gradually dissolves again, as the more sluggish complexation of Cu 2+ by the larger carbonate CO 3 2− and bicarbonate HCO 3 − ions progresses, leading to the formation of the amorphous malachite-like precursor phase. As follows from the structural analysis of malachite in ref. 24 , malachite itself shows not only selective bonding along [001] crystallographic directions by the interconnection of Cu(OH) 2 building blocks, but also produces strong Cu-O with the carbonate ions in the (001) planes 24 . Assuming a similar local ordering for the amorphous malachite-like precursor, the progressive complexation of Cu 2+ by carbonate CO 3 2− ions 46 then results in the observed change in the crystallite shape of the CuO NPs from anisotropic to equiaxed morphology with increasing stirring time: see Fig. 5. As follows from a comparison of Fig. 5c with 5g, the (agglomerated) CuO aggregates synthesized for stirring times of 2 h appear much less densely clustered for pH = 7.0 as compared to pH = 5.6. For longer stirring times, the degree of NP aggregation increases for both pH-values, hinting at increasing densification of the amorphous precursor phase. Hence, a stirring time of 2 h can be considered as the optimum aging time for production of nano-porous nanopowders.</p><p>The TEM micrographs of the calcinated CuO nanopowders obtained at pH = 5.6 and 7.0 for the optimum stirring time of 2 h are shown in Fig. 6. Deposition of the nanopowder dispersions on the electron-transparent TEM grid leads to a mixture of CuO aggregates and their agglomerates, which hinders a determination of the true CuO NP aggregate size (which is obtained by DSL in the present study). However, the TEM analysis clearly evidences that the agglomerated aggregates are constituted of clusters of much smaller primary crystallites. As is evident from comparison of Fig. 6c,d, the average size of the primary crystallites increases with increasing pH. The primary crystallite size of 23 ± 6 nm for pH = 5.6, as determined by TEM (see Fig. 6a), complies well with the CuO crystallite sizes of 20 ± 6 nm (from D-S) and 23 ± 9 nm (from W-H), as determined by XRD. Notably similar crystallite sizes were obtained in the in-situ HT XRD study of the calcination process (for pH = 5.8) as discussed above. For pH = 5.6, CuO particles with considerably larger crystallite and aggregate sizes could be observed by TEM in very few occasions. These much larger particles may be attributed to the observed formation of fewer primary CuO crystallites during the synthesis step. Investigation of the obtained particles from pH = 7.0 show a larger average crystallite size of 38 ± 8 nm which also complies reasonably well with the corresponding XRD values of 25 ± 7 nm (for D-S) and 28 ± 13 nm (for W-H). Notably, for the synthesis at pH = 7.0 the shape of the primary crystallites appears smoother and more spherical, and the resulting (agglomerated) CuO aggregates are more loosely packed, corroborating the observations made by SEM: compare Fig. 6a-d. A possible explanation for the observed difference in primary crystalline as well as in aggregate size and morphology between pH = 5.6 and pH = 7.0 is the following. For the pH = 5.6 solution, a [CO 3</p><p>2− ]/[Cu 2+ ] = 0.4 ratio induces a deficiency of CO 3 2− ions with respect to Cu 2+ ions (see FTIR analysis). It may be assumed that upon gel formation, i.e. condensation of the Cu-complexions towards the nominal malachite composition, a much denser network of Cu 2+ and CO 3 2− is developed than in case of an equimolar ratio of Cu 2+ and CO 3 2− , i.e. more Cu-carbonate-Cu-bridges are formed. As a consequence the formed gel is much more compact, and calcination of the gel then leads to larger primary CuO crystallites with higher cluster density, i.e. denser aggregate morphology.</p><p>The observed difference in primary crystalline and aggregate size between pH = 5.6 and pH = 7.0 can be rationalized as follows: upon mixing of the starting solutions, the concentrations of the competing Cu complexation species practically instantaneous reach critical supersaturation, which for the ideal case of a homogeneous mixed solution under thermodynamic equilibrium, would result in the instantaneous homogeneous nucleation of the most stable precursor phase (i.e. crystalline malachite). These initial nuclei can grow upon aging of the ]/[Cu 2+ ] = 2.5 and pH = 7.0 (compare Fig. 4a,c). The CuO aggregates (as obtained after calcination), as synthesized from a carbonate deficient solution at pH = 5.6, are much more compact than the ones synthesized in a carbonate enriched solution at pH = 7.0 (compare Fig. 6d,f, as well as Fig. 5a,c,d). Here we propose the following explanation for the much higher compactness of the CuO aggregates synthesized at pH = 5.6. The precipitating amorphous precursor phase consists of randomly-packed clusters and chains of the predicted copper-complexes with HCO 3 − and CO 3 2− ligands (see Fig. 4), resulting in an amorphous precursor phase. The lower the [CO 3</p><p>2− ]/[Cu 2+ ]-ratio, the closer the proximity between neighboring copper cores in the precipitated amorphous precursor gel, which upon calcination result in a more dense CuO aggregates. Finally it is noted that, according to LaMer's-Model, larger primary crystallites are formed at a lower supersaturation (i.e. pH = 5.6), which is not observed in this study: the primary crystallites formed at pH = 5.6 are smaller than those grown at pH = 7.0 (as evidenced by XRD and TEM). In this regard it is emphasized that a variation of the carbonate molarity not only affects the supersaturation, but also the pH of the solution and thereby the stability (i.e. solubility) of the solid precursor phase in the mixed solution as well as its composition (thus affecting its final size resulting from Oswald ripening during continuous stirring). Both the higher carbonate molarity and the higher stability of the malachite phase promote diffusion-limited growth of the primary malachite nuclei at pH = 7.0.</p><p>Determination of the CuO aggregate size by DLS. DLS was applied to determine the average size of the CuO aggregates, as dispersed in a water-based solution with 0.5% sodium dodecyl sulfate (SDS) following an ultrasound treatment. In this regard, it is emphasized that DLS records the intensity of the scattered light at very high temporal resolution from which the hydrodynamic diameter is calculated, which is the diameter of the particle or aggregate plus any ligands, ions or molecules that are attached to their surface (here: SDS). The ultrasonic Determination of the specific surface area (SSA) of the CuO nanopowder by BET analysis. The specific surface area (SSAs) of the synthesized CuO nanopowders after calcination, as well as after ultrasonic treatment and subsequent drying in air, were determined by BET analysis. The compact CuO aggregates synthesized at pH = 5.6 have an SSA of 16 m²/g, which can be compared with typical SSA values in the range of 10-15 m²/g, as indicated for commercially available high-purity (i.e. ≥99.99%) CuO nanopowders (e.g. US research Nanomaterials © , Plasmachem © , GetNanoMaterials © , Nanografi © and Nanoshel © ). In a next step, the same CuO nanopowder (i.e. synthesized at pH = 5.6) was dispersed in a water-based solution with 0.5% SDS using ultrasound waves and dried in air. Strikingly, a similar SSA of 15.33 m²/g was determined after the ultrasonic treatment and subsequent drying. This indicates that the ultrasonic treatment of the nanopowder dispersion is not affecting the effective surface area. In other words, although the more weakly linked agglomerates will be dispersed during the ultrasonic treatment, this has no distinct effect on the effective available surface area of compact aggregates.</p><p>The loosely-clustered CuO aggregates synthesized at pH = 7.0 (with an average size of 93 nm; see Fig. 7b) have an SSA of 18.73 m²/g before and 18.92 m²/g after the ultrasonic treatment, which is about 20% larger than the SSA of the compact CuO aggregates (with an average size of 135 nm; see Fig. 7a). The maximum theoretical SSA that can be achieved for spherical nanoparticles is plotted as function of the particle diameter in Fig. 8. The SSA's of the compact and nanoporous CuO aggregates, as synthesized in this study are compared with the indicated SSAs for commercially available CuO nanopowders. It follows that the SSA of ≈19 m²/g, as obtained for the nanoporous CuO NP aggregates formed from pH 7 solution in the present study, is not only significantly higher than for most commercial CuO nanopowders, but also much closer to the theoretical value of ≈25 m²/g for the respective primary crystallite size (by TEM) of 38 ± 8 nm. Strikingly, the SSA for the compact CuO aggregates with a smaller primary crystallite size of 23 ± 6 nm, as synthesized at pH = 5.6, falls far below the respective theoretical SSA. Although a smaller primary crystallite size should result in a higher SSA (see Fig. 8), this was not observed in the present study. It is thus concluded that the BET analysis effectively probes the enhanced nanoporosity of the loosely-clustered CuO aggregates synthesized at pH = 7.0: i.e. the nanoporosity of the CuO aggregates at pH = 7.0 is much better accessible (permeable) to the infiltration by a gas (or liquid) than in case of the denser aggregates for pH = 5.6.</p><p>These findings underline the importance of tuning the cluster density of the primary crystallites in stable NP aggregates (and not solely the primary crystallite size) for achieving a high SSA and thereby an enhanced chemical reactivity (often the main property aimed at with the use of nanoparticles) and sinterability of the nanopowder when processed in the form of a dispersed solution, paste or nanocomposite. First trials of low temperature joining with the less dense CuO NP aggregates processed as a nanopaste have confirmed that the sintering temperature and time for the bonding process can both be effectively lower by tuning the SSA of the CuO NP aggregates (work in progress).</p><!><p>The cost-effective and green sol-gel synthesis of CuO nanopowders via thermal decomposition of an amorphous malachite-like precursor phase was successfully implemented to tune size, shape and cluster density of the primary crystallites in the CuO NP aggregates. In-situ heating XRD measurements showed that the transformation of the amorphous malachite-like precursor phase into single-phase CuO upon heating in air becomes thermally activated at T ≥ 250 °C. The resulting CuO nanopowder is composed of (agglomerated) CuO aggregates, which are constituted of clusters of much smaller primary CuO crystallites. The size, shape and nanoporosity (i.e. aggregate cluster density) of these primary CuO crystallites can be tuned by controlled adjustment of the pH and the degree of supersaturation of the mixed solution with respect to the nucleation of malachite. To this end, the mixed solution was regulated to a specific pH value of either 5.6 or 7.0 by adjusting the added volume of ammonia carbonate solution, which resulted in an carbonate deficiency (i.e. [CO 3</p><p>2− ]/[Cu 2+ ] < 1 for pH < 5.83) or surplus (i.e. [CO 3</p><!><p>]/[Cu 2+ ] > 1 for pH > 5.83), respectively. This resulted in an average CuO aggregate size of 135 nm ± 43 nm at pH = 5.6 and of 93 nm ± 40 nm at pH = 7.0. The loosely clustered (and thus nanoporous) CuO aggregates synthesized at pH = 7.0 have a specific surface area of 18.73 m²/g, which is about 20% larger than the SSA of 15.97 m²/g for the compact CuO aggregates, despite a smaller NP crystallite size of the latter structure. It follows that the cluster density of the primary crystallites in the synthesized CuO NP aggregates can be tailored to enhance the specific surface area of the resulting CuO nanopowder for targeted applications in the field of e.g. catalysis, nanojoining, energy conversion and energy storage.</p><!><p>The datasets generated and analysed during the current study are available from the corresponding author upon reasonable request.</p>

Scientific Reports - Nature

A Universal Stamping Method of Graphene Transfer for Conducting Flexible and Transparent Polymers

transfer method of chemically vapor deposition graphene is an appealing issue to realize its application as flexible and transparent electrodes. A universal stamping method to transfer as grown graphene from copper onto different flexible and transparent polymers (FTPs) reported here ensures simple, robust, rapid, clean and low-cost. This method relies on coating ethylene vinyl acetate (EVA) onto the as grown graphene, binding EVA coated graphene/Cu with FTPs and delamination by hydrogen bubbling process, which is analogous to the method used by stamping process where ink carries the imprint of the object onto any materials. the fate of the stamping method depends on how strongly the adhesion of EVA coated graphene/Cu with target FTPs. Interestingly, we have found that the thin film of EVA/ graphene/Cu can only bind strongly with the FTPs of less than 25 µm in thickness and lower glass transition temperature value to the EVA while wide range of other FTPs are considered upon surface engineering to enhance the binding strength between FTPs and EVA. What's more, the electrical performance was investigated with a demonstration of triboelectric nanogenerators which confirmed the reliability of graphene transfer onto the FTPs and prospect for the development of flexible and transparent electronics.

a_universal_stamping_method_of_graphene_transfer_for_conducting_flexible_and_transparent_polymers

<!>Characterization.<!>Results<!>Conclusion

<p>Since the discovery of mechanically exfoliated graphene 1 , production of high quality chemically vapor deposition (CVD) graphene for the industrial scale has remained challenging [2][3][4] . Large domain size graphene grown on Cu may explore the possibilities of realistic application upon transfer onto the dielectric substrates [5][6][7][8][9] . Graphene on polymer substrates are especially appealing as replacement of Indium tin oxide (ITO) and also one of the essential flexible and transparent electrodes for a wide range of optoelectronics devices such as touch screen displays and solar cells [10][11][12] . Polymethyl methacrylate (PMMA) polymer is used to transfer graphene from Cu metal substrate onto the dielectric substrates 13,14 , but only limited to study for the fundamental properties. Till date, as grown CVD graphene was transferred onto many different types of flexible polymers [15][16][17][18][19][20][21] . In majority of transfer case, PMMA and thermal releasing polymers are used as intermediate polymer to transfer graphene onto different polymers which scarifies the Cu substrates and induces cracks 19,22 . Roll to roll transfer of graphene onto polyethylene ptherapthalate (PET) was also achieved via thermal release tape 10 , epoxy resin 15 and ethylene vinyl acetate(EVA) 23 as binding source between graphene and PET. Roll to roll graphene transferred onto the EVA/PET by bubbling method 24 and green transfer 23 is hopeful to achieve the low cost, light weight, flexible and transparent electrodes. To our best of knowledge there is no report found on graphene transfer onto many other polymers, except PET, using EVA as binding agent for the industrial scale. EVA is explored against PMMA mediated graphene transfer for conformal contact with the target substrates 25,26 . Graphene transfer onto different polymers of characteristics properties suits wider range of applications still remains challenging and not many efforts have been made in this direction 27 . Recent advances in flexible electronics expect to soon realize the industrial production of graphene based hybrid transparent electrodes 28 . Our study met the challenge to transfer graphene onto different polymers by a fast and clean process.</p><p>In this work, we designed a stamping method to transfer as grown graphene on Cu onto different flexible and transparent polymers (FTPs) which is applicable to the large scale graphene requirement as flexible and transparent electrode in modern electronics. This innovative method (Figs 1a and S1) relies on coating EVA onto the as grown graphene on Cu, binding EVA coated graphene/Cu with different polymers and delamination by hydrogen bubbling process, which is analogous to the method used by stamping process where ink carries the imprint of the object onto any materials. EVA has characteristic properties such as excellent transparency, flexibility and adhesivity as a function of temperature, which is already being used as encapsulating layers in solar panels 29 , is the major role in this transfer process. Next generation of wearable/bendable electronics demands the potential supply of flexible and transparent polymers as transparent electrode substrates in the areas of energy conversion, environmental monitoring, healthcare and communication and wireless network 30 . Although either big or small differences in polymer properties such as transmittance, flexibility, oxygen and water permeability, temperature resistance may largely impact on the functional properties desired for the respective applications. We chose thermoplastic transparent and flexible polymers, but not limited to, such as polyethylene terephthalate (PET), polyimide(PI), polycarbonate(PC), polyvinyl chloride(PVC), TOPAS (thermoplastic polymer mr-I T85) and CYTOP (an amorphous fluoropolymer type: CTL-809M, was purchased from AGC Asahi Glass) as target substrates for graphene based conductive film which could open the new avenue either in the modern electronics. The fate of the stamping method depends on how strongly the adhesion of EVA coated graphene/Cu with target FTPs. Interestingly, we have found that the thin film of EVA coated on graphene/Cu can only bind strongly with the FTPs of less than 25 µm in thickness and lower glass transition temperature (Tg) value to the EVA without any pretreatment (unmodified-FTPs) while wide range of other FTPs such as higher or lower Tg to that of EVA with low thermal conductivity in thicker polymer substrates are considered upon surface engineering (surface engineered-FTPs) to enhance the binding strength between FTPs and EVA. What's more, the electrical performance was investigated with triboelectric nanogenerators (TENG) [31][32][33][34] , which confirms that our transfer process is reliable for the different polymers and prospect for the development of flexible and transparent electronics. Graphene transfer by stamping method. After CVD grown graphene on copper, an EVA solution (1-4 wt% dissolved in cyclohexane, Aladdin Industrial Corporation, Shanghai) was coated onto the graphene side of the copper by spin/spray/blade coating process. Prior to coating process, Cu was flattened using glass rod; edges of the graphene/Cu were covered with scotch tape to avoid the coating of EVA onto the backside of the graphene/ Cu. Spin coating is done with two spinning steps; 200 rpm for 30 seconds and 400 rpm for 60 seconds. Two times coating was done to ensure the continuity of the thin film EVA. In spray coating process, EVA solution was kept at hot container to make solution free flow through the nozzle. The spray nozzle movement and pressure were adjusted to 20 mm/s and 0.2 Mpa, respectively. We also demonstrated blade coating process, which is most suitable for the large scale graphene transfer attributed to roll to roll process. Speed of the blade movement was adjusted to 10 mm/s while distance between the blade and sample to be coated was set to 0.5 mm. The temperature of the sample holder both in spray and blade coating were kept at 50 °C to make fast evaporation of the solvent. After complete evaporation of solvent after being kept in dry box for 60 minutes, the resultant EVA/graphene/ Cu was bound onto any of the flexible polymers (cleaned using isopropanol and blow-dried with a nitrogen gun) using hot lamination method to form FTP/EVA/graphene/Cu. Finally, graphene/EVA/FTPs is delaminated from the metal substrate by electrochemical hydrogen bubbling method 21,33 and dried in air.</p><!><p>The morphology was characterized by optical microscopy (Olympus BX51), scanning electron microscopy (ZEISS-Merlin) and atomic force microscopy (Bruker). Quality of the CVD grown graphene transferred onto Si/SiO 2 was evaluated using a Raman spectroscope (Horiba, LabRAM HR Evolution) with a laser excitation wavelength of 532 nm. Monochromatic Al X-ray (Physical Electronics 56000 multitechnique system) was used to analyze any metal residue remained after graphene being transferred onto FTPs. The transmittance measurement was examined using UV-Vis-NIR spectrophotometer (Perkin Elmer, Lambda 750 s, 190-3300 nm). Contact angle measurement was done by contact angle tester (AST VCA Optima XE). Sheet resistance measurement was carried out using four probe system (Guangzhou 4-probe Tech Co. Ltd., RTS-4) with probe spacings of ~1 mm.</p><p>Fabrication of triboelectric nanogenerator. The final product from this transfer process graphene/EVA/ FTPs film was used as the electrode of TENG. The fabricated device consists of a graphene/EVA/FTP electrode coated with CYTOP film (triboelectric layer) (18 × 20 mm) and a PET substrate coated with ITO films. Both two components were placed in an acrylic glass as a proof mass which can be driven by external vibration. The acrylic mass and transparent substrate were assembled in the elastic holder to make an arched structure with a dimension of 45 × 45 × 10 mm. The CYTOP polymer coated on the bottom plate is charged in a custom-built corona charging setup. The setup consists of a grounded electrode, a metal mesh grid (V g = −2000 V) and a high-voltage probe tip (V H = −6 kV). The device is driven by a mechanical shaker with controlled frequency and amplitude, where an accelerometer is used to monitor the acceleration during the measurement. The shaker is driven by an excitation signal generated from a signal generator (Brüel&Kjaer, LAN-XI 3160) and a power amplifier (Brüel&Kjaer, 2719).</p><!><p>Monolayer graphene is grown on electrochemically polished 25 µm thick copper foils by low-pressure chemical vapor deposition (LPCVD) 2 . Both optical microscopy (OM) (Fig. S2a) and scanning electron microscope (SEM) (Fig. S2b) of graphene transferred onto SiO 2 by PMMA revealed continuous monolayer graphene with bilayer or multilayer graphene domains. Figure S2c shows the Raman spectroscopy of the graphene on SiO 2 confirms the high quality graphene by showing the negligible D peak which is raised commonly due to the PMMA mediated transfer. Figure 1a shows the schematic of stamping method to transfer CVD-grown graphene from Cu substrate to a FTPs comprises three major steps; (i) EVA is coated onto the as grown graphene on Cu by any of the three modes such as spin coating for small sample, spray coating and blade coating for large samples (Fig. S1a). The coated sample represents as ethylene vinyl acetate/graphene/copper (EVA/graphene/Cu) thin film. Figure 1b shows the photograph which distinguishes the graphene on Cu before and after EVA coating, where bright contrast of Cu/graphene becomes dull upon EVA coating which revealed the EVA film formation. A thin close contact of EVA upon blade coating is evidenced with the vertical cross-sectional SEM (Fig. 1c). (ii) Binding the EVA/ graphene/Cu thin film with the target FTP substrate using hot lamination (Fig. S1b) to form FTP/EVA/graphene/ Cu stack as similar to the graphene transferred onto commercially available EVA/PET 23,24 . Hot lamination in our graphene transfer method is reliable for the flexible polymers as coated EVA layer on Cu/graphene unchanged the flexible property. Roller temperature was adjusted according to the Tg of EVA. Higher temperature than Tg of EVA makes the FTPs film deform. We found some interesting mechanisms of binding EVA coated graphene/Cu thin film with the smooth surface morphology FTPs (less than 25 µm) having similar Tg of EVA such as TOPAS, CYTOP, PET. 25 µm thin FTPs with lower Tg undergoes easy deformation at 120 °C, since as grown graphene on Cu has considerable roughness 35 , smoother FTPs adhered tighter with EVA/graphene/Cu due to roughness impression carried from graphene/Cu. But in the case of thicker FTPs of 100 µm such as PET, PC, PVC, we cannot www.nature.com/scientificreports www.nature.com/scientificreports/ found tight adhesion with EVA/graphene/Cu due to insufficient supply of required temperature to attain Tg value for thicker FTPs in lamination process. (iii) Delamination of FTP/EVA/graphene/Cu stacks to transfer graphene onto FTP/EVA from Cu substrate using electrochemical hydrogen bubble method by polarizing Cu/graphene/ FTP at 2 V 24,36,37 (Fig. S1c). Figure 1d shows the photographs of the graphene/EVA/FTP (bottom row) which are comparable to that of EVA/FTP (top row), indicating that transfer process has little effect on the transmittance of target FTP upon graphene transfer. Note that all FTPs are resistant to the NaOH (electrolytic solution) and the method ensures successful graphene transfer without damaging either Cu substrates or EVA layer coated on graphene/Cu. Since FTP/EVA/graphene/Cu stack was used as cathode in the delamination process, H 2 bubble generates between graphene and the Cu substrates which leads to the delamination of graphene onto the EVA/ FTPs. Oxidation of graphene/EVA/FTPs is avoided while Cu undergoes slight oxidation due to the basic electrolytic solution but it favors the growth of high quality graphene for the second time 23 . The graphene/FTPs film is rinsed with deionized water and blow dried with nitrogen to ensure it to be free of chemical residues.</p><p>Prior to the coating EVA solution onto the graphene on Cu, the method further comprises the step of flattening the CVD-grown graphene on Cu foil (Fig. S3a). Spin coating is popular and commonly used to transfer graphene using PMMA solution because it helps to control the thickness of the thin film formation of PMMA layer 13 . The innovation of our work relies on the thin film deposition of EVA onto as grown graphene on Cu and surface engineering of FTPs, where EVA acts as a binding agent between graphene and target substrates. In the process of coating, EVA solution preparation and optimization process is very important because we are the first using EVA to transfer graphene on to different FTPs. The preparation method is as follows: 1-4 wt.% EVA solution is made by dissolving EVA and stirring in cyclohexane at 75 °C. In fact, we demonstrated the coating method based on the sample size; spin coating is generally suitable for small size graphene since it is easier and makes the process fast. A thin layer of 1% EVA was spin coated twice to ensure a continuous layer, parameters were set similar to the PMMA coating 38 . On the other hand, for large area of graphene samples especially for roll to roll CVD graphene, either spray coating or blade coating process enables the process efficient and easier. Since the diameter of spray nozzle is small, 1% EVA is used in the spray coating method to avoid the blockage of nozzle. It should be noted that spray coating can control the tight contact of EVA with graphene upon adjusting the spray pressure to form the compact thin uniform EVA film. Multi spray coating deposition can also be done to control the thickness, which is attributed to achieve graphene transfer onto surface engineered polymers. To obtain thicker film on large area graphene, 4% EVA was coated by blade coating machine. By adjusting the distance between Cu/ graphene and blade, thickness can be controlled up to 20 µm. Detailed procedure of coating method is described in experimental part. The samples after deposition were kept at room temperature for 60 minutes to evaporate the solvent. By comparing the vertical cross-sectional SEM images of Cu/graphene (Fig. S3d) and EVA coated Cu/graphene (Fig. S3e), the continuity of EVA layer without any disruption was confirmed. Note that there is no reaction found between EVA solution and graphene/Cu metal substrate, though it helps the formation of thin film very fast due to fast evaporation of the solvent. Both blade coating and spray coating can be integrated in the roll-to-roll processing of graphene transfer. Interestingly, we found our EVA mediated stamping method of graphene transfer onto smooth FTPs can only be realized when their thickness and Tg is less than 25 µm and ~140 °C, respectively. Weak binding force between the two smooth surface polymers films were noticed in the fabrication of graphene/metal nanowire transparent electrodes 39 , which signalling that the surface engineering of the target FTPs are deciding factors to clamp thin film EVA to achieve efficient graphene transfer in our stamping method. It is also reported that EVA film shows excellent adhesive bonding to solar glass which are rough in surface 40 . In addition, our method can also extend to FTPs thicker than 25 µm or with higher Tg temperature than EVA after proper surface engineering. The basic prerequisite of the target FTPs substrate should be surface roughness and hydrophobicity. Surface engineering in our method comprises two steps; (i) FTPs are and blasted to make the surface rough with the formation of crest and trough in large scale. (ii) Fill the crest and trough by coating 1% EVA solution using various methods. Generally, graphene transfer is only valid onto smooth surface materials for electronic applications 41 . To transfer graphene onto a wide variety of FTPs having higher Tg value to that of EVA and low thermal conductivity in thicker polymer substrates, surface engineering is inevitable. EVA mediated transfer overcome the challenge to transfer graphene onto rough surface by providing smooth surface basement to graphene upon EVA coating onto the target rough surface polymers. Here, we used but not limited to PI as FTP target substrates to demonstrate graphene transfer onto the smooth surface with a higher Tg value to that of EVA. Graphene transfer from Cu/ graphene/EVA stack onto the PI is unsuccessful, where arrow in the Fig. 2a shows the EVA/graphene detachment from PI substrate. EVA/graphene thin film detachment from ultra-smooth PI substrate is clearly seen in SEM image (square mark of Fig. 2a), which signs that the graphene transfer can only be done upon surface engineering of target polymer substrates of higher Tg value and higher thickness (Fig. 2b).</p><p>Figure S4a shows the schematic illustration of FTPs surface modification, surface was thoroughly rubbed by sand paper, forming uneven surface with crest and trough and the same was made smooth by filling EVA solution by spin coating method. Crest and trough of rubbed FTPs were clamped tightly by the coated EVA which forms the flat surface, as shown in the vertical cross-sectional image of SEM (square mark of Fig. 2b). Further surface morphology characterizations of surface engineered-FTPs and graphene transfer onto the FTPs were carried out by OM and atomic force microscopy (AFM). OM shows that the crest and trough in FTPs ensures that the entire surface is uneven (Fig. 2c), while it turned to flat uniform surface upon EVA coating (Fig. 2d) and finally graphene transferred onto the EVA coated FTPs remains surface flat which is identified with the graphene grain boundary (arrow in Fig. 2e). Our AFM observations show that the FTPs of higher Tg rubbed with sandpaper bears the uneven surface (Fig. 2f; bottom) with random crest and trough of several micrometers (line section of the surface morphology; Fig. S5a), which helps to clamp the EVA very tight (Fig. 2b). Upon EVA coating, crest and trough were filled to form a thin uniform film (Fig. 2f; middle) to satisfy the surface requirement to transfer graphene, surface roughness was decreased to several nanometers (Fig. S5b). After graphene transferred onto the surface engineered-FTPs, surface roughness was found to be quite increased as compared to the EVA coated FTPs before transfer, which is due to the surface morphology of graphene grown on copper 16 (Fig. 2f; top and Fig. S5c). Contact angle (CA) measurement was carried out to measure to the surface behavior of surface engineered FTPs before and after graphene transfer. In fact, rough surface FTPs shows hydrophilicity due to the rough surface morphology (CA = 75°, Fig. 2c), while EVA coated FTPs before (CA = 105°, Fig. 2c) and after Gr transfer (CA = 97°, Fig. 2e) remains hydrophobic which strongly shows that the graphene transfer upon surface engineered polymer is successful by stamping method. In contrast, topographical AFM images shown in Fig. 2f distinguishes the surface roughness of FTPs rubbed with sand paper (bottom), smoother upon EVA coating (middle) and finally turns to be negligible rough after graphene transfer (top). It is noteworthy that graphene grown on Cu foil by CVD shows high quality (Fig. S2). Surface morphology of the graphene on different polymers by stamping method was characterized by SEM. The graphene grain boundaries (arrow mark in Fig. 3a-f) are observed as the characteristics confirmation of graphene transfer onto the different polymers such as TOPAS, CYTOP, PET, PVC, PC and PI. No voids or cracks are seen from our transfer, which confirms the good contact between the target substrates and the EVA/graphene transferred from the Cu substrates. Surface of graphene/EVA on polymers shows slight wavy morphology due to the surface engineered FTPs substrates and the graphene/Cu. In the hot lamination process, EVA softens at 120 °C and mimics the surface morphology of both the FTPs and Cu, which results in the rough graphene/EVA/FTPs surface compared to that of EVA/FTPs before graphene transfer, as shown in the AFM image of Fig. 2f (See Supplementary Fig. S5c). However, OM image of graphene/EVA/FTP shows the confirmation of graphene transfer in large area, where graphene grains on Cu can be compared with the graphene grains on EVA/FTPs (Fig. S6a). Following coating and binding process, electrochemical bubbling method results the final graphene on FTPs; we found no surface contamination on graphene/EVA/FTPs by XPS and UV-Visible spectra. Figure 3g shows the XPS full spectra of graphene/EVA/FTPs. Two predominant XPS peaks of C 1 s and O 2 s found at 284.5 eV and 530 eV, which are ascribed to the sp2 carbon of graphene and the oxygen in EVA, respectively. The deconvoluted XPS spectra of graphene/EVA/FTPs in inset of Fig. 3g shows no predominant Cu peak in the binding energy between 930 and 960 eV at the detection limit of XPS. This observation indicates that the coating-lamination-bubbling process in the stamping method is efficient to transfer graphene/EVA from Cu onto FTPs. We then evaluated the UV-Visible spectrum of graphene/EVA/FTPs to confirm whether the quality of transparency is affected in the transfer process. Fig. S6b shows the UV-Visible spectrum of graphene transferred onto EVA/FTP showing 97.4% transmittance close to the theoretical value of graphene/Quartz, which confirms that our transfer process is successful with no surface contamination in the entire transfer process; note that the substrate transmittance was subtracted. The transmittance of graphene/EVA on different FTPs such as TOPAS, CYTOP, PET, PC, PVC and PI were evaluated showing that original transmittances of the FTPs were unaffected upon graphene/EVA transfer. The total T% values were found for graphene on different FTPs such as TOPAS, CYTOP, PET, PC, PVC and PI are 96%, 96%, 86%, 84%, 83% and 55%, respectively (Fig. 3h). Conductivity is one of the main concerns of graphene transferred onto FTPs. Sheet resistance of graphene/EVA/FTPs was evaluated with four point probe system. The large area of graphene on different-FTPs was measured with sheet resistance and the result value lies between 1-10 kohm/sq, which is acceptable range of graphene on EVA/PET by green transfer 23 . Because of high sheet resistance of polycrystalline graphene on dielectric substrates, metal nanowires are used to fabricate www.nature.com/scientificreports www.nature.com/scientificreports/ graphene based hybrid transparent electrodes to perform equivalent to that of ITO electrodes 24,42 . Figure S6c shows the schematic illustration for the reason of higher Rs found at graphene/EVA/FTPs. Four probes shown in the Fig. S6c spotted at within the grain size or between the large grain size Rs value is considerable lower than that of four probes spotted at small grains or higher grain boundary region results in higher Rs of graphene/EVA/ FTPs. Either with the improved growth of large area domain size graphene on Cu 43 , metal nanowire networks 44 or chemical doping 45 are strong strategic direction that our transfer method is efficient to resolve the higher Rs value in graphene/FTPs. In contrast, graphene on unmodified-FTPs shows a Rs of 10 Kohm (Fig. 3i) while graphene on surface engineered-FTPs shows higher Rs of about 10-20 K Ohm (Fig. 3j), indicating the surface engineering process affects the electrical properties slightly. The distribution curve of sheet resistance of graphene on different unmodified FTPs is observed over the 3 × 4 cm graphene/EVA/FTPs is found to be 1-10 k ohm/sq, which are measured with the typical probe spacing 1 mm (Fig. S6d).</p><p>The Cu foil was not damaged by any process involved in stamping method (Fig. 1b), neither of chemical residue remained while EVA coating on graphene/Cu or tore Cu foil in the lamination and bubbling method. Vertical SEM images of Cu after graphene/EVA transferred onto FTPs show that there are no such mechanical damages found on Cu substrates (Fig. S3f). Figure 4a (left) shows the Cu foil after graphene/EVA thin film transferred onto the FTPs showing some part undergoes oxidation which is an advantage for the growth of high quality graphene 46 , while graphene grown upon reused Cu (right of Fig. 4a) turns to be shine which is similar to the Cu used to grow graphene at first time (Fig. 1b; left). The high-quality nature of graphene on reused Cu is evidenced by OM (Fig. 4b), SEM (Fig. 4c) and Raman spectrum (Fig. 4d). The enlarged domain size and less grain boundary may favor the quality improvement of graphene on FTPs, which is significant for economic industrial scale.</p><p>To confirm the graphene transfer performance on the electrical conductivity and flexibility, we have applied graphene/EVA/FTPs electrodes in the TENG. The fabricated device consists of a FTPs/EVA/graphene electrode coated with CYTOP film (18 × 20 mm2) and a PET substrate coated with ITO films sketched in Fig. 5a. The ITO plays dual roles as electrode and contact surface, while CYTOP plays the role as the other contact surface. Graphene is used as back electrode. Figure 5a (right) shows the photograph of the graphene/EVA/PI revealing the bending capability without causing any damage to graphene/EVA. To fabricate the device, cast acrylic glass was prepared as a proof mass which can be driven by external vibration. Both the acrylic mass and transparent substrate were assembled in the elastic holder to make an arched structure with a dimension of 45 × 45 × 10 mm 3 . The CYTOP polymer coated on the bottom plate is charged in a custom-built corona charging setup (Fig. 5b). The setup consists of a grounded electrode, a metal mesh grid (V g = −2000 V) and a high-voltage probe tip (V H = −5 kV). After charging for 15 min, the surface potential of the electrets layer is mapped in Fig. 5d. The energy harvesting performance of the device is characterized with a shaker setup shown in Fig. S7a. Our device is driven by a mechanical shaker with controlled frequency and amplitude, where an accelerometer is used to monitor the acceleration during the measurement. The shaker is driven by an excitation signal generated from a signal generator (Brüel&Kjaer, LAN-XI 3160) and a power amplifier (Brüel&Kjaer, 2719). Figure 5c shows a typical www.nature.com/scientificreports www.nature.com/scientificreports/ driven vibration with amplitude of 170 m/s 2 at 46 Hz. To find the optimal work frequency, we have explored the relationship between the output power of the TENG and the frequency of the vibration source under different amplitudes, as shown in Fig. S7b. An output power peak can be seen at 46 Hz, which is exactly the same as the resonant frequency of the device. At resonance, the output voltage of TENG based on FTPs/EVA/graphene/CYTOP VS ITO/PET was presented in Fig. 6a with a closed-up view shown in Fig. 6b. The voltage peak reaches 0.23 V. Output current was found with different values for the same area size of graphene on different FTPs shown in Fig. S8. Here we demonstrated graphene transfer onto the unmodified-FTPs, found output voltage was 0.008 V for graphene transferred onto the thin CYTOP, TOPAS and PET (Fig. S8a-c), while graphene on surface engineered FTPs such as thicker PET (Fig. S8d) and higher Tg value polymers such as PI (Fig. S8e) output voltage was found to be increased due to wavy surface morphology of graphene. CVD graphene surface roughness increases the triboelectric effect 47 . Furthermore, it is observed that the electric signal also can be generated by hand driving periodically; the output result of device FTP/EVA/graphene/CYTOP VS ITO/PET was shown in Fig. 6c. And it was proved that the electricity generated by our device is effective when a 10 μF capacitor is charged successfully from 0 V to 3 V in less than 20 min, as Fig. 6d presented. Meanwhile, we also proved the electricity generated by our device is powerful enough to power LEDs, as 6 LEDs can be powered successfully as showed in Fig. 6e,f. This demonstration showed that the binding energy between graphene/EVA and the FTPs are strong enough and as well materials could be used as flexible and transparent electrode in the modern electronics. Even though only five polymer examples are shown, the procedure worked consistently on all other target substrates such as PC and PVC which justifies that our graphene transfer is universal.</p><!><p>We have demonstrated a stamping method to transfer graphene onto different FTPs using EVA as binder only between graphene and target substrates without affecting the Cu substrates which could be used for repeated graphene growth. Surface modification of FTPs to alter the effective surface interaction with the EVA widens the choice for target substrates. Our transfer method is simple and fast, which ensures the clean and efficient transfer without inducing any damage either onto the graphene/substrates or Cu foil. What's more, the electrical output performance is demonstrated with the fabrication of TENG and implied that our transfer method is realistic to scale up the graphene on plastics for industrial-scale. We believe this approach may be further improved by adopting effective strategies like metal nanowire based graphene transparent and flexible electrode reported elsewhere to replace ITO in optoelectronic devices 24 .</p>

Scientific Reports - Nature

Room-Temperature Phosphorescent Co-Crystal Showing Direct White Light and Photo-Electric Conversion

The development of molecular crystalline materials with efficient room-temperature phosphorescence has been obtained much attention due to their fascinating photophysical properties and potential applications in the fields of data storage, bioimaging and photodynamic therapy. Herein, a new co-crystal complex [(DCPA) (AD)2] (DCPA = 9,10-di (4-carboxyphenyl)anthracene; AD = acridine) has been synthesized by a facile solvothermal process. Crystal structure analysis reveals that the co-crystal possesses orderly and alternant arrangement of DCPA donors and AD acceptors at molecular level. Fixed by strong hydrogen bonds, the DCPA molecule displays seriously twisty spatial conformation. Density functional theory (DFT) calculations show well separation of HOMO and LUMO for this co-crystal system, suggesting the efficient triplet excitons generation. Photoluminescence measurements show intensive cyan fluorescence (58.20 ns) and direct white phosphorescence (325 µs) emission at room-temperature. The transient current density–time curve reveals a typical switching electric response under the irradiation of simulated light, reveal that the [(DCPA) (AD)2] co-crystal has a high photoelectric response performance.

room-temperature_phosphorescent_co-crystal_showing_direct_white_light_and_photo-electric_conversion

Introduction<!>Materials and General Methods<!><!>Materials and General Methods<!>Synthesis of [(DCPA) (AD)2].<!>Crystal Structure Description<!><!>Crystal Structure Description<!>Powder X-ray Diffraction and Thermal Gravimetric Analysis<!><!>Photoluminescence Properties<!><!>Photoluminescence Properties<!>Density Functional Theory Calculations<!><!>Photo-Electronic Performance<!><!>Photo-Electronic Performance<!>Conclusion<!>Data Availability Statement<!>Author Contributions<!>Funding<!>Conflict of Interest<!>Publisher’s Note

<p>The rational design of molecular crystalline materials with long-lived room-temperature phosphorescence (RTP) has atracted tremendous attentions owing to their extended potential to create new opportunities in the development of photocatalytic reactions, photodynamic therapy, optical storage, organic light emitting diodes, and bioimaging (Bhattacharjee and Hirata, 2020; Jiang et al., 2018; Gao and Ma, 2021; Hirata, 2019; Gu et al., 2019; Lei et al., 2019; Wang et al., 2020; Wang et al., 2021; Li and Li, 2020; Yang et al., 2019; Yang et al., 2020; Gao et al., 2021; Chen et al., 2021). Besides the traditional noble-metal (ruthenium, platinum, iridium) based complexes (Xiang et al., 2013), breakthroughs have been achieved during the past decade on pure organics, polymers, metal–organic frameworks (MOFs), organic−inorganic hybrid perovskite and host–guest doping (Mu et al., 2017; Lu et al., 2019; Zhou and Yan, 2019; Lei et al., 2020; Liu et al., 2020; Lei et al., 2021; Wu et al., 2021; Zhou et al., 2021). Promising strategies (such as crystallization, H-aggregation, halogen bonding) have also been vastly accepted to obtain efficient RTP (Bolton et al., 2011; Gong et al., 2015; Kenry and Liu, 2019; Wang et al., 2020), and the inherent principle is absolutely focused on promoting triplet excitons generation.</p><p>Considering the spin-forbidden intersystem crossing (ISC) from excited singlet state to excited triplet state, the rate of ISC can be enhanced by reducing the energy gap (ΔE ST) between the lowest singlet excited state and a nearby triplet state. Small ΔE ST can be achieved by designing the charge transfer of donor-acceptor system with large spatial separation between the HOMO and LUMO (Parke and Rivard, 2018). To date, many single component organic molecules with twisted donor-acceptor spatial conformation have been demonstrated as efficient RTP materials (Xiao et al., 2020; Xu et al., 2021). However, triplet state excitons of multi-component co-crystal donor-acceptor systems are still relatively limited (Zhou and Yan, 2019; Zhou et al., 2020), and the systematical investigation is needed for well understanding the relationship between their structures and photophysical behaviors.</p><p>In this paper, one new type of co-crystal [(DCPA) (AD)2] has been obtained under solvothermal conditions by the selection of 9,10-di (4-carboxyphenyl)anthracene (DCPA) electron donor and acridine (AD) electron acceptor. The obtained donor-acceptor co-crystal system shows alternant arrangement of DCPA and AD components at the molecular level. The crystal structure and density functional theory (DFT) calculations reveal that the DCPA molecule fixed by strong hydrogen bonds displays seriously twisty spatial conformation. This structure feature affords well separation of HOMO-LUMO, promoting for the generation of triplet excitons. As a result, the formation of [(DCPA) (AD)2] co-crystal exhibits cyan fluorescence and direct white long-lived RTP under ambient condition.</p><!><p>9,10-di (4-carboxyphenyl)anthracene (DCPA), acridine (AD) and anhydrous ethanol were purchased commercially. Single-crystal X-ray diffraction data were collected by Oxford Diffraction SuperNova area-detector diffractometer with the program of CrysAlisPro. The crystal structure was solved by SHELXS-2014 and SHELXL-2014 software (Sheldrick 2008). The crystallographic data for [(DCPA) (AD)2] were listed in Table 1. The CIF file (CCDC No. 2104581) presented in this study can be downloaded free of charge via http://www.ccdc.cam.ac.uk/conts/retrieving.html.</p><!><p>Crystallographic data for [(DCPA) (AD)2].</p><!><p>Phase purity of co-crystal powders were tested by Bruker D8-ADVANCE X-ray diffractometer with Cu Kα radiation. Elemental analysis was performed by Perkin–Elmer Elementarvario elemental analysis instrument. Fourier transform infrared (FT-IR) spectra were measured by SHIMADZU IR Spirit-T spectrometer from 4,000 to 400 cm−1 with KBr pellet. UV-vis absorption spectra were detected by Shimadzu UV-3600 plus UV-vis-NIR spectrophotometer. Thermo gravimetric analysis (TGA) experiments were measured by SII EXSTAR6000 TG/DTA6300 thermal analyzer from room temperature to 800°C. The fluorescent and phosphorescent spectra were conducted on Edinburgh FLS1000 fluorescence spectrometer excited by xenon arc lamp (Xe900) and microsecond flash lamp, respectively. The time-resolved phosphorescent decay curves were measured by a microsecond flash lamp with a frequency of 100 Hz. Optoelectronic properties were tested on CHI 660 E electrochemical analyzer in a standard three-electrode system. The working electrode, counter electrode, reference electrode, and electrolyte is [(DCPA) (AD)2] powders modified indium tin oxide (ITO) glass, platinum wire, Ag/AgCl, and 0.5 M sodium sulfate aqueous solution, respectively. The linear sweep voltammetry (LSV) was recorded by the voltage rang of 0.2 to −1 V with a scan rate of 50 mV/s. The transient photocurrent were measured by on–off cycle's illumination of Xe lamp (300 W) with bias potential (vs Ag/AgCl) of 0 and −0.5 V.</p><!><p>A mixture of 9,10-di (4-carboxyphenyl)anthracene (0.1 mmol, 41.8 mg), acridine (0.2 mmol, 35.8 mg) and 8 ml EtOH was sealed into a Teflon reactor (23 ml), and heated at 120°C for 12 h. Then, the light yellow block crystals can be obtained after naturally cooled to room temperature. Anal. Calc (%) for C27H18NO2: C 83.48, H 4.67, N 3.61; found (%): C 83.12, H 4.33, N 3.46. IR (KBr pellet, cm−1): 3,415(w), 3,054(w), 1,947(w), 1,692(s), 1,607(m), 1,572(m), 1,524(m), 1,440(m), 1,401(m), 1,281(s), 1,100(m), 920(m), 853(w), 774(s), 735(s), 673(m), 505(m).</p><!><p>High-grade light yellow block single crystals of the two-component [(DCPA) (AD)2] co-crystal were synthesized under the solvothermal condition from the mixture of DCPA and AD (Figure 1A) with a 1:2 stoichiometry in ethanol solution. Single-crystal X-ray diffraction analysis reveals that [(DCPA) (AD)2] crystallizes in triclinic Pī space group, and the asymmetric unit consists of two AD and one DCPA molecules. In the co-crystal system, the DCPA molecules are linked together by C‒H···O hydrogen bonds (C6‒H6···O2: H6···O2 = 2.69 Å, ∠C6‒H6···O2 = 141.30°) to form a 1D chain (Figure 1B). Pairs of AD molecules arrange in a head-to-tail π-stacking mode with short interplanar distance of 3.66 Å, which extends the DCPA 1D chain into a 2D sheet with the alternant arrangement of DCPA and AD molecules (Figure 1C).</p><!><p>(A) Chemical structures of DCPA and AD molecules in this work. (B) 1D chain-like structure of DCPA extended by C‒H···O hydrogen bonds. (C) View of the 2D sheet constructed by the alternant arrangement of DCPA and AD molecules through C‒H···O and O‒H···N hydrogen bonds. (D) Torsion angles between benzene acid arm and the anthracene core.</p><!><p>Owing to above mentioned hydrogen bond interactions, the DCPA chromophores are highly fixed in an ordered arrangement at the molecular level, which exhibits a seriously twist conformation with torsion angles between benzene acid arm and the anthracene core up to 84.5° (Figure 1D). These supramolecular interactions also provide rigid environment to restrict the molecular motions/vibrations, minimizing the nonradiative loss of single/triplet excitons and facilitate for efficient emission (Yang et al., 2020).</p><!><p>Powder X-ray diffraction (PXRD) experiment was conducted to detect the phase purity of [(DCPA) (AD)2] co-crystal (Figure 2A). The experiment diffraction peaks match well with the simulated one, providing the high purity and good crystalline degree of the as-synthesized samples. Thermo gravimetric (Figure 2B) curve shows the first weight loss of about 44.50% in the range of 200–283°C, assigning to the loss of AD molecules (calculated: 46.14%). Additional heating results in the gradual decomposition of framework of co-crystal.</p><!><p>(A) PXRD patterns of simulated (black) and as synthesized DCPA-AD (red). (B) Thermo gravimetric analysis curve of DCPA-AD.</p><!><p>The steady-state, transient-state photoluminescence (PL) spectra and time-resolved PL decay curves of both [(DCPA) (AD)2] co-crystal, pure DCPA and AD in solid state were recorded at room temperture. Figure 3A illustrates the fluorescence spectra of DCPA in solid state, which shows strong dark-blue emission owing to the presence of the anthracene chromophore (λ ex = 329 nm). The fluorescence decay curve estimated at the maximal emission peak at 447 nm gives rise to a short lifetime of 1.01 ns(Figure 3B), whereas the single component AD has an emission peak at 396 nm and lifetime of 2.88 ns(Yang et al., 2020). By contrast, the formation of co-crystal presents a red-shift of the emission peak to long wavelength at 474 nm, attaching with a weak shoulder at about 443 nm when excited at 365 nm (Figure 3C), suggesting the charge transfer interaction between the DCPA donor and AD acceptor. Apart from the emission peak, the [(DCPA) (AD)2] co-crystal also shows much longer fluorescence lifetime up to 58.20 ns(Figure 3D), which is more than 50 times as long as that of free DCPA molecules in solid state.</p><!><p>Fluorescence spectra (A) decay curve (B) of DCPA in solid state (λex = 329 nm). Fluorescence spectra (C) decay curve (D) of [(DCPA) (AD)2] co-crystal in solid state (λex = 365 nm). Insert shows solid state samples under UV (365 nm) light radiation. Phosphorescence spectra (E) decay curve (F) of [(DCPA) (AD)2] co-crystal in solid state (λex = 365 nm). Insert shows CIE-1931 chromaticity diagram (0.33,0.34) of [(DCPA) (AD)2] co-crystal with white phosphorescence emission. All of these measurements were recorded under ambient condition.</p><!><p>The delayed PL spectrum shows a broad emission region spanning nearly the whole visible spectra with a maximum peak at 570 nm (Figure 3E). The time-resolved PL decay curve affords a long lifetime of 325 µs, indicating long-lived RTP emission of [(DCPA) (AD)2] co-crystal (Figure 3F). The inserts show the cyan emission of [(DCPA) (AD)2] crystalline powders irradiated under UV (365 nm) and the CIE-1931 chromaticity coordinate obtained from the phosphorescence spectra. The chromaticity coordinate of (0.33,0.34) is close to the optimum white-light with value of (0.33,0.33). The above results indicate that the fromation of co-crystal can largely tune the fluorescence emission of DCPA from dark-blue to cyan, and prolong the lifetime more than 50 times. In our opinion, the strong supramolecular interactions efficiently reduce the nonradiative loss of single/triplet excitons, and further enable prolonged PL lifetime.</p><!><p>Density functional theory (DFT) calculations were conducted by Dmol3 module in Material Studio software package (Delley 2000) based on the X-ray single crystal diffraction data of [(DCPA) (AD)2]. The results show the highest occupied molecular orbital (HOMO) is occupied by the anthracene core of DCPA molecules, whereas the lowest unoccupied molecular orbital (LUMO) is exclusively located on the benzene acid groups. The LUMO+1 mainly appears on AD molecules (Figure 4). Herein, the seriously twist conformation of DCPA molecule leads to large spatial separation of the HOMO and LUMO. The alternant arrangement of DCPA electron donor and AD electron acceptor further promotes the separation of molecular orbitals, boosting the spin-orbit coupling and intersystem crossing for efficient triplet state exciton generation.</p><!><p>The structure mode and selected molecular orbitals of [(DCPA) (AD)2].</p><!><p>It has been found that the long-lived triplet state excitons have more chance for the electron migration (Yang et al., 2019). Encouraged by the long-lived RTP of [(DCPA) (AD)2] co-crystal in this work, its photo-electronic properties have been further conducted by a three-electrode system in Na2SO4 aqueous solution. The UV-Vis absorption spectrum shows an optical band gap of 2.63 eV (471 nm), consisting with the fluorescence emission peak (Figure 5A). The linear sweep voltammetry (LSV) curve reveals that the [(DCPA) (AD)2] co-crystal material can generate large current with the addition of negative potential (Figure 5B). The absence of redox peak suggests that the [(DCPA) (AD)2] co-crystal is stable within the applied bias potential from 0.2 to −1 V.</p><!><p>(A) UV-VIS-NIR absorption of as synthesized [(DCPA) (AD)2]. (B) The linear sweep voltammetry curve of as synthesized [(DCPA) (AD)2] modified ITO electrode measured in 0.5 M Na2SO4 aqueous solution. Transient current density–time curve of [(DCPA) (AD)2] at bias potential of 0 V (C) and −0.5 V (D) with the periodic on-off cycles of light radiation.</p><!><p>The transient current density–time curve reveals a typical on/off switching response under the irradiation of simulated light. Without only bias potential, it generates high photocurrent up to −3.1 μA cm−2 with the momentary light radiation. Under the initiatory dark condition, extremely small dark current of about 0.002 μA cm−2 can be detected (Figure 5C). The rate of current between light radiation and dark conditions was calculated up to 1,550. By the addition of −0.5 V bias potential, it generates more large current of about −46.5 μA cm−2 under light radiation (Figure 5D). All these results reveal that the [(DCPA) (AD)2] co-crystal has superior photoelectric response performance, which can be applied in the future photoelectric detector device.</p><!><p>In summary, we report a rare example of direct white-light RTP co-crystal material [(DCPA) (AD)2], which can be synthesized under a facile solvothermal condition. The framework of [(DCPA) (AD)2] shows an orderly distribution of heterojuction at the molecular level: alternant arrangement of DCPA electronic donors and AD electron acceptors bonded together through strong C‒H···O and O‒H···N hydrogen bonds. Fixed by these supramolecular interactions, the molecular motions/vibrations can be restricted, which affords long-lasting singlet and triplet excitons through minimize the nonradiative loss. In addition, the seriously twist conformation of DCPA molecule is beneficial to the separation of molecule orbitals. Combined with the introduction of AD acceptor, it provides efficient platform for long distance exciton transfer and good electron-hole separation ability, possessing superior photoelectric response performance. Therefore, this work not only develops a new type of white-light RTP co-crystal, but also provides a perspective to deeply understand the relationship among molecular structure, stacking mode and photoelectric performance.</p><!><p>The data presented in the study are deposited in the (Cambridge Crystallographic Data Centre) repository, accession number (2104581).</p><!><p>XY, LF, and D conceived the idea and designed research. WJ, JR, XK, and X synthesized and characterized materials; all authors analyzed data and wrote the paper.</p><!><p>This work was supported by the National Natural Science Foundation of China (No. 21971100, 21771021, 21822501, and 22061130206), Project of Central Plains Science and Technology Innovation Leading Talents of Henan Province (No. 204200510001), Project for Science and Technology Innovation Talents in Universities of Henan Province (No. 21HASTIT006), and Key Scientific Research Projects of Higher Education of Henan Province (No. 20A150005).</p><!><p>The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.</p><!><p>All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors, and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.</p>

PubMed Open Access

WISB: Warwick Integrative Synthetic Biology Centre

Synthetic biology promises to create high-impact solutions to challenges in the areas of biotechnology, human/animal health, the environment, energy, materials and food security. Equally, synthetic biologists create tools and strategies that have the potential to help us answer important fundamental questions in biology. Warwick Integrative Synthetic Biology (WISB) pursues both of these mutually complementary ‘build to apply’ and ‘build to understand’ approaches. This is reflected in our research structure, in which a core theme on predictive biosystems engineering develops underpinning understanding as well as next-generation experimental/theoretical tools, and these are then incorporated into three applied themes in which we engineer biosynthetic pathways, microbial communities and microbial effector systems in plants. WISB takes a comprehensive approach to training, education and outreach. For example, WISB is a partner in the EPSRC/BBSRC-funded U.K. Doctoral Training Centre in synthetic biology, we have developed a new undergraduate module in the subject, and we have established five WISB Research Career Development Fellowships to support young group leaders. Research in Ethical, Legal and Societal Aspects (ELSA) of synthetic biology is embedded in our centre activities. WISB has been highly proactive in building an international research and training network that includes partners in Barcelona, Boston, Copenhagen, Madrid, Marburg, São Paulo, Tartu and Valencia.

wisb:_warwick_integrative_synthetic_biology_centre

WISB: personnel and activities<!>Predictive biosystems engineering<!>Engineering biosynthetic pathways<!>Engineering microbial communities<!>Engineering microbial effector systems in plants<!>Training and education in synthetic biology<!>Ethical, Legal and Societal Aspects of synthetic biology<!>National and international partnerships<!>WISB facilities<!>

<p>The core academic community comprises academics from the Departments of Life Sciences, Chemistry, Computer Science, Engineering and Education Studies. In addition, WISB comprises Associate Members from the University of Warwick, Honorary Members from other institutions, RCDFs, postdoctoral associates and PhD students and also trains and supports an iGEM team each year. More information can be found on our website: http://www.wisb-uow.co.uk.</p><!><p>One of the most important goals for synthetic biology is to make system design having predictable consequences in terms of biological function. It is currently the case that assembling multiple (genetic) components together frequently generates unpredicted or suboptimal behaviour because we know too little about inter-component interactions or about circuitry–host interactions. Since this is not a satisfactory platform on which to base the future development of the field, we are using experimental and computational methods to develop new and improved types of biomolecular circuitry for both prokaryotic and eukaryotic hosts, and to develop advanced CAD tools to make this task faster, cheaper and easier. Rather than relying on the standard types of transcriptional part, we are constructing new types of post-transcriptional component and circuitry that rely, for example on control at the levels of mRNA translation and protein modification or that facilitate computation directly using DNA strand displacement. We are also seeking ways to improve the robustness and scalability of synthetic circuitry by optimizing system stochasticity and minimizing the negative effects of host–circuit interactions.</p><!><p>Mankind has exploited microorganisms as sources of valuable metabolic products for millennia, but attempts to engineer biosynthetic pathways to create novel metabolic diversity are often hampered by insufficient understanding of underlying molecular mechanistic principles and can produce unpredictable outcomes. We aim to apply synthetic biology principles to develop a better understanding of biosynthetic pathways, which will enable more predictable pathway engineering. Computational design tools and yeast-mediated recombination are being applied to build refactored biosynthetic gene clusters for structurally complex metabolites with applications in medicine (e.g. antibiotics), agriculture (e.g. herbicides) and bioenergy. In parallel, orthogonal transcriptional and translational regulatory elements are being engineered with a view to fine-tuning expression from (refactored) biosynthetic gene clusters.</p><!><p>Microbial species have generally evolved in the context of communities. There are many ways in which naturally evolved or engineered, microbial communities can benefit mankind, but this has been a relatively poorly explored aspect of synthetic biology. At WISB, we are pursuing a number of complementary approaches to advancing the understanding and benefits of engineered microbial communities. We are engineering synthetic microbial communities from the bottom up using (multiple) defined species that are functionally interlinked using metabolic, genetic and physical interactions, and reengineering existing communities using rational manipulations at the species and community level. Examples of the former category are projects on (i) three-species systems involving a phototroph that convert light into high-value chemicals in a closed ecosystem and (ii) a spatially organized community designed to perform environmental bioremediation. An example of the latter category is work on reengineering gut communities towards reducing production of TMAO, a by-product of a high-meat diet that has been shown to increase the risk of atherosclerosis.</p><!><p>Ensuring food security for the world's population is one of the greatest challenges faced by mankind. WISB researchers are developing new types of synthetic control system that use synthetic effectors (SynEffectors), derived from natural effectors of plant pathogens and mutualists. Natural microbial effectors are targeted to bespoke pathways within plants in order to engineer (orthogonal) temporal and spatial control of plant responses, thus paving the way for the development of plants with enhanced resistance to stress and microbial attack. A key part of this work is the identification of effectors in pathogenic and mutualistic microbes that control developmental (senescence), immunity and abiotic stress (drought, high-light) pathways. Particular emphasis is given to mechanisms that can uncouple growth and immunity–pathways that typically act antagonistically. Synthetic versions of these effectors (SynEffectors) will be engineered into new plant regulatory systems.</p><!><p>WISB is committed to delivering a comprehensive and coherent programme of excellent education and training in synthetic biology in collaboration with academic and industrial partners. This starts with dedicated teaching elements (including a third-year module) at the undergraduate level that provides the platform for students to pursue the Master's degree (MBio) in synthetic biology in their fourth year of study. Each year, WISB academics train and support a team that competes in the international Genetically Engineered Machine (iGEM) competition. WISB is a partner in the U.K.'s only Centre of Doctoral Training in Synthetic Biology. We also provide extensive skills training for postdoctoral researchers working in synthetic biology and support a cohort of WISB Research Career Development Fellows (RCDFs). All WISB members benefit from bespoke training courses on transferable skills and on enterprise and impact. Moreover, the WISB community pursues an extensive and diverse programme of outreach and public engagement.</p><!><p>We believe that WISB can most effectively explore, and where appropriate implement principles related to Ethical, Legal and Societal Aspects (ELSA) of synthetic biology in the context of collaboration and partnership with other centres, both national and international. WISB is accordingly a member of the U.K. National Synthetic Biology network. We are also working together with the Department of Media, Cognition and Communication at the University of Copenhagen to develop new research and to deliver training courses, on science and society.</p><!><p>WISB is building strong partnerships in research and training with academic institutions across the globe. We collaborate in multiple areas with colleagues in the other U.K. Synthetic Biology Research Centres (SBRCs). In addition, our international partners already include the Biomass Systems and Synthetic Biology Centre (BSSB), São Paulo, Brazil; Boston University BioDesign Center; Experimental and Health Sciences, Universitat Pompeu Fabra, Barcelona, Spain; Löwe Centre for Synthetic Microbiology (SYNMIKRO), Philipps-Universität Marburg, Germany; SynBio Centre, Tartu University, Estonia; the Synthetic Biology Centre at the University of Copenhagen. This network is expected to be mutually beneficial for all partners, allowing us all to enhance the breadth and scale of our research and training activities. WISB is also collaborating with a range of companies, currently including Abolis, Algenuity, BASF, Biopolis, Evolva, Green Biologics, Ingenza, Isomerase, Microsoft Research, Syngenta and Tata Steel.</p><!><p>With funding from the BBSRC and EPSRC, WISB has established a state-of-the-art Research Technology Facility designed to support a diversity of synthetic biology projects, core staff to help manage its activities and a large cohort of postdoctoral researchers and PhD students. The instrumentation/technologies in the RTF include wide-field and spinning disk confocal microscopy, flow cytometry, cell sorting, microfluidics, surface plasmon resonance, LC–MS, ion chromatography, HPLC, 3D printers, micro- and macro-scale bioreactors, chemostats, gas analyser and robotics.</p><!><p>WISB (http://www.wisb-uow.co.uk)</p><p>Synthetic Biology CDT (www.warwick.ac.uk/synbiocdt)</p><p>Synthetic Biology UK 2015: Held at Kingsway Hall Hotel, London, U.K., 1–3 September 2015</p><p>Biotechnology and Biological Sciences Research Council</p><p>computer-aided design</p><p>Ethical, Legal and Societal Aspects</p><p>Engineering and Physical Sciences Research Council</p><p>international Genetically Engineered Machine</p><p>Research Career Development Fellow</p><p>responsible research and innovation</p><p>Research Technology Facility</p><p>trimethylamine N-oxide</p><p>Warwick Integrative Synthetic Biology</p>

PubMed Open Access

Fluorescent substrates for flow cytometric evaluation of efflux inhibition in ABCB1, ABCC1, and ABCG2 transporters

ATP binding cassette (ABC) transmembrane efflux pumps such as P-glycoprotein (ABCB1), multidrug resistance protein 1 (ABCC1), and breast cancer resistance protein (ABCG2) play an important role in anti-cancer drug resistance. A large number of structurally and functionally diverse compounds act as substrates or modulators of these pumps. In vitro assessment of the affinity of drug candidates for multidrug resistance proteins is central to predict in vivo pharmacokinetics and drug\xe2\x80\x93drug interactions. The objective of this study was to identify and characterize new substrates for these transporters. As part of a collaborative project with Life Technologies, 102 fluorescent probes were investigated in a flow cytometric screen of ABC transporters. The primary screen compared substrate efflux activity in parental cell lines with their corresponding highly expressing resistant counterparts. The fluorescent compound library included a range of excitation/emission profiles and required dual laser excitation as well as multiple fluorescence detection channels. A total of 31 substrates with active efflux in one or more pumps and practical fluorescence response ranges were identified and tested for interaction with eight known inhibitors. This screening approach provides an efficient tool for identification and characterization of new fluorescent substrates for ABCB1, ABCC1, and ABCG2.

fluorescent_substrates_for_flow_cytometric_evaluation_of_efflux_inhibition_in_abcb1,_abcc1,_and_abcg

<!>Reagents and instrumentation<!>ABCB1, ABCC1, and ABCG2 transporter-expressing cell lines<!>Uptake and efflux assay<!>Inhibition assay<!>Uptake and efflux of fluorescent dyes in parental and transporter-expressing cell lines<!>Substrate efflux inhibition<!>Discussion

<p>The transmembrane ATP binding cassette (ABC)2 efflux pumps ABCB1 (P-glycoprotein, P-gp), ABCC1 (multidrug resistance protein 1, MRP1), and ABCG2 (breast cancer resistance protein, BCRP) play an important role in the development of resistance against anticancer drugs [1,2]. To date, more than a dozen ABC transporter pumps have been observed to efflux chemotherapeutic agents in vitro [3]. ABCB1, ABCC1, and ABCG2, in particular, are highly expressed in the gut, liver, and kidneys, and they may restrict the oral bioavailability of administered drugs. ABCB1 and ABCG2 are also expressed in the epithelia of the brain and placenta as well as in stem cells, where they perform a barrier function [4]. The role played by ABC transporter pumps in protecting tissues from xenobiotics is now widely recognized, but their interplay, their relationship with other enzymes, and how they affect the disposition, distribution, and effect of individual drugs remain an active area of investigation.</p><p>Structural information for mammalian ABC transporter family members is relatively sparse, with ABCB1 being the most extensively studied. Recent investigations indicate that at least four distinct drug binding sites exist on ABCB1, which can be classified as both transport and modulation sites. At 3.8 Å resolution, the X-ray structure of mouse apo ABCB1a displays a 6000-Å3 cavity and two ATP-binding domains separated by approximately 30 Å. The apo and drug-bound ABCB1a structures show portals open to the cytoplasm and the inner leaflet of the lipid bilayer for drug entry as well as the ability to accommodate large and small substrates or even two substrates simultaneously [5]. Together, these facts can account for broad or even poly-specificity for unrelated chemical structures.</p><p>In addition, the substrate binding cavity can be formally partitioned into an upper portion with mostly hydrophobic and aromatic interactions, and a lower space containing polar interactions (with overlap in the middle). Binding of a substrate to one of the sites may induce conformational changes to adjacent binding site(s), which in turn alters experimental affinities [6]. The drug binding pocket of ABCG2 may function in a similar manner to that of ABCB1, with radioligand binding studies suggesting two or more symmetric substrate binding sites with overlapping specificity [7]. Drug–drug interactions resulting from transporter inhibition present a clinical concern [8,9]. The presence of multiple binding sites and interactions between them may account for diverse specificity of structurally and functionally unrelated modulators and substrates. Multiple binding site interactions also raise questions as to which substrate should be used to demonstrate inhibitory potential of a new chemical probe.</p><p>To understand the mechanism of action and to design more effective modulators, efforts have been made to study the interaction of substrates and modulators with these transporters [10]. For example, most ABCB1 inhibitors are also substrates of the efflux pump [11]. It is valuable not only to assess inhibitor potency for a given transporter but also to profile its activity with respect to other transporters as well as its interrelationship with substrate drugs. For instance, strong inhibition of ABCB1 by drugs such as cyclosporine and verapamil in vitro was of limited value in vivo due to toxic pharmacological effects of the inhibitors [1].</p><p>We previously reported a new platform for identification of substrates and inhibitors for three human ABC transporters using two fluorescent probes: J-aggregate-forming lipophilic cation 5,5′,6,6′-tetrachloro-1,1′,3,3′-tetraethylbenzimidazolcarbocyanine iodide (JC-1, T3168) and calcein acetoxymethyl ester (CaAM, C1430) as substrates [12]. We demonstrated differential activity of inhibitors for ABCB1, ABCC1, and ABCG2 transporters, and we noted cross-reactivity of both these substrates across the three transporters, which could help to explain such severe toxicity effects. Recently, we used a similar duplex system with T3168 to identify an ABCG2 inhibitor with high selectivity as compared with ABCB1 [13]. In the current study, we applied this high-throughput flow cytometric assay system to evaluate 102 fluorescent compounds for their substrate properties and their interactions with several known inhibitors for these transporters.</p><!><p>The collection of potential fluorescent substrates was obtained from Life Technologies (Eugene, OR, USA). The efflux pump inhibitors loxapine succinate, nicardipine hydrochloride, niclosamide, novobiocin sodium, pimozide, and verapamil hydrochloride were purchased from Sigma–Aldrich (St. Louis, MO, USA). Lasalocid sodium was purchased from Prestwick Chemical (Illkirch, France), and mometasone furoate was purchased from U.S. Pharmacopeia (Rockville, MD, USA). Unless otherwise indicated, all compound solutions were maintained and diluted in dimethyl sulfoxide (DMSO) prior to addition to assay wells. Final DMSO concentrations were no more than 1% (v/v). A Biomek NX Multichannel (Beckman Coulter, Fullerton, CA, USA) was used for all cell and compound solution transfers for volumes greater than 1 μl. Low-volume transfers (100 nl) were performed via pintool (V&P Scientific, San Diego, CA, USA). Compound dose–response plates were generated with the Biomek NX Span-8 (Beckman Coulter, Fullerton) or the Eppendorf epMotion 5070 (Westbury, NY, USA). The HyperCyt high-throughput flow cytometry platform (IntelliCyt, Albuquerque, NM, USA) was used to sequentially sample cells from 384-well microplates (2 μl/sample) for delivery to the flow cytometer at a rate of 40 samples per minute [14,15]. Flow cytometric analysis was performed with a CyAn flow cytometer (Beckman Coulter, Fort Collins, CO, USA). To cover the fluorescence range of all substrates, each sample was (i) excited at 488 nm and detected with 530/40 (FL 1), 575/25 (FL 2), 613/20 (FL 4), and 680/30 (FL 5) optical bandpass filters and (ii) excited at 635 nm and detected with 665/20 (FL 8) and 750 LP (FL 9) optical bandpass filters. The resulting time-separated data files were analyzed with HyperView software (IntelliCyt) to determine compound activity in each well. Inhibition–response curves were fitted by Prism software (Graph-Pad Software, San Diego, CA, USA) using nonlinear least-squares regression in a sigmoidal dose–response model with variable slope, also known as the four-parameter logistic equation. Analysis of time-separated flow cytometric data was detailed in a previous ABC transporter screen from our group [12].</p><!><p>The ABCB1-overexpressing drug-resistant cell line, CCRF-ADR 5000, and its parental CCRF-CEM cells were kindly provided by T. Efferth (Pharmaceutical Biology, German Cancer Research Center, Heidelberg, Germany). We have developed and previously characterized a SupT1-vincristine (Vin) drug-resistant cell line that selectively overexpresses ABCC1 [16]. Ovarian Ig-MXP3 (ABCG2) and its parental Igrov1-sensitive cells were kindly provided by D. Ross (Department of Medicine, University of Maryland Greenebaum Cancer Center, Baltimore, MD, USA). Cells were grown in RPMI-1640 medium supplemented with 10% fetal bovine serum (Hy-clone, Logan, UT, USA), 2 mM L-glutamine, 10 mM Hepes, 10 U/ml penicillin, 10 lg/ml streptomycin, and 4 μg/ml ciprofloxacin. To ensure ABCB1 up-regulation, CCRF-ADR 5000 cell lines were grown in 20 nM daunorubicin. To up-regulate ABCC1 expression, the SupT1-Vin cell line was grown in the presence of 150 nM vincristine. Further population enrichment of the CCRF-ADR 5000 and SupT1-Vin cell lines was achieved via fluorescence cell sorting with 250 nM CaAM and 1 μM JC-1, respectively, on a MoFlo cell sorter (Beckman Coulter, Fort Collins). Up-regulation of ABCG2 pump expression in Ig-MXP3 cells was achieved via the addition of 340 nM mitoxantrone to the cell media 1 h prior to cell harvest. Flow cytometric screening was done in calcium- and magnesium-free phosphate-buffered saline (PBS, Mediatech, Manassas, VA, USA) containing 10% fetal bovine serum.</p><!><p>To identify potential fluorescent substrates in ABCB1-, ABCC1-, and ABCG2-overexpressing cell lines, substrates were serially diluted 1:3 eight times, affording in-well dose–response ranges from 15 pM to 100 nM. Table 1 shows the catalog number and name of each substrate along with the plating protocol type and the observed fluorescence channel. Depending on the amount of each substrate available, mother plates of the substrate stocks were generated and stored at 10, 1, and 0.1 mM. Regardless of starting stock concentration, daughter dose–response plates were generated with a top final concentration of 10 μM. The screen was done in 384-well plates at 100 μl per well with a maximum of 16 substrates tested per plate, 1 per row. To avoid potential carryover from highly fluorescent substrates, the last 12 wells of each row were filled with buffer only, providing multiple wash wells between the end of one dose–response series and the beginning of the next one. Substrate (1 μl) was added to plates containing 99 μl of the transporter-expressing cell lines and separately to their parental counterparts (1.5 × 106 cells/ml). After compound addition, the plates were vortexed briefly to mix the compounds with the cell suspension and then rotated end over end at room temperature to avoid cell settling. After a 20-min incubation, cellular fluorescence representing substrate retention was measured on the flow cytometer. Median channel fluorescence (MCF) was plotted for each substrate for each cell line for direct comparison between parental and transporter-expressing values. Response was calculated for each substrate for each transporter-expressing/parental cell line pair as</p><p>where MCF_Test, MCF_HC-P, and MCF_NC represent the MCF of wells containing transporter-expressing cells and substrate, MCF of parental cells containing 100 nM substrate, and MCF of wells containing cells with no substrate, respectively. Although responses were calculated for each substrate concentration, only the 100-nM substrate data were used for comparison across cell lines. JC-1 (T3168) and CaAM (C1430) dose responses were used as efflux controls for each run.</p><!><p>Numerous ABC transporter efflux inhibitors have been described previously [17], eight of which were chosen for their specific efflux pump inhibition profiles here: ABCB1 (mometasone), ABCC1 (loxapine and pimozide), ABCG2 (niclosamide and novobiocin), ABCB1/ABCC1 (verapamil), ABCB1/ABCG2 (lasalocid), and ABCB1/ABCC1/ABCG2 (nicardipine) [12,18]. To investigate the inhibition profile of each potential substrate, a 9-point dilution series of each inhibitor (assay well concentrations ranging from 7.6 nM to 50 μM) was added in combination with an optimal concentration of each substrate to ABCB1, ABCC1, and ABCG2 transporter-expressing cells. The optimal concentration of each fluorescent substrate was based on the maximal efflux activity in the transporter-expressing cells. The assay protocol and order of addition were essentially the same as described in the efflux assay; however, the screening volume was reduced to 20 μl per well (19.8 μl of cells with 100 nl of substrate and inhibitor). Time-separated MCF values were plotted for each substrate/inhibitor dose response, resulting in eight inhibition curves per fluorescent sub-strate per cell line. IC50 values were calculated with the four-parameter logistic equation via Prism:</p><p>where Log[Inhib] is the log of the inhibitor concentration, MCF_Test is the MCF value indicating substrate retention in the transporter-expressing cell line, and Bottom and Top define the range of the fitted variable slope sigmoidal curve. Subsequently, each curve was subjected to a set of validation cutoff criteria (Fig. 1). Due to imperfect automated curve fitting, curves were excluded where the standard error for the logIC50 was greater than 15% of the value. Curves with negative Hill slope values were also excluded. Several fluorescence limits were established to exclude low-level or inconsistent response levels. Of the fitted curves, a minimum change from bottom to top of 50 MCF units was required. This helped to eliminate the background fluorescence for several inhibitors, including mometasone, verapamil, and pimozide. For substrates with lower fluorescence ranges (50–500 MCF units), a 2.5-fold change was required from bottom to top of the fitted curve. For fluorescent responses greater than 500 MCF units, a relaxed 2-fold change allowed inclusion of substrates with high baselines (i.e., 1500–3000 MCF inhibition levels). A total of four dose–response curves were analyzed for each inhibitor with each substrate. The IC50 values were averaged, and standard deviations were calculated.</p><!><p>The 102-member fluorescent library of potential efflux substrates included labeling agents and tracer dyes as well as fluorescently labeled small molecules. The fluorophores included fluorescein, rhodamine, rosamine, Alexa Fluor, BODIPY, cyanine, and several others. Each compound was incubated with both parental and transporter-expressing ABCB1, ABCC1, and ABCG2 cell lines using two lasers for excitation and seven fluorescence emission bands for analysis as specified in Materials and Methods. Because each dye had unique fluorescence and cell penetration qualities, it was challenging to find the optimal concentration range. An initial screen was conducted at a higher concentration range of 3 pM to 1 μM, and in some cases the fluorescence levels of parental cell lines were off-scale (data not shown). Therefore, we rescreened all samples at concentrations ranging from 0.3 pM to 100 nM. Although each fluorescent compound was tested in dose response, the 100-nM concentration response was used to compare across cell lines and between substrates. Although we were unable to characterize the full retention curve, optimal probe concentrations could be inferred for use in subsequent testing.</p><p>Fig. 2 illustrates dose–response data for two carbocyanine dyes, JC-1 (T3168) and DiOC2(3) (D14730) in the green fluorescence channel (FL 1, 530 nm) for each dye.3 Efficient T3168 efflux was observed for all three overexpressing cell lines, whereas selective efflux was noted with D14730 by ABCB1 and ABCC1. Fig. 2 also shows that normalized response data provided a sensitive index for comparing transporter-expressing cell efflux to parental retention. The single-point 100-nM MCF was extracted from each curve and plotted for the entire compound set (Fig. 3). A majority of tested compounds (61 of 102) failed the primary efflux assay based on low observed cellular fluorescence (as opposed to uptake without efflux). This may have been due to poor permeability of the dye or possibly to compound instability (i.e., hydrolysis). For example, several compounds such as the Alexa Fluor hydrazides A20501MP and A20502 were noted as membrane-impermeant dyes and acted as negative controls. In fact, all tested Alexa Fluor-containing fluorescent probes, including the Alexa Fluor maleimide labeling reagents A10254, A10255, A10256, A10258, and A20341, were determined to be membrane impermeant in the primary efflux screen, and none of them was considered for screening in the inhibitor assay. In addition, redox-sensitive dyes such as MitoTracker M7511 and M7513 and the dihydrorhodamines D632 and D633 were apparently not oxidized to their fluorescent counterparts, resulting in no observed cellular fluorescence.</p><p>Although some selectivity data have been compiled for commonly used dyes in flow cytometric transporter analysis as well as recently described detection kits [19,20], here we report a direct comparison of efflux activity in ABCB1, ABCC1, and ABCG2 within one screening protocol. For the full compound set, ABCB1 and ABCC1 shared similar uptake and efflux profiles. Although substrates of ABCC1 were also substrates of ABCB1 (i.e., the cyanines S7575 and S7578), the reverse was not necessarily true (i.e., the cyanines D22421, D273, and D378). The only exception was the ABCC1/ABCG2 efflux substrate CellTracker Green CMFDA (C2925), which was pumped poorly by ABCB1. ABCG2 pumped the fewest overlapping substrates with ABCB1 and ABCC1 and tended to have higher fluorescence baselines. Cellular fluorescence was not observed with any of the labeled methotrexate analogs regardless of fluorophore, including Alexa Fluor (M23271), fluorescein (M1198MP), rhodamine (M23273), and BODIPY (M23272). Six probes (B10250, B7447, D20350, L7526, R6479, and N1142) with membrane permeability were not substrates for any of the transporters.</p><p>Fluorescent dyes that were actively effluxed by one or more of the transporter-expressing cell lines (response >75%) and displayed appropriate fluorescence levels (∼100 MCF units or more) were chosen for the inhibition screening protocol. These criteria were based primarily on the reference substrate T3168, which has previously been used in high-throughput screening (HTS) and was effluxed by ABCB1, ABCC1, and ABCG2 (Fig. 2A and B). Fig. 4 shows the 31 selected substrates displayed on the basis of the efflux response in transporter-expressing cells along with the fluorescence retention in the parental cell line. Retention in the parental cell line was used to gauge the potential fluorescence response. For simplicity, we assume that full inhibition of efflux in the transporter-expressing cells affords similar cellular fluorescence retention levels as in the cells with low pump expression. This data subset demonstrates that high responses were observed for the majority of active pump substrates in ABCB1, with the only exceptions being C2925 and D20350. Fig. 4 also shows that the efflux activity of ABCC1 closely mirrors that of ABCB1, with the exception of substrate C2925 as noted above. The ABCG2 profile of effluxed substrates was significantly narrower than the other two transporters, indicating higher specificity in pump–substrate interactions. Fewer than a dozen substrates showed responses above 75%, including substrate D20350, which appeared to be ABCG2 selective at approximately 65% response.</p><!><p>The 31 fluorescence substrates from Fig. 4 were tested against a panel of eight inhibitors known for their activity versus the three transporters. Nicardipine is a strong cross-reactive inhibitor for all three transporters and was used as the reference standard [12]. MCF was used to plot the dose–response data for the eight inhibitors with each fluorescent substrate and to generate IC50 values. After curve fitting, Hill slopes ranged from 0.3 to 10.9 with a mean of 2.1 ± 1.3 (median = 2.1), with 42% between 0.5 and 1.5 and 82% between 0 and 3. Fig. 5 illustrates the results for the fully cross-reactive efflux substrate, BODIPY histamine (B22461), inhibited by both mometasone and nicardipine for ABCB1, ABCC1, and ABCG2.</p><p>Table 2 shows the IC50 values calculated for the active inhibitor/substrate pairs. Novobiocin was inactive and excluded from further consideration. Although efflux was noted for CellTracker Green CMFDA (C2925), D20350, DiOC7(3) (D378), and BODIPY taxol (P7500), no notable inhibition was observed and these substrates are also excluded from Table 2. For visualization, substrate/inhibitor pairs, clustered by fluorophore, are also depicted as a heat map in Fig. 6, showing the activity of each substrate in each cell line. It should be noted that for fluorescently labeled small molecules (i.e., the majority of active BODIPY probes), this clustering method only roughly indicates structure–activity relationships due to the variation in the nonfluorophore moiety. A distinction also needs to be made between efflux in the first round of screening and efflux inhibition by one or more inhibitors for each ABC transporter-expressing cell line. Although not depicted, substrates with noninhibited efflux activity are discussed by fluorophore. Of the 27 inhibitable substrates, 19 contained BODIPY or cyanine fluorophore, with 7 of the remaining 8 containing rhodamine, rosamine, or fluorescein.</p><p>Aside from the previously mentioned efflux-inactive M1198MP, the fluorescein probes CaAM (C1430) and C2925 demonstrated efflux activity in the primary screen for ABCB1/ABCC1/ABCG2 and ABCC1/ABCG2, respectively. Only C1430 was taken forward into the inhibition screen, where inhibition of ABCB1/ABCC1 responses with mometasone, nicardipine, and pimozide was observed. No significant selectivity was seen between ABCB1 and ABCC1, with all IC50 values being in the low micromolar (μM) range for mometasone (1.9 ± 1.6 and 5.4 ± 5.6 μM, respectively) and nicardipine (5.8 ± 2.8 and 4.3 ± 5.0 μM, respectively). In a flow cytometric fluorescence retention analysis, Wang and coworkers reported ABCB1 efflux inhibition of CaAM with nicardipine at an IC50 of 6.6 ± 0.4 μM [21], which correlated well with the IC50 value reported here.</p><p>A total of 34 rhodamine/rosamine-based compounds were represented in the collection. Unconjugated alkyl amine-substituted rhodamine probes tended to be active in ABCB1 or ABCB1/ABCC1 efflux and inhibitor assays provided that the carboxylic acid was ester protected (R634, R648MP, and T669). The exception was the membrane probe R18 (O246), with its octadecyl ester demonstrating no cellular fluorescence in the efflux assay. An exception to the ester-based activity rule was the free carboxylate-containing CellTracker Orange CMTMR (C2927), where the aryl amide substitution appears to maintain adequate lipophilicity to facilitate membrane permeability. All four of the rhodamine substrates tested in the inhibitor assay (R634, R648MP, T669, and C2927) showed quantifiable ABCB1 efflux inhibition with both mometasone and nicardipine. Although not fully illustrated in Fig. 6, each of these substrates was at least weakly inhibited by mometasone and nicardipine in ABCC1 as well (Table 2). However, the potential for high selectivity of ABCB1 over ABCC1 can be seen in the sub-μM ABCB1 efflux inhibition example of R648MP with nicardipine. Inhibitor-based substrate efflux variation can also be seen with R648MP, which was observed to have an ABCB1/ABCC1 cross-pump interaction with pimozide. C2927 efflux was inhibited by pimozide as well as verapamil in ABCB1 and ABCC1. No significant inhibition was seen for these four rhodamine substrates with lasa-locid, loxapine, or niclosamide. Despite a long history of use in transporter efflux assays [22], rhodamine 123 (R302) was observed to have comparatively low fluorescence levels at the available wavelengths and was not explored further in the inhibition protocol.</p><p>Rosamine-based tetramethylrosamine chloride (T639) and the MitoTracker dyes M7510 and M7512 showed ABCB1/ABCC1 efflux potential, albeit at lower than optimal fluorescence levels. Low-μM efflux inhibition of M7510 and T639 was observed in ABCB1 with mometasone, nicardipine, and pimozide. T639 also demonstrated similar ABCC1 efflux inhibition with mometasone, pimozide, and (to a lesser degree) verapamil. The ABCB1 T639 efflux inhibition result also correlated with low-μM nicardipine inhibition (IC50 = 11.7 μM) previously reported by Wang and coworkers [21].</p><p>A total of 37 BODIPY-based probes were tested in the primary efflux screen, with 8 going forward into the inhibition assay. Aqueous solubility of BODIPY analogs is often of concern and likely affected those compounds without polar functional groups, resulting in low cellular fluorescence in the efflux assay. Although efflux by ABCB1/ABCC1/ABCG2 was noted for the acidic compartment tracer LysoTracker Green DND-26 (L7526), the low-level efflux response coupled with less than optimal fluorescence ranges excluded it from further investigation. BODIPY EDA (D2390) demonstrated a low fluorescence but high response efflux activity across all three pumps, which translated into an interesting inhibition profile in the subsequent screen. D2390 showed full cross-pump inhibition in ABCB1, ABCB2, and ABCG2 with lasalocid and was one of only two probes inhibited by lasalocid. The efflux inhibition of D2390 by loxapine, mometasone, nicardipine, pimozide, and verapamil was noted only for ABCB1 and ABCC1 at low-μM IC50 values.</p><p>Although little activity was seen with the majority of BODIPY labeling agents, BODIPY-labeled small molecules demonstrated activities apparently related to the labeled molecule rather than the fluorophore. In the primary efflux screen, pan-ABCB1, -ABCC1, and -ABCG2 activity was observed for BODIPY prazosin (B7433), glibenclamide (E34251, ER-Tracker Green), verapamil HCl (B7431), and vinblastine (V12390). The inhibition profile of these four substrates was more varied. B7433 has previously been described as an ABCG2 substrate and used as a fluorescent probe for inhibitors in ABCG2-transfected HEK293 cells [23]. Here, B7433 demonstrated ABCB1, ABCC1, and ABCG2 inhibition with mometasone and nicardipine and showed moderate ABCB1 and ABCC1 efflux inhibition with loxapine, pimozide, and verapamil. Unlabeled glibenclamide is a competitive ABCB1 inhibitor [24]. Here, labeled glibenclamide (E34251) showed both ABCB1 and ABCC1 efflux inhibition with nicardipine. Verapamil is well known in multidrug resistance (MDR) reversal [25] and the BODIPY-labeled analog (B7431) has been described as a substrate for both ABCB1 and ABCC1 [26,27]. We observed low to moderate ABCB1 and ABCC1 efflux inhibition by B7431 with loxapine, mometasone, nicardipine, pimozide, and verapamil. Vinca alkaloids are ABCB1 and ABCC1 pump substrates [28], and low-μM ABCB1 and ABCC1 efflux inhibition was noted here for BODIPY vinblastine (V12390) with mometasone, nicardipine, verapamil, and (to a lesser degree) loxapine. Although active efflux of BODIPY histamine (B22461) was noted only for ABCB1 and ABCG2, inhibition was observed with mometasone and nicardipine across all three pumps, indicating that it was indeed an efflux substrate for ABCC1. Reduced inhibition was also noted for B22461 with pimozide (ABCC1 and ABCG2) and verapamil (ABCC1). ABCB1 and ABCC1 efflux activity was observed for BODIPY forskolin (B7469), BODIPY thapsigargin (B7487), and BODIPY taxol (P7500). B7469 efflux was inhibited by nicardipine across all three pumps along with ABCB1 inhibition by loxapine, mometasone, and pimozide. Efflux of B7487 was inhibited to varying degrees for ABCB1 and ABCC1 by loxapine, mometasone, and nicardipine. The cost and availability of P7500 made it less attractive, and it was not carried forward into the inhibition screen.</p><p>Of the 17 cyanine dyes tested in the efflux assay, 12 were taken into the inhibition screen. The monomeric cyanine dyes, the dead cell indicator TO-PRO-1 iodide (T3602), the apoptotic cell stain YO-PRO-1 iodide (Y3603), and the cell-impermeant nucleic acid stain YOYO-1 iodide (Y3601), not surprisingly, showed no parental cell uptake in the efflux screening protocol. The asymmetric cyanine dyes (M7514, 34854, S7575, and S7578) showed varying degrees of inhibition of ABCB1 and ABCC1 by loxapine, mometasone, nicardipine, pimozide, and verapamil. Limited cellular fluorescence was noted for the lipophilic tracer indocarbocya-nine dyes D383 and D7776. However, the potentiometric probes indodicarbocyanine (DiIC1(5), H14700) and thiadicarbocyanine (DiSC3(5), D306) both demonstrated inhibitable efflux responses. H14700 efflux was inhibited by mometasone and pimozide for ABCB1 and ABCC1 and by nicardipine for ABCB1 alone. ABCB1 efflux of D306 was inhibited by loxapine, nicardipine, pimozide, and verapamil along with ABCB1 and ABCC1 inhibition with mometasone. The oxacarbocyanine probes (D272, D273, D378, D14730, and D22421) exhibited ABCB1 or ABCB1/ABCC1 efflux activity, which was inhibited to varying degrees by mometasone, nicardipine, pimozide, and verapamil. The benzimidazolylcarbocy-anine JC-1 (T3168) [12,29] was shown to be an efflux substrate for all three ABC transporter pumps, and at an optimal 500-nM concentration efflux was inhibited by mometasone, nicardipine, and pimozide for ABCB1, ABCC1, and ABCG2. The nicardipine ABCB1 efflux response for T3168 matches earlier data (IC50 values of 5.8 ± 2.8 vs. our 7.1 ± 0.7 μM) [21]. T3168 efflux inhibition was also reported in ABCG2 alone for lasalocid (5.1 ± 3.9 μM) and niclosamide (0.8 ± 0.6 μM), demonstrating unique substrate/inhibitor specificity.</p><p>Of three probes (L7595, N1142, and P7581) not placed into fluorophore categories, the efflux of the ABCB1 substrate LDS 751 (L7595) [30,31] was inhibited with loxapine, mometasone, nicardipine, pimozide, and verapamil. Reported IC50 values for nicardipine (5.6 ± 0.5 μM) and verapamil (4.7 ± 1.3 μM) [21] compare with ours (1.7 ± 1.0 and 8.9 ± 5.7 μM, respectively).</p><!><p>Taken together, our results reveal a general correspondence of efflux substrates between ABCB1 and ABCC1, with ABCB1 exhibiting somewhat more diversity. In contrast, the list of efflux substrates for ABCG2 was more restricted, indicating higher selectivity in pump–substrate interactions. The majority of tested efflux substrates have BODIPY or cyanine fluorophores and, interestingly, BODIPY is analogous to a rigid monomethine cyanine dye. Moreover, the selectivity of efflux activities by BODIPY-labeled small molecules correlated more with the labeled molecule than the fluorophore. Evaluations of inhibitors and efflux substrate pairs yielded mometasone and nicardipine as pan-inhibitors for multiple efflux substrates across ABCB1 and ABCC1. Our control inhibitor nicardipine blocked efflux from all three transporters (ABCB1, ABCC1, and ABCG2) for four substrates: carbocyanine JC-1 (T3168) and BODIPY-labeled small molecules BODIPY prazosin (B7433), BODIPY forskolin (B7469), and BODIPY histamine (B22461). In an orthogonal view, the cyanine compound T3168 was a pan-substrate for ABCG2 across a majority of the inhibitors and was the sole ABCG2 substrate inhibited by niclosamide. The combination of BODIPY-labeled substrate D2390 with lasalocid was active for all three pumps. D2390 was the only substrate besides T3168 inhibited by lasalocid, suggesting related binding sites for this substrate/inhibitor pair.</p><p>Based on the response of the control inhibitor nicardipine, 12 substrates demonstrated ABCB1/ABCC1 efflux inhibition; however, no ABCC1, ABCG2, ABCB1/ABCG2, or ABCC1/ABCG2 selectivity was observed. Mometasone, pimozide, and (to a lesser degree) verapamil displayed similar selectivity profiles as nicardipine across the substrate set. Lasalocid was originally chosen as a specific inhibitor of T3168 efflux for ABCG2, and this interaction was replicated here with an IC50 value of 5.1 ± 3.9 μM, with limited activity across the rest of the substrates. Niclosamide, the other ABCG2-specific T3168 efflux inhibitor, was found to inhibit only T3168 efflux for ABCG2 (IC50 = 0.8 ± 0.6 μM). The two ABCC1 inhibitors, loxapine and verapamil, produced similar ABCB1 and ABCC1 inhibition profiles over the substrate set.</p><p>The three transporters (ABCB1, ABCC1, and ABCG2) reported on here are known to significantly influence the ADME-Tox (absorption, distribution, metabolism, excretion, and toxicity) properties of drugs [32]. Although a large number of compounds possess ABC transporter inhibitory properties, only a few of these agents are appropriate candidates for clinical use as MDR reversal agents [33]. Characterization of the transporter/substrate/inhibitor interactions might provide further clues about their structure and mechanism of action as well as aid in predicting potential drug interactions among different therapeutic agents. Ongoing clinical trials with third-generation modulators (e.g., biricodar, zosuquidar, laniquidar) [34] have not yet defined an ideal MDR reversal agent [35]. Main liabilities from cross-reactivity of these inhibitors with major ABC transporters involved in the body's physiological protection from xenobiotics and endogenous metabolites result in high toxicity and mortality in patients. Acquired mutations in transporter genes introduce more complexity, altering the pattern of resistance and improving the ability of the mutants to efflux new drugs [36]. Moreover, drug-resistant human tumor cell lines express different ABCG2 variants, suggested to be gain-of-function mutations acquired during the course of drug exposure [37]. Single amino acid changes alter the drug resistance profile and substrate specificity compared with wild-type ABCG2 [38]. Thus, daunorubicin, used to identify ABCB1 inhibitors, is a substrate for ABCG2 mutants [39].</p><p>A diverse set of compounds are substrates for efflux pumps, with many showing cross-pump activity. It stands to reason that, for each pump, more than one combination of substrate and inhibitor is required to properly characterize that particular pump's activity. For example, rhodamine 123 has been used in combination with Hoechst 33342 to describe two functional transport sites in ABCB1 with complex allosteric interactions [40]. In combination with LDS 751, it was also shown that the rhodamine 123 may bind to a different or overlapping region within the same large flexible binding site as LDS 751 [41]. It has also been shown that ABCB1 possesses two allosterically coupled drug acceptor sites where one binds vinblastine, doxorubucin, etoposide, and cyclosporin A and the other binds dexniguldipine-HCl and other 1,4-dihydropyri-dines [42]. Fluorescent substrates in combination with high-throughput flow cytometry can be a powerful tool for transporter interaction studies. The single-transporter fluorescent substrate, pheophorbide A, was shown to be ABCG2 specific. Its transport correlated with ABCG2 expression, and an HTS campaign identified ABCG2 inhibitors [43,44]. Further development of high-throughput assay systems to screen for potential transporter-interacting partners may be of particular interest to help elucidate structure and function within a given transporter.</p><p>The current study indicates that each substrate has not only diverse activity for multiple transporters but also unique interactions with different inhibitors, suggesting the involvement of different binding sites in substrate recognition and transport inhibition processes. Such widely reported promiscuity (or polyspecificity) does not allow a definitive correlation between substrate specificity and structural characteristics of the fluorophores investigated here. However, several trends in transporter/inhibitor/substrate interactions were identified. We found striking similarity in substrate specificity profiles between ABCB1 and ABCC1 transporters. All ABCC1 efflux compounds in our test system demonstrated variable levels of interaction with ABCB1. Moreover, mometasone and nicardipine, two structurally diverse compounds, inhibited most of the ABCB1/ABCC1-specific compounds at low concentration. On the other hand, only 10 of 102 tested fluorescent dyes were effluxed by ABCG2, and only 5 of them could be inhibited by the tested inhibitors. Interestingly, T3168 efflux by ABCG2 was inhibited by six of seven tested inhibitors. The other four ABCG2-specific dyes belong to the BODIPY family.</p><p>One of the major aims of the current research was to provide novel fluorescent tools for understanding drug resistance. Well-characterized fluorescent probes with variable profiles of selectivity can be used to identify new inhibitors as well as to characterize functional expression of ABCB1, ABCC1, and ABCG2 in tissue samples. Pan-efflux substrate probes such as T3168 and the BODIPY-FL-labeled small molecules histamine (B22461), prazosin (B7433), and forskolin (B7469) could help to define the expression of several transporters in combination with inhibitors such as nicardipine and mometasone. By varying the inhibitor, in principle, one could evaluate pump expression such as T3168 combined with lasalocid (IC50 = 5.1 μM) or niclosamide (IC50 = 0.8 μM) for ABCG2 expression or ABCB1 expression probed with LDS 751 (L7595) and inhibitors such as loxapine, mometasone, nicardipine, pimozide, and verapamil. Depending on the system of interest, combinations of such substrate/inhibitor pairs might describe the ABC transporter phenotype. These data could also be useful for predicting common drug inhibition/drug binding patterns of ABC transporters and contribute to a better understanding of the pharmacological mechanisms of transporter–reversal agent interactions. Further refinement of the data presented here could also lead to multicolor probe sets that identify unique binding motifs in a single pump or across pumps, allowing for concurrent elucidation of the mode of action for existing or new ABC transporter modulators and substrates.</p>

PubMed Author Manuscript

Effect of PKD1 gene missense mutations on polycystin-1 membrane topogenesis\xe2\x80\xa0

Polycystin-1 (PC1), the product of the Polycystic Kidney Disease-1 (PKD1) gene, has a number of reported missense mutations whose pathogenicity is indeterminate. Previously, we utilized N-linked glycosylation reporter tags along with membrane insertion and topology assays to define the eleven membrane-spanning domains (I-XI) of PC1. In this report, we utilize glycosylation assays to determine whether two reported human polymorphisms/missense mutations within transmembrane (TM) domains VI and X affect the membrane topology of PC1. M3677T within TM VI had no effect on the topology of this TM domain as shown by the ability of two native N-linked glycosylation sites within the extracellular loop following TM VI to be glycosylated. In contrast, G4031D, within TM X, decreased the glycosylation of TM X reporter constructs demonstrating that the substitution affected the C-terminal translocating activity of TM X. Furthermore, G4031D reduced the membrane association of TM X and XI together. These results suggest that G4031D affects the membrane insertion and topology of the C-terminal portion of polycystin-1 and represents a bona fide pathogenic mutation.

effect_of_pkd1_gene_missense_mutations_on_polycystin-1_membrane_topogenesis\xe2\x80\xa0

<!>PC1 membrane-associated construct lacking native N-linked glycosylation sites<!>Mutant TMVI (M3677T) expression construct<!>Mutant TMX (G4031D) expression constructs<!>In Vivo Expression and Glycosidase Assays<!>In Vitro Expression, Glycosylation, and Membrane Association Assays<!>TM VI mutation M3677T does not alter the membrane insertion or topology of TM VI<!>TM X mutation G4031D decreases the translocation efficiency of TM X<!>G4031D mutation affects the membrane association property of TM X<!>G4031D mutation may influence the membrane insertion of TM X and TM XI<!>Discussion<!>M3677T does not affect membrane integration of TM VI<!>In vivo glycosylation analysis of TM XG4031D mutant<!>In vitro translocation analyses of mutant TM X<!>Analysis of membrane association of mutant TM X with TM XI

<p>Polycystin-1 (PC1) is the protein product of the PKD1 gene, which when mutated is responsible for 85% of the cases of autosomal dominant polycystic kidney disease (ADPKD). Mutations within the PKD2 gene, encoding polycystin-2 (PC2), comprise the remainder of ADPKD cases. ADPKD is a systemic disease that is primarily characterized by fluid-filled, epithelial-lined cysts within both kidneys, and is associated with increased prevalence for hypertension, aneurysms, hernias, and cysts in other organs (e.g., liver and pancreas). ADPKD is highly prevalent, affecting one in every 500–1,000 individuals, and leads to end stage renal failure in approximately half of those affected. As such, ADPKD comprises nearly 5% of the costs for renal replacement therapy in the United States (1).</p><p>PC1 is an integral plasma membrane protein that has been localized to multiple sites within the cell including the primary cilium (2–4). PC1 is composed of a large N-terminal extracellular portion, eleven transmembrane (TM) domains, and a short intracellular C-terminal tail (5, 6). The N-terminal portion of PC1 consists of multiple domains proposed to be involved in both cell-cell and cell-matrix interactions and in sensing fluid shear stress (5, 7). The C-terminal tail interacts with multiple protein partners, is proteolytically cleaved in response to changes in mechanical stimuli, and initiates multiple signaling pathways (8–20). Evidence suggests a function for PC1 as a G protein coupled receptor (GPCR) (21, 22), including the ability of the C-tail to directly bind heterotrimeric G proteins (10, 23). The C-tail of PC1 also interacts with PC2 via a coiled coil domain (24). PC2 is a smaller protein with six TM domains and has been shown to form a cation-selective ion channel permeable to Ca2+ (25). Studies have shown an ability for PC1 and PC2 to sense fluid shear stress and initiate calcium-mediated signaling (26). Altogether, these observations suggest that PC1 is a complex, multi-functional protein capable of acting as a mechanosensor, receiving signals from the primary cilia, neighboring cells, and extracellular matrix, and transducing them across the plasma membrane to the interior of the cell.</p><p>Our previous work provided the first experimental evidence supporting an eleven TM domain (I-XI) structure for PC1 (27). Furthermore, those studies suggested that the membrane biogenesis of PC1 is somewhat complex, with TM domains I-IX inserting in a cotranslational and sequential manner, while the insertion of TM domains X and XI appeared to be non-cotranslational and cooperative. Currently, nothing is known regarding how the membrane-associated structure of PC1 influences its functions or how disease-associated mutations affect the membrane topology or biogenesis of PC1. In this report, we utilize glycosylation assays of endogenous N-linked glycosylation sites or engineered glycosylation reporter gene fusions to determine whether two reported human polymorphisms/mutations within TM domains VI and X affect the membrane topology of PC1. These studies demonstrate that the disease-associated missense mutation G4031D within TM X negatively impacts the membrane-integrated structure of TM X and possibly of the last two TM domains of PC1, suggesting that it represents a pathogenic mutation.</p><!><p>In order to identify native, N-linked glycosylation sites within the membrane-spanning portion of PC1, three potential N-linked glycosylation sites (N3139, N3728, and N3780) within the C-terminal 1,284 residues of murine polycystin-1 were sequentially removed by mutating asparagine (N) residues conforming to the N-linked consensus sequence, NxS/T, to serine (S) residues. The sites were mutated by a combination of site-directed mutagenesis (for the N3139 or N1 site; Gene Editor kit, Promega) and PCR/replacement cloning (for the N3728 or N2 and N3780 or N3 sites), as described previously (27). All mutations and the integrity of PC1 sequences were confirmed by DNA sequencing. Finally, the C-terminal 1,284 residues of mutants N1S, N12S, and N123S (ScaI-NotI fragment) were joined in frame with the CD5 protein signal sequence to generate CD5-11TMN1S, CD5-11TMN12S, and CD5-11TMN123S.</p><!><p>To generate an 11 TM domain expression construct with the M3677T mutation within the sixth TM domain, CD5-11TM was used as a template for Stratagene Quikchange kit using M3677TFor (5′-AGACTCTTGGTGTACACGCTTTTCTTACTGGTGACG-3′) and M3677T Rev (5′-CGTCACCAGTAAGAAAAGCGTGTACACCAAGAGTCT-3′) primers. The mutation was confirmed by DNA sequencing. In order to eliminate potential effects of offsite mutations, multiple CD5-11TMM3677T clones were analyzed.</p><!><p>To generate TM X mutant clones, CD5-I-X or flag-X (constructed as described (27)) were used as templates for Stratagene Quikchange kit with G4031D For (5′-GGGAGCCACCTTGGACCTGGTGCTGCTTG-3′) and G4031D Rev (5′-CAAGCAGCACCAGGTCCAAGGTGGCTCCC-3′) primers. Generation of the desired mutation was confirmed by DNA sequencing. Primer synthesis and DNA sequencing was performed by the staff at the Kansas Medical Center Biotechnology Support Facility.</p><!><p>HEK293T cells were maintained in DMEM/4.5 g/L glucose/10% serum with penicillin/streptomycin. Transient transfections were performed as previously described (27). Briefly, cells were incubated with CaHPO4-DNA precipitates consisting of 100 or 250 ng of PC1 fusion protein DNA unless otherwise noted and pBluescript (Stratagene) filler DNA for a total of 8 μg of DNA for 3.5 h. Cells were lysed 22–26 h later in passive lysis buffer (Promega). For analysis of N-linked glycosylation, lysates were supplemented with detergents at final concentrations of 1% Triton X-100, 0.5% sodium deoxycholate, and 0.1% SDS, and nuclei were pelleted. Supernatants were digested with N-glycosidase F (Glyko) as described previously (27). Glycosidase reactions or cell lysates were precipitated, solubilized in 2× SDS loading buffer at 65 °C, electrophoresed on either 7.5% or 5% polyacrylamide-SDS minigels, and transferred to Immobilon-P membrane (Millipore). Some blots were blocked in 5% milk-TBSN (10 mM Tris, pH 7.4, 0.9% NaCl, 0.1% NP-40) for probing with a rat anti-HA monoclonal antibody (Roche). Alternatively, for probing with antiserum raised against a C-terminal peptide of mouse PC1 (AS19, (10)), blots were blocked in 5% milk-TBST (10 mM Tris, pH 7.4, 0.9% NaCl, 0.1% Triton X-100). Blots were developed with either anti-rat-HRP secondary and ECL substrate or anti-rabbit-AP secondary antibody and CDP-Star and were exposed to film.</p><!><p>For in vitro analysis of glycosylation, flag-TM-glycosylation reporter clones were incubated in coupled transcription/translation (T7 Quick TNT) extracts (Promega) with 35S-methionine (Amersham) in the presence of canine pancreatic microsomes (Promega) (27). For assessment of membrane association of the flag-TM X clones, 1 μL of a TNT reaction was equilibrated in 0.5 mL of 100 mM Tris, pH 7.5, and 250 mM sucrose and then centrifuged for 30 min at 78,000 rpm, 4 C, in a TLA100.2 rotor. Resulting Tris-sucrose supernatants were precipitated in methanol, dissolved in 2× SDS loading buffer, and denatured 2 min at 100 C. Treatment of TNT reactions with N-glycosidase F was in glycosylation lysis buffer (150 mM NaCl, 50 mM Tris, pH 7.5, 2 mM EDTA, 0.5% Triton X-100, 0.5% sodium deoxycholate, 0.5% SDS) (27). Following electrophoresis, gels were enhanced in 1 M sodium salicylate, dried down, and exposed to film at −70° C.</p><!><p>A number of missense mutations/polymorphisms have been reported in the PKD1 gene (http://pkdb.mayo.edu/cgi-bin/mutations.cgi). One of these amino acid substitutions, M3677T, is located within TM VI, which has been shown to have an Ncytosolic-Cextracellular oriention (27). In order to analyze the effect of M3677T on the topology of TM VI within the context of the membrane-integrated portion of PC1, the C-terminal 1,284 amino acid residues of murine PC1 were cloned with the signal sequence of CD5 protein to generate CD5-11TM (Figure S1A). The authenticity of three consensus, N-linked glycosylation sites within the membrane-spanning portion of PC1 was determined by a combination of mobility shift and N-glycosidase assays with site-directed mutants of CD5-11TM (Figure S1B and C). These experiments showed that the loop following TM VI contained the only native N-linked glycosylation sites within this portion of PC1 (Figure S1D). Therefore, the glycosylation status of this loop was used as a topology reporter to determine if the M3677T substitution affected the membrane insertion of TM VI. The M3677T mutation was engineered into CD5-11TM to generate CD5-11TMM3677T, and the electrophoretic mobilities of CD5-11TM, CD5-11TMN123S, and CD5-11TMM3677T fusion proteins were compared (Figure 1). As observed previously (Figure S1B), the mobility of the wild type and the triple glycosylation mutant CD5-11TM proteins differ considerably, however, analysis of three independent CD5-11TMM3677T mutant clones showed electrophoretic mobility identical to that of wild type CD5-11TM. These results suggest that the Met to Thr substitution at residue 3677 does not prevent TM VI from inserting into the membrane and translocating the following loop into the lumen of the ER.</p><!><p>A human disease-associated amino acid substitution, G4031D, has been reported in TM X (28). TM X was previously shown to have an Ncytosolic-Cextracellular topology (27). To determine if G4031D affects the membrane integration and C-terminal translocation of TM X in vivo, the amino acid substitution was engineered into the CD5-I-X glycosylation reporter clone (Figure 2A) (27). This construct contains the signal sequence of CD5 protein followed by the first 10 TM domains of PC1 and ends in the glycosylation reporter tag consisting of the topologically neutral prolactin (PRL) "spacer" sequence (29), 3 consensus N-linked glycosylation sites (N3), and a hemagglutinin (HA) tag (30). All native N-linked glycosylation sites were removed from PC1 sequences in order to assess glycosylation of the C-terminal glycosylation reporter tag only. The wild type and mutant constructs were transfected into 293T cells and their glycosylation efficiencies were analyzed by Western blot (Figure 2B). The relative proportion of glycosylated versus non-glycosylated species of wild type CD5-I-X was less than 50% (upper band compared to lower band), which is consistent with previous results showing that TM X has somewhat weak signal anchor II activity (27). Importantly, the proportion of glycosylated to non-glycosylated species was significantly less for the CD5-I-XG4031D mutant protein. These results suggest that the TM X mutant has weaker translocation activity than wild type TM X.</p><p>To further assess the effects of the G4031D substitution, wild type and mutant TM X domains and their N- and C-terminal loop regions were cloned in frame with an N-terminal FLAG epitope tag (flag) and the C-terminal glycosylation reporter tag (PRL-N3-HA) (Figure 3A), and were expressed in in vitro coupled transcription/translation (TNT) extracts in the presence of rough microsomal membranes. Since these expression constructs lack a signal sequence, this assay reveals whether the TM domains can function to direct their own membrane-associated synthesis and integration (i.e., signal-anchor assay). Gel electrophoresis of the TNT reaction products (Figure 3B) shows that the wild type flag-X template resulted in synthesis of both glycosylated and non-glycosylated species, as demonstrated by treatment with NgF, while the flag-XG4031D template produced non-glycosylated product. Very long exposures revealed a minor amount of glycosylated flag-XG4031D protein (data not shown). These in vitro observations are consistent with the in vivo data obtained with wild type and mutant CD5-I-X fusion proteins.</p><!><p>To determine whether the reduced translocation activity of the TM X mutant fusion protein is due to the loss of signal sequence properties, an aliquot of the TNT plus canine microsomal membrane reaction was centrifuged in Tris sucrose buffer, and the resulting membrane pellet (P) and supernatant (S) fractions were analyzed for amounts of fusion protein (Figure 3C). More than half of the flag-XG4031D fusion protein was found in the supernatant (non-membrane associated) fraction. In contrast, the majority of wild type flag-X was present in the membrane pellet (the glycosylated and non-glycosylated forms together). This suggests that the G4031D mutation reduces the ability of TM X to act as a signal sequence and to direct its own synthesis at membrane-bound ribosomes.</p><!><p>Previously, we showed that the membrane association and integration of TM X was increased by the addition of C-terminal sequences containing the extracellular loop and TM XI (27). To determine whether the G4031D mutation affected the ability of the two TM domains together to associate with membranes, wild type and mutant flag-X-XI fusion proteins were synthesized in TNT extracts with microsomal membranes and subjected to centrifugation in Tris sucrose buffer. Analysis of the pellet and supernatant fractions (Figure 4B) showed that addition of TM XI resulted in approximately half of flag- XG4031D-XI being associated with the microsomal membrane pellet. In contrast, the major proportion of wild type flag-X-XI fusion protein was found in the membrane pellet. These results suggest that the G4031D mutation could negatively impact the membrane integration of the final two TM domains of PC1.</p><!><p>Disease-associated mutations within the human PKD1 gene include insertions, deletions, duplications, nonsense, and missense mutations (http://pkdb.mayo.edu/cgi-bin/mutations.cgi). Most of the missense mutations are classified as indeterminate since the effects of these amino acid substitutions have not been and/or cannot be easily assayed. However, a few PKD1 missense mutations have been demonstrated to impact properties and functions described for PC1. PKD1 family-related mutations within the receptor for egg jelly (REJ) domain were shown to prevent cleavage at the GPS domain, to abolish STAT1 activation, and to prevent in vitro tubulogenesis (31). Several missense mutations within the first PKD repeat were shown to alter the stability of the folding of this domain, which was proposed to affect its function in mechanosensing (7, 32). A missense mutation within the coiled-coil domain of the cytosolic C-terminal tail of PC1 was found to alter its ability to interact with, activate, and stabilize PC2 (33–35). Currently, nothing is known regarding how human disease-associated mutations affect the membrane topology or biogenesis of PC1. The Pkd1m1Bei mouse is the only known example of a missense mutation within a TM domain of PC1 that leads to cystic disease (36). In this instance, ENU-mutagenesis generated a Met to Arg substitution (M3083R) within TM domain I which resulted in a mouse model with a homozygous null phenotype. The molecular basis by which M3083R leads to the complete loss of PC1 function is not known however. Previously, we developed expression constructs and assays that enabled us to determine the location and topology of the 11 TM domains and intervening loop regions of PC1 (27). In this report, we utilize electrophoretic mobility shift, glycosidase, and membrane association assays to determine whether two reported human missense mutations within TM domains VI and X affect the membrane-associated structure of PC1.</p><p>Glycosidase and mobility shift assays with single, double and triple glycosylation mutants were used to demonstrate the presence of two, native N-linked glycosylation sites within the CD5-11TM expression construct of PC1. The results of these assays (Figure S1D) are consistent with the experimentally supported membrane topology model for PC1 (27). Both N2 (N3728), and N3 (N3780) glycosylation sites are located in the extracellular loop after TM VI where they would be exposed to modifying enzymes during synthesis of the protein in the ER. In contrast, N1 (N3139) is located within the first intracellular loop and, therefore, would not be expected to be glycosylated. Furthermore, the sequence for N1 is NPT, and there are reports that "x", in the N-linked glycosylation consensus sequence, NxS/T, can be any amino acid residue except proline (37). Confirmation of the authenticity of N2 and N3 as glycosylation sites allowed them to be used as topology reporters for a missense mutation in TM VI (see below). It is anticipated that these N-linked glycosylation sites will be useful for assaying the effects of other missense mutations within preceding TM domains (e.g., TM I – V) or within other regions of PC1 that might impact its topology and/or transport from the ER (i.e., by Endo H cleavage resistance-susceptibility assays).</p><p>One of the ADPKD-associated amino acid substitutions investigated was M3677T, located in TM domain VI. M3677T did not alter the electrophoretic mobility of CD5-11TM (Figure 1) indicating that sites N2 and N3 were glycosylated as in wild type CD5-11TM (Figure S1). The ability of sites N2 and N3 to be glycosylated suggests that this amino acid substitution did not alter the membrane insertion or orientation of TM VI. However, this approach is unable to discern if M3677T causes more a subtle change such as a shifting in the location of the membrane-embedded residues of TM domain VI. It is possible that M3677T may affect other properties of PC1 such as its stability, or the ability of TM VI to interact with other TM domains of PC1 or with interacting partners such as PC2. Recently, the TM domain portions of PC1- and PC2-related proteins PKD1L3 and PKD2L1 were shown to interact with each other in order for this complex to be transported to taste pores (38). Although the stability of the M3677T mutant of CD5-11TM did not appear to be altered (as compared to wild type CD5-11TM) in our studies, we realize that ours represents an overexpression system that might overwhelm endogenous degradative mechanisms. Finally, it is also possible that M3677T represents a harmless polymorphism based on the observation that although Met is present at residue 3677 in PC1 from human, mouse, rat, dog and chicken, the corresponding residue in Fugu PC1 is Thr.</p><p>The second ADPKD-associated missense mutation investigated was G4031D, located within TM domain X. Previous work demonstrated that TM X has the ability, albeit weak, to translocate its C-terminus and associated sequences into the ER lumen (i.e., signal anchor type II or SAII activity) (27). Assays with glycosylation reporter fusion proteins expressed either in vivo or in vitro showed that mutant TM X was less efficiently glycosylated than wild type TM X-containing proteins, suggesting that the G4031D mutation interferes with the SAII activity of TM X (Figures 2 and 3). In addition, previous work had shown that the addition of TM XI and intervening loop sequences to TM X significantly increased the membrane association of TM X. This was interpreted to suggest that the membrane integration of TM X requires multiple topogenic determinants and occurs via a cooperative mechanism involving TM XI (27). To determine whether the G4031D mutation affected this cooperative phenotype, loop-TM XI sequences were added to mutant TM X. Addition of TM XI only slightly improved the membrane association of mutant TM X, in contrast to its significantly positive effect on the membrane association of wild type TM X (Figure 4). Thus, the G4031D mutation not only reduces the membrane association and translocation of TM X alone, but also diminishes the membrane association of the last two TM domains together.</p><p>Predictive algorithms represent an alternative method to the experimental testing of the effect of a missense mutation on the topology of a transmembrane domain. Using the Phobius (www.ebi.ac.uk/Tools/phobius/) and TMHMM (www.cbs.dtu.dk/services/TMHMM-2.0/) protein topology prediction programs, neither the G4031D nor the M3677T variant is predicted to eliminate or to alter the topology of its respective TM domain (data not shown). For G4031D, however, the probability for residues within the domain to be membrane-embedded is reduced, and there is a suggestion that the membrane-spanning portion may be shifted C-terminal-ward, and shortened. In addition, the probabilities for the preceding and following loops to be located inside and outside of the cell, respectively, are less definitive with G4031D than with G4031. Any apparent discrepancy between experimental and predictive results may be due to their inherent strengths and weaknesses. For example, although our glycosylation reporter approach involves the overexpression, or the in vitro expression, of artificial, truncated PC1 constructs, it is a valid method to biochemically assess the membrane association and topology of a predicted TM domain. While topology prediction programs are invaluable bioinformatic tools, it is important to recognize that it is experimentally derived structural data (e.g., from NMR, X-ray crystallography, gene fusions, cysteine substitution, N-linked glycosylation, and protease protection assays) that are used as training sets for developing these algorithms, and that are also used as test sets to determine the accuracy of the program. Currently, there is no topology prediction program that is 100% accurate (39). Some of this inaccuracy stems from an inability to factor in the influence of one domain upon another, the distances between TM domains, the regulation of translocation by other factors, and the amphipathic nature of TM domains internal to the protein.</p><p>The data in this report support the view that G4031D would interfere with the membrane insertion of TM X and TM XI in vivo and thereby represents a pathogenic mutation. Several possibilities exist as to what the molecular pathogenic effects of G4031D could be. For example, inhibition of the membrane integration of TM X and TM XI would result in the cytosolic localization of these TM domains and intervening loop sequences and could alter the structure or environment of the cytosolic C-tail, potentially influencing its cleavage and disrupting the multiple signaling functions ascribed to this portion of PC1 (10, 13–19). Furthermore, the cytoplasmic exposure of sequences normally embedded within the membrane might lead to gratuitous activation of non-specific signaling effectors, resulting in a dominant gain of function. An intriguing idea is that the membrane integration of (wild type) TM X and XI may be "purposefully" inefficient and/or regulated (40). In this scenario, alternative, membrane-integrated conformations of PC1, consisting of 11 or 9 TM forms, each with different signaling properties, could be synthesized at the ER. G4031D would favor the predominance of the 9 TM conformation and its signaling activities, which could result in a pathological imbalance in the pathways activated. The L envelope protein of the hepatitis B virus is an example of a protein with dual membrane-integrated structures that are linked to different functions and whose biogenesis is regulated (41). Alternatively, G4031D might interfere with an entirely different biochemical property of TM X, such as its ability to interact with other TM domains. With regard to this, we have noticed that G4031D disrupts a potential GxxxG motif at the N-terminal end of TM X. GxxxG motifs are important for transmembrane domain packing and oligomerization of many types of integral membrane proteins, including GPCRs (42–44). Interestingly, the GxxxG motif is conserved in TM X of human, rat, mouse, and dog PC1, and is replaced by the GxxxG-like motif (44, 45), AxxxA, in chicken and Fugu PC1.</p><p>In summary, our studies provide biochemical evidence that is highly supportive of a pathological, rather than a neutral, effect of G4031D. Currently, we are unable to conclude whether G4031D represents an inactivating or a hypomorphic mutation. These data warrant future studies that will determine whether this amino acid substitution affects the biological properties of PC1 (e.g., cellular localization, signaling, and tubulogenesis) and leads to a disease phenotype.</p><!><p>(A) Illustration of the CD5-11TMM3677T fusion protein with Met to Thr mutation at residue 3677 within TM VI (hatched box). Boxes and roman numerals indicate the 11 TM domains of PC1. CD5, signal sequence; G pro; G protein activation domain; coil, coiled coil domain; N, N-linked glycosylation site. (B) AS19 Western blot analysis of lysates from 293T cells transiently transfected with CD5-11TM (WT), triple glycosylation mutant CD5-11TMN123S (N123S), or 3 different CD5-11TMM3677T mutant fusion protein clones (1–3). Arrow, glycosylated form; arrowhead, non-glycosylated form.</p><!><p>(A) Illustration of wild type CD5-I-X and mutant CD5-I-XG4031D glycosylation reporter constructs. Both fusion proteins contain TM domains I-X of PC1 and have the 3 potential N-linked glycosylation sites within PC1 removed by mutation (S). CD5-I-XG4031D also has the engineered Gly to Asp mutation within TM X (hatched box). CD5, signal sequence from CD5 protein; PRL-N3-HA, glycosylation reporter tag. (B) Anti-HA Western blot analysis of constructs transfected into 293T cells at a low DNA (10 ng) amount. The glycosylated (arrow) and non-glycosylated (arrowhead) forms are indicated.</p><!><p>(A) Illustration of wild type flag-X and mutant flag-XG4031D glycosylation reporter tag fusion constructs. Hatched box, G4031D mutant TM domain X; flag, N-terminal FLAG epitope tag; PRL-N3-HA, glycosylation reporter tag. (B) Gel analysis of flag-X and flag-XG4031D fusion proteins produced in TNT extracts with microsomal membranes and incubated with (+) or without (−) NgF. Arrow, glycosylated form. (C) Gel analysis of flag-X and flag-XG4031D TNT reactions following centrifugation in Tris sucrose buffer. Both the membrane pellet (P) and supernatant (S) fractions were analyzed along with an aliquot of the total TNT reaction (Tot). Arrow, glycosylated form. (D) Interpretation of the glycosidase and membrane association studies. The location and orientation of the predominant forms of the wild type and mutant flag-X fusion proteins are shown relative to the microsomal membrane (gray rectangle). N-linked glycosylation is illustrated by the branched structures on the N3 portion of the glycosylation reporter tag. SAII, signal anchor II orientation.</p><!><p>(A) Illustration of wild type flag-X-XI and mutant flag-XG4031D-XI glycosylation reporter tag fusion constructs. Hatched box, TM X with G4031D mutation; flag, N-terminal FLAG epitope tag; PRL-N3-HA, glycosylation reporter tag. (B) Gel analysis of flag-X-XI and flag-XG4031D-XI fusion proteins produced in TNT extracts with microsomal membranes following centrifugation in Tris sucrose buffer. Both the membrane pellet (P) and supernatant (S) fractions were analyzed along with an aliquot of the total TNT reaction (Tot).</p>

PubMed Author Manuscript

Translocator Protein Ligands Based on N-Methyl-(quinolin-4-yl)oxypropanamides with Properties Suitable for PET Radioligand Development

Modifications to an N-methyl-(quinolin-4-yl)oxypropanamide scaffold were explored to discover leads for developing new radioligands for PET imaging of brain TSPO (translocator protein), a biomarker of neuroinflammation. Whereas contraction of the quinolinyl portion of the scaffold or cyclization of the tertiary amido group abolished high TSPO affinity, insertion of an extra nitrogen atom into the 2-arylquinolinyl portion was effective in retaining sub-nanomolar affinity for rat TSPO, while also decreasing lipophilicity to within the moderate range deemed preferable for a PET radioligand. Replacement of a phenyl group on the amido nitrogen with an isopropyl group was similarly effective. Among others, compound 20 (N-methyl-N-phenyl-2-[2-(pyridin-2-yl)-1,8-naphthyridin-4-yloxy]propanamide) appears especially appealing for PET radioligand development, based on high selectivity and high affinity (Ki = 0.5 nM) for rat TSPO, moderate lipophilicity (logD = 2.48), and demonstrated amenability to labeling with carbon-11.

translocator_protein_ligands_based_on_n-methyl-(quinolin-4-yl)oxypropanamides_with_properties_suitab

INTRODUCTION<!>RESULTS and DISCUSSION<!>Chemistry<!>Ligand Pharmacology: TSPO Binding Affinity, cLogD, and LipE<!>Pharmacological screen<!>Radioligand Syntheses<!>Lipophilicity<!>CONCLUSIONS<!>Materials and Methods<!>N-(3-Acetylpyridin-2-yl)picolinamide (12)<!>N-(2-Acetylphenyl)pyrimidine-2-carboxamide (13)<!>2-(Pyridin-2-yl)-1,8-naphthyridin-4(1H)-one (14)<!>2-(Pyrimidin-2-yl)quinolin-4(1H)-one (15)<!>2-Bromo-N-isopropyl-N-methylpropanamide (16)<!>2-Bromo-N-(4-fluorophenyl)-N-methylpropanamide (17)<!>N-Phenyl-2-[2-(pyridin-2-yl)-1,8-naphthyridin-4-yloxy]propanamide (18)<!>N-Phenyl-2-[2-(pyridin-2-yl)quinolin-4-yloxy]propanamide (19)<!>N-Methyl-N-phenyl-2-[2-(pyridin-2-yl)-1,8-naphthyridin-4-yloxy]propanamide (20)<!>N-Methyl-N-phenyl-2-[2-(pyrimidin-2-yl)quinolin-4-yl]oxypropanamide (21)<!>N-Isopropyl-N-methyl-2-[2-(pyridin-2-yl)quinolin-4-yloxy]propanamide (22)<!>N-(4-Fluorophenyl)-N-methyl-2-[2-(pyridin-2-yl)quinolin-4-yloxy]propanamide (23)<!>Methyl 2-(1-methoxyethylideneamino)nicotinate (24)<!>2-Methoxy-1,8-naphthyridin-4(1H)-one (25)<!>4-Chloro-2-methoxy-1,8-naphthyridine (26)<!>2-(2-Methoxy-1,8-naphthyridin-4-yloxy)-N-methyl-N-phenylpropanamide (27)<!>2-(2-Aminoquinolin-4-yloxy)-N-methyl-N-phenylpropanamide (28)<!>2-[2-(1H-Pyrrol-1-yl)quinolin-4-yloxy]-N-methyl-N-phenylpropanamide (29)<!>3-Chloro-1,5,5-trimethylpyrrolidin-2-one (31)<!>3-Chloro-1-isopropylpyrrolidin-2-one (32)<!>1,5,5-Trimethyl-3-[2-(pyridin-2-yl)quinolin-4-yloxy]pyrrolidin-2-one (33)<!>1-Isopropyl-3-[2-(pyridin-2-yl)quinolin-4-yloxy]pyrrolidin-2-one (34)<!>2-Nitro-N-(pyridin-2-ylmethyl)aniline (35)<!>N-Isopropyl-N-methyl-2-(2-(pyridin-2-yl)-1H-benzo[d]imidazol-1-yloxy)propanamide (36)<!>[11C]Methyl triflate<!>[11C]11<!>[11C]20<!>LogD determinations<!>Determination of ligand binding affinities for rat brain TSPO<!>Pharmacological Screening<!>

<p>Translocator protein 18 kDa (TSPO), formerly known as the peripheral benzodiazepine receptor [1], is located predominantly at the outer mitochondrial membrane in association with a voltage-dependent anion channel and an adenine nucleotide transporter [2]. TSPO is present in several major organs, and is particularly dense in adrenal gland, heart, kidney, and testis [2]. Low amounts are present in normal human brain, primarily in microglia [3]. Activated microglia upregulate TSPO in instances of neuronal damage [4] as seen in many neurological disorders [5–7] including Alzheimer's disease, movement disorders, stroke, multiple sclerosis, and major depression [8]. Therefore, TSPO can serve as an important biomarker for neuroinflammation. Moreover, ligands for TSPO have also been explored as possible drugs, particularly for anxiety [9].</p><p>For more than three decades, PET imaging of human TSPO has been carried out with [11C]PK11195 ([11C]1) [10] or its (R)-enantiomer ([11C](R)-1) [11] (Chart 1) for biomedical investigations of neuroinflammation. [11C](R)-1 has been by far the most employed radioligand for this purpose despite limited brain uptake [12], low specific binding [12], and an undesirable metabolic profile [13]. Efforts to tackle these shortcomings of [11C](R)-1 have resulted in several new structural classes of TSPO radioligand with superior imaging characteristics (Chart 1). Examples include [11C]PBR28 ([11C]2) [14,15], [11C]DAA1106 ([11C]3) [16], [11C]DPA-713 ([11C]4) [17], [18F]DPA-714 ([18F]5) [18], [18F]FBR ([18F]6) [19], [18F]PBR111 ([18F]7) [20], [18F]FEPPA ([18F]8) [21], [18F]FEMPA ([18F]9) [22], and [11C]ER176 ([11C]10) [23]. Nonetheless, many of these new radioligands also suffer particular deficiencies, most prevalent of which is sensitivity to the rs6971 polymorphism in human subjects [20,21,24,25].</p><p>Successful PET radioligands for imaging specific proteins in brain are required to display a wide array of properties [26]. Among these properties are: i) high affinity and selectivity for the target protein; ii) low molecular weight; iii) intermediate polar surface area for blood-brain barrier penetration; iv) moderate lipophilicity for adequate brain entry in the absence of excessive non-specific binding; and v) amenability to labeling with a positron-emitter, either carbon-11 (t1/2 = 20.4 min) or fluorine-18 (t1/2 = 110 min). This study aimed to develop TSPO ligands as leads with a desirable combination of properties for PET radioligand development. We have previously explored a series of N-methyl-(quinolin-4-yl)oxypropanamides as prospective TSPO ligands [27], and encouragingly many of these ligands have shown low TSPO genotype sensitivity in vitro. Here, we further explore structure-affinity relationships in this structural class. High-affinity TSPO ligands emerged from this effort and a few of these are promising new leads to PET radioligands.</p><!><p>In this study, we identified new leads to PET radioligands for imaging TSPO based on modifications to the previously reported [27] N-methyl-(quinolin-4-yl)oxypropanamide TSPO ligand scaffold (Chart 2). These modifications were aimed at exploring, i) variation of substituents on the amide nitrogen, ii) introduction of nitrogen into the quinolin-4-yl group or pendant aryl ring, iii) replacement of the pendant aryl ring with methoxy, 2-pyrimidinyl or N-pyrrolidinyl, iv) the effect of cyclization to eliminate amide bond rotation, and v) contraction of the bicyclic quinolinyl nucleus. Generally, PET radioligands are required to have high affinity with KD in the low nM range, and moderate lipophilicity with measured (or computed) logD in the 2–4 range [26]. Most of the changes that we made to the lead ligand scaffold were intended to retain the very high TSPO affinity (Ki = 0.07 nM for rat TSPO) seen in the previously reported example 11 (Chart 2; scaffold with Y = 2-pyridinyl, and R = Ph), as well as to decrease ligand computed lipophilicity (clogD) from 4.73 towards the desirable range. Usually, the overall shape of the scaffold was modified little to retain high affinity, although the effects of scaffold pruning were also investigated. The main strategy for lowering lipophilicity was to introduce nitrogen into one or more of the aryl rings. The lipophilicity cost for high ligand affinity may be indexed as a lipophilicity efficiency parameter (LipE), defined as ligand pIC50 (or pKi) minus clogD [28]. Therefore, our overall aim encompasses the discovery of ligands with high LipE scores (>6). The amide N-methyl substituent was retained in all new ligands as a site that should be amenable to labeling with carbon-11 through 11C-methylation of N-desmethyl precursors.</p><!><p>As prospective TSPO ligands, the 2-heteroaryl-4-alkoxyquinolines 18–23 (Scheme 1) were synthesized in three steps, proceeding with acylations of the requisite 2-aminoarylethanones [29]. The resultant amides 12 and 13 were subjected to Camps cyclization [30] to yield the 2-heteroarylquinolin-4-ones 14 and 15, respectively, which were then chemoselectively O-alkylated to the desired ligands 18–23 (Scheme 1). The required α-bromoamides 16 and 17 were made under Schotten-Baumann conditions.</p><p>The 2,4-dialkoxy-1,8-naphthyridine ligand 27 was made by first accomplishing a regioselective methoxylation [31,32] at C2 of 2,4-dichloro-1,8-naphthyridine to yield 2-methoxy-4-chloro-1,8-naphthyridine 26 (Method A, Scheme 2). The position of methoxylation was confirmed by an alternative but much lower yielding regiospecific synthesis wherein methyl 2-aminopyridine-3-carboxylate was first converted into the methyl acetimidate 24. Dieckmann condensation [33] of 24 then gave 2-methoxy-1,8-naphthyridin-4-one 25. Compound 25 was then converted into 26 by treatment with phosphoryl chloride (Method B, Scheme 2), but in only 3% yield. Finally, alkoxylation at C4 of 26 gave 27 (Scheme 2).</p><p>The 2-pyrrolequinoline ligand 29 was made by chemoselective O-alkylation [34] of 2-amino-4-hydroxyquinoline to yield 28 followed by treatment of its free base with dimethoxytetrahydrofuran in acetic acid (Scheme 3) [35].</p><p>Two analogs of ligand 22, namely 33 and 34, in which rotation of the amide bond was eliminated through cyclization, were synthesized in two steps. First, an appropriate lactam was subjected to α-chlorination with hexachloroethane to give compounds 31 and 32 in moderate yields [36,37]. Alkylation of 2-(pyridin-2-yl)-quinolin-4(1H)-one with each of these α-chlorolactams required prolonged heating but eventually gave the desired ligands 33 and 34, respectively, in moderate yields (Scheme 4).</p><p>Finally, a ring-contracted analog of 11, namely 36, was made through nucleophilic amination [38] of 1-fluoro-2-nitrobenzene to yield 2-nitro-N-(pyridin-2-ylmethyl)aniline 35, followed by treatment of 35 with two equivalents of sodium hydride to give the cyclized intermediate benzoimidazole-N-oxide, and finally alkylation [39] of this oxide in situ with α-bromoamide 16 (Scheme 5).</p><!><p>The binding affinities (1/KD) of all TSPO ligands, including the previously reported ligand 11, were determined on rat brain homogenates (Table 1). We tested all ligands as racemates only, with the assumption that the high-affinity enantiomer has twofold higher affinity than that recorded for the racemate [27].</p><p>The N-(p-fluorophenyl) analog 23 of the previously described high-affinity TSPO ligand 11 maintained very high binding affinity and a comparably high LipE value (Table 1). Ligand 23 opens up the possibility to prepare an 18F-labeled ligand, according to modern 18F-labeling techniques [40,41].</p><p>Previously, we achieved an increase in LipE by introducing a nitrogen atom at either the 2' position of the pendant phenyl group or the 8 position of the quinolinyl scaffold (Chart 2) [26]. We presumed that by having nitrogen atoms at both positions in the same ligand that we might significantly improve LipE, and, indeed, this turned out to be the case in ligand 20 (Table 1). We obtained a similar result when the pendant phenyl group alone was modified to have nitrogen atoms at positions 2' and 6', as in ligand 21 (Table 1). Previously, we had established the tolerance of the TSPO binding pocket for ligands containing nitrogen atoms at positions 2', 3', and 4' of the pendant aryl group of the lead scaffold [26]. Nitrogen at the 1' position in a pyrrolyl ring was no exception, yielding ligand 29 with high affinity (Ki = 0.22 nM) and high LipE (6.3) (Table 1). In addition, we had found that removal of the pendant phenyl group in the scaffold does not seriously affect binding affinity and that LipE was maintained as a result of a decrease in lipophilicity (clogD) by about 2 units [27]. Seeking to exploit the apparent benefit of reducing the size of this substituent and concomitantly molecular weight, we swapped the large and lipophilic pendant aryl group in 20 for the small and polar methoxy group (Table 1, 27). This replacement conserved high LipE despite some decrease in TSPO affinity and established that TSPO is quite tolerant for such substitutions.</p><p>Tertiary amides are present in nearly all high-affinity TSPO ligands (e.g., see Chart 1). We sought further improvements to LipE by manipulation of substituents on the tertiary amido group. Previously, we also reported some TSPO ligands based on the scaffold in Chart 2 that incorporated a methylene tether to the aryl ring in place of the oxygen tether [27]. As a group, these methylene-tethered ligands had lower LipE scores than the analogous oxygen-tethered ligands. Because LipE improved by about 1 unit when a phenylamide was exchanged for an isopropylamide in the methylene-tethered series [27], we expected a similar improvement in the oxo-linked series, and indeed this was found when comparing the new ligand 22 with 11 (Table 1).</p><p>Previously, we had found that the elimination of amide bond rotation in a methylene-tethered TSPO ligand, 1-methyl-3- [(2-phenylquinolin-4-yl)methyl]pyrrolidin-2-one (30), increased LipE by about 1 unit [26]. Therefore, we were interested to exploit this effect again by making a locked amide rotamer of ligand 22. There are two possible ligands that may result from tethering the α-methyl group in 22 to one of its amide substituents, namely the ligands 34 and 35 (Table 1). Unlike the locked rotamer from our earlier study, both 34 and 35 lacked any appreciable binding affinity for TSPO (Table 1). These results add to previous observations of locked amide rotamers showing much reduced affinity for TSPO [42]. Moreover, evidence indicates that TSPO has a preference to bind the E-rotamer of PK11195 (1) [43,44]. Thus, it seems that TSPO is highly sensitive to the spatial arrangement of the substituents on the requisite tertiary amide in TSPO ligands.</p><p>Finally, we tested a truncated scaffold analog of 11. This ligand (Table 1, 36) has a fused bicyclic scaffold whose structure is rather similar to that found in many literature TSPO ligands, as exemplified by DPA-713 (4), PBR111 (7) and IGA-1 (37) [45] (Chart 3). Ligand 36 appears to occupy essentially the same chemical space as these compounds, so we were surprised that 36 showed no binding to TSPO (Ki > 1000 nM). However, simple comparison of the formal 2-dimensional structure of 36 with those of other known TSPO ligands with similar core scaffold shows that the amido carbonyl group, a key pharmacophoric element, is not aligned with the carbonyl groups of the other ligands. Hence, the inability of 36 to form a required directional hydrogen bond with TSPO may explain its lack of affinity.</p><!><p>Ligand 11, at 10 µM concentration, showed <50% inhibition of specific binding to all tested receptors and transporters, except the 5-HT1A receptor (Ki = 366 nM).</p><!><p>The radioligands [11C]11 and [11C]20 were produced in practically useful radiochemical yields in only five minutes at room temperature by treating their respective N-desmethyl precursors, 19 and 18, with [11C]methyl triflate [46] (Scheme 6). Each radioligand was separated with reversed phase HPLC and readily formulated in sterile ethanol with 10% saline, for possible use in PET scanning. Each radioligand was obtained with high radiochemical purity (>99%) and with high specific activities (244 and 126 GBq/µmol, respectively).</p><!><p>The accessibility of [11C]11 and [11C]20 afforded the opportunity to measure their lipophilicities accurately for comparison to their computed values. Prolonged incubation of these radioligands in phosphate buffer (pH 7.4) at room temperature left them unchanged chemically. Although the measured logD for [11C]11 (3.11) was appreciably lower than the computed value (4.18), the measured logD of [11C]20 (2.48) was quite similar to the computed value (2.8) (Table 1). The computed values are therefore likely to be reasonable estimates (within one log unit of true values), and are also expected to be in approximately correct rank order for the closely related compounds. The measured values for both radioligands placed them near or within the desirable logD range of 2–4.</p><!><p>Modifications to N-methyl-(quinolin-4-yl)oxypropanamides, especially introduction of nitrogen into the 2-arylquinolin-4-yl scaffold, and use of an isopropyl substituent instead of a phenyl substituent on the tertiary amido nitrogen, were effective in retaining LipE to within a desirable range for PET radioligands. Cyclization of the amido group or contraction of the quinolinyl ring abolished high affinity. Three compounds (20–22) showed favorable properties for PET radioligand development, including high TSPO affinity and moderate computed lipophilicity. Ligands 11 and 20 were shown further to be amenable to labeling with carbon-11, to have acceptable moderate measured lipophilicity, and to be selective for binding to TSPO.</p><!><p>Literature methods were used to prepare ligand 11 [27], 2-(pyridin-2-yl)-quinolin-4(1H)-one [30], 2-bromo-N-methyl-N-phenylpropanamide [47], 2-hydroxy-N-methyl-N-phenylpropanamide [48], 1,5,5-trimethylpyrrolidin-2-one [49], and 1-isopropylpyrrolidin-2-one [50]. All other reagents and solvents were purchased. Air-sensitive reagents were stored under N2 in a PureLab HE glovebox (Innovative Technology; Amesbury, MA). Melting points were determined on an SMP30 apparatus (Stuart; Staffordshire, UK). Boiling point vacuum pressures were determined on a DDR-1200 apparatus (J-Kem Scientific Inc.; St. Louis, MO). Reactions in dry solvent were performed with dried reagents under an inert atmosphere. Solutions were taken to dryness by treatment with MgSO4 (unless stated otherwise), followed by filtration and evaporation. 1H (400 MHz), 13C NMR (100 MHz), and 19F NMR (376 MHz) spectra were recorded on an Avance 400 instrument (Bruker; Billerica, MA). Chemical shifts for 19F are reported relative to neat TFA in a coaxial insert (δ = −76.6). HRMS data were obtained at the University of Illinois Urbana-Champaign (Mass Spectrometry Laboratory, School of Chemical Sciences) with a Micromass Q-Tof Ultima instrument for ESI (Waters Corp.; Milford, MA). Preparative HPLC was performed with elution at 30 mL/min on either a Luna PFP(2) column (5 µm; 100 Å; 30 × 250 mm; Phenomenex; Torrance, CA), or a Gemini C18 column (10 µm; 110 Å; 30 × 250 mm; Phenomenex). Mixtures were separated on either an XBridge C18 column (5 µm; 130 Å; 4.6 × 250 mm; Waters: Milford, MA), a Luna C18(2) column (10 µm; 100 Å; 4.6 × 250 mm; Phenomenex), or an XTerra RP18 column (10 µm; 125 Å; 7.8 × 300 mm; Waters). Compound purities were established on either a Luna PFP(2) column (5 µm; 100 Å; 4.6 × 250 mm; Phenomenex) or a Gemini C18 column (5 µm; 110 Å; 4.6 × 250 mm; Phenomenex). All compounds were >95% pure and typically >99% pure, as monitored by absorbance at 220 nm. Radioligands were isolated with HPLC on an XBridge C18 column (5 µm; 130 Å; 10 × 250 mm; Waters) or a Luna C18(2) column (10 µm; 100 Å; 10 × 250 mm; Phenomenex). Radiochemical purities were determined with HPLC on an on a Gold HPLC apparatus (Beckman Coulter, Inc.; Fullerton, CA) equipped with an in-line Flow-Count NaI scintillation detector (Bioscan, Inc.; Washington, DC) and a UV absorbance detector (Beckman Coulter, Inc.) operating at 254 nm. The response of the HPLC system was calibrated for mass of ligand to enable radioligand specific activity to be determined from the detected mass of carrier in a known activity of radioligand. Radioactive decay events were counted in an automatic gamma counter (1480 Wizard 3; PerkinElmer Life Sciences: Wallac Oy; Turku, Finland) with an electronic window set between 360 and 1,800 keV. cLogD values were computed with Pallas for Windows software version 3.8 in default option (CompuDrug; Bal Harbor, FL).</p><!><p>Pyridine-2-carbonyl chloride hydrochloride (4.89 g, 27.5 mmol) was added portion-wise to a solution of 1-(2-amino-3-pyridinyl)-1-ethanone (5.15 g, 37.8 mmol) and TEA (9.2 mL, 65.0 mmol) in dry CHCl3 (100 mL) at 4 °C. The mixture turned from tan to deep blue-green and the temperature rose to 15 °C. After 29 h, the mixture was diluted with CHCl3 (100 mL) and washed with hydrochloric acid (1 M; 100 mL × 2), water (100 mL), and brine (100 mL × 2), and then dried. The product was recrystallized (MTBE–CHCl3) to give a tan powder (714 mg, 11%). A second recrystallization (MTBE–MeOH) gave 12 as a cream-white powder. mp 151–154 °C. HRMS-ESI (m/z): [M + H]+ calcd for C13H12N3O2, 242.0930; found, 242.0929. 1H NMR (CDCl3): δ 13.22 (bs, 1H), 8.79 (dq, J = 4.8, 0.8 Hz, 1H), 8.74 (dd, J = 4.8, 2.0 Hz, 1H), 8.35 (td, J = 8.0, 1.2 Hz, 1H), 8.24 (dd, J = 8.0, 2.0 Hz, 1H), 7.91 (dt, J = 8.0, 2.0 Hz, 1H), 7.51 (ddd, J = 7.6, 4.8, 1.2 Hz, 1H), 7.18 (dd, J = 7.6, 4.8 Hz, 1H), 2.71 (s, 3H). 13C NMR (CDCl3): δ 200.2, 162.6, 153.2, 151.1, 150.1, 148.5, 139.8, 137.5, 126.7, 123.2, 119.3, 118.5, 28.0.</p><!><p>DMF (30 mL) was added dropwise to a slurry of sodium pyrimidine-2-carboxylate (5.03 g, 34.4 mmol), 2′-aminoacetophenone (4.6 mL, 37 mmol), DIPEA (6.5 mL, 37 mmol) and PyBroP (17.4 g, 37.4 mmol) in CH2Cl2 (40 mL) held in a water bath, wherein the temperature rose from 21 to 29 °C. As the colorless slurry dissolved, the solution turned from orange to a deep forest green over the course of 18 h. The solvent was removed. The residue was taken up in EtOAc (370 mL), then washed successively with aq KHSO4 (5% w/v, 370 mL × 3), brine (370 mL), aq NaHCO3 (5% w/v, 370 mL × 3), brine (370 mL), and dried (Na2SO4). The residue was washed with ether and dioxane and recrystallized (MTBE–CHCl3; charcoal to decolorize) to give 13 as golden-yellow plates (435 mg, 5%). mp 201–203 °C. HRMS-ESI (m/z): [M + H]+ calcd for C13H12N3O2, 242.0930; found, 242.0928. 1H NMR (CDCl3): δ 13.65 (bs, 1H), 9.09 (dd, J = 8.4, 0.8 Hz, 1H), 9.03 (d, J = 5.2 Hz, 2H), 7.98 (dd, J = 8.0, 1.6 Hz, 1H), 7.66 (ddd, J = 8.4, 7.6, 1.2 Hz, 1H), 7.50 (t, J = 4.8 Hz, 1H), 7.22 (ddd, J = 8.4, 7.6, 1.2 Hz, 1H), 2.74 (s, 3H). 13C NMR (CDCl3): δ 202.4, 161.3, 158.2, 157.8, 139.9, 135.1, 131.7, 123.2, 122.9, 122.6, 121.2, 28.6.</p><!><p>Compound 12 (716 mg, 2.97 mmol) and NaOH (355 mg, 8.88 mmol) in dry dioxane (40 mL) were heated to 95 °C for 3 h, whereupon the tan solution precipitated a brown solid. The mixture was cooled to rt, and the solid filtered off, washed with toluene (10 mL × 3), and taken up in H2O (20 mL). The blood-red solution was neutralized with hydrochloric acid (1 M, ~7 mL) to precipitate a tan solid, which was filtered off, washed with water (10 mL × 2) followed by ether (10 mL × 2), and then recrystallized (toluene) to give 14 as a pink-tan solid with a cotton candy texture (165 mg). More 14 (60 mg) was obtained from the mother liquor (225 mg total, 34%). mp 205–206 °C. HRMS-ESI (m/z): [M + H]+ calcd for C13H10N3O, 224.0824; found, 224.0829. 1H NMR (CDCl3): δ 10.78 (bs, 1H), 8.76 (td, J = 2.8, 1.2 Hz, 1H), 8.74 (dd, J = 4.4, 1.6 Hz, 1H), 8.68 (dd, J = 7.6, 1.6 Hz, 1H), 8.01 (td, J = 7.2, 0.8 Hz, 1H), 7.92 (dt, J = 7.6, 1.6 Hz, 1H), 7.47 (ddd, J = 7.2, 4.8, 0.8 Hz, 1H), 7.35 (dd, J = 8.0, 3.6 Hz, 1H), 7.00 (br s, 1H). 13C NMR (CDCl3): δ 179.8, 153.6, 150.1, 149.3, 148.2, 145.6, 137.7, 135.5, 125.5, 120.9, 120.5, 120.0, 107.2.</p><!><p>Compound 13 (407 mg, 1.69 mmol) and NaOH (202 mg, 5.05 mmol) in dry dioxane (15 mL) were heated in a pressure vessel to 110 °C for 3 h, during which time the pale yellow solution turned yellow-brown and a red-orange solid precipitated. After cooling the mixture to rt, the solid was filtered off, washed with dioxane, taken up in water (20 mL), and brought to pH 5 with acetic acid, whereupon the orange solution gave a salmon-pink precipitate. The solid was filtered off, washed with water, and recrystallized (toluene) to give 15 as a light pink solid (187 mg, 50%). mp 258 °C dec. HRMS-ESI (m/z): [M + H]+ calcd for C13H10N3O, 224.0824; found, 224.0827. 1H NMR (DMSO-d6, keto–enol; 1.4:1.0): δ 12.14 (bs, 1H, keto), 9.10 (d, J = 4.8 Hz, 2H, keto + enol), 8.12 (d, J = 8.8 Hz, 2H, keto + enol), 7.74–7.68 (m, 2H, keto + enol), 7.37 (dt, J = 6.8, 0.8 Hz, 1H, keto + enol), 7.15 (s, 1H, keto + enol), 3.34 (s, 1H, enol). 13C NMR (DMSO-d6): δ 177.7 (keto), 158.3 (enol), 158.0 (keto), 144.7 (enol), 144.6, 140.3 (enol), 140.2 (keto + enol), 132.2 (keto + enol), 125.8 (keto + enol), 124.7 (keto + enol), 123.6 (keto + enol), 122.3 (keto + enol), 119.6 (keto + enol), 108.0 (keto + enol).</p><!><p>N-Isopropylmethylamine (1.0 mL, 9.8 mmol) was added to a biphasic solution of KOH (1.67 g, 29.7 mmol) in water (10 mL) and EtOAc (10 mL). This mixture was stirred rapidly and cooled to 4 °C whereupon 2-bromopropanoyl chloride (1.5 mL, 15 mmol) was added dropwise with the temperature kept below 11 °C. The ice bath was removed and stirring continued for 1.5 h. The organic phase was separated off and the aqueous phase extracted with EtOAc (5 mL × 2). The combined extracts were washed with brine (10 mL) and dried. The residue was distilled (bp 48–50 °C at 0.2 mmHg), but was still contaminated with 2-bromopropanoic acid. The oil was taken up in CHCl3 (10 mL), washed with NH4OH (5 mL × 3), water (5 mL), and brine (5 mL × 2), and then dried to give 16 as a colorless oil (588 mg, 29%). d 1.38 g/mL. HRMS-ESI (m/z): [M + H]+ calcd for C7H15BrNO, 208.0337; found, 208.0341. 1H NMR (CDCl3, cis–trans; 1.4:1.0): δ 4.84 (sept, J = 6.8 Hz, 1H, cis), 4.60 (q, J = 6.4 Hz, 1H, trans), 4.53 (q, J = 6.4 Hz, 1H, cis), 4.20 (sept, J = 6.8 Hz, 1H, trans), 2.91 (s, 3H, cis), 2.81 (s, 3H, trans), 1.84 (d, J = 5.6 Hz, 3H, trans), 1.83 (d, J = 6.4 Hz, 3H, trans), 1.27 (d, J = 6.8 Hz, 3H, trans), 1.21 (d, J = 6.8 Hz, 3H, cis), 1.12 (d, J = 7.2 Hz, 3H, cis), 1.10 (d, J = 6.8 Hz, 3H, cis). 13C NMR (CDCl3): δ 168.6 (cis), 168.4 (trans), 48.3 (trans), 44.8 (trans), 39.5 (cis), 38.5 (trans), 28.3 (cis), 26.6 (trans), 22.1 (trans), 21.7 (cis), 20.8 (trans), 19.9 (trans), 19.4 (cis), 18.8 (cis).</p><!><p>The method for 16 was applied to 4-fluoro-N-methylaniline (3.5 mL, 29 mmol) and 2-bromopropanoyl chloride (4.4 mL, 44 mmol). The crude product oil was fractionally distilled. The forerun (bp 37–39 °C at 7.2 mmHg; 0.2 mL) was discarded and 17 (6.7 g, 88%) was collected as a yellow oil with bp 84–88 °C at 5.7 mmHg. d 1.5 g/mL. HRMS-ESI (m/z): [M + H]+ calcd for C10H12BrFNO, 260.0086; found, 260.0090. 1H NMR (CDCl3): δ 7.29 (m, 2H), 7.14 (m, 2H), 4.22 (q, J = 6.8 Hz, 1H), 3.28 (s, 3H), 1.74 (d, J = 6.4 Hz, 3H). 13C NMR (CDCl3): δ 169.6, 162.1 (d, J = 247 Hz), 138.8 (d, J = 3 Hz), 129.1 (d, J = 9 Hz), 116.9 (d, J = 22 Hz), 38.8, 38.2, 21.7. 19F NMR (CDCl3): δ −113.0.</p><!><p>NaH (20 mg, 0.50 mmol; 60% in mineral oil) was added to a solution of 14 (60 mg, 0.27 mmol) in dry DMSO (2 mL). The orange effervescent solution was stirred for 1 h, whereupon 2-bromo-N-phenylpropanamide (192 mg, 0.84 mmol) was added. The mixture was heated to 80 °C for 8 h, cooled to rt and quenched with water (15 mL). The precipitate was filtered off, washed with water and purified with HPLC [PFP column; MeOH–NH4CH3CO2 buffer (25 mM, pH 5); 70:30] to give a peach solid (14 mg, 14%), which was recrystallized (toluene–MTBE) to give 18. mp 222–223 °C. HRMS-ESI (m/z): [M + H]+ calcd for C22H19N4O2, 371.1508; found, 371.1503. 1H NMR (CDCl3): δ 9.13 (dd, J = 4.0, 2.0 Hz, 1H), 8.75 (d, J = 8.0 Hz, 1H), 8.74 (ddd, J = 4.8, 2.8, 0.8 Hz, 1H), 8.57 (dd, J = 8.4, 2.0 Hz, 1H), 8.44 (bs, 1H), 8.20 (s, 1H), 7.82 (dt, J = 7.6, 1.6 Hz, 1H), 7.61 (dd, J = 8.8, 1.2 Hz, 2H), 7.46 (dd, J = 8.4, 4.4 Hz, 1H), 7.38–7.31 (3H), 7.14 (tt, J = 7.6, 0.8 Hz, 1H), 5.37 (q, J = 6.8 Hz, 1H), 1.85 (d, J = 6.8 Hz, 3H). 13C NMR (CDCl3): δ 168.4, 160.5, 160.5, 157.0, 154.8, 154.0, 149.0, 137.0, 136.9, 131.2, 129.1, 125.0, 124.9, 122.4, 121.5, 120.4, 116.1, 100.2, 75.7, 18.2.</p><!><p>2-(Pyridin-2-yl)quinolin-4(1H)-one (222 mg, 1.00 mmol), 2-bromo-N-phenylpropanamide (255 mg, 1.12 mmol), and K2CO3 (834 mg, 6.03 mmol) in MeCN (35 mL) were heated to 55 °C for 8 h. The mixture was cooled to rt and poured into water (175 mL). The precipitate was filtered off and recrystallized (aq dioxane) to give 19 as a colorless powder (283 mg, 77%). mp 205–206 °C. 1H NMR (CDCl3): δ 8.70 (qd, J = 4.8, 0.8 Hz, 1H), 8.63 (md, J = 9.2, 0.8 Hz, 1H), 8.29 (dd, J = 8.4, 0.8 Hz, 1H), 8.22 (bs, 1H), 8.17 (d, J = 8.4 Hz, 1H), 8.13 (s, 1H), 7.85 (dt, J = 8.0, 2.0 Hz, 1H), 7.80 (ddd, J = 7.2, 5.6, 0.6 Hz, 1H), 7.61 (ddd, J = 8.4, 6.8, 0.6 Hz, 1H), 7.55–7.52 (2H), 7.35 (ddd, J = 7.6, 4.8, 1.2 Hz, 1H), 7.34–7.29 (2H), 7.12 (m, 1H), 5.40 (q, J = 6.8 Hz, 1H), 1.87 (d, J = 6.8 Hz, 3H). 13C NMR (CDCl3): δ 168.9, 159.9, 157.6, 155.7, 149.3, 149.1, 136.9, 136.9, 130.3, 129.8, 129.1, 126.5, 124.9, 124.3, 121.7, 121.2, 121.2, 120.2, 99.6, 75.2, 18.4.</p><!><p>2-Bromo-N-methyl-N-phenylpropanamide (122 mg, 0.50 mmol), 14 (100 mg, 0.45 mmol), and K2CO3 (371 mg, 2.69 mmol) in dry MeCN (10 mL) were heated to 55 °C for 18 h. The mixture was cooled to rt and filtered through diatomaceous earth. The solvent was removed, the residue taken up in CH2Cl2 (10 mL), extracted into hydrochloric acid (2 M; 30 mL × 2), and neutralized with satd. NaHCO3. The solution was extracted with CH2Cl2 (50 mL × 3) and the combined organic layers washed with water (50 mL) and brine (50 mL × 2), and then dried. The product was isolated by HPLC [PFP column; MeOH–phosphate buffer (25 mM, pH 6); 80:20] followed by recrystallization (cyclohexane–dioxane) to give 20 as a cream solid (17 mg, 10%). mp 193– 194 °C. HRMS-ESI (m/z): [M + H]+ calcd for C23H21N4O2, 385.1665; found, 385.1664. 1H NMR (CDCl3): δ 9.08 (dd, J = 4.4, 2.4 Hz, 1H), 8.90 (d, J = 8.0 Hz, 1H), 8.78 (d, J = 4.0 Hz, 1H), 8.66 (dd, J = 8.0, 2.0 Hz, 1H), 8.03 (s, 1H), 7.89 (dt, J = 8.0, 2.0 Hz, 1H), 7.68 (bs, 2H), 7.50 (t, J = 7.6 Hz, 2H), 7.45–7.38 (3H), 5.08 (q, J = 6.4 Hz, 1H), 3.33 (s, 3H), 1.69 (d, J = 6.4 Hz, 3H). 13C NMR (CDCl3): δ 169.9, 161.4, 159.9, 157.0, 155.5, 153.7, 148.6, 142.5, 137.0, 132.3, 130.3, 128.8, 127.8, 124.7, 122.5 121.0, 116.3, 99.7, 71.7, 50.7, 38.1, 18.4.</p><!><p>2-Bromo-N-methyl-N-phenylpropanamide (216 mg, 0.89 mmol), 15 (179 mg, 0.80 mmol), and K2CO3 (1.04 g, 7.53 mmol) in dry MeCN (10 mL) were heated to 55 °C for 4.5 h, whereupon the colorless slurry turned yellow. The mixture was cooled to rt and filtered through diatomaceous earth. The solvent was removed and the residue taken up in CH2Cl2 (100 mL), washed with water (50 mL × 2) followed by brine (50 mL), dried, and recrystallized (cyclohexane–dioxane) to give 21 as a colorless solid (263 mg, 86%). mp 211–212 °C. HRMS-ESI (m/z): [M + H]+ calcd for C23H21N4O2, 385.1665; found, 385.1665. 1H NMR (CDCl3): δ 9.02 (d, J = 4.8 Hz, 2H), 8.32 (d, J = 8.4 Hz, 1H), 8.26 (d, J = 8.4 Hz, 1H), 7.87 (s, 1H), 7.72 (dt, J = 6.8, 1.2 Hz, 1H), 7.55– 7.51 (3H), 7.41–7.37 (3H), 7.32 (t, J = 7.2 Hz, 1H), 5.09 (q, J = 6.8 Hz, 1H), 3.34 (s, 3H), 1.69 (d, J = 6.4 Hz, 3H). 13C NMR (CDCl3): δ 170.0, 163.7, 161.0, 157.6, 155.3, 149.4, 142.6, 130.2, 130.1, 130.0, 128.6, 127.7, 126.6, 122.1, 121.7, 120.6, 100.6, 71.5, 38.2, 18.3.</p><!><p>The method for 21 was applied to 2-(pyridin-2-yl)-quinolin-4(1H)-one (222 mg, 1.00 mmol) and 16 (170 µL, 1.12 mmol). The crude product was recrystallized (aq EtOH) to give 22 as white sea urchin-shaped clusters (219 mg, 63%). mp 175–176 °C. HRMS-ESI (m/z): [M + H]+ calcd for C21H24N3O2, 350.1869; found, 350.1868. 1H NMR (CDCl3, cis–trans; 1.1:1.0): δ 8.67–8.64 (4H, cis + trans), 8.31 (d, J = 8.4 Hz, 2H, cis + trans), 8.10 (d, J = 8.4 Hz, 2H, cis + trans), 7.95 (s, 1H, trans), 7.90 (s, 1H, cis), 7.85 (tt, J = 7.6, 2.0 Hz, 2H, cis + trans), 7.72 (dd, J = 6.8, 1.6 Hz, 1H, cis), 7.71 (dd, J = 6.8, 1.6 Hz, 1H, trans), 7.53 (dd, J = 6.8, 1.2 Hz, 1H, cis), 7.51 (dd, J = 6.8, 1.2 Hz, 1H, trans), 7.35–7.31 (2H, cis + trans), 5.55 (q, J = 6.8 Hz, 1H, trans), 5.41 (q, J = 6.4 Hz, 1H, cis), 4.91 (sept, J = 6.8 Hz, 1H, cis), 4.35 (sept, J = 6.4 Hz, 1H, trans), 3.02 (s, 3H, cis), 2.84 (s, 3H, trans), 1.78 (d, J = 6.4 Hz, 3H, trans), 1.77 (d, J = 6.8 Hz, 3H, cis), 1.38 (d, J = 6.4 Hz, 3H, trans), 1.17 (d, J = 6.4 Hz, 3H, trans), 1.16 (d, J = 6.8 Hz, 3H, cis), 1.10 (d, J = 6.8 Hz, 3H, cis). 13C NMR (CDCl3): δ 169.4 (cis), 169.2 (trans), 160.8 (cis), 160.7 (trans), 157.3 (trans), 157.2 (cis), 156.2 (trans), 156.1 (cis), 149.2 (cis + trans), 148.9 (cis + trans), 136.8 (cis + trans), 129.9 (cis), 129.3 (trans), 125.9 (cis + trans), 124.1 (cis + trans), 122.1 (cis + trans), 121.7 (trans), 121.6 (cis), 121.3 (cis + trans), 98.7 (cis), 98.6 (trans), 72.3 (cis), 72.1 (trans), 47.7 (trans), 44.8 (cis), 39.5 (cis), 38.5 (trans), 28.3 (cis), 26.6 (trans), 22.1 (trans), 21.7 (cis), 20.8 (trans), 19.9 (trans), 19.4 (cis), 18.8 (cis).</p><!><p>The method for 21 was applied to 2-(pyridin-2-yl)-quinolin-4(1H)-one (222 mg, 1.00 mmol) and 17 (195 µL, 1.12 mmol). The crude product was recrystallized (aq dioxane) to give white crystals. These were dried in an Abderhalden pistol under high vacuum (T = 110 °C) in the presence of P2O5 for 1 d to give 23 (293 mg, 73%). mp 179–180 °C. HRMS-ESI (m/z): [M + H]+ calcd for C24H21FN3O2, 402.1618; found, 402.1613. 1H NMR (CDCl3): δ 8.75 (d, J = 4.0 Hz, 1H), 8.70 (d, J = 8.0 Hz, 1H), 8.27 (dd, J = 8.4, 0.8 Hz, 1H), 8.08 (d, J = 8.0 Hz, 1H), 7.88 (dt, J = 7.6, 1.6 Hz, 1H), 7.82 (s, 1H), 7.70 (dt, J = 8.0, 1.2 Hz, 1H), 7.68 (br s, 2H), 7.50 (dt, J = 8.0, 0.8 Hz, 1H), 7.38 (ddd, J = 7.6, 4.8, 1.2 Hz, 1H), 7.15 (dt, J = 8.0, 1.2 Hz, 2H), 5.04 (q, J = 6.4 Hz, 1H), 3.31 (s, 3H), 1.87 (d, J = 6.4 Hz, 3H). 13C NMR (CDCl3): δ 170.3, 162.3 (d, J = 248 Hz), 160.7, 156.9, 156.2, 149.2, 148.6, 138.6 (d, J = 3 Hz), 137.0, 130.0, 129.8 (d, J = 8 Hz), 129.1, 125.9, 124.2, 122.3, 121.8, 121.3, 117.2 (d, J = 22 Hz), 98.6, 71.1, 38.3, 18.5. 19F NMR (CDCl3): δ −112.9.</p><!><p>Methyl 2-aminopyridine-3-carboxylate (5.17 g, 34.0 mmol) and acetic anhydride (20 mL) were heated to 110 °C for 1 h in a mixture of trimethyl orthoacetate (50 mL) and acetic anhydride (20 mL), whereupon the colorless solution turned yellow. After 1 h, methyl acetate began to distill off and heating was continued for 5 h. The excess reagents were removed in vacuo leaving a red oil and white syrup. This was taken up in Et2O (100 mL) and washed with aq Na2CO3 (2 M, 50 mL × 2), water (50 mL), and brine (50 mL × 2), and then dried. The residue was purified by Kugelrohr distillation (140–160 °C at 1.3 mmHg) to yield 24 as a yellow oil which smelled like sugar snap peas (2.00 g, 28%). HRMS-ESI (m/z): [M + H]+ calcd for C10H13N2O3, 209.0926; found, 209.0930. 1H NMR (CDCl3): δ 8.51 (dd, J = 4.8, 1.6 Hz, 1H), 8.21 (dd, J = 7.6, 2.0 Hz, 1H), 7.06 (dd, J = 8.0, 4.8 Hz, 1H), 3.87 (s, 3H), 3.86 (s, 3H), 1.87 (s, 3H). 13C NMR (CDCl3): δ 165.9, 163.1, 160.7, 152.3, 140.0, 118.3, 117.6, 53.9, 52.1, 17.2.</p><!><p>A solution of 24 (1.9 g, 9.1 mmol) in dry THF (10 mL) was added dropwise to a slurry of LDA (10 mmol) in THF–hexane (10 mL) at −50 °C under Ar, whereupon the solution turned bright orange. After 1 h, the temperature was raised to 0 °C. The solution was stirred for another 30 min and quenched with cold satd. NH4Cl solution (25 mL) followed by Na2CO3 solution (2 M, 50 mL). The aqueous layer was separated off, washed with Et2O (50 mL × 2), and carefully neutralized with hydrochloric acid (1 M) to give a white precipitate. Product remaining in the mother liquor was extracted into BuOH (50 mL × 4), which was washed with brine (50 mL × 2), and dried. The product was isolated by HPLC [PFP column; MeOH–NH4HCO2 buffer (25 mM, pH 4); 35:65] to give 25 as a cream solid (1.2 g, 76%). mp ~160 °C (dec; if ramping was omitted: ~200 °C dec). HRMS-ESI (m/z): [M + H]+ calcd for C9H9N2O2, 177.0664; found, 177.0665. 1H NMR (HFIP-d2): δ 8.82 (dd, J = 8.0, 1.2 Hz, 1H), 8.64 (dd, J = 4.8, 1.6 Hz, 1H), 7.57 (dd, J = 8.0, 5.2 Hz, 1H), 6.27 (s, 1H), 4.11 (s, 3H). 13C NMR (HFIP-d2): δ 164.2, 149.9, 146.8, 136.9, 119.5, 116.7, 90.6, 55.3.</p><!><p>Method A: A slurry of 2,4-dichloro-1,8-naphthyridine (4.95 g, 24.9 mmol) in dry toluene (50 mL) was added to a slurry of NaOMe (5.0 g, 93 mmol) in toluene (50 mL) at rt. The temperature rose to 32 °C as the yellow solid dissolved to give a brown solution [Note-NaOMe should be broken up periodically if needed]. After 17 h, the mixture was filtered through diatomaceous earth, and washed with toluene. The solvent was removed and the residue recrystallized (aq EtOH) to give 26 as fine, light yellow needles (3.87 g). Concentration of the mother liquor yielded more product (250 mg; 4.12 g; total, 85%). mp 134–135 °C. HRMS-ESI (m/z): [M + H]+ calcd for C9H8ClN2O, 195.0325; found, 195.0327. 1H NMR (CDCl3): δ 8.99 (dd, J = 4.4, 2.0 Hz, 1H), 8.48 (dd, J = 8.0, 2.0 Hz, 1H), 7.45 (dd, J = 8.0, 4.4 Hz, 1H), 7.12 (s, 1H), 4.16 (s, 3H). 13C NMR (CDCl3): δ 164.5, 155.4, 153.6, 143.7, 133.6, 120.5, 118.2, 114.0, 54.5. Method B: Compound 25 (1.20 g, 6.82 mmol) in POCl3 (10 mL) was heated to 110 °C under Ar for 19 h to give an orange solution. The solution was cooled and carefully quenched by slow addition to water (100 mL) and neutralized with NH4OH. This product was extracted into CHCl3 (100 mL), washed with water (100 mL) followed by brine (100 mL × 2), and then dried. The product was isolated by HPLC [Gemini column; MeOH– NH4HCO2 buffer (25 mM, pH 7); 65:35] of the residue gave 26 as a pale yellow solid (41 mg, 3%). mp 132–133 °C.</p><!><p>2-Hydroxy-N-methyl-N-phenylpropanamide (197 mg, 1.10 mmol) and NaH (44 mg, 1.10 mmol; 60%) were stirred in dry DMF (1.0 mL) at rt for 4 h. The red solution was transferred via cannula into a mixture of 26 (195 mg, 1.00 mmol) in DMSO–DMF (1:1; 2.0 mL). The reaction stalled after 30 min so the product was isolated with HPLC [PFP column; MeOH–NH4HCO2 buffer (25 mM, pH 7); 75:25] followed by [Gemini column; MeOH–NH4HCO2 buffer (25 mM, pH 7); 70:30] to yield 27 as small, colorless needles (112 mg, 33%). mp 151–152 °C. HRMS-ESI (m/z): [M + H]+ calcd for C19H20N3O3, 338.1505; found, 338.1499. 1H NMR (CDCl3): δ 8.88 (dd, J = 4.4, 2.0 Hz, 1H), 8.39 (dd, J = 8.4, 2.0 Hz, 1H), 7.43–7.34 (3H), 7.28–7.25 (3H), 5.94 (s, 1H), 4.88 (q, J = 6.8 Hz, 1H), 4.11 (s, 3H), 3.33 (s, 3H), 1.61 (d, J = 6.8 Hz, 3H). 13C NMR (CDCl3): δ 169.3, 165.9, 161.5, 156.0, 152.9, 142.3, 132.0, 130.2, 128.7, 127.2, 119.0, 113.8, 92.5, 71.9, 54.1, 38.2, 17.9.</p><!><p>2-Amino-4-hydroxyquinoline hydrate was dried in an Abderhalden pistol under high vacuum (T = 110 °C) in the presence of P2O5. A slurry of this quinoline (3.70 g, 23.1 mmol) with sodium ethoxide (24 mmol) in ethanol (9.1 mL) was heated to 70 °C, whereupon a red solution formed. The solvent was distilled off, leaving a white solid, which was dried under vacuum until the internal temperature returned to 70 °C. Dry DMF (20 mL) was added to the solid and the slurry heated to 90 °C. 2-Bromo-N-methyl-N-phenylpropanamide (5.87 g, 24.3 mmol) was added in one portion. Heating was continued for 30 min, whereupon NaBr gradually precipitated. The mixture was cooled to rt poured into water (200 mL), and extracted into CH2Cl2 (150 mL × 3). The organic extracts were washed with water (40 mL × 5), brine (250 mL × 2), and finally dried (K2CO3). Recrystallization (aq EtOH) of the white solid gave light yellow crystals, which were dried in an Abderhalden pistol under high vacuum (T = 110 °C) in the presence of P2O5 to give 28 (4.50 g, 61%). mp 182–184 °C. HRMS-ESI (m/z): [M + H]+ calcd for C19H20N3O2, 322.1556; found, 322.1552. 1H NMR (CDCl3): δ 7.83 (d, J = 8.0 Hz, 1H), 7.54 (q, J = 8.0 Hz, 1H), 7.54 (s, 1H), 7.30–7.28 (3H), 7.19–7.14 (3H), 5.67 (s, 1H), 4.90 (q, J = 6.4 Hz, 1H), 4.67 (s, 2H), 3.31 (s, 3H), 1.60 (d, J = 6.4 Hz, 3H). 13C NMR (CDCl3): δ 169.7, 161.0, 158.0, 148.6, 142.4, 130.2, 129.9, 128.4, 127.2, 125.3, 122.3, 121.9, 117.7, 90.5, 71.9, 38.4, 17.8.</p><!><p>2,5-Dimethoxytetrahydrofuran (195 µL, 1.50 mmol) was added to a pale yellow solution of 28 (321 mg, 1.00 mmol) in acetic acid (4 mL) at 120 °C (bath temp.), causing the solution to turn red. This solution was heated for 25 min, cooled to rt, poured into CH2Cl2 (100 mL), and washed with NH4OH (50 mL × 2), water (50 mL), and brine (50 mL × 2), and finally dried. The product was isolated with HPLC [PFP column; MeCN–NH4CH3CO2 buffer (25 mM, pH 7); 70:30] to give 29 as a pale pink solid (90 mg, 24%). mp 130 °C dec. HRMS-ESI (m/z): [M + H]+ calcd for C23H22N3O2, 372.1712; found, 372.1713. 1H NMR (CDCl3): δ 8.07 (dd, J = 8.0, 0.4 Hz, 1H), 7.88 (d, J = 8.4 Hz, 1H), 7.88 (ddd, J = 8.4, 7.2, 1.6 Hz, 1H), 7.52 (d, J = 4.4 Hz, 1H), 7.52 (s, 1H), 7.40 (ddd, J = 8.0, 6.8, 1.2 Hz, 1H), 7.37–7.31 (3H), 7.23–7.21 (2H), 6.41 (s, 1H), 6.40 (t, J = 2.0 Hz, 2H), 5.04 (q, J = 6.4 Hz, 1H), 3.32 (s, 3H), 1.67 (d, J = 6.4 Hz, 3H). 13C NMR (CDCl3): δ 169.4, 161.9, 150.7, 148.0, 142.3, 130.7, 130.2, 128.7, 127.9, 127.3, 124.7, 122.4, 120.0, 118.6, 111.4, 91.8, 71.8, 38.3, 18.1.</p><!><p>t-Butyl lithium (1.64 M; 17 mmol) in pentane (2.3 mL) was added dropwise to a solution of 2-bromomesitylene (fractionally distilled from CaH2) (2.8 mL, 18 mmol) in dry THF (30 mL) held between −70 °C and −60 °C, whereupon a white precipitate formed. The suspension was warmed to −20 °C and then cooled back to −70 °C. A solution of 1,5,5-trimethylpyrrolidin-2-one (fractionally distilled from BaO) (2.3 mL, 17 mmol) in dry THF (25 mL) was added dropwise. The reaction mixture was warmed to 0 °C for 30 min, and then transferred dropwise via a cannula into a solution of hexachloroethane (4.17 g, 17.6 mmol) in dry THF (40 mL) held between −78 °C and −70 °C. After 1 h, the solution was warmed to rt and stirred for another 1 h. The solvent was removed and the residue taken up in CH2Cl2 (200 mL), and then filtered through diatomaceous earth several times until the solution was no longer cloudy. The solvent was removed and the residual yellow oil shaken with water (100 mL), causing separation of a white oil. The liquid layer was decanted from this oil, filtered through diatomaceous earth, and extracted with CHCl3 (100 mL × 2). The extracts were washed with brine (100 mL ×2) and dried to give a white waxy solid that was contaminated with 3-bromo-1,5,5-trimethylpyrrolidin-2-one (~13%). This byproduct could not be removed either by recrystallization (hexanes) or by sublimation (65 °C at 0.2 mmHg). Therefore, the solid was taken up in DMF (2 mL) and stirred with LiCl (510 mg, 12 mmol) for several hours. The mixture was diluted with water (10 mL), extracted into CHCl3 (10 mL × 2), washed with brine (10 mL × 2), and dried to give white crystals which were triturated with cold hexanes to give 31 (1.12 g, 41%). mp 74–75 °C. HRMS-ESI (m/z): [M + H]+ calcd for C7H13ClNO, 162.0686; found, 162.0691. 1H NMR (CDCl3): δ 4.45 (ddd, J = 8.4, 5.6, 0.4 Hz, 1H), 2.81 (s, 3H), 2.48 (dd, J = 14.0, 8.4 Hz, 1H), 2.17 (dd, J = 14.0, 5.6 Hz, 1H), 1.37 (s, 3H), 1.24 (s, 3H). 13C NMR (CDCl3): δ 169.2, 58.8, 53.9, 44.3, 26.5, 26.3, 24.9.</p><!><p>This compound was synthesized in the same manner as 31 using 1-isopropylpyrrolidin-2-one (9.3 g, 73 mmol). The workup was modified slightly in that the reaction mixture was quenched with satd. NaHCO3 (100 mL) and filtered through diatomaceous earth. The organic layer was separated off, and the aqueous layer extracted with CHCl3 (100 mL × 3). The combined extracts were washed with brine (300 mL × 2) and dried. The residual oil was worked up as for 31. No brominated byproduct was present so the product was fractionally distilled (bp 68–70 °C at 0.9 mmHg) to give 32 as a colorless oil (8.2 g, 70%). HRMS-ESI (m/z): [M + H]+ calcd for C7H13ClNO, 162.0686; found, 162.0690. 1H NMR (CDCl3): δ 4.38 (dd, J = 7.6, 4.4 Hz, 1H), 4.36 (sept, J = 6.8 Hz, 1H), 3.48 (ddd, J = −13.6, 6.8, 6.4 Hz, 1H), 3.32 (ddd, J = −13.6, 7.6, 4.0 Hz, 1H), 2.53 (dddd, J = −14.0, 7.6, 7.6, 6.4 Hz, 1H), 2.23 (dddd, J = −14.0, 7.6, 4.0, 4.0 Hz, 1H), 1.18 (d, J = 6.8 Hz, 3H), 1.16 (d, J = 6.8 Hz, 3H). 13C NMR (CDCl3): δ 169.4, 55.7, 43.4, 39.4, 29.8, 19.8, 19.4.</p><!><p>2-(Pyridin-2-yl)-quinolin-4(1H)-one (222 mg, 1.00 mmol), 31 (210 mg, 1.40 mmol), and K2CO3 (866 mg, 6.28 mmol) in dry MeCN (35 mL) were heated to 80 °C for 70 h. The mixture was cooled to rt, poured into water (300 mL) and kept at 4 °C overnight. The precipitate that formed was recrystallized (hexanes–Et2O) to give 33 as large white prisms (201 mg, 58%). mp 131–133 °C. HRMS-ESI (m/z): [M + H]+ calcd for C21H22N3O2, 348.1712; found, 348.1712. 1H NMR (CDCl3): δ 8.72 (ddd, J = 4.8, 1,6, 0.8 Hz, 1H), 8.66 (d, J = 8.0 Hz, 1H), 8.26 (dd, J = 8.4, 0.8 Hz, 1H), 8.12 (s, 1H), 8.12 (d, J = 8.4 Hz, 1H), 7.90 (dt, J = 8.0, 2.0 Hz, 1H), 7.71 (ddd, J = 8.4, 6.8, 1.6 Hz, 1H), 7.50 (ddd, J = 8.4, 6.8, 1.2 Hz, 1H), 7.36 (ddd, J = 7.6, 4.8, 1.2 Hz, 1H), 5.40 (dd, J = 7.6, 5.6 Hz, 1H), 2.90 (s, 3H), 2.67 (dd, J = 13.6, 8.0 Hz, 1H), 2.19 (dd, J = 13.6, 8.0 Hz, 1H), 1.40 (s, 3H), 1.37 (s, 3H). 13C NMR (CDCl3): δ 169.3, 161.1, 157.2, 156.3, 149.1, 149.0, 136.9, 130.0, 129.2, 125.9, 124.1, 122.2, 121.9, 121.5, 99.5, 75.0, 58.3, 41.3, 27.3, 26.6, 24.6.</p><!><p>2-(Pyridin-2-yl)-quinolin-4(1H)-one (222 mg, 1.00 mmol), 32 (160 µL, 1.12 mmol), and K2CO3 (834 mg, 6.03 mmol) in dry MeCN (35 mL) were heated to 80 °C for 7 d. The red mixture was cooled to rt, filtered through diatomaceous earth, taken up into CH2Cl2 (100 mL), washed with water (50 mL), and brine (50 mL × 2), and then dried. The product was isolated with HPLC [PFP column; MeOH–NH4CH3CO2 buffer (25 mM, pH 7); 75:25] followed by trituration with aq MeOH (50% v/v) to give 34 as a white solid (167 mg, 48%). mp 108–110 °C. HRMS-ESI (m/z): [M + H]+ calcd for C21H22N3O2, 348.1712; found, 348.1712. 1H NMR (CDCl3): δ 8.70 (ddd, J = 4.8, 1,6, 0.8 Hz, 1H), 8.66 (ddd, J = 8.0, 1.2, 1.2 Hz, 1H), 8.28 (dd, J = 8.4, 1.2 Hz, 1H), 8.14 (s, 1H), 8.10 (d, J = 8.4 Hz, 1H), 7.86 (dt, J = 8.0, 2.0 Hz, 1H), 7.71 (ddd, J = 8.4, 6.8, 1.2 Hz, 1H), 7.50 (ddd, J = 8.0, 6.8, 0.8 Hz, 1H), 7.34 (ddd, J = 7.2, 4.8, 1.2 Hz, 1H), 5.38 (dd, J = 7.6, 7.2 Hz, 1H), 4.50 (sept, J = 6.8 Hz, 1H), 3.48 (ddd, J = −13.2, 10.0, 3.6 Hz, 1H), 3.42 (ddd, J = −13.6, 7.2, 2.4 Hz, 1H), 2.83 (dddd, J = −13.6, 10.0, 7.6, 3.6 Hz, 1H), 2.26 (dddd, J = −13.6, 8.8, 7.2, 1.6 Hz, 1H), 1.26 (d, J = 6.8 Hz, 3H), 1.22 (d, J = 6.8 Hz, 3H). 13C NMR (CDCl3): δ 169.1, 161.2, 157.3, 156.3, 149.2, 148.9, 136.9, 130.0, 129.2, 125.9, 124.1, 122.3, 121.8, 121.5, 99.5, 75.9, 43.2, 38.5, 26.2, 19.8, 19.5.</p><!><p>2-(Aminomethyl)pyridine (4.9 mL, 48 mmol), 1-fluoro-2-nitrobenzene (5.5 mL, 52 mmol), and K2CO3 (13 g, 95 mmol) in DMF (35 mL) were heated for 19 h (90 °C, bath temp.). The blood red mixture was then allowed to cool, poured into citrate buffer (pH 7; 450 mL), and extracted with CH2Cl2 (250 mL × 2). The organic layer was then washed with water (350 mL × 5), followed by brine (350 mL), and then dried. The residue was sonicated in Et2O (100 mL × 2), filtered through diatomaceous earth, and taken up into rapidly stirring hydrochloric acid (0.1 M; 500 mL). The biphasic mixture was filtered through diatomaceous earth, and the aqueous layer was separated off, washed with Et2O (200 mL × 2), and neutralized with NH4OH. The resultant precipitate was recrystallized twice (aq EtOH) to yield 35 as tiny golden needles (862 mg, 8%). mp 92–93 °C. HRMS-ESI (m/z): [M + H]+ calcd for C12H12N3O2, 230.0930; found, 230.0937. 1H NMR (CDCl3): δ 8.87 (s, 1H), 8.64 (ddd, J = 4.8, 1.6, 0.8 Hz, 1H), 8.21 (dd, J = 8.4, 1.6 Hz, 1H), 7.68 (ddd, J = 7.6, 7.6, 1.6 Hz, 1H), 7.40 (ddd, J = 8.4, 6.8 Hz, 1H), 7.32 (d, J = 8.0 Hz, 1H), 7.23 (ddd, J = 7.6, 4.8, 0.8 Hz, 1H), 6.83 (dd, J = 8.4, 0.8 Hz, 1H), 6.68 (ddd, J = 8.4, 6.8, 1.2 Hz, 1H), 4.68 (d, J = 5.2 Hz, 1H). 13C NMR (CDCl3): δ 156.6, 149.6, 145.0, 136.9, 136.2, 132.4, 126.9, 122.6, 121.2, 115.8, 114.3, 48.4.</p><!><p>NaH (480 mg, 2.00 mmol) in mineral oil (60% w/v) was added to a solution of 35 (229 mg, 1.00 mmol) in DMF (20 mL) at rt, which turned from bright yellow to deep red-brown. The solution was heated to 55 °C for 1 h. A solution of 16 (315 µL, 2.00 mmol) in DMF (5 mL) was then added and heating continued for another 1 h. The solution was then cooled and the solvent removed under reduced pressure. The residue was taken up into water (100 mL) and extracted with CHCl3 (50 mL × 3). The combined extracts were then washed with water (50 mL) followed by brine (50 mL × 2), and then dried. The product was isolated with HPLC [PFP column; MeOH–NH4HCO2 buffer (25 mM, pH 7); 70:30] to give 36 as an orange syrup (240 mg, 71%). HRMS-ESI (m/z): [M + H]+ calcd for C19H23N4O2, 339.1821; found, 339.1825. 1H NMR (CDCl3, cis–trans; 1.3:1.0): δ 8.76–8.73 (m, 2H, cis + trans), 8.32 (s, 1H, cis), 8.30 (s, 1H, trans), 7.90–7.85 (m, 2H, cis + trans), 7.76–7.72 (m, 3H, 2 cis + trans), 7.68 (m, 1H, trans) 7.41 (ddd, J = 7.6, 4.0, 1.2 Hz, 1H, trans), 7.40 (ddd, J = 7.6, 4.0, 1.2 Hz, 1H, cis), 7.35–7.28 (4H, 2 cis + 2 trans), 6.06 (q, J = 6.4 Hz, 1H, trans), 5.99 (q, J = 6.4 Hz, 1H, cis), 4.83 (sept, J = 6.8 Hz, 1H, cis), 4.13 (sept, J = 6.8 Hz, 1H, trans), 2.74 (s, 3H, trans), 2.60 (s, 3H, cis), 1.63 (d, J = 6.0 Hz, 3H, cis), 1.62 (d, J = 6.0 Hz, 3H, trans), 1.10 (d, J = 6.4 Hz, 3H, trans), 1.04 (d, J = 6.8 Hz, 3H, cis), 0.78 (d, J = 6.8 Hz, 3H, cis), 0.63 (d, J = 6.4 Hz, 3H, trans). 13C NMR (CDCl3): δ 169.5 (cis), 169.4 (trans), 149.2 (cis), 148.7 (trans), 147.3 (trans), 147.2 (cis), 137.8 (cis + trans), 136.9 (cis + trans), 133.7 (trans), 133.5 (cis), 124.7 (trans), 124.6 (cis), 124.3 (cis + trans), 123.3 (cis + trans), 120.2 (trans), 120.1 (cis), 111.1 (cis), 111.0 (trans), 80.8 (cis), 80.1 (trans), 47.7 (trans), 44.5 (cis), 27.8 (cis), 26.6 (trans), 20.8 (trans), 19.9 (trans), 19.3 (cis), 18.7 (cis), 16.8 (trans), 16.6 (cis).</p><!><p>[11C]Methyl triflate was prepared from cyclotron-produced [11C]carbon dioxide, via conversion into [11C]methane by reduction with hydrogen over palladium, direct iodination of [11C]methane, and passage of the generated [11C]methyl iodide over heated silver triflate [43].</p><!><p>Compound 19 (0.55 mg, 1.5 µmol) was treated with [11C]methyl triflate in the presence of aq NaOH (0.5 M, 1 eq.) in dry MeCN (300 µL) at rt for 5 min. The radioligand was isolated with HPLC [XBridge column; MeCN–NH4OH buffer (1 mM, pH 7.7); 65:35] at 6 mL/min (tR = 11 min). The isolated product was taken up in ethanol–saline (10:90, v/v) containing polysorbate 80 (12 mg) and sterile filtered (Millex-MP 0.22 µm, 25 mm). The radiochemical purity of [11C]11 was >99% as established by HPLC [Xbridge column; MeCN– NH4HCO2 buffer (0.1 M, pH 6.3); 50:50] at 2 mL/min (tR = 9.2 min). Product identity was also confirmed by LC-MS of associated carrier. The average decay-corrected radiochemical yield of [11C]11 was 17% from cyclotron-produced [11C]carbon dioxide and the average specific activity was 244 GBq/µmol at the end of synthesis (n = 7), corresponding to about 40 min from the end of radionuclide production.</p><!><p>Compound 18 (0.80 mg, 2.2 µmol) was treated with [11C]MeOTf in the presence of excess NaH in dry MeCN (300 µL) at rt for 5 min The radioligand was isolated by HPLC [Luna column; MeCN–NH4HCO2 (0.1 mM); 45:55] at 6 mL/min (tR = 12 min). The isolated product was formulated as for [11C]11. HPLC analysis of [11C]20, as for [11C]11, showed > 99% radiochemical purity (tR = 6.1 min). LC-MS of associated carrier confirmed product identity. The average decay-corrected radiochemical yield of [11C]20 was 21% from cyclotron-produced [11C]carbon dioxide and the average specific activity was 126 GBq/µmol at the end of radiosynthesis (n = 4).</p><!><p>These were performed on [11C]11 and [11C]20 of high radiochemical purity (>99.4%), for distribution between n-octanol and sodium phosphate buffer (pH 7.4, 0.15 M) at rt, essentially by the methodology that we have described previously [14]. The chemical stabilities of these radioligands in the buffer for the duration of the measurements were verified by HPLC analysis. Neither radioligand was adsorbed on the wall of the test tube during distribution between phases.</p><!><p>All animal experiments were performed in accordance with the Guide for the Care and Use of Laboratory Animals and were approved by the NIMH Animal Care and Use Committee.</p><p>Binding assays were performed as previously described [27], except that crude rat brain homogenates were used instead of mitochondrial fractions. Data were analyzed with Prism 5 nonlinear regression curve-fitting software (GraphPad Prism; San Diego, CA). Briefly, whole rat brains from Sprague-Dawley rats were homogenized in cold HEPES buffer (50 mM; pH 7.4) with a Teflon pestle and Glas-Col Homogenizing System. The homogenates were centrifuged at 20,000g for 15 min at 4 °C. The pellets were then resuspended, aliquotted into various vials, and stored at −80 °C. A self-displacement assay on 1 was used as a control along with each assay of test ligand with [3H]1 as reference radioligand. The individually calculated control KD values for 1 were compared to the reported value of 0.707 nM [27] as an assurance of the correctness of results obtained on test ligands. The KD value of 0.707 nM for 1 was used as the dissociation constant to calculate Ki values for test ligands.</p><!><p>Compound 11 was screened at the Psychoactive Drug Screening Program [51] for inhibition of binding at 10 µM concentration (n = 4) to a wide range of human receptors: AMY 1A, 1B, 1D, 2A, 2B and 2C; benzodiazepine; β2; serotonin 1A, 1B, 1D, 1E, 2A, 2B, 2C, 3, 5A, 6, and 7; dopamine 2–5; opiate δ, κ, and µ; muscarinic 1–5; σ1R and σ2R; GABAA; histamine 1 and 4; and human transporters (DAT, NET, SERT).</p><!><p>This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.</p><p> Author Contributions </p><p>This paper was composed with contributions from all authors. All authors approved the final version of the manuscript. CB, synthesized, and characterized the compounds; KJJ tested the ligands in vitro; SSZ measured the lipophilicities; CLM prepared the radioligands; RBI supervised in vitro tests; VWP initiated and supervised the project.</p><p> Conflict of Interest </p><p>Each author declares there were not any actual or potential conflicts of interest that could have influenced this study.</p><p> Appendix A. Supplementary data </p><p>Supplementary data related to this article can be found at</p>

PubMed Author Manuscript

Electrophilic reactivities of cyclic enones and α,β-unsaturated lactones

The reactivities of cyclic enones and a,b-unsaturated lactones were characterized by following the kinetics of their reactions with colored carbon-centered reference nucleophiles in DMSO at 20 C. The experimentally determined second-order rate constants k 2 were analyzed with the Mayr-Patz equation, lg k ¼ s N (N + E), to furnish the electrophilicity descriptors E for the Michael acceptors. Cyclic enones and lactones show different reactivity trends than their acyclic analogs. While cyclization reduces the reactivity of enones slightly, a,b-unsaturated lactones are significantly more reactive Michael acceptorsthan analogously substituted open-chain esters. The observed reactivity trends were rationalized through quantum-chemically calculated Gibbs energy profiles (at the SMD(DMSO)/M06-2X/6-31+G(d,p) level of theory) and distortion interaction analysis for the reactions of the cyclic Michael acceptors with a sulfonium ylide. The electrophilicities of simplified electrophilic fragments reflect the general reactivity pattern of structurally more complex terpene-derived cyclic enones and sesquiterpene lactones, such as parthenolide.

electrophilic_reactivities_of_cyclic_enones_and_α,β-unsaturated_lactones

Introduction<!>Product studies<!>Kinetics<!>Application of electrophilicity parameters in synthesis<!>Quantum chemical calculations<!>Rate constants toward glutathione (GSH)<!>Conclusions

<p>Cyclic carbonyl compounds with a,b-unsaturated positions are important motifs within many natural products (Chart 1). [1][2][3] Previous studies of their cellular reactivities with endogenous proteins revealed intriguing insights into their target proles. 4,5 The ability of these biomolecules to react as electrophiles with nucleophilic sites furnishes them with a multitude of biological functions, 6,7 e.g. the recently reported inhibition of focal adhesion kinase 1 by parthenolide, 5a,b the cytotoxic activity of dehydroleucodine against human leukemia cells, 8 or the ability of nimbolide to inhibit metastasis. 5c Nature has structurally tailored the reactivity of a,b-unsaturated cyclic carbonyl compounds in different variants. In particular, a-methylene-gbutyrolactones exhibit superior cellular protein binding compared to lactones with endocyclic p-system, likely associated with an elevated reactivity. 4a,9 For the sesquiterpene lactones costunolide and dehydrocostus lactone, 10 a,b-unsaturated d-lactones such as leptomycin, fostriecin or the anguinomycins, [11][12][13][14][15] as well as for simple fragments, such as tulipalin A, 16 it has been analyzed that their biological activities mainly depend on the ability to alkylate biomacromolecules through Michael additions. Sometimes these Michael additions are coupled with subsequent steps to achieve irreversible covalent enzyme inhibition. 15 On the other hand, in modied rugulactones the a,b-unsaturated d-lactone unit does not Chart 1 Examples for electrophilic natural products.</p><p>contribute to the antibacterial effects and bioactivities of rugulactone were instead assigned to the reactivity of the a,b-unsaturated ketone unit. 17 Despite these insights into proteome reactivity, a systematic analysis of the individual electrophilicity of the Michael acceptor moieties in different natural products or their truncated analogs is lacking. Knowledge of the reactivity of such biologically occurring electrophilic fragments would facilitate the identication of pharmacophores and is, therefore, of fundamental interest in biochemistry, toxicology, medicinal chemistry, and drug discovery. 9,18 Moreover, Michael acceptors with endo-and exocyclic unsaturation are also a structural motif of signicant importance for synthetic chemists. 3 In life-sciences, rate constants for the reactions of electrophiles with glutathione (GSH) are frequently used for estimating the reactivity and potential toxicity of various electrophilic compounds. [19][20][21][22][23] However, the most comprehensive overview of polar organic reactivity is currently given by Mayr and co-workers who used eqn (1) to characterize the reactivities of more than 1200 nucleophiles and over 300 electrophiles in solution phase. 24 lg k 2 (20 C) ¼ s N (N + E)</p><p>Eqn ( 1) is a linear free energy relationship that allows for the semi-quantitative prediction of second-order rate constants k 2 for the reactions of electrophiles with nucleophiles from three parameters: the electrophilicity parameter E and the solventdependent nucleophilicity parameters N and s N (susceptibility).</p><p>Recently, we determined the nucleophilic reactivity parameters N and s N of GSH in aqueous solution, which facilitates to interconnect both approaches. Bioassay-derived GSH kinetics can now be used to roughly estimate Mayr electrophilicity parameters E, and vice versa. In this way, Mayr electrophilicities E for more than 70 acyclic Michael acceptors were estimated based on their previously determined kinetics toward GSH. 25 More precise electrophilicities E for a series of structurally simple acyclic Michael acceptors were determined from the kinetics of their reactions with carbon-centered one-bond nucleophiles (reference nucleophiles), that is, mainly with pyridinium and sulfonium ylides. 26,27 We now set out to determine the Mayr electrophilicity parameters E of cyclic enones 1-3 and a,b-unsaturated lactones 4-5 by studying the kinetics of their reactions with the reference nucleophiles 6-7 (Chart 2). We then tested whether the Mayr E parameters obtained for the electrophilic core structures 1-5 are also representative of the reactivity prole of structurally more complex natural products that bear these fragments in their molecular scaffold. Quantum-chemical calculations were used to rationalize the observed reactivity trends which signicantly differ from those for analogous acyclic (open chain) ketones and esters.</p><!><p>The formal 1,3-dipolar cycloadditions (Huisgen reactions) of simple electron-decient alkenes with pyridinium ylides, generated from N-alkylated pyridinium salts under basic conditions, are well-known to yield tetrahydroindolizines. Subsequent oxidation (e.g. with air or chloranil) efficiently aromatizes the newly formed heterocycles to afford diversely substituted indolizines. [27][28][29][30][31] In contrast, formation of the analogous tricyclic cyclopenta-, cyclohexa-, or cyclohepta-indolizines has rarely been studied. Only Tamura reported the formation of cyclohexaindolizines in low yield (10%) in a vinylic substitution reaction that used the pyridinium ylide 6c (R ¼ CO 2 Et) and 3chlorocyclohexanone as educts. 32 Direct 1,3-dipolar cycloaddition reactions of pyridinium ylides with cyclic enones or a,bunsaturated lactones have not been reported to the best of our knowledge.</p><p>We planned to use the pyridinium ylides 6 as colored reference nucleophiles to follow the kinetics of their reactions with cyclic Michael acceptors by photometric methods. Given the lack of knowledge about the outcome of these reactions, we decided to characterize the products of a subset of the electrophile/nucleophile combinations under the conditions of the kinetic experiments, that is, in DMSO at 20 C (Scheme 1).</p><p>Treatment of a 1 : 1-mixture of the pyridinium salt 6b$HY (HY ¼ HCl, HBr) and sodium carbonate with a DMSO solution of cyclopentenone (1a, 2 equiv.) resulted in a (3+2)-cycloaddition to give a mixture of diastereomeric tetrahydroindolizines. Due to their high sensitivity toward oxidation 33 and to facilitate the product purication, we oxidized these initial adducts to the aromatic indolizine 8, which was isolated in 18% yield and characterized by single-crystal X-ray diffraction. We were delighted to nd that analogous reactions of 6b with cyclohexenone (2a), cycloheptenone (3) as well as with the lactone 5a gave the corresponding indolizines 9b, 10, and 11, respectively, in signicantly higher yields (72-86% of isolated products). Furthermore, cyclohexaindolizines 9a and 9c were isolated in high yields from the reactions of the ester-and keto-stabilized pyridinium ylides 6a and 6c with cyclohexenone (2a).</p><p>To diversify the types of reference nucleophiles in our kinetic studies, we also investigated the reactions of the cyclic electrophiles with the sulfonium ylide 7. Treatment of a solution of the sulfonium tetrauoroborate 7$HBF 4 and a cyclic Michael acceptor in DMSO with potassium tert-butoxide generated the sulfonium ylide 7 which then underwent cyclopropanation reactions with the electrophiles 1a, 2a, 4a, and 5a (Scheme 2). The cyclopropanes 12-15 were obtained as mixtures of diastereomers. Separation of the diastereomers by column chromatography was not always possible. However, puried diastereomers of 12 and 15 could be crystallized and characterized by single-crystal X-ray crystallography, providing unequivocal evidence for the cyclopropanation reaction.</p><p>The lactones 4b and 5b reacted with the sulfonium ylide 7 at their exo-methylene groups to give diastereomeric mixtures of 16 (46%) and 17 (69%), respectively. Owing to their sufficiently different polarity these diastereomeric mixtures were separable by column chromatography. One diastereomer of 16 and one of 17 were crystallized and analyzed by single-crystal X-ray diffraction (Scheme 2).</p><p>As a general trend, the yields of the cyclopropanes 12-17 depended on two factors: (a) the excess and (b) the absolute concentration of the electrophiles. A survey of the reaction conditions showed that highest yields were obtained when the cyclic Michael acceptor was present in excess (up to 10 equiv.) over the pronucleophile 7 and/or at low concentrations (<0.01 M). Experimental protocols with higher concentrations of the Michael acceptors or reduced excess (1.5 equiv.) resulted in complete consumption of the colored ylide 7, too, but the cyclopropanes were only formed as minor products under these conditions. Instead, 7 isomerized in a background reaction to furnish the sulde 18, 34,35 presumably through a Sommelet-Hauser type of rearrangement (Scheme 3). 36 The sulde 18 is the starting material for BAY 85-8501, a candidate for the treatment of inammatory diseases such as acute lung injury. 35 We characterized 18 by single-crystal X-ray diffraction. Solutions with low concentrations of 7 in DMSO isomerized slower (t 1/2 ¼ 30 min at [7] 0 ¼ 1  10 À4 M) than solutions with higher concentrations of 7 (t 1/2 ¼ 4 min at [7] 0 ¼ 0.057 M, monitored by time-resolved 1 H NMR spectroscopy). Thus, the isomerization of 7 into 18 partially consumed the nucleophile when 7 was combined with weakly or only moderately reactive electrophiles, whose cyclopropanations proceeded at comparable time scale as the Sommelet-Hauser rearrangement of 7.</p><!><p>The kinetics of the reactions of the colorless electrophiles 1-5 with the pyridinium (6) and sulfonium (7) ylides were determined by following the decay of the UV-vis absorbance of the colored nucleophiles. Reactions at the seconds to minutes timescale were followed by conventional photometry. Stopped-ow photometric methods were employed for faster reactions in the millisecond regime. DMSO was used as the solvent for all Scheme 1 Products for the reactions of the pyridinium ylides 6 (generated by deprotonation of 6$HY) with cyclic Michael acceptors (yields of isolated products after chromatography, see ESI † for details). Insert: Single-crystal X-ray structure of 8. Thermal ellipsoids drawn at a 50% probability level.</p><p>Scheme 2 Products for the reactions of the sulfonium ylide 7 with the electrophiles 1-5 (yields of isolated products after column chromatography, thermal ellipsoids drawn at 50% probability level, see ESI † for details). [a] 5 equiv., [b] 10 equiv., [c] electrophile-nucleophile combinations, which were uniformly studied at 20 C.</p><p>Solutions of the ylides 6 and 7 in DMSO were generated by adding stoichiometric amounts of potassium tert-butoxide to the corresponding pyridinium or sulfonium salts. In the next step, these DMSO solutions were mixed with an excess (>10 equiv.) of the electrophiles 1-5. With this ratio of reactants, the concentration of the excess compound can be assumed to remain practically constant during the kinetic measurements, which simplies the kinetics and makes it possible to determine rate constants k obs under pseudo-rst order conditions. In general, the timedependent change of the nucleophile's absorbance followed a mono-exponential decay. The rst-order rate constants k obs were then determined by a least-squares tting of the monoexponential decay function A t ¼ A 0 exp(Àk obs  t) + C to the experimental absorbances A t (Fig. 1A).</p><p>The correlation of k obs with the concentration of the electrophiles 1-5 revealed a linear relationship, the slope of which corresponds to the second-order rate constant k exp 2 (Fig. 1B, Table 1). Isomerization (of 7) and/or decomposition of the colored reference nucleophiles proceed concurrently and impede the kinetic study of slower reactions. Therefore, only the highly reactive nucleophile 6c was available to study the rather unreactive 2-and 3-methylated cyclic enones 1c, 2b, and 2c.</p><p>Based on the set of experimental second-order rate constants k exp 2 , we calculated the electrophilicity parameters E for compounds 1-5 by applying eqn (1) and the reported Mayr nucleophilicity parameters N and s N of the reference nucleophiles. 24d, 30,37 Electrophilicity of natural products Natural products. Cyclic Michael acceptors are frequent moieties in natural products (NPs), such as sesquiterpene lactones. 13,14 We, therefore, set out to assess whether the electrophilicity parameters determined for the simple cyclic Michael acceptors 1-5 (Chart 2) also hold to estimate the correct order of reactivity for analogous, but more complex, natural products. We made use of the reference nucleophiles 6c and 7 to investigate the kinetics of their reactions with different classes of electrophilic natural products with embodied cyclic enone or exo-methylene lactone units (Chart 3). The monoterpenes carvone (19) and verbenone (20) as well as the sesquiterpene nootkatone (21) were used to test the reactivity of naturally occurring cyclic enones. Four sesquiterpene lactones (parthenolide 22, costunolide 23, dehydroleucodine 24, dehydrocostus lactone 25) were chosen to gain insight into the reactivity of exomethylene lactones.</p><p>Product studies. 1 H NMR spectroscopic studies of the products of the reactions of the natural products 22-25 with the nucleophile 7 indicated exclusive cyclopropanation at the amethylene lactone fragments. Neither was nucleophilic opening of the epoxide ring in parthenolide ( 22) nor cyclopropanation of the 3-methylcyclopentenone moiety in dehydroleucodine (24) observed. However, the cyclopropanations of 22-25 by 7 do not proceed with noticeable stereoselectivity and mixtures of up to four diastereomers were obtained, e.g. with dehydroleucodine (24). Luckily, the reaction of 7 with parthenolide (22) furnished a mixture of only two major diastereomeric products aer separation by preparative thin layer chromatography, and single-crystal X-ray crystallography of 26 (arbitrarily taken from the diastereomeric mixture of crystalline material) corroborated the structural assignment on the fundament of our NMR spectroscopic analysis (Scheme 4).</p><p>Reactivity studies. Direct kinetic measurements in DMSO at 20 C and the determination of second-order rate constants k 2 in analogy to those for the fragments 1-5 require access to sufficient quantities of the electrophilic reaction partner used in excess over the colored reference nucleophiles. The available quantities were sufficient to follow this strategy for the natural products 19-22, and the kinetics of their reactions were studied toward the pyridinium ylide 6c as reference nucleophile. The Mayr electrophilicities E of 19-22 (Table 2) were estimated by substituting the experimental second-order rate constants k exp 2 and the known N and s N for 6c in eqn (1). Competition experiments were performed to estimate the Mayr electrophilicities E of the exo-methylene lactones 22-25 (¼ natural products, NP). The lactone 4b (E ¼ À19.4) was chosen as the competition partner because it contains the entire core structure of the electrophilic moiety in the natural products. As outlined in Scheme 5, 38 the experiments were performed such that the reference nucleophile 7 (generated in solution from 7$HBF 4 with KOtBu) was completely consumed in reactions with an excess of the two competing electrophiles NP and 4b. Consequently, the product mixture contained the remaining electrophiles NP and 4b as well as both products, the respective cyclopropanated natural product CNP (from NP + 7) and 16 (from 4b + 7). The reaction mixture was analyzed by 1 H NMR spectroscopy to determine the competition constant k according to eqn (3). The competition constants k were then used to estimate the E parameters for the electrophilic NPs 22-25 (Table 2).</p><p>The reactivity of parthenolide (22) was characterized by both approaches. An electrophilicity E ¼ À19.0 was determined from the direct kinetic measurements with 6c and E ¼ À18.5 resulted from the competition experiment (vs. 4b) with the nucleophile 7. Thus, the E values determined by the two different experimental methods agreed within one order of magnitude, and an averaged E ¼ À18.8 is a realistic semiquantitative estimate for the electrophilicity of parthenolide (22). Substituents remote from the electrophilic p-system have only a minor impact on the observed reactivity. Carvone ( 19) is almost as reactive as 2-methylcyclohexenone (2b) and verbenone (20) has a similar reactivity as 3-methylcyclohexenone (2c). Only a signicant increase of the steric hindrance in the vicinity to the reaction center, for example in nootkatone (21), causes another slight decrease in electrophilicity in comparison with the model fragment 2c.</p><!><p>The levels of electrophilicity derived from the ranking of 1-5 in the Mayr electrophilicity scale (Fig. 2) facilitate assessing the reaction times and experimental conditions required for successful reactions with C-nucleophiles (see comment box in Fig. 2). 24d Usually, reactions with predicted second-order rate constants of k 2 < 10 À5 M À1 s À1 (at 20 C) will need catalytic activation, heating or signicantly extended reaction times to furnish products. In the subsequent summary, the reaction conditions of reported procedures are compared with predictions based on the Mayr-Patz eqn (1).</p><p>In accord with the determined electrophilicity, tulipalin A (4b, a-methylene-g-butyrolactone, E ¼ À19.4) was reported to undergo high yielding DBU-catalyzed Michael reactions with the nitromethane-(N/s N ¼ 20.7/0.60 in DMSO) (À25 to 20 C, 16 h) 39 and 2-nitropropane-derived carbanions (N/s N ¼ 20.6/0.69 in DMSO) (20 C, 48 h). 40 The Michael adduct from the reaction of 4b with the deprotonated diethyl 2-chloromalonate (N/s N ¼ 18.2/0.74 in DMSO) (in THF, r.t., 6 h, 85% yield) was accompanied by traces of the corresponding 4-oxo-5-oxaspiro [2,4] heptane, generated via a cyclopropanation reaction. This sequence of nucleophilic attack at 4b with subsequent ring closure was exclusively observed when diethyl 2-bromomalonate was used as the pronucleophile in the analogous reaction with 4b (in THF, r.t., 10 h, 75% yield). 41 Furthermore, the piperidine-catalyzed Michael addition of malononitrile (N/s N ¼ 18.2/0.69 in MeOH) at 4b was reported to be facile at ambient temperature in ethanol. The reaction did not stop at the 1 : 1 stage and furnished the two-fold alkylated malononitrile (piperidine cat, EtOH, 2-3 min, product precipitates, 75% yield). 42 The carbon-carbon bond-formation between 4b and the weakly nucleophilic Meldrum's acid-derived enolate ion (N/ s N ¼ 13.9/0.86 in DMSO) is predicted by eqn (1) to be very slow at 20 C (k eqn (1) 2 ¼ 4  10 À7 M À1 s À1 ), and effective product formation required phase transfer catalysis and elevated reaction temperatures (TEBA-Cl in MeCN, 50 C, 10 h, 64% yield). 43 Reactions of dehydrocostus lactone (25, E ¼ À19.0) with the anion of nitromethane (N/s N ¼ 20.7/0.60 in DMSO, 90% yield) were carried out under the same experimental conditions as applied for 4b, the core electrophilic fragment of 25. 39 Given the almost identical electrophilic reactivities, it is unsurprising that reported Michael additions or cyclopropanation reactions of a-methylene-pyranone (5b, E ¼ À19.5) cover the same spectrum of carbon-nucleophiles as for 4b. Carbanions generated by deprotonation of diethyl 2-chloroand 2-bromomalonate (for 2-Br-malonate: in THF, r.t., 10 h, 85% yield) 41 and 2-nitropropane (DBU-catalyzed in MeCN, r.t., 4.5 h, 81% yield) 44 were successfully used to functionalize 5b. Michael additions of nitroethane to cyclopentenone (1a, E ¼ À20.6) and cyclohexenone (2a, E ¼ À22.1) under basic conditions were reported. 45 Enantioselective additions of the anion generated by deprotonation of dimethyl malonate (N/s N ¼ 20.2/ 0.65 in DMSO) to 1a were carried out in the presence of a bifunctional amine-thiourea catalysts (toluene, 50 C, 20 h, 84% yield). 46 Alkylations of the cyclic enones 1a, 2a, and 3 (E ¼ À22.0) at their b-positions were also reported when dimethyl malonate was deprotonated by potassium tert-butoxide (in THF, r.t., 92-95% yield) 47a or when 3 reacted with the slightly less reactive ethyl acetoacetate-derived carbanion (in ethanol, 25 C, 21 h, 52% yield). 47b Furthermore, the cyclic enones 1a and 2a were used as substrates for cyclopropanation reactions with the sulfonium ylide generated from trimethylsulfoxonium iodide (N/s N ¼ 21.3/0.47 in DMSO). 48 Elongated reaction times were needed (DBU, CHCl 3 , r.t., overnight, 82% yield), however, when a bicyclic framework was constructed from cyclopentenone 1a with the less nucleophilic sulfonium ylide derived from ethyl (dimethylsulfonium)acetate bromide (N/s N ¼ 15.9/0.61 in DMSO). 49 With the same sulfonium ylide as the nucleophile, the butenolide 4a (E ¼ À20.7) was reported to produce only a poor yield (22%) of the attempted cyclopropanation product (Cs 2 CO 3 , DMF, r.t., reaction time not given). 50 Conjugate additions of silyl ketene acetals (N/s N ¼ 9.0/0.98 in CH 2 Cl 2 for Me 2 C]C(OMe) OSiMe 3 ) to 4a and 5,6-dihydro-2H-pyran-2-one 5a (E ¼ À21.8) required activation e.g. by Lewis acid catalysts to be productive. 51 The highly reactive phenyl lithium reacted with 2-methyl cyclohexenone (2b) through 1,2-addition at the carbon atom of the carbonyl group. 52 The electron-poor olen 2b (E ¼ À27.5) underwent conjugate additions, however, with the anion of dimethyl malonate (N/s N ¼ 18.2/0.64 in MeOH) in methanol or ethanol aer initial heating and long overall reaction times (MeOH, 16 h (ref. 53) and EtOH, 60 C for 5 h + 20 C, 12 h). 54 The Michael addition of nitroethane (N/s N ¼ 21.5/0.62 in DMSO) to 2b was accomplished by deprotonation of the pronucleophile with N,N,N 0 ,N 0 -tetramethylguanidine and stirring the acetonitrile solution for 3 days at ambient temperature (62% yield). 45 Analogous reactions of deprotonated nitroethane with the even less electrophilic 3-methylcycloalkenones 1c (E ¼ À28.9) and 2c (E ¼ À29.6) were carried out under phase transfer catalysis to avoid too long reaction times (K 2 CO 3 /TEBA-Cl in benzene, r.t. for 4 days, 51% yield from 1c). 45 Accordingly, 2c (E ¼ À29.6) requires a reaction time of 9 days for the Michael addition of the diethyl malonate-derived anion (N/s N ¼ 18.2/0.64 in MeOH) in ethanol at ambient temperature (74-76% yield). 55a In an alternative procedure, the diethyl 2methylmalonate-derived carbanion (N/s N ¼ 21.1/0.68 in DMSO) added to 2c under 15 kbar pressure (DBN, MeCN, 45 C, 36 h) in a yield of 50%. 55b The reaction of 2c with the highly nucleophilic lithiated phenylacetonitrile (N/s N ¼ 29.0/0.58 in DMSO, estimated based on data for 2-phenylpropionitrile) 24d delivers within a few minutes the allyl alcohols via kinetically controlled 1,2-addition (À90 C in THF). The 1,2-addition is reversible, however, and extended reaction times or slightly higher temperatures furnish the corresponding ketone via the thermodynamically favored 1,4-attack (THF, À60 C, 120 min, 95%). 56 This survey of reported reactions of C-nucleophiles with the cyclic Michael acceptors characterized in this work shows, that the determined Mayr electrophilicities E for the electrophiles 1-5 and the dehydrocostus lactone (25) are well in accord with practical experience in organic synthesis.</p><p>Structure reactivity relationships. Embedding the cyclic electrophiles 1-5 and electrophilic natural products 19-25 in the Mayr electrophilicity scale makes it possible to compare their reactivities with those of acyclic Michael acceptors (Fig. 2). The analysis in Fig. 3 reveals that cyclization changes the reactivity of enones and a,b-unsaturated esters in a way that is difficult to predict by intuition. Cyclic enones are by 2-3 E units weaker electrophiles than acyclic b-substituted enones. The opposite trend is observed for lactones: a,b-unsaturated lactones 4-5 are more reactive by 2-3 E units compared to their acyclic counterparts. We performed quantum-chemical calculations to rationalize these antipodal reactivity trends.</p><!><p>Energy proles. To gain further insight into the observed reactivity ranking and structural factors that inuence the observed reactivity of cyclic Michael acceptors, we calculated the reaction proles for the addition of the sulfonium ylide 7 to the electrophiles 1-5 at the SMD(DMSO)/M06-2X/6-31+G(d,p) level of theory using the Gaussian soware package. 57 As depicted in Fig. 4 for the reaction of 7 with cyclopentenone 1a, zwitterionic intermediates IM are generated in the rst step of the reaction mechanism (via TS1). The newly formed C-C bond connects two stereocenters, and the reaction can proceed through a cis-and a trans-attack. As displayed in Table 3, the computational results indicate that the trans-attack is slightly favored over the cis-attack for the cyclic Michael acceptors, except for the 2-and 3-methyl substituted electrophiles 1b, 2b and 2c. In general, however, the computed differences between the cis-and the trans-pathways are small, in accord with the experimental observation that mixtures of diastereomeric products were isolated in moderate yields (Scheme 2). Hence, we refrained from interpreting the stereoselectivity of the cyclopropanation reactions and used the most favorable pathway for our subsequent analyses (if not stated otherwise).</p><p>In the nal step, an intramolecular S N 2 reaction eliminates dimethyl sulde from IM via TS2 to yield the highly exergonic products, namely, dimethyl sulde and cyclopropanes with cisor trans-conguration. For all entries in Table 3, the relative Gibbs activation energies for TS1 and TS2 indicate that the addition (via TS1) is the rate-determining step in the reactions of 7 with 1a.</p><p>As shown in Table 3 and graphically in Fig. 5A, the quantumchemically calculated activation barriers DG ‡ (TS1) agree reasonably well (AE11 kJ mol À1 ; mean deviation: AE3.3 kJ mol À1 ) with the experimental DG ‡ determined either by experiment (k exp 2 ) or by utilizing eqn (1) (k eqn (1) 2</p><p>). Accordingly, there is also a reasonable correlation of DG ‡ (TS1) with the electrophilicity parameters E from Table 1 (Fig. 5B).</p><p>Enone conformation. If compared to analogous acyclic Michael acceptors, the cyclic enones studied in this work experience a signicantly reduced conformational exibility. Experimental electrophilicities were so far only determined for (E)-congured acyclic Michael acceptors. However, relevant information about the reactivity of (Z)-congured conformers, which is required for the discussion of stereoelectronic effects in cyclic enones, is missing. To get insights into the effects of locked conformations on transition state energetics, we set out to perform quantum-chemical calculations for the reaction of 7 with both (E)-and (Z)-pentenone.</p><p>As discussed by Bienvenüe on the basis of UV and IR spectroscopic data, (E)-and (Z)-enones exist in both the s-trans and scis form owing to the hindered rotation around the central carbon-carbon s-bond (Fig. 6, top). 58 Experimental data 58 as well as computations (this work) agree that for (E)-pentenone both s-cis and s-trans conformers are of comparable energy. For (Z)-pentenone, however, the calculations indicate a signicant preference for the s-cis form (cis/trans ¼ 93 : 7). We then computed Gibbs energies for the transition states of the addition of 7 at both (E)-and (Z)-pentenone. We found that the s-cis conformers of (E)-and (Z)-pentenone both react with 7 via lower energy barriers than the respective s-trans conformers. As shown in Fig. 6 (bottom, le), the transition state energy for the s-cis-(E)-pentenone is 8.6 kJ mol À1 lower than that for the strans-(E)-conformer. The difference between the transition states for s-trans-(Z)-and s-cis-(Z)-pentenone amounts to 5.0 kJ mol À1 (Fig. 6, bottom, right). When the most favored transition states for (E)-and (Z)-pentenone are compared, the (E)-isomer of pentenone can be expected to be by approximately one order of magnitude more reactive than the (Z)-congured isomer (DDG ‡ ¼ 7.6 kJ mol À1 ).</p><p>Furthermore, the calculations suggest that the experimentally characterized (E)-pentenone reacts via the s-cis transition state (DG ‡ ¼ 69.7 kJ mol À1 ) with nucleophiles (such as 7). Conformationally locked cyclic species, such as 1a or 2a, adopt transition states similar to the unfavorable s-trans pathway for (Z)-pentenone (DG ‡ ¼ 82.3 kJ mol À1 ). Thus, we can roughly estimate that cyclic enones are at minimum by two orders of magnitude less reactive than analogously substituted a,bunsaturated open-chain ketones. The Mayr E values for (E)pentenone (E ¼ À18.8), cyclopentenone (1a, E ¼ À20.6), Ester vs. lactone. Due to the analogous conjugated p-systems of unsaturated ketones and esters, (E/Z)-congurations and scis/s-trans conformations should inuence the reactivity of esters in a similar manner as in ketones. Counterintuitively, (Z)lactones are more electrophilic than their open-chain ester analogs with (E)-congured CC double bond (cf. Fig. 2), and other stereoelectronic effects seem to dominate their reactivity.</p><p>In line with the relative reactivity ranking in our work, lactones are well-known to undergo signicantly faster alkaline hydrolysis than acyclic esters. This nding was explained by unfavorable orbital interactions in the transition state 59 or through differences in the dipole moments leading to ground state destabilization of (Z)-congured ester units. 60 More recently, stereoelectronic effects were suggested to explain the higher reactivity of unsaturated lactones. 61 Fig. 4 Gibbs energy profile for the reaction of 7 with cyclopentenone (1a) at the SMD(DMSO)/M06-2X/6-31+G(d,p) level of theory (see ESI, Fig. S1, † for a distortion/interaction analysis).</p><p>Table 3 Quantum-chemically calculated energy profiles (in kJ mol À1 ) for the addition of the sulfonium ylide 7 to the electrophiles 1-5 at the SMD(DMSO)/M06-2X/6-31+G(d,p) level of theory</p><p>Trans-pathway Cis-pathway For acyclic esters, the s-(Z) conformation is generally preferred, in which the n / s* interaction donates electron density from the oxygen lone pair into the antiperiplanar antibonding s * CO orbital (Fig. 7). This negative hyperconjugation reduces the electron-deciency of the p-system and, in consequence, electrophilicity. In contrast, the locked s-(E) conformation in lactones impedes such a transfer of electron density and gives rise to an unattenuated electrophilic reactivity of the conjugated p-system (Fig. 7). 61 The oxygen atom of the alkoxy group can affect the reactivity of the p-system only through a minor inductive effect. In line with this interpretation, the quantum-chemically calculated transition state structures for the addition of 7 at the ketone 2a and the lactone 5a are highly similar in geometry and energetics (Table 3, Fig. 8) in agreement with the almost identical experimentally determined second-order rate constants k exp 2 for both reactions (Table 1).</p><p>The unsaturated lactones 4b and 5b bearing an exo-methylene group are more electrophilic than the lactones 4a and 5a with endocyclic unsaturation. The higher reactivity can be attributed to the favorable interplay of two effects. First, the s-trans geometry is locked in lactones 4b and 5b in both the reactants and the transition states. Additionally, we assumed that the absence of substituents at the site of nucleophilic attack introduces less steric constraints in 4b/5b than in 4a/5a.</p><p>To assess this hypothesis, a distortion interaction analysis (DIA) 62 was performed, which compared the transition states of the rst step in the reactions of the S-ylide 7 with 2a, 5a, and 5b, respectively (Fig. 8A). While the distortion energy of ylide 7 is identical in the reactions with 2a, 5a and 5b, the distortion energy of the electrophile is signicantly lower for 5b than for 2a or 5a. It can be expected, that variable demand for geometrical changes at the electrophiles' reactive carbon atom upon C-C bond formation is key for the observed distortion energy difference in the comparison of 5b vs. 5a. We used the distance of the attacked C-atom of the electrophile from the plane dened by the three surrounding atoms in the transition state, as depicted in Fig. 8B, to describe the degree of pyramidalization D in the transition state. In line with Hine's principle of least nuclear motion (PLNM), which predicts 'that those elementary reactions will be favored that involve the least change in atomic position', 63 we observed a higher requirement for pyramidalization in the transition state of the reaction of 7 with 5a (D ¼ 0.227 Å, distortion energy: +39.7 kJ mol À1 ) than in the analogous transition state for the faster reaction of 7 with 5b (D ¼ 0.174 Å, distortion energy: +34.3 kJ mol À1 ).</p><p>Let's now analyze the higher interaction energy in the reaction of 7 with 5b (À40.6 kJ mol À1 ) than in the reaction of 7 with 5a (À33.1 kJ mol À1 ). It has previously been shown, that interaction energies can be further decomposed by energy decomposition analysis (EDA). 64 In this work, we applied symmetryadapted perturbation theory (SAPT) at the sSAPT0/jun-cc-pVDZ level of theory, which decomposes an interaction into its electrostatic, exchange, induction, and dispersion components. 64 The SAPT analysis was performed in gas-phase with an entirely different theoretical method and, therefore, absolute numbers of the interaction energies differ from the results of our DFT method. Nevertheless, we expected the relative trends to hold. As shown in Fig. 8C, the interaction energy (DE sSAPT0 ) for 5b is generally more negative than for 5a. Depending on the extent of bond formation, this is due to different origins. (1) During the approach to the transition state (located at 2.20 Å), it is the stronger electrostatic interaction that favors 5b over 5a. (2) At the transition state and in the further course of the reaction, however, the induction component becomes the decisive factor. In the transition state, the LUMO energy of the distorted 5b is lower (3 LUMO ¼ À0.03527 Hartree) than that of the distorted 5a (3 LUMO ¼ À0.03309 Hartree) while the HOMO energies of 7 are essentially identical (with 5b: 3 HOMO ¼ À0.21875 Hartree; with 5a: 3 HOMO ¼ À0.21867 Hartree). The smaller energetic gap for 7 + 5b (4.99 eV) indicates a more favorable HOMO-LUMO interaction for the couple 7 + 5b than for the combination 7 + 5a (5.05 eV), in accord with the relative DE induction for 5b and 5a in the SAPT analysis.</p><p>Effects of 2-and 3-methyl substitution. Methyl substituents in the aor b-position strongly inuence the reactivity of enones. In a similar but more distinct way than in acyclic systems (Fig. 9A), 25,27 alkyl substitution of the C]C double bond drastically lowers electrophilicity of cyclic enones. A methyl group in the a-position reduces the electrophilicity E of cycloenones by 2 to 6 units (cf. Fig. 2). For substituents placed in the b-position this effect is even more pronounced: the reactivity of b-methyl cycloenones 1c and 2c is reduced by approximately 8 units on the Mayr E scale if compared to the unsubstituted analogs 1a and 2a, respectively. The retarding effect of aand balkyl substituents at the cyclic enones may be caused by steric constraints and/or the electron donating ability of the alkyl group.</p><p>Again, DIA was used to quantify the effects. To keep the transition-state conformations comparable (Fig. 9B), the trans-TS for the reaction 1b + 7 was evaluated in the DIA instead of the (by 1.6 kJ mol À1 ) preferred cis-TS. As the C-C bond length in the transition state of the reaction 1a + 7 differs from that of the reaction 1b + 7, the entire pathways of the reactions of 7 with 1a, 1b, and 1c, respectively, were analyzed. The positions of the distortion and interaction energy curves of 1c (Fig. 9C) and 1b (Fig. 9D) relative to those of 1a reveal the reasons responsible for the reduced reactivity in both cases. Let us rst discuss the effect of 3-substitution (Fig. 9C): the distortion energy for 1c is signicantly more positive than that for 1a while the interaction energy is slightly more negative for 1c than for 1a.</p><p>As the C-C bond lengths in the transition states are similar for the reactions of 7 with the 3-substituted cycloenones and their unsubstituted analogs, these observations can also be assessed in a DIA of the respective transition state geometries of 1a and 1c (and analogously for 2a and 2c). As shown in Fig. 9E, the signicant decrease of reactivity of b-substituted enones is mostly due to an increase of the distortion energy in both the enone fragments and the ylide 7. As previously discussed for 5a and 5b, pyramidalization D and Hine's PLNM can be utilized to rationalize the higher distortion energies of the 3-methyl substituted cycloenones. The reaction of 1a with 7 requires a minor extent of pyramidalization (D ¼ 0.222 Å) than for the much less electrophilic 1c (D ¼ 0.303 Å). 65 Moreover, the higher distortion energy of 7 in the reaction with 1c than in that with 1a can be rationalized by comparing the structures of the transition states (Fig. 9B). Different from the transition states for the reactions of 7 with 1a or 1b, the SMe 2 group of 7 is rotated in the transition state of the reaction with 1c to avoid a clash with the methyl group of the electrophile. Also 2-substituted cyclic enones were found to be weaker electrophiles than the unsubstituted analogs, though, for a different reason. The distortion energy curves for 1a/1b (Fig. 9D) are highly similar or, for 2a/2b (not depicted) even indicate a lower distortion component for 2b. Hence, the nucleophilic b-attack is not sterically hindered by the presence of an a-methyl group. However, the reaction path for 1b suffers from a slightly less negative interaction energy than for the analogous reactions of 7 with 1a. 66 Analysis of the involved HOMO/LUMO interactions of the fragments in the transition state resulted in a slightly stronger orbital interaction in the reaction of 7 with 1a (5.14 eV) than in the reaction of 7 with 1b (5.19 eV). Moreover, Hirshfeld atomic charge analysis of the fragments showed that in the transition state the reactive center of 1b (+0.0391) is less positively charged than that of 1a (+0.0472), presumably due to the electron-donating effect of the methyl group in 1b (ESI, Fig. S3 †).</p><!><p>Helenalin is a sesquiterpene lactone isolated, e.g., from Arnica montana, which embodies two different electrophilic units. 67 GSH was reported to attack faster, yet reversible, [67][68][69] at the cyclopentenone moiety of helenalin than at the a-methylene butyrolactone part. 68 This kinetic preference differs from the ordering of electrophilicities derived from our measurements, which predict higher reactivity for the unsaturated lactone from E ¼ À19.4 for 4b and E ¼ À20.6 for 1a (Fig. 10).</p><p>We, therefore, set out to evaluate whether the electrophilicity parameters E for the cyclic Michael acceptors, which we determined from their reactions with carbon-centered nucleophiles in DMSO solution, would enable us to also predict their reactivity toward glutathione (GSH) in aqueous solution. The rate constants for the reaction of GSH with the selected electrophiles were measured in aqueous, buffered solution at pH 7.4 by utilizing a modied bioassay. Typically, eqn (1) allows one to calculate second-order rate constants within a precision of factor <100 for reactions in which one new s-bond is formed. Table 4 shows that eqn (1) estimated the second-order rate constants for the additions of GSH at cyclopentenone (1a), cyclohexenone (2a), the dihydropyranone 5a and the a-methylene-pyranone 5b within a factor of 20. For 2-methyl-cyclopentenone (1b) and the exoand endocyclic lactones 4a and 4b k exp 2 and the calculated k eqn (1) 2 agreed within a factor of 2. It can, thus, be concluded that the general reactivity pattern of cyclic electrophiles toward GSH is represented by their E parameters.</p><p>In agreement with previous studies by Schmidt on the dual electrophilicity of helenalin toward GSH, 67,68 we determined k GSH (1a) > k GSH (4b). Hence, we have to note that small reactivity differences within one or two orders of magnitude are not unequivocally resolved by the simple three-parameter eqn (1). Changing the experimental method for determining the kinetics, swapping from a C-to an S-centered reference nucleophile as well as the neglect of constraint conformational space in natural products by the fragment approach may twist the relative reactivity order of similarly reactive Michael acceptors. Also the inuence of solvents on the reactivity of carbonyl compounds needs further investigation.</p><p>The relative position of the lactone 5a (E ¼ À21.8) and (E)pent-3-en-2-one (E ¼ À18.8) or (E)-hex-4-en-3-one (E ¼ À18.9) in the electrophilicity scale (Fig. 2) is in accord with the preferential binding of the ketone unit of rugulactone by nucleophilic sites in the course of covalent enzyme inhibition. 17 This illustrates that DE > 2.5 enables a safe prognosis of the reactive site in a natural product with dual electrophilicity.</p><!><p>In summary, sulfonium and pyridinium ylides were utilized as one-bond reference nucleophiles in kinetic experiments to characterize the Mayr electrophilicity parameters E for various cyclic enones and a,b-unsaturated lactones in DMSO at 20 C. By combining the electrophilicity parameters E with tabulated nucleophilicity descriptors N (and s N ) eqn (1) can be used to predict the rate constants for the reactions of 1-5 with various C-nucleophiles, as demonstrated by comparison with reported synthetic protocols.</p><p>Most valuable, the reactivities of cyclic core fragments of the Michael acceptors 1-5 agree with the observed electrophilicities of natural products (terpenes) of more complex structure and considerably higher molecular weight that contain the same reactive moiety. The distinct different reactivity of cyclic enones and unsaturated lactones compared to their acyclic analogs was analyzed by quantum-chemical calculations, distortion interaction analysis, and by considering stereoelectronic effects.</p><p>The most important structural effects on the reactivity of a,bunsaturated carbonyl compounds are summarized in Fig. 11. The locked conformations of cyclic Michael acceptors have a signicant impact on their electrophilic reactivities. If compared to analogous open-chain enones, the electrophilicity of cyclic enones is signicantly reduced by the xed (Z)-geometry of the s-trans congured p-system. Alkyl groups in either aor b-position of the cyclic enones further attenuate the electrophilicity of cyclic enones by positive inductive effects and steric bulk in vicinity or directly at the electrophilic reaction center. Thus, b-alkylated cyclohexenones are among the least electrophilic species characterized so far in Mayr's reactivity scales. 24 In contrast, the rigid cyclic structures of a-methylene-gbutyrolactones facilitate synergistic stereoelectronic effects which favorably combine with a lack of steric hindrance at the reactive site to furnish a privileged class of highly potent electrophiles. In contrast to simple alkyl acrylates of comparable electrophilic reactivity, the cyclic scaffold of sesquiterpene lactones can be loaded with stereochemical information needed for recognition processes and selective reactions in living organisms. It is, therefore, not surprising that plants have chosen a-methylene-g-butyrolactones as most abundant electrophilic fragment in biologically active sesquiterpene lactones.</p><p>The reactivity parameters determined in this work, together with those of previously characterized acyclic Michael acceptors, now provide an extensive basis for the systematic development of reactions with various classes of nucleophiles. Derivatization of natural products with the studied electron-decient cyclic core fragments can in future be exploited in a more straightforward manner, thus saving limited natural resources, energy, and human effort. Knowledge of the electrophilic potential of these cyclic Michael acceptors to undergo</p>

Royal Society of Chemistry (RSC)

MODELING VARIANCE: A VARIANCE-MOTIVATED APPROACH TO MOLECULAR PREDICTION

Using machine learning to predict molecular properties is an exciting research area at the interface of computer science, statistics, chemistry, and physics. Thus far, a great deal of work has been done on training various ML models to predict ground state energy, using the so-called 'Coulomb matrix', a global molecular descriptor as inputs. In this variance-motivated study, the variance of a multi-thousand molecular dataset of Coulomb matrices is analyzed; a variance analysis. This paper presents novel statistical methods and models that can aide in prediction and analysis of molecular properties and molecules using the Coulomb matrix. Analysis is performed after a detailed literature review of molecular prediction and Normality of data assessment. It is also hoped that the models introduced in this variance analysis can be generalized to other areas of interest.

modeling_variance:_a_variance-motivated_approach_to_molecular_prediction

INTRODUCTION<!>COMPUTATIONAL METHODS<!>RESULTS AND DISCUSSION<!>Figure 2. Density histogram of atomic energy variance for the 477 molecule sample<!>Figure 8. Output of Normality classification algorithm, classifying 477 size atomic energy sample as Normal<!>FIGURE 9. Scatter plot of atomic energy variance against mean atomic energy for the 477 sized random sample<!>Conclusion

<p>Quantum mechanics has been combined with machine learning to predict molecular properties. 1 Various ML models for this purpose have been introduced 1,2,3 ; they generally share one thing in common: models are trained to predict ground state energy. One common input for such an ML model is the Coulomb Matrix (CM), made popular by Dr. Matthias Rupp. CM is a global 3D molecular descriptor. 1D molecular descriptors capture composition, 2D captures the molecular graph, and 3D captures shape. Global descriptors characterize the entire molecule.</p><p>CM is an alternative to using the Schrodinger equation to determine molecular properties, which is rather cumbersome. Using the Hamiltonian, a component of the Schrodinger equation, and the Schrodinger equation to determine molecular properties is the classical quantum mechanical approach. The Coulomb matrix is a square and symmetric matrix with size equal to the number of atoms in the molecule, squared. It represents electronic interactions of the atoms of a molecule with themselves and each other. Diagonal elements represent a polynomial fit of atomic energies to nuclear charge, hence, they represent the fitted atomic energies of the atoms. Off-diagonal elements represent Coulomb repulsion between atoms. As can be seen in figure 1, the number of elements is the number of atoms, squared, hence there are 9 elements. The diagonal elements represent an atom interacting with itself, which is represented in the CM as a polynomial fit: .5*Zatom 2.4 . Where Z is the atomic number of the atom. Hence element 1,1 (row 1, column 1) has a fitted atomic energy of 36.86. All elements that are not diagonal represent Coulomb repulsion, as stated earlier; the numeric representation is: ZIZJ/|RI-RJ|, where RI-RJ is the distance between atoms I and J.</p><p>It has been shown computationally 1 that a plot of an ML model that used the Coulomb matrix as inputs, based on thousands of molecules, showed a 1:1 relationship between predicted atomization energies and reference atomization energies. The model was said to be a better predictor of atomization energy than semi-empirical quantum chemistry. The energy estimates used in the ML model are a weighted Gaussian sum. Molecules used in the study were from a Molecular Generated Database (MGD) of nearly 1 billion stable organic molecules.</p><p>Regression tree algorithms have also been used to predict ground state energies 2 , using CM as input-"features". Using a dataset that contained over 16,000 molecules, ground state energies were computed on trained ML models. The model, boosted regression tree, was shown to have increased accuracy and reduced computational cost. The computational based study has potential applications in molecular discovery and informatics. The actual prediction is pseudo-atomization energy, which is a function of ground state energy and pseudo atomic energy. The absolute value of pseudo-atomization energy for the sample size of 16,242 molecules was shown to have a Normal distribution. Not all seek alternatives to solving the Schrodinger equation for molecular prediction. AFLOW 4 , a high throughput (HT) framework uses ab initio calculations to solve the Schrodinger equation to yield information such as energies and electron densities. AB initio calculations use only restrains and coulomb interactions as inputs. AFLOW is also used to calculate crystal structure properties of alloys, inorganic compounds, and intermetallic compounds. AFLOW is designed to run atop structure energy software (DFT is common). After selection of starting structures from a database, AFLOW adjust lattice parameters, and creates an input file with all parameters necessary for relaxations, and static and bond structure runs. AFLOW computes structure total energies and electronic bond structures. It also can perform Monte Carlo calculations, generate nanoparticle structure files, and identify interstitial sites inside any crystallographic structure (from an input of atomic positions).</p><p>Whether AFLOW, which uses classical quantum mechanics, or the CM is used, both can perform powerful molecular characterizations. Both however require a detailed and descriptive list of inputs to function. In addition to using the CM in ML to predict atomization energies, it has also been stated that the CM can be used to for rotational spectra interpretation 3 . In the computationally driven study revolving around molecular isomerism, the author used an ETKDG algorithm to generate 1000 random conformations of 309 isomers. The prediction task was to distinguish between a given isomer from all other isomers. The misclassification rate was defined as proportion of incorrect assigned labels (incorrectly identifying an isomer). Various ML models were used: logistic regression, decision tree, random forest, gradient boosted trees, support vector machine, and k-nearest neighbor. Models were trained and cross-validated. The decision tree ML model had the highest misclassification rate and support vector machine had the lowest. The mean largest eigenvalue of CM (averaged over conformers) was said to be inefficient at distinguishing isomers. Now it's time to discuss a blended computational-laboratory study on the using ML to label the degree of peptide reactivity with chemicals that contain allergens 5 . Chemical allergens react with proteins, inducing skin sensitization 5 . The majority of allergens are electrophilic and react with nucleophilic amino acids. The purpose of the study was to determine whether and to what extent reactivity correlates with skin sensitization potential. They evaluated 82 allergen containing chemicals (of different potencies) and non-allergen containing chemicals for their ability to react with amino-acid containing molecules. After a set reaction time, UV detection was used to quantify the depleted amount (reacted amount). The reactivity data and existing data was used to build a classification tree that allowed ranking of reactivity: minimal, low, moderate, high. The classification tree ML model had 89% prediction accuracy (based on cysteine and lysine amino acids). The splitting rule for the tree was based on average peptide depletion exceeding a threshold.</p><p>Thus far various ML models have been presented that predicted molecular properties. In addition, a non-ML (quantum mechanic) model was presented (AFLOW) that had a much broader spectrum of molecular characterization, though it relied heavily on input files. Predictions for the ML models were usually limited to atomization energy, though several other molecular predictions were presented as well. Isomerism was predicted via comparison of various ML models and the reactivity of chemicals containing allergens was predicted using classifications trees. Except for the last study and AFLOW, all others presented used the CM as input for the ML model.</p><p>One may wonder how this relates to This study of CM variance. The purpose of This study is to model the variance of the fitted atomic energies for a random sample of molecules from a large molecular database, in an effort to make inference to broader populations of molecules, using current and novel models. The goal also is to use smaller and simpler inputs to predict molecular properties. To analyze and model variance, parametrically, it is necessary to assign a distribution to variances of fitted atomic energies. Of note is the fact that fitted atomic energy variance is not the same as sampling variance. To begin, let's briefly review some literature on characterizing and assessing probability distributions and data.</p><p>The Normal distribution is a common distribution for modeling data, especially of a large size. Hence, it can model and statistically summarize Normal data. Common methods to assess normality are graphical, usually the data quantiles are plotted against the standard Normal quantiles. Another common test, that is non-graphical is the Shapiro-Wilk test, which is generally only recommended for sample sizes 50 and under. Box-plots can also be used as a simple quantile graphical tool to assess normality. A histogram plot of the data is yet another common graphical method for assessing normality. It is often that analysts will use these graphical tools to deem data that 'roughly' fits as normal or 'approximately Normal'. Let us look at a study that presents an alternative graphical method to assess Normality that was compared with various tests using Monte Carlo testing 6 . The results of the study suggest a potential more evident way to reject un-Normal data.</p><p>According to the study, in finance literature, a plot of empirical and fitted normal densities on the log scale is regularly preferred as a graphical means to assess normality. The study argues that interpretation of quantile-quantile plots can be compromised; assessing degree of curvature in the plot to designate data as not Normal is largely subjective. They then discuss a graphical alternative, the logdensity (empirical density) plot. The graphical procedure involves plotting empirical density alongside fitted Normal density. The log-scale is used to clearly display the tails of non-Normal data. One weakness of this technique is that it was said to be possibly misleading for small sample sizes (which shouldn't be a surprise). In the study, a 1,000 simulation Monte Carlo test was performed to simulate a p-value for the log density method, which was compared to other Normality assessing tests (Shapiro-Wilk, Anderson-Darling, and Cremer von-Mises). Power was used to compare the various tests. Data from three distributions were used: Cauchy, t(with 6 degrees of freedom), and Gumbel (extreme value). In all but the extreme value distribution, the Monte Carlo log density test had the highest power. The author argues this is due to the sensitivity of the Monte Carlo Log Density to fat-tailed departures from normality, but being weaker for skewed (extreme value) distributions. Another test comparison between quantile-quantile plots and log density plots showed a Normal appearing quantile-quantile plot, but an obvious non-Normal log density plot. One of the biggest drawbacks of the log density method presented in the study is that it is cumbersome to compute empirical density, hence most would likely prefer the more easily implemented quantile-quantile plots for approximating normality.</p><p>A brief literature review of skewness 7 will conclude the data characterization review. The paper is essentially a review of skewness and argues in favor of incorporating it into statistical education. There are various ways to assess skewness, graphical and quantitative. One quantitative method is in the form of a skewness statistic, which compares the mean to the median. Visual tools to assess skewness include beam-and-fulcrum plots, boxplots, and dotplots; dotplots being deemed the best visual tool to assess skewness. The Fisher-Pearson coefficient of skewness, measures skewness and is the ratio of average distance cubed from the mean to average distance squared from the mean (the denominator is raised to the 3/2 power). Of note is that the statistic presented in the paper consists of sample distance from the mean; the mean being the sample average. The brief review of skewness was necessary, because it is often the case that data exhibits non-Normal deviation in the quantile-quantile plots at the extreme ends (tails), and it is necessary to have a means to measure tail deviation as significant skewness, or not. Now that the literature review has been completed, it is time to discuss the experiments and data analysis.</p><!><p>The QM7 database 8 was used to obtain the CMs. The dataset contains 7165 molecules and is a subset of the GDB-13 database (contains nearly 1 billion stable organic molecules). QM7 contain molecules no larger than 23 atoms, and contains the CMs for each molecule and atomization energies. ML models commonly aim to predict the atomization energies using CMs as input.</p><p>CMs were stored as row elements in the QM7 database, so it was necessary to regenerate the CMs in matrix form. A random sample of size 500 was taken from the 7165 molecule database, each sample corresponding to a molecule (stored as a CM and atomization energy). Atomic energies (diagonals of CM) were extracted for each CM in the sample. Mean atomic energy and variances were computed for each molecule and stored in vectors. Only molecules with variance of no more than 1000 (units: protons 5.76 ) were considered, reducing the sample size to 477. Unless noted otherwise, all calculations are rounded to the nearest whole number. Unless stated otherwise, variance will refer to sample variance. Unless stated otherwise, atomic energy refers to CM model fitted atomic energy (.5Z 2.4 )</p><!><p>The density histogram of atomic energy (fitted) variance of the random sample of size 477 molecules is shown in Figure 2. The sample mean for this distribution is 562 protons 5.76 and the corresponding standard error is 118 protons 5.76 . Note that the units here are the same, because the measured quantity is variance. The sample median for this distribution (557 protons 5.76 ) is very near in value to the sample mean, indicating the symmetry of this distribution and providing good justification to further assess normality of data.</p><!><p>Figure 3 shows the density histogram of the mean fitted atomic energies for the random sample of size 477. For this distribution, the sample mean, median, and variance are 20 protons 2.4 , 20 protons 2.4 , and 19 protons 5.76 . Though the mean and median are equal, the distribution is visually not symmetric with high variance relative to the sample mean; they are approximately equal. Both distributions are visibly continuous. Figure 4 shows the density histogram of atomic energy variance with fitted normal density superimposed. From the histogram, the Normal distribution is a good approximation of the sample data. To further assess normality of atomic energy variance of the sample, a quantile-quantile plot was graphed (Figure 5), which showed bulk normality, but normality deviation for lower and higher quantiles. To assess normality further, an algorithm was created that superimposed Normal density over the density of atomic energy variance. The Normal density represented a population of atomic energy variance with mean(atomic energy variance) equal to the estimated mean (562 protons 5.76 ) and standard deviation equal to estimated standard deviation (128 protons 5.76 ); note that the units are the same due to the measured quantity being variance. The size of the population was set equal to the original population size of the QM7 dataset, 7,165 molecules. Each simulation (algorithm run) superimposed randomly generated Normal density with mean variance and standard deviation given by the estimates. The density histogram of a simulation is shown in Figure 6. The results further support that the Normal probability distribution is a good model to summarize and approximate the atomic energy variance sample data.</p><p>Next, an alternative model to classify data as approximately Normal will be introduced. The model is a combination of a classification tree and statistical hypothesis testing, and will be referred to as TDT (Test-performing Decision Tree). Figure 7 displays the decision tree. The advantage of the test over visual (subjective) methods to assess and 'measure' normality is that TDT performs a hypothesis test for Normality. The test ends in a composite test statistic, which is used to make a decision that the data is Normal or not Normal. The test is also fully automated; user interpretation is not a component of the test. Note that Normal refers to being Normal from an input population (or parent data). The decision tree splits are based on the data proportions of a Normal distribution: 68% of the data falls within 1 standard deviation units of the mean, 95% of the data falls within 2 standard deviation units of the mean, and 99.7% of the data falls within 3 standard deviation units of the mean. The actual acceptance criteria for proportions is relaxed, since the TDT tests for approximate Normality: 58% to 83% of the data must fall within 1 standard deviation unit of the mean, 85% or more of the data must fall within 2 standard deviation units of the mean, and 95% or more of the data must fall within 3 standard deviation units of the mean. The test tallies the counts in each class and computes proportions. The proportions form the composite test statistic. Data is deemed Normal only if the proportions fall within the percentage ranges discussed prior. TDT performs a hypothesis test, the null hypothesis is that the Data is Normal (with mean equal to the population mean and standard deviation equal to the population standard deviation). The alternate (research) hypothesis is that the data is not Normal (with mean and standard deviation equal to that of the population). The model inputs are the sample and the parent sample (or population).</p><p>To assess the strength of the model at correctly identifying Normal data from a given population as Normal, 10,000 simulations were performed to measure type 1 error (rejecting the null hypothesis when the null hypothesis is true). TDT inputs were a population (size 7,165) of randomly generated realizations from the standard normal distribution and a random sample (size 477) from the population. The sizes of the population and sample correspond to the size of the QM7 dataset and variance random sample used in This study. Three iterations of 10,000 simulations were performed, each having a type 1 error rate of 0. TDT was then used on the atomic variance sample of interest, which classified the data as Normal. Figure 8 shows the R output of the TDT test. test statisticA, p_1sd, is: 0.6771488 .test statisticB, p_2sd, is: 0.9979036 . test statisticC, p_3sd, is: 1 . Accept null. X ~ Normal(mean_population,sd_population)</p><!><p>The next part of the analysis is determining if the atomic energy variance can be modeled as a linear function of mean atomic energy. Figure 9 is a scatter plot of the 477 molecule sized sample.</p><!><p>The plot shows a clear positive correlation between atomic energy variance and mean atomic energy, however it also shows a fanning pattern and it is not hard to see that a line fit through the data would have residuals of nonconstant variance, a violation of linear regression. To reduce the fanning so that a linear model could be fit to the data, only molecules with a maximum variance of 350 protons 5.76 were considered, shrinking the sample size to 25. Of the size 25 random sample, only 8 data points were unique, in other words multiple molecules in the sample shared the same atomic energy variance and mean atomic energy. This is understandable, since the population (QM7 dataset of 7,165 molecules) contain small organic molecules, isomerism will result in numerous molecules having the same CM diagonals (fitted atomic energies), thus they will have the same mean atomic energies and same atomic energy variance. The linear model that will be discussed next consists of the 8 molecules with unique CM diagonals; isomers are not included. Figure 10 shows the scatter plot, showing a linear relationship between atomic energy variance and mean atomic energy. Figure 11 is the same plot with the best fit regression line shown. Figure 12 shows the R output of the linear model, and figure 13 is a plot of the residuals against mean atomic energy. To summarize the regression results, for a random sample of 8 small organic molecules with maximum atomic energy variance of 350 protons 5.76 , the atomic energy variance can be expressed as a linear function of mean atomic energy. Though the sample consists only of 8 molecules, the results support that for a small number of molecules with low atomic energy variance, the mean atomic energy can be used to predict atomic energy variance, and may potentially be an alternative molecular descriptor (for a small number of molecules). The final portion of the data analysis concludes with variance visualization. Though commonly used to characterize data in statistical analysis, visualization of variance is not regularly used in statistical analysis and data visualization. This portion of the analysis presents a model to visual variance. Atomic energy sample variance is known for each of the 477 molecules in the random sample of interest. Sample variance is the average squared distance from the sample mean; emphasis here is put on the fact that it is a distance. The distance between two points on a line can be calculated as the hypotenuse of a triangle: c=(a 2 +b 2 ) 1/2 , where a is the horizontal segment length between the two points and b is the vertical segment length between the two points. The variance visualization technique involves computing the vertical component of sample variance for a fixed horizontal component (this set represents point B), then plotting a line containing 0,0 (point A) and point B. The length of this line is the sample variance. The variance, represented as a vector, is translated to the origin (all variance vectors use the origin as point A). Figure 14 is a plot of a random sample of size 30 from the atomic energy variance sample (size 477). The mean, median, and standard deviation of this sample is 584 protons 5.76 , 583 protons 5.76 , and 121 protons 5.76 , which are near in value to the values of the parent sample. The horizontal component of the variance was fixed at 1. Figure 14 shows the distribution of variance, visualized as vectors representing atomic energy sample variances. Figure 15</p><!><p>Variance was modeled for a random sample of molecules from the QM7 database. Modeled variance represented fitted atomic energy variance. A rigorous assessment of normality was performed using simulation and multiple visualization methods, including quantile-quantile plots. An alternate model to assess normality was introduced, the TDT (Test-performing Decision Tree), which computes a test statistic for accepting or rejecting data as Normal based on proportion amounts of data that are within 1,2, and 3 standard deviations of the mean. TDT is a combination of statistical hypothesis testing and machine learning/statistical learning classification trees that uses sample and population data (or parent sample data) as inputs. Three iterations of 10,000 simulations were performed using TDT, each resulting in a type 1 error rate of 0. TDT classified the fitted atomic energy variance sample as Normal. TDT is a fully automated model for assessing normality that is independent of user interpretation. Linear regression was performed on a small random sample and showed that atomic energy variance can be expressed as a linear function of mean atomic energy. This will be further explored to see if mean atomic energy can be used as a molecular descriptor for some molecules, which would greatly reduce the dimensions of molecular datasets like the one in This study. For future analysis, similar groups will be considered from QM7 (example atmospheric or greenhouse gases, greenhouse related gases) and linear regression will be reattempted. The variance-driven analysis concluded with the introduction of a variance visualization nonparametric model, which provided visualization of variance about a fixed location (the origin). It is hoped that the models in this paper will be considered and used to augment statistical analysis, in numerous areas of application.</p>

ChemRxiv

Modified carbazoles destabilize microtubules and kill glioblastoma multiform cells

Small molecules that target microtubules (MTs) represent promising therapeutics to treat certain types of cancer, including glioblastoma multiform (GBM). We synthesized modified carbazoles and evaluated their antitumor activity in GBM cells in culture. Modified carbazoles with an ethyl moiety linked to the nitrogen of the carbazole and a carbonyl moiety linked to distinct biaromatic rings exhibited remarkably different killing activities in human GBM cell lines and patient-derived GBM cells, with IC50 values from 67 to > 10,000 nM. Measures of the activity of modified carbazoles with tubulin and microtubules coupled to molecular docking studies show that these compounds bind to the colchicine site of tubulin in a unique low interaction space that inhibits tubulin assembly. The modified carbazoles reported here represent novel chemical tools to better understand how small molecules disrupt MT functions and kill devastating cancers such as GBM.

modified_carbazoles_destabilize_microtubules_and_kill_glioblastoma_multiform_cells

Introduction<!>Chemistry<!>Antitumor activity of reference MTAs and modified carbazoles in the human GBM cell line T98G.<!>Killing activity of select modified carbazoles in patient-derived GBM cells and the human liver cell line HepG2<!>Cell-killing activities and MOA of modified carbazoles in the NCI-60 cancer cell line panel<!>Activity of modified carbazoles with tubulin and microtubules<!>Gas Phase Pharmacophore Study<!>Molecular Docking Studies<!>Conclusion<!>General experimental methods.<!>9-ethyl-9H-carbazole (5).<!>9-ethyl-9H-carbazole-3-carbaldehyde (6).<!>9-ethyl-9H-carbazole-3-carboxylic acid (7).<!>9-ethyl-3-[(4-methylnaphthalen-1-yl)carbonyl]-9H-carbazole (8).<!>9-ethyl-3-(4-methylpiperazine-1-carbonyl)-9H-carbazole (9).<!>3-(4-chlorobenzoyl)-9-ethyl-9H-carbazole (10).<!>9-ethyl-3-(4-fluorobenzoyl)-9H-carbazole (11).<!>9-ethyl-3-(4-methylbenzoyl)-9H-carbazole (12).<!>1-(9-ethyl-9H-carbazol-3-yl)-2-phenylethan-1-one (13).<!>9-ethyl-3-(naphthalene-1-carbonyl)-9H-carbazole (14).<!>9-propyl-9H-carbazole (15).<!>9-propyl-3-[(4-methylnaphthalen-1-yl)carbonyl]-9H-carbazole (16).<!>9-(2,2,2-trifluoroethyl)-9H-carbazole (17).<!>3-(4-methylnaphthalene-1-carbonyl)-9-(2,2,2-trifluoroethyl)-9H-carbazole (18).<!>(9-ethyl-9H-carbazol-3-yl)(quinolin-5-yl)methanol (19).<!>9-ethyl-3-(quinoline-5-carbonyl)-9H-carbazole (20).<!>(9-ethyl-9H-fluoren-3-yl)(quinolin-7-yl)methanol (21).<!>9-ethyl-3-(quinoline-8-carbonyl)-9H-carbazole (22).<!>(9-ethyl-9H-carbazol-3-yl)(1,2,3,4-tetrahydroquinolin-6-yl)methanone (23).<!>9-ethyl-3-(1,2,3,4-tetrahydroquinoline-8-carbonyl)-9H-carbazole (24).<!>9-ethyl-3-[(1,2,3,4-tetrahydroquinolin-1-yl)carbonyl]-9H-carbazole (25).<!>9-ethyl-3-(4-methylnaphthalene-1-carbothioyl)-9H-carbazole (26).<!>(9-ethyl-9H-carbazol-3-yl)(4-methylnaphthalen-1-yl)methanol (27).<!>9-(4-methylnaphthalene-1-carbonyl)-9H-carbazole (28).<!>9-ethyl-2-methoxy-9H-carbazole (29).<!>9-ethyl-7-methoxy-9H-carbazole-3-carboxylic acid (30):<!>9-ethyl-2-methoxy-6-(4-methylnaphthalene-1-carbonyl)-9H-carbazole (31).<!>Ethyl 1H,2H,3H,4H,5H-pyrido[4,3-b]indole-2-carboxylate (32).<!>Ethyl 5-ethyl-1H,2H,3H,4H,5H-pyrido[4,3-b]indole-2-carboxylate (33).<!>5-ethyl-1H,2H,3H,4H,5H-pyrido[4,3-b]indole (34).<!>5-ethyl-2-[(4-methylnaphthalen-1-yl)carbonyl]-1H,2H,3H,4H,5H-pyrido[4,3-b]indole (35).<!>Cell culture.<!>Cell viability.<!>[3H]Colchicine binding to purified tubulin and tubulin assembly.<!>MT-tubulin partitioning.

<p>Many subtypes of cancers are successfully treated with microtubule-targeting agents (MTAs), including breast, lung and head and neck tumors, lymphoma and melanoma [1]. MTAs act by binding to tubulin or MTs and disrupting MT assembly, disassembly and dynamics, all of which impairs the precise orchestration of mitosis and triggers cell death by engaging the MT spindle check-point [2]. Starting in the 2000s, evidence suggested that GBM, the most common, devastating and incurable type of brain cancer, is particularly sensitive to MTAs [3–6]. The sensitivity of GBM cells likely results from the expression of mutated proteins that erroneously control MT dynamics and their interactions with kinetochores [7, 8]. Accordingly, several studies have shown that GBM cells treated with MTAs undergo death through MT spindle check-point arrest and ensuing apoptosis [4, 5]. Human clinical trials showed that MTA treatment reduces tumor burden in patients diagnosed with GBM [9], and the recent advent of tumor-treating fields for the treatment of patients diagnosed with GBM provide additional evidence that GBMs are sensitive to the disruption of MT function [10]. Specifically, the polarizing field established by tumor-treating fields affects tubulin assembly into MTs, disrupts mitosis and triggers the spindle check-point and ensuing cell death [11]. Together, these genetic, pharmacological and tumor-treating field findings show that GBM cells are particularly sensitive to alterations in MT function and highlights the need to develop novel MTAs for the treatment of this devastating disease.</p><p>There are several classes of MTAs that disrupt MT function through distinct mechanisms of action (MOA). Three of the most extensively studied classes of MTAs are vinca alkaloids, taxanes and colchicinoids, which bind to distinct sites on tubulin and differentially affect MT dynamics [1, 12]. Prototypical MTAs that bind to the colchicine sites include combretastatin A-4 (1), the carbazole-based analogues of combretastatin (2) that weakly inhibit tubulin polymerization and yet triggers pronounced apoptosis, and nocodazole (3), which rapidly and fully inhibits tubulin polymerization and triggers apoptosis in cancer cells (Figure 1) [13, 14].</p><p>In a previous study, we identified a series of indole-based MTAs that bind to the colchicine site of tubulin, reduce MT polymerization in vitro, reduce MT dynamics in GBM cell lines in culture and kill GBM cells by apoptosis when tested in an orthotopic syngeneic mouse model [14]. Here, we used these results as the starting point for the development of a series of carbazole-based MTAs (4) and studied their antitumor activity in both GBM cell lines and patient-derived GBM cells in culture. We also studied how the antitumor activity of carbazole analogues relate to their ability to disrupt MT. Our studies emphasize the importance of substituting the carbonyl linker with either carbinol or thiocarbonyl functional groups, as well as the importance of substitutions at the nitrogen atom of the carbazole moiety (4). Finally, we provide a modeling study of the docking of the most potent modified carbazoles to the colchicine site and outline its unique binding mode and how it affects MT assembly.</p><!><p>The route for the synthesis of the carbazole derivatives is outlined in Scheme 1. Compounds 8, 10, 11, 12, 13, 14, 16, 17, 25 and 31 were prepared from commercially available carbazoles via consecutive substitution with the corresponding n-alkyl bromide under alkaline conditions, followed by a Friedel-Crafts reaction using the corresponding acyl chloride derivatives (Scheme 1). Carbazole 28 was obtained by alkylation of carbazole with 4-methylnaphthalene-1-carbonyl chloride under basic conditions. Bromine-lithium exchange of bromo-carbazole derivative (t-BuLi, −78 °C), followed by the addition of quinolone aldehyde derivatives, furnished the corresponding alcohols, 19 and 21, which were oxidized to form 20 and 22 (Scheme 2).</p><p>Quinoline derivatives 23 and 24 were conveniently prepared after reduction of compounds 20 and 22, respectively.</p><p>The thione and the carbinol derivatives (26 and 27) were obtained by heating amide 8 with Lawesson's reagent under microwave conditions and by reducing its carbonyl with NaBH4, respectively (Scheme 3).</p><p>To validate the importance of the planar carbazole ring, we synthesized the carboline derivative 35 (Scheme 4) from phenylhydrazine and ethyl-4-oxocyclohexanecarboxylate.</p><!><p>The two prototypical, colchicine site targeting agents, combretastatin A-4 (1) and nocodazole (3), killed human GBM cells line, T98G, with the expected antitumor EC50 values (0.6 and 29 nM, respectively), but only killed T98G cells by a maximum of 30.4% ± 2.1 (at 10 nM) and 45.6% ± 1.7 (at 10 μM), respectively (Figure 2A). Supplementary Figure 1 shows that antitumor activity of 3 additional references MTAs also only partially killed T98G cells: EC50 and maximal killing values for colcemide were 113 nM and 35.9% ± 6.4 at 10 μM, for vinblastine were 184 nM and 36.0% ± 2.3 and for paclitaxel were 148 μM and 63.2% ± 2.5 at 30 μM. These results show that reference MTAs that both inhibit assembly and destabilize MTs exhibit moderate antitumor activities in T98G cells.</p><p>We then treated T98G cells with the 19 modified-carbazoles synthesized for this study and report the EC50 values and maximal killing activity values of their antitumor activities in Tables 1 and 2.</p><p>Figure 2B and 2C show the antitumor activities of five modified carbazoles (8, 20, 23, 25 and 27) that hold key SAR characteristics. The antitumor activities for the remaining compounds are in Supplementary Figure 2. Thus, a comparison of the chemical structures of the modified carbazoles and their antitumor activities expressed as EC50 values (potency) and maximal killing activities (efficacy) allows for multiple conclusions around the SAR of this response.</p><p>First, we found that fourteen modified carbazoles that all have an ethyl moiety linked to the nitrogen of the carbazole and a carbonyl moiety linked to distinct biaromatic rings exhibited remarkably different EC50s. Specifically, five such compounds killed T98G cells with EC50 values below 1 μM [8 (87 nM) < 20 (184 nM) < 27 (560 nM) < 25 (757 nM) < 12 (993 nM)], two compounds killed T98G cells with EC50s between 1 and 2 μM [10 (1.5 μM) and 14 (1.7 μM)] and seven compounds were only weakly active at concentrations above five micromolar or essentially inactive [9, 13, 16, 18, 23, 24 and 26]. Note that the replacement of the methylnaphthyl group present in the most potent compound 8 (87 nM) by either toluyl (as in 12, 993 nM) or a non-substituted naphthyl moiety (as in 14, 1.7 μM) or a benzyl moiety (as in 13, inactive), increased EC50s values by 11-fold, 20-fold and more than 115-fold, respectively.</p><p>Second, modified carbazoles and carboline exhibited remarkably different maximal killing activities. Specifically, nine compounds killed T98G cells by more than 50% (i.e. more than the maximal killing activities reached by reference MTAs). Thus, the rank order of these compounds is: 25 (89%) > 12 (84%) > 20 (74%) > 27 (73%) > 14 (72%) = 31 (72%) > 10 (71%) > 22 (60%) > 11 (55%). Three compounds killed T98G cells by 35–45% (8, 18 and 35) and six compounds did not significantly affect T98G cell viability (9, 13, 16, 23, 24 and 26). Note that the position of the nitrogen atom in the quinoline portion of modified carbazoles controls their maximal killing activities, as exemplified by the reduction in maximal killing activity when moving the nitrogen atom from position 1 (as in 25, 89% killing) to either position 5 (as in 20, 60% killing) or position 8 (as in 22, 60% killing).</p><p>Third, introduction of a nitrogen atom in position 6 of the carbazole scaffold increases the EC50 value by more than 115-fold as exemplified when comparing the response of 8 (87 nM) and the carboline analogue 35 (> 10 μM). A possible explanation for this change in EC50 value is that both the carbazole moiety and the aromatic moiety attached to the carbonyl will impact the positioning of the lipophilic moiety attached to the carbonyl. Thus, this result suggest a need for modified carbazoles to be planar for optimal activity [15]. Importantly, the nitrogen atom in position 6 of the carbazole scaffold only affect the EC50 values of 8 and 35, as the maximal killing reached by both compounds was 45%, indicating that a nitrogen atom in position 6 of the carbazole scaffold will control the EC50 value of the compound independently of the maximal killing activity.</p><p>Fourth, the carbonyl linker determined the compound's EC50 values as evidenced by comparing 8 (87 nM) with both the carbinol 27 (560 nM) and the thiocarbonyl compound 26 (> 10 μM), which results in 6- and 115- fold increase in EC50 values, respectively. This result can be explained by a loss of a hydrogen bond interaction since thiocarbonyl derivatives are known to destabilize hydrogen bond formation through higher steric demands that are imposed by both the larger sulfur atom and the lower electronegativity compared with the oxygen atom [15]. Remarkably, changes in the carbonyl linker had an opposite effect on the maximal killing activity of these compound, as the carbinol 27 killed more T98G cells than the carbonyl 8 (72% versus 45%, respectively).</p><p>Fifth, the importance of the chain born from the endocyclic nitrogen is evident by comparing the EC50 values of the ethyl compound 8 (87 nM) with both the trifluoroethyl compound 18 (5.6 μM) and the propyl compound 16 (> 10 μM), corresponding to 64- and 115-fold increases in the EC50s values, respectively. Remarkably, the chain linked to the endocyclic nitrogen has less impact on the maximal killing activity: 8 (45%) and 18 (36%). This result suggests a critical interaction site between the chain borne by the endocyclic nitrogen of modified carbazoles and the target ((i.e. colchicine site of tubulin) that controls antitumor activity.</p><p>Sixth, a small methoxy R1 substituent in position 7 of the carbazole moiety affects the EC50 value by 30-fold (as exemplified by 8 (87 nM) and 31 (2.6 μM)), suggesting an essential interaction site between the R1 substituent in position 7 of the carbazole and the target. Remarkably, the R1 substituent in position 7 of the carbazole moiety had an opposite effect on their maximal killing activities: 31 (72%) killed more T98G cells compared with 8 (45%). Note that the positioning of the methylnaphthyl moiety has the inverse impact on the EC50 values (8 (87 nM) and 28 (957 nM)) and the maximal killing activities (8 (45%) and 28 (89%)). Together, these results show that the antitumor EC50 values (potency) and maximal killing activities (efficacy) of modified carbazoles can be mechanistically separated by specific chemical substitutions.</p><p>In conclusion, the SAR analysis of the antitumor activities of novel modified carbazoles measured in T98G cells identified chemical features that independently control their antitumor EC50s values and maximal killing activities, including a carbonyl moiety that links the carbazole to aromatic moieties, the ethyl moiety linked to the nitrogen atom of the carbazole and select chemical modifications in the second aromatic moiety.</p><!><p>To explore the therapeutic potential of the modified carbazoles, we measured their antitumor activity in human patient-derived GBM (PD-GBM) cells in culture. Specifically, these cells represent a recognized preclinical model system for testing for the efficacy of novel treatments of GBM as they are categorized in three subtypes based on their mutation profile (i.e., proneural, mesenchymal and classical) that differentially respond to treatment [16–20]. To test for potential liver toxicity, we treated the human liver cell line HepG2, a commonly used cell culture model system [21–24]. Therefore, we treated the PD-GBM cells in culture with the four most potent modified carbazoles (8, 20, 27 and 25) from our initial screen and found that they kill the PD-GBM cells with EC50 values and maximal efficacy values that both mirrored their activity on T98G cells and remained within similar values irrespective of the PD-GBM subtype (Table 3). Importantly, the four compounds exhibited lower cell-killing activity in HepG2 cells, reaching only 8–41% maximal killing activities as compared with 44–99% maximal killing activities reached in PD-GBM cells (Table 3).</p><p>The preferential sensitivity of PD-GBM cells to the killing activity of modified carbazoles compared with HepG2 cells suggests a promising therapeutic index (i.e., the ratio of the amount of agent that causes the therapeutic effect to the amount that causes toxicity). One compound, 25, demonstrated a clear-cut result by killing 95–100% of the PD-GBM cells when applied at 1–10 μM while remaining inactive at these concentrations in HepG2 cells (Table 3 and Supplementary Figure 3). Another advantage of the modified carbazoles developed here is that they kill all PD-GBM subtypes independent of their genetic make-up. This addresses one of the most thorny and extensively highlighted issues concerning the selectivity of targeting GBM tumors, which are known to rapidly become heterogeneous in their genetic make-up, often encompassing all subtypes (proneural, mesenchymal and classical) [25, 26]. Therefore, the equally potent and efficacious killing activities of modified carbazole measured in PD-GBM subtypes shows their potential to kill the majority of GBM cells forming the heterogeneous tumor mass.</p><!><p>We worked with The Developmental Therapeutics Program's NCI 60 cell screen to study the MOA by which the three most potent modified carbazoles (8, 20 and 27) kill cancer cell lines and included the inactive analogue 23 as negative control. This screening platform measures the antitumor activities of compounds expressed as percent growth inhibition, which differentiates the compound's activity at inhibiting cell proliferation and reducing cell number (i.e. killing cells) [27]. As expected, the inactive compound 23 did not significantly affect the viability and proliferation of any of the cell lines (growth inhibition: Mean 102.05; Delta 10.42, and Range 28.77) (Supplementary Figure 4). By contrast, the three active compounds 8, 20 and 27 killed 4 cell lines (13% to 60% loss in cell number), halted the proliferation of 11 cell lines (from 9% to −1% growth inhibition) and reduced the proliferation rate of 41 cell lines (from 67% to 10% percent growth inhibition) (Supplementary Figure 4). Of note, two GBM cells lines that are included in the NCI-60 panel, SF-295 and SF-539, were among the top ten cell lines that were the most susceptible to 8, 20 and 27 treatments (Figure 3 and Supplementary Table 1). The respective EC50s of the compounds measured in SF-295 and SF-530 cells mirrored their EC50s measured in T98G cells. Additional cell lines among the top ten cell lines that were sensitive to these three compounds were three cell lines of colon cancer origin: Colo205, HT-29 and HCC-2998 (Supplementary Figure 5).</p><p>The COMPARE and CellMiner web applications allow for both the statistical analysis of the antitumor activities of drugs tested in the NCI-60 cell line panel and the identification of other compounds that similarly impact the proliferation and viability of NCI 60 cells [27, 28]. Using these applications, we found three MTAs (paclitaxel, thalicarpine and maytansine) among the top eight compounds that exhibit antitumor activities in the NCI-60 cell line panel that correlate with the antitumor activities of 8, 20 and 27 (Supplementary Table 2). While this result suggests that 8, 20 and 27 kill cancer cells by disrupting MTs, the respective correlation parameters were relatively low, pointing to a somewhat different MAO as compared to that of paclitaxel, thalicarpine and maytansine, and all other MTAs tested on NCI 60 cells. Based on these results, it was reasonable to expect that the modified carbazoles would interact with tubulin or MTs through a different MOA.</p><!><p>First, we determined whether modified carbazoles compete for [3H]colchicine binding to native purified tubulin as previously described [29]. In this assay, the reference MTAs 1 (5 μM) and 3 (5 μM) competed for binding by 98% and 74%, respectively (Table 4).</p><p>Under these experimental conditions, modified carbazoles competed for colchicine binding with increasing affinity (23 << 25 < 8 < 27 < 20) that mirrored their respective EC50s at killing GBM cells. We next tested their ability to inhibit the assembly of purified tubulin using an in vitro turbidity assay as previously described [29]. In this assay, where tubulin is present at 10 μM in 0.8 M glutamate with 0.4 mM GTP and the extent of assembly is measured after 20 min at 30 °C, 1 and 3 have IC50s of 0.6 μM and 0.5 μM, respectively (Table 4). Under these experimental conditions, 8, 20 and 27 exhibited IC50s of 1.4, 0.9 and 1.6 μM, respectively, and 23 was inactive at concentrations up to 20 μM (Table 4). To build on these results, we measured the ability of modified carbazoles to influence the partitioning of MT dimers and polymers, an index of MT assembly and disassembly as previously described [14]. In this assay, compounds are incubated with sheared MTs at 37°C and the resulting amounts of tubulin dimers (free tubulin) and polymer (MT) after 15 min is measured on polyacrylamide gels [30]. Under these experimental conditions, 1 and 3 have IC50s of 0.6 and 2.4 μM, respectively; 8, 20 and 27 have IC50s of 2.1, 3.4 and 6.6 μM, respectively; and 23 was inactive when tested up to 20 μM (Table 4). Thus, the IC50s values of these modified carbazoles to bind to the colchicine site on tubulin parallels their EC50s to kill GBM cells, whereas their IC50s to trigger MT disassembly does not correlate as well and remained in the micromolar range as measured in two in vitro assays.</p><p>The analysis of the SARs of modified carbazoles to bind to the colchicine site, disrupt MT assembly and favor disassembly provide initial mechanistic insights on how their binding to the colchicine site in tubulin might favor conformational changes in this protein that might affect MT dynamics. The potency of modified carbazole derivatives are in the micromolar range when affecting MT assembly and disassembly. A possible explanation is that the affinity of modified carbazoles for the colchicine site depends on specific interactions with a subset of amino acid within the binding pocket, whereas the conformational changes in tubulin that are stabilized by the modified carbazoles are only reached at high concentrations. Thus, the modified carbazoles that we developed here may represent new chemical tools to study how binding to the colchicine site might affect tubulin conformation and how this affects MT end dynamics.</p><!><p>To understand the structural basis of the MT-disrupting and antitumor activities of the modified carbazole series, we initiated a gas phase study to determine common pharmacophoric features with sterically-relevant colchicine site agent. The structure of 20, the most potent of the series to bind to the colchicine site, was used to compare complementary features and volumetric space filling models of ligand co-crystal structures that currently exist in the Protein Data Bank (PDB) for the colchicine site on tubulin. Two MTAs that target the colchicine site, colchicine and podophyllotoxin, were found to have the most steric, hydrophilic and hydrophobic spatial similarity with the carbazole analogues. Using methodology employed previously [31, 32], the X-ray based configurations and conformations of the two chiral molecules (colchicine and podophyllotoxin, taken from the X-ray co-crystals 402B [31], and 1SA1 [33], respectively) were employed as templates for the translational and conformational overlap modeling of 20. Figure 4 displays the results of the gas phase pharmacophore overlap study for colchicine, podophyllotoxin and 20.</p><p>The results from the gas-phase pharmacophore overlap study identified the common steric features between 20, podophyllotoxin and colchicine. The atoms of 20 are almost completely subsumed within the steric space of podophyllotoxin and approximately 80% of the steric space of colchicine. The common steric overlap between these three molecules is largely composed of hydrophobic atoms. However, the two key polar atoms of 20 are the basic quinoline N and its ketone O. The quinoline N atom of 20 shows remarkable spatial consistency with the central methoxy O's of the trimethoxylated-aryl moieties of podophyllotoxin and colchicine (Figure 4), whereas the ketone O of 20 is isosteric with the lactone carbonyl of podophyllotoxin. Unique to the 20 scaffold is the relatively strong hydrogen bond acceptor quinoline N in spatial proximity to the comparatively much weaker hydrogen bond accepting ether O's of podophyllotoxin and colchicine. Furthermore, the 10-membered aromatic quinoline ring system of 20 offers unique topology to satisfy the binding requirements of the subsite normally occupied by the trimethoxylated-aryl systems of podophyllotoxins and colchicine. An additional unique feature of the 20 scaffold is the ethyl moiety attached to the carbazole N (Figure 4). The consistent polar, hydrophobic and steric overlap identified by the gas phase pharmacophore study provides a template to conduct molecular docking studies to determine a detailed, all-atom rationalization of the modified carbazole series' structure activity relationships in the colchicine site.</p><!><p>Given the close steric congruency with podophyllotoxin, we employed the alpha and beta tubulin subunits of the recently solved 2.30 Å co-crystal of the podophyllotoxin analogue: 4β-(1,2,4-triazol-3-ylthio)-4- deoxypodophyllotoxin (PDB entry code =5JCB) [34] as the initial geometry to conduct the molecular docking studies of the modified carbazole series. Maestro [34] protein preparation tools (utilizing the OPLS force field) were utilized for X-ray co-crystal comparisons and to prepare the structure for molecular mechanics energy refinement simulations. The 5JCB coordinates were optimized in a stepwise fashion (hydrogens first, followed by side chain and backbone) resulting in an all-atom energy-refined structure used as the starting structure for the carbazole molecular docking studies. Superimposition of the 5JCB X-ray coordinates onto the resulting energy refined structure revealed only a 0.98 Å RMS deviation between the 13,646 heavy atoms. Of the 6880 backbone atoms compared, only a 0.74 Å deviation was observed. These data signify that very close geometries were obtained between the X-ray and energy refined structures that are well within the expected thermodynamic variation and resolution of the 2.3 Å 5JCB X-ray structure.</p><p>Carbazole structures were built and initially optimized with the MM2 force field and transferred into the cff91/cvff force fields for potential assignment [35]. For molecular docking studies, we used methodology employed previously with added refinements [36]. Briefly, the carbazoles were docked into the energy refined 5JCB X-ray co-crystal structure to examine all translation and rotational steric complementarity between the most potent ligand of the carbazole series, 20, its conformational isomers, and the side chains of the energy-refined podophyllotoxin binding site. The resulting all-atom model was derived from iterative constrained optimizations to refine a maximum complementarity between the atoms of 20 and its contacts with the amino acid side chains of the colchicine site. Figure 5 shows the optimally docked conformation of 20 and its overlap with the triazole-podophyllotoxin analogue in the model.</p><p>An overlap that includes colchicine can be found in Supplementary Figures 6 and 7. The optimally docked model of 20 served as a template for molecular docking studies with the rest of the carbazole series to rationalize the protein-atom ligand contacts' SARs. Constrained optimizations were performed in an iterative fashion to eliminate bad contacts and biochemically unfeasible hydrophobic-polar interactions as identified by the HINT program [37]. When available, the in vitro binding data were prioritized to rank-order the carbazoles; otherwise the percent inhibition of the rank-ordered congeneric series served as the biological activity data for the SAR. A total of twenty structures (including the R and S enantiomers of 27) provided a structure-based explanation that rationalized the SAR of the carbazole series.</p><p>The modified carbazoles described herein exemplify a new chemotype for tubulin depolymerizers that interact at the colchicine site. Our modeling studies indicate that the 20 scaffold is almost entirely subsumed by the steric space of podophyllotoxin and approximately 80% within the steric space of colchicine. Potent carbazoles bind more efficiently (requiring fewer atoms) in the same steric space than podophyllotoxin because of the conformational restriction imposed by the 13-membered (carbazole) and 10-membered (quinoline/napthalene) rings systems when both are bridged by a methanone. The methanone carbonyl, which functions as a strong hydrogen bond acceptor, is important for potent activity in the carbazole series, as evidenced by a substantial loss of activity when the carbonyl is replaced by a thiocarbonyl (26). Thus, 26 forms a weaker hydrogen bond r and imposes the bulkier S atom in the sterically restricted space where the lactone carbonyl O of podophyllotoxins and carbazole Os overlap.</p><p>Although most of the atoms that are sterically congruent with podophyllotoxin and colchicine are hydrophobic, 20 introduces several unique structural elements to its colchicine site interactions. 20 poses a 10-membered quinoline aromatic ring system in the colchicine subsite normally occupied by the colchicine/podophyllotoxin trimethoxy-aryl system. With respect to enthalpic interactions, two hydrogen bond acceptors: the methanone O, and the quinoline N account for most of the potency of 20. Specifically, the methanone O can hydrogen bond with the backbone N-H of β-Asp-251, and the quinoline N located at the 5 position enables the formation of a strong can hydrogen bond with the S-H of β-Cys-241) (Supplementary Figure 6). Perhaps this occurs as the tubulin dimers polymerize and could therefore play a significant role in the prevention of MT assembly. By contrast, 22 possesses the quinoline N in the 8 position which greatly reduces activity due to its inability to form the hydrogen bonds characteristic of 20, podophyllotoxin, and colchicine (Supplementary Figure 7).</p><p>Our modeling studies indicate that the 8-quinoline N enables the formation a water-mediated intramolecular hydrogen bond that stabilizes a sterically-dissimilar conformation from both 20 and podophyllotoxin, which potentially also contributes to a loss of activity. Removal of the N atom altogether (as in in 8) to form a naphthalene, also reduces activity due to the loss of the hydrogen bonding N altogether. Our docking studies further indicate that the 4-methyl group of 8 forces the naphthalene deeper into the trimethoxy-aryl subsite to avoid unfavorable hydrophobic-polar contacts with backbone polar atoms. This then places the methanone O slightly further from the ketone O in 20's optimal orientation. The net effect is that 8's methanone O is less conformationally available to function as a strong hydrogen bond acceptor. This is also supported by 8's close congener 14, which only differs structurally by the lack of the methyl at the napthalene 4 position, and 14 is markedly less active due to loss of a hydrogen bond acceptor. Conversely, adding a hydrophobic bulk to the carbazole ring system reduces activity, as exemplified by the 7-methoxy substituted carbazole of 31 where the methyl portion of its methoxy forms unfavorable hydrophobic-polar contacts with the backbone N-H atoms of b-Lys-352, b-Val-351–349 and b-Asn-350 (Supplementary Figure 8, Panel A).</p><p>A second unique structural feature of the most potent analogue 20 is the ethyl group attached to the carbazole. Both conformational analysis and docking studies indicate that this ethyl group helps stabilize the scaffold's conformation through steric hindrance during binding. An additional attribute of the ethyl group is that it forms favorable hydrophobic contacts with the aliphatic component of the b-Lys-352 side chain (Supplementary Figure 8, Panel A). Accordingly, the less active compound 18, in which the hydrophobic distal carbon of the ethyl is replaced with an isosteric polar trifluoromethyl group, is closer to the hydrocarbon component of β-Lys-352 side chain (Supplementary Figure 8, Panel B). Moreover, replacement of this ethyl group with a bulkier hydrophobic propyl group (as with 16) reduces activity because of unfavorable hydrophobic-polar liabilities with the cationic NH3 of the β-Lys-352 side chain (Supplementary Figure 8, Panel B).</p><p>Compound 25 features a 10-membered system linked to the methanone by a N atom. Since it is a non-planar dihydroquinoline derivative, it also lacks ring aromaticity in this part of the molecule. There are several possible explanations for the reduced activity of 25. First, it could be due to the formation of a more rigid amide-like linkage that produces a less optimal orientation of the hydrogen bond accepting O due to the rigidity of the amide. Second, it could be due to the bulkier unsaturated ring that induces a twisted binding mode that potentially introduces an unfavorable hydrophobic-polar interaction liability near the side chain group of β-Ala-250. Third, it could be due to a loss of a hydrogen bond with β-Cys-241 (Supplementary Figure 9).</p><p>Several conclusions can be drawn from several less active compounds. For example, a feature to highlight is how removal of one the aromatic rings from the naphthalene/quinoline systems of the carbazoles reduces activity. Specifically, despite its p-tolyl methyl group, the single ring of 12 sterically occupies almost half of the trimethoxy-aryl subsite compared with colchicine, podophyllotoxins and 20 and yet it is less potent. An implication for this finding could be that binding to the colchicine site has a minimum steric requirement of occupation at the trimethoxy-aryl subsite to incur significant activity. However, replacement of the 10-membered ring with a 6-membered ring (as in 12 and 10) also increases the number of conformational isomers, which reduces conformational access compared to the more active carbazoles possessing double aromatic ring systems. Another example is 13, which also loses activity by reducing conformational access while increasing conformation isomerism as it extends the single ring linkage to the methanone by a methylene group. Thus, replacement of the p-tolyl methyl of 12 with a hydrophobic and isosteric Cl atom (as in 10) reduces activity by introducing similar disadvantages. The poor activity of 9 is due to the presence of a positive charge introduced by the likely protonated piperazinyl single ring system, reinforcing the finding that the trimethoxy-aryl subsite favors hydrophobic occupation with carefully positioned hydrogen bond acceptors. Accordingly, both 23 and 24 contain hydrogen bond donating N-Hs that are potentially charged.</p><p>Finally, our molecular docking studies indicate that the structural configuration of 28 requires a unique binding mode to form a biochemically-feasible complementarity with the colchicine site that accounts for its reduced biological activity compared to 20 (Supplementary Figure 10, Panel A). Specifically, the 4-tolyl-naphthalene methanone of 28 is attached directly to the carbazole N, which causes the carbazole moiety of 28 to bind less congruently with the dioxolane component of podophyllotoxin's tetracycle. Instead, one of the carbazole's aromatic rings finds steric overlap with the hydrophobic portion of the 5-membered lactone ring of the podophyllotoxin analog. Compared to the other carbazoles, 28 forms unique, favorable hydrophobic contacts with the side chain of β-Leu-248. Interestingly, the methanone O of 28 is still positioned close to both the other methanone O's of the more potent modified carbazole and the lactone carbonyl O's of podophyllotoxin (Supplemental Figure 10, Panel B). The significance of this novel binding mode is that 28 can be modified to include key features from the optimal carbazole 20, while retaining its novel interactions with the colchicine site.</p><p>Together, the structure-activity relationships of these carbazoles signify that there are key features for them to retain optimal activity in the colchicine site: 1) steric occupation of a moiety of similar size to a naphthyl, quinolinyl or trimethoxy-aryl system with preferably a hydrogen bond interacting with the β-Cys-241 sulfhydral, 2) lack of cationic ionizability for moieties that occupy the trimethoxy-aryl subsite, 3) reduction of conformational isomerism by either intramolecular steric hindrance or inherent rigidity, and 4) a strong hydrogen bond acceptor with an overlapping pose of the methanone O of 20 and the lactone carbonyl O of podophyllotoxins. Thus, the most potent 20 appears to share similar binding features with both podophyllotoxins and colchicine but provides better atom economy to achieve a comparable potency. In 20, the larger quinoline aromatic system fills the trimethoxy-aryl subsite in a sterically more compact manner than colchicine and podophyllotoxin. Moreover, the quinoline N serves as a strong hydrogen bond acceptor capable of forming diverse hydrogen bonding interactions either directly with the side chain of β-Cys-241 (Supplementary Figure 6) or via water bridges (models not shown). We have shown that the short N-ethyl substituent on the carbazole part of 20 is involved in key hydrophobic interactions with the side chain of β-Lys-352 (Supplementary Figure 8, Panel A). Our molecular docking studies provide a unified structure-based explanation for both positive and negative substitutions according to the SAR of the carbazole analogs.</p><!><p>We report here a novel series of modified carbazoles that destabilize MTs by binding to the colchicine site of tubulin in a similar mode to a podophyllotoxin analogue and appears to interact with a unique low interaction binding space. Several modified carbazoles trigger marked cell death in multiple GBM model systems while exhibiting a much less lower activity in the HepG2 liver cells, suggesting a promising therapeutic index. The use of the heterocyclic carbazole scaffold provides several advantages when considering the future optimization of small-molecule for therapeutic use. For example, this scaffold can be modified in a variety of ways using readily available and simple starting materials, and their chemical modification typically requires only as few steps of well-established chemistry, thereby providing a versatile scaffold for medicinal chemistry optimization. In summary, this work increases our understanding of how targeting the colchicine site by small molecules affects tubulin assembly and disrupts MT function, and provides a reference design approach to develop the next generation of MTAs for treating devastating cancers, such as GBM.</p><!><p>Moisture sensitive reactions were performed in an inert, dry atmosphere of Ar in flame-dried glassware. Air sensitive liquids were transferred via syringe or cannula through rubber septa. Reagent grade solvents were used for extraction and flash chromatography. THF was distilled from Na/benzophenone under Ar. All commercially obtained reagents and solvents were used directly without further purification. The progress of reactions was checked by analytical thin-layer chromatography (Silica G TLC plates w/UV 254). Flash column chromatography was performed using prepacked Biotage SNAP/ZIP cartridges on a Biotage Isolera One instrument. The solvent compositions reported for all chromatographic separations are on a volume/volume (v/v) basis. 1H NMR spectra were recorded at 400 MHz and are reported in parts per million (ppm) on the δ scale relative to tetramethylsilane as an internal standard. 13C NMR spectra were acquired at 101 MHz operating with 1H decoupling and are reported in parts per million (ppm) on the δ scale relative to CDCl3. Melting points were determined on a Stuart melting point apparatus from Bibby Scientific Limited and are uncorrected. LC/MS and HRMS analyses were obtained on a Waters ACQUITY UPLC-series liquid chromatography system equipped with a diode array detector and coupled to a LCT PREMIER XE™ time of flight (TOF) mass spectrometer with electrospray ionization (ESI). The liquid chromatography conditions were as follows: a Phenomenex column (NX, 3μ, C18, 110A, 50.0×4.60 mm) was used, and bound compounds were eluted with the following gradient over 15 min at a rate of 0.4mL/min (water (0.1% formic acid)/acetonitrile): (90/10 to 2/98, 0–6.6 min), (2/98 isocratic, 6.6–13 min), (2/98 to 90/10, 13–15 min). Compound purity was assigned based on 254 nM detection data assessed by comparing relative peak areas of the signals. All final compounds were more than 95% pure.</p><!><p>Under Ar, a solution of carbazole (5.0 g, 29.9 mmol), bromoethane (4.45 mL, 35.9 mmol), and Cs2CO3 (14.6 g, 44.8 mmol) in DMF (20 mL) was stirred at 80°C for 16 h. The reaction mixture was cooled, diluted with EtOAc (50 mL), and filtered. The organic solvents were evaporated in vacuo. The resultant dark oil was purified by column chromatography on silica gel using heptanes/EtOAc in different proportions to afford the title compound as a light-yellow oil (5.075 g, 87%). 1H NMR (500 MHz, CDCl3) δ ppm 8.18 (d, J=7.79 Hz, 2H), 7.54 (t, J=7.66 Hz, 2H), 7.46 (d, J=8.06 Hz, 2H), 7.26 – 7.35 (m, 2H), 4.40 (q, J=7.25 Hz, 3H), 1.41 – 1.52 (m, 3H) 13C NMR (126 MHz, CDCl3) δ ppm 139.85, 125.54, 122.85, 120.37, 118.68, 108.37, 37.42, 13.75.</p><!><p>POC13 (1.4 mL, 10.2 mmol) was added, over a period of 10 min, to ice-cooled, stirred DMF (3.2 mL, 40.8 mmol) under Ar. The reddish solution was stirred at room temperature for 1 h. 9-ethyl-9H-carbazole 5 (1.0 g, 5.1 mmol) was added over 10 min, and the mixture was subjected to microwave irradiation at 100 °C for 1 h. The reaction mixture was cooled and then poured into crushed ice. After warming to room temperature, the resultant product was extracted with EtOAc. The organic phase was washed with water, and brine, dried (MgSO4), filtered, and evaporated in vacuo. The residue was purified by column chromatography on silica gel using heptanes/EtOAc in different proportions to afford the title compound as a white solid (1.054 g, 93%); 1H NMR (400 MHz, CDCl3) δ ppm 10.03 (s, 1H), 8.57 (s, 1H), 8.12 (d, J = 7.7 Hz, 1H), 7.94 (d, J = 8.5 Hz, 1H), 7.52 (t, J = 7.6 Hz, 1H), 7.41 (d, J = 8.6 Hz, 2H), 7.30 (t, J = 7.4 Hz, 1H), 4.26 (t, J=7.2 Hz, 2H), 1.45 (t, J = 7.2 Hz, 3H). 13C NMR (101 MHz, CDCl3) δ ppm 191.83, 144.18, 141.27, 128.56, 127.23, 126.82, 124.06, 123.14, 123.08, 120.84, 120.40, 109.53, 109.05, 42.76, 13.91.</p><!><p>To an ice-cold solution of 9-ethyl-3-carbaldehyde 6 (1.0 g, 4.5 mmol) in water/acetone (50 mL, 1:1) was added dropwise with stirring a solution of potassium permanganate (711 mg, 4.5 mmol) in acetone (25 mL). The mixture was heated 3 h at reflux and then allowed to cool to room temperature. After that, the reaction mixture was quenched with ethanol (20 mL) and then stirred for 30 min at reflux. After cooling to room temperature, the mixture was filtered through a pad of Celite© and concentrated in vacuo. The concentrated solution was diluted with water (100 mL), basified with NaOH to pH ca. 10, and extracted with heptane/ether (4:1, 50 mL × 3) to remove the unreacted starting material. The aqueous solution was cooled on an ice-water bath and then acidified with an ice-cold solution of 2 N HCl to pH ca. 2. The resultant precipitate was extracted with EtOAc (150 mL). The organic layer was washed with brine (30 mL), dried over MgSO4, filtered, and concentrated in vacuo. The precipitated product was collected by filtration, washed with heptanes (20 mL), and dried overnight to produce the title compound 3 (527.2 mg, 49%) as a yellow oil; 1H NMR (500 MHz, CDCl3) δ ppm 12.34 (s, 1H), 8.89 (d, J = 1.6 Hz, 1H), 8.31 (dd, J = 8.6, 1.6 Hz, 1H), 8.12 (d, J = 7.7 Hz, 1H), 7.51 (ddd, J = 8.3, 7.1, 1.3 Hz, 1H), 7.37 (d, J = 8.2 Hz, 1H), 7.34 (d, J = 8.7 Hz, 1H), 7.28 (ddd, 7.9, 6.9, 1 Hz, 1H), 4.32 (t, J = 7.2 Hz, 2H), 1.45 (t, J = 7.2 Hz, 3H). 13C NMR (126 MHz, CDCl3) δ ppm 173.30, 143.47, 141.17, 128.13, 126.62, 123.87, 123.14, 122.67, 120.88, 120.33, 119.82, 109.34, 108.45, 42.57, 14.01.</p><!><p>Under Ar, AlCl3 (199 mg, 1.5 mmol) was added to a solution of 9-ethyl-9H-carbazole 7 (300 mg, 1.25 mmol) in anhydrous benzene (30 mL), and the solution was cooled by an ice bath for 20 min. 4-methyl-1-naphthoyl chloride (282 mL, 2.43 mmol) was added dropwise via a syringe to the solution, and the reaction mixture was stirred for 16 h while warming at room temperature. The reaction mixture was cooled on an ice-water bath, then poured onto a mixture of ice and a 4 M NaOH solution (50 mL) and extracted with ethyl acetate (150 mL). The obtained residue was purified by column chromatography on silica gel eluting with EtOAc/heptanes in different proportions to yield the target product (296.3 mg, 64%) as a light yellow solid: 1H NMR (400 MHz, CDCl3) δ ppm 8.63 (s, 1H), 8.00 – 8.18 (m, 4H), 7.37 −7.61 (m, 7H), 7.24 – 7.27 (m, 1H), 4.41 (q, J = 7.28 Hz, 2H), 2.81 (s, 3H), 1.47 (t, J = 7.28 Hz, 3H) 13C NMR (101 MHz, CDCl3) δ ppm 197.9, 142.9, 140.7, 137.3, 136.2, 132.9, 131.3, 129.8, 128.6, 126.9, 126.6, 126.6, 126.5, 126.2, 125.3, 124.4, 124.3, 123.3, 122.7, 120.9, 120.0, 109.0, 108.1, 37.9, 19.9, 13.9. HRMS calcd for C26H22NO (M+H)+: 364.1684, found HRMS: 364.1701.</p><!><p>9-ethyl-9H-carbazole-3-carboxylic acid 7 (200 mg, 0.84 mmol), 1-methylpiperazine (171 mg, 1.29 mmol), DIPEA (284 μL, 1.7 mmol), and DMAP (122 mg, 1 mmol) were added to DCM (26 mL) under N2. The obtained solution was cooled on an ice-water bath. EDC (275 mg, 1.44 mmol) was added to the solution, and the reaction mixture was then allowed to warm to room temperature and stirred for 16 h. The solvent was removed in vacuo, and the obtained residue was extracted into ethyl acetate (100 mL). The organic layer was washed consecutively with 5% citric acid solution (50 mL × 3), concentrated NaHCO3 (50 mL × 3), brine (50 mL), dried over MgSO4, filtered, and concentrated in vacuo. The residue was purified on silica gel using heptanes/ethyl acetate in different proportions to afford 199 mg (74%) of 9 as a yellow glass. 322.1952 1H NMR (400 MHz, CDCl3) δ ppm 8.24 (s, 1H), 8.10 (d, J=7.45 Hz, 1H), 7.50 – 7.59 (m, 2H), 7.41 – 7.48 (m, 2H), 7.29 (t, J=7.53 Hz, 1H), 4.40 (q, J=7.19 Hz, 2H), 3.72 (br s, 1H), 2.52 (br s, 1H), 2.44 (s, 3H), 2.17 (s, 1H), 1.40 – 1.49 (m, 4H) 13C NMR (101 MHz, CDCl3) δ ppm 167.78, 141.17, 140.49, 126.63, 125.27, 122.83, 122.51, 120.72, 120.66, 119.79, 108.93, 108.46, 53.47, 43.58, 37.79, 26.89, 13.78. HRMS calcd for C20H24 N3 O (M+H)+: 322.1913, found: 322.1952.</p><!><p>Under Ar, AlCl3 (1.13 g, 8.44 mmol) was added to a solution of 5 (1.5 g, 7.68 mmol) in anhydrous benzene (20 mL), and the solution was cooled by an ice bath for 20 min. 4-chlorobenzoic acid (1.12 mL, 8.83 mmol) was added dropwise via a syringe to the solution, which was tightly capped in a microwave vessel and subjected to microwave irradiation at 100 °C for 1 h and then cooled to room temperature. The reaction mixture was cooled on an ice-water bath, then poured onto a mixture of ice and a 4 M NaOH solution (50 mL) and extracted with ethyl acetate (150 mL). The organic phase was washed with saturated aqueous NaHCO3 and brine, dried (MgSO4), filtered and evaporated in vacuo. The residue was purified by column chromatography on silica gel eluting with EtOAc/heptanes in different proportions to yield the target product (1.374 g, 54%) as a white solid: 1H NMR (400 MHz, CDCl3) δ ppm 8.63 (d, J=1.30 Hz, 1H), 8.16 (d, J=7.77 Hz, 1H), 8.04 (dd, J=8.58, 1.67 Hz, 1H), 7.87 – 7.98 (m, 2H), 7.56 (td, J=7.64, 1.03 Hz, 1H), 7.45 – 7.53 (m, 2H), 7.29 – 7.37 (m, 1H), 7.17 – 7.28 (m, 2H), 4.43 (q, J=7.23 Hz, 2H), 1.51 (t, J=7.23 Hz, 3H) 13C NMR (101 MHz, CDCl3) δ ppm 195.06, 166.11, 163.60, 142.43, 140.57, 135.12, 135.09, 132.36, 132.28, 128.25, 128.22, 126.46, 123.76, 123.04, 122.51, 120.68, 119.92, 115.30, 115.08, 108.95, 107.99, 37.75, 13.74. ESI: m/z 334.1 (M + H)+. HRMS calcd for C21H17ClNO (M + H)+ 334.0999, found 334.0981.</p><!><p>Using 9-ethyl-9H-carbazole 5 (1.57 g, 8.05 mmol) and 4-fluorobenzoic acid (1.1 mL, 9.26 mmol), as starting compounds, the title compound was prepared by the procedures described in the preparation of compound 10 to yield 1.07 g (42%) of 11 as an orange oil. 1H NMR (400 MHz, CDCl3) δ ppm 8.56 (d, J = 1.7 Hz, 1H), 8.09 (d, J = 7.8 Hz, 1H), 7.97 (dd, J = 8.6, 1.7 Hz, 1H), 7.86 (dd, J = 8.6, 5.6 Hz, 2H), 7.50 (ddd, J = 8.3, 7.0, 1.2 Hz, 1H), 7.43 – 7.40 (m, 2H), 7.33 – 7.23 (m, 1H), 7.17 (t, J = 8.6 Hz, 2H), 4.36 (q, J = 7.2 Hz, 2H), 1.44 (t, J = 7.2 Hz, 3H). 13C NMR (101 MHz, CDCl3) δ ppm 195.28, 142.56, 140.62, 137.93, 137.30, 131.28, 128.60, 128.45, 128.26, 128.06, 126.53, 123.89, 123.08, 122.58, 120.75, 120.01, 109.00, 108.07, 37.83, 26.87, 13.79. HRMS calcd for C21H17FNO (M + H)+ 318.1294, found 318.0766.</p><!><p>Using 9-ethyl-9H-carbazole 5 (325 mg, 1.67 mmol) and p-toluoyl chloride (246 mg, 2.00 mmol) as starting compounds, the title compound was prepared by the procedures described in the preparation of compound 10 to yield 241 mg (46%) of 12 as a yellow glass. 1H NMR (400 MHz, CDCl3) δ ppm 8.63 (d, J=1.30 Hz, 1H), 8.16 (d, J=7.77 Hz, 1H), 8.04 (dd, J=8.58, 1.67 Hz, 1H), 7.64 – 7.73 (m, 2H), 7.56 (td, J=7.64, 1.03 Hz, 1H), 7.29 – 7.37 (m, 1H), 7.17 – 7.28 (m, 4H), 4.43 (q, J=7.23 Hz, 2H), 2.31 (s, 3H) 1.51 (t, J=7.23 Hz, 3H) 13C NMR (101 MHz, CDCl3) δ ppm 197.02, 143.46, 141.52, 135.19, 135.07, 131.36, 131.28, 128.15, 128.12, 126.46, 123.76, 122.64, 122.22, 120.66, 119.72, 115.90, 115.68, 108.95, 107.54, 36.98, 21.43, 13.13. HRMS calcd for C22H20NO (M+H)+: 314.1539, found: 314.1545.</p><!><p>Using 9-ethyl-9H-carbazole 5 (325 mg, 1.67 mmol) and phenylacetyl chloride (246 mg, 2.00 mmol) as starting compounds, the title compound was prepared by the procedures described in the preparation of compound 10 to yield 299 mg (57%) of 13 as a yellow glass. 1H NMR (400 MHz, CDCl3) δ ppm 8.81 (d, J=1.30 Hz, 1H), 8.24 (d, J=7.77 Hz, 1H), 8.09 (dd, J=8.58, 1.67 Hz, 1H), 7.59 – 7.69 (m, 2H), 7.29 – 7.37 (m, 2H), 7.23 – 7.36 (m, 5H), 4.43 (q, J=7.23 Hz, 2H), 4.12 (s, 2H), 2.31 (s, 3H), 1.51 (t, J=7.23 Hz, 3H) 13C NMR (101 MHz, CDCl3) δ ppm 195.38, 141.72, 135.66, 129.74, 129.13, 128.19, 128.42, 126.35, 124.32, 123.64, 122.22, 120.63, 119.79, 115.57, 115.35, 108.78, 107.54, 44.43, 36.98, 13.13. HRMS calcd for C22H20NO (M-H)− 314.1539, found: 314.1530.</p><!><p>Using 9-ethyl-9H-carbazole 4 (100 mg, 0.40 mmol) and 1-naphthoyl chloride (74 mL, 0.71 mmol) as starting compounds, the title compound was prepared following the procedures described in the preparation of compound 10. A yellowish viscous oil was obtained. Yield: 141.1 mg (41.5%). 1H NMR (400 MHz, CDCl3) δ 8.65 (s, 1H), 8.16(d, 1H), 8.04 (t, 2H), 7.96 (d, 1H), 7.51(d, 2H), 7.42(m, 2H), 7.33 (m, 4H), 7.18(t, 1H), 4.18 (q, 2H), 1.45 (t, 3H). 13C NMR (101 MHz, CDCl3) δ 197.79, 143.02, 140.73, 137.27, 133.77, 130.44, 128.64, 128.37, 126.56, 126.38, 125.97, 124.58, 124.33, 120.89, 120.12, 109.05, 108.13, 37.92, 13.87. HRMS calcd for C25 H19 NO (M+H)+: 350.1466, found: 350.1536.</p><!><p>Using carbazole (550 mg, 2.52 mmol) and 1-bromopropane (369 mg, 3 mmol), as the starting compounds, the title compound was prepared by the procedures described in the preparation of compound 10 to yield 410 mg (74%) of 15 as a yellow oil. 1H NMR (400 MHz, CDCl3) δ 8.09 (d, J = 7.8 Hz, 2H), 7.44 (ddd, J = 8.2, 6.9, 1.2 Hz, 2H), 7.39 (d, J = 8.1 Hz, 2H), 7.21 (ddd, J = 8.0, 6.8, 1.1 Hz, 2H), 4.24 (t, J = 7.2 Hz, 2H), (m, 2H), 0.95 (t, J = 7.4 Hz, 3H). 13C NMR (101 MHz, CDCl3) δ 140.71, 125.75, 123.00, 120.53, 118.90, 108.89, 44.80, 22.51, 12.02.</p><!><p>Using 9-propyl-9H-carbazole (350 mg, 1.67 mmol) and 4-methyl-1-naphthoyl chloride (246 mg, 2.00 mmol) as starting compounds, the title compound was prepared by the procedures described in the preparation of compound 10 to yield 327 mg (52%) of 16 as a yellow glass. 1H NMR (400 MHz, CDCl3) δ ppm 8.62 (s, 1H), 8.14 (d, J=8.53 Hz, 1H), 8.04 (t, J=9.03 Hz, 2H), 7.96 (d, J=7.78 Hz, 1H), 7.47 – 7.55 (m, 2H), 7.38 – 7.47 (m, 2H), 7.27 – 7.38 (m, 3H), 7.18 (t, J=7.40 Hz, 1H), 4.18 (t, J=7.03 Hz, 2H), 2.73 (s, 3H), 1.78 – 1.90 (m, 2H), 0.97 (t, J=7.40 Hz, 3H) 13C NMR (101 MHz, CDCl3) δ ppm 197.63, 143.30, 141.06, 137.08, 132.71, 129.58, 128.35, 126.76, 126.41, 126.04, 125.42, 124.24, 122.97, 122.41, 120.53, 119.84, 118.56, 109.16, 108.57, 44.62, 60.38, 22.72, 19.67, 11.83. HRMS calcd for C27H24NO (M+H)+: 378.1852, found: 378.1851.</p><!><p>Using carbazole (300 mg, 1.80 mmol) and 2-chloro-1,1,1-trifluoroethane (0.25 mL, 1.28 mmol) as starting compounds, the title compound was prepared by the procedures described in the preparation of compound 10 to yield 372 mg (83%) of 17 as a yellow oil. 1H NMR (500 MHz, CDCl3) δ ppm 8.17 (d, J=7.86 Hz, 2H), 7.53 (t, J=7.63 Hz, 2H), 7.37 (d, J=7.86 Hz, 2H), 7.22 – 7.32 (m, 2H), 4.35 (s, 2H), 13C NMR (126 MHz, CDCl3) δ ppm 140.14, 127.44, 122.91, 120.53, 118.77, 111.23, 108.37, 72.14, 19.7.</p><!><p>Using 17 (350 mg, 1.41 mmol) and 4-methyl-1-naphthoyl chloride (246 mg, 2.00 mmol) as starting compounds, the title compound was prepared by the procedures described in the preparation of compound 10 to yield 363 mg (62%) of 18 as a yellow glass.1H NMR (400 MHz, CDCl3) δ 8.64 (s, 1H), 8.00 – 8.19 (m, 4H), 7.36 – 7.59 (m, 7H), 7.24 – 7.27 (m, 1H), 4.37 (s, 2H), 2.79 (s, 3H), 13C NMR (101 MHz, CDCl3) δ ppm 197.6, 142.8, 139.7, 136.6, 136.1, 133.2, 131.2, 129.5, 128.8, 126.7, 126.9, 126.7, 126.5, 126.2, 125.2, 124.9, 124.6, 123.3, 122.4, 120.9, 120.0, 112.7 109.8, 108.6, 71.8, 19.7. HRMS calcd for C26H19F3NO (M+H)+: 418.1413, found: 418.1345.</p><!><p>A solution of 1.5 M tert-buthyllithium in pentane (9 mL) was added dropwise in 20 min. at −78 °C to a solution of 3-bromo-9-ethyl-9H-carbazole (1.9 g; 0.007 mol) dissolved in dry THF (50 mL). The solution was stirred at −78 °C for 1 h. A brownish precipitate formed. TLC in cyclohexane/DCM 9/1: 100 % conversion. A solution of quinoline-5-carbaldehyde (1.1 g; 0.007 mol) dissolved in 50 mL of dry THF was added dropwise at −78 °C in 15 min. The resulting solution was stirred at −78 °C for 1.5 h. 220 mL of a saturated NH4Cl solution was added. The product was then extracted with EtOAc (200 mL and then 50 mL). The organic phases were combined and washed with water (2×100 mL) and dried over MgSO4. The solution was concentrated under reduced pressure. The residue was purified by column chromatography on silica gel eluting with EtOAc/cyclohexanes to afford the pure product as a pale yellow solid. Yield: 1.54 g (62.4%). 1H NMR (400 MHz, CDCl3) δ 8.78 (d, J = 4.0, 1H), 8.40 (d, J = 8.6, 1H), 8.12 (s, 1H), 8.06 (d, J = 8.4, 1H), 8.02 (d, J = 7.8, 1H), 7.82 (d, J = 7.1, 1H), 7.78 – 7.67 (m, 1H), 7.45 (t, J = 7.6, 1H), 7.39 (td, J = 4.7, 2.3, 2H), 7.31 (d, J = 8.5, 1H), 7.23 (d, J =4.2, 1H), 7.20 (t, J = 7.3, 1H), 6.65 (s, 1H), 4.32 (q, J = 7.2, 2H), 1.39 (t, J = 7.2, 3H). 13C NMR (101 MHz, CDCl3) δ 149.94, 139.74, 139.66, 133.53, 132.82, 129.69, 128.90, 125.94, 124.93, 124.89, 120.80, 120.53, 119.17, 119.01, 108.70, 108.61, 74.18, 37.65, 13.83.</p><!><p>A suspension of 19 (1.4g; 3.97 mmol) dissolved in DCM (20 mL), PDC (2.95 g, 7.84 mmol) and molecular sieves 4 Å (2.95 g) was stirred for 2 h at r.t. The resulting solution was filtered over a pad of SiO2 and eluted with AcOEt. 350 mL of solvent (AcOEt and DCM) was collected, and the solutions were combined and concentrated under reduced pressure to afford the pure product as a pale yellow solid. Yield: 995 mg (78.7%). 1H NMR (400 MHz, CDCl3) δ 9.04 – 8.92 (m, 1H), 8.62 (s, 1H), 8.49 (d, J = 8.5, 1H), 8.31 (d, J = 8.3, 1H), 8.06 (d, J = 7.4, 2H), 7.87 – 7.77 (m, 1H), 7.75 (d, J = 6.9, 1H), 7.52 (t, J = 7.6, 1H), 7.49 – 7.43 (m, 2H), 7.43 – 7.39 (m, 1H), 7.31 – 7.25 (m, 2H), 4.42 (q, J = 7.2, 2H), 1.48 (t, J = 7.2, 3H). 13C NMR (101 MHz, CDCl3) δ 196.38, 150.91, 148.35, 143.07, 140.74, 137.72, 134.31, 132.13, 129.16, 128.54, 128.02, 127.78, 126.69, 126.63, 124.37, 123.19, 122.84, 121.95, 120.87, 120.24, 109.11, 108.24, 37.96, 13.86. HRMS calcd for C24H19N2O (M+H)+: 351.1491, found: 351.1487.</p><!><p>To an oven-dried round-bottom flask flushed with N2 was added 3-bromo-9-ethyl-9H-carbazole (508 mg, 1.85 mmol) in 13 mL dry THF. The mixture was cooled to −78°C and tert-BuLi (2M in heptane) (1.85 mL, 3.7 mmol) was added dropwise. The mixture was stirred at −78°C for 1 hr, then 8-Quinolinecarboxaldehyde (291 mg, 1.85 mmol) was added. The resulting mixture was stirred at −78°C for 1.5 hr, then allowed to warm to 0°C. 50 mL sat. NH Cl was added, then organics were extracted with EtOAc (2×30 mL), washed with water and brine, and dried over MgSO4. Solvents were removed in vacuo, and the resulting oil was purified by column chromatography eluting with a gradient of 12–100% EtOAc in Heptane to yield the title compound as a dark purple oil (303 mg, 46.4% yield). 1H NMR (400 MHz, Chloroform-d) δ 8.90 (dd, J = 4.3, 1.8 Hz, 1H), 8.26 – 8.24 (m, 1H), 8.22 (dd, J = 8.4, 1.8 Hz, 1H), 8.06 (dt, J = 7.8, 1.0 Hz, 1H), 7.75 (dd, J = 8.0, 1.7 Hz, 1H), 7.60 (dd, J = 8.4, 1.7 Hz, 1H), 7.45 (dd, J = 6.1, 1.8 Hz, 2H), 7.39 (d, J = 5.0 Hz, 1H), 7.19 (ddd, J = 7.9, 6.9, 1.2 Hz, 1H), 6.97 (s, 1H), 6.67 (s, 1H), 4.36 (q, J = 7.2 Hz, 2H), 1.42 (t, J = 7.2 Hz, 3H).</p><!><p>A suspension of 21 (300 mg, 0.85 mmol) dissolved in DCM (5 mL), PDC (641 mg, 1.70 mmol) and molecular sieves 4 Å (632 mg) was stirred for 3 h. at r.t. The resulting solution was filtered under a pad of SiO2, and eluted with AcOEt. 350 mL of solvent (AcOEt and DCM) were collected and the solutions were combined and then concentrated in vacuo. The resulting oil was purified using column chromatography eluting with a gradient of 7–50% EtOAc in Heptane to afford the pure product as a bright-orange solid (73 mg, 24.4% yield). 1H NMR (400 MHz, Chloroform-d) δ 8.87 (dd, J = 4.2, 1.8 Hz, 1H), 8.60 (d, J = 1.7 Hz, 1H), 8.25 (dd, J = 8.3, 1.8 Hz, 1H), 8.05 (dd, J = 8.7, 1.7 Hz, 1H), 8.03 – 7.97 (m, 2H), 7.81 (dd, J = 7.0, 1.5 Hz, 1H), 7.67 (dd, J = 8.2, 7.0 Hz, 1H), 7.48 (ddd, J = 8.3, 7.0, 1.2 Hz, 1H), 7.44 – 7.40 (m, 2H), 7.38 (d, J = 8.7 Hz, 1H), 7.23 (ddd, J = 8.0, 7.0, 1.1 Hz, 1H), 4.39 (q, J = 7.2 Hz, 2H), 1.45 (t, J = 7.2 Hz, 3H). 13C NMR (101 MHz, CDCl3) δ 197.39, 150.97, 146.36, 143.05, 140.69, 140.37, 136.04, 129.42, 129.26, 128.56, 128.37, 128.06, 126.37, 125.92, 124.20, 123.44, 122.79, 121.58, 120.83, 119.96, 108.97, 108.02, 77.36, 77.25, 77.05, 76.73, 37.86, 13.83. HRMS calcd for C24H18N2O (M+H)+: 351.1492, found: 351.1656.</p><!><p>To a solution of 20 (106.3 mg, 0.30 mmol) in MeOH (1.5 mL) and THF (1 mL) was added sodium cyanoborohydride (95.46 mg, 1.52 mmol). Boron trifluoride diethyl etherate (190.9 L, 1.52 mmol) was added dropwise to the resulting solution, with evolution of gas. The solution was stirred and refluxed at 63 °C for 3.5 h under N2. 3 mL NH3 (25% in water) was added to the reaction mixture, then diluted with another 10 mL of water. The product was extracted with EtOAc (2 × 15 mL), and washed with brine (1 × 30 mL). The organic layers were dried over anhydrous MgSO, and the solvent was removed in vacuo. Raw product was purified with flash chromatography (16 – 40% EtOAc in heptane) to yield 23 as a sticky yellow solid (43.9 mg, 41.3% yield). 1H NMR (400 MHz, Chloroform-d) δ 8.68 – 8.52 (m, 1H), 8.11 (d, J = 7.7 Hz, 0H), 8.08 – 8.00 (m, 1H), 7.51 (ddd, J = 8.1, 6.3, 2.6 Hz, 1H), 7.42 (dd, J = 15.0, 8.5 Hz, 1H), 7.32 – 7.27 (m, 0H), 7.05 (t, J = 7.7 Hz, 1H), 6.75 – 6.65 (m, 1H), 6.61 (dt, J = 8.1, 2.2 Hz, 1H), 4.40 (q, J = 7.2 Hz, 1H), 4.10 (s, 1H), 3.32 (dd, J = 7.0, 3.9 Hz, 1H), 2.72 (t, J = 6.2 Hz, 1H), 1.88 (p, J = 6.0 Hz, 1H), 1.47 (t, J = 7.2 Hz, 2H). 13C NMR (101 MHz, CDCl3) δ 198.70, 145.01, 142.90, 140.68, 140.54, 129.03, 128.31, 126.44, 126.14, 124.07, 123.36, 122.73, 120.90, 120.03, 119.09, 116.93, 115.51, 108.97, 108.01, 41.73, 37.88, 24.76, 21.89, 13.85. HRMS calcd for C24H22N2O (M+H)+: 355.1804, found: 355.1810.</p><!><p>To a solution of 22 (60 mg, 0.171 mmol) in MeOH (0.75 mL) and THF (0.5 mL) was added sodium cyanoborohydride (54.8 mg, 0.856 mmol). Boron trifluoride diethyl etherate (108 L, 0.856 mmol) was added dropwise to the resulting solution, with evolution of gas. The solution was stirred and refluxed at 60⁰C for 3.5 hours under nitrogen. 2 mL ammonia (25% in water) was added to the reaction mixture, which was then diluted with another 10 mL of water. The product was extracted with EtOAc (2 × 15 mL) and washed with brine (1 × 30 mL). The organic layers were dried over anhydrous MgSO4 and the solvent was removed in vacuo. Raw product was purified with flash chromatography twice eluting with a gradient of 15–55% EtOAc in Heptane and then 10–100% DCM in heptane to yield 24 as a yellow oil (12.6 mg, 20.8% Yield). 1H NMR (400 MHz, Chloroform-d) δ 8.43 (t, J = 2.5 Hz, 2H), 8.10 (d, J = 7.7 Hz, 1H), 7.84 (dd, J = 8.5, 1.7 Hz, 1H), 7.50 (ddd, J = 8.2, 6.9, 1.2 Hz, 1H), 7.47 – 7.40 (m, 3H), 7.27 (d, J = 8.1 Hz, 1H), 7.08 (dd, J = 7.1, 1.4 Hz, 1H), 6.47 – 6.39 (m, 1H), 4.41 (q, J = 7.2 Hz, 2H), 3.49 (td, J = 5.9, 2.7 Hz, 2H), 2.86 (t, J = 6.3 Hz, 2H), 2.02 – 1.94 (m, 2H), 1.47 (t, J = 7.2 Hz, 3H). 13C NMR (101 MHz, CDCl3) δ 199.08, 148.76, 141.63, 140.59, 133.38, 133.30, 131.50, 127.75, 126.19, 123.24, 122.92, 122.46, 122.28, 120.77, 119.61, 117.18, 112.80, 108.83, 107.77, 77.35, 77.23, 77.03, 76.71, 41.26, 39.00, 37.80, 35.45, 34.15, 31.91, 29.52, 29.45, 29.05, 27.93, 25.05, 22.71, 22.68, 20.81, 20.18, 19.19, 14.42, 14.12, 13.85, 11.41, 10.97.HRMS calcd C24H22N2 O (M+H)+: 355.1805, found 355.1815.</p><!><p>Using 9-ethyl-9H-carbazole-3-carboxylic acid 7 (200 mg, 0.84 mmol) and 1,2,3,4-tetrahydroquinoline (171 mg, 1.29 mmol) as starting compounds, the title compound was prepared following the procedures described for the preparation of compound 9 to yield 241 mg (81%) of 25 as a yellow glass. 1H NMR (400 MHz, CDCl3) δ ppm 8.62 (s, 1H), 8.04 (t, J=9.03 Hz, 2H), 7.96 (d, J=7.78 Hz, 1H), 7.41 – 7.54 (m, 3H), 7.08 (t, J=7.40 Hz, 1H), 6.89 – 7.01 (m, 3H), 4.38 (t, J=7.03 Hz, 2H), 2.73–2.84 (m, 2H), 1.54 – 1.70 (m, 4H), 1.28 (t, J=7.40 Hz, 3H). 13C NMR (101 MHz, CDCl3) δ ppm 166.05, 141.72, 135.50, 131.12, 129.53, 128.00, 126.77, 125.17, 124.85, 124.46, 123.83, 123.56, 122.45, 121.39, 118.95, 118.55, 118.26, 107.58, 106.98, 51.10, 42.14, 28.87, 25.32, 20.09, 14.72. HRMS calcd for C24H23N2O (M+H)+: 355.1804, found: 355.1831.</p><!><p>Under Ar, a solution of carbazole 8 (62 mg, 0.17 mmol) and Lawesson's reagent (49 mg, 0.12 mmol) in toluene (3 mL) was tightly capped in a 5 mL microwave vessel. The mixture was subjected to microwave irradiation at 140 °C for 4 h and then cooled to r.t. The organic solvent was evaporated in vacuo, and the residue was purified by column chromatography on silica gel using heptanes/ethyl acetate in different proportions to yield thioamide 26 as a yellow glass. Yield: 20 mg (31%). 1H NMR (400 MHz, CDCl3) δ ppm 8.63 (s, 1H), 8.10 (m, 4H), 7.49 (m, 7H), 7.26 (t, 1H), 4.41 (q, 2H), 2.82 (s, 3H), 1.46 (t, 3H). 13C NMR (101 MHz, CDCl3) δ ppm 218.89, 142.92, 140.70, 137.27, 129.82, 128.61, 126.94, 126.60, 126.49, 126.21, 125.32, 124.35, 120.88, 120.05, 109.01, 108.06, 37.91, 19.92, 13.87. HRMS calcd for C26H22NS (M+H)+: 380.1467, found: 380.1498.</p><!><p>Carbazole 8 (2.59 g; 0.071 mol) was dissolved in 62 mL of dioxane, and 100 mL of MeOH was added. NaBH4 (400 mg; 0.01 mol) was added to the solution over 10 min. The solution was stirred for 14 h at room temperature. 110 mg of NaBH4 was added to complete the reduction, and the reaction mixture was stirred at r.t. for 14 h. 250 mL of water was added. The product was then extracted with EtOAc (250 mL) from water. The organic phases were combined and washed with water (2×100 mL) and dried over MgSO4. The solution was concentrated under reduced pressure. Flash-chromatography of the crude mixture using a DCM/Cyclohexane gradient afforded the pure product as a pale yellow solid. Yield: 1.735 g (66.6%). 1H NMR (400 MHz, CDCl3) δ 8.18 (s, 1H), 8.12 (d, J=8.2, 1H), 8.05 (dd, J=15.5, 8.8, 2H), 7.65 (d, J=7.3, 1H), 7.53 – 7.37 (m, 6H), 7.37 – 7.29 (m, 2H), 7.23 – 7.15 (m, 1H), 6.74 (d, J=3.8, 1H), 4.34 (q, J=7.2, 2H), 2.72 (s, 3H), 2.32 (d, J=3.9, 1H), 1.44 – 1.36 (m, 8H). 13C NMR (101 MHz, CDCl3) δ 140.31, 139.56, 137.57, 134.38, 134.12, 133.06, 130.84, 126.16, 125.72, 125.37, 125.11, 124.80, 124.69, 124.18, 122.99, 122.89, 120.56, 119.26, 118.84, 108.51, 74.17, 37.62, 26.95, 19.66, 13.84.</p><!><p>Using carbazole (300 mg, 1.80 mmol), and 4-methyl-1-naphthoyl chloride (0.25 mL, 1.28 mmol) as starting compounds, the title compound was prepared by the procedures described for the preparation of compound 5 to yield 96 mg (16%) of 28 as a yellow oil. (400 MHz, CDCl3) δ ppm 8.14 (d, J = 8.5 Hz, 1H), 8.05 – 7.93 (m, 3H), 7.60 (ddd, J = 8.4, 6.8, 1.3 Hz, 1H), 7.54 (d, J = 7.2 Hz, 1H), 7.48 (ddd, J = 8.5, 6.9, 1.4 Hz, 1H), 7.42 (t, J = 8.3 Hz, 2H), 7.36 – 7.30 (m, 2H), 7.25 – 7.18 (m, 1H), 2.84 (s, 3H), 13C NMR (101 MHz, CDCl3) δ ppm 167.81, 140.09, 139.36, 133.72, 132.29, 131.84, 129.61, 128.79, 126.76, 126.41, 126.04, 125.42, 120.53, 120.14, 119.49, 115.17, 19.64. HRMS calcd for C24H18 NO (M+H)+: 336.1383, found: 336.1737.</p><!><p>Using 7-methoxycarbazole [15], (1.2 g, 4.7 mmol), bromoethane (0.697 mL, 9.4 mmol) and Cs2CO3 (3.46 mg, 10 mmol) in DMF (40 mL) as starting compounds, the title compound was prepared by the procedures described for the preparation of compound 5 to yield 1.03 g (77%) of 29 as a greenish solid. 1H NMR (400 MHz, CDCl3) δ 8.69 (dd, J = 1.6, 0.6 Hz, 1H), 8.08 (dd, J = 8.7, 1.7 Hz, 1H), 7.99 (d, J = 8.4 Hz, 1H), 7.31 (d, J = 8.6 Hz, 1H), 6.89 (dd, J = 8.5, 2.2 Hz, 1H), 6.85 (d, J = 2.1 Hz, 1H), 4.28 (q, J = 7.2 Hz, 2H), 3.96 (s, 3H), 3.93 (s, 3H), 1.41 (t, J = 7.2 Hz, 3H). 13C NMR (101 MHz, CDCl3) δ 168.02, 159.55, 142.81, 141.95, 130.20, 127.23, 126.14, 122.90, 121.91, 121.47, 120.78, 116.88, 108.06, 107.67, 93.43, 55.75, 51.91, 37.76, 13.68.</p><!><p>Using 29 (1.413 g, 4.99 mmol), as the starting compound, the title compound was prepared by the procedures described for the preparation of compound 6 to yield 990 mg (52%) of 29 as an orange solid. 1H NMR (400 MHz, DMSO-d6) δ ppm 12.60 (br s, 1H), 8.68 (s, 1H), 8.12 (d, J=8.53 Hz, 1H), 7.99 (dd, J=8.53, 1.51 Hz, 1H), 7.58 (d, J=8.53 Hz, 1H), 7.18 (d, J=1.76 Hz, 1H), 6.86 (dd, J=8.53, 1.76 Hz, 1H), 4.42 (q, J=6.86 Hz, 2H), 3.89 (s, 3H), 1.24 – 1.37 (m, 4H) 13C NMR (101 MHz, DMSO-d6) δ ppm 168.11, 159.27, 142.18, 141.72, 125.76, 122.22, 121.53, 121.47, 121.16, 115.80, 108.56, 108.40, 93.58, 55.52, 37.13, 13.56.</p><!><p>Using 7-methoxy-9-pentyl-9H-carbazole-3-carboxylic acid (1.57 g, 8.05 mmol), and 1-methylnaphthalene (1.1 mL, 9.26 mmol) as starting compounds, the title compound was prepared by the procedures described for the preparation of compound 10 to yield 133 mg (42%) of 31 as an orange oil. 1H NMR (400 MHz, Chloroform-d) δ 8.53 (d, J = 1.7 Hz, 1H), 8.16 – 8.04 (m, 2H), 7.98 (dd, J = 8.6, 1.7 Hz, 1H), 7.91 (d, J = 8.4 Hz, 1H), 7.62 – 7.55 (m, 1H), 7.54 (d, J = 7.1 Hz, 1H), 7.48 (ddd, J = 8.2, 6.7, 1.3 Hz, 1H), 7.42 – 7.38 (m, 1H), 7.35 (d, J = 8.5 Hz, 1H), 6.86 (dd, J = 8.5, 2.2 Hz, 2H), 4.34 (q, J = 7.3 Hz, 2H), 3.94 (s, 3H), 2.80 (d, J = 1.0 Hz, 3H), 1.45 (t, J = 7.2 Hz, 3H). 13C NMR (101 MHz, DMSO-d6) δ ppm 168.10, 159.29, 142.19, 141.74, 125.77, 122.23, 121.55, 121.48, 121.16, 115.81, 108.58, 108.43, 93.60, 55.54, 37.15, 30.67, 13.58. HRMS calcd for C27H24NO2 (M + H)+: 394.1807, found 394.0875.</p><!><p>12 M HCl (1.2 mL) was added to a solution of phenylhydrazine (3.64 mL, 36.99 mmol) and ethyl-4-oxocyclohexanecarboxylate (0.98 mL, 5.73 mmol) in EtOH (10 mL). The solution was microwaved at 140 °C for 3 h. The reaction mixture was cooled on an ice-water bath, then poured onto a mixture of ice and a 4 M NaOH solution (50 mL) and extracted with ethyl acetate (150 mL). The organic phase was washed with saturated aqueous NaHCO3 and brine, dried (MgSO4), filtered and evaporated in vacuo. The residue was purified by column chromatography on silica gel eluting with EtOAc/heptanes in different proportions to yield 731 mg (52%) of 32 as an orange solid: 1H NMR (500 MHz, CDCl3) δ ppm 7.77 (br. s., 1H), 7.28 (t, J=7.80 Hz, 1H), 7.14 (t, J=7.80 Hz, 1H), 7.02 (t, J=8.08 Hz, 1H), 6.98 (d, J=8.08 Hz, 1H), 4.15 – 4.29 (m, 2H), 3.90 (s, 2H), 3.39 (dd, J=16.01, 5.15 Hz, 2H), 2.92 (m, 1H), 2.22 – 2.31 (m, 1H), 1.32 (t, J=7.10 Hz, 3H) 13C NMR (126 MHz, CDCl3) δ ppm 175.89, 137.20, 131.02, 127.33, 121.85, 117.32, 108.25, 103.93, 60.39, 45.13, 40.67, 25.86, 14.23.</p><!><p>Under Ar, a solution of 320020(1.5 g, 6.15 mmol), bromoethane (0.20 mL, 26.8 mmol), and NaH (131 mg, 44.8 mmol) in DMF (30 mL) was stirred for 30 min at r.t. The reaction mixture was diluted with DCM (50 mL) and filtered through Celite©. The organic solvents were evaporated in vacuo. The concentrated residue was extracted with tBuOMe/EtOAc (5:1, v/v, 150 mL). The organic layer was washed with NaHCO3 (100 mL), dried over MgSO4, filtered, and concentrated in vacuo. The resultant oil was purified by column chromatography on silica gel using heptanes/EtOAc (4:1, v/v) to afford the title compound as an orange oil, 1455 mg (87%) 1H NMR (500 MHz, CDCl3) δ ppm 7.77 (br. s., 1H), 7.28 (t, J=7.80 Hz, 1H), 7.14 (t, J=7.80 Hz, 1H), 7.02 (t, J=8.08 Hz, 1H), 6.98 (d, J=8.08 Hz, 1H), 4.15 – 4.29 (m, 2H), 3.90 (s, 2H), 3.39 (dd, J=16.01, 5.15 Hz, 2H), 2.92 (m, 1H), 2.22 – 2.31 (m, 1H), 1.32 (t, J=7.10 Hz, 3H) 13C NMR (126 MHz, CDCl3) δ ppm 175.74, 136.24, 131.72, 127.12, 121.58, 117.21, 108.32, 103.83, 60.27, 45.42, 40.55, 37.22, 25.79, 15.25, 14.12.</p><!><p>A solution of 33 (1400 mg, 5.15 mmol) and KOH (726 mg, 7.73 mmol) in a mixture of H2O (2 mL) and EtOH (9 mL) was refluxed for 48 h. The reaction mixture was cooled and diluted with EtOAc (50 mL), and the organic phase was washed with water and brine, dried (MgSO4), filtered, and evaporated in vacuo. The resultant oil was purified by column chromatography on silica gel using heptanes/EtOAc in different proportions to afford the title compound as a light yellow oil, 988 mg (96%). δ ppm 7.50–7.56 (m, 1H), 7.09 – 7.15 (m, 3H), 4.09 (q, J=7.06 Hz, 2H), 3.66 (br. s.,2 H), 3.26 – 3.34 (m, 2H), 2.87 – 2.96 (m, 2H), 1.16 (t, J=7.06 Hz, 3H) 13C NMR (101 MHz, DMSO-d6) δ ppm 135.11, 132.14, 129.12, 122.15, 119.24, 118.56, 108.33, 108.10, 44.99, 41.59, 37.18, 25.77, 15.15.</p><!><p>Using 34 (250 mg, 1.25 mmol), and 4-methyl-1-naphthoic acid (348 mg, 1.87 mmol) as starting compounds, the title compound was prepared by the procedures described for the preparation of compound 7 to yield 193 mg (42%) of 35 as an orange oil. 1H NMR (400 MHz, DMSO-d6) δ ppm 8.71 (d, J=8.20 Hz, 1H), 8.07 (d, J=8.42 Hz, 1H), 7.72 – 7.81 (m, 2H), 7.52 – 7.66 (m, 2H), 7.35 – 7.50 (m, 2H), 7.29 – 7.35 (m, 2H), 6.98 – 7.02 (m, 1H), 6.91 – 6.96 (m, 1H), 4.36 (br s, 2H), 4.02 (q, J=7.01 Hz, 2H), 3.41 – 3.53 (m, 2H), 3.04 – 3.14 (m, 2H), 2.64 (s, 3H), 1.18 (t, J=7.01 Hz, 3H) 13C NMR (101 MHz, DMSO-d6) δ ppm 168.86, 135.21, 135.14, 133.16, 132.09, 132.04, 129.15, 126.62, 126.32, 125.96, 125.91, 125.12, 124.65, 124.58, 123.15, 119.31, 118.70, 108.30, 108.00, 44.20, 42.78, 37.05, 25.47, 19.07, 15.15. HRMS calcd for C25H25N2O (M+H)+: 369.1961, found: 369.1936.</p><!><p>Cells were cultured in DMEM supplemented with 10% FBS, 100 U/mL penicillin, and 100 μg/mL streptomycin at 37 °C in a 5% CO2 humidified atmosphere. T98G and HepG2 cells (ATCC, Manassas, VA) were authenticated by ATCC when purchased using human short tandem repeat analysis and maintained in culture for less than 6 months. Tumors were obtained from surgeries performed at the Swedish Medical Center (Seattle, WA) according to Institutional Review Board guidelines. Patient samples used in this study were diagnosed as WHO grade IV glioblastoma multiforme. Patient-derived GBM cells were established from the freshly resected tumor tissues and maintained in NeuroCult® NSA medium (Stem Cell Technologies) with B-27 serum-free supplement, 20 ng/mL epidermal growth factor and 20 ng/mL fibroblast growth factor 2 as previously described [18, 38].</p><!><p>WST-1 (Roche, Pleasanton, CA) was used to evaluate cell viability as described [14] 72 h following drug treatment according to the manufacturer's protocol. Maximal killing activity was identified as the maximal % reduction in cell viability measured at drug concentrations tested. EC50 values were calculated and reported only when a curve was reliably extrapolated by Prism software. The solvent for all modified carbazoles was DMSO (0.1% final). This solvent (negative reference, vehicle control) had no effect on cell viability.</p><!><p>[3H]Colchicine binding to purified tubulin and tubulin assembly (as assessed by turbidity development in purified bovine tubulin solutions) were both measured as previously described[29].</p><!><p>Bovine brain tubulin was purified, polymerized into MTs and sheared as previously described[39]. Sheared MTs and drugs were incubated at 37 °C for 15 min and centrifuged for 10 min at 42,000 rpm at 37 °C prior to separation into supernatants containing free tubulin and pellets containing polymerized MTs as described[29]. Supernatants and pellets were run on a 4–12% polyacrylamide gel and stained with Coomassie G-250 for quantification. Peak intensities were quantified using ImageJ.</p>

PubMed Author Manuscript

Crystal Structures of the Glycopeptide Sulfotransferase Teg12 Complexed with the Teicoplanin Aglycone

The TEG gene cluster, a glycopeptide biosynthetic gene cluster that is predicted to encode the biosynthesis of a polysulfated glycopeptide congener, was recently cloned from DNA extracted directly from desert soil. This predicted glycopeptide gene cluster contains three closely related sulfotransferases (Teg12, 13, and 14) that sulfate teicoplanin-like glycopeptides at three unique sites. Here we report a series of structures including: an apo structure of Teg12, Teg12 bound to the desulfated co-substrate 3\'-phosphoadenosine 5\'-phosphate and Teg12 bound to the teicoplanin aglycone. Teg12 appears to undergo a series of significant conformational rearrangements during glycopeptide recruitment, binding and catalysis. Loop regions that exhibit the most conformational flexibility show the least sequence conservation between TEG sulfotransferases. Site directed mutagenesis guided by our structural studies confirmed the importance of key catalytic residues as well as the importance of residues found throughout the conformationally flexible loop regions.

crystal_structures_of_the_glycopeptide_sulfotransferase_teg12_complexed_with_the_teicoplanin_aglycon

<!>Teg12 Expression and Purification<!>Teg12 Crystallization<!>Data Collection and Structure Solving<!>Site-Directed Mutagenesis<!>Mutant Teg12 Expression and Purification<!>Teg12 Activity Assays<!>Teg12 apo structure<!>Teg12 co-crystal structures<!>Teg12 Ternary Structure<!>Teicoplanin aglycone bound in the ternary structure<!>Teg12 binary structure and the Teg12 active site<!>Discussion

<p>During the 20th century, widespread use of antibiotics significantly reduced the threat of many once lethal infectious diseases. However, the success of these wonder drugs may soon become their Achilles' heel. Bacterial pathogens that have developed resistance to most widely used antibiotics are now regularly seen in clinical settings. Vancomycin and teicoplanin are glycopeptide antibiotics used in the treatment of many gram-positive bacterial infections, including methicillin-resistant Staphylococcus aureus (MRSA). With the appearance of vancomycin resistant Enterococci in the late 1980s and resistant Staphylococci in the early 1990s, these traditional antibiotics of last resort are in danger of becoming clinically compromised (1–3). As with many bacterial natural products, the discovery of additional glycopeptide congeners that might combat the growing problem of antibiotic resistance has slowed as it has become increasingly difficult to identify new biodiversity from which novel molecules might be characterized.</p><p>The vast majority of bacteria present in the environment remain recalcitrant to culturing (4). This uncultured majority no doubt contains previously inaccessible glycopeptide biosynthetic gene clusters, many of which could encode the biosynthesis of novel glycopeptide congeners. Although metabolites produced by bacteria that are difficult to culture in the laboratory cannot be characterized using standard microbiological methods, it is possible to extract DNA directly from environmental samples and then analyze this DNA for sequences that might encode the biosynthesis of new natural products. In a recent analysis of DNA extracted directly from desert soil, we uncovered a new glycopeptide biosynthetic gene cluster (the TEG gene cluster) that is predicted to encode the biosynthesis of the first polysulfated glycopeptide congeners (5). The TEG gene cluster contains three closely related 3'-phosphoadenosine 5'-phosphosulfate (PAPS) dependent sulfotransferases (Teg12, 13 and 14). In vitro, these three sulfotransferase finishing enzymes sulfate teicoplanin-like glycopeptides at three unique sites and, in combination, can be used to produce seven different glycopeptide sulfation patterns (Figure 1).</p><p>Vancomycin- and teicoplanin-like glycopeptides are structurally defined by the presence of an oxidatively cross-linked heptapeptide core (6, 7). The peptide core is initially produced as a linear polymer by nonribosmal peptide synthetases, which is then oxidatively cross-linked into either three or four large macrocycles by conserved cytochrome P450 oxidases. During their biosynthesis, each member of this family of antibiotics is functionalized by a unique collection of finishing enzymes that includes glycosyl transferases, halogenases, acyl transferases and sulfotransferases. Although glycopeptides show some variation in the sequence of the heptapeptide core, the bulk of the structural diversity seen within this class of antibiotics arises from the functionality added by finishing enzymes. Over 150 different glycosylated, halogenated, and alkylated glycopeptide congeners have been characterized from cultured bacteria (8). Only three naturally occurring sulfated congeners have been identified to date from studying this same pool of bacteria (8, 9). While anionic glycopeptides have rarely been reported as natural products, telavancin, a semisynthetic phosphono congener that proved to be a very effective antibiotic in clinical trials, was recently approved for use in humans by the FDA (10, 11). Increasing the hydrophilicity with the addition of the negatively charged phosphono group was found to significantly improve the adsorption, distribution, metabolism, and excretion profile of this class of antibiotics (12). The enzymatic synthesis of anionic glycopeptides may provide a facile means to access additional anionic congeners with improved pharmacological properties.</p><p>Here we report a series of Teg12 sulfotransferase structures, including an apo structure, a binary structure complexed with the teicoplanin aglycone substrate, as well as a ternary structure containing both PAP and the teicoplanin aglycone. In the binary and ternary structures, the glycopeptide substrate is observed bound at two different locations. Teg12 appears to undergo a series of conformational rearrangements during glycopeptide recruitment, binding and catalysis. These studies provide insight into the sulfotransferase mechanism, as well as insights into how this rarely seen class of finishing enzymes might be engineered to produce novel anionic glycopeptides.</p><!><p>Teg12 was cloned and expressed as previously described (5). Briefly, teg12 was amplified (30 cycles of 95 °C for 30 s, 60 °C for 30 s, and 72 °C for 90 s; FailSafe system from Epicentre) from eDNA cosmid clone D30 using the following primers: Teg12FWD(BclI):GCGCTGATCAATGAACGGAATTCGATGG, Teg12REV(HindIII):GCGCAAGCTTTCCTTAACCGGCATACCCGTA. Restriction enzyme sites used for cloning are shown in bold. The resulting product was doubly digested with BclI and HindIII and subsequently ligated into pET28a, which had been BamHI/HindIII doubly digested. The resulting construct was then transformed into E. coli BL21(DE3) for protein expression. Expression cultures were grown to OD600=0.6, followed by IPTG induction, and overnight growth at 20 °C. The culture was pelleted by centrifugation (3,200 × g for 30 min), the supernatant was discarded, and the cell pellet was resuspended in 40 mL lysis buffer [50 mM HEPES, pH 7.5, 0.5 M NaCl, 5% (vol/vol) glycerol, 20 mM imidazole, pH 8, 10 mM β-mercaptoethanol and 0.5% (vol/vol) Triton X-100]. The resuspended cell pellet was lysed by sonication, and the insoluble portion was removed by centrifugation (15,000 × g for 30 min). The cleared cell lysate was incubated with 1 mL Ni-NTA resin for 15 min. The slurry was loaded onto a column, allowed to empty by gravity flow, washed with 40 mL lysis buffer, and finally washed with 40 mL wash buffer [50 mM HEPES, pH 7.5, 0.5 M NaCl, 5% (vol/vol) glycerol, 20 mM imidazole, pH 8.0, and 10 mM β-mercaptoethanol]. The protein was eluted by the addition of 15 mL of elution buffer [50 mM HEPES, pH 7.5, 0.5 M NaCl, 5% (vol/vol) glycerol, 125 mM imidazole, pH 8, and 10 mM β-mercaptoethanol]. No attempt was made to remove the vector derived 6X-histidine tag, resulting in a Teg12 protein plus 34 additional residues N-terminal to the start methionine. Protein was concentrated using Vivascience Vivaspin 30,000 MWCO ultrafiltration concentrators, and was buffer exchanged 3 times into protein buffer [200 mM NaCl, 20 mM HEPES, pH 7.5, 5% glycerol, and 1 mM DTT].</p><!><p>Concentrated protein was centrifuged at 14,000 rpm (4 °C, 30 min) in a microcentrifuge to remove any insoluble material prior to crystallization. All crystals were grown using the hanging drop vapor diffusion method. Initial Teg12-apo crystals were obtained by mixing 1 µl of protein (7.5 mg/ml in protein buffer) with 1 µl of reservoir solution (1.0 M sodium citrate, 0.1 M sodium cacodylate, pH 6.5, JCSG core III-48, Qiagen) over a 500 µl reservoir. Blade-like crystals grew overnight at 22 °C and reached a maximal size of 400 µm × 50 µm × 20 µm in approximately one week. To improve crystal thickness, Teg12-apo crystals were optimized by microseeding, in addition to mixing 1 µl of protein at 7.5 mg/ml with 0.5 µl of reservoir and 0.5 µl of Silver Bullets™ reagent 29 (Hampton Research). A component of the Silver Bullets™ screen, aspartame, was modeled into the PAPS binding site of one of the monomers of the Teg12 dimer. Crystals were soaked in a 20 µl drop containing reservoir solution plus 10% ethylene glycol. The drop was allowed to dehydrate by exposure to open air at room temperature for approximately 5 hours before flash cooling the crystals in liquid ethane.</p><p>Teg12-ternary crystals were co-crystallized at 4 °C in the presence of 2 mM PAP and 1 mM teicoplanin aglycone. Teg12 was first concentrated to 20 mg/ml in protein buffer. The protein was then diluted 1:1 with 50 mM CHES, pH 9.1, 2 mM teicoplanin aglycone, 4 mM PAP, achieving a final concentration of 10 mg/ml Teg12 in 0.5× protein buffer, 25 mM CHES, pH 9.1, 1 mM teicoplanin aglycone, 2 mM PAP. 1 µl of protein was mixed 1:1 with reservoir solution (0.2 M ammonium acetate and 20% w/v PEG 3350, JCSG, core I-25, Qiagen). Crystals appeared in 2–3 days and grew to a maximal size of 100 µm × 50 µm × 50 µm in approximately 1 week. These crystals were of an irregular chunk-like morphology and had cracks throughout. Crystals were cryo-protected by quickly dunking in reservoir solution plus 15% ethylene glycol and were flash cooled in liquid nitrogen.</p><p>Teg12-binary crystals were co-crystallized at 4 °C in the presence of 1 mM teicoplanin aglycone. Similar to Teg12-ternary crystallization, the protein was first concentrated to 20 mg/ml in protein buffer, then diluted to 10 mg/ml with 50 mM CHES, pH 9.1, 2 mM aglycone (final 0.5× protein buffer, 25 mM CHES, 1 mM teicoplanin aglycone). 1 µl of protein solution was mixed with 1 µl of reservoir solution (2.0 M sodium formate, 0.1 M sodium acetate, pH 4.6, JCSG core III-85, Qiagen). Cubic crystals grew between 2 and 3 weeks, and were approximately 50 µm × 50 µm × 50 µm. Crystals were soaked in 6.0 M sodium formate, 0.1 M sodium acetate, pH 4.6, 1 mM teicoplanin aglycone overnight, prior to flash cooling in liquid nitrogen.</p><!><p>All data sets were reduced and scaled using the HKL2000 package (18). Data for Teg12-apo crystals were collected at the NSLS, beamline X29A. All but one of the crystals screened diffracted poorly to approximately 4 Å resolution. The crystal from which the 2.91 Å dataset was collected rotated briefly out of the cryostream, and thereby had gone through a room temperature annealing cycle of several seconds. Diffraction from this crystal was dramatically improved compared with other crystals taken from the same drop. Data for Teg12-apo was reduced and scaled in space group C2221. Phase information was obtained by molecular replacement using the program Phaser and StaL (GenBank accession number AAM80529, PDB code 2OV8), devoid of all flexible loops, as the search model (19). The initial molecular replacement model was refined against the Teg12-apo dataset using rigid body refinement in Refmac (20). Additional features of the map were enhanced through density modification, and 2-fold ncs averaging in CNS (21, 22). The model was rebuilt manually using the program Coot (23). Full restrained refinement was carried out using the translation/libration/screw model in Refmac, with the addition of hydrogen atoms, converging to a final Rwork and Rfree of 21.96 and 27.19, respectively (24). NCS restraints were not used during refinement. The final model comprises residues 1–129, 136–203, and 251–285 for monomer A, and 1–27, 42–128, 137–210, 247–285 for monomer B. The Teg12-apo model was used as a molecular replacement model for all subsequent structures.</p><p>Teg12-ternary and Teg12-binary data sets were collected at the APS, microfocus beamline 24-IDE. Teg12-ternary data was reduced and scaled in space group P212121. Teg12-binary was scaled in space group I212121. The crystal structure of glycopeptide aglycone A-40926 was used as a starting point to generate a restraint definition file for teicoplanin aglycone using the program Phenix Elbow (25). Geometry optimization was achieved using the semi-empirical quantum mechanical AM1 method. Teg12-binary and Teg12-ternary models were refined using the translation/libration/screw model in Phenix Refine to a final Rwork and Rfree of 17.30 and 22.61, and 17.12 and 22.47, respectively (26). The final Teg12-binary model comprises residues 1–129, 135–216, and 247–285. The final Teg12-ternary model comprises residues 1–215, 220–224, and 231–285 for monomer A, and residues 1–129, 136–224, and 240–285. All structures were validated using the Molprobity server from the Richardson laboratory at Duke University (27).</p><!><p>Teg12 point mutants were generated using the "megaprimer" method, with slight modifications (28). Oligonucleotide primers were designed for each mutant (Table 2), and a megaprimer was generated by PCR amplification from the Teg12/pET28a construct using the Pfx Accuprime System (Invitrogen), the relevant mutant oligonucleotide primer, and either the T7 promoter (for mutations at residues 9–108) or the T7 terminator (for mutations at residues 167–251) as the second oligonucleotide primer, (30 rounds of amplification: 95 °C for 30 s, 55°C for 30 s, 68 °C for 30 s). The full length mutant Teg12 gene was amplified from the Teg12/pET28a construct, using the megaprimer, which then contained the bases that code for the specific mutant residue, and either the T7 terminator (for mutations at residues 9–108) or the T7 promoter (for mutations at residues 167–251) as the second oligonucleotide primer (30 rounds of amplification: 95 °C for 30 s, 55 °C for 30 s, 68 °C for 80 s). Full-length mutant amplicons were then sequentially digested with BamHI and HindIII, and subsequently ligated into BamHI/HindIII doubly digested pET28a. Ligated constructs were transformed into E. coli EC100 (Epicentre), and sequenced to identify successfully mutated constructs. Mutant constructs containing the desired point mutation were then transformed into E. coli BL21 (DE3) for protein expression.</p><!><p>Mutant proteins were expressed and purified in a manner similar to the native Teg12, except on a reduced scale. 100 mL overnight expression cultures were pelleted and resuspended in 4 mL lysis buffer. After sonication to lyse the cells, the crude lysates were centrifuged to remove insoluble material (10 min at 15,000 × g). The cleared lysates were incubated with 100 µl Ni-NTA resin for 15 min. The slurry was then loaded onto a column, allowed to empty by gravity flow, washed with 4 mL lysis buffer, followed by a second wash with 4 mL wash buffer. The protein was eluted by the addition of 1.5 mL elution buffer. All Teg12 mutants used in activity assays appeared to be homogeneous by polyacrylamide gel electrophoresis</p><!><p>All soluble Teg12 mutants were assayed for activity using the teicoplanin aglycone as a substrate. 50 µL reactions were run in duplicate, as follows: 15 mM HEPES, pH 7.5, 1 mM 3'-phosphoadenosine-5'-phosphosulfate (PAPS), 0.1 mM DTT, 1.2 mM teicoplanin aglycone (in DMSO), and 500 ng purified protein in elution buffer. Reactions were carried out at 30 °C for each of the four time points (10, 15, 20, 25 min), followed by heat inactivation at 99 °C for 10 min, and a further 10 min in an ice water bath. Vmax and Km values were determined under the same reaction conditions using the teicoplanin aglycone as substrate (5 µM to 100 µM). 25 µL of each reaction was run on a Waters analytical HPLC system (C18 (4.6 × 150 mm)). A linear gradient (1.5 ml/min) was run from an initial condition of 95:5 20 mM ammonium acetate:acetonitrile to 70:30 20 mM ammonium acetate:acetonitrile over twenty minutes. The area under the UV peak (Diode Array, 240 nm–400 nm) was determined for both the monosulfated product and the teicoplanin aglycone substrate at each time point. The percent substrate conversion for duplicate time points was averaged. The slope of the graph derived from the four time points for Teg12 and each mutant was then determined. Relative activity of each mutant is reported as a percent of the slope for wild-type Teg12.</p><!><p>N-terminally His tagged Teg12 was affinity purified using nickel NTA resin and crystallized without the need for further purification. The apo structure was solved to 2.91 Å resolution by molecular replacement using StaL (PDB code 2OV8) as a search model (13). StaL is a related sulfotransferase involved in the biosynthesis of the monosulfated glycopeptide congener A47934 (9, 14). StaL has 52.9% sequence identity with Teg12. The Teg12 structure superimposes well on StaL (rmsd of 1.053 Å over 210 Cα atoms). Teg12 crystallizes as a dimer and the overall structure resembles that of StaL. Running Teg12 through the DALI server shows that its closest eukaryotic sulfotransferase relative is the human cytosolic sulfotransferase 1C1 (SULT1C1, pdb code 1zhe), with an rmsd of 2.8 Å over all Cα atoms (15). As with other sulfotransferases in this family, Teg12 consists of a single globular α/β domain composed of a parallel beta sheet core surrounded by alpha helices (Figure 2A). The beta sheet core of both StaL and Teg12 contains four strands. The dimer interface, which resembles that seen in StaL, consists of a symmetrical interaction between a short helix-loop motif from one monomer and the same helix-loop motif from the other monomer. In addition to the hydrophobic contacts that exist between the two helices there is a hydrogen bond between the carbonyl oxygen of Gly51 at the end of the helix from one monomer and the backbone nitrogen of Val74 of the other monomer. The active site cavity of each monomer faces the dimer interface.</p><p>Minor differences exist between the two monomers of the apo structure. The pET28a vector derived N-terminal tail, including the 6X-His, thrombin cleavage site and T7 tags, from monomer A is ordered and involved in crystal packing interactions. The arrangement of these residues within the crystal causes distortions in the region Gly28 through Ser41 when compared with StaL, and subsequent Teg12 complex structures. In monomer B, residues Gly28 through Ser41 could not be modeled in the electron density map, further underscoring the conformational flexibility of this region. From our co-crystal structures and that of StaL, it is known that residues Glu216 through Asp250 constitute a largely disordered loop. In the Teg12 apo structure, this disorder extends even further N-terminally into what is observed as a helix in the Teg12 binary and ternary structures and StaL. Electron density falls off in monomer A after Ser203, but extends through Glu210 in monomer B. Ser203 corresponds to Cys196 in StaL. In StaL Cys196 makes a disulfide bond with Cys20 (residue 20 is also a cysteine in Teg12). The absence of a disulfide in Teg12 could impart a greater flexibility in this region and explain the additional observed disorder. In subsequent Teg12 co-crystal structures, portions of this helix-loop region (Thr204-Asp250) play key roles in binding the glycopeptide substrate. Because of its role in binding the glycopeptide, we have termed this entire conformationally flexible region the GHL (glycopeptide-helix-loop).</p><p>ClustalW alignment of the three TEG sulfotransferases shows that most of the sequence variability seen within this family of sulfotransferases is concentrated on three short loops (variable regions V1, V2, V3) that surround the predicted active site (Figure 3). The large disordered loop from the GHL corresponds to the longest of these three variable regions, V3. V2 encompasses residues Gly127 through Gly137. Much of V2 (Ala130 to Gly135 in monomer A and Asn129 through Gly136 in monomer B) is disordered in the Teg12 apo structure. V1, the shortest of the three variable sequences, corresponds to an ordered region (though disordered in monomer B) that spans Ile37 through Thr43.</p><p>Electron density was observed in the PAPS binding site of monomer A (this binding site was empty in monomer B). However, neither PAP nor PAPS could be modeled into the density. We were able to model the dipeptide aspartame (N-(l-a-aspartyl)-l-phenylalanine,1-methyl ester) into this density (Figure 4). Aspartame is a component of the Silver Bullets™screen (Hampton) used during crystallization. The orientation of aspartame in the structure mimics that of PAP. Its phenylalanine ring stacks with Trp17, as does the adenine ring of PAP, and its N-terminal nitrogen hydrogen bonds with the hydroxyl of Ser98. Eukaryotic sulfotransferases are known to tightly bind ribose, adenine and other nucleotides in the PAPS binding pocket (16).</p><!><p>The TEG gene cluster is predicted to encode the biosynthesis of a heptapeptide (hydroxyphenylglycine (Hpg)-betahydroxytyrosine (Bht)-dihydroxyphenylglycine (Dpg)-Hpg-Hpg-Bht-Dpg) that is oxidatively cross-linked into the four macrocycles seen in teicoplanin-like glycopeptides (Figure 1). This TEG derived, oxidatively cross-linked heptapeptide skeleton only differs from the teicoplanin aglycone by the substitution of Bht for Tyr at the 2nd position in the peptide core. All three TEG sulfotranferases can use the teicoplanin aglycone as a substrate. We therefore used this molecule in Teg12 co-crystallization experiments. The Vmax and Km values for the teicoplanin aglycone were determined to be 215.3 ± 25.2 nmol/min mg and 59.6 ± 1.1 uM, respectively. Attempts to obtain co-crystals by soaking Teg12 apo crystals with either the teicoplanin aglycone or the co-substrate PAP were unsuccessful. The addition of PAP to drops containing apo crystals caused the crystals to rapidly dissolve into the mother liquor. Therefore, we co-crystallized Teg12 and the teicoplanin aglycone with and without PAP. PAP is the desulfated byproduct of PAPS from the sulfonation reaction. Our co-crystallization experiments led to a Teg12 binary structure bound to teicoplanin, and a ternary structure bound to PAP and teicoplanin. These structures represent the first examples of sulfotransferase structures containing glycopeptide substrates.</p><!><p>We obtained a 2.05 Å structure of a Teg12-PAP-teicoplanin aglycone ternary complex by co-crystallization (Figure 2B and 2D). Several regions of the Teg12 sequence not seen in the other Teg12 structures could be modeled into the electron density. Loop region Gly127 to Val137, which is primarily disordered in the other Teg12 structures, is fully ordered in one of the monomers of the ternary complex. This short loop contains the V2 region, one of the three regions that are highly variable within the three TEG sulfotransferase sequences. The V2 loop in Teg12 is extended by four amino acids compared with StaL, and by two amino acids when compared with Teg13 and Teg14. A much larger portion of the GHL helix and loop could be modeled in the ternary structure. The GHL helix extends through Ser15 in one monomer. In the opposite monomer the helix extends to Gln218 and transitions uninterrupted to the GHL loop.</p><p>Both monomers of the ternary structure contain a single molecule of PAP bound in the predicted active site location. Three highly conserved structural motifs in sulfotransferases are known to play a role in binding PAPS, all of which are seen in Teg12. Lys12 from the 5'-phosphate binding loop (PSB) (Pro11-Thr16) hydrogen bonds with the 5'-phosphate of PAP, Ser98 from the 3'-phosphate binding loop (PB) (Val89-Ser99) hydrogen bonds with the 3'-phosphate of PAP, and Trp17 parallel stacks with the adenine base. A number of additional residues from the PSB loop region, the PB loop region and the conformationally flexible GHL region are within hydrogen bonding distance of PAP (Figure 4B). Thr16 is only ~2.6 Å from the 5' phosphate, to which it hydrogen bonds. This residue corresponds to a histidine in StaL, although threonine is the more common side chain at this position in eukaryotic sulfotransferases (13). A general hydrogen bond donor at position 16 is likely to be important for properly coordinating PAPS in the active site. Indeed, an alanine replacement mutant of Thr16 was inactive for the sulfation reaction. A K12A mutant was also inactive, while S98A was partially active (27% compared with wild-type). A W17A mutant did not express as soluble protein, indicating a role not only in binding PAP, but also in protein folding and stability.</p><p>PAPS dependent sulfotransferases are believed to use a histidine as a general base to activate the hydroxyl or amine used in the sulfate exchange reaction. Teg12 contains three histidine residues, His67, His226 and His269. Of these, only His67 is located near the co-substrate binding site. The 5'-phosphate of PAP is ~5 Å from the imidazole of His67, and when PAPS is modeled in its place, the sulfate is positioned directly adjacent to His67 (Figure 7). In this model, both Lys12 and Lys65 are close enough to coordinate with the sulfate. In place of the sulfate from PAPS, a water molecule occupies this space in the ternary structure. Ser9, which is within hydrogen bonding distance of the imidazole ring, was proposed to help activate the catalytic histidine in StaL. Attempts to mutate the predicted catalytic histidine in StaL resulted in an insoluble protein and therefore its role in catalysis had not been previously confirmed (13). Alanine exchange mutants were generated for His67, Ser9, and Lys65. H67A and S9A did not express as soluble proteins, while K65A was partially active (65%). Additional glutamate and glutamine mutants of His67 were generated to potentially provide better expression. H67E also expressed as an insoluble protein. However, H67Q, which was soluble, produced a completely inactive sulfotransferase.</p><!><p>One of the monomers in the ternary structure, in addition to PAP, contains two molecules of teicoplanin aglycone. Both molecules of the aglycone are located outside of the predicted active site pocket, and appear to be organized primarily by crystallographic forces (Figure 5A). One of the molecules of aglycone makes numerous contacts with the GHL loop (Figure 5B). Beginning with the N-terminal section of the GHL loop, residues Gly220 through Ile224 fold back towards the dimer interface. Pro223 stacks with Trp134 and helps to stabilize this region. The subsequent six amino acids (Arg 225-Arg230) are disordered, before density picks up again at Met231. However, these residues occupy a region of space where they could easily make additional contacts with the glycopeptides. Lys233 through Gln242 closely trace the outer surface of the N-terminal three amino acids of the aglycone. These ten amino acids immediately precede the five residues (Phe243-Gly247) that interact with PAP. Lys233 intercalates between the Dpg at position 3 and the Hpg at position 5, and it hydrogen bonds with hydroxyls from the Hpg at position 1 and the Dpg at position 7. The carboxylate from Asp271 of a symmetry mate makes several contacts with backbone nitrogens in the concave cleft of the glycopeptides. The C-terminal end of the teicoplanin aglycone extends out from the surface of the protein and interacts with a second glycopeptide molecule that is bound to the His-tag from an adjacent monomer. Despite the assumption that these aglycone-GHL interactions are mediated primarily by crystal packing, we generated a series of mutants in GHL loop residues that contact the glycopeptides. All mutants in this loop region, with the exception of G235A, had deleterious effects on the rate of substrate conversion, supporting the notion that the V3 loop interacts with the substrate and is involved in recruitment. The P241A mutation abolished Teg12's activity completely. While the observed glycopeptide-GHL interactions observed in the ternary structure are clearly stabilized by crystal packing interactions, our mutagenesis experiments make it so we cannot rule out the possibility that this represents a snapshot of the substrate recruitment path. A summary of all mutations made in Teg12 and their corresponding activities is given in Table 3.</p><!><p>We solved a Teg12-teicoplanin aglycone structure to 2.27 Å resolution (Figure 2C). A single monomer constitutes the asymmetric unit and the dimer is generated by crystallographic symmetry. The active site contains a single molecule of teicoplanin aglycone. Two sets of side chain hydrogen bonds play key roles in creating the glycopeptide binding cavity. The base of the binding pocket is largely defined by the outstretched side chain of Arg107 coordinating with the side chain of Gln251. The back of the cavity is largely dSefined by two outstretched arginine side chains (Arg214 and Arg248) that interact with each other via a mediating water molecule. Numerous hydrogen bonds exist between amino acid side chains from residues found in the GHL helix (Glu206, Arg207, Glu210 and Lys213) and the glycopeptide (Figure 6). Several of the residues that make contacts with PAP also interact with teicoplanin (Lys12, Arg90, Arg101, and Tyr167). Side chains from Tyr167 and Lys12 coordinate the C-terminal carboxylate, while a backbone carbonyl oxygen from Lys213 and water hydrogen bonded to Glu216 coordinate the N-terminal amine.</p><p>The GHL helix and the helix immediately N-terminal (residues Ile193 to Ser203) form a single contiguous helix that runs along the top surface of the aglycone (Figure 2C). In the ternary structure with an empty active site pocket, these two helices adopt a bent conformation that results from unwinding of the long helix at Thr204 (compare orange GHL helix, Figures 2B and 2C). This bent conformation is also observed in StaL. In the bent conformation, the helices are at a nearly perpendicular angle to one another. The long, largely disordered loop region of the GHL that interacts with PAP in the ternary structure is completely displaced by the bound glycopeptide in the binary structure. This disordered loop is now positioned at the back of Teg12, behind the bound glycopeptide. The front surface of the glycopeptide, where the sulfotransferase reaction is predicted to occur on residue 3 (see red arrow in Figure 2C), is largely exposed to solvent. With the exception of an additional small compensatory movement of a short loop (Gln31 through Ile37) into the void left by the straightening of the flexible helix from the GHL, the remainder of this Teg12 binary structure is essentially identical to that seen in the other Teg12 structures.</p><p>The C-terminus of the glycopeptide is buried in the PSB loop from the strand-loop-helix motif and the helix from the PB loop containing the strand-turn-helix motif runs along the under side of the glycopeptide. While most residues only contact the outer surface of the glycopeptide, one residue, Arg101, extends into the core of the molecule. The side chain of Arg101 intercalates between the Dpg at position 3 and the Hpg at position 5. The guanidinium from Arg101 hydrogen bonds with the backbone carbonyl from the Hpg at position 4 as well as with hydroxyls from the Dpg at position 3 and the Dpg at position 7. In this respect, Arg101 behaves in much the same way as Lys233 in the ternary structure. We generated a series of alanine replacement mutants for residues contacting the aglycone to determine which are important for substrate conversion. The results are summarized in Table 3. R101A showed a reduction of activity to 27% compared with wild-type. Interestingly, an E212A mutation also showed a reduction of activity to 26%. Glu212 is part of the GHL helix, but does not make contacts with the aglycone in the binary structure. Instead, this residue is found on the upper, solvent exposed face of the helix. Arg248, along with Arg214 forms the back of the active site cavity. An alanine placed at position 248 resulted in a two-fold increase in sulfation. Removing the bulky side chain at this position may impart more flexibility on the active site, allowing the glycopeptide to shift more easily towards His67 (see discussion).</p><p>Although the glycopeptide substrate interacts with different sets of GHL residues in the binary and ternary structures, the general glycopeptide binding motif seen in these structures is very similar. In both cases, a long positively charged amino acid side chain (Lys233 in the ternary, Arg101 in the binary) extends into the core of the glycopeptide, where it coordinates with the same set of side chain hydroxyls. All other glycopeptide-protein contacts observed in these structures involve the outer surface of the glycopeptide.</p><p>In addition to the glycopeptide bound in the active site cavity, there is a second glycopeptide molecule bound on the surface of the protein (not shown in Figure 2C). The location of the second teicoplanin aglycone in the crystal is likely not biologically relevant. Instead, it appears to mediate crystal contacts between symmetry partners. A stretch of electron density, large enough to contain four residues, is located in the D-ala-D-ala binding cleft of the second aglycone. Although it is presumed that this stretch of residues comes from the N-terminal vector derived His-tag of an adjacent symmetry mate, we were not able to determine its sequence, and hence it was modeled as polyalanine.</p><!><p>Teg12 sulfates the hydroxyl on the dihydroxyphenyl glycine at position 3 of the teicoplanin aglycone (Figure 2C, red arrow). In an alignment of the binary and ternary structures, this hydroxyl is 16.4 Å from where the sulfate of PAPS would be and there is significant overlap between the C-terminus of the glycopeptide and PAP (Figure 7). Before the sulfation reaction can proceed, there must be a substantial conformational rearrangement in the active site. In the binary structure, the interaction of the glycopeptide with Arg101 appears to preclude the substrate from shifting to accommodate PAPS or from its shifting into the vicinity of His67. Arg101 makes mutually exclusive contacts with the substrates present in the binary and ternary structures. In the binary structure it is inserted into the peptide core and hydrogen bonds with residues 1 and 7, while in the ternary structure it forms a key salt bridge with the 3-phosphate of PAP. If, when both substrates are present in the active site, Arg101 were to adopt the conformation seen in the apo structure, the glycopeptide would largely be free to move out of the PAPS binding pocket and towards His67. By pivoting towards His67, the glycopeptide would occupy the same general location that substrates bind in eukaryotic sulfotransferases (17), and it would also be positioned so that it could interact with all three variable loops.</p><p>The movement the glycopeptide must undergo to reach the PAPS sulfate traces a path similar to that which the GHL helix undergoes when it bends. This conformational rearrangement may serve not only as a mechanism to allow large glycopeptide substrates into the active site, but it could also play a role in positioning a glycopeptide for catalysis once it enters the active site. Alanine exchange mutagenesis of residues found on both the lower glycopeptide associated face (E206A, R207A, R214A) and the upper solvent exposed face of the helix (E212A) led to decreases in substrate conversion. Residues around the entire helix likely make key contacts with the glycopeptide at different points in time as a result of the rotation in the helix as it bends. In fact, all residues that contact the glycopeptide in the binary structure face out from the active site cavity in the ternary structure. One alanine exchange mutant, R248A, resulted in a two-fold increase in activity. R248, along with R214 of the GHL helix, and a mediating water molecule, forms a cage-like enclosure over the top of the glycopeptide (Figure 6). Introduction of an alanine side chain at this position would open up the back side of the active site to the solvent, and could provide substrates in the active site more room for positional rearrangement. The sequence of the GHL helix is conserved throughout the TEG sulfotransferases and a large portion of it is also conserved in StaL. The mechanics of the conformational change from the straight to bent structures may therefore be conserved throughout this family of finishing enzymes.</p><p>Sequence and structural similarities between eukaryotic and prokaryotic sulfotransferases indicate that all members have likely arisen from one common ancestor. In eukaryotic structures the loop that would correspond to the V3 loop of the GHL behaves in much the same way as it does in Teg12; it is disordered in apo structures and associated with PAP and the substrate in co-crystal structures (17). The open active site conformation that results from the straightening of the GHL helix was not observed in previous StaL structures. While a GHL-like helix-loop region is present in eukaryotic sulfotransferases, the helix from these structures does not appear to exhibit the same conformational rearrangements from a single straight helix to two helices with a bent conformation during substrate binding as it does in Teg12. At over 1000 daltons, glycopeptides are significantly larger than the substrates used by most eukaryotic sulfotransferases. The increased conformational plasticity that results from the ability of the GHL helix to easily flex may help TEG-like sulfotransferases accommodate larger substrates.</p><p>In addition to the conformational flexibility seen in the V3 loop, both the V1 and V2 loops adopt different conformations in this series of structures, suggesting they could easily reorganize to accommodate an incoming glycopeptide substrate. The sequence differences seen between Teg12, 13, and 14 suggest that these three short variable regions likely control the selection and orientation of the glycopeptide substrate bound in the active site. Systematically altering the residues found in the three TEG variable regions may provide a means to generate new glycopeptide finishing enzymes that sulfate a broader collection of glycopeptide congeners than is currently possible with the small number of sulfotransferases that have been identified naturally. The sulfated teicoplainin aglycone derivatives produced by the native TEG sulfotransferase retain potent in vitro antibacterial activity (5).</p><p>The cloning and characterization of biosynthetic gene clusters derived from uncultured bacteria provides a means to access both novel small molecules and new biosynthetic enzymes. Teg12 is one of the first enzymes discovered using culture independent methodologies to be characterized structurally, and the Teg12-teicoplanin aglycone co-crystal structures are the first examples of a substrate complexed with a member of this family of glycopeptide finishing enzymes. This series of Teg12 structures provides key insights into how sulfotransferases might be engineered to generate additional anionic glycopeptides that could be evaluated against clinically relevant drug resistant bacteria.</p>

PubMed Author Manuscript

The blind men and the filament: Understanding structures and functions of microbial nanowires

Extracellular electron transfer via filamentous protein appendages called \xe2\x80\x98microbial nanowires\xe2\x80\x99 has long been studied in Geobacter and other bacteria because of their crucial role in globally-important environmental processes and their applications for bioenergy, biofuels, and bioelectronics. Thousands of papers thought these nanowires as pili without direct evidence. Here, we summarize recent discoveries that could help resolve two decades of confounding observations. Using cryo-electron microscopy with multimodal functional imaging and a suite of electrical, biochemical, and physiological studies, we find that rather than pili, nanowires are composed of cytochromes OmcS and OmcZ that transport electrons via seamless stacking of hemes over micrometers. We discuss the physiological need for two different nanowires and their potential applications for sensing, synthesis, and energy production.

the_blind_men_and_the_filament:_understanding_structures_and_functions_of_microbial_nanowires

Introduction\xe2\x80\x94blind men and an elephant: critical need for multidisciplinary approaches<!>Critical need to correlate nanowire structure with function<!>Chemically tuning the nanowire structure and conductivity<!>Why two nanowires?\xe2\x80\x94physiologically distinct roles of OmcS and OmcZ<!>The regulatory role of pili in secretion of cytochrome nanowires<!><!>Resolving the controversy about nanowires by reconciling conflicting results<!><!>Resolving the controversy about nanowires by reconciling conflicting results<!><!>Resolving the controversy about nanowires by reconciling conflicting results<!><!>Resolving the controversy about nanowires by reconciling conflicting results<!><!>Resolving the controversy about nanowires by reconciling conflicting results<!><!>Resolving the controversy about nanowires by reconciling conflicting results<!>Outlook

<p>In the ancient Indian tale of blind men encountering an elephant, each man approaches the creature from a different direction [1]. One finds the trunk, another a leg, the third the tail, and so on, whereupon they naturally disagree with each other as to the true appearance of the elephant. Each person experiences something different depending on their angle of approach.</p><p>And so it is with microbial nanowires, genetic and electrical properties of nanowires seem to depend on the discipline from which you view them. In 2002, Childers et al. [2] pioneered the field by finding that Geobacter produce filaments during extracellular electron transfer. Since then, microbiologists thought that these 'microbial nanowires' are pili filaments made up of PilA protein [3] and that monomeric cytochromes bind to pili [4] (Fig. 1a). Electrochemists found a cytochrome signal in biofilms and suggested that pili are a scaffold for these monomeric cytochromes [5] (Fig. 1a). Biophysicists, like us, found that the filaments are intrinsically conductive, and not a scaffold [6,7], but did not know the mechanism. Lack of nanowire structure was everybody's elephant in the room. It had become clear that understanding the nanowire structure and function would require an unusually wide range of interdisciplinary knowledge.</p><p>Geobacter sulfurreducens produces current densities in microbial fuel cells that are among the highest known for pure cultures [8]. As discovered recently [9], such high current density is possible because of the unique ability of Geobacter to produce nanowires in biofilms with conductivity rivaling synthetic polymers [6,10,11]. Nanowires enable bacteria to transport electrons over hundreds of cell lengths [6,12]. However, the nanowire composition, structure, and conduction mechanism had remained unknown.</p><p>The recent discoveries of microbial nanowire structures have forced us to rethink the aforementioned decade-old beliefs [9,13]. Rather than pili, we found nanowires made up of cytochromes with seamlessly stacked hemes over the entire nanowire length, providing a continuous path for electrons [9,13] (Figs. 1 and 2).</p><p>We have combined cryo-electron microscopy (cryo-EM) with multimodal atomic force microscopy (AFM), to determine composition and structure of nanowires in biofilms to correlate with their conductivity and stiffness [9,13] (Figs. 2 and 3). Using mass spectrometry and near-atomic resolution, cryo-EM has allowed 'sequencing' the protein forming the nanowires without knowing its identity a priori [13,14]. Many other studies are now using this approach [15,16]. Multimodal imaging enabled the discovery of OmcZ nanowires by identifying their composition using infrared nanospectroscopy–based chemical imaging [17] and prediction of their structure with computational modeling [9] (Figs. 1–2). As proteins are widely considered as nonconductors [18], these studies will help understand the electron transport mechanism and why these two different nanowires are essential for various physiological roles such as iron reduction [19], current production [20], and direct interspecies electron transfer [21] (Fig. 4).</p><!><p>Previous studies presumed that nanowires are pili based on indirect genetic evidence that pilA mutant did not produce filaments and because of a lack of high-resolution structural methods [3,4]. Therefore, it is critical to correlate nanowire structure with functional studies. We have correlated atomic structure with AFM imaging to confirm that the same OmcS and OmcZ nanowires were studied for both conductivity measurements and atomic structure determination [9,13]. For example, AFM revealed an axial height periodicity with a 20-nm pitch for OmcS nanowires, consistent with the helical pitch determined by cryo-EM, whereas no such pitch was observed for noncytochrome filaments (Fig. 3a,c) [13]. Furthermore, at pH 7, the nanowire heights for OmcS and OmcZ are 3.6 and 2.5 nm, respectively, and the nanowires undergo very large conformational changes to beta sheets at pH 2 that reduce their diameter to 2.4 and 1.5 nm, respectively [9] (Fig. 3k). This distinct axial periodicity and the substantial thickness difference observed for OmcS and OmcZ nanowires versus other filaments were used to confirm that the same nanowires were studied for both structural as well as conductivity and stiffness studies. These studies revealed that in comparison with OmcS nanowires, noncytochrome filaments show 100-fold lower conductivity [13] (Fig. 3e–g), whereas OmcZ nanowires show 1000-fold higher conductivity [9] (Fig. 2d). Such correlative imaging studies will ensure that identical filaments are examined via multiple methods by first mapping their structural features and then linking them with different functional properties.</p><!><p>Our previous studies had shown that lowering the pH enhances nanowire conductivity [6] and alters their conformation [22]. However, the identity of nanowires and the underlying mechanism for low pH–enhanced conductivity were unknown. We have found that lowering the pH enhances conductivity of OmcS and OmcZ nanowires by 100-fold because of protein conformational changes to a β sheet–rich structure [9] (Fig. 3l and m). X-ray diffraction studies showed that this structural change improves the stacking of hemes in nanowires [9] (Fig. 1c–d). This enhanced π stacking between hemes can increase the effective conjugation length, yielding a longer mean free path for electrons that enhances conductivity [24].</p><!><p>The role of OmcS as nanowires was overlooked because ΔomcS biofilms were conductive and produced high current densities in microbial fuel cells when grown over prolonged growth conditions. Therefore, we evaluated the possibility of proteins other than OmcS capable of forming nanowires in biofilms [9]. Using our AFM-based multimodal imaging platform, we have found that growing G. sulfurreducens biofilms, under current-producing conditions that require an electric field, stimulates production of OmcZ nanowires that exhibit 1000-fold higher conductivity than OmcS nanowires (Fig. 2d) [9]. The electric field is maximum near the biofilm–electrode interface and decreases away from the electrode. Therefore, OmcZ expression will be maximum at the interface. This could explain the maximum accumulation of OmcZ [9] and the highest metabolic activity [9] observed near the biofilm–electrode interface (Fig. 4b).</p><p>Both OmcS and OmcZ are important for electricity production: deletion of omcS inhibits electricity production during the early stages of biofilm growth [6,26], whereas deletion of omcZ precludes formation of thick, high–current density biofilms [20]. In wild-type biofilms, OmcZ accumulates near the electrode, whereas OmcS is distributed throughout the biofilm [25]. Based on all these findings, we propose a new model that OmcS nanowires are involved in the biofilm growth during early stages, whereas OmcZ nanowires help bacteria form 100 μm–thick biofilms because of their high conductivity (Fig. 4a and b).</p><p>The OmcS is also essential for Fe (III) oxide reduction [19] and direct interspecies electron transfer between Geobacter co-cultures [21,27], with cells connecting each other via anti-OmcS–labeled filaments [21]. Analysis of such anti-OmcS–labeled filaments revealed structure similar to OmcS nanowires [13]. We therefore propose that G. sulfurreducens use OmcS nanowires to transfer electrons to Fe(III) oxide (Fig. 4c) and to accept electrons from Geobacter metallireducens (Fig. 4d).</p><!><p>The nanowires were thought to be type IV pili composed of PilA protein [3] because ΔpilA cells did not produce conductive filaments [6] and could not transfer electrons to extracellular acceptors such as iron [3] or electrodes in microbial fuel cells [28]. However, the presence of PilA in a filament of wild-type cells has not been established, only inferred from indirect evidence such as the presence of PilA monomer in filament preparations that also contain OmcS nanowires [13]. It is important to note that the conduction along the length of a single bona fide PilA filament has not been demonstrated and theoretical studies did not find conductivity in modeled PilA filaments [29,30], except when hypothesized that aromatic residues are within 3–4 Å of each other [31]. It is possible that synthetic pilA could assemble into a filament under artificial conditions [32–34]. However, these individual synthetic filaments' conductivity has not been shown along their length, only across their diameter, and their exact composition and structure is unknown.</p><p>We propose that, rather than serving as nanowires, pili are involved in the translocation of OmcS and OmcZ nanowires to the outer surface. The deletion of pilA inhibits the extracellular translocation of OmcS [23,27] and OmcZ [23] nanowires, which are essential for extracellular electron transfer to iron [19] and high–current density biofilms [20], respectively. Furthermore, overexpression of PilA is accompanied by over-production of OmcS [25], OmcZ [20], and extracellular filaments that result in the formation of highly conductive biofilms with enhanced current density [25]. Cryo-EM studies did not find any filaments with structure consistent with type IV pili either in filaments from current-producing wild-type biofilms or in previously published images of intact, cell-attached filaments [13]. Analysis of previously published filament images, that were thought to be pili, showed structure similar to OmcS nanowires [13]. Furthermore, conductivity measurements along the length of individual OmcS and OmcZ nanowires showed values similar to conductivity values [35,36] for filaments of wild-type [13] and the W51W57 strain [9], respectively. All these results suggest that these previous studies, including some of our own [7], interpreted OmcS and OmcZ nanowires as pili. It is, therefore, important to identify the conditions under which G. sulfurreducens can naturally show pili and determine their composition, structure, and conductivity to evaluate their exact function.</p><!><p>Long-range electron transport requires the formation of thick (>50 μm) electrically-conductive biofilms. The omcS gene deletion has no obvious impact on current production in biofilms.</p><!><p>The omcS deletion inhibits current production during early stages of biofilm growth [26], suggesting that OmcS nanowires are involved in the current production. Our discovery of OmcZ nanowires helps to explain how cells could compensate for the loss of OmcS nanowires during later stages of thick biofilm growth.</p><!><p>OmcS filaments do not participate in long-range electron transport as the expression of a pilin gene in which 5 aromatics were mutated ("Aro5') generates a less conductive biofilm; heterologous expression of pilin genes from other bacteria in G. sulfurreducens yielded strains expressing pili with low conductivity, but that expressed even more outer-surface OmcS.</p><!><p>Pili are required for the secretion of both OmcS and OmcZ cytochromes and not just OmcS alone [23]. Therefore, it is necessary to evaluate the expression and localization of both OmcS and OmcZ nanowires, as well as other outer-surface cytochromes, before attributing the aforementioned phenotypes solely to pili.</p><p>G. sulfurreducens strain KN400, which expresses much less OmcS and much more PilA than wild-type G. sulfurreducens, generates higher current and much more conductive biofilms.</p><p>We found that this strain produces OmcZ nanowires [9] with 1000-fold higher conductivity than OmcS nanowires. This could also explain the ability of KN400 strain to generate higher current and more conductive biofilms. Therefore, it is not surprising that OmcS is not essential for the KN400 strain.</p><!><p>E-pili expression is required for Fe(III) oxide reduction, but there are substitutes for the deletion of OmcS, such as magnetite.</p><!><p>Without an atomic resolution structure of a Geobacter filament composed of PilA, there seems to be no conclusive proof that Geobacter can make such a filament. As the deletion of pilA inhibits the extracellular translocation of OmcS [23,27] nanowires, it is not possible to attribute lack of Fe(III) oxide reduction in a pilA mutant to pili alone. Furthermore, magnetite is electrically conductive and is shown to facilitate extracellular electron transfer [39]. We have previously shown that in a sediment environment conductive minerals can transport electrons over centimeters, 10,000 times the size of a cell [40]. Therefore, introducing such conductive minerals can compensate for the loss of OmcS nanowires.</p><!><p>Iron corrosion, current production, and syntrophic growth do not require OmcS filaments emanating at distances from the cell.</p><!><p>If gene deletion studies show that a cytochrome is not essential, it does not mean that it has no function under wild-type growth conditions. Bacteria use multiple approaches and redundant pathways for growth. The discovery of OmcZ nanowires shows that many cytochromes could form nanowires, not just OmcS.</p><!><p>There is no correlation between the expression level of PilA and the secretion of OmcS.</p><!><p>The lack of correlation does not necessarily mean a lack of causation. Deletion of pilA inhibits the secretion of OmcS and OmcZ, establishing that PilA is required for nanowire secretion [23,27]. No prior studies have ever quantified the density of either OmcS or OmcZ nanowires in current-producing biofilms or visualized their network to determine the percolation threshold that determines biofilm conductivity. The lack of correlation between the total amount of OmcS/OmcZ in biofilms and measured biofilm conductivity does not mean that cytochrome nanowires do not confer conductivity to biofilms [41].</p><!><p>The culture conditions of Wang and co-workers' research [13] are not good for the expression of e-pili [37]. That is the reason that PilA is barely detectable in their filament preparation.</p><!><p>The culture conditions for current-producing biofilms [42], that led to discovery of OmcS and OmcZ nanowires [9,13], are identical to those previously used by Lovley and colleagues to evaluate long-range electron transport in biofilms where they found overexpression of PilA and OmcZ [20]. Both SDS-PAGE gel and western immunoblotting confirm the presence of abundant PilA monomer in our filament preparations from these biofilms [13]. However, cryo-EM revealed that all the filaments in current-producing highly-conductive biofilms are OmcS and OmcZ nanowires and not pili [9,13].</p><!><p>When asked about the importance of discovering conducting polymers that started the field of plastic electronics, Nobel laureate Alan Heeger offered two basic answers: (i) they did not previously exist and (ii) they offer a unique combination of properties not available from any other known materials [43]. The first expressed an intellectual challenge; the second expressed a promise for a wide range of applications.</p><p>The discovery and many properties of microbial nanowires are just like conducting polymers. Without interdisciplinary approaches, none of these discoveries would have been possible. Moreover, the ability to modulate their conductivity by targeted changes in the sequence or environment is particularly exciting because it can provide a foundation for a new field of research on the boundaries between molecular biology, microbiology, biophysical chemistry, and physics.</p><p>These discoveries are creating many opportunities:</p><p>First, microbial nanowires are opening the way for understanding the fundamental chemistry and physics of electron transport in proteins over micrometer distances at rates not previously known in biomolecules. Recent theoretical studies have suggested quantum-coherence effects in the conductivity of OmcS nanowires [44] that need to be examined using the atomic structure of nanowires.</p><p>Second, microbial nanowires are providing an opportunity to address questions that had been of interest, such as polymerization of cytochrome c in apoptosis [45] and design of synthetic metalloprotein nanowires [46–48].</p><p>Third, the ability to function at a low pH is a unique strength of these materials [9]. There is no other protein-based electronic material that shows such high electronic conductivity at low pH, to our knowledge. Improved conductivity at low pH in polyaniline, discovered by Nobel laureate Alan MacDiarmid, was critical for the development of conducting polymer–based sensors [49]. Therefore, we anticipate that the discovery of protein-based electronic materials that can withstand and function in extreme environments will serve as a foundation for future developments of biosensor and pH sensors. Improved conductivity will enhance the performance of protein nanowire–based devices used for energy harvesting [50,51], sensors [52], ultra-low power computing [53], and bioelectronics such as living transistors [6] and supercapacitors [54].</p><p>Finally, microbial nanowires offer promise for achieving a new class of electronic materials that could exhibit the electrical and optical properties of metals and semiconductors but retain mechanical properties and demonstrate versatile functionalities of proteins to bring together synthetic biology with semiconducting technology [55].</p>

PubMed Author Manuscript

Simultaneous morphology manipulation and upconversion luminescence enhancement of β-NaYF4:Yb3+/Er3+ microcrystals by simply tuning the KF dosage

A strategy has been adopted for simultaneous morphology manipulation and upconversion luminescence enhancement of β-NaYF 4 :Yb 3+ /Er 3+ microcrystals by simply tuning the KF dosage. X-ray power diffraction (XRD), field emission scanning electron microscopy (FE-SEM), transmission electron microscopy (TEM), X-ray photoelectron spectroscopy (XPS) and photoluminescence spectra (PL) were used to characterize the samples. The influence of molar ratio of KF to Y 3+ on the crystal phase and morphology has been systematically investigated and discussed. It is found that the molar ratio of KF to Y 3+ can strongly control the morphology of the as-synthesized β-NaYF 4 samples because of the different capping effect of F − ions on the different crystal faces. The possible formation mechanism has been proposed on the basis of a series of time-dependent experiments. More importantly, the upconversion luminescence of β-NaYF 4 :Yb 3+ /Er 3+ was greatly enhanced by increasing the molar ratio of KF to RE 3+ (RE = Y, Yb, Er), which is attributed to the distortion of local crystal field symmetry around lanthanide ions through K + ions doping. This synthetic methodology is expected to provide a new strategy for simultaneous morphology control and remarkable upconversion luminescence enhancement of yttrium fluorides, which may be applicable for other rare earth fluorides.In recent years, the synthesis of inorganic nano-/microstructures with controllable morphologies and accurately tunable sizes has attracted much attention not only for fundamental scientific interest but also for their potential applications in the fields of photoelectric device, sensor, catalysis, biological labeling, imaging and drug delivery [1][2][3][4] . It is generally accepted that most of the applications of such materials strongly depend on various parameters, including crystal structure, morphology, size, and dimensionality. Subsequently, simultaneous control over shape, size and phase purity of crystals has been becoming the research focus and one of the challenging issues. Until now, a variety of inorganic crystals, such as oxides, oxyfluorides, fluorides, sulfides, hydrates and other compounds, have been prepared with different shapes and sizes by various methods [5][6][7][8] . However, the precisely architectural manipulation of inorganic functional materials with predictable size, shape and crystal phase is still a challenging and urgent task, owing to the complexity of crystal structures and compositions of materials. To clarify these issues clearly, a deep understanding on the nature of shape evolution and phase transition is still needed. As a

simultaneous_morphology_manipulation_and_upconversion_luminescence_enhancement_of_β-nayf4:yb3+/er3+_

<!>Results and Discussion<!>Upconversion luminescence properties.<!>Methods

<p>result, it is very important for us to establish the relationship between the observed complex phenomena of crystal growth with the underlying fundamental theories and principles, which could be regarded as a reference to controllable synthesis of other inorganic materials.</p><p>As a significant class of rare earth compounds, rare earth (RE) fluorides have been become a research focus in the material field due to their unique applications in optical communications, three-dimensional displays, solid-state laser, photocatalysis, solar cells, biochemical probes and medical diagnostics [9][10][11][12][13] . Among them, NaYF 4 has been regarded as one of the most excellent host lattices for performing multicolor upconversion (UC) luminescence of the doped RE ions, due to its low phonon energy, high chemical stability and good optical transparency over a wide wavelength range [14][15][16][17] . As we known, the crystal structure of NaYF 4 exhibits two crystallographic forms, namely, cubic (α -) and hexagonal (β -) phases, depending on the synthesis conditions and methods 18 . Previous studies have indicated that the hexagonal polymorph exhibits considerable enhanced UC emissions compared with the cubic one 14,15 . Consequently, how to obtain pure β -NaYF 4 is crucial in successfully achieving high luminescence performance. Till now, many efforts have been dedicated to exploring excellent routes to the synthesis of hexagonal NaYF 4 with various sizes and shapes, such as nanospheres, nanoplates, nanorods, nanotubes, microrods, microtubes, microshpheres, micro-bipyramids, microplates and microprisms [19][20][21][22] . However, it is still limited on the investigation of the mechanism underlying the shape and phase evolution of NaYF 4 microcrystals. A deep understanding on the dynamic process governing nucleation and growth of the complex fluoride microcrystals is further needed.</p><p>Compared with other phosphors, such as organic fluorophores and quantum dots, lanthanide ions doped β -NaYF 4 crystals have many advantages, including sharp emission peaks, large anti-Stokes shifts, long-lived excited electronic states and high photostability [23][24][25] . But in spite of these advances, improvements are still needed to optimize UC luminescence properties for further potential commercialization. The remarkable challenge for us is how to further enhance the UC intensities of RE ions doped β -NaYF 4 crystals, which has considerable significance to their applications. So far, several attempts have been devoted to improving UC intensity via internal adjustment and external approaches, such as sensitizing mechanisms 26 , the formation of core-shell structure 27 , the introduction of non-lanthanide ions 28,29 and the incorporation of noble metals 30,31 . Among these methods, co-doping with non-lanthanide ions provides an alternative approach to enhance UC luminescence intensity by adjusting the crystal field symmetry.</p><p>Herein, we demonstrate a facile and effective hydrothermal process to synthesize β -NaYF 4 microcrystals using KF as fluoride source. In our experiments, the KF serves two purposes: (1) to tune morphology of the final products based on different capping effect of F − ions on the different crystal faces;</p><p>(2) to tailor the local crystal field of host lattice by K + ions doping. By simply tuning the molar ratio of KF to Y 3+ , regular β -NaYF 4 crystals with controllable morphologies can be obtained. In addition, the phase and morphology evolution process as well as the formation mechanism have been systematically investigated and discussed in detail. Meanwhile, significant enhancement of UC luminescence intensity was also observed in β -NaYF 4 :Yb, Er microparticles by simply tuning the KF dosage. To the best of our knowledge, simultaneous morphology control and UC luminescence enhancement for Yb, Er co-doped β -NaYF 4 microcrystals has been reported for the first time, and KF is rarely used as fluoride source.</p><!><p>Structures and Morphologies of As-prepared NaYF 4 samples. Here, we mainly focus on the effect of the KF/Y 3+ molar ratios on the phases and morphologies of the final products. In our experiments, the other synthetic parameters were set as reaction temperature 220 °C, reaction time 24 h, Na 3 Cit 1 mmol. Figure 1 shows the XRD patterns of these samples obtained at different molar ratio of KF/Y 3+ as well as standard data of pure hexagonal NaYF 4 phase for comparison. As shown, all crystals exhibit diffraction patterns corresponding to the hexagonal phase of NaYF 4 according to JCPDS No. 28-1192. The crystal structure of the hexagonal phase is determined with lattice parameters of a = 0.596 and c = 0.353 nm, space group P6 3 /mmc 15 . No trace of characteristic peaks is detected for other impurity peaks such as KYF 4 , YF 3 , indicating that the simple hydrothermal method is a feasible route to synthesize pure β -NaYF 4 microcrystals using KF as fluoride source. Moreover, careful observation reveals that the relative diffraction peak intensity of XRD patterns varies with different molar ratio of KF to Y 3+ , implying the morphology evolution of as-prepared β -NaYF 4 crystals.</p><p>The typical morphology evolution of β -NaYF 4 microcrystals obtained with increasing KF/Y 3+ molar ratio is presented in Fig. 2. As shown, the molar ratio of KF to Y 3+ has a profound influence on the morphologies of the as-synthesized samples. At low molar ratio of KF to Y 3+ (KF/Y 3+ = 16), the SEM images in Fig. 2a reveal that the sample is composed of a large quantity of microrods with uniform size of 11.8 μ m in length and 2.3 μ m in diameter. From the images of a higher magnification (Fig. 2b) and a typical individual microrod (Fig. 2c), we can see that the products are prismatic microrods with smooth and flat surfaces as well as sharp ends. Furthermore, the ends of rods are of hexagonal pyramid structure, as shown in Fig. 2c. At a medium molar ratio of KF to Y 3+ (KF/Y 3+ = 30), the general images of β -NaYF 4 sample are shown in Fig. 2d. It clearly indicates that the as-prepared product consists of a great deal of hexagonal microprisms with prefect uniformity, monodispersity and well-defined crystallographic facets. The mean size of microprism is calculated to be about 2.6 μ m in diameter and 12.1 μ m in length. Further investigation under higher magnification (Fig. 2e,f) indicates that both tops and bottoms of these microprisms exhibit flat planes. At high molar ratio of KF to Y 3+ (KF/Y 3+ = 50), the regular hexagonal prism-shaped β -NaYF 4 with an average size of 11.5 μ m in length and 2.8 μ m in diameter are observed from Fig. 2g. A magnified SEM image (Fig. 2h) reveals that the large-scale, regular and monodisperse prismatic microrods with soomth and flat surfaces are obtained in this experimental condition. Interestingly, the surfaces of top/bottom have very regular concave centers, as depicted in Fig. 2i. It's worth mentioning that the above-mentioned experiments have been repeated three times at least and the same conclusion could be drawn from these experiments. From the above investigations, it can be concluded that the morphology of products can be tuned accordingly with no further change in particle size by changing the molar ratio of KF/Y 3+ .</p><p>Growth Mechanism. To understand the formation process of β -NaYF 4 microcrystals with different morphologies, reaction samples have been carefully investigated by quenching the reaction at different time intervals. Although we know that the reaction doesn't stop immediately after the autoclave is removed from the heater because of heat transfer, we do believe that the products synthesized at that time represent certain stage in the formation process. Figure 3 shows the XRD patterns of the NaYF 4 samples synthesized with 50:1 KF/Y 3+ at different reaction times as well as standard data of α -NaYF 4 (JCPDS No. 77-2042) and β -NaYF 4 (JCPDS No. 28-1192) phases for comparison. It reveals that the samples exhibit distinctively different XRD patterns at different reaction times. The sample obtained at t = 1 h is pure cubic NaYF 4 (Fig. 3a). A new hexagonal NaYF 4 phase emerges in addition to cubic NaYF 4 phase with the reaction proceeding for 2 h (Fig. 3b). With the further reaction from 2 h to 4 h, the fraction of β -NaYF 4 increases dramatically while the amount of α -NaYF 4 decreases. The result indicates that the phase transformation (α → β ) takes place through a dissolution-renucleation process [32][33][34] . When the reaction time increases from 4 h to 24 h, pure β -NaYF 4 can be successfully obtained, as shown in Fig. 3c-f. Based on the above analysis, it can be concluded that the crystal evolves from cubic phase to mixed phase and ultimately to hexagonal phase with the prolonged reaction time.</p><p>At the meantime, the morphologies of the products are carefully investigated by quenching the reaction at different time intervals. Figure 4 shows the corresponding FE-SEM images of the different intermediate samples at different reaction stages. It clearly reveals that the six samples exhibit dramatically different morphologies in the process of crystal growth. At a short reaction time of 1 h, the α -NaYF 4 sample consists of spherical-like nanoparticles with a mean diameter of 40 nm (Fig. 4a and Fig. S1 in Supplementary Information). But in the present situation, the α -phase of NaYF 4 is unstable and these nanoparticles would serve as seeds for the growth of β -NaYF 4 hexagonal microrods by a dissolution-renucleation process. With further reaction, these unstable α -NaYF 4 nanoparticles convert to β -NaYF 4 microprisms gradually. After 2 h of growth, the regular and well-defined microprisms begin to appear in the intermediate product. Moreover, a large amount of nanoparticles are attached on the surface of hexagonal microprisms, as shown in Fig. 4b. According to the corresponding XRD pattern, the coexisting of two shapes results from the presence of the mixture crystal phase (α + β ). As we known, the cubic NaYF 4 has isotropic unit cell structure, resulting in an isotropic growth of particles. As a result, spherical-like particles are observed. By comparison, hexagonal NaYF 4 has anisotropic unit cell structure, which can induce anisotropic growth along crystallographically reactive directions, leading to the formation of hexagonal-shaped structure 35 . On the basis of above analysis, it can inferred that these crystals with different morphologies should be cubic NaYF 4 (nanoparticles) and hexagonal NaYF 4 (microprisms), respectively. As reaction time extended to 4 h, the α -NaYF 4 nanoparticles disappear completely and only the fairly uniform well-defined β -NaYF 4 microprisms exist. The result also reveals that the phase transition (α → β ) can directly induce obvious change in the morphology of NaYF 4 crystals. As shown in Fig. 4c, the mean length and diameter of rods could be estimated to be 5.2 μ m and 1.0 μ m, respectively. With the further reaction from 8 h to 24 h, there are no further change in morphology, but the size of microprisms increases from 8.8 μ m to 11.6 μ m in length and from 1.9 μ m to 2.8 μ m in diameter (Fig. 4d-f), indicating the longitudinal and transversal growth of NaYF 4 microprisms along with the reaction time. We also investigate the growth process of β -NaYF 4 crystals synthesized with other KF/Y 3+ molar ratios (KF/Y 3+ = 20, 25, and 40). As can be seen from the XRD patterns (Fig. S2-S4 in Supplementary Information), these crystals exhibit very similar phase transformation process to that of these three samples. In addition, the FE-SEM images (Fig. S5-S7 in Supplementary Information) also reveal a similar phase transformation process (α → β ) to these three products.</p><p>Based on the above results, a possible phase and morphology evolution mechanism is shown in Fig. 5 and described as follows. At the beginning, the citrate anions (Cit 3− ) introduced into the reaction system can form complexes with Y 3+ ions through strong coordination interaction. At the same time, KF dissolves in the aqueous solution to form K + and F − ions.</p><p>Under the conditions of high temperature and high pressure, the chelating ability of the Y 3+ − Cit 3− complexes would be weakened by slow degrees during hydrothermal process, resulting that the Y 3+ ions could be released gradually. Then Na + , K + and F − ions in this solution react with Y 3+ ions to generate small nuclei. In a very short reaction time, these nuclei quickly aggregate together and grow into cubic-phased K x Na (1−x) YF 4 nanoparticles.</p><p>However, these grown up α -K x Na (1−x) YF 4 nanoparticles are thermodynamically unstable and evolve inevitably to hexagonal K x Na (1−x) YF 4 seeds through dissolution-renucleation process. Meanwhile, this phase transformation (α → β ) results in dramatic morphology change of the samples, which could be related to different characteristic unit cell structures for varying crystallographic phases. The dissolution-reconstruction process of cubic-phased nanoparticles preferentially happens at the circumferential edges of each prismy microrod along crystallographically reactive direction, resulting in the formation of rod-like β -K x Na (1−x) YF 4 with a well-defined cross-section. With the further reaction, the morphology of products changes from spherical nanoparticles into short hexagonal microrods.</p><p>As we known, the shape evolution of β -NaYF 4 microcrystals is significantly dependent on external factors such as the molar ratios of KF to Y 3+ and Na 3 Cit to Y 3+ , the pH values in the solution 36 . In this experiment, the subsequent crystal growth of β -K x Na (1−x) YF 4 seeds is significantly affected by the KF/ Y 3+ molar ratio, resulting in different morphologies of hexagonal K x Na (1−x) YF 4 microcrystals. Although the exact mechanism is not very clear at present, the explanation for the change of morphology can be provided as follows. According to the general principle of crystal growth, the growth of crystals is related to the relative growth rate of different crystal facets 33,37,38 . The different growth rate of various crystal planes results in diverse appearance of the crystallite. Generally speaking, crystal facets perpendicular to the fast directions of growth have smaller surface area and show growing faces therefore dominate the morphology of the final crystal. The growth velocity in different crystallographic facets of β -K x Na (1−x) YF 4 crystals could be influenced by the coordination effect between F − and Y 3+ ions. According to the Gibbs-Thompson theory, the relative chemical potential of crystal plane is simply proportional to its surface-atom ratio, determined by the average number of dangling bonds per atom over the entire crystal facet 20 . The capping effect of F − ions could decrease the average number of dangling bonds and further reduce the chemical potential of the crystal plane. Moreover, the different density of Y 3+ on various crystal planes leads to the difference in chemical potential of crystal facets. Consequently, the chemical potential of different crystal facets could be modified, and the relative growth rates could be affected by different molar ratio of KF to Y 3+ , finally leading to different crystal shapes. For β -K x Na (1−x) YF 4 crystals, the density of Y 3+ on the prismatic planes ({10-10} crystal planes) is bigger than that on the top/bottom facets ({0001} crystal planes), resulting in the selective adsorption ability of F − ions on the prismatic facets being bigger than that on the top/bottom planes. Finally, the relative growth rate along [0001] is much quicker than that of along 10-10, resulting in the hexagonal microrods with long length and high aspect ratio. The difference in the top ends of hexagonal microrods is relate to the different capping effect of F − on the {10-11}, {-101-1} and {0001} crystal planes, as depicted in Fig. 6. At low F − concentration, the capping effect of F − on the {10-11} and {-101-1} crystal planes is greater than it on the {0001} plane. Consequently, the growth rate of {10-11} and {-101-1} crystal facets is faster than that of {0001} planes, resulting in the formation of sharp ends. At medium F − concentration, the capping effect of F − ions on the {0001} facets is greater and the fast growing faces ({0001} crystal planes) therefore induce to the flat ends. The concave structure is observed at the top/bottom facets by further increasing the F − concentration in the solution. The presence of the concave ends demonstrates that the growth rate of the prismatic side facets ({10-10} planes) is a little faster than that of the top/bottom facets ({0001} planes).</p><!><p>To investigate the UC luminescence properties of β 4 synthesized with different molar ratios of KF to RE 3+ , Yb 3+ /Er 3+ are selected as co-doped ion pairs to form β -NaYF:Yb 3+ , Er 3+ microcrystals. Figure 7 shows the UC emission spectra of β -NaYF 4 :20%Yb 3+ , 2%Er 3+ samples (20 mg powder is dispersed in 10 mL ethanol under ultrasound treatment) synthesized by different molar ratio of KF to RE 3+ under 980 nm laser diode excitation (power density: 0.2 W/mm 2 ). As can be seen clearly, the six samples show the same emission peaks yet with quite different emission intensity. As shown in Fig. 7a, the characteristic UC emission bands centered at 521 nm, 540 nm and 656 nm can be ascribed to 2 H 11/2 → 4 I 15/2 (green), 4 S 3/2 → 4 I 15/2 (green) and 4 F 9/2 → 4 I 15/2 (red) transitions of Er 3+ , respectively 39 . It is worth noticing that the UC luminescence intensities in both the green and red regions increase notably with the increment of molar ratio of KF to RE 3+ . Figure 7b exhibits the integral intensity of green and red emissions as a function of the KF dosage. By increasing the molar ratio of KF to RE 3+ , the integral intensities of 521 nm, 540 nm and 656 nm emissions are enhanced dramatically. The integrated green (500-600 nm) and red (600-700 nm) emissions in KF6 sample are measured to be about 15 and 12 times as high as that of KF1 sample (Table S1 in Supporting Information). In addition, the similar trend of green and red emissions also suggests the same upconversion pathways for them. To visualize the enhancement of UC emission, the corresponding luminescence photographs of β -NaYF 4 :20%Yb 3+ , 2%Er 3+ crystals synthesized with different KF dosage are provided in Fig. 7c. The green emission of six samples can be seen clearly by naked eyes. Moreover, the emission intensity of KF6 sample is the strongest, which agrees with the results in Fig. 7a,b. To accurately demonstrate enhancement of UC luminescence, the quantum yields of the KF1 and KF6 samples were measured. Importantly, the UC efficiencies of KF1 and KF6 are roughly estimated to be ~0.3% and ~1.6%, respectively. The detailed measuring procedure for quantum yield has been presented in supporting information (Fig. S8). Additionally, the obvious UC enhancement can also observed in Yb 3+ /Tm 3+ co-doped β -NaYF 4 samples (Fig. S9 in Supplementary Information). Notably, the integrated blue emission of the β -NaYF 4 :20%Yb 3+ , 0.5%Tm 3+ sample obtained with KF/RE 3+ molar ratio 50 are enhanced by about 80 times (Table S2 in Supporting information), resulting in the strongest blue emission under the excitation of 980 nm laser diode. The results are quite similar to the Yb 3+ /Er 3+ co-doped β -NaYF 4 samples, revealing the generality of the approach. In order to deeply investigate the relevant UC mechanism in the as-synthesized β -NaYF 4 :20%Yb 3+ , 2%Er 3+ crystals, the excitation power-dependent UC emissions of green and red are calculated accordingly. It is generally known that the output UC emission intensity (I uc ) is proportional to the infrared excitation power (I IR ): I uc ∝ (I IR ) n , where n is the absorbed photon numbers per visible photon emitted, and its values can be acquired from the slope of the fitted line of the plot of log(I uc ) versus log(I IR ) 40,41 . The pump power dependence of the UC emissions in KF6 sample under 980 nm LD excitation is presented in Fig. 7d. As shown, the slopes of the linear fit of log(I uc ) versus log(I IR ) for 521 nm, 540 nm and 656 nm are 1.67, 1.86 and 1.94, respectively. The result indicates that only two-photon process is involved to produce the green and red UC emissions, whereas a saturation effect can be observed at relatively higher excitation power. Based on the above results, the proposed UC mechanism in β -NaYF 4 :Yb 3+ /Er 3+ under 980 nm LD excitation is shown in Fig. 7e and briefly described as follows 42 . Firstly, the electron of Yb 3+ is excited from 2 F 7 to 2 F 5 level in β -NaYF 4 : Yb 3+ /Er 3+ microcrystals under 980 nm LD excitation. An initial energy transferred from Yb 3+ ions in the 2 F 5/2 state to Er 3+ ions populates the 4 I 11/2 level of Er 3+ ions. Then, a second 980 nm photon transferred by the adjacent Yb 3+ ions can populate the 4 F 7/2 level of Er 3+ ions, whose energy lies in the visible region. The Er 3+ ions can relax nonradiatively to the level of 2 H 11/2 , 4 S 3/2 and 4 F 9/2 . Through a two-photon UC process, the dominant green and red emissions are observed by these transitions from the aforementioned states to 4 I 15/2 level. In order to demonstrate the UC enhancement more theoretically, the emission decay curves of 4 S 3/2 → 4 I 15/2 (540 nm) and 4 F 9/2 → 4 I 15/2 (656 nm) transitions in the six samples were measured at the excitation wavelength of 980 nm, as shown in Fig. 8. The effective experimental lifetime is evaluated using</p><p>eff where I(t) represents the luminescence intensity at time t after the cutoff of the excitation light 43 . It can be seen clearly that the lifetimes of 4 S 3/2 and 4 F 9/2 states in β -NaYF 4 :Yb 3+ /Er 3+ samples are prolonged gradually with the increase of molar ratio of KF to RE 3+ . Furthermore, the variation trend of lifetimes is also consistent with the enhancement of UC luminescence intensity. The average lifetimes of 4</p><p>According to the UC mechanism and experimental results, it can be concluded that the energy transfer process between Yb 3+ and Er 3+ hasn't been changed by tuning the KF dosage. Moreover, the nonradiative transition rate should not have obvious effect on the great enhancement of UC luminescence because of the same experimental and excitation conditions for these samples. Therefore, we can ascribe the as-observed much longer lifetimes (τ) to the increase of theoretical lifetimes (τ rad ), namely the modification of lattice parameters. The introduction of K + ions can tailor the local crystal field of the host lattice, and therefore can modify the theoretical lifetimes of Er 3+ ions through slightly changing their wave functions.</p><p>Enhancement mechanism for UC emissions. What accounts for the obvious enhancement of UC luminescence when tuning the KF dosage in our experiments? Three important factors for UC enhancement should be in consideration: crystal phase, particle size and morphology, local crystal field. First of all we should investigate the crystal phase, particle size and morphology of β -NaYF 4 :Yb 3+ /Er 3+ crystals synthesized with the addition of different KF dosage. The XRD patterns and FE-SEM images of as-prepared β -NaYF 4 :Yb 3+ /Er 3+ samples are presented in Fig. S12 and Fig. S13, respectively. It is evident that all the peaks can be indexed to pure hexagonal-phased NaYF 4 according to JCPDS card (No. 28-1192). No additional peaks can be detected, indicating that the addition of KF has not lead to the formation of other impurity phases. Although the morphology of crystals has changed a lot, the specific surface areas of the six samples show a little distinction from each other (Table S3 in Supplementary Information). Moreover, the luminescent property of phosphors is relate to their specific surface area of the materials. Consequently, it is obvious that the first two factors should not have an obvious effect on marked enhancement of UC luminescence. However, careful observation reveals that the position of the main diffraction peak at 17.1° in the XRD pattern shifts slightly towards small angles as the molar ratio of KF to RE 3+ increases (Fig. 9a). The variation of lattice parameters and unit cell volume with the KF dosage were calculated form the observed XRD data and are presented in Fig. 9b. According to the Bragg's law (nλ = 2dsinθ), the decrease of Bragg angle (θ) indicates that the spacing between the planes in atomic lattice (d) increases, resulting in the expansion of unit cell. Considering that KF is used as fluorine source and r K (1.51-1.57 Å) > r Na (1.13-1.53 Å), we can easily infer that some K + ions are doped in β -NaYF 4 host lattice and may occupy lattice sites by the substitution of Na + ions. Furthermore, X-ray photoelectron spectroscopy (XPS) was used to determine the successful incorporation of rare earth ions (Yb 3+ , Er 3+ ) and K + ions into the β -NaYF 4 host matrix. As shown in Fig. 9c, the peaks observed at 285.5 and 302.0 eV can be assigned to the binding energy of K2p 1/2 and K2p 3/2 , respectively 45 . The characteristic peaks at 198.8 and 186.5 eV can be assigned to the binding energy of Yb 4d and Er 4d, respectively (Fig. 9d) 46 . The EDS and ICP-AES analyses (Table 2 and Fig. S15 in Supplementary Information) are also used to confirm the successful incorporation of K + , Yb 3+ , Er 3+ ions into β -NaYF 4 matrix. Evidently, with increasing of the KF dosage in the product, K content increases gradually, and the Yb and Er contents keep unchanged. Therefore, it is believed that the introduction of K + ions into the host lattice leads to the UC enhancement. Based on the above results, we can provide the following explanation for this obvious enhancement of UC luminescence. According to the single-crystal X-ray powder diffraction data, the crystal structure of hexagonal NaYF 4 with the P6 3 /mmc has three types of cation sites in a unit cell: one for rare earth ions (site 1a), one for both rare earth and sodium ions (site 1f), and the third for sodium ions (site 2h). Site 1a and 1f both have C 3h symmetry, whereas site 2h has C s symmetry 18,47 . When trivalent lanthanide ions (Yb 3+ , Er 3+ ) are doped into β -NaYF 4 lattice by isomorphic substitution of Y 3+ ions without charge compensation, only one kind of substitutional site with a crystallographic site symmetry of C 3h could be observed in the unit cell. It is well known that the electric-dipole transitions between 4f n configuration (f-f) with the same parity are parity-forbidden for free lanthanide ions according to the quantum selection rules. However, such prohibition can be broken by mixing of opposite-parity configuration, resulting that the electric-dipole transitions can be weakly allowed in crystal lattice. In order to greatly increase the electric diploe transitions probability, an asymmetric crystal field is required. When K + ions are doped in crystal lattice, in view of the bigger ionic radius of K + relative to that of Na + , K + ions only occupy the lattice sites by the substitution of Na + ions in the crystal, corresponding to the situation presented in Fig. 10 48 . As a result, the coordination shell around site 1f originally statistically distributed by Y 3+ and Na + ions could be perturbed seriously. Accordingly, the displacement patterns of various Y/Na coordination shell around each subset of lanthanide ion can be slightly different. According to the microscopic model of disorder, the local site symmetry around the lanthanide ions may descend from C 3h to lower symmetries C s . For rare earth ions embedded in solid materials, lower crystal symmetry caused by tailoring the crystal structure is generally favorable for higher UC emission intensity. Therefore, the introduction of K + ions into β -NaYF 4 crystal lattice would change the crystal field and lower the local crystal field symmetry around lanthanide ions, resulting in the enhancement of UC luminescence intensity 49 . Tailoring the local crystal field of host lattice has become a general explanation for increasing UC emission intensity when introducing non-luminescent ions, such as Li + , Bi 3+ , Sc 3+ , Fe 3+ ions 29,46,50,51 . However, the direct evidence of tailoring local crystal field is still lacking. So, investigation on the local structure and site symmetry of lanthanide ions, for example, by using high-resolution photoluminescence spectroscopy at low temperature (10 K), are further needed to provide such proof.</p><p>In conclusion, we have described a facile and general strategy for simultaneous morphology manipulation and UC luminescence enhancement of β -NaYF 4 :Yb 3+ /Er 3+ samples using KF as fluoride source. Through the simple manipulation of the KF/Y 3+ molar ratio, regular β -NaYF 4 microcrystals with predictable shapes have been synthesized. A mechanism for how the molar ratio of KF to Y 3+ influences the anisotropic growth and morphology evolution of β -NaYF 4 crystals is proposed. Based on the phase and morphology evolution, the possible formation mechanism for hexagonal NaYF 4 is discussed in detail. Notably, the UC luminescence intensity of β -NaYF 4 :Yb 3+ /Er 3+ sample is significantly enhanced by increasing the KF dosage. It is found that the doping of K + ions into β -NaYF 4 crystal lattice can tailor the local crystal field and lower the local crystal field symmetry around lanthanide ions, which is the main reason for the UC enhancement. This study provides a reference for simultaneous morphology control and UC luminescence enhancement of rare earth fluorides, which will have great potential in fields of sensors, solar cells and photocatalysis. Preparation. β -NaYF 4 :Yb 3+ , Er 3+ microcrystals have been synthesized via a facile hydrothermal method assisted with trisodium citrate. In a typical procedure for the synthesis of β -NaYF 4 : 20%Yb 3+ , 2% Er 3+ , 1.56 mmol YCl 3 .6H 2 O, 0.40 mmol YbCl 3 .6H 2 O, 0.04 mmol ErCl 3 .6H 2 O were firstly dissolved in 10 mL H 2 O with magnetic stirring to form rare earth chloride solution. Then this solution was added into 20 mL aqueous solution containing 2 mmol trisodium citrate (Na 3 Cit) to form the metal-citrate complex. After vigorous stirring for 30 min, 30 mL aqueous solution containing different amounts (32, 40, 50, 60, 80, 100 mmol) of KF.2H 2 O was introduced into the above solution. After additional agitation for 15 min, the as-obtained mixing solution (60 mL) was transferred into a Teflon bottle (100 mL) held in a stainless steel autoclave, sealed, and maintained at 220 °C for 24 h. As the autoclave was cooled to room temperature naturally, the precipitates were separated by centrifugation, washed with deionized water and ethanol in sequence, and then dired in air at 80 °C for 12 h. In addition, different molar ratios (16:1, 20:1, 25:1, 30:1, 40:1, 50:1) of KF to RE 3+ (RE = Y, Yb, Er) and hydrothermal treatment times (0.5 h, 1 h, 3 h, 8 h, 12 h, 24 h) were selected to investigate the effects of these factors on the morphological, structural and luminescent properties of the as-obtained sampleaccs. According to the KF/RE 3+ molar ratio, the resulted samples were denoted as KF1, KF2, KF3, KF4, KF5, and KF6, resspectively. These as-syntheszied products were used to characterize without any further purification.</p><!><p>Characterization. Powder X-ray diffraction (XRD) measurements were performed on a ARL X' TRA diffractometer at a scanning rate of 10°/min in the 2θ range from 10° to 80° with Cu Kα radiation (λ = 0.15406 nm). SEM micrographs were obtained using a field emission scanning electron transmission microscope (FE-SEM, SU8010, Hitachi). Transmission electron microscopy (TEM) was recorded on a JEM-200CX with a field emission gun operating at 200 kV. Images were acquired digitally on a Gatan multiple CCD camera. The chemical compositions were determined by inductively coupled plasma (ICP) technique using a Perkin-Elmer Optima 3300DV spectrometer. X-ray photoelectron spectroscopy (XPS) spectra were performed with a PHI5000 VersaProbe system, using a monochromatic Al Kα X-ray source. UC emission spectra were recorded with a Jobinyvon FL3-221 fluorescence spectrophotometer equipped with a 980 nm LD (laser diode) Module (K98SA3M-54W, China) as the excitation source. Upconversion decay lifetimes were measured with a customized UV to mid-infrared steady-state and phosphorescence lifetime spectrometer (FSP920-C, Edinburgh) equipped with a digital oscilloscope (TDS3052B, Tektronix) and a tunable mid-band OPO pulse laser as excitation source (410-2400 nm, 10 Hz, pulse width ≤ 5 ns, Vibrant 355II, OPOTEK). All the measurements were performed at room temperature.</p>

Scientific Reports - Nature

Degradation and Detection of Endocrine Disruptors by Laccase-Mimetic Polyoxometalates

Endocrine disruptors are newly identified water contaminants and immediately caught worldwide concern. An effort has been made to degrade endocrine disruptors in the water body by relying on laccase-assisted approaches, including laccase-mediated catalytic systems, immobilized laccase catalytic systems, and nano-catalytic systems based on atypical protein enzymes. Analogous to laccases, polyoxometalates (POMs) have a similar size as these enzymes. They are also capable of using oxygen as an electron acceptor, which could assist the removal of endocrine disruptors in water. This perspective begins with a brief introduction to endocrine disruptors and laccases, summarizes current approaches employing laccases, and focuses on the nano-catalytic systems that mimic the function of laccases. Among the inorganic nanoparticles, POMs meet the design requirements and are easy for large-scale production. The catalytic performance of POMs in water treatment is highlighted, and an example of using polyoxovanadates for endocrine disruptor degradation is given at the end of this perspective. Exploring laccase-mimetic POMs will give key insights into the degradation of emergent water contaminants.

degradation_and_detection_of_endocrine_disruptors_by_laccase-mimetic_polyoxometalates

Introduction<!><!>Introduction<!>Endocrine Disruptors in Water<!>Laccase-Assisted Detection and Degradation<!>Laccase-Mimicking Nanozymes<!>Enzyme-Mimicking POMs and POMs in Water Treatment<!><!>Enzyme-Mimicking POMs and POMs in Water Treatment<!>Challenges and Perspective<!>Discussion<!>Data Availability Statement<!>Author Contributions<!>Funding<!>Conflict of Interest<!>Publisher’s Note

<p>Endocrine disruptors are emergent water contaminants which frequently found in every aspect of human life, including some plastic bottles and containers, fungicides, disinfectants, anti-viral agents, pharmaceutical drugs for oral analgesic and mild anesthetic, and cosmetics and skin care products (Figure 1A) (Bilal et al., 2019a). Some of the endocrine disruptors are phenol products that are analogous to natural hormones due to their capability of acting like natural hormones and disrupting the endocrine system (Barrios-Estrada et al., 2018). Endocrine disruptors receive more attention than other phenols because of their appearance in various sources and unique interferences with natural hormones in interaction with corresponding receptors that result in an altered cellular signal and subsequent a failure in the body. The persistence of endocrine disruptors in water bodies has raised particular concern because of the widespread and continuing sewage discharge.</p><!><p>(A) Common sources and typical structures of endocrine disruptors in water environment. (B) Scheme of immobilized laccase-assisted degradation of phenolic contaminants and the representative nanozymes that have been reported the catalytic activities of mimicking laccase.</p><!><p>Enzymatic degradation approaches have been developed for addressing this growing issue (Bilal et al., 2019b). Laccase catalyzed reaction effectively removes many kinds of trace pollutants, which are difficult to degrade in wastewater, especially some phenolic endocrine disruptors (Mate and Alcalde, 2017; Janusz et al., 2020). Laccases catalyze one-electron oxidation of a broad range of these phenol substrates and release water as the by-product (Figure 1B). As an eco-friendly and versatile biocatalyst, laccases have been applied in enzymatic bioremediation of the water bodies (Sharma et al., 2018; Morsi et al., 2020). To extend the using range in the harsh environment, a set of nanosystems, mostly nanozymes, are designed to mimic the function of laccases in wastewater treatment (Zhou et al., 2017). Polyoxometalates (POMs) are nanosized soluble molecular metal-oxo clusters with well-defined structures. Each metal-oxo cluster comprises an array of corner-sharing and edge-sharing pseudo-octahedral MO6 (M = Mo, W, V, Nb, etc.) units (Gumerova and Rompel, 2018). POMs have demonstrated promising catalytic activities over a wide range of catalysis fields (Anyushin et al., 2018). POMs for constructing the nanocatalytic system can be directly used to catalyze the same substrates as laccases. Meanwhile, POMs are much more cost-effective for production and robust for working at high temperature, high pressure, and extreme pH conditions (Liu et al., 2020). In this perspective, the current development in this field is briefly summarized, and the scientific and technological challenges are outlined. Exploring laccase-mimetic POMs would spark future research interest in advancing the technic for the removal of emergent water contaminants.</p><!><p>Environmental endocrine disruptors, a type of pollutant in water, are a classification of exotic chemicals that alter or interact with the endocrine systems of vertebrates and invertebrates (Figure 1A). Endocrine disruptors belong to the endocrine organic chemicals, but differ from natural phytoestrogens in that they mimic, block, or alter the actions of normal hormones (Landrigan et al., 2018). Endocrine disruptors have shown the potential to interact with biological systems, such as the hormonal system and nervous system, and cause various complications, such as neurodevelopmental toxicity and Parkinson's disease (Barrios-Estrada et al., 2018). Endocrine disruptors have a variety of mechanisms of action. One pathway is through the direct interaction with a given estrogen receptor, which may interfere with or regulate downstream gene expression. For example, most endocrine disruptor-related reproductive and developmental defects are thought to be due to endocrine disruptors interfering with the function of estrogen receptors and androgen receptors, thereby interfering with the regular activity of estrogen and androgen ligands (Barrios-Estrada et al., 2018). In addition to sex steroid receptors, the estrogen receptor superfamily includes transcription factors that play a key role in integrating complex metabolic homeostasis and development. The ability of endocrine disruptors to interact with these estrogen receptors is supported and explained by metabolic disorders in experimental and epidemiological studies (Barrios-Estrada et al., 2018). During early development, the exposure of even extremely low doses of endocrine disruptors will likely lead to permanent impairments in fetus organ function and increase their disease risk. In addition, many endocrine disruptors are also developmental neurotoxicants and can be stored in animal and human fats for years (Schug et al., 2011). These chemicals include bisphenol A, phthalates, and dioxins.</p><p>Water pollution has become a pressing issue as populations grow and industrial production expands. This growing problem is often linked to poor wastewater management, outdated infrastructure, factories, and limited treatment strategies. Most endocrine disruptors come from products used to combat unfavorable wildlife and agricultural threats, such as pesticides and fungicides, as well as various synthetic products used in the plastics industry (bisphenol A or phthalates), insulation materials (polychlorinated biphenyls), and brominated flame retardants (Bilal et al., 2019a). These chemicals are manufactured in the high output of millions of kilograms per year and have caused substantial impacts on our daily life. They are everywhere around people, in consumer products such as perfumes, shampoos, soaps, plastics, and food containers. Another problem associated with such chemicals is that they degrade very slowly or are not photodegradable (Bilal et al., 2019b). Therefore, endocrine disruptors have become a significant public health problem globally due to their high stability, low degradation, high toxicity, and persistence in the environment. Biological technics for the degradation of these pollutants using oxidoreductases are a promising area of research (Cabana et al., 2007a; Cabana et al., 2007b; Torres-Duarte et al., 2012).</p><!><p>Laccase is a group of enzymes with a wide taxonomic distribution that belongs to the copper oxidases (MCOs) superfamily (Mate and Alcalde, 2017; Janusz et al., 2020). MCOs reduce oxygen molecules to water without releasing harmful substances, including those species often generated during oxygen reduction (Figure 1B), such as the partially reduced products of O2, reactive oxygen species (ROS). Laccases are widely distributed in nature. Higher plants, bacteria, lichens, sponges, and fungi, especially white rot fungi, can produce laccases with different biological functions and substrate diversity (Mate and Alcalde, 2017). Aromatic compounds (e.g., catechol and hydroquinone, methoxy substituted phenols, diamines, and phenylthiols), organometals ([Fe(EDTA)]2− and [W(CN)8]4−), and metal ions are all the substrates of laccases.</p><p>The remarkable broad substrate specificity of laccase aroused the attention of those who are worried about the environment. Over the past few decades, laccases have been used as a biocatalyst to detect and reduce pollutants by removing electrons from these organic substrates and blocking the entry of these contaminants into the water bodies. The laccase-involving enzymolysis approach has been used in different industrial applications to replace traditional chemical processes in the paper, textile, cosmetics, paint, pulp, furniture, and pharmaceutical factories (Sharma et al., 2018; Morsi et al., 2020). To develop a robust laccase-based biocatalytic platform, the enzymes are normally immobilized on a support matrix to address the limitations related to enzyme reusability and recycling (Lassouane et al., 2019; Zhou et al., 2021). The stability and resistance of laccase (e.g., isolated from the basidiomycete Trametes versicolor) to protease is increased by the immobilization on a solid carrier (Sharma et al., 2018). Several methods for immobilizing enzymes have been developed, such as covalently attaching to solid carriers, solid carrier adsorption methods, embedding in polymeric gels, cross-linking with biofunctional reagents, and embedding in solid carriers (Bilal et al., 2017; Shakerian et al., 2020). Laccase immobilization is a promising water purification technology. Compared with free laccase, the reusability of immobilized laccase makes it more advantageous in the practical application of water purification (Lassouane et al., 2019; Lou et al., 2020; Masjoudi et al., 2021).</p><p>In recent years, nanocarriers have been engineered to immobilize and support enzymes, which greatly advanced traditional enzyme-immobilization methods. Many enzyme systems based on nanostructures are designed and used to detect a variety of organic pollutants and degrade them efficiently into harmless smaller intermediates (Alarcón-Payán et al., 2017; Chen C. et al., 2018; Masjoudi et al., 2021; Qiu et al., 2021). Furthermore, the attachment of laccase to its nanocarrier not only reduces its mobilization, but also increases activity and stability of the enzyme (Koyani and Vazquez-Duhalt, 2016; Costa et al., 2019). It was reported that the electrode modified with laccase-immobilizing polyaniline/magnetic graphene exhibited superior electrical properties, high detection sensitivity to hydroquinone, low detection limit, and wide linear range (Lou et al., 2020). Many carrier nanomaterials for laccase immobilization have been engineered. Co-immobilization of laccase and 2,2-binamine-di-3-ethylbenzothiazolin-6-sulfonic acid (ABTS) onto amino-functionalized ionic liquid-modified magnetic chitosan nanoparticles improves the capability of biocatalyst for the pollutant removal of bisphenol A, indole, and anthracene (Qiu et al., 2021). Functionalized multi-walled carbon nanotubes (CNTs) are used as nanocarriers for laccase immobilization to enhance the biocatalytic sustainability of laccase (Costa et al., 2019). In another work, laccase was also cross-linked onto hollow mesoporous carbon spheres (HMCs) for antibiotic degradation and removal from the aqueous phase (Shao et al., 2019). A biomimetic dynamic membrane (BDM) fabricated by using carbon nanotubes (CNTs) and laccases (Chen et al., 2019; Bilal et al., 2021), was proved to be very effective for wastewater treatment (Shao et al., 2019). Quite a few reviews have summarized the progress in nanoengineered laccases-advanced biotechnology (Bilal et al., 2021).</p><p>The support nanomaterials used for enzyme-immobilization are expected to be low cost and have a large enough surface area to avoid diffusion limitations of substrates and products of enzyme reaction. Meanwhile, the catalytic efficiency of the enzyme is anticipated to be improved by the immobilization to a solid surface. In general, enzymes are immobilized in various ways: binding to affinity labels, adsorption on substrates, and covalently anchoring to carriers (Shakerian et al., 2020). The immobilization model of laccase with its carriers greatly impacts on the properties of the enzyme (Bilal et al., 2017). At least, immobilization should not affect the conformation and activity of the enzyme, and the activity of immobilized enzymes should be retained for a longer time than that of free enzymes. The degradation of foreign biological compounds using immobilized enzymes may prove economical because of their enhanced stability and reusability. However, the immobilized enzyme cannot be recycled and reused too many times. Sometimes, intracellular enzymes do not work well in cell-free systems. Therefore, a group of nanomaterials was engineered to mimic the function of protein enzymes (Chen et al., 2019; Bilal et al., 2021), a relatively new strategy we are going to discuss in the next part.</p><!><p>Nanozymes are a class of nanomaterials that mimic and achieve the function of natural enzymes (Figure 1B) (Zhou et al., 2017). Nanozyme-based water treatment methods have many advantages over protein enzymes (Tian et al., 2020). Nanozymes are able to operate both at high and low pollutant concentrations, which reduces sludge generation. In addition, nanozymes can work catalytically to a wide range of pollutants with low energy inputs. Although protein enzyme has many advantages, it should be pointed out that it also has some challenges, such as the high catalyst cost, low reusability, and tendence to deactivation. Compared with protein enzymes, nanozymes are attractive for both applied and fundamental research. A nanomaterial based on guanosine monophosphate coordinated copper mimics the activity of laccase and converts a diverse range of phenol-containing substrates as laccase, including catechol, hydroquinone, epinephrine, and naphthol (Liang et al., 2017). While the cost of this nanomaterial is about 2400-fold lower than that of laccase, its stability is overwhelming. Normally, these laccase mimics are nanozymes formed by copper and biological molecules, such as guanosine monophosphate, dipeptide, guanine-rich ssDNA, and proteins (Wang et al., 2019; Rashtbari and Dehghan, 2021; Tran et al., 2021; Xu et al., 2021). In the last two years, inorganic nanozymes appeared in the form of Cu-base metal-organic framework (MOF) (Shams et al., 2019; Hu et al., 2021) or CuO nanorods (Alizadeh et al., 2020) for efficient dye detection and phenolic pollutant degradation. Apparently, inorganic nanozymes are more robust than the ones involving biomolecules. They are abiotic and much stable at high salt, high temperature, and extreme pH and could be stored for a very long time.</p><!><p>Polyoxometalates (POMs) are a kind of inorganic molecular materials with well-defined structures and mono-dispersity (Figure 2A). These nano-scaled clusters are composed of up to hundreds of face-, edge-, and/or angular shared bonds of metal-oxo polyhedral units. Most POMs can undergo reversible multi-electron redox reactions without structural change, which is a fantastic property in catalysis. Structures of POMs are rich and varying in morphisms and are versatile for co-assembly with other building blocks to construct diverse catalytic platforms. Some of them have nanopore or nanochannel structures with strong and/or selective affinity for guest molecules and/or ions, which makes them advantageous for size-selective catalysis. At the same time, POM surfaces are rich in oxygen, hydroxyl, and/or water ligands, suggesting that POM-based materials are excellent candidates for biomimetic applications. Over the last decades, POMs have demonstrated their promising biological activities, such as antitumoral (Bijelic et al., 2019), antimicrobial (Xu et al., 2020; Chen et al., 2021; Chen et al., 2022), insulin-sensitizing (Chen et al., 2020), immune-enhancing activities (Li et al., 2021; Li et al., 2022). In virtue of their structural diversity and physicochemical properties, POM-based nanomaterials exert oxidoreductase-mimicking activities, including oxidase, peroxidase (Chen K. et al., 2018; Jia et al., 2019), and catalase (Yadav and Singh, 2021). Compared with other nanomaterials, POMs are advantageous in their well-defined composition and molecular structure, as well as their economical production costs and mass production.</p><!><p>(A) Conventional types of polyoxometalates (M = Mo, W, V, Nb; X = heteroatom). (B) Scheme of degradation of phenolic contaminants by laccase-mimicking POMs. (C) Scheme of a batch reactor with POV hybrid as a quasi-homogeneous catalyst for degradation of emergent water contaminants. Scheme of laccase-like catalytic activity of POV hybrids for oxidating p-phenylenediamine (D) and hydroquinone (E).</p><!><p>In addition, POM-based nanomaterials have been successfully applied in wastewater treatment. An electrocatalyst was prepared with polyoxometalate (POM) as a molecular platform on a large scale by a one-step pyrolysis method and showed its prospects in directly usage in seawater (Ma et al., 2017). A series of Ln/Cu-POMs were used for wastewater treatment with rapid adsorption and excellent selective separation of cationic dyes from aqueous solutions (Yi et al., 2015). Efficient heterogeneous photocatalytic materials, nanosized and bimodal porous polyoxotungstate-anatase TiO2 composites, were prepared and exhibited visible-light photocatalytic activities in degrading organophosphorus pesticides in aqueous solution (Li et al., 2005). POM-based hybrid materials showed great potential for the removal of contaminants, such as phthalates and bisphenol A (Figure 2B), from wastewater (Zou et al., 2021; Huo et al., 2022). These POM-based hybrid materials showed high stability and long duration in either continuous or separated modes for effective water remediation (Galiano et al., 2021; Lai et al., 2021). In addition, a voltametric sensor was prepared with the aid of POMs for the determination of simazine in wastewater samples (Ertan et al., 2016). Such hybrid materials combined with POMs offer new insights for designing functional materials with low cost and high efficiency for wastewater treatment.</p><!><p>In recent years, POM has played a salient role in catalysis and showed its successful application in industrial catalysis, especially in the catalytic degradation of emerging contaminants in wastewater treatment. Structural diversity and excellent redox properties make POMs a vast treasure trove of active catalysts with intrinsic enzyme mimicking activities. The recent development of POM-based nanomaterials has created a new way for the development of artificial enzymes with high catalytic activity. Yet, people need to realize that nanozymes, including POMs, have not reached the catalytic activity as high as natural enzymes. There are several scientific and technologic challenges faced by POMs in catalytic degradation of endocrine disruptors. Since catalytic reactions involving nanozymes take place on the surface of the nanomaterials, surface modification represents an effective way to improve nanozyme activity. POMs are highly tailorable. Specifically, the capacity for POMs will be enlarged by the covalent or non-covalent interaction of POMs with a limitless range of organic moieties (Jia et al., 2017; Jia et al., 2018). POM organic derivatives have been shown to be able to assemble into a variety of hierarchical nanostructures applicable to different needs while maintaining their catalytic properties (Chen K. et al., 2018). Companied with low activity, the lack of reaction-specificity is another concern related to nanozymes, including POMs. POMs are able to catalyze the oxidation of a wide range of substrates and act as enzymes. However, the catalytic reactions involving nanozyme systems are typically more complex than natural enzymes. Laccase catalyzes substrate oxidation coupled to the four-electron reduction of molecular oxygen to water without releasing these partially reduced O2 products ROS. While the oxidation catalyzed by POMs may be coupled to the oxygen reduction to produce superoxides, •O2 − anions (Lai et al., 2021), in the real and complicated water environment. POMs are suitable for facile post-functionalization with other organic or inorganic molecules, which is an effective approach to design advanced catalytic materials. With rationale functionalization, POMs are expected to gain novel and improved physicochemical properties relevant to the development of novel catalysts for wastewater treatment in the near future.</p><!><p>Polyoxovanadates (POVs) are a subclass of POMs and have been described as bioinorganic drugs (Aureliano et al., 2021). POVs have shown different bioactivities not observed for monovanadate alone. Hybrid-type hexavanadate is one of the earliest organometallic POV derivatives that has been attracting research attention since its isolation. POV derivatives [V6 VO13{(OCH2)3CCH2OH}2]2− are redox stable and have been applied in homogeneous catalysis, materials science, and energy storage (Anyushin et al., 2020). Hybrid-type hexavanadates could be obtained through a simple, nontoxic, and one-pot method and showed favorable enzyme-like catalytic activity for oxidating phenylenediamine and hydroquinone (Figure 2). This oxidation could be conducted in a batch reactor with POV hybrid as a quasi-homogeneous catalyst (Figure 2C). [V6 VO13{(OCH2)3CCH2OH}2]2− is a versatile platform that can undergo DMAP-catalyzed esterification reactions with acid anhydrides to generate functional hybrid materials in catalysis. Following the etherate method for preparation and separation, solution stable POV hybrids were obtained and showed intrinsic laccase-like activities for catalyzing the oxidation of laccase substrate endocrine-disrupting p-phenylenediamine and hydroquinone to produce typical color changes (Figures 2D,E). These features POM-based hybrid catalysis as a potentially cost-effective approach for degradation of emergent water contaminants.</p><!><p>The original contributions presented in the study are included in the article, further inquiries can be directed to the corresponding author.</p><!><p>KC and SL participated in the preparation of the original draft. KC and QZ participated in revision and editing. All authors listed have made a substantial, direct, and intellectual contribution to the work and approved it for publication.</p><!><p>The work is supported financially by the National Natural Science Foundation of China (22101086) and the Natural Science Foundation of Guangdong Province (2021A1515010271).</p><!><p>The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.</p><!><p>All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.</p>

PubMed Open Access

Synthesis of Indole Analogues of the Natural Schweinfurthins

An interest in the schweinfurthins, natural stilbenes with significant anti-proliferative activity, has prompted efforts to prepare a set of indole analogues. To approach the desired compounds through a Horner-Wadsworth-Emmonds condensation, new indole derivatives bearing a phosphonomethyl substituent in the B-ring were required. The parent indole system with the necessary substitution pattern was obtained through a Stobbe condensation and cyclization. A prenyl substituent was incorporated at the C-3 position of a 4,6-disubstituted indole through a highly regioselective electrophilic aromatic substitution reaction, while metalation and alkylation provided the C-2 prenylated indole. After introduction of the phosphonate group through classical reactions, the new indole phosphonates were found to undergo the desired condensation with nonracemic aldehydes representing the schweinfurthin left half. This approach gives facile access to new heteroaromatic analogues of the natural schweinfurthins, and should be applicable to many other natural stilbenes as well.

synthesis_of_indole_analogues_of_the_natural_schweinfurthins

Introduction<!>Results and Discussion<!>General Experimental Procedures<!>General Procedure for Stobbe Condensations<!>4-Acetoxy-1H-indole-6-carboxylic acid ethyl ester (20) and 5-Acetoxy-indolizine-7-carboxylic acid ethyl ester (21)<!>Alternative route to indole 20 and indolizine 21<!>2-(4-Bromo-1H-pyrrol-2-ylmethylene)-succinic acid 1-ethyl ester (22)<!>5-Acetoxy-2-bromo-indolizine-7-carboxylic acid ethyl ester (24)<!>2-(4-Bromo-1-methoxymethyl-1H-pyrrol-2-ylmethylene)-succinic acid 1-ethyl ester (25)<!>4-Acetoxy-3-bromo-1-methoxymethyl-1H-indole-6-carboxylic acid ethyl ester (26)<!>2-(1-Methoxymethyl-1H-pyrrol-2-ylmethylene)-succinic acid 1-ethyl ester (27)<!>4-Acetoxy-1-methoxymethyl-1H-indole-6-carboxylic acid ethyl ester (28)<!>4-Hydroxy-1H-indole-6-carboxylic acid ethyl ester (29)<!>4-Methoxymethoxy-1-methoxymethyl-1H-indole-6-carboxylic acid ethyl ester (30) and 4-methoxymethoxy-1,5-bis-methoxymethyl-1H-indole-6-carboxylic acid ethyl ester (31)<!>(4-Methoxymethoxy-1-methoxymethyl-1H-indol-6-yl)-methanol (32)<!>(4-Methoxymethoxy-1-methoxymethyl-1H-indol-6-ylmethyl)-phosphonic acid diethyl ester (33)<!>5-Methoxymethoxy-7-[2-(4-methoxymethoxy-1-methoxymethyl-1H-indol-6-yl)-vinyl]-1,1,4a-trimethyl-(2R,4aR,9aR)-2,3,4,4a,9,9a-hexahydro-1H-xanthen-2-ol (34)<!>7-[2-(4-Hydroxy-1-methoxymethyl-1H-indol-6-yl)-vinyl]-1,1,4a-trimethyl-(2R,4aR,9aR)-2,3,4,4a,9,9a-hexahydro-1H-xanthene-2,5-diol (35)<!>Preparation of 4-Methoxymethoxy-1H-indole-6-carboxylic acid ethyl ester (36)<!>4-Methoxymethoxy-indole-1,6-dicarboxylic acid 1-tert-butyl ester, 6-ethyl ester (37)<!>6-Hydroxymethyl-4-methoxymethoxy-indole-1-carboxylic acid tert-butyl ester (38)<!>Preparation of (4-Methoxymethoxy-1H-indol-6-ylmethyl)-phosphonic acid diethyl ester (39) and 6-(Diethoxy-phosphorylmethyl)-4-methoxymethoxy-indole-1-carboxylic acid tert-butyl ester (40)<!>Boc protection of phosphonate 39<!>Preparation of phosphonate 40 at reduced temperature<!>[4-Methoxymethoxy-1-(toluene-4-sulfonyl)-1H-indol-6-yl]-methanol (41)<!>[4-Methoxymethoxy-1-(toluene-4-sulfonyl)-1H-indol-6-ylmethyl]-phosphonic acid diethyl ester (42)<!>Preparation of Toluene-4-sulfonic acid 5-methoxymethoxy-7-[2-(4-methoxymethoxy-1H-indol-6-yl)-vinyl]-1,1,4a-trimethyl-(2R,4aR,9aR)-2,3,4,4a,9,9a-hexahydro-1H-xanthen-2-yl ester (43) and 5-methoxymethoxy-7-[2-(4-methoxymethoxy-1H-indol-6-yl)-vinyl]-1,1,4a-trimethyl-(2R,4aR,9aR)-2,3,4,4a,9,9a-hexahydro-1H-xanthen-2-ol (44)<!>Reduction of tosylate 43<!>7-[ 2-(4-Hydroxy-1H-indol-6-yl)-vinyl]-1,1,4a-trimethyl-(2R,4aR,9aR)-2,3,4,4a,9,9a-hexahydro-1H-xanthene-2,5-diol (7)<!>Toluene-4-sulfonic acid 5-methoxy-7-[2-(4-methoxymethoxy-1H-indol-6-yl)-vinyl]-1,1,4a-trimethyl-(2R,4aR,9aR)-2,3,4,4a,9,9a-hexahydro-1H-xanthen-2-yl ester (45)<!>Toluene-4-sulfonic acid 5-methoxy-7-[2-(4-methoxymethoxy-1H-indol-6-yl)-vinyl]-1,1,4a-trimethyl-(2R,4aR,9aR)-2,3,4,4a,9,9a-hexahydro-1H-xanthen-2-yl ester (46)<!>6-[2-(7-Hydroxy-4-methoxy-8,8,10a-trimethyl-(5R,8aR,10aR)5,7,8,8a,9a,10a- hexahydro-6H-xanthen-2-yl)-vinyl]-1H-indol-4-ol (8)<!>4-Methoxymethoxy-3-(3-methyl-but-2-enyl)-1H-indole-6-carboxylic acid ethyl ester (47)<!>[4-Methoxymethoxy-3-(3-methyl-but-2-enyl)-1-(toluene-4-sulfonyl)-1H-indol-6-yl]-methanol (48)<!>[4-Methoxymethoxy-3-(3-methyl-but-2-enyl)-1-(toluene-4-sulfonyl)-1H-indol-6-ylmethyl]-phosphonic acid diethyl ester (49)<!>5-Methoxy-7-{2-[4-methoxymethoxy-3-(3-methyl-but-2-enyl)-1H-indol-6-yl]-vinyl}-1,1,4a-trimethyl-(2R,4aR,9aR)-2,3,4,4a,9,9a-hexahydro-1H-xanthen-2-ol (51)<!>6-[2-(7-Hydroxy-4-methoxy-8,8,10a-trimethyl-(5R,8aR,10aR)-5,7,8,8a,9,10a-hexahydro-6H-xanthen-2-yl)-vinyl]-3-(3-methyl-but-2-enyl)-1H-indol-4-ol (9)<!>6-(tert-Butyl-dimethyl-silanyloxymethyl)-4-methoxymethoxy-1-(toluene-4-sulfonyl)-1H-indole (52)<!>6-(tert-Butyl-dimethyl-silanyloxymethyl)-4-methoxymethoxy-2-(3-methyl-but-2-enyl)-1-(toluene-4-sulfonyl)-1H-indole (53)<!>[4-Methoxymethoxy-2-(3-methyl-but-2-enyl)-1-(toluene-4-sulfonyl)-1H-indol-6-yl]-methanol (54)<!>[4-Methoxymethoxy-2-(3-methyl-but-2-enyl)-1-(toluene-4-sulfonyl)-1H-indol-6-ylmethyl]-phosphonic acid diethyl ester (55)<!>5-Methoxy-7-{2-[4-methoxymethoxy-2-(3-methyl-but-2-enyl)-1H-indol-6-yl]-vinyl}-1,1,4a-trimethyl-(2R,4aR,9aR)-2,3,4,4a,9,9a-hexahydro-1H-xanthen-2-ol (57)<!>6-[2-(7-Hydroxy-4-methoxy-8,8,10a-trimethyl-(5R,8aR,10aR)-5,7,8,8a,9,10a-hexahydro-6H-xanthen-2-yl)-vinyl]-2-(3-methyl-but-2-enyl)-1H-indol-4-ol (10)

<p>The schweinfurthins (Figure 1), a small group of rare natural products,1,2 display a novel pattern of differential activity in the National Cancer Institute's (NCI) 60 cell line screen. Their activity pattern suggests that these compounds act on a novel target or through a new mechanism,1 and thus these compounds can be viewed as potential leads for further drug development. To alleviate the scarcity of these natural products, to access novel analogues, and to explore the limits of the pharmacophore, we have undertaken the synthesis of both natural schweinfurthins and a range of analogues.3–9 After an analysis of new compounds of potential interest, we considered the possibility of incorporating an indole in the stilbene system. The indole substructure is so common in both natural products and pharmaceutical agents that it often is considered a privileged scaffold.10,11 Incorporation of an indole motif might afford analogues with comparable or improved activity while at the same time increasing bioavailability.12,13 Furthermore, the D-ring resorcinol of the natural schweinfurthins may limit the schweinfurthins' stability, and proper placement of an indole system might improve the chemical stability as well. Based on this rationale, synthesis of indole analogues of the schweinfurthins became a goal of our program.</p><p>There are multiple ways that an indole moiety could be superimposed upon the D-ring of the natural schweinfurthins. The pattern pursued in this study would view the indole nitrogen as a replacement for one of the resorcinol oxygens, and incorporate the remainder of the indole ring as a substituent on the position para to the stilbene olefin (Figure 2). These structures would exploit the known flexibility of the para position to modification with preservation of biological activity.4,7,8 Furthermore, preparation of intermediates leading to structures 7 and 8 might be readily modified to allow addition of isoprenoid substituents to the 5-membered ring, via electrophilic aromatic substitution (which is favored at C-3 of indole itself14 and would lead to compound 9) or via anion chemistry (which can be directed to C-2 in N-substituted indoles and would provide compound 10).15,16 Because both compounds 9 and 10 represent modest deviations from the natural products in terms of the position of the prenyl group both series were of interest, and a strategy that could diverge to both isomers at a later stage would be particularly attractive.</p><p>Our foray into schweinfurthin studies began with synthesis of schweinfurthin C,17 and that early effort established the strategy of a late stage Horner-Wadsworth-Emmons (HWE) condensation for construction of the trans-stilbene olefin. To take advantage of intermediates already in hand from previous research, especially the now readily available R,R,R-aldehydes 11 and 12 that carry all of the schweinfurthin stereogenic centers (Scheme 1), would require an indole phosphonate such as compound 13. Given the vast number of known indoles it was somewhat surprising to find that apparently only C-218 and C-319 phosphonomethyl compounds have been prepared. Based on the assumption that phosphonate 13 could be prepared from the corresponding alcohol 14, which in turn should be available from the ester 15, routes to these two potential intermediates were considered. The presence of the "benzylic" alcohol of compound 14 might not be tolerated by many of the classical methods20 for de novo indole synthesis because of their reliance on acidic conditions, and the recent Kraus indole synthesis appears to be better suited for preparation of 2-substituted or 2,3-disubstituted compounds.21,22 However, preparation of the substituted indole 15 has been reported through an approach based on a Stobbe condensation of a succinate diester (17) and 2-pyrrole carboxaldehyde (18) followed by cyclization of the intermediate acid 16.23 While the initial report did not provide a complete characterization of the product, a more recent study from the Vedejs labs placed this approach on a solid foundation and proved that it does afford the desired substitution pattern.24 Therefore we began an effort to obtain the targeted schweinfurthin analogues with preparation of several indoles based on this strategy.</p><!><p>The Stobbe condensation of diethyl succinate with 2-pyrrole carboxaldehyde (18) smoothly gave the half ester 19 as expected.24 Without extensive purification, this material was treated with a mixture of acetic anhydride and acetic acid (6:1) in refluxing toluene to induce cyclization (Scheme 2). These conditions resulted in formation of the acetate-protected indole 20 (74%) accompanied by small amounts of the indolizine 21 (~1%), also as expected,24 while a parallel reaction in THF at reflux gave a less favorable product ratio (42% and 19%, respectively). Attempts to extend this approach to the brominated pyrrole 22, which might be useful for elaboration of the final products through halogen-metal exchange or cross coupling reactions,25 were more complex. While the desired half ester 22 was readily prepared by a Stobbe condensation, treatment of compound 22 under standard cyclization conditions gave only trace amounts of the desired indole 23 and afforded the indolizine 24 as the major product instead. Compound 24 is highly fluorescent and might be useful for synthesis of new types of fluorescent schweinfurthin analogues.26 However, for the immediate goal, N-protection of the pyrrole aldehyde would circumvent this issue as observed with N-methyl pyrrole.27 Because previous syntheses of schweinfurthin analogues employed MOM-protected phenols, the half ester 25 was prepared by Stobbe condensation of the MOM-protected aldehyde. In this case, cyclization under the standard conditions afforded only the desired indole product 26. In a similar sense, after the pyrrole 18 was protected as its MOM derivative 27, cyclization of the Stobbe product now gave only the desired indole 28. Because a late stage deprotection of the indole MOM group ultimately proved more difficult than expected (vide infra), pyrrole aldehyde 18 also was protected as its tosyl derivative. However, in this case attempted Stobbe condensation proved problematic, so introduction of this group at this stage of the sequence was not pursued further.</p><p>After hydrolysis of the acetate group of indole 20, treatment of the resulting phenol 29 with NaH and MOMCl in THF gave the desired MOM-protected indole 30 along with a significant amount of a C-alkylated product, tentatively assigned as the C-5 isomer 31 (Scheme 3). Addition of DMF to the solvent system improved the ratio of desired to undesired product from ~1.3:1 to ~9:1. Reduction of ester 30 proceeded in quantitative yield, but attempts at conversion to the phosphonate were somewhat frustrating. The reaction proceeded via the corresponding bromide, although the Arbuzov reaction of that bromide with (EtO)3P in refluxing toluene gave the desired phosphonate 33 in modest yield.</p><p>The HWE coupling of the hexahydroxanthene aldehyde 1228 with phosphonate 33 smoothly gave the protected analogue 34. Unfortunately, attempted hydrolysis of the three MOM groups by treatment with TsOH/MeOH gave compound 35, where both of the phenolic MOM groups had been cleaved but the indole nitrogen was still protected. Attempts to remove this remaining MOM group under more vigorous conditions29–31 proved unsuccessful, and gave only decomposition.</p><p>To circumvent this difficult hydrolysis, a new strategy based upon early formation of a differentially protected indole was explored. Selective MOM protection of the phenol 29 gave indole 36 (Scheme 4) and different N-protecting groups then could be introduced easily. For example, treatment of compound 36 with base and Boc2O gave the carbamate 37, and selective reduction of the ethyl ester gave the primary alcohol 38 in good yield. Under standard conditions for formation of the phosphonates (i.e. initial formation of the mesylate followed by treatment with LiBr and then neat (EtO)3P at reflux), formation of the C-P bond was accompanied by cleavage of the Boc group32 to afford phosphonate 39 as the major product. The Boc group was easily re-installed through treatment of phosphonate 39 with Boc2O to give phosphonate 40, or phosphonate 40 could be obtained more directly from the alcohol 38 in a reasonable yield (61%) if the Arbuzov reaction were conducted at a lower temperature (~95 °C) instead of reflux (~165 °C). Alternatively, a tosylate protecting group could be installed through treatment of indole 36 with TsCl and base, and the intermediate carboxylic acid ester was reduced selectively to the alcohol 41 in good yield. The tosyl group proved stable to standard conditions for formation of the phosphonate, and compound 42 was obtained smoothly.</p><p>Of the new indole phosphonates 39, 40, and 42, the HWE condensation of compound 39 with an aldehyde representing the schweinfurthin left half would be most advantageous because it would avoid an N-deprotection step of the product at a later stage. In the limited number of condensations between an indole phosphonate and an aldehyde, an N-protected indole always was employed.33–36 Nevertheless, because aldehyde 12 has been used in similar HWE reactions,3,6,13 condensations were attempted between this aldehyde and phosphonate 39. At best just trace amounts of a possible stilbene product were observed in this case, even though p-methoxybenzaldehyde reacted smoothly with phosphonate 39.37 Attempted condensation of aldehyde 12 with phosphonate 40 also was problematic. In this case, little or no condensation was observed and TLC analysis suggested that Boc cleavage had taken place instead. Fortunately, the HWE condensation of phosphonate 42 with aldehyde 12 at reflux gave a mixture of stilbene products in very good total yield (Scheme 5). Somewhat to our surprise, analysis of the 1H and 13C NMR spectra showed that the major product 43 carried a tosylate as an A-ring ester, while the minor product 44 did not have an A-ring tosylate, but already had undergone cleavage of the N-tosyl group. The hindered tosylate ester 43 proved resistant to standard hydrolysis,38–45 but reduction with LiAlH446,47 converted the major HWE product (43) to the minor product (44) in low yield. Final hydrolysis of the MOM groups gave the stilbene 7, the first schweinfurthin G analogue that incorporates an indole system.</p><p>To prepare the analogous schweinfurthin F analogue, phosphonate 42 was allowed to react with aldehyde 113 and base (Scheme 6). When the reaction was conducted at reflux in THF, the only stilbene product (56%) again reflected transfer of the tosyl group from the indole nitrogen to the A-ring alcohol. Treatment of this hindered tosylate ester with LiAlH4 did afford the free alcohol 46 in modest yield. Compound 46 undergoes hydrolysis of the phenolic MOM group under standard conditions to afford the schweinfurthin F analogue 8.</p><p>Because the natural schweinfurthins contain an isoprene chain as a D-ring substituent, installation of an isoprenoid chain on the indole would afford analogues more closely parallel to the natural products. Our original plan had been to incorporate this chain in a regiospecific manner through halogen-metal exchange on a protected indole derived from bromide 26, but this sequence would become unappealing if the MOM hydrolysis were problematic or the SN2′ product was formed during alkylation with prenyl bromide.48–52 An attractive alternative might be based on an extension of the methodology of Ganesan,53 which relies upon Zn(OTf)2 activation of an allylic halide to bring about only C-3 alkylation through electrophilic aromatic substitution. Among the attractive features of the original study, alkylation of indole itself with prenyl halides generally gave only the C-3 alkylated product, proceeded in ~60% yield, and did not give the products of SN2′ reaction (i.e. "reversed" prenyl substituents) that are frequently observed with other methods.48,49 However, it was unclear whether this approach could be applied to access the substituted indole required here, where both C-6 and C-4 groups that might impact reactivity were required. In particular, a C-6 ethoxycarbonyl group would add an electron withdrawing substituent system, while reduction of this group to the corresponding alcohol might invite polymerization reactions given the known reactivity of benzyl alcohol under these conditions.53 Furthermore, a MOM substituent at the C-4 position might compete with an isoprenoid halide for complexation with the Zn(OTf)2 or introduce a degree of steric hindrance to the C-3 position. Nevertheless, the brevity of this approach led us to study the process with indole 36. To our delight, the reaction of indole 36 with prenyl bromide in the presence of Zn(OTf)2 gave the desired product 47 in 65% yield (Scheme 7). This yield is comparable to those obtained on indole itself,53 despite the presence of the B-ring substituents.</p><p>Once ester 47 was in hand, the remaining steps in the sequence proceeded in a fashion parallel to those employed for preparation of the earlier analogues. Protection of the indole nitrogen as the tosylate proceeded smoothly. Then, after selective reduction of the carboxylic acid ester with DIBAL, the resulting alcohol 48 was readily converted to phosphonate 49. An HWE condensation with aldehyde 11 afforded a mixture of N-tosyl intermediate 50 and the free indole 51. After partial purification, treatment with NaH in a mixture of THF and i-PrOH afforded only compound 51. Final hydrolysis of the MOM group proceeded in low yield, but did afford the desired target compound, the schweinfurthin F analogue 9.</p><p>To access compound 10 from an intermediate already in hand, indole 41 was protected as its silyl ether 52, and then treated with n-BuLi and prenyl bromide. Despite the presence in the B-ring of two substituents that might participate in directed ortho metallation,54 this sequence gave a single product identified as the C-2 alkylated indole 53. After deprotection to the alcohol 54, and formation of the phosphonate 55 through standard reactions, condensation of phosphonate 55 with aldehyde 11 provided a mixture of the new stilbenes 56 and 57. After partial purification, treatment with 2-propanol and base completed conversion to compound 57, and final deprotection gave the desired schweinfurthin analogue 10.</p><p>In preliminary bioassays, compounds 7–10 were tested for their activity against the SF-295 cell line, which is one of those more sensitive to the natural schweinfurthins.1 These new schweinfurthin analogues did show activity in these assays, with EC50's ranging from ~200 nM to 2.5 μM (Table 1).37 Because the more active compounds show potency comparable to some of the natural schweinfurthins, preparation of additional indole analogues as well as more extensive testing in the 60 cell line assay of the National Cancer Institute would be warranted.</p><p>In conclusion, we have developed a strategy for synthesis of indole analogues of the natural schweinfurthins. This effort included preparation of several new indoles by cyclization after a Stobbe condensation, and ultimately led to preparation of the first indoles bearing a phosphonomethyl substituent in the indole B-ring. These B-ring phosphonates have been used in HWE reactions with the complex aldehydes 11 and 12, and undergo these condensations smoothly as long as the indole nitrogen is securely protected. With a tosyl group on the indole nitrogen, an unexpected transfer of the tosyl group to an unprotected alcohol was observed. While this transfer undoubtedly could be avoided through use of an alcohol protecting group, instead, because this transfer also deprotected the indole nitrogen, the tosylate ester was isolated and cleaved to the free alcohol, which allowed preparation of indole analogues of the schweinfurthin G and F cores. These studies also have shown that the Zn(OTf)2 mediated alkylation of a 4,6-disubstituted indole is a facile way to introduce a prenyl substituent to C-3 of the indole system, which in turn allowed preparation of a schweinfurthin F analogue complete with a side chain. In this more hindered prenyl indole, an HWE condensation at room temperature did afford the desired stilbene without transfer of the tosyl group, and reductive cleavage of the N-tosyl group was more efficient. Finally, a C-2 prenylated indole was obtained through metalation and alkylation of a tosyl indole intermediate, which allows divergent use of intermediate 35 to obtain either the C-2 or C-3 alkylated compounds. Together these studies have afforded four new indole analogues (7–10) of the natural schweinfurthins, and they define procedures that could be used to prepare analogues of many other natural stilbenes including resveratrol,55 the chiricanines,56 the arachidins and arahypins,57 and the pawhuskins.58 Further research on the biological activity of the new schweinfurthin analogues is underway, and will be reported in due course.</p><!><p>THF was freshly distilled from sodium/benzophenone, while CH2Cl2 and Et3N were freshly distilled from CaH2. All reactions in non-aqueous solvents were conducted in oven dried glassware under a positive pressure of argon with magnetic stirring. All commercial reagents were used without further purification unless otherwise stated. NMR spectra were recorded at 300 MHz for 1H, and 75 MHz for 13C or higher with CDCl3 as solvent and (CH3)4Si (1H, 0.00 ppm) or CDCl3 (13C, 77.0 ppm) as internal standards unless otherwise noted. High resolution mass spectra were run with magnet detection unless another method is noted. Elemental analyses were performed by a commercial facility.</p><!><p>According to the procedure of Vedejs24 but in THF (60 mL) instead of benzene, NaH (4.2 g, 105 mmol, 60% dispersion oil) was added slowly to aldehyde 18 (5.01 g, 52.6 mmol) and diethylsuccinate (13.3 mL, 80.2 mmol) at 0 °C. The reaction mixture was allowed to stir overnight and warm to rt. The reaction mixture was cooled to 0 °C, quenched by addition of water and Et2O was added and then extracted with 5% KOH. The combined aqueous layers were acidified with HCl (6 M) and extracted with Et2O. The combined organic extracts were washed with brine, dried (MgSO4), filtered, and the solvent was removed in vacuo to afford acid 19 (11.2 g, 96%) as a red-brown solid: 1H NMR ((CD3)2CO, 400 MHz) δ 10.83 (br s, 1H), 10.63 (br s, 1H), 7.75 (s, 1H), 7.07 – 7.06 (m, 1H), 6.61 – 6.59 (m, 1H), 6.30 – 6.27 (m, 1H), 4.20 (q, J = 7.1 Hz, 2H), 3.65 (s, 2H), 1.28 (t, J = 7.1 Hz, 3H); 13C NMR ((CD3)2CO, 400 MHz) δ 172.4, 168.5, 131.8, 128.8, 123.1, 119.2, 114.4, 111.9, 61.3, 34.4, 14.9; HRMS (TOF MS EI) m/z calcd for C11H13NO4 (M+) 223.0845, found 223.0851.</p><!><p>To acid 19 (17.1 g, 76.7 mmol) in toluene (800 mL) was added Ac2O (48 mL, 506 mmol) and glacial AcOH (4.62 mL, 80.5 mmol) and the reaction was heated to reflux. The next day the reaction mixture was allowed to cool to rt, quenched by addition of K2CO3 (sat), washed with brine, dried (MgSO4), and filtered, and the filtrate was concentrated in vacuo. Final purification by flash column chromatography (0 to 50% ethyl acetate in hexanes) afforded indole 20 (14.0 g, 74%) as a light brown solid and indolizine 21 (201 mg, 1%) as a yellow-brown oil. For indole 20: 1H NMR δ 8.98 (br s, 1H), 7.95 (s, 1H), 7.53 (s, 1H), 7.20 – 7.18 (m, 1H), 6.40 (m, 1H), 4.37 (q, J = 7.1 Hz, 2H), 2.40 (s, 3H), 1.38 (t, J = 7.2 Hz, 3H); 13C NMR 169.5, 167.0, 142.8, 136.7, 127.9, 124.8, 124.2, 112.6, 111.8, 99.4, 60.9, 20.9, 14.3. Anal. calcd. for C13H13NO4: C, 63.15; H; 5.30; N, 5.66. Found: C, 62.97; H, 5.31; N, 5.61.</p><p>For indolizine 21: 1H NMR δ 8.14 (s, 1H), 7.33 – 7.31 (m, 1H), 6.94 (d, J = 1.4 Hz, 1H), 6.90 (dd, J = 3.9, 2.8 Hz, 1H), 6.79 (dd, J = 4.0, 1.2 Hz, 1H), 4.36 (q, J = 7.2 Hz, 2H), 2.45 (s, 3H), 1.39 (t, J =7.2 Hz, 3H); 13C NMR 166.9, 165.6, 138.8, 133.4, 120.2, 119.2, 115.7, 110.5, 105.6, 99.0, 60.9, 20.6, 14.3; HRMS (TOF MS EI) m/z calcd for C13H13NO4 (M+) 247.0845, found 247.0849.</p><!><p>To acid 19 (1.00 g, 4.48 mmol) in THF was added Ac2O (5.4 mL, 57.5 mmol) and glacial AcOH (2.2 mL, 5.76 mmol) and the reaction mixture was heated to reflux. The next day the reaction mixture was allowed to cool to rt, poured into Et2O and water, washed with NaHCO3 (sat), dried (MgSO4), and filtered, and the filtrate was concentrated in vacuo. Final purification by flash column chromatography (15% to 50% Et2O in hexanes) afforded indole 20 (461 mg, 42%) and indolizine 21 (212 mg, 19%).</p><!><p>According to the general procedure, a solution of 4-bromo-2-pyrrolecarboxaldehyde (502 mg, 2.89 mmol) and diethyl succinate (0.72 mL, 4.29 mmol) in THF (4 mL) at 0 °C was treated with NaH (266 mg, 6.65 mmol, 60 % dispersion oil). Standard work-up and final purification by flash column chromatography (30% to 40% ethyl acetate in hexanes) afforded acid 22 (316 mg, 36%) as a light brown solid: 1H NMR ((CD3)2CO) δ 10.87 (br s, 1H), 7.67 (s, 1H), 7.14 (dd, J = 2.9, 1.4 Hz, 1H), 6.63 – 6.62 (m, 1H), 4.21 (q, J = 7.1 Hz, 2H), 3.65 (s, 2H), 1.28 (t, J = 7.1 Hz, 3H); 13C NMR ((CD3)2CO) δ 172.2, 167.9, 130.7, 129.3, 122.5, 121.5, 115.0, 98.8, 61.3, 34.1, 14.5; HRMS (TOF MS EI) m/z calcd for C11H12Br NO4 (M+) 300.9950, found 300.9954.</p><!><p>To acid 22 (811 mg, 2.68 mmol) in THF was added glacial AcOH (0.19 mL, 3.3 mmol), and Ac2O (3.2 mL 33.8 mmol) and the solution was heated at reflux overnight. The reaction mixture was then allowed to cool to rt, quenched by addition of Na2CO3 (sat), and extracted with ethyl acetate. The combined organic extracts were washed with water and brine, dried (MgSO4), and filtered, and the filtrate was concentrated in vacuo. Final purification by flash column chromatography (20% Et2O in hexanes) afforded indolizine 24 (687 mg, 79%): 1H NMR δ 8.00 (d, J = 1.4 Hz, 1H), 7.32 (dd, J = 1.5, 0.5 Hz, 1H), 6.96 (d, J = 1.4 Hz, 1H), 6.78 (d, J = 1.5 Hz, 1H), 4.35 (q, J = 7.1 Hz, 2H), 2.44 (s, 3H), 1.28 (t, J = 7.2 Hz, 3H); 13C NMR δ 166.6, 165.1, 138.1, 133.3, 120.6, 118.6, 110.3, 107.2, 105.4, 99.4, 61.2, 20.6, 14.3. Anal. calcd. for C13H12BrNO4: C, 47.88; H; 3.71; N, 4.29. Found: C, 48.10; H, 3.73; N, 4.22.</p><!><p>To 4-bromo-1H-pyrrole-2-carboxaldehyde (1.84 g, 10.6 mmol) in 10:1 THF/DMF (55 mL) at 0 °C was added NaH (525 mg, 7.5 mmol, 60% dispersion oil) and the reaction was allowed to stir for 5 min. To the resulting solution was added MOMCl (0.97 mL, 12.8 mmol) and the reaction was allowed to stir for 2 h and then quenched by addition of NH4Cl (sat), diluted with water, and extracted with Et2O. The combined organic extracts were washed with water and the brine, dried (MgSO4), and filtered, and the filtrate was concentrated in vacuo. Final purification by flash column chromatography (25% Et2O in hexanes) afforded the protected aldehyde (1.97 g, 86%) as a white solid: 1HMR δ 9.53 (d, J = 1.0 Hz, 1H), 7.13 (dd, J = 1.7, 1.0 Hz, 1H), 6.97 (d, J = 1.9 Hz, 1H), 5.62 (s, 2H), 3.31 (s, 3H); 13C NMR δ 179.0, 131.8, 130.1, 125.8, 98.0, 78.4, 56.3; HRMS (EI) m/z calcd for C7H8BrNO2 (M+) 216.9738, found 216.9740. According to the general procedure, the MOM-protected bromopyrrole aldehyde (1.01 g, 4.63 mmol) in THF (9 mL) at 0 °C was treated with diethyl succinate (1.2 mL, 1.54 mmol), followed by NaH (310 mg, 7.75 mmol). Standard work-up and final purification by flash column chromatography (25% to 40% ethyl acetate in hexanes) afforded acid 25 (425 mg, 27%) as a brown-yellow solid: 1H NMR δ 7.77 (s, 1H), 6.92 (d, J = 1.4 Hz, 1H), 6.61 (d, J = 1.1 Hz, 1H), 5.23 (s, 2H), 4.29 (q, J = 7.1 Hz, 2H), 3.68 (s, 2H), 3.26 (s, 3H), 1.33 (t, J = 7.1 Hz, 3H); 13C NMR 175.4, 167.6, 128.3, 128.2, 125.2, 122.1, 116.6, 97.9, 78.1, 61.5, 56.0, 34.0, 14.2. Anal. calcd for C13H16BrNO5: C, 45.10; H; 4.66; N, 4.05. Found: C, 45.19; H, 4.69; N, 3.93.</p><!><p>To acid 25 (1.084 g, 3.13 mmol) in Ac2O (20 mL) was added KOAc (0.49 g, 5.0 mmol) and the reaction was heated to reflux for 1 h and then allowed to cool to rt. The solution was diluted with ethyl acetate, washed with Na2CO3 (sat), water, and brine, dried (MgSO4), and filtered, and the filtrate was concentrated in vacuo. Final purification by flash column chromatography (20% ethyl acetate in hexanes) afforded indole 26 (911 mg, 79%) as a brown solid: 1H NMR δ 8.14 (d, J = 1.2 Hz, 1H), 7.56 (d, J = 1.2 Hz, 1H), 7.33 (s, 1H), 5.45 (s, 2H), 4.40 (q, J = 7.1 Hz, 2H), 3.28 (s, 3H), 2.43 (s, 3H), 1.41 (t, J = 7.1 Hz. 3H); 13C NMR δ 179.9, 166.2, 142.9, 137.1, 130.8, 126.1, 123.1, 115.2, 110.8, 88.2, 77.7, 61.2, 56.4, 21.0 14.4. Anal. calcd for C15H16BrNO5: C, 48.67; H, 4.36; N, 3.78. Found: C, 48.84; H, 4.60; N, 3.58.</p><!><p>A solution of N-MOM-2-pyrrolecarboxaldehyde (100 mg, 0.72 mmol) and diethyl succinate (145 mg, 0.84 mmol) in THF at 0 °C was treated with KOt-Bu (120 mg, 1.07 mmol). The solution was allowed to warm to rt overnight and the next day was heated to reflux for one h. The solution was cooled to 0 °C, quenched by addition of water, diluted with Et2O, and extracted with 5% KOH. The combined aqueous extracts were acidified (6M HCl) and extracted with Et2O. The combined organic layers were washed with brine, dried (MgSO4), and filtered, and then the filtrate was concentrated in vacuo. Final purification by flash column chromatograph (30% ethyl acetate in hexanes) afforded acid 27 (60 mg, 31%) as a yellow solid: 1H NMR δ 7.87 (s, 1H), 6.93 (dd, J = 2.7, 1.5 Hz, 1H), 6.67 – 6.66 (m, 1H)), 6.29 – 6.27 (m, 1H), 5.29 (s, 2H), 4.29 (q, J = 7.1 Hz, 2H), 3.72 (s, 2H), 3.25 (s, 3H), 1.33 (t, J = 7.1 Hz, 3H); 13C NMR δ 176.1, 168.1, 129.3, 127.7, 126.2, 120.0, 115.6, 110.1, 78.0, 61.3, 55.7, 34.2, 14.2. Anal. calcd for C13H17NO5: C, 58.42; H, 6.41. Found: C, 58.49; H, 6.43.</p><!><p>To acid 27 (333 mg, 1.25 mmol) in Ac2O (10 mL) was added KOAc (153 mg, 1.56 mol) and the solution was heated at reflux until the reaction was complete as judged by TLC analysis. The solution was allowed to cool to rt and then poured into NaHCO3 (sat) and diluted with Et2O. Once bubbling had ceased, the aqueous layer was extracted with Et2O and the combined organic extracts were washed with NaHCO3 (sat), water, and brine, dried (MgSO4), and filtered, and the filtrate was concentrated in vacuo. Final purification by flash column chromatography (40% ethyl acetate in hexanes) afforded indole 28 (298 mg, 82%) as a brown-yellow solid: 1H NMR δ 8.14 (dd, J = 1.0, 1.0 Hz, 1H), 7.60 (d, J = 1.1 Hz, 1H), 7.31 (d, J = 3.3 Hz, 1H), 6.46 (dd, J = 3.3, 0.8 Hz, 1H), 5.47 (s, 2H), 4.39 (q, J = 7.1 Hz, 2H), 3.25 (s, 3H), 2.40 (s, 3H), 1.41 (t, J = 7.1 Hz, 3H); 13C NMR δ 169.0, 166.6, 143.0, 137.2, 131.2, 125.9, 124.9, 113.4, 110.3, 99.7, 77.5, 60.9, 56.0, 22.0, 14.3; HRMS (TOF MS EI) m/z calcd for C15H17NO5 (M+) 291.1107, found 291.1104.</p><!><p>To a solution of acetate 20 (201 mg, 0.81 mmol) in EtOH (20 mL) was added K2CO3 (210 mg, 1.52 mmol) and the resulting mixture was heated to reflux for 90 min. The reaction mixture was cooled to 0 °C, filtered through celite, and then concentrated in vacuo. The resulting residue was dissolved in Et2O and extracted with 2N NaOH. The aqueous extracts were acidified and extracted with Et2O, dried (MgSO4), and filtered, and the filtrate was concentrated in vacuo. Final purification by flash column chromatography (50% Et2O in hexanes) afforded phenol 29 (136 mg, 82%) as a light brown solid: 1H NMR (CD3)2CO δ 10.5 (br s, 1H), 8.60 (br s, 1H), 7.76 (dd, J = 1.2, 1.2 Hz, 1H), 7.42 (dd, J = 3.2, 2.5 Hz, 1H), 7.18 (d, J = 1. 3 Hz, 1H), 6.67 (m, 1H), 4.32 (q, J = 7.1 Hz, 2H), 1.36 (t, J = 7.2 Hz, 3H); 13C NMR δ 167.8, 150.9, 138.2, 127.2, 125.5, 122.8, 107.0, 104.5, 100.1, 60.9, 14.7. Anal. calcd for C11H11NO3 : C, 64.38; H, 5.40; N, 6.83. Found: C, 64.39; H, 5.49; N, 6.66.</p><!><p>To a stirring suspension of NaH (800 mg, 20 mmol, 60% dispersion in oil) in a 6:1 mixture of THF and DMF (35 mL) at 0 °C was added indole 29 (1.61 g, 7.86 mmol) as a THF solution. Next MOMCl (1.5 mL, 20 mmol) was added dropwise and the reaction mixture was allowed to stir for 50 min. The reaction was quenched by addition of water and extracted with Et2O. The combined organic extracts were dried (MgSO4) and filtered, and the filtrate was concentrated in vacuo. Final purification by flash column chromatography (25 to 50% Et2O in hexanes) afforded indoles 30 (1.82 g, 79%) and 31 (227 mg, 9%). For compound 30: 1H NMR δ 7.94 (dd, J = 0.9, 0.9 Hz, 1H), 7.47 (d, J = 1.1 Hz, 1H), 7.25 (d, J = 3.3 Hz, 1H), 6.69 (dd, J = 3.2, 0.8 Hz, 1H), 5.47 (s, 2H), 5.38 (s, 2H), 4.40 (q, J = 7.2 Hz, 2H), 3.55 (s, 3H), 3.25 (s, 3H), 1.44 (t, J = 7.2 Hz, 3H); 13C NMR δ 167.3, 150.0, 137.1, 129.8, 125.4, 124.0, 106.8, 104.6, 100.2, 94.7, 77.4, 60.8, 56.2, 55.9, 14.4. Anal. calcd for C15H19NO5: C, 61.42; H, 6.53; Found: C, 61.59; H, 6.62. For compound 31: 1H NMR δ 7.82 (d J = 0.6 Hz, 1H), 7.25 (d, J = 3.3 Hz, 1H), 6.68 (dd, J = 3.3, 0.8 Hz, 1H), 5.46 (s, 2H), 5.28 (s, 2 H), 4.93 (s, 2H), 4.40 (q, J = 7.1 Hz, 2H), 3.66 (s, 3H), 3.39 (s, 3H), 3.22 (s, 3H), 1.42 (t, J = 7.1 Hz, 3H); 13C NMR δ 168.4, 150.0, 136.7, 130.1, 126.9, 124.6, 121.1, 109.1, 100.9, 99.5, 77.4, 65.7, 61.0, 58.0, 57.4, 56.0, 14.3. Anal. calcd for C17H23NO6: C, 60.52; H, 6.87; N, 4.15. Found: C, 60.40; H, 7.00; N, 4.00.</p><!><p>To ester 30 (668 mg, 2.28 mmol) in THF at 0 °C was added LiAlH4 (190 mg, 5.0 mmol) and the resulting mixture was allowed to stir for 2 h. The reaction mixture was then quenched by addition of water, acidified, and extracted with Et2O. The combined organic extracts were washed with water, dried (MgSO4), and filtered, and the filtrate was concentrated in vacuo. Final purification by flash column chromatography (50% ethyl acetate in hexanes) afforded alcohol 32 (566 mg, 99%) as a white solid: 1H NMR δ 7.17 (s, 1H), 7.09 (d, J = 3.3 Hz, 1H), 6.80 (d, J = 0.9 Hz, 1H), 6.63 (dd, J = 3.2, 0.7 Hz, 1H), 5.39 (s, 2H), 5.32 (s, 2 H), 4.75 (s, 2H), 3.53 (s, 3H), 3.22 (s, 3H), 2.02 (br s, 1H); 13C NMR δ 150.7, 137.9, 136.6, 127.3, 119.9, 103.7, 102.8, 99.8, 94.7, 77.5, 66.1, 56.1, 55.8; HRMS (EI) m/z calcd for C13H17NO4 (M+) 251.1158; found 251.1152.</p><!><p>To a solution of alcohol 32 (12 mg, 0.048 mmol) in CH2Cl2 (5 mL) at 0 °C was added Et3N (0.05 mL, 0.38 mmol) and MsCl (0.02 mL, 0.24 mmol) and the reaction was allowed to warm to rt. The following day the reaction was quenched by addition of NH4Cl (sat) and extracted with CH2Cl2. The combined organic extracts were washed with brine, dried (MgSO4), and filtered, and the filtrate was concentrated in vacuo. The resulting residue was dissolved in acetone (5 mL) at rt, LiBr (33 mg, 0.38 mmol) was added, and the reaction mixture was allowed to stir overnight. The following day the reaction mixture was poured into Et2O, quenched by addition of water, and extracted with Et2O. The combined organic extracts were washed with brine, dried (MgSO4), and filtered, and the filtrate was concentrated in vacuo. The resulting residue was dissolved in P(OEt)3 (0.5 mL) and toluene (3 mL) and the solution was heated at reflux overnight. The following day the solution was allowed to cool to rt, poured into Et2O, and then quenched by addition of water and extracted with Et2O. The combined organic extracts were washed with brine, dried (MgSO4), and filtered, and the filtrate was concentrated in vacuo. Final purification by flash column chromatography (80% ethyl acetate in hexanes) afforded phosphonate 33 (7 mg, 39% yield) as an oil: 1H NMR δ 7.12 (d, J = 3.2 Hz, 1H), 7.07 (dd, J = 3.2 Hz, 1.0 Hz, 1H), 6.75 (dd, J = 1.7, 1.3 Hz, 1H), 6.61 (dd, J = 3.2 Hz, 0.7 Hz, 1H), 5.40 (s, 2H), 5.32 (s, 2H), 4.06 – 3.96 (m, 4H), 3.53 (s, 3H), 3.25 (d, JHP = 21.3 Hz, 2H), 3.23 (s, 3H), 1.26 (td, J = 7.1 Hz, 0.3 Hz, 6H); 13C NMR δ 150.4 (d, JCP = 2.8 Hz), 138.0 (d, JCP = 3.0 Hz), 127.0 (d, JCP = 1.2 Hz), 126.4 (d, JCP = 9.2 Hz), 119.3 (d, JCP = 2.9 Hz), 106.5 (d, JCP = 5.9 Hz), 105.5 (d, JCP = 7.7 Hz), 99.7 (d, JCP = 1.5 Hz), 94.7, 77.4, 62.0 (d, JCP = 6.6 Hz, 2C), 56.1, 55.8, 34.2 (d, JCP = 138 Hz), 16.3 (d, JCP 6.1 Hz, 2C); 31P NMR δ 27.4; HRMS (EI) m/z calcd for C17H26NO6P (M+) 371.1498; found 371.1497.</p><!><p>To a suspension of NaH (45 mg, 1.13 mol, 60% dispersion in oil) in THF at 0 °C was added phosphonate 33 (37 mg, 0.10 mmol) as a THF solution followed by aldehyde 1228 (17.6 mg, 0.052 mmol) as a THF solution and the reaction was allowed to warm slowly to rt. The following day the reaction mixture was quenched by addition of water and extracted with ethyl acetate. The combined organic extracts were washed with brine, dried (MgSO4), and filtered, and the filtrate was concentrated in vacuo. Final purification by flash column chromatography (50 to 70% ethyl acetate in hexanes) afforded stilbene 34 (16 mg, 55%) as an oil: 1H NMR δ 7.24 (s, 1H), 7.16 (d, J = 1.9 Hz, 1H), 7.43 (d, J = 3.2 Hz, 1H), 7.03 – 6.97 (m, 4H), 6.62 (d, J = 3.2 Hz, 1H), 5.44 (s, 2H), 5.39 (s, 2H), 5.25 (d, J = 6.5 Hz, 1H), 5.21 (d, J = 6.6 Hz, 1H), 3.57, (s, 3H) 3.55 (s, 3H), 3.47 – 3.42 (m, 1H), 3.27 (s, 3H), 2.75 – 2.72 (m, 2H), 2.13 – 2.08 (m, 1H), 1.91 – 1.64 (m, 5H), 1.25 (s, 3H), 1.12 (s, 3H), 0.90 (s, 3H); 13C NMR δ 150.8, 146.2, 143.6, 138.2, 133.5, 129.5, 127.7, 127.0,125.5, 123.1, 121.9, 120.1, 113.4, 102.9, 102.5, 100.0, 95.9, 94.8, 78.0, 77.6, 76.9, 56.2, 56.2, 55.9, 46.8, 38.4, 37.7, 28.3, 27.3, 23.2, 19.9, 14.3; HRMS (EI) m/z calcd for C32H41NO7 (M+) 551.2883 found 551.2891.</p><!><p>To MOM-protected compound 34 (16 mg, 0.029 mmol) in MeOH (3 mL) was added TsOH (80 mg, 0.42 mmol) and the solution was allowed to stir at rt. The next day the solution was quenched by addition of NH4Cl (sat), diluted with water, and extracted with ethyl acetate. The combined organics extracts were washed with water, dried (MgSO4) and filtered, and the filtrate was concentrated in vacuo. Final purification by flash column chromatography (50% ethyl acetate in hexanes) afforded the schweinfurthin analogue 35 (9 mg, 67%) as a yellow oil: 1H NMR (CD3OD) δ 7.16 (d, J = 3.3 Hz, 1H), 7.11 (m, 1H), 6.98 (d, J = 16.0 Hz, 1H), 6.91 (d, J = 16.4 Hz, 1H), 6.86 (d, J = 1.9 Hz, 1H), 6.77 (d, J = 1.8 Hz, 1H), 6.73 (d, J = 1.0 Hz, 1H), 6.55 (dd, J = 3.3, 0.7 Hz, 1H), 5.47 (s, 2H), 3.40–3.35 (m, 1H), 3.25 (s, 3H), 2.75–2.71 (m, 2H), 2.09–2.04 (m, 1H), 1.85–1.63 (m, 4H), 1.24 (s, 3H), 1.11 (s, 3H), 0.89 (s, 3H); 13C NMR δ 151.6, 147.0, 142.1, 140.1, 134.9, 131.4, 128.6, 128.4, 128.0, 124.0, 120.3, 120.2, 111.1, 103.3, 102.2, 100.5, 78.8, 78.3, 78.2, 56.0, ~49 (obscured by solvent), 39.5, 38.9, 29.0, 27.9, 24.0, 20.3, 14.8; HRMS (EI) m/z calcd for C28H33NO5 (M+) 463.2359 found 463.2353.</p><!><p>To a suspension of phenol 29 (1.18 g, 5.74 mmol) in CH2Cl2 (100 mL) at rt was added DIPEA (4.0 mL, 23.0 mmol) and MOMCl (0.7 mL, 9.2 mmol) and the reaction mixture was allowed to stir overnight. The reaction was quenched by addition of water and extracted with CH2Cl2. The combined organic extracts were washed with brine, dried (MgSO4), and filtered, and the filtrate was concentrated in vacuo. Final purification by flash column chromatography (15 to 25% ethyl acetate in hexanes) afforded indole 36 (1.10 g, 77%) as a light yellow solid: 1H NMR δ 8.95 (br s, 1H), 7.89 (dd, J = 1.0, 1.0 Hz, 1H), 7.43 (d, J = 1.1 Hz, 1H), 7.26 (dd, J = 3.1, 2.5 Hz, 1H), 6.69 (m, 1H), 5.38 (s, 2H), 4.38 (q, J = 7.1 Hz, 2H), 3.54 (s, 3H), 1.38 (t, J = 7.2 Hz, 3H); 13C NMR δ 167.7, 149.9, 136.5, 126.4, 124.7, 123.0, 108.4, 103.8, 100.0, 94.7, 60.8, 56.2, 14.3. Anal. calcd for C13H15NO4: C, 62.64; H, 6.07; N, 5.62. Found: C, 62.83; H, 6.12; N, 5.42.</p><!><p>To a solution of indole 36 (1.00 g, 4.01 mmol) in THF (20 mL) at 0 °C was added NaH (200 mg, 5 mmol, 60% dispersion in oil) and Boc2O (960 mg, 4.40 mmol). An additional aliquot of THF was added (8 mL) and after 1 h the reaction mixture was quenched by addition of NH4Cl (sat) and extracted with ethyl acetate. The combined organic extracts were washed with brine, dried (MgSO4), and filtered, and the solvent was removed in vacuo. Final purification of the resulting material by flash column chromatography (12.5 to 15% Et2O in hexanes) afforded indole 37 (1.23 g, 87%): 1H NMR δ 8.54 (br s, 1H), 7.67 (d, J = 3.7 Hz, 1H), 7.57 (d, J = 1.2 Hz, 1H), 6.74 (dd, J = 3.7, 0.7 Hz, 1H), 5.36 (s, 2H), 4.39 (q, J = 7.1 Hz, 2H), 3.53 (s, 3H), 1.70 (s, 9H), 1.41 (t, J = 7.1 Hz, 3H) 13C NMR 167.0, 149.8, 149.4, 135.7, 127.5, 127.4, 125.4, 111.6, 107.5, 104.2, 94.8, 84.4, 60.9, 56.3, 28.1 (3C), 14.4. Anal. calcd for C18H23NO6: C, 61.88; H, 6.64; N, 4.01. Found: C, 62.00; H, 6.68; N, 4.02.</p><!><p>To ester 37 (434 mg, 1.24 mmol) in THF (30 mL) at 0 °C was added DIBAL (4.1 mL, 1M in THF). When judged complete by TLC analysis, the reaction was quenched by addition of NH4Cl (sat), poured into ethyl acetate, acidified, and then extracted with ethyl acetate. The combined organic extracts were washed with NaHCO3 (sat) and brine, dried (MgSO4), and filtered, and the filtrate was concentrated in vacuo. Final purification by flash column chromatography (25% ethyl acetate in hexanes) afforded alcohol 38 (345 mg, 91%) as a colorless oil: 1H NMR δ 7.84 (s, 1H), 7.48 (d, J = 3.8 Hz, 1H), 6.93 (d, J = 0.9 Hz, 1H), 6.67 (dd, J = 3.8, 0.7 Hz, 1H), 5.30 (s, 2H), 4.75 (s, 2H), 3.51 (s, 3H), 2.16 (br s, 1H), 1.66 (s, 9H) 13C 150.3, 149.7, 138.7, 136.6, 124.8, 121.0, 108.0, 106.3, 104.1, 94.7, 83.7, 66.0, 56.1, 28.1 (3C). Anal. calcd for C16H21NO5: C, 62.53; H, 6.89; N, 4.56. Found: C, 62.30; H, 7.13; N, 4.56.</p><!><p>To LiBr (450 mg, 5.18 mmol) and NEt3 (0.43 mL, 3.09 mmol) in THF at 0 °C was added the benzylic alcohol 38 (312 mg, 1.02 mmol) as a THF solution. The solution was stirred for 5 min and then MsCl (0.16 mL, 2.07 mmol) was added dropwise. The reaction mixture was allowed to stir for 1 h and more LiBr (400 mg, 4.61 mmol) was added. After the reaction was judged complete by TLC analysis it was quenched by addition of NaHCO3 (sat), diluted with water, and extracted with ethyl acetate. The combined organic extracts were washed with brine, dried (MgSO4), and filtered, and the filtrate was concentrated in vacuo. To the resulting residue was added P(OEt)3 (4 mL) and the solution was heated at reflux overnight. The next day the solution was allowed to cool to rt and then poured into water and extracted with ethyl acetate. The organic extracts were washed with brine, dried (MgSO4), and filtered, and the filtrate was concentrated in vacuo. Final purification by flash column chromatography (50 to 70% ethyl acetate in hexanes) afforded indole phosphonate 40 (18 mg, 4%) as an oil and the parent indole phosphonate 39 (194 mg, 58%) as an oil.</p><p>For phosphonate 39: 1H NMR δ 9.61 (s, 1H), 7.05 (d, J = 2.9 Hz, 1H), 6.99 (t, J = 2.3 Hz, 1H), 6.66 (s, 1H), 6.54, (t, J = 2.2 Hz, 1H), 5.29 (s, 2H), 4.44 – 3.96 (m, 4H), 3.50 (s, 3H), 3.21 (d, JPH = 21.1 Hz, 2H), 1.24 (t, J = 7.0 Hz, 6H); 13C NMR δ 150.2 (d, JCP = 2.7 Hz), 137.7 (d, JCP = 2.9 Hz), 124.8 (d, JCP = 9.4 Hz), 123.5, 118.2 (d, JCP = 2.7 Hz), 107.1 (d, JCP = 7.4 Hz), 105.6 (d, JCP = 5.8 Hz), 98.7, 94.7, 62.1 (d, JCP = 6.8 Hz, 2C), 55.9, 33.9 (d, JCP = 138 Hz), 16.2 (d, JCP = 6.1 Hz, 2C); 31P NMR δ 28.2; HRMS (EI) m/z calcd for C15H22NO5P (M+) 327.1236; found 327.1229.</p><!><p>To phosphonate 39 (194 mg, 0.593 mmol) in CH2Cl2 (10 mL) was added DMAP (8 mg, 0.065 mmol) and Boc2O (150 mg, 0.69 mmol). The reaction was allowed to stir for 2 h and then checked by TLC analysis. After an additional amount of Boc2O was added (50 mg, 0.23 mmol), the reaction was allowed to proceed for another hour. The reaction mixture was quenched by addition of water and extracted with CH2Cl2. The combined organic extracts were dried (MgSO4), and filtered, and the filtrate was concentrated in vacuo. Final purification by flash column chromatography (80% ethyl acetate in hexanes) afforded the Boc-protected indole 40 (183 mg, 72%) with 1H and 13C NMR spectra consistent with material prepared via the route below.</p><!><p>To alcohol 38 (147 mg, 0.48 mmol) in THF (10 mL) was added LiBr (250 mg, 2.9 mmol) and NEt3 (0.2 mL, 1.4 mmol), the solution was cooled to 0 °C, and then was allowed to stir. After 10 min, MsCl (0.08 mL, 2.07 mmol) was added dropwise and the reaction mixture was allowed to stir for 2 h. The reaction was then quenched by addition of NH4Cl (sat), diluted with water, and extracted with ethyl acetate. The combined organic extracts were dried (MgSO4) and filtered, and the filtrate was concentrated in vacuo. To the residue was added P(OEt)3 and the resulting solution was heated to 95 °C and allowed to stir overnight. The next day the solution was allowed to cool to rt, and then concentrated in vacuo. Final purification by flash column chromatography (1.5% EtOH in Et2O) afforded phosphonate 40 (125 mg, 61%) as an oil: 1H NMR δ 7.78 (br s, 1H), 7.48 (d, J = 3.5 Hz, 1H), 6.88 (m, 1H), 6.66 (d, J = 3.7 Hz, 1H), 5.30 (s, 2H), 4.09–4.00 (m, 4H), 3.51 (s, 3H), 3.26 (d, JPH = 21.6 Hz, 2H), 1.66, (s 9H), 1.27 (t, J = 7.1 Hz, 6H); 13C NMR δ 150.0 (d, JCP = 2.9 Hz), 149.6, 128.7 (d, JCP = 9.5 Hz), 124.6, 120.3, 110.7 (d, JCP = 7.9 Hz), 108.9 (d, JCP = 5.7 Hz), 104.0 (d, JCP =1.6 Hz), 94.7, 83.6, 62.0 (d, JCP = 6.6 Hz, 2C), 56.3, 34.3 (d, JCP = 138 Hz), 28.1 (3C), 16.3 (d, JCP = 6.3 Hz, 2C); 31P NMR δ 27.3; HRMS (EI) m/z calcd for C20H30NO7P (M+) 427.1760; found 427.1757</p><!><p>To indole 36 (805 mg, 3.23 mmol) in THF (30 mL) at 0 °C was added NaH (170 mg, 4.2 mmol, 60% dispersion in oil) followed after 10 min by TsCl (700 mg, 3.61 mmol). After 30 min, DIBAL (1.45 mL, 8.1 mmol) was added and the reaction was allowed to stir for an additional 30 min. It then was quenched by addition of NH4Cl (sat), poured into ethyl acetate, acidified, and extracted with ethyl acetate. The combined organic extracts were washed with brine, dried (MgSO4), and filtered, and the filtrate was concentrated in vacuo. Final purification by flash column chromatography (50% ethyl acetate in hexanes) afforded benzylic alcohol 41 (1.02 g, 87% overall yield): 1H NMR ((CD3)2CO) δ 7.84 (d, J = 8.3 Hz, 2H), 7.78 (s, 1H), 7.59 (d, J = 3.6 Hz, 1H), 7.22, (d, J = 8.5 Hz, 2H), 6.99 (s, 1H), 6.81, (dd, J = 3.7, 0.7 Hz, 1H), 5.27 (s, 2H), 4.78 (s, 2H), 4.53 (br s, 1H), 3.41 (s, 3H), 2.23 (s, 3H); 13C NMR δ 151.2, 146.0, 141.8, 136.9, 135.8, 130.7 (2C), 127.5 (2C), 126.0, 121.5, 107.2, 106.7, 105.9, 95.2, 65.0, 56.2, 21.3; HRMS (EI) m/z calcd for C18H19NO5S (M+) 361.0984; found 361.0992.</p><!><p>To alcohol 41 (118 mg, 0.33 mmol) in THF (10 mL) at 0 °C was added LiBr (226 mg, 2.62 mmol) and NEt3 (0.18 mL, 1.30 mmol). The reaction was allowed to stir for 5 min and then MsCl (0.06 mL, 0.78 mmol) was added dropwise. The reaction was allowed to warm to rt and after 3 h it was quenched by addition of NaHCO3 (sat) and extracted with ethyl acetate. The organic extracts were washed with brine, dried (MgSO4), and filtered, and the filtrate was concentrated in vacuo. The resulting residue was dissolved in P(OEt)3 (3 mL) and heated to reflux. The next day the reaction was allowed to cool to rt, poured into water, and extracted with ethyl acetate. The organic extracts were washed with brine, dried (MgSO4), and filtered, and the filtrate was concentrated in vacuo. Final purification by flash column chromatography (2.5 to 3% EtOH in Et2O) afforded phosphonate 42 (133 mg, 85%) as a white solid: 1H NMR δ 7.78 (d, J = 8.3 Hz, 2H), 7.62 (d, J = 2.8 Hz, 1H), 7.44 (dd, J = 3.7, 0.9 Hz, 1H), 7.22, (d, J = 8.0 Hz, 2H), 6.86 (m, 1H), 6.73 (d, J = 3.7 Hz, 1H), 5.25 (s, 2H), 4.05 – 3.95 (m, 4H), 3.47 (s, 3H), 3.25 (d, JPH = 21.5 Hz, 2H), 2.33 (s, 3H), 1.24 (t, J = 7.1 Hz, 6H); 13C NMR δ 150.2 (d, JCP = 2.9 Hz), 144.8, 136.1 (d, JCP = 3.1 Hz), 135.1, 129.7 (2C), 129.3 (d, JCP = 9.2 Hz), 126.8 (2C), 125.0 (d, JCP = 1.4 Hz), 120.6 (d, JCP = 3.1 Hz), 109.3 (d, JCP = 6.0 Hz), 108.6 (d, JCP = 7.5 Hz), 105.8 (d, JCP = 1.5 Hz), 94.6, 62.0 (d, JCP = 6.7 Hz, 2C), 56.2, 34.2 (d, JCP = 138.1 Hz), 21.5, 16.3 (d, JCP = 6.1 Hz, 2C); 31P NMR δ 27.3; HRMS (EI) m/z calcd for C22H28NO7PS (M+) 481.1324; found 481.1315.</p><!><p>To phosphonate 42 (40 mg, 0.83 mmol) and aldehyde 1228 (18 mg, 0.54 mmol) in THF (3 mL) at rt was added NaH (60 mg, 1.5 mmol, 60% dispersion in oil) and 15-crown-5 (3 drops) and the resulting solution was heated to reflux. After 30 min the reaction mixture was allowed to cool to rt and quenched by addition of NH4Cl (sat), diluted with water, and extracted with Et2O. The combined organic extracts were washed with brine, dried (MgSO4), and filtered, and the filtrate was concentrated in vacuo. Final purification by flash column chromatography (20 to 40% ethyl acetate in hexanes) afforded the tosylate 43 (24 mg, 67%) along with the alcohol 44 (5 mg, 22%). For tosylate 43: 1H NMR δ 8.24 (br s, 1H), 7.82 (d, J = 8.3 Hz, 2H), 7.34 (d, J = 8.0 Hz, 2H) 7.14 – 7.11 (m, 3H), 6.98 – 6.92 (m, 4H), 6.63 (m, 1H), 5.38 (s, 2H), 5.23 (d, J = 6.6 Hz, 1H), 5.19 (d, J = 6.6 Hz, 1H), 4.33 (dd, J = 10.6, 4.8 Hz, 1H), 3.57 (s, 3H), 3.53 (s, 3H), 2.69–2.66 (m, 2H), 2.45 (s, 3H), 2.10–2.04 (m, 1H), 1.82 – 1.60 (m, 4H), 1.22 (s, 3H), 0.91 (s, 3H), 0.90 (s, 3H); 13C NMR δ 150.8, 146.1, 144.7, 143.3, 137.7, 134.3, 133.1, 129.8 (3C), 127.9, 127.7 (2C), 126.5, 123.5, 122.6, 121.7, 119.1, 113.4, 104.1, 101.9, 100.1, 95.9, 94.8, 88.4, 76.0, 56.2, 56.2, 47.0, 38.2, 37.4, 27.0, 25.8, 23.1, 21.6, 19.8, 15.1; HRMS (TOF MS ES) m/z calcd for C37H44NO8S ((M+H)+) 662.2788; found 662.2797.</p><p>For alcohol 44: 1H NMR δ 8.30 (br s, 1H), 7.15 – 7.11 (m, 3H), 7.05 – 6.92 (m, 4H), 6.64 (m, 1H), 5.39 (s, 2H), 5.24 (d, J = 6.4 Hz, 1H), 5.20 (d, J = 6.5 Hz, 1H), 3.57 (s, 3H), 3.35 (s, 3H), 3.43 (dd, J = 11.5, 3.8 Hz, 1H), 2.75 – 2.71 (m, 2H), 2.11 – 2.04 (m, 1H), 1.90–1.54 (m, 5H), 1.25 (s, 3H), 1.11 (s, 3H), 0.89 (s, 3H); 13C NMR δ 150.8, 146.1, 143.6, 137.7, 133.2, 129.6, 127.8, 126.7, 123.5, 123.2, 121.9, 119.1, 113.5, 104.1, 102.0, 100.1, 96.0, 94.8, 78.0, 76.9. 56.2, 56.2, 46.8, 38.4, 37.7, 28.3, 27.3, 23.2, 19.9, 14.2; HRMS (EI) m/z calcd for C30H37NO6 (M+) 507.2621; found 507.2620.</p><!><p>To the MOM-protected tosylate 43 (19.0 mg, 0.03 mmol) in THF (3 mL) at 0 °C was added LiAlH4 (14 mg, 0.40 mmol) and the reaction mixture was allowed to warm to rt overnight. The following morning the reaction was quenched by addition of NH4Cl (sat), diluted with water, and extracted with Et2O. The combined organic layers were washed with brine, dried (MgSO4) and filtered, and the solvent was removed in vacuo. Final purification by preparative TLC (70% ethyl acetate in hexanes) afforded the desired indole 44 (4.4 mg, 30%) along with recovered starting material (2.7 mg, 14%). The 1H NMR spectra was consistent with that of material prepared above.</p><!><p>To a methanol solution of protected indole 44 (6 mg, 0.012 mmol) at 0 °C was added TsOH (25 mg, 0.145 mmol). The reaction was allowed to stir overnight, then quenched by addition of water and extracted with ethyl acetate. The combined organic extracts were dried (Mg2SO4), filtered, and concentrated in vacuo. Final purification of the residue by preparative TLC (70% ethyl acetate in hexanes) afforded schweinfurthin analogue 7 (2.9 mg, 58%):1H NMR (CD3OD) δ 7.09 (d, J = 3.3 Hz, 1H), 7.00 (s, 1H), 6.95 (d, J = 16.2 Hz, 1H), 6.87 (d, J = 16.2 Hz, 1H), 6.84 (d, J = 1.6 Hz, 1H), 6.75 (d, J = 1.6 Hz, 1H), 6.66, (d, J = 1.0 Hz, 1H), 6.50 (dd, J = 3.2, 0.9 Hz, 1H), 3.43 (dd, J = 11.5. 3.8 Hz, 1H), 2.74 – 2.71 (m, 2H), 2.09 – 2.04 (m, 1H), 1.83 – 1.63 (m, 4H), 1.24 (s, 3H), 1.11 (s, 3H), 0.89 (s, 3H); 13C NMR δ 151.2, 147.0, 141.9, 139.8, 133.9, 131.5, 128.9, 127.2, 124.4, 124.0, 120.2, 119.3, 111.0, 103.8, 101.8, 99.7, 78.8, 78.2, 39.5, 38.9, 29.0, 27.9, 24.0, 20.3, 14.9; HRMS (EI) m/z calcd for C26H29NO4 (M+) 419.2097; found 419.2096.</p><!><p>To aldehyde 113,28 (63 mg, 0.21 mmol) and phosphonate 42 (156 mg, 0.32 mmol in THF (5 mL) at rt was added NaH (80 mg, 2.0 mmol, 60% dispersion in oil) and 15-crown-5 (3 drops). The reaction mixture was slowly heated to reflux for 40 min and then allowed to cool to rt. After the reaction was quenched by addition of NaHCO3 (sat), it was diluted with water, and extracted with ethyl acetate. The combined organic extracts were washed with brine, dried (MgSO4), and filtered, and the filtrate was concentrated in vacuo. Final purification by flash column chromatography (30% ethyl acetate in hexanes) afforded the tosylate 45 (73 mg, 56%): 1H NMR δ 8.25 (br s, 1H), 7.82 (d, J = 8.3 Hz, 2H), 7.34 (d, J = 8.0 Hz, 2H), 7.14 (s, 1H), 7.12 (dd, J = 3.2, 2.4 Hz, 1H), 7.03 (d, J = 16.2 Hz, 1H), 6.99 (d, J = 1.1 Hz, 1H), 6.95 (d, J = 16.3 Hz, 1H), 6.90 (d, J = 1.6 Hz, 1H), 6.83 (d, J = 1.6 Hz, 1H), 6.65 – 6.63 (m, 1H), 5.39 (s, 2H), 4.36 – 4.31 (m, 1H), 3.89 (s, 3H), 3.57 (s, 3H), 2.70 – 2.67 (m, 2H), 2.45 (s, 3H), 2.14 – 2.09 (m, 1H), 2.01 – 1.96 (m, 1H), 1.87 – 1.68 (m, 3H), 1.56 (br s, 1H), 1.23 (s, 3H), 0.91 (m, 6H); 13C δ 150.8, 148.9, 144.6, 142.0, 137.7, 134.3, 133.1, 129.8 (2C), 129.6, 127.8, 127.7 (2C), 126.8, 123.6, 122.0, 120.1, 119.2, 107.0, 104.0, 102.0, 100.1, 94.8, 88.5, 76.0, 56.2, 56.0, 47.0, 38.2, 37.3, 27.1, 25.7, 23.1, 21.6, 19.7, 15.1; HRMS (TOF MS ES) m/z calcd for C36H42NO7S ((M+H)+) 632.2682; found 632.2684.</p><!><p>To the tosylate 45 (73 mg, 0.12 mmol) in THF (3 mL) was added LiAlH4 (45 mg, 1.18 mmol) and the reaction mixture was allowed to stir overnight. The reaction then was quenched by addition of NH4Cl (sat) and extracted with Et2O. The combined organic layers were washed with brine, dried (MgSO4), and filtered, and the filtrate was concentrated in vacuo. Final purification by flash column chromatography (30 to 50% ethyl acetate in hexanes) yielded alcohol 46 (24 mg, 43%): 1H NMR δ 8.25 (br s, 1H), 7.15 (s, 1H), 7.12 (dd, J = 3.1, 2.5 Hz, 1H), 7.04 (d, J = 16.2 Hz, 1H), 7.00 (s, 1H), 6.97 (d, J = 16.2 Hz, 1H), 6.91 (d, J = 2.3 Hz, 1H), 6.88 (d, J = 2.3 Hz, 1H), 6.63 (m, 1H), 5.39 (s, 2H), 3.90 (s, 3H), 3.58 (s, 3H), 3.45 – 3.40 (m, 1H), 2.74 – 2.71 (m, 2H), 2.15 – 2.10 (m, 1H), 1.90 – 1.80 (m, 2H), 1.74 – 1.50 (m, 3H), 1.26 (s, 3H), 1.11 (s, 3H), 0.89 (s, 3H); 13C NMR δ 150.8, 148.9, 142.3, 137.7, 133.2, 129.4, 127.6, 127.0, 123.5, 122.6, 120.2, 119.1, 106.9, 104.0, 102.0, 100.1, 94.8, 78.0, 77.0, 56.2, 56.0, 46.8, 38.4, 37.7, 28.3, 27.3, 23.2, 19.9, 14.3; HRMS (EI) m/z calcd for C29H35NO5 (M+) 477.2515; found 477.2512.</p><!><p>To the MOM-protected indole 46 (16.0 mg, 0.033 mmol) in MeOH (3 mL) was added HCl (0.15 mL, 6M). The reaction was stirred in a warm water bath for 8.5 h, quenched by dropwise addition of NaHCO3 (sat), and then extracted with Et2O. The combined organic extracts were washed with brine, dried (MgSO4), and filtered through basic alumina, and the filtrate was concentrated in vacuo. Final purification by preparative TLC (70% ethyl acetate in hexanes) afforded indole 8 (9 mg, 62%); 1H NMR δ 8.2 (br s, 1H), 7.13 (dd, J = 3.1, 2.5 Hz, 1H), 7.07 (s, 1H), 7.00 (d, J = 16.2 Hz, 1H), 6.94 (d, J = 16.4 Hz, 1H), 6.90 (d, J = 1.8 Hz, 1H), 6.85 (d, J = 1.7 Hz, 1H), 6.77 (d, J = 0.9 Hz, 1H), 6.59 – 6.57 (m, 1H), 5.22 (br s, 1H), 3.90 (s, 3H), 3.43 (dd, J = 11.5, 3.7 Hz, 1H), 2.75 – 2.72 (m, 2H), 2.16 – 2.10 (m, 1H), 1.90 – 1.80 (m, 2H), 1.75 – 1.60 (m, 3H), 1.26 (s, 3H), 1.11 (s, 3H), 0.89 (s, 3H); 13C δ 149.0, 148.9, 142.4, 138.0, 133.4, 129.4, 127.3, 127.2, 123.5, 122.7, 120.4, 117.4, 106.9, 103.1, 102.1, 99.2, 78.1, 77.0, 56.0, 46.8, 38.4, 37.6, 28.3, 27.4, 23.2, 19.9, 14.3; HRMS (EI) m/z calcd for C27H31NO4 (M+) 433.2253; found 433.2245.</p><!><p>To indole 36 (1.00 g, 4.01 mmol), TBAI (739 mg, 2.00 mmol), and Zn(OTf)2 (878 mg, 2.41 mmol) in a 9:2 mixture of toluene and CH2Cl2 (22 mL) at rt was added DIPEA (0.77 mL, 4.41 mmol). After the reaction mixture was allowed to stir for 10 min, prenyl bromide (298 mg, 2.00 mmol) was added dropwise. After 3 h the reaction mixture was quenched by addition of NH4Cl (sat) and extracted with ethyl acetate. The combined organic extracts were washed with water, dried (MgSO4), and filtered, and the filtrate was concentrated in vacuo. Final purification by flash column chromatography (10 to 15% ethyl acetate in hexanes) afforded prenylated indole 47 (415 mg, 65%) along with recovered starting material 36 (540 mg): 1H NMR δ 8.47 (br s, 1H), 7.79 (d, J = 1.2 Hz, 1H), 7.34 (d, J = 1.1 Hz, 1H), 6.96 (m, 1H), 5.46 (m, 1H), 5.35 (s, 2H), 4.37 (q, J = 7.1 Hz, 2H), 3.65 (d, J = 6.6 Hz, 2H), 3.53 (s, 3H), 1.74 (d, J = 1.0 Hz, 3H), 1.72 (s, 3H), 1.38 (t, J = 7.1 Hz, 3H); 13C δ 167.6, 151.4, 137.4, 131.5, 124.6, 123.8, 123.7, 121.3, 116.7, 108.2, 102.8, 94.2, 60.7, 56.2, 25.7, 25.4, 17.7, 14.4; HRMS (EI) m/z calcd for C18H23NO4 (M+) 317.1627; found 317.1631.</p><!><p>To indole 47 (315 mg, 0.99 mmol) in THF at 0 °C was added NaH (50 mg, 1.25 mmol, 60% dispersion oil) and the reaction mixture was allowed to stir for 10 min. After TsCl (230 mg, 1.21 mmol) was added, the solution was stirred for 30 min and then DIBAL (0.71 mL, 4.0 mmol) was added dropwise. After an additional 30 min the reaction was quenched by addition of NH4Cl (sat), acidified with HCl, and extracted with ethyl acetate. The combined organic extracts were washed with Na2CO3 (sat) and brine, dried (MgSO4), and filtered, and the filtrate was concentrated in vacuo. Purification by flash column chromatography (34% ethyl acetate in hexanes) afforded benzylic alcohol 48 (348 mg, 82%): 1H NMR δ 7.71 (d, J = 8.4 Hz, 2H), 7.60 (s, 1H), 7.16 (d, J = 8.2 Hz, 2H), 7.13 (m, 1H), 6.85 (d, J = 0.6 Hz, 1H), 5.41 – 5.39 (m, 1H), 5.22 (s, 2H), 4.71 (s, 2H), 3.51 (d, J = 7.1 Hz, 2H) 3.46 (s, 3H), 2.37 (br s, 1H), 2.30 (s, 3H), 1.76 (d, J = 0.8 Hz, 3H), 1.68 (s, 3H); 13C NMR δ 151.8, 144.6, 139.1, 137.0, 135.2, 132.9, 129.7 (2C), 126.6 (2C), 122.7, 121.9, 121.8, 120.2, 105.9, 105.7, 94.1, 65.5, 56.1, 25.7, 25.6, 21.4, 17.7; HRMS (EI) m/z calcd for C23H27NO5S (M+) 429.1610; found 429.1609.</p><!><p>To alcohol 48 (332 mg, 0.77 mmol) in THF (15 mL) at 0 °C was added LiBr (537 mg, 6.18 mmol) and NEt3 (0.43 mL, 3.09 mmol). The solution was stirred for 5 min and then MsCl (0.18 mL, 2.32 mmol) was added dropwise. The reaction was allowed to warm to rt, and after 2 h it was quenched by addition of NaHCO3 (sat.) and extracted with ethyl acetate. The combined organic extracts were washed with brine, dried (MgSO4), and filtered, and the filtrate was concentrated in vacuo. Without further purification, the resulting residue was dissolved in P(OEt)3 (3 mL) and heated to reflux. The next day the solution was allowed to cool to rt and then poured into water and extracted with ethyl acetate. The organic extracts were washed with brine, dried (MgSO4), and concentrated in vacuo. Final purification by flash column chromatography (2% EtOH in Et2O) afforded indole phosphonate 49 (374 mg, 88%) as a waxy white solid: 1H NMR δ 7.75 (d, J = 8.4 Hz, 2H), 7.57 (m, 1H), 7.21 (d, J = 8.1 Hz, 2H), 7.10, (d, J = 1.1 Hz, 1H), 6.80 (m, 1H), 5.41 – 5.36 (m, 1H), 5.23 (s, 2H), 4.00 (m, 4H), 3.51 – 3.47 (m, 5H), 3.22 (d, JPH = 21.5 Hz, 2H), 2.33 (s, 3H), 1.77 (s 3H), 1.68 (s, 3H), 1.25 (t, J = 7.0 Hz, 6H); 13C NMR δ 151.6 (d, JCP = 2.9 Hz) 144.8, 137.1 (d, JCP = 3.1 Hz), 135.4, 133.0, 129.7 (2C), 129.2 (d, JCP = 9.3 Hz), 126.8 (2C), 122.7 (d, JCP = 1.6 Hz), 121.8, 121.7 (d, JCP = 1.8 Hz), 119.7 (d, JCP = 3.2 Hz), 108.9 (d, JCP = 5.9 Hz), 108.7 (d, JCP = 7.6 Hz), 94.3, 62.1 (d, JCP = 6.7 Hz, 2C), 56.1, 34.2 (d, JCP = 138.3 Hz), 25.7, 25.6, 21.4, 17.7, 16.3 (d, JCP = 6.0 Hz, 2C); 31P NMR δ 26.9; HRMS (EI) m/z calcd for C27H36NO7PS (M+) 549.1950; found 549.1959.</p><!><p>To aldehyde 11 (44 mg, 0.15 mmol) and phosphonate 49 (100 mg, 0.18 mmol) in THF (4 mL) at 0 °C was added NaH (80 mg, 2.0 mmol, 60% dispersion oil) and 15-crown-5 (2 drops), and the reaction mixture was allowed warm to rt. After 2 hrs it was quenched by addition of NH4Cl (sat) and extracted with ethyl acetate. The combined organic layers were washed with brine, dried (MgSO4), and filtered, and the filtrate was concentrated in vacuo. Purification by flash column chromatography (50% ethyl acetate in hexanes) afforded a mixture of the N-Ts compound 50 and free indole 51 (55 mg) as an oil. This material was dissolved in a mixture of THF and i-PrOH (5 mL, 1:1 mixture) at 0 °C, NaH (150 mg, excess) was added, and the reaction mixture was allowed to warm to rt. The following day the reaction mixture was quenched by addition of water and extracted with ethyl acetate. The combined organic extracts were washed with brine, dried (MgSO4), and filtered, and the filtrate was concentrated in vacuo. Final purification by flash column chromatography (50% ethyl acetate in hexanes) afforded indole 51 (35 mg, 44% for 2 steps) as an oil: 1H NMR δ 7.95 (br s, 1H), 7.07 (s, 1H), 6.99 – 6.98 (m, 2H), 6.92 – 6.90 (m, 2H), 6.87 (m, 1H), 6.81 (s, 1H), 5.51 – 5.46 (m, 1H), 5.36 (s, 2H), 3.90 (s, 3H), 3.62 (d, J = 7.0 Hz, 2H), 3.57 (s, 3H), 3.43 (dd, J = 11.6, 3.8 Hz, 1H), 2.74 – 2.71 (m, 2H), 2.15 – 2.10 (m, 1H), 1.89 – 1.56 (m, 11H), 1.26 (s, 3H), 1.11 (s, 3H), 0.89 (s, 3H); 13C NMR δ 152.1, 148.9, 142.3, 138.6, 133.0, 131.2, 129.4, 127.6, 126.8, 124.1, 122.6, 120.9, 120.2, 117.5, 116.7, 106.9, 103.8, 100.9, 94.3, 78.0, 77.0, 56.1, 56.0, 46.8, 38.4, 37.7, 28.3, 27.3, 25.7, 25.6, 23.2, 19.8, 17.7, 14.3; HRMS (EI) m/z calcd for C34H43NO5 (M+) 545.3141; found 545.3135.</p><!><p>To compound 51 (31 mg, 0.057 mmol) in MeOH (2 mL) at rt was added TsOH (75 mg, 0.39 mmol) and the reaction flask was wrapped in foil. After 10 h the reaction was quenched by addition to NaHCO3 (sat) and extracted with ethyl acetate. The combined organic extracts were washed with Na2CO3 (sat), brine, and dried (MgSO4), filtered, and the filtrate was concentrated in vacuo. Final purification by flash column chromatography (50% ethyl acetate in hexanes) afforded stilbene 9 (8 mg, 28%) as a light yellow oil: 1H NMR δ 7.90 (br s, 1H), 6.99 – 6.96 (m, 3H), 6.89 – 6.85 (m, 3H), 6.74 (s, 1H), 5.91 (br s, 1H), 5.54 (m, 1H), 3.90 (s, 3H), 3.58 (d, J = 6.6 Hz, 2H), 3.44 (dd, J = 11.6, 3.7 Hz, 1H), 2.75 – 2.72 (m, 2H), 2.16 – 2.10 (m, 1H), 1.90 – 1.55 (m, 5H), 1.84 (s, 3H), 1.82 (s, 3H), 1.26 (s, 3H), 1.11 (s, 3H), 0.89 (s, 3H); 13C NMR δ 150.1, 148.9, 139.2, 135.1, 133.6, 129.8, 129.4, 127.3, 127.1, 125.1, 122.6, 121.0, 120.3, 116.4, 115.2, 106.9, 102.8, 102.8, 78.1, 56.0, 46.8, 38.4, 37.7, 28.3, 27.4, 25.8, 25.7, 23.2, 19.8, 17.7, 14.3; HRMS (EI) m/z calcd for C32H39NO4 (M+) 501.2879; found 501.2874.</p><!><p>To alcohol 41 (1.09 g, 3.01 mmol) in CH2Cl2 (50 mL) at 0 °C was added imidazole (502 mg, 7.53 mmol) and TBSCl (500 mg, 3.31 mmol), and then the solution was allowed to warm to rt. The next day the reaction was quenched by addition of NH4Cl (sat) and extracted with CH2Cl2. The combined organic extracts were washed with brine, dried (MgSO4), and filtered, and the filtrate was concentrated in vacuo. Final purification by flash column chromatography (8% ethyl acetate in hexanes) afforded silyl ether 52 (1.39 g, 97%): 1H NMR δ 7.75 (d, J = 8.4 Hz, 2H), 7.63 (m, 1H), 7.45 (d, J = 3.7 Hz, 1H), 7.20, (dd, J = 8.5, 0.6 Hz, 2H), 6.88 (m, 1H), 6.73 (dd, J = 3.7, 0.8 Hz, 1H), 5.24 (s, 2H), 4.81 (s, 2H), 3.47 (s, 3H), 2.33 (s, 3H), 0.97 (s, 9H), 0.12 (s, 6H); 13C δ 150.3, 144.8, 139.8, 136.1, 135.3, 129.8 (2C), 128.8 (2C), 124.9, 120.7, 105.8, 105.9, 104.9, 94.7, 65.2, 56.1, 25.9 (3C), 21.5, 18.3, −5.2 (2C); HRMS (EI) m/z calcd for C24H33NO5SSi (M+) 475.1849; found 475.1856.</p><!><p>To the silyl-protected indole 52 (724 mg, 1.52 mmol) in THF was added a few 4 Å molecular sieves and the mixture was cooled to −78 °C. After n-BuLi (0.75 mL, 2.3M in hexanes) was added, the mixture was stirred for 20 min and then prenyl bromide (420 mg, 2.82 mmol) was added. The next day the reaction mixture was quenched by addition of NH4Cl (sat), and extracted with Et2O. The combined organic layers were washed with brine, dried (MgSO4), and filtered, and the filtrated was concentrated in vacuo. Final purification by flash column chromatography (5% ethyl acetate in hexanes) afforded the prenyl indole 53 (560 mg, 68%) as well as recovered starting material 52 (76 mg, 10%): 1H NMR δ 7.91 (d, J = 0.8 Hz, 1H), 7.73 (d, J = 8.4 Hz, 2H), 7.25, (d, J = 8.5 Hz, 2H), 6.99 (s, 1H), 6.52 (d, J = 0.8 Hz, 1H), 5.47 (m, 1H), 5.31 (s, 2H), 4.90 (s, 2H), 3.74 (d, J = 7.2 Hz, 2H), 3.55 (s, 3H), 2.40 (s, 3H), 1.86 (s, 3H), 1.71 (s, 3H) 1.05 (s, 9H), 0.20 (s, 6H); 13C NMR δ 149.5, 144.5, 139.9, 138.7, 138.6, 136.5, 134.5, 129.7 (2C), 126.3 (2C), 119.8, 119.6, 106.5, 106.3, 105.3, 94.8, 65.5, 56.0, 27.9, 25.9 (3C), 25.7, 21.4, 18.3, 17.7, −5.2 (2C); HRMS (EI) m/z calcd for C29H41NO5SSi (M+) 543.2475; found 543.2476.</p><!><p>To the silyl ether 53 (682 mg, 1.26 mmol) in THF (20 mL) at rt was added TBAF (1.88 mL, 1.0 M in THF). After 2 h the reaction was quenched by addition of water and extracted with ethyl acetate. The combined organic extracts were washed with brine, dried (MgSO4), and filtered, and the filtrate was concentrated in vacuo. Purification by flash column chromatography (30 to 45% ethyl acetate in hexanes) afforded alcohol 54 (461 mg, 85%): 1H NMR δ 7.84 (s, 1H), 7.74 (d, J = 8.3 Hz, 2H), 7.17 (d, J = 8.4 Hz, 2H), 6.93 (s, 1H), 6.44 (s, 1H), 5.38 (m, 1H), 5.24 (s, 2H), 4.74 (s, 2H), 3.64 (d, J = 7.1 Hz, 2H), 3.46 (s, 3H), 2.60 (br s, 1H), 2.31 (s, 3H), 1.78 (s, 3H), 1.61 (s, 3H); 13C δ 149.5, 144.6, 140.1, 138.5, 138.1, 136.2, 134.7, 129.7 (2C), 126.2 (2C), 119.9, 119.5, 107.2, 106.7, 105.2, 94.5, 65.7, 56.1, 27.8, 25.7, 21.4, 17.6; HRMS (TOF MS EI) calcd m/z for C23H27NO5S (M+) 429.1610; found 429.1622.</p><!><p>To benzylic alcohol 54 (333 mg, 0.78 mmol) in THF was added LiBr (540 mg, 6.20 mmol) and NEt3 (0.44 mL, 3.10 mmol) and the solution was cooled to 0 °C. After 15 min MsCl (0.19 mL, 2.46 mmol) was added dropwise and the reaction was allowed to stir and slowly warm to rt. After 2 h, when the reaction was complete by TLC analysis, it was quenched by addition of water and extracted with Et2O. The organic extracts were washed with brine, dried (MgSO4), and filtered, and the filtrate was concentrated in vacuo. To the resulting residue was added P(OEt)3 (3 mL) and the solution was heated at reflux overnight. The next day the solution was allowed to cool to rt and then poured into water and extracted with ethyl acetate. The organic extract was washed with brine, dried (MgSO4), and filtered, and the filtrate was concentrated in vacuo. Final purification by flash column chromatography (50 to 70% ethyl acetate in hexanes) afforded indole phosphonate 55 (384 mg, 90%): 1H NMR δ 7.82 (d, J = 2.8 Hz, 1H), 7.69 (d, J = 8.4 Hz, 2H), 7.21 (d, J = 8.5 Hz, 2H), 6.87 (s, 1H), 6.43 (s, 1H), 5.40 – 5.35 (m, 1H), 5.25 (s, 2H), 4.07 – 3.94 (m, 4H), 3.64 (d, J = 7.2 Hz, 2H), 3.48 (s, 3H), 3.26 (d, JPH = 21.3 Hz, 2H), 2.34 (s, 3H), 1.78 (s, 3H), 1.62 (s, 3H), 1.26 (t, J = 7.1 Hz, 6H); 13C NMR δ 149.3 (d, JCP = 3.1 Hz) 144.6, 140.0 (d, JCP = 1.9 Hz), 138.5 (d, JCP = 3.1 Hz), 136.2, 134.7, 129.9 (2C), 128.1 (d, JCP = 9.3 Hz), 126.3 (2C), 119.5, 119.4 (d, JCP = 3.1 Hz), 109.9 (d, JCP = 7.4 Hz), 109.5 (d, JCP = 6.1 Hz), 105.2, 94.8, 62.2 (d, JCP = 6.9 Hz, 2C), 56.2, 34.2 (d, JCP = 137.7 Hz), 27.8, 25.6, 21.4, 17.7, 16.2 (d, JCP = 5.9 Hz, 2C); 31P NMR δ 27.3; HRMS (EI) m/z calcd for C27H36NO7PS (M+) 549.1950; found 549.1943.</p><!><p>To phosphonate 55 (74 mg, 0.14 mmol) and aldehyde 11 (30 mg, 0.10 mmol) in THF (2 mL) at 0 °C was added NaH (50 mg, 1.25 mmol, 60% dispersion oil) and 15-crown-5 (3 drops). The reaction mixture was allowed to stir for 4 h, then quenched by addition of NH4Cl (sat) and extracted with ethyl acetate. The combined organic extracts were washed with brine, dried (MgSO4), and filtered, and the filtrate was concentrated in vacuo. Purification by flash column chromatography (50% ethyl acetate in hexanes) afforded a mixture of N-tosyl indole 56 and the unprotected indole 57. To the mixed residue in 1:1 THF and i-PrOH (3 mL) at 0 °C was added NaH (120 mg, 3.0 mmol) and the reaction mixture allowed to warm to rt overnight. The next day the reaction mixture was quenched by addition of NH4Cl (sat), diluted with water, and extracted with ethyl acetate. The combined organic extracts were washed with water, brine, and dried (MgSO4), filtered, and then the filtrate was concentrated in vacuo. Final purification by flash column chromatography (50% ethyl acetate in hexanes) afforded indole 57 (20 mg, 37% for the 2 steps) as an oil: 1H NMR δ 7.92 (br s, 1H), 7.08 (m, 1H), 7.02 (d, J = 16.1 Hz, 1H), 6.96 (m, 1H), 6.94 (d, J = 16.1 Hz, 1H), 6.89 (m, 1H), 6.86 (m, 1H), 6.31 (m, 1H), 5.40 (m, 1H), 5.36 (s, 2H), 3.90 (s, 3H), 3.56 (s, 3H), 3.49 – 3.39 (m, 3H), 2.74 – 2.71 (m, 2H), 2.18 – 2.10 (m, 1H), 1.90 – 1.60 (m, 5H), 1.79 (s, 3H), 1.74 (s, 3H), 1.26 (s, 3H), 1.11 (s, 3H), 0.89 (s, 3H); 13C NMR δ 150.1, 148.9, 142.3, 138.3, 137.5, 134.6, 132.1, 129.5, 127.8, 126.4, 122.6, 120.1, 120.1, 119.9, 107.1, 106.9, 103.5, 102.3, 95.0, 78.1, 77.0, 56.1, 56.0, 46.8, 38.4, 37.7, 28.3, 27.4, 27.1, 25.7, 23.2, 19.9, 17.8, 14.3; HRMS (EI) m/z calcd for C34H43NO5 (M+) 545.3141; found 545.3135.</p><!><p>To compound 57 (8 mg, 0.015 mmol) in MeOH (0.8 mL) in a foil-wrapped flask was added TsOH (25 mg, 0.13 mmol) and the reaction was allowed to stir at rt. After 10 h the reaction was quenched by addition of NaHCO3 (sat) and extracted with ethyl acetate. The combined organic extracts were washed with brine, dried (MgSO4), and filtered, and the filtrate was concentrated in vacuo. Final purification by radial chromatography (50% ethyl acetate in hexanes) afforded compound 10 (5 mg, 68%) as a light yellow oil: 1H NMR (CD3OD) δ 6.99 (d, J = 16.4 Hz, 1H), 6.95 (m, 2H), 6.90 (d, J = 16.2 Hz, 1H), 6.82 (m, 1H), 6.63 (s, 1H), 6.17 (s, 1H), 5.46 – 5.41 (m, 1H), 3.85 (s, 3H), 3.44 (d, J = 7.3 Hz, 2H), 3.37 (dd, J = 10.8, 3.9 Hz, 1H), 2.76 – 2.73 (m, 2H), 2.07 – 2.02 (m, 1H), 1.85 – 1.60 (m, 4H), 1.79 (s, 3H), 1.75 (s, 3H), 1.23 (s, 3H), 1.11 (s, 3H), 0.88 (s, 3H); 13C NMR δ 150.5, 150.1, 143.2, 140.1, 139.4, 134.3, 132.9, 131.4, 129.3, 126.6, 124.0, 122.2, 121.4, 119.4, 108.0, 103.4, 102.0, 96.7, 78.7, 78.1, 56.4, ~49 (obscured by solvent), 39.5, 38.9, 29.0, 28.0, 27.9, 25.9, 24.1, 20.2, 17.8, 14.9; HRMS (TOF MS ES) m/z calcd for C32H39NO4 (M+H)+ 502.2957; found 502.2956.</p>

PubMed Author Manuscript

Characterization of Fibrinogen Binding by Glycoproteins Srr1 and Srr2 of Streptococcus agalactiae*

Background: The serine-rich repeat glycoproteins Srr1 and Srr2 are surface adhesins of Streptococcus agalactiae important for pathogenicity.Results: Both Srrs bind tandem repeats of the fibrinogen Aα chain, but Srr2 has greater affinity explained by structure-function analysis of the Srrs.Conclusion: A dock, lock, and latch mechanism describes the Srr-fibrinogen interaction.Significance: The higher affinity of Srr2 may contribute to the hypervirulence of Srr2-expressing strains.

characterization_of_fibrinogen_binding_by_glycoproteins_srr1_and_srr2_of_streptococcus_agalactiae*

Introduction<!>Reagents<!>Strains and Growth Conditions<!><!>Cloning and Expression of Srr1-BR and Srr2-BR<!>Cloning and Expression of Fibrinogen Chains<!>Site-directed Mutagenesis<!>Construction of Plasmids for Gene Complementation<!>Analysis of Srr2-BR Binding to Fibrinogen by Far Western Blotting<!>Analysis of Srr2-BRs Binding to Fibrinogen by Enzyme-linked Immunosorbent Assay (ELISA)<!>Lectin Blot Analysis of Srr2 Expression by GBS<!>GBS Adherence Assay<!>Binding of GBS to Immobilized Fibrinogen and Recombinant Proteins<!>Surface Plasmon Resonance (SPR) Spectroscopy<!>Isothermal Titration Calorimetry (ITC)<!>Crystallization of Srr Binding Regions<!><!>Data Analysis<!>Srr2 Mediates GBS Binding to Fibrinogen<!><!>Srr2 Mediates GBS Binding to Fibrinogen<!><!>Structures of S. agalactiae Srr1 and Srr2 Binding Regions<!><!>Structures of S. agalactiae Srr1 and Srr2 Binding Regions<!>Srr2-BR Binding to Fibrinogen Occurs through a Variant DLL Mechanism<!><!>Srr2-BR Binding to Fibrinogen Occurs through a Variant DLL Mechanism<!><!>Identification of the Fibrinogen Region Bound by Srr2-BR<!><!>Identification of the Fibrinogen Region Bound by Srr2-BR<!><!>Quantitative Assessment of SRR Protein Binding to Fibrinogen by SPR and ITC<!><!>Quantitative Assessment of SRR Protein Binding to Fibrinogen by SPR and ITC<!>GBS Strains Expressing Srr2 Have Higher Levels of Fibrinogen and Endothelial Cell Binding<!><!>GBS Strains Expressing Srr2 Have Higher Levels of Fibrinogen and Endothelial Cell Binding<!><!>DISCUSSION<!><!>DISCUSSION<!>

<p>The serine-rich repeat (SRR)2 glycoproteins of Gram-positive bacteria are a family of adhesins that are important virulence factors for their respective pathogens (1–3). These bacterial surface components are encoded within large loci that also encode proteins mediating their glycosylation and export. Each SRR protein consists of a long and specialized signal sequence, a short serine-rich region (SRR1), a ligand binding region, a second lengthy SRR region, and a typical LPXTG cell wall anchoring motif at the C terminus (4, 5). Although relatively few of the SRR proteins have been studied in detail, the binding regions of the SRR glycoproteins appear to vary significantly in predicted structure and binding properties. Among the best characterized SRR proteins is GspB of Streptococcus gordonii, which binds human platelets through its interaction with sialyl-T antigen on the platelet receptor GPIb (6, 7). This interaction appears to be an important event in the pathogenesis of endocarditis, because disruption of GspB binding is associated with a marked reduction in virulence, as tested by animal models of endocardial infection (7, 8). A number of other SRR proteins have been shown to contribute to virulence, including SraP of Staphylococcus aureus, PsrP of Streptococcus pneumoniae, and UafB of Staphylococcus saprophyticus (9–11), although the molecular basis for binding by these other adhesins is somewhat less well defined.</p><p>Streptococcus agalactiae (group B Streptococcus, GBS) is a leading cause of neonatal sepsis, pneumonia, and meningitis (12, 13). In recent decades, this organism has also become a significant cause of invasive infections among adults (14). GBS strains express either one of two SRR proteins, Srr1 or Srr2. Expression of Srr1 by GBS has been shown to contribute to colonization and virulence in models of infection (15–17). Srr1 mediates bacterial binding to cytokeratin 4, which is likely to be important for colonization of the female genital tract and is a risk factor for subsequent invasive disease (17, 18). In addition, we have recently shown that Srr1 binds to human fibrinogen via its interaction with the Aα chain of the protein. Srr1-mediated binding to fibrinogen is important for the attachment of GBS to human brain microvascular endothelial cells (hBMEC), where fibrinogen served as a bridging molecule between Srr1 and the endovascular surface (4).</p><p>Sequence comparisons and deletion mutagenesis studies (4) suggest that the interaction between Srr1 and fibrinogen could employ the "dock, lock, and latch" (DLL) mechanism described for several other fibrinogen-binding adhesins, such as ClfB of S. aureus and SdrG of Staphylococcus epidermidis (19–21). During this binding process, fibrinogen engages a cleft between two IgG-like folds (the N2 and N3 domains) of the binding region. This docking event results in a conformational change of the adhesin, such that the flexible C terminus of the N3 domain (the "latch") forms a β-strand and completes a β-sheet within the N2 domain, thereby "locking" the ligand in place. Deletion of the latch region of Srr1 is associated with reduced GBS binding in vitro to fibrinogen and hBMEC and resulted in attenuated virulence in a mouse model of bacteremia and meningitis (4). These findings indicate that fibrinogen binding via Srr1 may occur via a DLL mechanism and that this interaction enhances pathogenicity.</p><p>As compared with Srr1, relatively little is known about the binding properties of Srr2 or its contribution toward GBS virulence. Srr2 has been detected in serotype III strains exclusively and only in isolates belonging to sequence multilocus sequence type 17 (ST-17), a genotype linked epidemiologically to increased invasive disease (16, 22–28). In addition, strains expressing Srr2 were significantly more virulent in a mouse model of neonatal sepsis, as compared with Srr1-expressing strains (16), suggesting that this surface component may at least in part explain the increased virulence associated with ST-17 isolates. ST-17 strains also have higher levels of fibrinogen binding, but the molecular basis for this has not been well defined (28). Delineating the molecular differences between Srr1 and Srr2 could improve our understanding of how Srr2 confers hypervirulence in S. agalactiae. We now report that both Srr1 and Srr2 bind to a specific tandem repeat region of fibrinogen Aα chain. Crystal structures and mutagenesis studies indicate that both proteins employ a DLL mechanism for host binding. Moreover, Srr2 has significantly higher binding affinity for fibrinogen as compared with Srr1, and analysis of their structures suggests that the physical positioning of the latch region may underlie this enhanced affinity.</p><!><p>Purified human fibrinogen was obtained from Hematologic Technologies. Rabbit anti-fibrinogen IgG was purchased from Aniara. Rabbit anti-Srr2 IgG was generated by NeoPeptide, using purified recombinant protein corresponding to the binding region (BR) of Srr2.</p><!><p>The bacteria and plasmids used in this study are listed in Tables 1 and 2. S. agalactiae strains were grown in Todd-Hewitt broth (Difco) supplemented with 0.5% yeast extract (THY). All mutant strains grew comparably well in vitro, as compared with parent strains (data not shown). Escherichia coli strains DH5α, BL21, and BL21(DE3) were grown at 37 °C under aeration in Luria broth (LB; Difco). Antibiotics were added to the media as required.</p><!><p>Strains</p><p>a ErmR, erythromycin resistance; CmR, chloramphenicol resistance.</p><p>Plasmids</p><!><p>Genomic DNA was isolated from GBS NCTC 10/84 and COH1 using Wizard Genomic DNA purification kits (Promega), according to the manufacturer's instructions. PCR products were cloned into pET28-FLAG to express FLAG-tagged versions of Srr1-BR (amino acids 303–641), Srr2-BR (amino acids 303–641), or the latch deletion variant of Srr2-BR (amino acids 303–628). DNA encoding Srr1-BR, Srr2-BR, Srr1-BRΔlatch, Srr2-BRΔlatch, or ClfA-BR (N2N3) were cloned into pET22b(+) (Novagen) or pET28-FLAG. Proteins were purified by either nickel-nitrilotriacetic acid (Promega) or anti-FLAG M2-agarose affinity chromatography (Sigma), according to the manufacturers' instructions.</p><!><p>DNA of each chain was amplified from cDNA encoding the Aα-, Bβ-, and γ-chains of human fibrinogen and cloned into pMAL-C2X (New England Laboratory) as described previously (29–31). The recombinant proteins were purified by affinity chromatography with amylose resin, according to the manufacturer's instructions (New England Biolabs).</p><!><p>Cysteine replacement mutations were made within latch and latching cleft domains of Srr-BRs by a two-stage PCR procedure. For codon conversion to cysteine in the latching cleft, overlapping primers were used with either primer 3006(NotI)/5003 (N423C) or 3003(N423C)/5006(XhoI) for Srr1-BR and either 3012(NotI)/5009 (N336C) or 3009(N339C)/5012(XhoI) for Srr2-BR to generate overlapping DNA fragments spanning the entire Srr1-BR and Srr2-BR. The two DNA fragments were combined for the second stage PCR and then amplified using primers 3006(NotI)/5006(XhoI, K639C) for the Srr1-BR and 3012(NotI)/5012(XhoI, N541C) for Srr2-BR. Amplified products were digested with the appropriate restriction enzymes and ligated into pET28-FLAG. The constructs were sequenced to confirm that the mutations were correctly positioned and then expressed in E. coli, as described above.</p><!><p>Genomic DNA was isolated from COH1 and NCTC 10/84 strains, using Wizard Genomic DNA purification kits (Promega). Polymerase chain reaction (PCR) was performed with the primers (Srr1 forward, AAT CTA GAT AGA TTT CTA ATC ACT TAA TTT TAC, and Srr1 reverse, GCT CTA GAA GAA TTC AAA GTA GGT TTA GTC; Srr2 forward, TTT CTA GAT AGC ATT ATT TTT TAA ATA TGG, and Srr2 reverse, TTC TGC AGT TAA TCT TTT TTC TTC TTG C) to amplify srr1 or srr2 genes. PCR products were purified, digested, and ligated into pDE123 to express the full-length Srr1 and Srr2.</p><!><p>Purified human fibrinogen (0.1 μg) and recombinant fibrinogen truncations (0.5 μg) were separated by electrophoresis in 3–8% NuPAGE Tris acetate gels (Invitrogen) and transferred onto nitrocellulose membranes. The membranes were treated with a casein-based blocking solution (Western Blocking Reagent, Roche Applied Science) for 1 h at room temperature and then incubated for 1 h with FLAG-tagged Srr2-BR(5 μg/ml) suspended in PBS, 0.05% Tween 20 (PBS-T). The membranes were then washed three times for 15 min in PBS-T, and bound proteins were detected with mouse anti-FLAG antibody (Sigma).</p><!><p>Purified fibrinogen (0.1 μm) or recombinant fibrinogen truncations were immobilized overnight in 96-well microtiter plates at 4 °C. The wells were blocked with 300 μl of casein blocking solution (Roche Applied Science) for 1 h at room temperature (32, 33). The plates were washed three times with PBS-T and either FLAG-Srr2-BR or FLAG-Srr2-BRΔlatch in PBS-T was added over a range of concentrations for 1 h. The plates were incubated for 1 h at 37 °C, washed with PBS-T to remove unbound protein, and incubated with mouse anti-FLAG antibodies (1:4000) in PBS-T for 1 h at 37 °C. Wells were washed and incubated with HRP-conjugated rabbit anti-mouse IgG (Sigma) diluted 1:5000 in PBS-T for 1 h at 37 °C. For some studies, wells containing immobilized fibrinogen were pretreated with rabbit anti-fibrinogen IgG or recombinant untagged proteins, followed by washing prior to the addition of FLAG-Srr2-BR. Levels of binding were assessed by absorbance at 450 or 495 nm, using 3,3′,5,5′-tetramethylbenzidine or o-phenylenediamine dihydrochloride as chromogenic substrates.</p><!><p>Cell wall proteins were released from whole bacteria using mutanolysin, as described previously (4). The proteins were separated by SDS-PAGE in 3–8% Tris acetate gels (Invitrogen) under reducing conditions (0.5 m dithiothreitol) and transferred to nitrocellulose membranes. After incubating for 1 h at room temperature with the casein blocking reagent (Roche Applied Science), the membranes were incubated with biotin-conjugated wheat germ agglutinin (Vector Laboratories) (0.2 μg/ml) followed by incubation with HRP-conjugated streptavidin (0.2 μg/ml) (4).</p><!><p>Primary hBMEC were purchased from ScienCell (34). Bacterial adherence assays were performed as described (15). In brief, bacteria were grown to mid-log phase and adjusted to the concentration of 105 CFU/ml in PBS. Bacterial suspensions were added to confluent hBMEC monolayers and incubated for 30 min. The wells were washed to remove unbound bacteria and treated with 100 μl of trypsin (2.5 mg/ml) for 10 min at 37 °C to release attached bacteria. The number of bound bacteria was determined by plating serial dilutions of the recovered bacterial suspensions onto THY agar. After 24 h, the number of bacteria were counted, and bacterial adherence was calculated as recovered CFU/initial inoculum CFU × 100%. In the indicated experiments, exogenous fibrinogen (20 μg/ml) was added directly to bacteria and incubated for 30 min with rotation at 37 °C prior to addition to hBMEC monolayers.</p><!><p>Overnight cultures of GBS were harvested by centrifugation and suspended in PBS (final concentration, 106 CFU/ml). Purified fibrinogen (0.1 μm) or recombinant truncated fibrinogen polypeptides were immobilized in 96-well microtiter plates and then incubated with 100 μl of GBS suspension for 30 min at 37 °C. Unbound bacteria were removed from the plates by washing with PBS, and the number of bound bacteria was determined by treating the wells with trypsin and plating serial dilutions of the recovered bacteria onto THY agar plates as described above or staining with crystal violet (0.5% v/v) for 1 min, as described previously (32).</p><!><p>SPR spectroscopy was performed using a BIAcore T100 system (GE Healthcare). Purified human fibrinogen (10 μg/ml in sodium citrate buffer, pH 5.5) was covalently immobilized on CM5 sensor chips using amine coupling as described previously (35, 36). Increasing 2-fold concentrations (range, 1.25–160 μm) of Srr1-BR and Srr2-BR were flowed over fibrinogen or block reagent (ethanolamine) at a rate of 10 μl/min. The sensorgram data were subtracted from the corresponding data from the reference flow cell and analyzed using the BIAevaluation software version 3.0. A plot of the level of binding (response units) at equilibrium against a concentration of analyte was used to determine the KD.</p><!><p>ITC was performed with a MicroCal ITC200 microcalorimeter at 25 °C as described previously for ClfA (a fibrinogen-binding protein of S. aureus) (35). All recombinant proteins were dialyzed against HBS buffer (10 mm HEPES, 150 mm NaCl, pH 7.4). The reaction cell contained 50 μm fibrinogen Aα RU678 or RU789 (expressed as MBP fusion proteins), and the syringe contained 0.5 mm recombinant Srr1-BR or Srr2-BR in HBS buffer. These concentrations were based on the above-published studies with ClfA. The data were analyzed using MicroCal Origin software (version 5.0), with results fitted to a single binding mode (35, 36).</p><!><p>Purified recombinant Srr1-BR (7.6 mg/ml in 0.5 m NaCl, 0.01 m Tris-HCl, pH 8.3, 5 mm β-mercaptoethanol) was crystallized using the sitting drop vapor diffusion method with 1 μl of protein incubated with 0.2 mm peptide (NPGSPRPGSTGTWNPGSSERGSAGHWTSESSVSGSTGQWHSESGSFRPDSPG) and 1 μl of reservoir solution (0.2 m CaCl2, 0.1 m Tris-HCl, pH 6.0, 20% (w/v) PEG 6000) at room temperature. Srr2-BR was crystallized using the sitting drop vapor diffusion method with 1 μl of protein (7.5 mg/ml, 0.25 m NaCl, 0.01 m Tris-HCl, pH 8.3, 5 mm β-mercaptoethanol) and 1 μl of reservoir solution (5 m NaCl) at room temperature.</p><p>Crystals of S. agalactiae Srr1-BR were cryo-cooled from the reservoir solution without additional cryoprotectant. Data were collected at 100 K using beamline 21-ID-D of the Life Sciences Collaborative Access Team (LS-CAT) at the Advanced Photon Source (Argonne, IL) using a wavelength of 0.97928 Å and a MarMosaic 300 CCD detector. Crystals of S. agalactiae Srr2-BR were removed from the crystallization droplet and cryo-protected in 4 m NaHCOO prior to cryo-cooling. Data were collected at 100 K on beamline 21-ID-G of the LS-CAT using a wavelength of 0.97956 Å and a MarMosiac 300 CCD detector. All data were processed using the HKL3000 suite (37). The structure of Srr1-BR was determined by molecular replacement using PHASER (38) and the structure of S. epidermidis adhesin SdrG (Protein Data Bank code 1R17) as a search model (39). The structure of Srr2-BR was determined by molecular replacement with the program PHASER (38, 40) using the refined coordinates of Srr1 as the search model. Both models were improved using iterative rounds of model building in COOT and refinement in REFMAC (41, 42). Details of data collection, structure determination, refinement, and model quality are provided in Table 3.</p><!><p>Crystallographic data collection and refinement statistics</p><p>a Values in parentheses are for the highest resolution shell.</p><p>b Rsym = Σ(Ii − 〈I〉)/Σ (〈I〉), where i is the ith measurement and 〈I〉 is the weighted mean of I.</p><p>c Rwork = Σ ‖Fobs| − |Fcalc‖/S|Fobs|.</p><p>d Rfree is calculated using the same equation as Rwork using a subset of reflections omitted from refinement and reserved in the test set of the data.</p><!><p>Binding data are expressed as means ± S.D. and were compared for statistical significance by the unpaired t test.</p><!><p>We have previously shown that the Srr1 glycoprotein of GBS can bind fibrinogen in vitro and that this interaction mediates bacterial binding to the host in vivo (4). To assess whether Srr2 has a similar role, we first measured the adherence of two GBS strains expressing Srr2 to immobilized fibrinogen. Strain NCTC 10/84, which expresses Srr1, served as a control for fibrinogen binding. As shown in Fig. 1A, strains COH1 adhered to immobilized human fibrinogen at levels that were significantly higher than those seen with a negative control (casein). Similar results were seen with strain J48 (data not shown). Binding of the strains was significantly inhibited by pretreatment of immobilized fibrinogen with anti-fibrinogen IgG, indicating that the interaction between GBS and fibrinogen was specific (Fig. 1B).</p><!><p>GBS binding to fibrinogen is mediated by glycoprotein Srr2. A, suspensions of GBS strain COH1 were incubated in microtiter wells with immobilized fibrinogen (blue) or a casein-based blocking reagent (red). Binding of bacteria was assessed by crystal violet staining and expressed as the means ± S.D. absorbance. B, immobilized fibrinogen was pre-incubated with anti-rabbit fibrinogen (Fg) IgG (100 μg/ml) or rabbit IgG (100 μg/ml) prior to testing for binding by GBS strains. Unbound IgG was removed by washing, and GBS binding was assessed. Values represent percent of GBS binding, as compared with untreated fibrinogen. NCTC 10/84, which expresses Srr1, served as a control. C, binding of GBS strain COH1 or its srr2 variant (Δsrr2) to immobilized to fibronectin (Fn) or fibrinogen (Fg). D, inhibition of GBS COH1 binding to fibrinogen by anti-Srr2 IgG. Strain COH1 was co-incubated with rabbit anti-Srr2 IgG or normal rabbit IgG, and relative binding to immobilized fibrinogen was assessed. Values are mean ± S.D. of relative binding, normalized for WT levels of binding to fibrinogen. *, p < 0.01.</p><!><p>To more directly assess the impact of Srr2 on bacterial binding, we compared the binding of COH1 and an Srr2-deficient strain (COH1Δsrr2) to immobilized fibrinogen. As shown in Fig. 1C, loss of srr2 expression significantly reduced GBS binding to fibrinogen but had no effect on bacterial binding to immobilized fibronectin. Expression of the srr2 gene in trans restored binding to wild type (COH1) levels, demonstrating that the loss of binding observed with srr2 disruption was not due to polar or pleiotropic effects (Fig. 9B). In addition, binding by COH1 to fibrinogen was inhibited by rabbit anti-Srr2 IgG but not by normal (preimmune) rabbit IgG (Fig. 1D). The level of inhibition was concentration-dependent, with 100 μg/ml of anti-Srr2 IgG being sufficient to reduce WT GBS binding to levels comparable with those seen with GBSΔsrr2. These results indicate that the binding of GBS COH1 to immobilized fibrinogen is predominantly mediated by Srr2.</p><p>To confirm that the putative binding region of Srr2 interacts with fibrinogen, we assessed the interaction of the purified FLAG-tagged binding region (FLAG-Srr2-BR) with immobilized human fibrinogen (Fig. 2). In control studies, no significant binding by FLAG-Srr2-BR to immobilized casein was detected. In contrast, FLAG-Srr2-BR showed significant levels of binding to fibrinogen, which increased in direct proportion to the amount of protein applied. At concentrations above 3.3 μm of FLAG-Srr2-BR, binding reached a plateau, consistent with saturation. Binding of Srr2-BR was significantly inhibited by anti-fibrinogen IgG, indicating that this interaction was specific (Fig. 2C).</p><!><p>Interaction of the BR of Srr2 with fibrinogen. A, schematic diagram of the serine-rich repeat proteins Srr1 and Srr2. Level of identity (%) between regions is indicated. SS, signal sequence; Srr1-BR and Srr2-BR, binding domains; SRR1 and SRR2, serine-rich regions; LPxTG, cell wall anchoring motif; AA, amino acids. B, binding of FLAG-Srr2-BR and FLAG-Srr2-BRΔlatch proteins to immobilized fibrinogen. Indicated concentrations of FLAG-Srr2-BR and FLAG-Srr2-BRΔlatch were added to wells coated with fibrinogen or casein blocking reagent. C, inhibition of FLAG-Srr2-BR binding to immobilized fibrinogen by anti-fibrinogen IgG. Values represent percent of FLAG-Srr2-BR binding to the wells treated with fibrinogen. Bars indicate the means ± S.D. *, p < 0.01.</p><!><p>To assess whether Srr1 and Srr2 could support a DLL mechanism of ligand binding, we determined the crystal structures of S. agalactiae Srr1-BR and Srr2-BR at resolutions of 1.65 and 3.65 Å, respectively (Fig. 3). As was previously predicted from sequence and functional analyses (4), Srr1-BR and Srr2-BR each adopt an overall fold that resembles the binding regions of "microbial surface components recognizing adhesive matrix molecules" (MSCRAMMs), including the well characterized ClfA (43, 44), ClfB (20), SdrG (21), and the likely MSCRAMM UafA (45). However, Srr1-BR and Srr2-BR are more similar to each other than they are to these structurally characterized MSCRAMMs, with a root mean square deviation of 1.6 Å between Srr1-BR and Srr2-BR and root mean square deviations between 2.1 and 2.5 Å when either Srr1 or Srr2 is placed in a pairwise structural alignment with ClfA, ClfB, SdrG, or UafA (20, 21, 43–45).</p><!><p>Structures of Srr1-BR and Srr2-BR. A, structure of Srr1-BR with secondary structural elements colored green and turns colored in gray. The latch region is disordered in the structure. B, structure of Srr2-BR with secondary structural elements colored yellow and turns colored gray. The latch region is highlighted in cerulean blue. C, structure of ClfB without peptide ligand mimetic shows a latching region that is open and an unoccupied latching cleft ((Protein Data Bank entry 4F24 (19)). The ordered region of the latch is highlighted in cerulean blue. D, structure of ClfB with peptide ligand mimetic identifies peptide binding to the cleft between the N2 and N3 domains and shows the latch in the locked position. (Protein Data Bank entry 4F27 (19)). The peptide is shown in red, and the latch is shown in cerulean blue.</p><!><p>Like other MSCRAMMs, Srr1-BR and Srr2-BR contain two domains, termed N2 and N3, with each adopting the DE variation of the Ig fold (44). Between the N2 and N3 domains is a cleft, demonstrated to be the ligand-binding site in other MSCRAMMs. The size and shape of the interdomain cleft is consistent with this being the binding site. An unusual feature of the Srr2-BR structure, not observed in Srr1-BR or in any previous structures of MSCRAMMs, is the conformation of the C terminus of the N3 domain. To date, the structures of MSCRAMMs crystallized in the absence of a ligand have the C-terminal extension of the N3 domain either completely or partially disordered (21, 44, 45). This C-terminal extension, known as the latch, has been shown to close over the peptide mimetic of a host ligand in co-crystal structures. Once closed, the latch forms a β-strand that completes a fully hydrogen-bonded β-sheet within the N2 domain and locks the host ligand in place (21, 43) (Fig. 3, C and D). In contrast to other structurally characterized DLL proteins, the latch region in S. agalactiae Srr2-BR adopted a distinct conformation in the absence of ligand. Although the assignment of amino acids is somewhat tenuous at this resolution, electron density is consistent with the latch being nearly closed or "ajar."</p><!><p>To reveal whether the latch domain of Srr2-BR was important for binding, we generated a variant of the protein (Srr2-BRΔlatch), in which the terminal 13 residues of this domain were deleted. We have previously shown that this type of deletion in Srr1-BR abrogates binding, without altering the overall conformation of the protein (4). As shown in Fig. 2B, removing the C-terminal amino acids of the Srr2 binding region abolished fibrinogen binding. In addition, untagged Srr2-BR inhibited the binding of FLAG-Srr2-BR to immobilized fibrinogen, whereas untagged Srr2-BRΔlatch failed to block this interaction (Fig. 4B). These data indicate that the fibrinogen binding domain of Srr2 is indeed located in the predicted binding region and that Srr2-BR binds fibrinogen by a DLL mechanism.</p><!><p>Identification of the Srr2 binding domain on fibrinogen. A, schematic drawing of human fibrinogen. The 10 tandem repeating units of the Aα chain are shown in purple. B, inhibition of FLAG-Srr2-BR (0.05 μm) binding to immobilized fibrinogen by purified untagged proteins (10 μm). C, binding of FLAG-Srr1-BR (FLAGSrr1-BR) (25 μg/ml) or FLAG-Srr2-BR (FLAGSrr1-BR (5 μg/ml) to MBP fused with full-length recombinant Aα (MBP-Aα) or subdomains of the Aα chain. Subscripts indicate amino acids contained within each fragment. Bars represent the mean binding levels (± S.D.). D, Srr2-BR binding to the fibrinogen (Fg) Aα chain. Purified human fibrinogen was separated by SDS-PAGE and stained with Coomassie Blue (left panel). Far Western blotting of fibrinogen with Srr2-BR is shown in the right panel. E, recombinant MBP-Aα, Bβ, and γ chains probed with FLAG-Srr2-BR (5 μg/ml). F, recombinant MBP-Aα and its truncated variants probed with FLAG-Srr2-BR (5 μg/ml; right). Lane 1, MBP:Aα(1–610); lane 2, MBP:Aα(1–197); lane 3, MBP:Aα(198–610); lane 4, MBP:Aα(198–282); lane 5, MBP:Aα(283–410); and lane 6, MBP:Aα(198–282 + 411–610).</p><!><p>Based on the above findings, we hypothesized that the affinity of Srr-BRs to fibrinogen is affected by the location of the latch prior to ligand binding, with a more closed conformation of the latch associated with enhanced affinity. To address this possibility, we constructed variants of the Srr1 and Srr2 BRs, in which residues in both the latch and latching cleft regions were replaced with a cysteine, such that a disulfide bond would be formed, thus fixing the spatial position of the latch. We used the crystal structures to guide the location of the cysteine insertions, to generate BRs with the latch adopting a closed conformation, even in the absence of ligand. When assessed by SDS-PAGE under nonreducing conditions, the mutated Srr1 and Srr2 BRs migrated faster than their respective WT proteins, presumably due to the more compact folding of the cross-linked proteins (Fig. 5). Under reducing conditions (1 mm DTT), the mutant proteins had mobilities identical to their WT counterparts As compared with the WT Srr1-BR and the Srr2 BR, both cross-linked proteins showed enhanced binding to fibrinogen (Fig. 5). These data indicate that the affinities of Srr1 and Srr2 for fibrinogen are enhanced by a pre-closed latch, consistent with the location of the latch influencing binding affinity.</p><!><p>Effect of cross-linking the latch and latching cleft domains on fibrinogen binding. Left panels, cysteine substitutions were made in the indicated residues of Srr1-BR and Srr2-BR, and the resultant recombinant proteins were compared with their respective WT proteins for binding to immobilized fibrinogen, as measured by ELISA. Values shown are the means ± S.D. Right panels, SDS-PAGE of WT versus cysteine-cross-linked Srr1-BR and Srr-2 BR under reducing (+) or nonreducing (−) conditions (Coomassie Blue staining).</p><!><p>Fibrinogen is a 340-kDa hexameric glycoprotein composed of three pairs of chains (α, β, and γ) linked by disulfide bonds (Fig. 4A). To identify the binding site on fibrinogen for Srr2, we first examined whether binding of FLAG-Srr2-BR to immobilized fibrinogen could be inhibited by untagged Srr1-BR (which binds the tandem repeat region of the fibrinogen Aα chain) or untagged ClfA-BR (which binds the N terminus of the fibrinogen Bβ chain) (4, 46). As shown in Fig. 4B, binding of FLAG-Srr2-BR was significantly reduced by pretreatment of immobilized fibrinogen with either untagged Srr1-BR or Srr2-BR. In contrast, no inhibition was seen with ClfA-BR. To further characterize the binding site for Srr2 on fibrinogen, we analyzed by far Western blotting the interaction of FLAG-Srr2-BR with human fibrinogen and recombinant fibrinogen Aα, Bβ, and γ chains. As was found for Srr1 (4), Srr2 binding was only detected to the Aα chain (Fig. 4, C and D). To better define the binding site on this chain for the SRR proteins, we then measured their binding to recombinant Aα chain fragments. When assessed by ELISA, we found no significant binding of either FLAG-Srr1-BR or FLAG-Srr2-BR to MBP:Aα(198–282) or MBP:Aα(198–282 + 411–610). In contrast, both SRR proteins exhibited levels of binding to MBP:Aα(283–410) that were comparable with those seen with the recombinant full-length Aα chain (MBP:Aα(1–610); Fig. 4E). Far Western blotting analysis confirmed that the Srr1-BR- and Srr2-BR-binding sites are indeed contained within the tandem repeat region (amino acids 283–410) of the Aα chain of fibrinogen (Fig. 4F).</p><p>We then sought to characterize further which subdomain within the tandem repeat region is the receptor for the SRR proteins. This region of the fibrinogen Aα chain is composed of 10 repeating units, each containing 13 amino acids (Fig. 6). We therefore expressed various portions of this region as maltose-binding protein fusions and assessed SRR protein binding to these peptides by far Western blotting (Fig. 6A). Peptides composed of repeat units 1–8 (RU1–8) and 1–9 (RU1–9) were bound by both Srr1-BR and Srr2-BR. In addition, tandem repeat units 5–9 (RU5–9) and 6–9 (RU6–9) were bound by both binding regions, indicating that the binding site of the Srr proteins is located within tandem repeats 6–8 of the fibrinogen Aα chain. To directly confirm these findings, we assessed Srr1-BR or Srr2-BR binding to peptides composed of tandem repeat units 1–10 (RU1–10), 6–8 (RU678), and 7–9 (RU789), as measured by ELISA (Fig. 6B). Binding levels of Srr1-BR and Srr2-BR to RU678 were comparable with those seen with the full-length tandem repeat region (RU1–10). In contrast, no binding was detected with immobilized RU789, suggesting that the RU678 comprises the minimum target for Srr1-BR and Srr2-BR binding to the fibrinogen Aα chain.</p><!><p>Srr1-BR and Srr2-BR bind the repeat domain of the fibrinogen Aα chain. A, recombinant MBP-RU fusion proteins were separated by SDS-PAGE and stained with Coomassie Blue (top) or transferred to nitrocellulose and probed with FLAG-Srr1-BR (25 μg/ml, middle) or FLAG-Srr2-BR (5 μg/ml, bottom). B, FLAG-Srr1-BR or FLAG-Srr2-BR (5 μg/ml) was incubated with immobilized recombinant MBP-RU1–10, MBP-RU789, or MBP-RU6789. Binding was detected by ELISA using anti-FLAG antibody. Bars indicate the means (± S.D.).</p><!><p>To examine whether the binding of GBS NCTC 10/84 and COH1 to fibrinogen was mediated by the interaction of Srr1-BR and Srr2-BR with RU678, the above WT strains and their respective Δsrr1 or Δsrr2 isogenic mutants were incubated with immobilized fibrinogen, RU1–10 (amino acids 283–410), RU678, or RU789 (Fig. 7). The WT GBS strains exhibited levels of binding to RU678 that were comparable with those seen with fibrinogen, but both strains had only minimal levels of binding to immobilized RU789. In contrast, the GBSΔsrr1 and Δsrr2 mutant strains exhibited low levels of binding to all immobilized proteins. To further investigate whether GBS binding to fibrinogen was mediated by the interaction of Srr1-BR or Srr2-BR with RU678, we tested the ability of the RU678 peptide to inhibit GBS binding to immobilized fibrinogen. Preincubation of GBS strains with 10 μm RU678 resulted in a significant reduction in binding (Fig. 7B), further indicating that bacterial binding to fibrinogen is mediated by the interaction of the Srr proteins with repeating units 6–8 of the fibrinogen Aα chain.</p><!><p>GBS binds RU678 of the fibrinogen Aα chain. A, GBS strains NCTC 10/84 and COH1 were compared with their Δsrr1 or Δsrr2 isogenic variants for binding to fibrinogen, MBP-RU1–10, MBP-RU678, and MBP-RU789. Values are levels of binding relative to the WT strain, expressed as mean ± S.D. *, p < 0.01. B, inhibition of GBS COH1 or NCTC 10/84 binding to fibrinogen by purified MBP-RU678. 106 CFU of WT GBS were co-incubated with MBP-RU678 in 96-well plates coated with fibrinogen. Values represent percent of WT GBS binding. *, p < 0.01.</p><!><p>We assessed the relative binding of Srr1-BR and Srr2-BR to immobilized fibrinogen (0.1 μm) or MBP-RU678 (0.1 μm), by ELISA as described previously (4). As shown in Fig. 8, A and B, both proteins bound to immobilized fibrinogen and MBP-RU678 in a concentration-dependent manner. Next, the binding of Srr1-BR and Srr2-BR to fibrinogen was analyzed via SPR (Fig. 8C). Increasing concentrations of Srr1-BR or Srr2-BR (1.25–160 μm) were flowed over fibrinogen immobilized on a CM5 chip, and the dissociation constant (KD) of binding was determined from analysis of the equilibrium binding data. Srr1-BR and Srr2-BR showed specific and concentration-dependent binding to human fibrinogen. The KD values of Srr1-BR and Srr2-BR were determined to be 2.1 × 10−5 and 3.7 × 10−6 m, respectively. These values are within the range reported for fibrinogen-binding proteins of Gram-positive bacteria (43, 47). However, the above KD value for Srr1-BR was considerably larger than the 7.51 × 10−8 we had previously calculated, based on ELISA data (4). A similar variation in KD values has been reported for ClfA binding to fibrinogen (43), although the reason for this variability is unclear.</p><!><p>Interaction of the Srr binding regions with fibrinogen and RU678. Binding is shown of purified Srr1-BR and Srr2-BR proteins (FLAG-Srr1-BR and FLAG-Srr2-BR) to immobilized fibrinogen (A) or MBP-RU678 (B). Bound proteins were detected with anti-FLAG antibody. Bars indicate the means (± S.D.). No binding was seen to casein-coated wells (data not shown). C, surface plasmon resonance analysis of SRR binding to fibrinogen. Sensorgrams of binding to fibrinogen were obtained by passing 1.25–160 μm of Srr1-BR (A) or Srr2-BR (B) over fibrinogen immobilized on the surface of a CM5 sensor chip. Injections began at 0 s and ended at 90 s. The results shown are representative of two independent experiments. D, ITC analysis of Srr-BR binding to RU678. Srr1-BR (A) or Srr2-BR (B) were injected into an ITC chamber containing MBP-RU678. The upper panels show enthalpic heat released per s at 25 °C during titration, and the lower panels show integrated binding isotherms of the titration fitted to a one-site model.</p><!><p>We next sought to confirm these results by ITC. Because of the limited solubility of fibrinogen, we were unable to assess its binding to the Srr binding regions by this method. For that reason, we assessed the binding of Srr1-BR and Srr2-BR to RU678 (Fig. 8D). In control studies, no significant interaction of Srr1-BR or Srr2-BR with RU789 was detected (data not shown). However, Srr1-BR and Srr2-BR bound RU678 with dissociation constants (KD) of 6.9 × 10−5 and 1.2 × 10−5 m, respectively. Both binding reactions were exothermic, and the stoichiometry (n) of the binding reaction with both proteins was close to 1.</p><!><p>Because Srr2-BR exhibited higher binding affinity to fibrinogen than Srr1-BR, we next compared fibrinogen binding by Srr1-expressing strains with strains expressing Srr2. As shown in Fig. 9, the Srr2 strains (COH1 and J48) had significantly higher levels of fibrinogen binding, as compared with five strains expressing Srr1. We then examined the impact of expressing either Srr1 or Srr2 in strain COH1Δsrr2 (Fig. 9B). Of note, expression levels of the SRR glycoproteins on the cell surface of the complemented strains were comparable, as measured by binding to wheat germ agglutinin, although somewhat lower than those seen with the WT strain (data not shown). Complementation with the srr1 gene in trans significantly increased fibrinogen binding by COH1 Δsrr2, but not to levels observed with the WT strain. However, complementation of the same mutant with srr2 gene in trans restored binding of COH1 Δsrr2 to WT levels. These results indicate that the higher affinity of Srr2 for fibrinogen, as compared with Srr1, can result in higher levels of bacterial binding to the protein.</p><!><p>GBS binding to immobilized fibrinogen. A, GBS strains were incubated with wells pre-treated with fibrinogen (0.1 μm) or a casein blocking reagent. B, fibrinogen binding by strain COH1, COH1Δsrr2 ("Δsrr2"), and the mutant complemented with a plasmid encoding either srr1 (pSrr1) or srr2 (pSrr2). *, p < 0.01.</p><!><p>The attachment of GBS to hBMEC is thought to be important for the invasion of the organism into the central nervous system (48, 49). Our previous study indicates that Srr1-fibrinogen binding is important for the attachment of GBS to hBMEC. To determine whether Srr2 has a similar role, we assessed the impact of Srr2 on GBS attachment to hBMEC pretreated with purified fibrinogen. WT GBS and isogenic Δsrr1 and Δsrr2 variants were incubated with hBMEC. After 30 min, WT GBS efficiently adhered to these cells, whereas the Δsrr1 and Δsrr2 mutants were significantly reduced in binding (p < 0.01) (Fig. 10). Preincubation of bacteria with purified human fibrinogen (20 μg/ml) enhanced the binding of the WT strains to hBMEC but had no effect on the binding of the Δsrr1 and Δsrr2 mutant strains. Of note, strain COH1, which expresses Srr2, had higher levels of binding to hBMEC, as compared with the Srr1-expressing strain (NCTC 10/84), which was further increased by the addition of fibrinogen.</p><!><p>GBS adherence to hBMEC is mediated by the interaction of the Srr protein and fibrinogen. GBS strains NCTC10/84 (A) or COH1 (B) or their Δsrr mutants were incubated with hBMEC, with or without fibrinogen pretreatment (20 μg/ml). Values represent percent (mean ± S.D.) of inoculum bound to the monolayers. *, p < 0.01.</p><!><p>S. agalactiae is a leading cause of neonatal bacteremia and meningitis. Infection is initiated by colonization of the lower genital tract of pregnant women, followed by bacterial invasion and neonatal involvement. Srr1 has been shown to enhance the attachment of bacteria to vaginal and cervical epithelial cells in vitro and to augment genital colonization in mice (17). In addition, expression of the protein is associated with increased pathogenicity in animal models of infection. This enhanced virulence appears to be due at least in part to the binding of fibrinogen via Srr1, resulting in increased microvascular invasion and CNS penetration.</p><p>In contrast, Srr2 is associated with hypervirulence, but the ligand for this adhesin was previously unknown. Although the binding regions of Srr1 and Srr2 have limited amino acid sequence homology, our findings demonstrate that Srr2 also binds human fibrinogen. In addition, the crystal structures of the binding regions of Srr1 and Srr2 indicate that these both resemble those of several other fibrinogen-binding proteins, including ClfA and ClfB of S. aureus, and SdrG of S. epidermidis (19, 21, 43, 50). These and a number of other adhesins of Gram-positive bacteria bind fibrinogen through a DLL-like mechanism (35, 47, 51, 52). Co-crystal structures of MSCRAMMs with peptide mimetics of host ligands have revealed that an extended conformation of the polypeptide binds within a cleft between the N2 and N3 domains and forms a β-strand that hydrogen bonds to (and completes) a β-sheet of the N3 domain. This docking of host peptide to the trench in turn induces the C-terminal extension of the MSCRAMM to fold over the host ligand and lock it down. Variations of the DLL mechanism include whether the MSCRAMM has an inactivated state that requires a conformational change prior to binding the host ligand, whether the host ligand binds parallel or antiparallel to the strands of the MSCRAMM, and whether the latch can be closed before (latch and dock) or after (dock, lock, and latch) binding of the peptide mimetic of the host ligand (20, 21, 35, 39, 43, 51).</p><p>Srr1 and Srr2 appear to bind fibrinogen by a DLL process, because these share a fold with DLL proteins and because deletion of the predicted latch domains of the SRR proteins significantly reduced fibrinogen binding. As noted above, a shared feature of DLL adhesins is the binding of fibrinogen through a trench formed between two IgG-like folds within the binding domain. Although sequence alignment of the binding regions of Srr1 and Srr2 revealed little homology between their putative binding trenches or with the binding subdomain of ClfA (data not shown), the structures of Srr1 and Srr2 suggest that the binding trenches are more similar to each other than to other DLL proteins, consistent with the trench sequence conferring ligand selectivity. Notably, a bulky amino acid of the N3 domain (Tyr-623 of Srr1 and His-528 of Srr2) significantly constricts the center of each trench. In other structurally characterized DLL proteins, this location harbors a conserved asparagine.</p><p>One interesting finding is that, although the Srr proteins interact with the same region of fibrinogen, the binding affinity of Srr2 was higher, both when measured by SPR (using whole fibrinogen as the ligand) and by ITC (using recombinant tandem repeats 6–8). Examination of the structures of the S. agalactiae Srr1 and Srr2 binding regions suggests how Srr2 might have greater affinity for the host ligand. In Srr2, the C-terminal latch of N3 appears to be pre-ordered in a conformation that might best be called "open, but ajar" that is poised to close over host ligand quickly. Experimentally reported time scales for protein conformational changes can vary widely, and the time scales of the conformational changes in Srr1 and Srr2 have not been measured. Although the cysteine cross-linking studies suggest that both Srr1 and Srr2 C termini are able to close even in the absence of ligand, the pre-ordered conformation of the C terminus in Srr2 would almost certainly require less time to do so. This would theoretically increase the kon and reduce the koff, and it could be numerically reflected as improved ligand affinity. This is consistent with previous studies of the MSCRAMM ClfA (43), where the introduction of a disulfide bond in the latch resulted in an increase in affinity for peptides corresponding to host ligand, which bind to the cleft in an antiparallel fashion. Interestingly, the converse was observed when disulfide bonds were added to the latch of SdrG (39), which binds the host peptide mimetic in a parallel orientation (21).</p><p>Although the biological implications of this difference in affinity are unclear, it is noteworthy that Srr2 is expressed exclusively by GBS serotype III strains belonging to the ST-17 clonal complex. These organisms have been strongly associated with neonatal invasive infections (16, 22–26, 53). Few ST-17-specific virulence factors have been described to explain this enhanced pathogenicity of ST-17 strains, although it is likely that more than one virulence determinant contributes to the hypervirulence of this sequence type (27, 53). Of note, the surface protein HvgA is exclusively expressed by ST-17 strains, and its expression has been shown to promote GBS attachment to endothelial and epithelial cell lines (53). However the host receptor for HvgA remains to be elucidated.</p><p>Intriguingly, in contrast to the binding trenches, there is significant sequence homology among the latch-binding subdomain ("latching cleft") of Srr1 and Srr2 (Table 4), with both having a conserved cleft motif. Alignment of the Srr1- and Srr2-predicted latching clefts with those of other DLL adhesins showed that Srr1 has closer homology to these other DLL proteins, as compared with Srr2. Thus although the structure of the latch may control binding affinity, the influence of sequence differences in the latching clefts on the structure of the latch is unclear.</p><!><p>Putative latching cleft and latch sequences of DLL proteins</p><!><p>The binding site on fibrinogen for staphylococcal adhesins is typically composed of about 20 amino acids on one of the fibrinogen chains. For example, ClfA recognizes the C-terminal 17 residues of the γ chain (GEGQQHHLGGAKQAGDV); ClfB binds to 16 residues (tandem repeat 5; GSWNSGSSGTGSTGNQ) in the αC domain of the Aα chain, and SdrG binds the N-terminal 20 residues of the β chain (NEEGFFSARGHRPLDKKREE) (19, 43, 50). Although Srr1 and Srr2 also interact with the C terminus of the Aα chain, the binding site for both adhesins is contained within the adjacent tandem repeats 6–8 of the protein (NPGSPRPGSTGTWNPGSSERGSAGHWTSESSVSGSTGQW). Thus, not only do these bacterial surface proteins bind fibrinogen via a DLL-like mechanism, but these adhesins can interact with different regions of fibrinogen.</p><p>In vivo, GBS may interact with fibrinogen through several pathways, in addition to Srr1- or Srr2-mediated binding. Studies by Harris et al. (54) identified a cell wall-anchored protease of GBS (CspA) that both cleaved the fibrinogen Aα chain in vitro and appeared to mediate fibrinogen-dependent aggregation of whole bacteria. Deletion or disruption of cspA was associated with both increased opsonophagocytosis by neutrophils and decreased virulence in an animal model of neonatal sepsis. In addition, GBS may complement binding of Srr1 and Srr2 with binding by other adhesins. FbsA and FbsB are two additional surface proteins that also mediate GBS binding to human fibrinogen (55–59). These adhesins appear to be structurally unrelated to each other or the SRR proteins, and neither protein is associated with a DLL binding mechanism. The binding site on fibrinogen for FbsA is contained within the D fragment (60) but has not been further characterized. No binding site for FbsB has as yet been identified. Expression of FbsA enhances GBS binding to human epithelial and endothelial cells (61, 62), but it does not appear to contribute to cell invasion (63). In addition, fibrinogen binding via FbsA reduced uptake by a macrophage cell line (64), indicating that it may block phagocytosis. Deletion of fbsA attenuated the virulence of GBS in animal models of arthritis and septicemia (65), indicating that this protein contributes to pathogenicity of the organism. FbsB promotes GBS invasion of human brain microvascular cells (58), although the in vivo relevance of this phenotype remains to be examined. Recent proteomic screening identified two additional fibrinogen-binding proteins expressed by GBS, the fibronectin-binding protein Fib and a predicted ABC transporter (SAG0242) (66). The mechanisms for fibrinogen binding by these proteins are unknown, and the importance of these interactions for colonization or virulence has not as yet been described.</p><p>The results presented in this study show that Srr2 is an ST-17-specific surface protein that, like Srr1 (present in other GBS sequence types), interacts with fibrinogen through a DLL mechanism and promotes GBS attachment to human brain endothelial cells. Previous in vivo studies have shown that strains expressing Srr2 are more virulent than Srr1-expressing strains, as measured by a mouse model of sepsis (16). These findings suggest that, although both Srr proteins interact with fibrinogen, the increased affinity of Srr2 for the protein may be one factor contributing to the enhanced pathogenicity and invasive disease associated with ST-17 strains. Studies to address these issues are now in progress.</p><!><p>This work was supported, in whole or in part, by National Institutes of Health Grants R01-AI41513 and R01-AI057433 (to P. M. S.), R01-NS051247 (to K. S. D.), R01-GM09563 and R01-GM079419 (to T. M. I.), and HHSN272201200026C from NIAID (Department of Health and Human Services). This work was also supported by the Department of Veterans Affairs and the Veterans Affairs Merit Review program, the Northern California Institute for Research and Education, a fellowship award from the American Heart Association, Western Affiliate (to H. S. S.), and Grant 12GRNT11920011 from the American Heart Association.</p><p>The atomic coordinates and structure factors (codes 4MBO and 4MBR) have been deposited in the Protein Data Bank (http://wwpdb.org/).</p><p>serine-rich repeat</p><p>group B streptococci (Streptococcus agalactiae)</p><p>dock, lock, and latch</p><p>isothermal calorimetry</p><p>human brain microvascular endothelial cells</p><p>microbial surface components recognizing adhesive matrix molecules</p><p>colony-forming unit</p><p>maltose-binding protein</p><p>surface plasmon resonance</p><p>binding region</p><p>repeating unit</p><p>Life Sciences Collaborative Access Team.</p>

PubMed Open Access