Datasets:

griffin

/

ChemSum

like 12

Proton-conduction photomodulation in spiropyran-functionalized MOFs with large on–off ratio

Proton conduction in nanopores is important for applications in fuel cells, chemical sensors and information processing devices inspired by nature. Here, we present a nanoporous material, a metal-organic framework (MOF) thin film, allowing photomodulation of the aqueous and alcoholic proton conduction of the guests by almost two orders of magnitude. The MOF film possesses spiropyran groups which undergo reversible UV-light induced isomerization to the merocyanine form, a highly polar, zwitterionic molecule, where the strong binding of the guests to the merocyanine isomer efficiently suppresses the proton conduction. Such materials with photomodulated ionic conduction contribute to the development of advanced, remotecontrollable chemical sensors and to switchable devices interfacing with biological systems.

proton-conduction_photomodulation_in_spiropyran-functionalized_mofs_with_large_on–off_ratio

Introduction<!>Results and discussion<!>SURMOF synthesis<!>Sample characterization<!>Measurement of proton conduction properties<!>Conclusions<!>Conflicts of interest

<p>Ion, in particular proton, conduction is a widely spread phenomenon in nature, 1 e.g. in transmembrane proton pumps 2 and sensory receptors. 3 Moreover, proton conduction is the pivotal process in many advanced applications, like efficient and clean electric energy production by proton-exchangemembrane fuel cells 1,4 or in sensors. 5 Therefore, protonconducting materials attract particular attention. In addition to the established proton-conducting materials, like per-uorosulfonic acid polymers (like Naon) 6 and oxides, 7 functional nanoporous materials like metal-organic frameworks, MOFs, are investigated for advanced proton-conduction applications. 8 MOFs are nanoporous solid crystals, composed of metal clusters or ions connected by organic linker molecules. 9 MOFs possess various exclusive properties like high porosities with very large specic surface areas, dened structures and a great chemical and structural variety. The MOF structure can be designed by pre-and postsynthetic methods, enabling various functionalities. 10 In this way, MOF materials with embedded water or other proton-conducting molecules in the pores have been realized with proton conductivities in the range of z10 À3 to 1 S m À1 at room temperature, 8c,11 comparable to the state of the art polymeric materials. In addition to applications in fuel cells, 8c, 12 proton-conducting MOFs are investigated for applications in other elds, like in chemical sensors 13 and catalysis. 14 A particular aim in the research for advanced materials is the option of dynamic control of the material's properties. Because light is a fast, usually noninvasive signal for dynamic remote-control with a high spatial resolution, using photo-responsive molecules as functional components attract considerable attention. 15 Such photochromic molecules undergo reversible isomerization when irradiated with light of different wavelengths.</p><p>Recently, we presented the rst nanoporous material with photoswitchable proton-conduction properties. 16 In this proofof-principle study, the proton conductivity of the alcoholic and triazole guest molecule was reversibly decreased by 35% as a result of the photoisomerization of the azobenzene pendant to the MOF structure. The effect is based on the increased attractive polar interaction between the proton-conducting guests and the photoswitchable azobenzene moieties, whose dipole moment can be switched between 0 and 3 Debye. A major aim is to increase the on-off ratio from a few ten percent to many orders of magnitude, allowing their application in devices.</p><p>Here, we present a photoswitchable crystalline material, a MOF thin lm, whose proton-conduction properties can be massively altered. This MOF lm, prepared in a layer-by-layer fashion resulting in surface-mounted MOF (SURMOF), possesses photoswitchable spiropyran moieties. Spiropyran 17 is a stimuli-responsive molecule, which may reversibly isomerize to its open merocyanine form upon irradiation with UV light. It is intensively investigated for various applications such as lightresponsive glasses 18 and for information storage. 19 Since merocyanine is a zwitterionic molecule, the spiropyran-tomerocyanine isomerization goes along with a change of the dipole moment from about 5 to 16 Debye. 17b MOFs with spiropyran, attached to the scaffold or embedded in the pores, have been used for photoswitching the color of the material, 20 the uptake amount of the guest molecules 21 as well as the electron (hole) conductivity. 22 For the rst time, we take advantage of the large dipole moment change to dramatically modify the MOF properties, here the proton-conductance of the guest molecules. In addition to switching the alcoholic proton conduction, the concept is extended to water, the most important molecule for proton conduction applications. We demonstrate that the aqueous proton conductivity can be photomodulated by two orders of magnitude.</p><!><p>The surface-mounted metal-organic framework (SURMOF) thin lm was prepared on the substrate in a layer-by-layer fashion. The crystallinity of the sample was investigated by X-ray diffraction (Fig. 1). The X-ray diffractogram shows that the lm is crystalline with the targeted pillared-layer Cu 2 (e-BPDC) 2 (dabco) structure, 23 where e-BPDC refers to the 2ethynyl-[1,1 0 -biphenyl]-4,4 0 -dicarboxylic acid layer linker and dabco refers to the 1,4-diazabicyclo[2.2.2]octane pillar linker. In addition, the lm is grown mainly in [100] direction perpendicular to the substrate surface. Upon incorporating spiropyran in the parent SURMOF by post-synthetic modications (PSM), the X-ray diffractogram of the sample shows no signicant change in the positions and the intensity ratios of the diffraction peaks, indicating that the MOF lattice is unaffected by the PSM process. Noteworthy, the X-ray diffractograms of the sample before and aer performing the proton-conduction experiments (including loading with methanol, ethanol and water for 400 min each) are virtually identical (see Fig. 1c). This indicates that the SURMOF is stable under the used conditions.</p><p>The reaction yield of the PSM was investigated by infrared spectroscopy (Fig. SI2 †). The intensity of the ethynyl vibrational band at 3300 cm À1 decreased by 41.5%, which is correlated to the ethynyl groups that underwent ethynyl-azide click reaction anchoring the spiropyran moiety. Since each parent Cu 2 (e-BPDC) 2 (dabco) SURMOF unit cell possesses two ethynyl groups, in average, there are 0.83 photoswitchable moiety anchored in each MOF pore, i.e. almost 1 per pore, referred to as Cu 2 (SP-BPDC) 2 (dabco) SURMOF.</p><p>The distance between the van der Waals-surfaces of two opposing e-BPDC-linkers is approximately 1.0 nm, which is a measure for the size of the pore and pore window (see Fig. 1b). The total free pore volume of each MOF unit cell is approximately 1.5 nm 3 before the PSM. Aer PSM (with one spiropyran moiety per pore) the free pore volume of each unit cell decreases to approximately 1.2 nm 3 . Since the spiropyran moieties in the pore are exible and can rotate (or wiggle), e.g. around the bonding axis of the linker, the pore diameter and size of the pore window are not strongly affected by the PSM.</p><p>The photoswitching of the functional molecules is investigated by UV-vis spectroscopy (Fig. 2). The spiropyran-tomerocyanine isomerization of the molecules in ethanolic solution upon UV light irradiation (Fig. 2a) can be seen by the increase of the absorption band at approximately 550 nm. This band is correlated to the merocyanine isomer, being a clear indication of the photoisomerization. 17b,c The UV irradiation of the Cu 2 (SP-BPDC) 2 (dabco) SURMOF results in the evolution of an absorption band at approximately 580 nm, which is correlated to the merocyanine isomer (Fig. 2b). As a result of the different molecular environment in the SURMOF, the merocyanine absorption band is slightly red shied in comparison to the absorption band of the ethanolic solution. The photoisomerization is investigated in more details by infrared vibrational spectroscopy. The CO-spiro vibrational band only occurs for the spiropyran isomer and not for the merocyanine form, thus allowing the quantication of the photostationary state (PSS) upon UV irradiation (Fig. 2c). Based on the data, a switching yield of approximately 80% merocyanine was achieved upon 365 nm-UV-light irradiation.</p><p>The SEM images (Fig. SI3 †) show that the SURMOF homogenously covers the substrate and has a thickness of approximately 0.7 mm.</p><p>The ionic conduction properties of the Cu 2 (SP-BPDC) 2 (dabco) SURMOF are investigated by impedance spectroscopy. The empty sample shows a very small conductivity (Fig. SI4 †). Upon loading the sample with water from the gas phase, resulting in H 2 O@Cu 2 (SP-BPDC) 2 (dabco), the conductivity signicantly increases. The analysis of the impedance data in the Nyquist plot (Fig. 3) shows that the non-irradiated H 2 O@Cu 2 (SP-BPDC) 2 (dabco) sample, i.e. in the spiropyran form, has an ohmic resistance of approximately 3.45 AE 0.5 MU. This corresponds to a conductivity of 2.5 Â 10 À6 S m À1 . Upon UV-induced switching to the merocyanine form, the resistance of the H 2 O@Cu 2 (SP-BPDC) 2 (dabco) sample increases to 279.5 AE 19 MU. This means the spiropyran-to-merocyanine isomerization results to a conductivity decrease by a factor of 82. The conduction switching is fully reversible and the initial Nyquist plot is obtained aer thermal relaxation of the photoswitches.</p><p>In addition to the aqueous proton conduction, the charge transfer of other common proton-conducting molecules was also investigated. The Nyquist plots of the impedance of ethanol and of methanol loaded in Cu 2 (SP-BPDC) 2 (dabco) SURMOF are shown in Fig. SI5. † Both molecules show strong proton conduction properties in the SURMOF. By UV-light induced spiropyran-to-merocyanine isomerization of the host SURMOF, the ohmic resistance increases by more than one order of magnitude (Fig. SI5 †). The resulting proton conductivities of the sample in the spiropyran and merocyanine form are summarized in Fig. 4. It shows that the proton conductivity of , is a common model to analyse the proton conduction in nanoporous materials. 24 The determined parameters are presented in Table SI1. † all studied guest molecules can be photomodulated by more than one order of magnitude. The highest on-off ratio is reached for aqueous proton conduction.</p><p>The time course during the switching of 3 subsequent cycles is shown in Fig. 5. The absolute value of the impedance quickly increases upon starting the UV irradiation, resulting in the spiropyran-to-merocyanine isomerization. Upon 10 min irradiation, the impedance value increased by almost 2 orders of magnitude. Aer the irradiation, the sample relaxes back to the spiropyran state and the impedance value decreases to its initial value. The data shows that the cycling can be repeated and the behavior is reversible.</p><p>The 3 subsequent cycles have similar switching behavior. Thus, prominent photobleaching as previously observed for embedded spiropyran 22 seems to be oppressed. We speculate that the dimerization, which typically leads to the photobleaching of many spiropyran compounds, 17c is hindered by the anchoring at the MOF scaffold. Note, UV-induced photoconduction processes 25 or signicant electronic conduction due to spiropyran-to-merocyanine isomerization, 22 both increasing the conductivity, cannot be observed and have only a negligible inuence.</p><p>Infrared spectroscopy of the sample loaded with the guests was performed to gain deeper insights into the molecular mechanism of the proton-conduction photomodulation. To distinguish the vibration bands of the guests from the vibrations of the host MOF, D 2 O was used as guest molecules. Aer loading with D 2 O, apart from the D 2 O band at 2640 cm À1 , the corresponding infrared data do not show any signicant changes to the empty SURMOF in the spiropyran form (Fig. 6). However, upon UV-light induced spiropyran-to-merocyanine isomerization, a rather broad feature was observed at 1900-2300 cm À1 , which is characteristic for the formation of hydrogen bonds. 26 The broadening effect due to strong coupling makes an unambiguous identication of different types of hydrogen bonds (OD/O, OD/N, and ND/O) extremely difficult. Note, such hydrogen bonds are also not observed for the empty merocyanine-SURMOF (see Fig. 2c). The assignment of the hydrogen bonds is further supported by the observation of H 2 O-related hydrogen bonding at the high-frequency range. As shown in Fig. 6, a broad IR feature appeared at 2800-3300 cm À1 , which are typical for the hydrogen-bonded O-H and N-H stretching vibrations. Overall, the present IR results provide direct spectroscopic evidence that water guest molecules are strongly adsorbed via hydrogen-bonding formed only in the presence of the merocyanine-SURMOF.</p><p>Based on the small pore size, we tentatively assume that the proton conduction follows the Grotthuss mechanism. 27 The alignment of the water molecules, which allow relatively efficient Grotthuss-like proton transfer in the spiropyran-SURMOF, is disturbed by the strong hydrogen-bonding to the merocyanine-SURMOF, resulting in a substantially decreased proton mobility and, hence, in the smaller proton conductivity of water. These results are in agreement with previous infrared vibrational spectroscopy and density-functional theory calculation of butanediol and triazole in azobenzene-containing SUR-MOFs. 16 It should be note that a dense hydrogen-bond network generally would favor efficient, long-range proton transport, which is in contrast to the experimental ndings. In addition, the MOF, in both spiropyran and merocyanine forms, contains no ionizable protons. Both ndings support the fact that the proton-conduction-change results not from changes in hydrogen-bond networks but from the strong short-range binding of the guests.</p><p>The presented MOF with spiropyran moieties shows a tremendously larger switching effect in comparison to stateof-the-art photoswitchable MOFs, namely MOFs with azobenzene moieties. 16 Since the modulation of aqueous proton conduction was not yet published, we compare the protonconduction of alcohol guest molecules. While the conductivity of butanediol in azobenzene-MOFs is switched only by a factor of 1.5 (corresponding to a decrease by 35%), 16 the conductivity of ethanol in spiropyran-MOFs is photomodulated by a factor of 20. The 13-times-higher switching effect is explained by the different dipole moment changes of the photoswitchable components. While the dipole moment m of azobenzene changes only from 0 D (trans) to 3 D (cis), the dipole moment of spiropyran changes approximately from 5 D (spiropyran form) to 16 D (merocyanine form). The substantially larger dipole moment of merocyanine results in much stronger hydrogen bonds in comparison to the rather weak hydrogen bonds in the cis-azobenzene-MOF. As rough estimation of the bond-strengthchanges using Keesom dipole-dipole interaction with E $ m 2 (ref. 28) (and disregarding any further interaction of the guests with the host framework as well as the structural differences), the energy change in the spiropyran-MOF is 26 times larger than in the azobenzene-MOF. This value is in the same order of magnitude as the observed proton-conduction-photomodulation difference.</p><!><p>The thin MOF lms, referred to as surface-mounted MOFs (SURMOFs), 29 were prepared in a layer-by-layer fashion in a twostep process. First, the parent MOF thin lm with a pillaredlayer structure of type Cu 2 (e-BPDC) 2 (dabco) was prepared, where e-BPDC refers to 2-ethynyl-[1,1 0 -biphenyl]-4,4 0 -dicarboxylic acid 23b and dabco refers to 1,4-diazabicyclo(2.2.2)octan. The sample was synthesized by alternative immersion in the synthesis solutions, i.e. ethanolic 1 mM copper acetate solution and ethanolic 0.2 mM e-BPDC and dabco solution. The synthesis was performed by 100 cycles using a dipping robot. 30 By post-synthetic modications (PSM), the light-responsive spiropyran moieties, i.e. 2-(3 0 ,3 0 -dimethyl-6-nitrospiro[chromene-2,2 0 -indolin]-1 0 -yl)ethyl 2-azidoacetate (Fig. 1a and SI1 †), were anchored in the MOF pores by ethynyl-azide click reactions, 23b,31 resulting in Cu 2 (SP-BPDC) 2 (dabco) (Fig. 1d). PSM was carried out by immersing the SURMOF in the reaction solution of spiropyran in toluene with a concentration of 3 g per L for 24 hours at room temperature. Aer the initial immersion, the temperature of the solution was increased to 80 C for 9 days. Aerwards the sample was rinsed thoroughly with toluene and acetone to remove residual reactants.</p><!><p>XRD. X-ray diffraction (XRD) measurements in out-of-plane geometry with a wavelength of l ¼ 0.154 nm were carried out using a Bruker D8-Advance diffractometer equipped with a position-sensitive detector in q-q geometry.</p><p>UV-vis spectroscopy. The UV-vis transmission spectra were recorded by means of a Cary5000 spectrometer equipped with a UMA unit from Agilent. The spiropyran solution (Fig. 2a) had a concentration of 0.05 mg per ml and the cuvette thickness is 10 mm.</p><p>SEM. Scanning electron microscopy (SEM) images were recorded with a TESCAN VEGA3 tungsten heated lament scanning electron microscope. To avoid charging effects, the samples were coated with a thin ($2 nm) platinum lm using a LEICA EM ACE600 device.</p><p>Light irradiation. The samples were illuminated with a 365 nm-LED from PrizMatix with a power of about 40 mW cm À2 .</p><p>IRRAS. The infrared spectra were recorded with a Fourier-Transform Infrared Reection Absorption Spectrometer (FT-IRRAS) Bruker Vertex 80. The spectra were recorded in grazing incidence reection mode at an angle of incidence of 80 relative to the surface normal.</p><!><p>For the conduction measurements, interdigitated gold electrodes with a gap width of 10 mm and a total gap length of 1.69 m deposited on a quartz glass plate were used as SURMOF substrates. These substrates were purchased from Metrohm. The impedance spectra were measured using a Zurich Instruments MFIA Impedance Analyzer for a frequency range of 5 MHz to 1 Hz. The sample was placed in a home-made cell where the interdigitated gold electrodes on the substrate were contacted in a 2-probe way. The amplitude of the electric eld between the interdigitated electrodes is approximately 0.03 V mm À1 . The cell was purged with pure nitrogen or with nitrogen enriched with the vapor of the guest molecules with a ow rate of 100 ml min À1 . The water vapor had a relative humidity of 93%, corresponding a vapor pressure of 29 mbar. The vapor pressure of ethanol and methanol were approximately 75 mbar and 160 mbar, respectively. All experiments were performed at room temperature (298 K). Further details on the setup can be found in ref. 16. All conduction experiments were performed 3 times. The arrhythmic average value and the standard deviation is presented.</p><!><p>A nanoporous metal-organic framework (MOF) thin lm with photoswitchable spiropyran moieties is presented and the aqueous and alcoholic proton conduction of the guest molecules is investigated. As a result of the UV-light-induced reversible spiropyran-to-merocyanine isomerization in the host MOF, the bonding strength of the guest to the framework is substantially increased, decreasing their mobility. As a result, the proton conductivity of the guest is photomodulated by up to two orders of magnitude.</p><p>The study shows that by using more appropriate photoswitches than azobenzene, which is used for many proof-ofprinciple photoswitchable-MOF demonstrations, 32 the switching performance can be signicantly increased. Such materials with massively photomodulatable proton-conduction may nd application in switchable sensors 33 and allow the remote control of the interface to biological materials, in particular to ion conduction channels. 34 We foresee that further functionalization increasing the dipole moment of the photoswitch as well as MOF structure optimization will further increase the conduction on-off ratio.</p><!><p>There are no conicts to declare.</p>

Royal Society of Chemistry (RSC)

Biomarker Signatures of Mitochondrial NDUFS3 in Invasive Breast Carcinoma

We present evidence for potential biomarker utility of a mitochondrial complex I subunit, (NDUFS3) in discriminating normal and highly invasive breast carcinoma specimens obtained from clinical patients. Besides being a robust indicator of breast cancer aggressiveness, NDUFS3 also shows clear signatures of a hypoxia/necrosis marker in invasive ductal carcinoma specimens. Statistically significant positive correlation was observed between nuclear grade and NDUFS3 expression level in the tumor specimens analyzed. We support these findings with a plausible mechanism involving mitochondrial complex I assembly defects and/or redox buffering induced mitochondrial dysfunction during the process of cancer cell transformation. From a clinical standpoint, this novel observation adds value in augmenting the current receptor-based biomarkers for better accuracy in diagnosis and predicting survival rate in patients with breast carcinoma.

biomarker_signatures_of_mitochondrial_ndufs3_in_invasive_breast_carcinoma

1. Introduction<!>2.1 Tissue Immunofluorescence<!>2.2 Metabolic measurements<!>2.3 Image Acquisition and Analysis<!>2.4 Statistical Analysis<!>3.1. NDUFS3 expression is significantly higher in breast cancer tissues<!>3.2 NDUFS3 expression is preferentially upregulated in hypoxic/necrotic tumor sites<!>3.3. Increased NDUFS3 expression correlates with the extent of tumor aggressiveness<!>3.4 Redox Buffering / Mitochondrial assembly defects could contribute to the observed increased in NDUFS3 expression<!>4. Discussion<!>

<p>Ductal Carcinoma In situ (DCIS) is a histologically distinct, pre-invasive stage of breast carcinoma. Early detection of DCIS-associated precancerous lesions can significantly enable the complete removal of the lesions by lumpectomy or mastectomy – which can be further helped by radiation therapy where needed.[1,2] However from a clinical standpoint, even for the women who have been treated for DCIS, there is still difficulty in predicting the likelihood of DCIS recurrence and/or the likelihood of DCIS-invasive carcinoma transition. Biomarkers are qualitative and quantitative predictors of clinical outcomes based on a molecular understanding of the disease as well as the technological resources to interrogate the disease status with high sensitivity.[3,4,5] Current biomarkers for breast cancer include those that are either specific to receptor status (e.g., estrogen or herugulin receptors) and those that are generic to all type of cancer pathology (e.g., Ki67 positivity).[6,7] Despite the availability of these biomarkers, their utility is limited to cases which show specific receptor status and are not applicable to certain class of invasive breast cancers (e.g., triple negative breast cancers). A more prudent strategy will be to identify putative metabolic biomarkers that can target a larger ensemble of cancer transformation phenotype independent of their genetic background and/or receptor status – but targeting a more fundamental bioenergetic pathway of cancers.[8,9,10,11,12,13] In this short communication, we present evidence for potential metabolic biomarker utility of a mitochondrial complex I assembly subunit. NADH dehydrogenase [ubiquinone] iron-sulfur protein 3 (NDUFS3) is one of the precursor subunits in the 45-subunit mitochondrial complex I. As a catalytic subunit, it has been shown to play a vital role in the proper assembly of intact, functional complex I in the human mitochondrial respiratory chain.[14,15,16,17] Implications of complex I deficiency/dysfunction in a variety of neurodegenerative disorders have been documented.[14,18,19,20] However, there is no published evidence for its role in cancer pathology. In an attempt to identify mitochondrial role in breast cancer etiology and progression, we discovered that NDUFS3 subunit expression is a robust indicator of invasiveness in ductal and lobular carcinoma. Furthermore, NDUFS3 expression also was found to be significantly higher in hypoxic/necrotic regions of the breast cancer specimens from clinical patients (tissue arrays) thereby revealing its potential biomarker utility as a hypoxia marker in breast cancer and quite possibly, other cancer types as well. To the best of our knowledge, this is one of the first reports of mitochondrial/metabolic biomarker in breast carcinoma. We postulate a potential mechanism for our observations based on a cell culture studies in this paper.</p><!><p>Human breast cancer tissue arrays were purchased from Biomax US Inc., (Rockville, MD, USA). Tissue array slides were deparrafinized in xylene/ethanol and then heated in citrate buffer for antigen retrieval (10mM Sodium Citrate, pH 6) at 98°C for 45 minutes. After blocking (2%BSA/PBS at 37°C) for 30 minutes, primary antibody staining was done over night with the following working concentrations: NDUFS3 1:200; NDUFB8 1:200; Porin 1:1000 (Mitosciences, Eugene, OR). Fluorescent visualization was enabled by anti-mouse Alexa594 conjugated secondary antibody counterstained with nuclear dye, DAPI (50ng/ml). Similar protocol was adopted for two-dimensional cell cultures. For specific metabolic perturbation experiments reported in this study, MDA-MB-231and MCF-10A cells were seeded on 18 mm glass cover slips in 12 well plates (100,000–150,000 cells/well) and then treated with 50µg/ml and 100ug/ml chloramphenicol for 24 hr and in another experiment with 1µM ROT for 24 hr.</p><!><p>Cells grown in 12-well plate were treated with 0 –100µM H2O2 for 24 hours after which they were trypsinized and stained with trypan blue. Cell viability was determined by calculating the cell count and for each concentration and expressing as a percentage of total number of cells. Lactate measurement was done by measuring the extracellular lacte after incubating the cells for 1 hour in 100µl serum free media. Mitochondrial membrane potential was measured by incubating the cells with 200nM TMRM in PBS for 20 min and then analyzed by FACScan flow cytometer (BD Biosciences). For immunoblotting, 25µg protein was loaded on SDS-PAGE gel and transferred to PVDF membrane overnight. Membrane was then blocked with primary antibodies mouse NDUFS3 (1:2000; Mitosciences) and rabbit β-actin (1:5000; Abcam) overnight at 4°C. Next day, IR Dye labeled secondary antibodies (Antimouse 1:5000; Anti Rabbit 1:5000; 926–32221; Licor Biosciences) were used for visualization of bands using the Odyssey Scanner (Licor Biosciences). For SiRNA transfection, cells were seeded in 12 well plate (50,000 cells/well). Following 24 hour culture, 75ng SiRNA solution (NDUFS3_3; Gene ID 4722; Qiagen,USA), (NDUFS3_5; Gene ID 4722; Qiagen,USA), was added according to the manufacturer's instruction. The cells were incubated for 48 hours and then lactate measurement, TMRM FACS and western blot were done as described above.</p><!><p>Wide-field epifluorescence microscopy imaging systems (Olympus IX 70, Orca ER camera & Nikon AZ100, Nikon Qi camera) were employed in collecting all the images reported in this study. Appropriate filter cubes for collecting fluorescence from the specimen (DAPI filter: 360/40 nm excitation; 400 nm LP dichroic; 460/50 nm emission & Alexa 594 filter :595/30 nm excitation; 600 nm LP dichroic; 620/50 nm emission). Nuclear grade analysis was done by optimizing the DAPI labeling protocol for good signal-to-noise ratio as well as for rapid readout of the images. Supporting tumor cell proliferation studies were carried out by labeling the tissue arrays with Ki67 tagged with Alexa 488 fluorophore. Typical time of acquisition per image (1392 × 1040 pixels) was under 2 seconds. Tissue fluorescence images obtained by the aforementioned protocols were analyzed for three morphometric parameters namely, nuclear size, circularity and nuclear count. We recently demonstrated that nuclear area fraction is a reliable indicator of tumor aggressiveness.[21] This parameter yields a comprehensive picture of nuclear distribution which takes into account both the nuclear size/shape and the nuclear count. We extended this parameter in validating breast tissues arrays in the present study.</p><!><p>All statistical analysis was done using the open source statistical software R. Two sided t-test is used to compare two independent means. The assumption of normal distribution of the data is tested with the Shapiro-Wilk test. The distribution of NDUFS3 expression values by group was summarized in a box-plot diagram. The boxes represent 25th–75th percentiles of the NDUFS3 expression values in each group. The whisker bars represent the lowest and highest datum still within 1.5 interquartile range. The median values are indicated by the darker horizontal band. Pearson's correlation coefficient was used to estimate the correlation between nuclear area fraction and NDUFS3 expression levels.</p><!><p>Immunofluorescence studies in a series of breast tumor specimens obtained in a microarray format revealed a significantly higher NDUFS3 expression in invasive breast cancer specimens. Figure 1 shows representative immunofluorescence images from human breast cancer tissues. These initial studies were then extended to breast tissue array containing a larger ensemble of invasive ductal carcinoma (IDC) specimens. Supplemental Table 1 summarizes the complete list of tumor characteristics, patient age and tumor staging. The invasive ductal carcinoma specimens had the highest NDUFS3 expression whereas only moderate increase was observed in mildly aggressive mucinous carcinoma cases‥ Detailed statistical analysis of this data set (Figure 1b) further confirmed that invasive ductal carcinoma is associated with a significantly higher NDUFS3 expression. No specific exclusion of the available data was carried out so that the candid statistical analysis will reveal significance regardless of the heterogeneity in patient and tumor characteristics.</p><!><p>In order to understand the further ramifications of our initial observations, we analyzed all the IDC specimens and made a striking observation that the most significant increase in NDUFS3 expression in these specimens occurred only in hypoxic or necrotic regions of the tumor tissues. Figure 2 shows a representative set of images demonstrating this observation. Hypoxic and necrotic regions can be easily identified by the distinct nuclear morphometry where the apoptotic nuclei show a highly condensed nuclei (smaller size and increased DAPI fluorescence) whereas severely necrotic regions of the tumor show a high degree of nuclear fragmentation and a significant loss of DAPI fluorescence. As can be seen from these images, NDUFS3 expression has a preferential increase in the comedo-DCIS type and in particular, the hypoxic and necrotic regions. It is important to note that such necrotic regions do not show corresponding increase in another non-catalytic mitochondrial complex I subunit, NDUFB8 [inset of Figure 2(l)]. This comparison further confirms that observed increase in NDUFS3 is not an artifact of labeling but a genuine effect that has a plausible origin in the catalytic activity of mitochondrial complex I. Reduced oxygen availability in hypoxic regions can have either of two effects on the tumor cell mitochondria: (a) initiation of apoptosis due to hypoxic signals and (b) mitochondrial adaptation by reprogramming the translational and transcriptional apparatus so that the concomitant onset of mitochondrial assembly defects/dysfunction further exacerbate tumor glycolysis (Warburg phenotype).[20,22,23,24,25,26] Our data in Figure 2 suggests that mitochondrial defects and the resulting NDUFS3 expression profile could be associated with the progressive increase in tumor apoptosis and necrosis. Together these observations point out to the fact that NDUFS3 expression levels are strong indicators of tumor aggressiveness and also tumor hypoxia.</p><!><p>We recently reported a nuclear grade imaging approach to quantify the extent of tumor aggressiveness.[21] As shown in Figure 2 (and supplemental Figure S1), an increase in nuclear grade and Ki67 expression coincided with the extent of NDUFS3 expression profile thereby suggesting that NDUFS3 expression is linked with the extent of tumor aggressiveness. Figure 3a shows the nuclear grade analysis performed from the images in Figure 2. These data support the currently used nuclear grade as a morphological parameter in tumor staging. By considering the nuclear grade (nuclear area fraction) as the commonly used "gold standard", we further established that the NDUFS3 expression in these specimens correlates very well with the nuclear area fraction (Pearson's correlation coefficient −0.849). Together these results validate NDUFS3 expression as a robust indicator of aggressive tumor phenotype in clinical specimens.</p><!><p>In order to understand the plausible sources of the observed NDUFS3 expression profile in aggressive cancer tissues, we resorted to cell culture based system where we compared the non-tumorigenic, normal epithelial cells (MCF10A) and tumorigenic, metastatic breast carcinoma cells (MDA-MB-231). As shown in Figure 4a, the untreated breast cancer cells (MDA231) have significantly higher NDUFS3 expression as compared with the untreated normal epithelial cells. Since NDUFS3 is a precursor subunit in the mitochondrial Complex I assembly, we suspected that an increase in this subunit expression could either arise from translational defects and/or mitochondrial assembly defects. In order to verify if inhibition of mitochondrial translation could lead to an accumulation of NDUFS3, we treated the normal and breast cancer cells with 100µg/ml chloramphenicol for 24 hours and monitored the NDUFS3 expression by immunofluorescence. Inhibition of mitochondrial translation led only to a modest increase in NDUFS3 expression. Gene silencing of NDUFS3 with SiRNA led to distinct metabolic changes in MDA231 cells further confirming that mitochondrial function in the cancer cells are more susceptible to NDUFS3 levels and vice versa (Figure 4b).</p><p>Another source of constitutive mitochondrial dysfunction is the altered redox poise in cancer cells. In fact, chemoresistant cancer cells escape cell death induced by reactive oxygen species (ROS) by modulating their redox status (e.g., glutathione-induced redox buffering) which in turn offers the cancer cells an increased tolerance to ROS.[25,28] To exacerbate this ROS-induced mitochondrial damage, cancer cells also have another distinct biochemical hallmark namely, the apparent increase in glycolytic activity over the mitochondrial activity owing to their metabolic transformation. This phenomenon, Warburg effect, has been well documented in earlier publications.[20,27] An immediate consequence of Warburg phenotype in cancer cells is their ability to evade mitochondrial apoptosis and/or upregulation of their glycolytic potential even in the presence of oxygen (aerobic glycolysis). Figures 4c&d show representative data demonstrating that MDA231 cells have an increased tolerance to ROS insult (both exogenous and endogenous) as compared with normal MCF10A cells. Exogenous ROS inducer hydrogen peroxide could decrease the viability of MCF10A cells more drastically than the cancer cells. Interestingly, mitochondrial complex I inhibition by rotenone (endogenous ROS inducer) leads to significant change in mitochondrial structure (assembly defects and/or functional defects) in normal cells. On the other hand, even the untreated cancer cells have altered structure as detected with mitochondrial porin immunofluorescence and rotenone treatment did not cause any significant change in mitochondrial structure suggesting an increased tolerance of the cancer cells to ROS. This data further illustrates that an altered mitochondrial structure/function in cancer cells could be the potential mechanism by which these cells show a constitutively high NDUFS3 expression as a result of mitochondrial assembly defects.</p><!><p>Common wisdom predicts an intricate relationship between structure and function of any cellular component and mitochondrial paradigm is no exception. Another layer of complexity arises from the fact that mitochondrial biogenesis is governed both by nuclear and mitochondrial genomes. A compromise in proper assembly of these individual complexes can lead to mitochondrial dysfunctions. Last few decades of research in understanding Warburg effect and the concomitant metabolic alterations in cancers have led to a consensus that mitochondrial dysfunction could be vital in contributing to breast cancer etiology. NDUFS3 has been shown earlier to be a critical regulator of mitochondrial complex I assembly. Our results indicate that NDUFS3 has a novel role in discriminating invasive ductal carcinoma from normal breast tissues. Since our data were obtained from clinically relevant, human breast cancer patients, these observations add further credibility to the hypothesis that mitochondrial dysfunction arising from complex I assembly defects could be a viable source of metabolic deregulation in cancer. A recent study reported that mitochondrial translation inhibition led to defects in mitochondrial ND subunits which in turn, led to a significant accumulation of the complex I precursor subunit, NDUFS3.[29] Even though our studies did not address the role of mitochondrial ND subunits in the normal and cancer cells, the cell-based studies certainly point to the direction where NDUFS3 expression can be modulated by the altered metabolic status of the cancer cells as well as by alterations in mitochondrial assembly. We speculate that with an increase in tumor aggressivenss and the onset of hypoxia, mitochondrial assembly defects can be more common thereby leading to the observed increase in NDUFS3 expression in invasive ductal carcinoma specimens.</p><p>Current clinical practice for staging the tumors involves nuclear grade and the extent of hypoxia. The finding that catalytic NDUFS3 expression (but not the non-catalytic NDUFB8 expression) is significantly high in hypoxic/necrotic regions of the tumor bears an unique place in tumor staging. We recently demonstrated that it is possible to directly image nuclear morphometry and tissue topology for tumor margin assessment in lumpectomy specimens in preclinical animal models.[21] The distinct correlation between nuclear area fraction ( a measure of cancer invasiveness) and NDUFS3 expression (Pearson's correlation coefficient =0.84) further validates our conclusions. It will be intriguing to see if the observed hypoxic profile of NDUFS3 expression is an adaptive response of mitochondria to oxygen availability in comedo-necrosis phase of the ductal carcinoma. We believe that a systematic understanding of molecular mechanisms behind the observed NDUFS3 upregulation (and possibly mitochondrial complex I deregulation) could enable us in predicting the progressive transition from precancerous ductal carcinoma in situ (DCIS) to invasive forms of breast carcinoma thereby adding value to the current methods of estimating survial rate and prognosis in patients being treated for DCIS. A mechanistic study of our present observations is beyond the scope of this short communication. However, our preliminary in vitro metabolic measurements in normal epithelial cells and metastatic carcinoma cell lines point out to plausible roles of redox buffering and/or mitochondrial translation defects in contributing to the high expression in NDUFS3 in invasive breast carcinoma. A more detailed mechanistic exploration of the present findings will be the subject of a future preclinical animal study in the laboratory.</p><p>In conclusion, we have uncovered a novel, biomarker potential of a mitochondrial complex I subunit protein, NDUFS3 – as a robust indicator of breast cancer progression and invasiveness as well as of hypoxia/necrosis in clinical specimens of invasive ductal carcinoma. We envision that NDUFS3 and other potential metabolic biomarkers can be augmented with the currently available biomarkers so that a multiparmetric, biomarker assay panel (genetic and metabolic) can be developed to yield a better accuracy in prediction of the different stages of breast cancer, therapeutic monitoring and better prognosis.</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

Protein fold determined by paramagnetic magic-angle spinning solid-state NMR spectroscopy

Biomacromolecules that are challenging for the usual structural techniques can be studied with atomic resolution by solid-state nuclear magnetic resonance. However, the paucity of >5 \xc3\x85 distance restraints, traditionally derived from measurements of magnetic dipole-dipole couplings between protein nuclei, is a major bottleneck that hampers such structure elucidation efforts. Here we describe a general approach that enables the rapid determination of global protein fold in the solid phase via measurements of nuclear paramagnetic relaxation enhancements (PREs) in several analogs of the protein of interest containing covalently-attached paramagnetic tags, without the use of conventional internuclear distance restraints. The method is demonstrated using six cysteine-EDTA-Cu2+ mutants of the 56-residue B1 immunoglobulin-binding domain of protein G, for which ~230 longitudinal backbone 15N PREs corresponding to ~10-20 \xc3\x85 distances were obtained. The mean protein fold determined in this manner agrees with the X-ray structure with a backbone atom root-mean-square deviation of 1.8 \xc3\x85.

protein_fold_determined_by_paramagnetic_magic-angle_spinning_solid-state_nmr_spectroscopy

<!>Determination of 15N paramagnetic relaxation enhancements in GB1 mutants by solid-state NMR spectroscopy<!>Validation of the utility of solid-state NMR 15N PRE restraints for protein structure determination<!>De novo determination of the GB1 backbone fold<!>Discussion<!>Sample preparation<!>NMR spectroscopy<!>Structure calculations

<p>The determination of three-dimensional (3D) structures of biological macromolecules is critical to understanding their physiological function and is also the primary focus of structural genomics initiatives.1 Magic-angle spinning (MAS) solid-state nuclear magnetic resonance (NMR) provides a set of spectroscopic tools for structural studies of biological systems that cannot be induced to form sufficiently diffracting crystals and are also not amenable to solution NMR due to size limitations, such as certain membrane-bound polypeptides2,3 and amyloids.4,5 However, in spite of the remarkable recent progress in determining relatively high-resolution structural models for proteins with molecular weights up to ~20 kDa by MAS solid-state NMR,5-19 the routine generation of such structures continues to present considerable challenges.</p><p>Even for small to medium-size uniformly 13C,15N or uniformly 15N and extensively 13C labeled proteins—the latter generally prepared by using complementary [1,3-13C] and [2-13C]glycerol based expression media6 in order to achieve improved spectral resolution and magnetization transfer efficiencies6,15 by suppressing many one-bond 13C–13C dipolar and J-couplings—the 2D 13C-13C and 15N-13C chemical shift correlation spectra, which form the basis for the vast majority of the current solid-state NMR structure determination protocols, are typically highly congested and contain many tens to hundreds (or even thousands) of resonances. These spectra rely on magnetization transfers using radiofrequency pulse schemes such as transferred echo double resonance (TEDOR),20,21 proton driven spin diffusion (PDSD) or dipolar assisted rotational resonance (DARR),22 CHHC, NHHC or NHHN,23 or proton assisted recoupling (PAR),12 and report on the magnitudes of through-space 13C-13C, 15N-13C and 1H-1H magnetic dipole-dipole couplings, and thereby distances, among the protein nuclei. Though the use of ultrahigh-field (≥ ~20 Tesla) NMR spectrometers, additional spectral dimensions and tailored isotope labeling schemes (e.g., perdeuterated proteins containing sparsely distributed proton nuclei at a limited number of amide and methyl sites),18-20 as well as structure calculation algorithms that are able to deal with ambiguous distance restraints8,9 has certainly helped to alleviate some of the problems related to NMR signal overlap, the fundamental limitation in most solid-state NMR structure determination endeavors has been the relative scarceness of unambiguous long-range interatomic distance restraints. This is primarily because dipolar couplings between 1H, 13C and 15N spins, which scale with the inverse third power of the internuclear distance, become vanishingly small for distances in the ~5-10 Å regime and beyond (typical coupling magnitudes are on the order of a few Hertz to tens of Hertz). The detection and reliable quantification of these couplings is often further complicated by the fact that they are part of tightly coupled multiple spin networks. Consequently, correlations that encode information about the most structurally-relevant inter-residue dipolar couplings typically comprise only a small fraction of the observable resonances in multidimensional solid-state NMR datasets, and are also usually associated with some of the weakest cross-peak intensities.</p><p>In this article, we describe a general approach that has the potential to substantially accelerate the solid-state NMR protein structure determination process. In broad terms, this methodology involves the measurement of nuclear paramagnetic relaxation enhancements (PREs)24,25 in several point mutants of the protein of interest modified with covalently-attached paramagnetic tags, followed by the application of these PREs as structural restraints to establish the global protein backbone fold. The PRE phenomenon is well-known in magnetic resonance—the magnitude of the effect scales with the inverse sixth power of the electron-nucleus distance24,25 and it can be non-negligible for distances up to ~20-25 Å or longer depending on the observed nucleus and paramagnetic center (i.e., distances that are at least a factor of ~3-4 larger than those obtained by using traditional dipolar coupling based solid-state NMR techniques). Indeed, PRE-based methods of this kind, employing nitroxide spin labels and measurements of longitudinal or transverse amide 1H relaxation enhancements, have previously been successfully used to define the fold of soluble proteins with limited nuclear Overhauser effect (NOE) data.26-28</p><p>The latest advances in MAS NMR spectroscopy of paramagnetic proteins29-36 have laid the foundation for application of the PRE-based approach to the structural analysis of biological macromolecules in the solid phase. Using several 13C,15N-labeled analogs of a model 56-residue protein, B1 immunoglobulin binding domain of protein G (GB1), containing EDTA-Cu2+ tags attached to cysteine residues introduced into the protein especially for this purpose,37,38 we have recently demonstrated that residue-specific longitudinal PREs can be rapidly determined for most backbone amide 15N nuclei by multidimensional solid-state NMR methods.34,35 These 15N PREs, which correspond to ~10-20 Å 15N-Cu2+ distances, were found to be in good quantitative agreement with those predicted for protein structural models based on the wild-type (WT) GB1 fold. Here we show that a set of ~230 such PREs recorded for six Cys-EDTA-Cu2+ mutants of GB1 (i.e., ~4-5 restraints per residue), supplemented only by standard chemical shift based restraints on secondary structure and, importantly, in the absence of any dipolar coupling based internuclear distance restraints, is sufficient to derive the correct global fold for GB1 in a de novo fashion. Notably, since PRE restraints can be easily extracted with high-sensitivity from routine 2D or 3D solid-state NMR chemical shift correlation spectra, this methodology is expected to become widely applicable to larger proteins available in limited quantities.</p><!><p>Residue-specific backbone amide 15N longitudinal PREs, Γ N1, were determined from measurements of longitudinal 15N relaxation rate constants using series of 2D 15N-13CO (NCO) spectra recorded with different values of the relaxation delay, τrelax (see Fig. 1a and Supplementary Fig. S1), as described in the Methods section. Relaxation data were collected for six pairs of GB1 mutants containing non-native Cys-EDTA-Cu2+ or Zn2+ sidechains incorporated at solvent exposed sites corresponding to residues N8, E19, K28, E42, D46 or T53 in the WT protein (Fig. 1b); all of these GB1 analogs exhibit the WT fold as judged by the similarity of solution and solid-state NMR backbone chemical shifts to those of WT GB1.33-35 For brevity, the GB1 mutants used in this study are labeled as 8EDTA-Cu2+/Zn2+, 19EDTA-Cu2+/Zn2+, 28EDTA-Cu2+/Zn2+, 42EDTA-Cu2+/Zn2+, 46EDTA-Cu2+/Zn2+ and 53EDTA-Cu2+/Zn2+, according to location of the Cys-EDTA-Cu2+/Zn2+ tag.</p><p>In Fig. 1c and 1e we show representative small regions of 2D NCO spectra for 8EDTA-Cu2+/Zn2+ and 28EDTA-Cu2+/Zn2+ recorded with τrelax values of 100 μs (reference spectra) and 4 s. While the reference spectra for the corresponding EDTA-Cu2+ and Zn2+ proteins are effectively indistinguishable from one another and contain all of the expected intense one-bond 15N-13CO correlations, a number of cross-peaks are severely attenuated or altogether missing in spectra recorded for the EDTA-Cu2+ proteins with longer relaxation delays (τrelax ≈ 2-4 s). The most attenuated resonances (for example G9, G14 and T53 in 8EDTA-Cu2+ and T25, C28 and Q32 in 28EDTA-Cu2+) are invariably associated with residues located in the spatial proximity of the paramagnetic tag.34,35 This is to be expected based on the Solomon dipolar relaxation mechanism,24,25 and previously we have demonstrated that most experimental solid-state NMR 15N PREs and 15N-Cu2+ distances determined in this fashion are in quantitative agreement with the corresponding values predicted using protein structural models based on the WT GB1 fold.34,35 Fig. 1d and 1f show the complete experimental relaxation trajectories and best fits to decaying single exponentials (i.e., I = A·exp{–R1 × τrelax}, where I is the instantaneous cross-peak intensity, A is the peak intensity at τrelax = 0, and R1 the 15N longitudinal relaxation rate constant; the two fit parameters are A and R1) for representative residues in 8EDTA-Cu2+/Zn2+ and 28EDTA-Cu2+/Zn2+, respectively.</p><p>Collectively, these experiments yield 231 amide 15N longitudinal PRE restraints (Supplementary Table S1) for the six paramagnetic GB1 mutants (i.e., an average of ~38 out of 55 possible restraints per mutant or ~4-5 restraints per residue; the remaining PREs could not be determined reliably due to peak overlap in 2D NCO spectra). Approximately half of the PRE restraints (110 of 231) correspond to Γ1N values greater than 0.1 s-1, and were used as is in the course of the protein structure calculations discussed below. The remaining 121 restraints were converted to purely repulsive 'NOE-type' distance restraints preventing the associated atoms from approaching closer than 15.1 Å. This was done in order to avoid potential complications with the convergence of the calculated structures, associated with small but detectable effects on the extracted Γ1N values stemming from the presence of intermolecular 15N-Cu2+ couplings and secondary protein Cu2+ binding sites.36</p><!><p>In order to evaluate the ability of the longitudinal solid-state NMR 15N PREs to determine the correct protein fold in the absence of traditional dipolar coupling based distance restraints, we performed a set of preliminary, 'ideal case' scenario calculations for GB1 in Xplor-NIH39 using protocols where all native sidechains as well as the backbone atoms for regular secondary structure elements were frozen in rigid bodies corresponding to their conformations in the 1.14 Å GB1 X-ray structure40,41 (PDB entry 2GI9), and the Cys-EDTA-Cu2+ sidechain conformations (initially optimized using PRE data and 2GI9 atomic coordinates) were also fixed. The regular secondary structure elements included the α-helix (residues 22-37) and the four β-strands, β1 to β4, spanning amino acids (aa) 2-8, 13-19, 42-46 and 51-55, respectively. The backbone torsion angles for the remaining residues were randomized. Moreover, to enable the simultaneous use of all PRE restraints within the same structure calculation protocol, the Cys-EDTA-Cu2+ tags were placed on all six modified residues (even though separate NMR experiments were used to record PRE data for each of the GB1 mutants) with the concurrent disabling of interactions between atoms associated with the different tags.</p><p>In Fig. 2a and 2b we show representative sets of 10 lowest energy structures (out of a total of 1000 structures) calculated without and with the use of solid-state NMR longitudinal 15N PRE restraints, respectively, in addition to the more conventional potential energy terms described in the Methods section. Not surprisingly, when no experimental NMR data were incorporated into the calculations, the lowest energy protein structures obtained in Xplor-NIH are effectively random and show little resemblance to one another or to the GB1 X-ray structure (Fig. 2a). Within this set of 10 lowest energy structures, no structure, including the one displaying the smallest backbone atom (C′, Cα, N and O) coordinate root-mean-square deviation (RMSD) of 3.4 Å relative to 2GI9, correctly predicts the protein fold. Though occasionally (3 times out of 1000 for this dataset) the calculations with no experimental restraints included were able to yield protein structures with the correct GB1 fold (backbone RMSDs versus 2GI9 of ~1.5-2 Å), these 'correct' structures were associated with energies that were considerably higher than those obtained for numerous 'incorrect' structures having backbone atom RMSDs in the ~3-9 Å range. In stark contrast, analogous calculations which also include the 231 experimental 15N PRE restraints yield lowest energy protein structures that cluster tightly together, and, most importantly, correspond to the correct GB1 fold with backbone RMSDs of 0.9-1.8 Å relative to 2GI9 (Fig. 2b).</p><p>The complete results for both sets of calculations employing large ensembles of 1000 random starting protein structures are summarized in Fig. 2c and 2d. These data clearly illustrate that in the absence of 15N PRE restraints there is effectively no correlation between the total energy of a particular protein structure obtained in Xplor-NIH and that structure displaying the correct GB1 fold (Fig. 2c). Indeed, nearly all structures obtained in this manner bear little resemblance to the reference GB1 X-ray structure. On the other hand, the addition of 15N PRE restraints to the structure calculation protocol leads to a funnel-like effect where a relatively strong correlation emerges between the total energy of a calculated structure and its RMSD relative to the 2GI9 backbone atom coordinates (Fig. 2d). While the convergence properties of these calculations may not be ideal and the calculation procedure generates multiple incorrect protein structures, it is important to note that all of these incorrect structures have high energies and can be readily identified. Most significantly, however, the lowest energy structures (associated with total energies of ~3000 kcal/mol or less in this case; the 10 lowest energy structures are indicated by red circles in Fig. 2c and 2d) all possess similar backbone folds that also closely match the GB1 X-ray structure (Fig. 2b).</p><!><p>Having established the utility of solid-state NMR 15N PRE restraints for protein structure calculations, we proceeded to investigate whether the fold of GB1 could be determined in a de novo manner by using only 15N PRE data (Supplementary Table S1) and standard solid-state NMR chemical shift based backbone dihedral angle restraints for WT GB1 (Supplementary Table S2) to define local secondary structure. As described in detail in the Methods section, these realistic de novo calculations consisted of a two stage protocol. In the initial stage, backbone atoms for residues comprising regular secondary structure elements (identified within the torsion angle likelihood obtained from shifts and sequence similarity, TALOS+, software42 as having α or β conformation with greater than 85% confidence) were frozen in rigid bodies with their dihedral angles fixed to the ideal TALOS+ predicted values, while all other backbone and sidechain torsion angles (including the Cys-EDTA-Cu2+ tags) were randomized. We found that fixing the TALOS+ identified secondary structure regions during this initial stage resulted in significantly improved convergence relative to the standard protocols which allow more degrees of freedom, as well as improved accuracy of the final calculated structures. The initial calculation stage was followed by a second stage, in which the 10 lowest energy structures from the initial calculations (indicated by red circles in Fig. 3a) were further refined using 10 different sets of randomized velocities in a protocol that allowed all sidechains to move as well as all backbone dihedral degrees of freedom constrained only by the TALOS+ dihedral angle potential with minimum dihedral angle uncertainties of ±20°, to yield a final set of 100 structures. From this set, the 20 lowest energy structures, shown in Fig. 3b and indicated by yellow circles in Fig. 3a, were used to obtain a regularized mean structure.</p><p>Remarkably, using these largely unrestricted de novo calculations, which did not utilize any conventional dipolar coupling based interatomic distance restraints or structural information derived from X-ray diffraction data, we were able to determine a backbone fold for GB1 that is in good agreement with the 1.14 Å X-ray structure (Fig. 3c; structure indicated by the magenta circle in Fig. 3a). The mean solid-state NMR structure obtained in this manner (blue) displays a backbone atom RMSD of 1.8 Å and an all heavy atom RMSD (excluding sidechains of residues 8, 19, 28, 42, 46 and 53) of 2.7 Å relative to the 2GI9 reference structure (red), and the backbone atoms for the vast majority of residues in the X-ray structure (with exception of several aa located in the loop regions or near the N-terminus) fall within the conformational space occupied by the ensemble of the 20 lowest energy conformers (gray cloud). Finally, in the course of these calculations we have also considered the possible influence of Cys-EDTA-Cu2+ sidechain disorder on the quality of the derived structures. The PRE refinement facilities within Xplor-NIH allow such structural heterogeneity to be modeled using ensembles of several sidechains, by either representing the PRE as a sum of contributions from each ensemble member (corresponding to slow interconversion between ensemble members) or 1/r6 averaging of the different ensemble members (for fast interconversion), and both approaches were attempted using ensembles of two and three Cys-EDTA-Cu2+ tags. We found that the use of such ensembles did not improve the results relative to using a single Cys-EDTA-Cu2+ conformer, presumably due to the presence of additional degrees of freedom for which there are insufficient data in this study.</p><!><p>Notwithstanding the tremendous current progress,5-19 the routine determination of protein structures by solid-state NMR remains a formidable task mainly due to challenges associated with the detection of large numbers of correlations that report on >5 Å internuclear distances. Measurements of such long-distance restraints in paramagnetic proteins, including PREs33,34 and pseudocontact shifts (PCS),30,31 have the potential to significantly accelerate and enhance solid-state NMR protein structure elucidation efforts. In fact, the utility of paramagnetic restraints of this type was recently illustrated by Bertini and co-workers for a 159-residue (~17.6 kDa) native metalloprotein, cobalt(II)-substituted matrix metalloproteinase 12,31,32 where over 300 13C PCS restraints were used in conjunction with ~30031 and ~80032 internuclear distances obtained from PDSD/DARR, CHHC and PAR experiments to arrive at 3D protein structures.</p><p>The approach toward global protein fold determination described here utilizes PRE restraints recorded for structural analogs of the protein of interest containing covalently-linked paramagnetic tags, and is readily applicable to a wide variety of protein molecules in the solid phase, including natively diamagnetic proteins. A major advantage of this approach, in addition to the fact that nuclear PREs yield almost exclusively structurally-relevant inter-residue restraints generally corresponding to >10 Å distances, is that these PRE restraints are very straightforward to extract by monitoring cross-peak intensities as a function of a relaxation delay in the simplest 2D or 3D solid-state NMR chemical shift correlation spectra that contain a minimum number of signals arising from adjacent, strongly dipolar coupled nuclei in the protein. As such, this methodology is pertinent to larger protein systems. While the implementation of this method requires the preparation of several point mutants of the protein under study, such biochemical manipulations are routine nowadays. It is also important to note that although the known 3D structure of GB1 was helpful in identifying suitable sites for introduction of the Cys-EDTA-Cu2+ tags in this study, the incorporation of such tags into proteins of unknown structure, with minimal influence on the backbone fold,38 should be readily implementable using an approach discussed by Battiste and Wagner27 that utilizes secondary structure information obtained from chemical shifts for the WT protein and the analysis of hydrophilicity patterns to identify the most probable solvent-accessible residues.43</p><p>The current study demonstrates that ~4-5 longitudinal amide 15N PRE restraints per residue, collected using a condensed NMR data acquisition approach35 for a set of six Cys-EDTA-Cu2+ mutants available in small quantities (~150-200 nanomoles of 13C,15N-labeled protein per sample), are sufficient to obtain a protein backbone fold for GB1 that is in close agreement with the X-ray structure without the use of any conventional internuclear distance restraints. Although the global protein fold determined in this manner does not correspond to the highest achievable resolution structure, it can be derived very rapidly and provides a reasonable structural model that is likely to be sufficient for many applications. Moreover, such a PRE-based model can serve as a major aid for resolving assignment ambiguities in conventional dipolar coupling based solid-state NMR spectra, and can readily be further refined to higher resolution, if desired, by incorporating additional interatomic distance restraints or by combining the PRE restraints with more advanced fragment-based chemical shift/molecular mechanics (CSMM) based structure calculation protocols such as CHESHIRE13 or CS-ROSETTA.14 It should be noted that the structure of GB1 has, in fact, been successfully solved (and to higher resolution) using the CHESHIRE approach with chemical shifts as the sole experimental input,13 whereas in this work we only extracted broad backbone torsion angle ranges from chemical shift data. While such CSMM methods, which utilize molecular mechanics force fields to obtain the proper protein fold, have chalked up an impressive record of correctly determined protein structures and we strongly advocate their use, they are also known to fail at times due to inadequacies of the MM force fields and the fact that chemical shift data contain essentially no long-distance information. In contrast, the PRE-based approach reported here uses direct distance restraints simply related to experimental observables, and thus serves as an independent means to solve or validate a protein fold.</p><p>Importantly, the number of paramagnetic-based structural restraints in these studies can be dramatically increased, by at least ~2-3-fold, with minimal effort by measuring other types of PREs, including longitudinal backbone and sidechain 13C PREs and transverse PREs for the different nuclei, in the same protein samples, as well as PCS restraints for 13C, 15N and potentially 1H nuclei in analogous Cys-EDTA-Co2+ samples. This is expected to substantially enhance the resolution of the derived structures, and bodes well for the extension of this paramagnetic solid-state NMR methodology to larger proteins for which a significant fraction of resonances can be assigned. Indeed, such sequential resonance assignments in larger systems may be facilitated by using spectral editing approaches that take advantage of the presence of the paramagnetic tags.</p><!><p>Plasmid cDNAs encoding for N8C, E19C, K28C, E42C, D46C or T53C mutants were constructed as described previously,33 using T2Q-GB1 cDNA (referred to as GB1) and Stratagene QuikChange II site-directed mutagenesis protocol. Proteins were expressed in E. coli BL21(DE3) with Luria-Bertani medium or M9 minimal medium containing 1 g/L 15NH4Cl and 3 g/L 13C-glucose (Cambridge Isotope Laboratories) and purified by gel filtration chromatography.33,34 Cysteines were reacted with the sulfhydryl-specific reagent N-[S-(2-pyridylthio)cysteaminyl]EDTA37 (Toronto Research Chemicals), pre-loaded with 1.1 equivalents of Cu2+ or Zn2+ (diamagnetic control). Protein microcrystals for solid-state NMR were prepared by co-precipitating the 13C,15N-EDTA-Cu2+/Zn2+ proteins with natural abundance GB1 in a 1:3 molar ratio using microdialysis at 4 °C and a 2:1:1 (v/v) 2-methylpentane-2,4-diol/isopropanol/deionized water precipitant solution.34,36 Samples, each containing ~1.0-1.2 mg (~150-200 nanomoles) of 13C,15N-labeled protein, were packed into 1.6 mm zirconia rotors (Agilent Technologies) by centrifugation.</p><!><p>NMR spectra were recorded on a 500 MHz Varian spectrometer equipped with a 1.6 mm FastMAS™ probe. MAS rate and sample temperature were regulated at 40,000 ± 20 Hz and ~5 °C, respectively. Residue-specific amide 15N longitudinal relaxation rate constants, R1, for the EDTA-Cu2+/Zn2+ GB1 mutants were determined from series of 2D NCO spectra recorded using the pulse scheme in Fig. 1a with relaxation delays, τrelax, of 100 μs to 4 s. Spectra were collected with acquisition times of 25.6 ms (t1) and 30 ms (t2) and recycle delays of 0.4 s and 1.3 s for EDTA-Cu2+ and EDTA-Zn2+ proteins, respectively. The total experiment time required to record 15N R1 data for all EDTA-Cu2+/Zn2+ GB1 mutants was ~18 days. Spectra were processed using NMRPipe44 and analyzed with home-built software (available at http://code.google.com/p/nmrglue). 15N PREs, Γ1N, were calculated as Γ1N = R1(Cu2+) – R1(Zn2+), where 15N R1 values for Cu2+ and Zn2+ proteins were obtained by fitting residue-specific relaxation trajectories to decaying single exponentials.</p><!><p>Structure calculations were carried out in Xplor-NIH39 with a protocol in which backbone atoms of regular secondary structure elements were frozen in rigid bodies corresponding to their X-ray (during initial calculations in Fig. 2) or ideal target conformations as predicted from TALOS+42 (during the realistic de novo calculations in Fig. 3). Regions of secondary structure were identified by inspection of the GB1 X-ray structure40,41 (PDB entry 2GI9) for the initial calculations and from TALOS+ hits predicted with >85% confidence in the α or β Ramachandran regions for the realistic calculations. The remaining backbone torsion angles were randomized. Sidechain torsion angles were fixed in their X-ray conformation for the initial calculations and randomized for the realistic calculations. The EDTA-Cu2+ tags were present on all six residues which were modified, but inter-residue interactions of the associated atoms were disabled. In the initial calculations, conformations of the EDTA-Cu2+ sidechains were first optimized using PRE data and the 2GI9 coordinates and held fixed during structure calculations. For the realistic calculations, EDTA-Cu2+ sidechains were treated as other sidechains with torsion angles randomized.</p><p>PRE restraints consisted of explicit restraints using the PREPot term45 with an electron correlation time of 3 ns25 for Γ N1 values above a cutoff of 0.1 s-1 (Supplementary Table S1). Smaller PREs were included using a purely repulsive 'NOE-type' term restraining the associated 15N-Cu2+ distances to values >15.1 Å. TALOS+ was used to derive 102 (of 110 possible) backbone torsion angle restraints from GB1 13C and 15N solid-state NMR chemical shifts (Supplementary Table S2). The radius of gyration term46 was used to achieve appropriate protein packing density. Additional knowledge-based energy terms included the torsion angle potential of mean force,47 the hydrogen bond potential of mean force,48 and the low resolution residue contact term.49 Standard bond, bond angle and improper torsion angle restraints were used to maintain proper covalent geometry, and a repulsive quartic van der Waals term used to prevent atomic overlap. During calculations in which EDTA-Cu2+ sidechains were allowed to move, the associated dihedral angles were restrained by a dihedral energy term.</p><p>The protocol consisted of initial conjugate gradient minimization without the PRE and repulsive distance terms, followed by minimization including all energy terms. This was followed by variable step-size dynamics for the smaller of 500 ps or 20,000 steps at 3000 K. Next, simulated annealing was performed from 3000 K to 25 K in 12.5 K increments, and at each temperature dynamics for the smaller of 1.6 ps or 400 steps was run. During simulated annealing the force constants of various energy terms were ramped to their final values as specified in Supplementary Table S3. For the realistic calculations, the initial structure calculation was followed by an additional calculation stage in which the 10 lowest energy structures from the initial calculation were refined using 10 different sets of randomized velocities (for a total of 100 structures) in a protocol which allowed all backbone dihedral degrees of freedom and omitted the initial minimization step. A regularized mean structure was calculated from the 20 lowest energy refined structures.</p>

PubMed Author Manuscript

On Catalytic Preorganization in Oxyanion Holes: Highlighting the Problems With the Gas Phase Modeling of Oxyanion Holes and Illustrating the Need for Complete Enzyme Models

Oxyanion holes play a major role in catalyzing enzymatic reactions, yet the corresponding energetics is frequently misunderstood. The main problem may be associated with the non-trivial nature of the electrostatic preorganization effect, without following the relevant formulation. That is, although the energetics of oxyanion holes have been fully quantified in early studies (which include both the enzymatic and reference solution reactions), the findings of these studies are sometimes overlooked, and, in some cases, it is assumed that gas-phase calculations with a fixed model of an oxyanion hole are sufficient for assessing the corresponding effect in the protein. Herein, we present a systematic analysis of this issue, clarifying the problems associated with modeling oxyanions by means of two fixed water molecules (or related constructs). We then re-emphasize the point that the effect of the oxyanion hole is mainly due to the fact that the relevant dipoles are already set in an orientation that stabilizes the TS charges, whereas the corresponding dipoles in solution are randomly oriented, resulting in the need to pay a very large reorganization energy. Simply calculating interaction energies with relatively fixed species cannot capture this crucial point, and considering it may help in advancing rational enzyme design.

on_catalytic_preorganization_in_oxyanion_holes:_highlighting_the_problems_with_the_gas_phase_modelin

I. INTRODUCTION<!>II. BACKGROUND<!>III. RESULTS AND DISCUSSION<!>III.1. ASSESSING THE FREE ENERGY OF THE FORMATION OF THE OXYANION: PROBLEMS AND PITFALLS<!>III.2 STUDIES USING A MINIMAL MODEL<!>III.3 THE FREE ENERGY IN AN ACTUAL ENZYME, RELATIVE TO THE CORRESPONDING REACTION IN SOLUTION<!>III.4 THE RISKS ASSOCIATED WITH A GAS PHASE ANALYSIS OF ENZYMES IN ENZYME DESIGN<!>III.5. REVISITING THE GENERAL GROUND STATE DESTABILIZATION IDEA<!>IV. CONCLUDING DISCUSSION

<p>Electrostatic catalysis due to polar preorganization is arguably the most important factor in enzymatic rate enhancment1,2. Despite this, however, this effect is still not widely appreciated, as a result of which it is not being used for enzyme design. A large part of this is associated with the non-trivial fact that the reorganization energy is being stored in the enzyme, and not in the enzyme-substrate interaction. Perhaps the best illustration of the problems with the greater acceptance of this concept are oxyanion holes, where overlooking preorganization led to a variety of different proposals, such as the low-barrier hydrogen bond (LBHB) proposal3–7 (for detailed consideration of the problem with this proposal see Refs. 8,9), as well as the problems in attempts to asses the preorganization contribution experimentally10–12 (see discussion in Ref. 13). Clearly, attempting to explain enzyme catalysis by overlooking the proper energetics needed to assess the LBHB catalytic proposal14 (which was discussed in e.g. Refs. 8,9), and instead relying on studies that use for instance minimal models or gas-phase calculations (e.g. Refs. 15,16) is a problem that should be taken note of. Additionally, perhaps the popularity of some of the dynamical ideas, which are presented as alternative explanations for enzyme catalysis (see the review in Ref. 17) is due to the difficulty of appreciating the preorganization effect.</p><p>The above difficulty becomes more relevant as a wider part of the chemical community is starting to appreciate the importance of oxyanion holes in possible design strategies18, but yet, still largely overlooks the polar preorganization idea. Based on the apparent complexity of this issue, it would seem to us that it is important to further clarify the preorganization idea, especially to those who are interested in calculating this effect. Thus, we will provide a detailed explanation of the preorganization proposal, and demonstrate why unfamiliarity with it can lead to incorrect conclusions. That is, we will expand on recent attempts to use a gas-phase study of oxyanion models as a way to gain an understanding of such systems. It will be illustrated that such studies cannot estimate the preorganization effect, as they are unable to estimate the protein constraints.</p><!><p>In light of the apparent difficulty of the wider acceptance of the polar preorganization concept (despite it's crucial role in enzyme catalysis), we will start our discussion by familiarizing the reader with this concept. That is, as has been discussed elsewhere2, there exist multiple cases where most of the catalytic effect of the enzyme is clearly due to electrostatic interactions. Specifically, quantifying the change in the polarization of the solvent (or protein) dipoles during the reaction in an enzyme and in solution led to an awareness of the importance of electrostatic reorganization1. Here, it was found that in the case of reactions occurring in solution, the solvent must pay a major free energy penalty when reorienting its dipoles towards the transition state (TS) charges as it moves from the reactant state (RS) to the TS, whereas, as the protein active site dipoles are already partially oriented toward the TS the protein has to pay much less reorganization energy. This point is illustrated schematically in Fig. 1, for the case of an ion pair type transition state, where the negative part is stabilized by an oxyanion hole.</p><p>Now one of the best ways to consider the preorganization concept is to focus on the specific case where oxyanion sites contribute to catalysis. The existence of special oxyanion sites in proteins was pointed out quite early in the literature19, although the energetic origin of such sites, as well as their relationship to the preorganization concept was only recognized in 19781 and deduced by mutational studies in 198620. The energetics of oxyanion holes have been quantified in many of our studies (e.g. Refs. 13,21–24), as well as those of others (e.g. Ref. 25), but the realization that proper quantum mechanical / molecular mechanical (QM/MM), and, particularly, empirical valence bond (EVB) calculations are probably the best way to quantify the effect of the oxyanions has not yet been widely accepted (e.g. Refs. 10–12). At any rate, here, we will provide a demonstration of the preorganization effect in the oxyanion hole, for a typical reaction of serine proteases of the type described in Fig. 2.</p><p>Now, when dealing with serine proteases, one should, in principle, model the complete reacting system (e.g. the catalytic triad) in the RS and TS, and then use the linear response approximation (LRA)27,28 (see also Section III) in order to extract the oxyanion contribution (as was done in e.g. Refs. 22,29,30). However, the use of the LRA with the complete catalytic triad has not been widely appreciated (or practiced). Thus, although we will also use the LRA here, we will mainly provide a simpler analysis with a subsystem of the reacting system, by taking a model that is sufficiently close to the model used by those who still work with gas-phase models, making it possible to clarify the problems with the oversimplified gas-phase model. In other words, we will focus on a simple nucleophilic attack at a C=O center (Fig. 3). However, we would also like to point out in passing that models containing the full catalytic triad (Asp, His and Ser) in the gas-phase without solvation are very problematic (as we have pointed out since 198623), as the Asp will not be negatively charged in the gas-phase. At any rate, here, we will focus on the nucleophilic attack step, where the change in interaction between the simulated system and the farther parts of the system are small relative to those in the full reaction of Fig. 2. We must also mention that the selection of such a model is essential if one wants to examine incomplete gas-phase models.</p><p>We start by schematically illustrating the energetic difference between the system in the enzyme and water in Fig. 4. With this scheme, it is possible to appreciate the relevant energetics. That is, computer simulation studies31 have indicated that the actual free energy electrostatic interaction between the enzyme dipoles and the charges of the substrate TS, ΔGQμp, is similar to that between water and the corresponding TS in solution (i.e. ΔGQμw). Now, clearly, if the interactions between the two were similar, one would expect the enzyme to have no catalytic effect. However, an early work1 provided the solution to this fundamental problem, by pointing out that, in fact, in water, roughly half the energy gained from the charge-dipole interactions is spent on changing the dipole-dipole interaction, ΔGμμw, such that the free energy of solvation of the transition state is given by32: (1)ΔGsolw≅ΔGQμw+ΔGμμw≅ΔGQμq−12ΔGQμw;12ΔGQμw where, for simplicity, we assume that the reactant state is nonpolar, and we refer here to the value obtained when the solvent dipoles are already oriented toward the solute charges as the "interaction free energy". Here, ΔGμμw is the reorganization energy mentioned above for the process of forming the transition state charges. In contrast, in a protein, where the active site dipoles associated with polar groups, internal water molecules and ionized residues are already partially oriented towards the charge center of the transition state, (2)ΔGsolp;ΔGμμp+ΔGQμp (where the superscripts "w" and "p" in Eqs. (1) and (2) denote water and protein respectively), and, of course, ΔGμμp<<ΔGμμw, with less free energy being spent on creating the oriented dipoles of the protein transition state. Now as (3)ΔG≠=ΔGgas≠+ΔGsol≠ therefore, following this discussion, (4)(ΔG≠)p−(ΔG≠)w;ΔGμμw which is the major source of the catalytic effect of the enzyme.</p><p>In order to quantify the above issue, we focus on a realistic estimate of the oxyanion contribution to the process of moving from C=O to C-O− (note that, as stated above, more complex studies which include the proton acceptor region have been performed by us, but this may be less clear to those who are not familiar with studies of a complete solvated system). As was already determined by us in 198933, the solvation free energies of C=O to C-O− are around −6 and −80 kcal/mol respectively (see Table 1). This means that the reorganization cost in water is ~3 and 40 kcal/mol for C=O and C-O− respectively. Now the maximum effect of the reorganization in Eq. 4 could be about 18 kcal/mol, but in reality there is also significant reorganization energy in the enzyme (the enzyme is not rigid) and also, it is important to consider the full system rather than just the C=O system (see Section IV). Nevertheless, the preorganization effect of the oxyanion is captured in the discussion above. Apparently, this effect has been overlooked in Ref. 34, where the reorganization has not been considered.</p><p>To summarize the above discussion, the solvent dipoles in the water reaction, will randomly orient themselves around the uncharged RS (or, if the RS is polar but uncharged, this will have a small non-random component), which means that the total activation free energy will also include a contribution from the free energy cost associated with the reorganization of these solvent dipoles towards the charged TS. On the other hand, in the protein, the active site dipoles (which can come from either polar groups, charged groups and/or internal water molecules) are already partially oriented towards the TS charge1. This means that the reaction costs far less reorganization energy compared to its counterpart in water. The analysis presented above allows us to obtain a quantitative estimate and to see that the main factor overlooked by analyses that only examine the interaction energy is overlooking the penalty of reorganizing the solvent dipoles upon going from the RS to the TS (see Section IV).</p><p>Now the analysis above may look formal to some readers, and thus, in order to provide a solid background for the reader, we will provide a quantitative estimate of the change in solvation energy during the process described in Fig. 4. However, we will also clarify below that the only way to quantify the result is by examining the complete solvation free energy, and not some specific structural parameters. Therefore, we will explain why the analysis of Ref. 34 and its conclusions are unjustified.</p><!><p>In this section, we will consider several approximated treatments of the energetics of oxyanion holes, starting from a very approximated treatment and moving to more complete models. We believe that the comparison of the different treatments will clarify what is needed for a proper description and will provide farther clarification of the nature of enzyme catalysis.</p><!><p>As a first step, we will attempt to move from a general analysis of the solvation energies in water and proteins to a specific analysis of different levels of approximation. We will start by considering treatments of the type used by e.g. Simón and Goodman34. That is, Ref. 34 has performed an extensive examination of the crystal structures of both enzymatic oxyanion holes, gathered from the Protein Data Bank35, as well as small molecule interactions obtained from the Cambridge Crystallographic Database36, and used these structures (in combination with gas phase ab initio calculations with an oversimplified model, which will be discussed in Section III.2) to explore the energetics of the oxyanion hole. Significantly, it was found that hydrogen bonding to carbonyls in molecular crystals does not seem to have the same directionality as those in enzyme active sites. Thus, it was suggested that oxyanion holes in enzymes do not work by transition state stabilization2 (TSS), but rather, that enzymes arrange hydrogen bonds such that the oxyanions are sub-optimally stabilized, in order to "avoid overstabilization of the "ground state"34, or, alternately, that "unlike water, enzymes can choose to orient their hydrogen bonds to stabilize the transition state slightly less well than is optimal, in order to stabilize the substrate much less well than is possible with the same number of hydrogen bonds"34 (with the authors then proceeding to argue that "this suboptimal arrangement of stabilizing groups to provide a greater reduction of the energy barrier (instead of higher transition state stabilization) can be readily introduced as a factor in the design of artificial catalysts"). It seems to us that such arguments can be understood as variations on the reactant state destabilization (RSD) idea, which was clearly and uniquely defined in e.g. Refs. 2,37. It must be pointed out here that while it is very unlikely that enzymes work by RSD, one could wonder whether the requirement of preorganization towards the TS would result in some RS destabilization. Thus, it is important to explore the validity of the proposal above.</p><p>Obviously, the hypothesis of Ref. 34 is interesting (see Ref. 38) since it is based on the observation that the RS in the enzymes might be less stable than in molecular crystals (but not necessarily in solution). However, the authors' actual conclusions involve some problems: Firstly, the main focus was placed on studying unrelaxed protein structures, which amounts to ignoring the reorganization in the protein. Secondly (and more importantly), the large reorganization in the reference reaction in water was not considered by the authors. Finally, the stabilizing effect of the oxyanion hole has not been analyzed properly. For instance, it was argued that "catalysis with two water molecules, even if they are optimally placed to stabilize the transition state, is less effective than catalysis by the oxyanion hole, because water will not be able to stabilize the ground state less well than the transition state". This suggestion that the enzyme destabilizes the RS more than in water overlooks the fact that two optimally placed water molecules in the gas-phase cannot provide a realistic model for examining catalysis. That is, in solution, the water molecules can even be pointing in the opposite direction in the ground state, but then rotate towards the charged TS (hence the reorganization penalty), or, as another example, it is possible that the ground state destabilization effect that the authors considered is an artifact which would have disappeared had the authors used a smaller, physically reasonable force constant (see the discussion in Section III.2). Since we are dealing with a general problem and conclusions that may be easily obtained by workers who do not use full enzyme models, we therefore find it useful to provide below a step-by-step "tutorial" on how to obtain the effect of an oxyanion in an oxyanion hole.</p><!><p>As our starting point, we will consider a minimal model (see Fig. 5) that involves a gas-phase "oxyanion hole" comprised of two water molecules, which may be assumed to be a reasonable model of the active site. We start by examining the relevant energetics obtained when the water molecules are allowed to relax to a geometry that most stabilizes the C=O state, and then to a geometry that most stabilizes the C-O− state. In doing so, we demonstrate that the oxyanion hole provides much more stabilization for the TS or intermediate (IS) states than the RS (see the cycle presented in Fig. 6). In other words, it is quite obvious that the oxyanion dipoles are designed in such a way as to stabilize the developing negative charge. This point is not apparent from the study of Ref. 34, since the reported study has not evaluated the energy of bringing the water molecules from infinity towards the oxyanion, and only examined the comparatively irrelevant effect of deforming the oxyanion hole. At any rate, clearly, the model oxyanion stabilizes the TS far more than the RS.</p><p>Next, we address the abovementioned issue of deforming the oxyanion hole, by fixing the water molecules in an orientation that provides less stabilization than the relaxed structure. The assumed deformation is similar but not identical to that assumed in Ref. 34, since the results obtained (see below) do not depend on the exact deformation. The results for the "good" configuration, as well as six different "deformed" configurations are complied in Table 2, and, as seen from the table, we of course reduce the stability of the system once we move the water molecules away from the optimal configuration (here, in most cases we lose more stability from the the C-O− system, with the two systems occasionally being affected similarly by the deformation, but the overall trend is clear). However, this point has little to do with RSD by the enzyme. That is, first of all, this issue has to be assessed relative to the reference state in solution (see below), but, more importantly, Ref. 34 puts too much weight into the importance of unrelaxed protein structures. This is demonstrated by taking any two of the deformed structures considered in Table 2 (in this example Structures 1 and 6), and placing strong constraints on the water molecules in order to try to "fix" them in the assumed deformed conformation in the protein. We therefore performed 1000 steps of QM/MM minimization (for details see SI, note that we chose 1000 steps simply because our aim is not to obtain perfect results, but rather to see how much effect the constraint has over a reasonable simulation length) using different constraints to examine the effect of constraints of different magnitudes. Now, as is shown in Table 3 and in Fig. 7, only an unrealistically large constraint of >>50 kcal mol−1 Å−2 can maintain the deformation effect, and with smaller constraints, the system very quickly minimizes back to an "optimal structure". For comparison, typical protein force constants are, at most, approximately 5 kcal mol−1 Å−2 for small deformations, and, the fact that trend this holds for two very different water conformations suggests that our finding is independent of the precise starting conformation of the water molecules, as long as they are not in the optimal position to begin with. Thus, as touched on in the previous section, one can obtain very problematic conclusions as a consequence of having used an infinite constraint in the calculation. In fact, the simulations of Ref. 34 do not examine the actual oxyanion hole in an enzyme, but rather consider unrelaxed crystal structures, and small models (i.e. either in the gas-phase with two water molecules, or theozymes). However, in cases with negatively charged oxyanions, such as in the case of TS analogues in the oxyanion hole of ketosteroid isomerase, one does find an oxyanion hole which is almost perfectly oriented to stabilize the oxyanion (see Ref. 12).</p><p>Most importantly, what is established here is that one cannot use fixed water molecules to examine or establish the energy of oxyanion holes in proteins, and that, as will be shown below, doing this requires performing full QM/MM calculations in which the energy in the protein and in water (not in a molecular crystal) is explored with full relaxation and proper free energy calculations (see the next section). Further dwelling on what the exact constraint in the water model should be is not useful, as this is not a proper model.</p><p>After considering the problems with a minimal model of the protein, we should consider the risks associated with using a model comprised of two fixed water molecules to examine the energetics of the reference reaction in water. Here, one may overlook the point that any study in bulk water must reproduce the fact that the solvent molecules are more or less randomly oriented around the C=O system (with some limited orientation in the first solvation shell), and then have to reorient upon the formation of the oxyanion. This reorganization penalty cannot be captured by a solvation model involving two fixed water molecules which are placed in the best orientation, a point which can be realized by anybody who searches for the average solvent orientation in actual simulations, but it can be much better quantified either by the estimate of Table 1, or by direct calculations of the reorganization energy which are summarized in Table 4 (which will be considered below). Thus, clearly, modeling an oxyanion hole using two gas-phase water molecules is not a fully realistic approach, and such an approach leads to, for instance, the incorrect conclusions reached with regard to LBHB (for discussion of this issue, see the problems with Ref. 42, which are discussed in Ref. 8). In conclusion, using two relatively fixed water molecules as a model for an oxyanion hole may lead one to conclude that such a system does not to stabilize the TS, but rather destabilizes the RS. However, in the case considered above, this conclusion is due to the effect of simply changing a dihedral angle, whereas, with a realistic constraint, the effect of these water molecules is primarily to interact with and stabilize the developing negative charge on the oxyanion.</p><!><p>At this stage, we will change our strategy and consider the entire protein/substrate system, and also consider the contributions from the oxyanion hole to the activation energy. We will then compare the results to the corresponding results in solution. There are many examples of such studies of oxyanion holes from our group, which include the reproduction of absolute catalysis and the effect of mutations, starting in 198623, quantifying this in 198831, later even doing this by ab initio QM/MM in 199843, and then also quantifying the problem in ketosteroid isomerase (KSI)13,22. This was also confirmed by others (e.g. Ref. 25).</p><p>Here, we repeat a part of our analysis of subtilisin, going beyond the well-defined and complete free energy analysis of Ref. 31, and evaluating the separate free energy contributions of the oxyanion in the presence of the rest of the system. This type of study (Table 4) involves LRA calculations27,44, using the same procedure that we applied in the study of the separate contributions in the reaction of vitamin B12 enzymes29 and the ribosome22. More specifically, the LRA calculations determine the free energy of moving from the C=O to the C-O− state by: (5)ΔG(C=O→C−O−)=0.5(<U(C−O−)−U(C=O)>C−O−+<U(C−O−)−U(C=O)>C=O) where U is the potential energy of the given state, obtained by the corresponding EVB energy, and <>x designates an average over the indicated state (where X is either C=O or C-O−). The reorganization energy is given by: (6)λ=0.5(<U(C−O−)−U(C=O)>C=O−<U(C−O−)−U(C=O)>C−O−) (note that the precise details of the EVB energy terms are given elsewhere, see e.g. citations in Ref. 45).</p><p>Since the LRA provides additive contributions, it is possible to extract the individual contributions from the oxyanion, and the calculated results are given in Table 4. We also explore the contribution of two water molecules in bulk water (this requires exploring the effect of leaving the specific molecules near the oxyanion with an artificial constraint). As seen from the table, a consistently modeled oxyanion stabilizes the C-O− more than the C=O, and also stabilizes the C=O more than the bulk water does. This finding is in apparent contrast to the conclusion that can be drawn from the work of Ref. 34, though this work (i.e. Ref. 34) attempted to consider the C=O energy in the model active site relative to that in molecular crystals, rather than the relevant reference state, which is the C=O species in water.</p><p>Now since examining the contribution from two water molecules in water is problematic (since we actually have to consider the entire solvent system) we have also performed LRA calculations which consider the complete environment (both in protein and in water), while evaluating only the free energy associated with the formation of the C=O polar group and the C-O− ionized group, and while keeping the rest of the substrate without its residual charges. This type of study (Table 5) can tell us whether or not we have an RSD mechanism. The evaluation of the energy of the C-O− is more complex, since in serine proteases the overall TS involves the [His+ C-O−] system, which is stabilized by both the oxyanion hole and the Asp 102 (which is, in turn, stabilized by the Asp hole31), so we cannot consider just the C-O− system when evaluating the observed catalytic effect. It should be noted that we have evaluated the overall correct contribution of forming the TS many times before (as was shown in the abovementioned references). Thus, we only focus here on the energy of the C=O system, where the effect of having or not having an ionized His does not change the solvation of the C=O system. Furthermore, the true ground state has an unprotonated His, which makes the analysis of the C=O species relevant to the actual reaction. At any rate, the calculations summarized in Table 5 establish that the enzyme stabilizes the C=O state more than water does.</p><!><p>After having established the energetics in actual oxyanion holes, and in oversimplified gas-phase models, we are ready to move on to a comprehensive discussion of the problems with oversimplified treatments of oxyanion holes. That is, basically, the analysis above has established the problems with performing oversimplified studies of artificial oxyanion holes, but, in order to prevent further errors, it is useful to be specific here.</p><p>The first issue is the risk of overinterpreting the hydrogen bonding orientations found in fixed X-ray structures. Here, one has to not only consider the poor resolution of the hydrogen positions, but, also (and far more importantly), the consequences of not allowing the oxyanion to relax, and thus not realizing that an infinitely rigid protein leads to completely unrealistic findings that do not reflect the preorganization correctly.</p><p>Additionally, choosing an incorrect reference state for evaluating the relative stability is a crucial problem. That is, not considering the energy of moving the "oxyanion hole" (i.e. the two water molecules) from infinity towards the oxyanion makes it difficult to determine the actual stabilization of the negative charge, which is perhaps reflected in the results shown in Fig. 8 of Ref. 34 (see also the SI of this paper). Following from this, comparison to the stability of small molecules in crystals (which was not done by any worker with actual calculations of the crystal) instead of the stability in water may lead to incorrect conclusions, as the crystal is already relaxed in such a way as to try to provide optimal stabilization to the C=O system. This fact has been demonstrated above, in addition to showing that the protein has not been designed to destabilize the C=O system. In any case, attempts to introduce structural constraints on an unrelaxed X-ray structure are problematic, and unlikely to reproduce any observed energy. Similarly, it should be noted that the active site arrangement can be quite different in the case of an inhibitor than in the case of the actual TS. Finally, with the focus given to enzyme design (e.g. Ref. 38), and the importance of small molecule catalysts for progress in this area, it would be remarkable to see gas phase models that can reproduce the known catalytic effect of an enzyme (a task which has been achieved very effectively by our approaches46).</p><p>An additional problematic issue we would like to touch on here is the idea presented in e.g. Ref. 34 that the preorganization effect cannot be important because even "optimally placed" water molecules will not be able to stabilize the reactant state less well than the ground state. That is, disproving the conclusions of quantitative earlier work (e.g. Ref. 22) about preorganization in oxyanion holes cannot be accomplished by modeling some gas-phase system without the enzyme (especially when doing this using the problematic assumption that the enzyme should destabilize the ground state in order to achieve catalysis). It is basically hard to accept the idea that gas-phase calculations are somehow superior to QM/MM calculations that have consistently reproduced the effect of mutations.</p><p>After discussing the general problems with gas-phase studies of enzymes, we would like to point out that this issue becomes more serious when being used to draw general conclusions about enzyme design. Here, the problem is two-fold. First and foremost, even though the use of models where the TS features are determined in the gas-phase are quite popular (e.g. Refs. 47,48), such an approach is not so effective for enzyme design. That is, since it is not capable of reproducing the TS binding free energy or the catalytic effect, it cannot be used to rank the different design constructs in a quantitative manner (even if an approach like that used in Refs. 47,48, where the emphasis was on the generation of protein structures with reasonable interactions with the TS model, can be an effective way to generate structural candidates). In other words, an approach that does not treat the protein-TS electrostatic free energy in a consistent way, will, as discussed in Section III.1, miss the preorganization effect and the reorganization penalty, and thus the corresponding catalytic effect. Another problem is the fact that such gas-phase models have no dielectric screening, and the absence of this can lead to irrelevant results such as the LBHB proposal, as was for example the case in Refs. 49,50. Thus, in short, a consistent treatment of electrostatics which captures the protein preorganization effect (by including the protein reorganization during the simulations) is crucial for successful enzyme design, and recently, we made great progress on this front46,51. Without this, the only way to reach a highly catalytic enzyme would be through extensive and costly experimental trial and error, and using gas-phase models with two water molecules is clearly not the way to go about this addressing this issue.</p><!><p>In light of the fact that the RSD idea still reappears in different incarnations in the literature (including the works that have been discussed above), it is quite important to reclarify (in greater detail) some of the problems with this proposal. Now while we could of course refer the reader to several careful studies of this issue (see Ref. 2, and references cited therein), we nevertheless find it useful to address the recent appearances of this proposal, in the particular case of orotidine 5'-monophosphate decarboxylase (ODCase)52. That is, the catalytic effect of ODCase was first proposed to involve the desolvation effect53. However, this was shown to involve an incomplete thermodynamic cycle (see e.g. Ref. 52). The elucidation of the structure of this enzyme showed that its active site is extremely polar (highly charged), however, this led to a new RSD proposal, where it was argued that the negatively charged protein group destabilizes the carboxylate of the orotate group of the substrate54. This proposal was shown to be inconsistent with the nature of the system, since a destabilized orotate will accept a proton and become stable52. Furthermore, a careful computational study has illustrated that the protein works by TSS and not by RSD (see discussion in Ref. 52 and below). Moreover, studies by Wolfenden and co-workers53,55 have provided clear evidence against the RSD proposal (see Ref. 2, as well as the discussion below of the experimental work of Ref. 56).</p><p>The problems with the RSD proposal were dramatically emphasized by the experimental work of Amyes and coworkers57, who explored the origin of the catalytic power of ODCase (trying to support the RSD proposal) by studying the decarboxylation of a truncated substrate (called EO), which lacks the 5'-phosphodianion part. These workers found that while the reaction of this substrate is quite slow, binding an exogenous phosphate dianion to ODCase results in an 80000-fold increase in kcat/Km. This appeared to be in a clear conflict with the proposal that the presumed RSD is due to the binding free energy of the 5'-phosphodianion part of the substrate, which is supposed to induce extremely large reactant RSD, and thus to catalyze the reaction (see e.g. Ref. 54). In this proposal, the negatively charged groups of the protein (i.e. Asp70 and Asp75B in Fig. 8) are used to destabilize the carboxylate of the orotate. In fact, this view has been presented as confirmation of Jencks' proposal that enzymes work by using binding energies to destabilize the ground state of the reactive part of the substrate. However, a logical analysis of the work of Ref. 57 indicated that the RSD idea is incorrect, and thus confirmed our careful analysis52. That is, the RSD proposal of Ref. 54 is based on the idea that the phosphate part is bound so strongly that it pulls the chemical part (thorough the R-C bond in Fig. 8) towards its destabilizing environment, and leads to about 20 kcal/mol RSD. Unfortunately, not only was this shown to be problematic in Ref. 52, but also, the experiments of Ref. 57 showed that catalysis occurs in the absence of a bond between the phosphate and EO parts, such that the presumed strain cannot be transferred between these parts. Of course, the binding of the negative part of the substrate does help the active site reach its proper preorganization, and thus, to use this for the electrostatic stabilization of the TS. This, however, has little to do with the classical Jencks idea of using the binding energy for RSD.</p><p>Now a more recent work58 focused on a part of the reverse reaction, starting from the product uridine 5'-monophosphase (UMP) without the CO2 part, and asked how much it "costs" to remove the proton from the C6 carbon in the presence of the protonated Lys (or what the energy of the transition state with a protonated lysine and negatively charged carbanion actually is). This energy, which is directly related to the pKa for the deprotonation of the C6 carbon, was determined by use of deuterium exchange. It was found that the pKa is reduced by at least 10 units, or more than 14 kcal/mol, which means that the corresponding system, that strongly resembles our TS, is stabilized by at least 14 kcal/mol. This provides another overwhelming specific example that counters the RSD idea and provides direct proof of the TSS idea.</p><p>Despite these convincing points, and the illustrations that theoretical support for the RSD idea in the case of ODCase is fundamentally flawed (including contradictions of the first law of thermodynamics by Ref. 59, see the discussion in Ref. 2), we nevertheless still see conflicting analyses and significant confusion about the meaning of the very clear experimental findings. An illustrative example is given in a very recent work60, where a D70N mutation reduced kcat by about 200, while leaving KM almost unchanged. This finding essentially proved that the ionized acid is not involved in any ground state destabilization (otherwise its mutation would have led to a larger binding energy as well as a reduction in KM), and yet the authors ventured to say that their finding supports the RSD idea of Ref. 54. In an effort to find a way around this convoluted argument, and again, perhaps in light of the difficulties with understanding the preorganization idea, the authors suggested a new definition of RSD (see footnote 4 in Ref. 60). They also suggested that somehow, RSD is related to an undefined destabilization of the path to the TS. Of course one has to define such a proposal, since the issue is the free energies of the RS and TS, and destabilization at any point on the path between them that cannot change the activation energy and the rate constant (unless such a point is a transition state, and we have TS destabilization).</p><p>The misunderstandings of the authors of Ref. 60 may be due to the difficulty of conceptualizing the preorganization effect, which thus leads them to misinterpret the observation60 that the mutation does not change the effect of the protein on the above proton transfer to the C6 of UMP. This finding seems to increase the difficulty of analyzing unique information about RSD. That is, without a clear energy based concept, it is basically impossible to define, analyze and understand the key experiments (see a related discussion in Ref. 17). The problem starts with the assumption that the role of D70 is to destabilize the RS, a point which was already shown to be incorrect by any of the points above (including the fact that such an effect would lead to protonation of the orotate), and, of course, if this point were correct, we would have observed an increase rather than a decrease in the pKa of C6. Now the only model that has up to now succeeded to rationalize and quantify the catalysis in ODCase is the model presented in Fig. 8. In this model, the role of D70 is to stabilize the TS formed by Lys72 and the negatively charged C6. Of course, this stabilization depends on the exact preorganization, and is not expected to be identical in UMP and in the TS of ODCase, but rather, to exhibit a similar trend. Our point is that the exact trend in the mutation cannot be predicted (and even understood) without the calculation of clear energy based concepts, but, in any case, starting from an incorrect premise and reformulating the meaning of key concepts such as RSD is not so useful.</p><p>Finally, despite the attempts to use clear evidence against the RSD proposal as support for this idea, we see that in their most recent paper56, the authors now propose that TSS is entropically helped by putting together residues (in this case Arg235), which participate in electrostatic stabilization (relating this to Jencks' entropic idea61). Of course, this is another example of the difficulty in understanding the preorganization idea. That is, having electrostatic stabilization by an enzyme has nothing to do with Jencks' entropic idea (which is about the entropy of placing the substrate fragments together). Folding the enzyme (Step 1 in Fig. 1b) is part of the formation of the catalyst, and not the explanation for its catalytic effect.</p><!><p>This work has re-analyzed the contribution of preorganized hydrogen bonds to the catalysis of oxyanion formation in enzymes, and further established their role in TS stabilization. It has also clarified the problems associated with an oversimplified "theozyme" analysis for preorganized systems. The first important issue we have addressed is the idea that in the oxyanion hole, the hydrogen bonds are arranged in such a way that the oxyanions are reasonably but sub-optimally stabilized, in order to avoid overstabilizing the ground state34. This idea seems to result in an apparently new version of the popular ground state destabilization (GSD) (i.e. RSD) proposal, which suggests that enzymes reduce the activation barrier of the reaction by destabilizing the ground states of their reactant fragments (as was put forward by many workers, such as, for instance, the works of Ref. 62–64 amongst others). However, this proposal has not been supported by systematic computational studies, which have demonstrated that the TS is in fact "solvated" much more strongly in the enzyme than in the reference solution reaction (see discussion in e.g. Refs. 2,32,65, and references cited therein, as well as in Section V). In Section III, we demonstrate once again that in this case, there is also much larger TS stabilization than an RS destabilization effect, and that the water molecules provide a comparatively much smaller interation with the RS. This issue ties in directly with the most risky problem with treatments that ignore the reorganization energy, which (in an ideally preorganized enzyme) can reach half of the final stabilization. That is, the misunderstanding of the nature of solvation effects, which are due to the total environment rather than just one or two water molecules, prevents one from comprehending the reorganization effect, which, in water, can be estimated by taking half of the solvation free energy as the upper possible limit.</p><p>Now, of course, there are other major problems that are associated with using high-level calculations and then forgetting about the entropy, which is a crucial part of the free energy, and automatically included in proper EVB calculations. Clearly, calculating the entropy correctly is a non-trivial issue, as we discussed at length in e.g. Ref. 66. However, ignoring this issue compromises the obtained results (for a related example of this problem, see Ref. 67). Additionally, it is not useful to focus on the secondary issue, namely, the effect of changing the angle of the hydrogen bond, rather than the change in the free energy interaction between the hydrogen bonds and the oxygen upon the formation of the oxyanion charge (see the demonstration of this in Section III.2). Yet another fundamental problem can be found in the idea that the enzyme is not designed to provide optimal stabilization to the TS. This conclusion may come from not trying to optimize the hydrogen bonding interactions at the TS in the protein. Of course, the initial structure is not perfect, and there is some reorganization. However, this reorganization is much smaller than that for the same reaction in water, which is the origin of enzyme catalysis.</p><p>These oversimplified views are problematic, not only because they reflect a misunderstanding of the reorganization concept, but also because of the problems it leads to in attempts to understand deeper issues, such as was the case in the argument about KSI. This issue clearly cannot be examined by those who trivialize the detailed analysis of Ref. 22 and assume that this work is about charge delocalization in water and in the protein, rather than a careful quantitative analysis of the preorganization effect. That is, as discussed in detail above, it is not possible to provide a meaningful judgment about problems that are on the level of those we addressed in our most recent work on the topic13 by using a model involving two water molecules in the gas-phase as a presumably physically meaningful representation of the oxyanion hole, and assuming that this can conclusively resolve the argument about KSI.</p><p>What we now see in the literature is a progression from workers who proposed the study of gas-phase reacting systems (e.g. Refs. 47,48,68,69) to now including also a model oxyanion hole34,70,71. However, perhaps it would be better to give more credit to studies that include all the protein from the beginning. In other words, this is similar to progress in studies of proton transfer in bacteriorhodopsin, where irrelevant studies with a single water molecule and a gas-phase proton such as that of Ref. 72 were viewed by the community as being much more reasonable than calculations of the whole solvation environment73–75, and leads to problematic concept such as proton wires. In any case, we have already discussed the enormous problems associated with the belief that gas-phase calculations, which completely miss the preorganization effect, are useful in enzyme design. Here, we would like to conclude by pointing out that even larger problem will arise if one were to follow studies that reach conclusions about catalysis by calculations that do not involve the enzyme, or are "solution" calculations without the solvent</p>

PubMed Author Manuscript

DNA Micelle Flares for Intracellular mRNA Imaging and Gene Therapy

Multifunctional DNA micelles: Molecular beacon micelle flares (MBMFs), based on diacyllipid-molecular beacon conjugate (L-MB) self-assembly, have been developed for combined mRNA detection and gene therapy. The advantages of these micelle flares include easy probe synthesis, efficient cellular uptake, enhanced enzymatic stability, high signal-to-background ratio, excellent target selectivity, and superior biocompatibility. In addition, these probes possess a hydrophobic cavity that can be used for additional hydrophobic agents, holding great promise for constructing an all-in-one nucleic acid probe.

dna_micelle_flares_for_intracellular_mrna_imaging_and_gene_therapy

<p>The hybridization between a nucleic acid strand and its complementary sequence, one of the strongest and most specific molecular recognition events,[1] has greatly facilitated the development of disease diagnosis and gene therapy. For example, both linear[2] and hairpin nucleic acid probes[3] have been used to visualize and detect specific messenger RNAs (mRNAs) in living cells. Many mRNAs are disease-related and can be used as specific biomarkers to assess the stage of diseases, including cancer. Through molecular engineering, these probes can effectively translate an mRNA binding event into a fluorescence signal change without the need to remove unbound free probes. In addition, most human diseases, even cancer, can be treated with the introduction of genetic materials – plasmid DNAs[4], antisense oligonucleotides[5], small interference RNAs[6], small hairpin RNAs[7], and microRNAs[8] – into somatic tissues. These genetic materials can either enhance gene expression[9] or inhibit the production of deleterious protein, thus making nucleic acid probes excellent candidates for gene therapy[5a]. The advantages of nucleic acid probes lie in the simplicity of their synthesis, the suitability of their modification, and the selectivity of their binding.[1b]</p><p>However, their potential has not been fully realized due to the following reasons. First, as negatively charged hydrophilic biomacromolecules, nucleic acid probes cannot freely traverse cell membrane[10], thus requiring additional instruments (such as microinjection and electroporation) or materials (such as transfection reagents including cationic lipids/polymers and nanomaterials) for efficient cellular internalization.[5b] Second, nucleic acid probes can be unstable even after successful cellular delivery because of endogenous nuclease digestion,[11] leading to high false positive signals or decreased therapeutic efficiency. Third, most applications for nucleic acid probes focus on either mRNA detection or gene therapy, while a better strategy that can improve patient outcome is the combining of mRNA imaging[3c, 12] and gene therapy[13] into one biomolecular tool. Through mRNA imaging, real-time spatiotemporal evaluation of nucleic acid probe delivery and target gene expression can be realized non-invasively, providing useful information for assessing therapeutic efficiency, adjusting treatment protocols, and refining probe design.[14] Even though nucleic acid functionalized gold nanoparticles (AuNPs) with efficient cellular uptake[15] and enhanced enzymatic stability[16] have been developed to solve the first two challenges, they suffer non-negligible cytotoxicity at relatively high concentrations as a result of AuNP incorporation.[17] In addition, the preparation of these probes is very time-consuming, requiring more than 24 hours even after obtaining AuNPs and nucleic acids.[5b] Therefore, an ideal nucleic acid probe should be easy to synthesize, possess self-delivery capability, be highly biocompatible, and be sufficiently stable in cellular environment, while at the same time, performing multiple functions in living cells.</p><p>Here we present a sensitive and selective approach for combined mRNA detection and gene therapy using molecular beacon micelle flares (MBMFs). MBMFs are easily prepared by diacyllipid-molecular beacon conjugate (L-MB) self-assembly, not requiring any materials of potential biohazard. Just like pyrotechnic flares that produce brilliant light when activated, MBMFs undergo a significant burst of fluorescence enhancement upon target binding. This hybridization event subsequently induces gene silencing, leading to cancer cell apoptosis. The advantages of MBMFs include easy probe synthesis, efficient cellular uptake, enhanced enzymatic stability, high signal-to-background (S/B) ratio, excellent target selectivity, and superior biocompatibility. In this approach (Scheme 1), L-MBs spontaneously self-assemble into MBMFs with diacyllipid core and MB corona in aqueous solutions due to hydrophobic interactions. The MB part is a DNA sequence composed of a target-recognition loop flanked by two short complementary stem sequences. The formation of the stem-loop (hairpin) structure brings the quencher and fluorophore, which are located at the opposite ends of MB, into close proximity, thus effectively quenching fluorescence ("OFF state"). When hybridized to target mRNA, the MB in MBMFs experience a conformation change that opens the hairpin structure, physically separating the fluorophore from the quencher to allow fluorescence to be emitted upon excitation ("ON" state). In addition, the hybridization of MBMFs to the target mRNA can specifically inhibit gene expression through different mechanisms, including translational arrest by steric hindrance of ribosomal activity and the induction of RNase H endonuclease activity,[18] leading to the suppression of cancer cell growth.</p><p>L-MBs with illustrated structure (Figure 1a) were prepared by directly coupling diacyllipid phosphoramidite onto the 5'-end of MBs on a fully automated DNA/RNA synthesizer, purified by reverse-phase high-pressure liquid chromatography (HPLC) (Figure S1a and S1b), and characterized by ESI-MSn (Figure S1c). Diacyllipid phosphoramidite was synthesized through a three-step reaction according to our previously reported procedure.[19] After purification, L-MBs spontaneously form MBMFs in aqueous solution with very low critical micelle concentration (CMC) (below 10 nM, Figure S2), indicating their excellent stability compared to polymer-micelle systems.[20] The formation of MBMFs was further confirmed by both agarose gel electrophoresis (Figure 1b) and dynamic light scattering (DLS) (Figure 1c). MBMFs migrated much slower than MBs without diacyllipid, suggesting the successful formation of larger micelle nanostructures. DLS measurements showed that MBMFs had a diameter of 17.1 nm. After adding synthetic complementary target (cDNA), the diameter increased to 29.4 nm, while incubating MBMFs with synthetic random control (rDNA) resulted in negligible size increase. Detailed size distribution information can be obtained from the Supporting Information (Figure S3). These results indicated that MBMFs maintained target recognition capability after the formation of a micellar structure and that the binding event did not disrupt the structural integrity of the micelle. Consistent results were also obtained from zeta-potential measurements: values of −6.2, −10.8, and −7.1 mV were obtained for MBMFs only, MBMFs treated with cDNA, and MBMFs treated with rDNA, respectively (Figure 1d). Detailed sequence information for all the probes used can be found in Supporting Information (Table S1)</p><p>The performance of MBMFs was first evaluated in buffer system. According to fluorescence spectroscopy results, MBMFs were specific to target sequences, with approximately 10-fold signal increase for cDNA which is much higher than previously mentioned AuNP-nucleic acid conjugates[5b, 17, 21], but only minimal enhancement for rDNA (Figure 2a). In addition, MBMFs were able to differentiate between perfectly complementary target and mismatched targets (Figure 2b). The fluorescence signal of MBMFs exhibited dose-dependent increases in response to cDNA concentrations from 0 to 1 µM (Figure 3c and Figure S4) with a wide dynamic range from 0 to 200 nM (Figure 3c, Inset). These results demonstrated that MBMFs could effectively signal the presence of target with excellent selectivity and sensitivity. To compare the stability of MBMFs and MBs towards enzymatic digestion, we incubated each with endonuclease DNase I (1U/mL, significantly greater than what would be found in the cellular environment) and monitored the fluorescence signal increase as a function of time. The results revealed a much slower increase in fluorescence signal for MBMFs compared to MBs, indicating their enhanced stability due to increased resistance to enzymatic digestion (Figure 2d). Similar phenomenon was also observed for MBMFs and MBs in cell lysate (Figure S5).</p><p>After testing the feasibility of the MBMF approach with a synthetic target, the ability of MBMFs to permeate cell membrane and detect target mRNA was further investigated. The loop region of L-MBs was designed to be perfectly complementary to c-raf-1 mRNA, a cancer biomarker and antisense therapeutic target of great significance in cancer diagnostics and theranostics. Noncomplementary MBMFs with similar background signal, but little response to target (Figure S6), were used as control. A549 cells were used to verify the ability of MBMFs to detect intracellular mRNA in the tumor microenvironment. These adenocarcinoma human alveolar basal epithelial cells come from cancerous lung tissue and have a high expression level of c-raf-1 mRNA.[5a]</p><p>In order to obtain optimum results for intracellular detection of c-raf-1 mRNA, we optimized both probe concentration and incubation time for all cell experiments. A549 cells cultured on coverglass-bottom confocal dishes were incubated with 150, 300, and 600 nM MBMFs, respectively, and then imaged under a confocal laser scanning microscope. Increasing fluorescence signal was observed for cells treated with increasing concentration of MBMFs (Figure S7). We noticed that cells treated with 150 nM MBMFs did not generate sufficient fluorescence signal to illuminate c-raf-1 mRNA, while cells treated with 600 nM MBMFs resulted in poor S/B ratio. Therefore, the optimal probe concentration was 300 nM, which had the best combined fluorescence signal and S/B ratio. We also studied the influence of incubation time on fluorescence signal by incubating A549 cells with 300 nM MBMFs for 2 and 4 h, respectively. Because similar fluorescence intensity was observed for both times (data not shown), 2 h was chosen as the assay time for the remaining cell experiments. In addition, the co-localization assay demonstrated that most fluorescence came from the cytoplasm, instead of endosomes or lysosomes (Figure S8), indicating that the signal was caused by the specific binding of MBMFs to c-raf-1 mRNA.</p><p>Under optimized conditions, confocal laser scanning microscopy results revealed that A549 cells treated with MBMFs (Figure 3a) displayed much more fluorescence than the population treated with noncomplementary control (Figure 3b) and MBs (Figure 3c), demonstrating the selectivity of the system and the need of diacyllipid for efficient self-delivery, respectively. In comparison, normal bronchial epithelial cell line HBE135 from healthy lung tissue, which expresses significantly less c-raf-1 mRNA,[22] displayed very low fluorescence (Figure S9). We also used flow cytometry to collect the fluorescence data for cells treated with MBMFs. Compared to confocal laser scanning microscopy, which allows imaging of only a small number of cells, flow cytometry can analyze thousands of cells per second, generating a quantifiable statistical average for a large population of cells, while eliminating cell-to-cell variation and experimental artifacts. The flow cytometry results were in excellent agreement with confocal imaging: 2.61 and 1.08 times signal enhancement were observed for A549 and HBE135 cells, respectively (Figure 3b). Thus, while MBMFs work for synthetic target detection in buffer system, these results also demonstrate that MBMFs are feasible for intracellular mRNA detection in living cells. In addition, the MBMF approach can also differentiate cell lines with distinct mRNA expression levels, such as the cancerous and normal cells used here.</p><p>Before they can be used for gene therapy, MBMFs must first hybridize with mRNA. The probe acts either by providing a translation block, preventing translation from the targeted mRNA, or by forming a DNA/RNA hybrid with the target mRNA, causing it to be degraded by enzyme RNase H.[18] Using these mechanisms, the MBMFs can be used for imaging guided gene therapy.[14a] In this effort, we used Raf genes which code for serine threonine-specific protein kinases that play pivotal regulatory roles in the development and maintenance of certain human malignancies. Substantial evidence supports the theory that antisense oligonucleotides targeted against c-raf-1 kinase can specifically inhibit c-raf-1 mRNA expression and tumor progression through aforementioned mechanisms when properly delivered. Therefore, we also tested the antiproliferative effect of MBMFs on cancer cells. Since gene therapy based on antisense oligonucleotides requires a long treatment period, phosphorothioate MBMFs (S-MBMFs) were used to avoid any potential nuclease digestion in living cells that would diminish therapeutic efficiency. Experiment results showed that this backbone modification didn't significantly affect the performance of S-MBMFs compared to MBMFs since they had similar background fluorescence without target and maximal fluorescence with excess target or non-target (Figure S10). In addition, S-MBMFs had comparable CMC (below 20 nM, Figure S11) as MBMFs (below 10 nM, Figure S2). According to the cytotoxicity assay (Figure 4), A549 cell growth was negligibly influenced by the treatment of cells with noncomplementary S-MBMFs, indicating the superior biocompatibility of our system compared to some metal nanoparticle systems.[17] However, their treatment with S-MBMFs resulted in marked inhibition of cell proliferation in a dose-dependent manner, suggesting that MBMFs can be applied as an antisense therapy for cancer cells with high expression level of c-raf-1 mRNA.</p><p>In summary, we have presented a novel nanoprobe based on molecular assembly that can be used for combined mRNA detection and gene therapy. Current advantages for this approach include easy probe synthesis, efficient cellular uptake, enhanced enzymatic stability, high S/B ratio, excellent target selectivity, and superior biocompatibility. Instead of incorporating potentially biohazardous materials for efficient nucleic acid probe delivery, easy molecular engineering of MBs with diacyllipid provided the resulting MBMFs with new properties that MBs do not have, such as self-delivery capability and enhanced intracellular stability. In addition to their use in the context of mRNA imaging and gene therapy, MBMFs possess a hydrophobic cavity for additional hydrophobic materials, such as magnetic contrast agents and anticancer drugs, showing great promise for constructing an all-in-one nucleic acid probe capable of imaging, diagnosis, and therapy at the same time.</p>

PubMed Author Manuscript

Second Contact Shell Mutation Diminishes Streptavidin-Biotin Binding Affinity Through Transmitted Effects on Equilibrium Dynamics

We report a point mutation in the second contact shell of the high-affinity streptavidin-biotin complex that appears to reduce binding affinity through transmitted effects on equilibrium dynamics. The Y54F streptavidin mutation causes a 75-fold loss of binding affinity with 73-fold faster dissociation, a large loss of binding enthalpy (\xce\x94\xce\x94H, 3.4 kcal/mol at 37 \xc2\xb0C) and a small gain in binding entropy (T\xce\x94\xce\x94S, 0.7 kcal/mol). The removed Y54 hydroxyl is replaced by a water molecule in the bound structure, but there are no observable changes in structure in the first contact shell and no additional changes surrounding the mutation. Molecular dynamics simulations reveal a large increase in atomic fluctuations for W79, a key biotin contact residue, compared to the wild type complex. The increased W79 fluctuations are caused by loss of water-mediated hydrogen bonds between the Y54 hydroxyl group and peptide backbone atoms in and near W79. We propose that the increased fluctuations diminish the integrity of the W79-biotin interaction and represent a loosening of the \xe2\x80\x9ctryptophan collar\xe2\x80\x9d which is critical to the slow dissociation and high affinity of streptavidin-biotin binding. These results illustrate how changes in protein dynamics distal to the ligand binding pocket can have a profound impact on ligand binding, even when equilibrium structure is unperturbed.

second_contact_shell_mutation_diminishes_streptavidin-biotin_binding_affinity_through_transmitted_ef

<!>Protein expression and purification<!>Kinetic measurements<!>Equilibrium measurements<!>Crystallization<!>Diffraction data collection<!>Structure solution and refinement<!>Molecular dynamics simulations<!>Screen of second contact shell mutations<!>Effects of Y54F on equilibrium complex structure<!>Effect of Y54F on binding energetics<!>Effects of Y54F on equilibrium dynamics<!>DISCUSSION

<p>Protein structural fluctuations (fluctuations about the mean conformation) are thought to play a major role in ligand binding and catalysis, but measuring and predicting the impact of these fluctuations on binding affinity and catalytic efficiency remains an extraordinary challenge. Point mutations and binding events that are distal from protein active sites can dramatically affect binding and catalysis through transmitted effects on protein dynamics rather than conformation, as seen in well-characterized enzymes such as dihydrofolate reductase,1-4 in thermophilic enzymes compared to their mesophilic homologs,5-7 and in "dynamically-driven allostery" in cAMP-CAP8 and ligand-PDZ domain binding.9 Loss of conformational entropy and side chain mobility during protein-ligand binding is well-documented,10,11 and in some cases compensated for by a gain in conformational fluctuations away from the binding site.12,13 However, there are few examples where the effects of a point mutation on protein-ligand binding energetics have been directly related to an observed change in fluctuations, with no equilibrium structural changes in the binding pocket.</p><p>In previous crystallographic and molecular dynamics simulation studies of wild type streptavidin, we observed that biotin binding caused notable decreases in protein dynamics – reduced temperature factors in refined crystal structures and diminished atomic fluctuations in MD simulations – at sites far removed from the biotin binding pocket. We hypothesized that it might be possible to impact biotin binding by introducing conservative point mutations at sites outside the first contact shell that would preserve equilibrium structure but alter local protein dynamics. To test this hypothesis, we designed a set of eight streptavidin mutants, each with a single point mutation of a second contact shell residue, in an attempt to increase local protein structural fluctuations by disrupting a hydrogen bonding network and/or displacing an adjacent bound water molecule. We screened the mutants based on biotin dissociation rate changes, and selected two mutants, Y54F and F130L, with koff(mutant) / koff(wild type) > 50 for further analysis. We refined x-ray structures for both mutants to confirm that the mutations had no significant effect on biotin-contacting side chain positions in the bound structure. We performed molecular dynamics simulations to first confirm the conservation of equilibrium structure in the simulations and then to compare the structural fluctuations of these mutants with wild type streptavidin. In parallel, we characterized the biotin binding thermodynamics.</p><p>We have reported preliminary results for the F130L mutation previously.14 Here we present a crystallographic, computational and biophysical study for Y54F, a streptavidin mutation that causes a 75-fold loss of binding affinity, with no observable changes in equilibrium binding pocket structure, but a large increase in fluctuations of a key biotin-binding residue, W79. The removal of the Y54 hydroxyl group disrupts a water-mediated hydrogen bond network involving protein backbone atoms of N49, W79, and R84, which leads to significantly increased motion for loop L5,6, and much larger side chain fluctuations for W79, a key biotin aromatic contact. The large loss of binding enthalpy and small gain of binding entropy relative to wild type are consistent with a mutation that loosens the "tryptophan collar" that traps biotin during binding.15-17 These observations directly relate changes in ligand binding thermodynamics with altered protein dynamics at a distal site, and demonstrate the importance of including dynamic effects originating outside the binding pocket in structure-based drug and enzyme design.</p><!><p>The Y54F mutation was created in the synthetic core streptavidin gene in pET21a (Novagen, San Diego, CA) using the QuikChange protocol (Stratagene, La Jolla, CA) as described previously,18 and confirmed by sequencing. Protein was expressed using the T7 expression system in BL21(DE3) E. coli and purified as described previously.17 The expected mass and purity of the mutant were confirmed using electrospray mass spectrometry.</p><!><p>The rates of biotin dissociation from streptavidin variants were measured using a cold-chase radiometric method described previously.19 Briefly, 10 nM 3H-biotin and 30 μM Y54F or WT streptavidin in 50 mM sodium phosphate buffer, pH 7.0, 100 mM NaCl, were equilibrated at the experimental temperature for 2 h, then a large excess of unlabeled biotin (50 μM final concentration) was added and mixed rapidly. Aliquots of 200 μL were removed periodically and immediately ultrafiltered using chilled, 30k Microcon filters (Millipore, Billerica, MA). The filtrate was counted to quantify the amount of 3H-biotin released as a function of time.</p><p>Dissociation rate constants for protein-ligand complexes at each temperature were determined by fitting each data set by a one-term exponential decay. These koff values were used to calculate initial estimates of ΔH± and ΔS± in a global fit of all data, with ΔH± and ΔS± as the only adjustable parameters, using the equation (1)It=I0{1−exp[−kBThexp(TΔS≠−ΔH≠RT)⋅t]} where It is the measured 3H count at time t, Io is the initial 3H count, kB is Boltzmann's constant, h is Planck's constant, and R is the gas constant. Dissociation experiments at each temperature were performed on two separate days.</p><!><p>Equilibrium binding enthalpies were measured at 12, 25 and 37 °C using a VP-ITC isothermal titration calorimeter (Microcal, Northampton, MA). Streptavidin at 30 μM in 50 mM sodium phosphate, pH 7.0, 100 mM NaCl, was titrated with 25 5-μL injections of 500 μM biotin. Heat flow was integrated and data were fit using Origin software. Two titrations were performed at each temperature.</p><p>Equilibrium binding affinity relative to WT streptavidin was determined using a radiometric competitive binding method described previously.19 Y54F and WT streptavidin at a range of concentrations competed with 50 nM WT streptavidin with a polyhistidine tag for 20 nM 3H-biotin, in 50 mM sodium phosphate, pH 7.0, 100 mM NaCl, for 24 h at 37 °C. The partitioning of 3H-biotin between proteins was measured by precipitating the his-tagged WT protein using nickel-nitrilotriacetic acid agarose resin (Qiagen, Valencia, CA) and counting 3H-biotin remaining in solution. These counts, corrected for unbound 3H-biotin, were fit as the root of the competitive binding equation (2)KCKP⋅[C⋅L]2+(LT+CT+KCKP(PT−LT))⋅[C⋅L]−LTCT=0 where KC and KP are the equilibrium dissociation constants for the competitor and WT, respectively, and [C·L], LT, CT and PT are the concentrations of competitor-bound ligand, total ligand, total competitor, and total WT protein, respectively. The ΔKd value in Table IV is the average value from two experiments on separate days.</p><!><p>The Y54F mutant of "core" streptavidin20 was co-crystallized with biotin using hanging drop vapor diffusion techniques. Crystals were obtained by mixing protein (12.5 mg/mL in water) with a two-fold molar excess of biotin. The reservoir solution for Y54F was 60% saturated ammonium sulfate, 5% isopropanol. Drops of protein and ligand solution were mixed with an equal volume of reservoir solution before equilibration. Crystals were transferred to a crystallization solution containing 30% glycerol as a cryoprotectant before freezing at 100 K in a nitrogen stream for diffraction data collection.</p><p>Crystals of uncomplexed Y54F were obtained using similar techniques, but with a reservoir solution containing 2.5 M sodium chloride, 0.1 M sodium / potassium phosphate, pH 6.2. The cryoprotectant for these crystals was 30% ethylene glycol.</p><!><p>Diffraction data for the Y54F-biotin complex were collected at Stanford Synchrotron Radiation Lightsource (SSRL) beamline 9-2 (λ = 0.97946 Å) at 100 K using a Mar 325 CCD detector and were processed using HKL2000.21 The space group for the Y54F crystals is I222 with two subunits in the asymmetric unit. Data for the uncomplexed protein were collected at SSRL beamline 12-2 (λ = 1.0 Å) at 100 K using a Pilatus detector and were processed using XDS.22 The space group is I4122 with one subunit in the asymmetric unit. Data set statistics are shown in Table 1a.</p><!><p>The initial structural model for Y54F was obtained from an isomorphous structure – the biotin complex of wild type streptavidin (PDB ident 1MK5). The structural model was refined using REFMAC-523 in the CCP4 suite.24 Rfree25 was calculated using 5% of the data in the test sets. All atoms were refined with anisotropic temperature factors. Riding hydrogen atoms were added to the models, and Babinet scaling was used to account for bulk solvent effects.</p><p>Sigma A weighted |Fo|-|Fc| and 2|Fo|-|Fc| electron density maps26 were viewed with XtalView27 and COOT28 for graphical evaluation of the model and electron density maps. XtalView, MOLSCRIPT29, and Raster3d30 were used to produce the structural figures for this paper.</p><p>In the biotin complex, a small peak about 1.4 Å from the biotin sulfur was seen in difference electron density maps indicating that a small portion of the biotin bound to the protein was oxidized. Reasonable atomic displacement parameters were obtained for an oxygen atom with 0.2 occupancy at this position. Comparison of the wild type biotin complex indicates that this minor oxidized component can be accommodated with no noticeable distortion of the equilibrium structure.</p><p>The final structural model for Y54F consists of two streptavidin subunits (chain A: residues 14-134, B: 15-136), two biotin molecules, three sulfate ions, 215 fully-occupied water molecules, 33 partially occupied waters, and four glycerol molecules. MolProbity31 was used for model validation.</p><p>The structure of Y54F in the absence of biotin was determined and processed in similar ways. The initial model was obtained using the BALBES molecular replacement pipeline32 and PDB entry 1MM9. The final model consists of one streptavidin subunit (residues 16-135), one chloride ion, 89 fully-occupied water molecules, 30 partially occupied waters, and one ethylene glycol molecule.</p><p>Table 1b contains refinement statistics for both structures. Coordinates and structure factors for the biotin complex of the Y54F mutant have been deposited in the Protein Data Bank with identifier 3T6F. The uncomplexed structure has PDB identifier 3T6L.</p><!><p>Starting coordinates for the Y54F-biotin complex simulations were taken from the current x-ray structure. The eight histidine residues in the tetramer were singly protonated to model the ionization state expected for a neutral solution. Hydrogen atoms were added to all protein heavy atoms using the Leap module in AMBER 9.33 The full complex was solvated in a truncated octahedral box containing 19,584 water molecules, and eight sodium counterions were added to maintain charge neutrality for the system.</p><p>The simulation methods and protocol are comparable to that reported previously for solution phase simulations of WT streptavidin.34 Briefly, all calculations were performed using the AMBER ff99 force field35,36 with modifications by Simmerling and co-workers,37 the SPC/E water model38 and a sodium cation model from Åqvist.39 Biotin parameters were taken from previous work by Israilev and co-workers.40 Force calculations were performed with periodic boundary conditions, a 9.0 Å cutoff on real space interactions, a homogeneity assumption to approximate the contributions of long-range Lennard-Jones forces to the virial tensor, and smooth particle-mesh Ewald for long-range electrostatics.41 The SHAKE algorithm42 was used to constrain the lengths of all bonds to hydrogen atoms and the SETTLE algorithm43 was used to constrain the internal geometry of all rigid SPC/E water molecules. A Langevin thermostat44 with collision frequency 3 ps-1 was used to maintain the system temperature. All energy minimizations and dynamics were performed with the PMEMD module of AMBER 9.31 To avoid artifacts arising from reuse of particular sequences of random numbers,45 the random number generator seed was incremented with every restart of the dynamics.</p><p>To prepare the system for equilibrium MD simulations, hydrogen atoms, water molecules and sodium atoms were first relaxed by 2000 steps of steepest-descent energy minimization while crystallographically resolved protein atoms were held in place by 1000 kcal/mol-Å2 position restraints. The protein heavy atoms were then energy minimized while solvent particles were tightly restrained to their new positions, and finally all components of the system were energy-minimized with no restraints. Restrained dynamics of the system were conducted for a total of 450 ps, beginning with a 0.5 fs time step in the constant volume, constant temperature ensemble and 16.0 kcal/mol-Å2 restraints on all crystallographically observed protein atoms. The restraints were gradually reduced to 1.0 kcal/mol-Å2 over the first 150 ps before switching to the constant pressure ensemble, increasing the time step to 1.5 fs, and reducing the restraints to 0.0625 kcal/mol-Å2 over the next 300 ps. Production dynamics were propagated in the constant pressure ensemble with a 1.5 fs time step for 500 ns with no position restraints.</p><p>All MM-GBSA and MM-PBSA calculations were performed using the SANDER module of AMBER 9. In the Generalized Born calculations, the electrostatic solvation energy was calculated using a model developed by Onufriev et al.48 with a protein dielectric constant of 1.0 and solvent dielectric constant of 80. The nonpolar solvation contribution was computed using the LCPO method.49 For the Poisson-Boltzmann calculations, a finite-difference algorithm implemented in Sander was used, with a 0.5Å grid spacing, a 1.4Å solvent probe radius, a protein dielectric constant of 1.0 and solvent dielectric constant of 80. The nonpolar solvation contribution was computed as described above for the Generalized Born calculations. We did not calculate explicit entropy contributions for the binding free energies, since we know from our experimental measurements that the ΔΔS contribution is small (TΔΔS ~ 0.7 kcal/mol). For each method, we employed the "single trajectory approach", I.e., we extracted configurations for the complex, the unbound protein, and the free ligand from our equilibrium trajectories of the complexes, rather than running separate MD simulations for free streptavidin and free ligand to generate configurations for the unliganded protein and free ligand independently. Since the available crystal structures all show that biotin-bound and unliganded streptavidin structures are nearly identical, and since the biotin molecule has only limited conformational flexibility (in the valeric acid side chain) and is bound in essentially identical conformations in both WT and Y54F mutant complexes, we believe that the single trajectory approach is quite reasonable.</p><!><p>Eight streptavidin mutations in the second contact shell were screened based on the change in biotin dissociation rate, Δkoff = koff(mutant) / koff(wild type) (Table II). Two mutations with Δkoff > 50 were selected for crystallography and further analysis: F130L and Y54F. F130L had by far the largest effect on dissociation rate; this mutation and its effects on biotin binding, which were not attributed to changes in dynamics, have been described previously.14 Y54F caused 73-fold faster biotin dissociation at 37 °C and is the focus of this report.</p><!><p>The Y54F mutation has minimal effect on the equilibrium structure of the streptavidin-biotin complex (Figure 1). Minor structural changes are observed in the vicinity of the mutation site (Figure 2, A), but no significant changes occur in the positions of side chains contacting biotin. Table III lists observed streptavidin-biotin hydrogen bond distances in wild type and Y54F streptavidin.</p><p>The removal of the Y54 hydroxyl group disrupts hydrogen bonds to neighboring main-chain atoms and to a bound water molecule (Water 1 in Figure 2, A) and creates a small cavity in the protein which is filled by an additional water (Water 2) in one of the two subunits in the liganded Y54F structure. The molecules in the deposited PDB entry in the Water 1 site are numbers 6248 (A chain) and 6139 (B chain). The Water 2 site is occupied by number 6140 (B chain).</p><!><p>The Y54F mutation causes a loss of binding affinity of 75 ± 10 relative to WT streptavidin at 37 °C, measured using a competitive binding assay for 3H-biotin, corresponding to a decrease in binding free energy of 2.7 ± 0.1 kcal/mol (Table IV; Figure 3 shows competitive binding, calorimetric and kinetic data). The loss of binding energy is due to a large loss of binding enthalpy (ΔΔH°, 3.4 kcal/mol), partially compensated by a more favorable entropy of binding (TΔΔS°, 0.7 kcal/mol, calculated from ΔΔG° = ΔΔH° - TΔΔS°). Activation thermodynamic parameters for Y54F were calculated by fitting all kinetic data to an Eyring model using ΔH≠ and ΔS≠ as the only adjustable parameters; activation parameters for WT were based on previously published values.18</p><!><p>We generated a 500 ns trajectory of the liganded Y54F mutant for analysis and comparison to a corresponding wild type complex trajectory34 and to trajectories for two other mutants we are studying, F130L14 and T106V. The structural fluctuations for Y54F appear to stabilize after ~ 75 ns, as assessed by RMSD values for trajectory snapshots from the starting x-ray coordinates. The RMSD value for core atoms – all atoms excluding loop residues and the amino and carboxy termini – stabilizes at ~ 0.7 Å, and the RMSD for all atoms plateaus at ~ 1.6 Å. We therefore performed all subsequent analyses using the final 425 ns of the MD trajectory.</p><p>The time-averaged structures for both Y54F and the wild type complex are quite similar, indicating that the Y54F mutation has negligible impact on equilibrium structure, consistent with the high-resolution crystallographic results. The simulations reveal a significant increase in structural fluctuations for loop L5,6, located at the edge of the β-barrel core, in liganded Y54F relative to wild type and all other mutants (Figure 4). This small loop is formed by residues K80, N81, N82, Y83 and R84, and its increased mobility is due to the loss of a water-mediated hydrogen bond network between the Y54 hydroxyl group and protein backbone atoms in residues W79 and N49. There are also transient hydrogen bonding interactions between Y54 and backbone atoms from residues N81 and R84 over the course of the wild type MD trajectory. In wild type streptavidin and all other mutants we have studied, this water-mediated hydrogen bond network restricts the range of motion for loop L5,6 in all four streptavidin subunits. However, in the Y54 mutant, loss of the hydroxyl group disrupts the hydrogen bond network (Figure 2, A) and removes restraints that limit motion of the loop residues.</p><p>In an attempt to assess whether the increased structural fluctuation for loop L5,6 in the mutant is simply an artifact of limited configurational sampling, we compared fluctuations for wild type versus F130L and T106V complexes, each computed from independent 500 ns trajectories (Figure 4, B and C). As seen in these plots, only the Y54F mutant exhibits increased fluctuations in loop L5,6. These trends, coupled with the clear structural basis for the increased fluctuations, suggest that the increased loop mobility observed in Y54F is real and not a simulation artifact.</p><p>The increased mobility for loop L5,6 appears to impact biotin binding thermodynamics even though these loop residues are well removed from the ligand binding pocket. The dramatically increased backbone mobility observed for residues K80-N82 contributes to the increased range of motion observed for W79 via direct mechanical coupling through the backbone atoms. In addition, increased mobility of loop residues, particularly N81, reduces packing constraints for side chains that contact W79, thus reducing packing constraints for W79 itself, allowing increased side chain mobility. In the MD simulation, all other streptavidin-biotin contacts besides W79 are perfectly maintained, keeping biotin firmly "locked" in position; i.e., biotin is immobilized and cannot move together with the W79 side chain to maintain consistent interactions. As a result, W79-biotin interactions are disrupted frequently over the course of the MD trajectory.</p><p>We have shown previously that in wild type streptavidin, W79 forms a crucial contact with biotin.17 Mutation of this residue to smaller side chains reduces biotin binding affinity significantly (W79F: ΔΔG = 1.0 kcal/mol; W79A: ΔΔG = 7.9 kcal/mol at 37 °C). Our simulation results suggest that we have reduced biotin binding affinity by degrading the strength of the W79-biotin interaction through increased fluctuations that diminish the contact time of this side chain with biotin. The magnitude of the effect for the Y54F mutation (ΔΔG = 2.7 kcal/mol) is reasonable relative to earlier mutagenesis studies for W79.16-17 In W79F mutant, the F79 side chain maintained good contact with biotin,16 and this mutation had a modest impact on binding affinity.17 In the W79A mutant, the alanine side chain was too small to maintain contact with biotin, and the impact on binding affinity was dramatic.17 The Y54F mutant has an intermediate effect: the increased W79 side chain fluctuations significantly diminish but do not eliminate the biotin interaction as a function of time over the trajectory, and the impact on binding affinity is greater than that for W79F but less dramatic than W79A.</p><p>The increased atomic fluctuations observed for the W79 side chain and adjacent L5,6 loop suggest that Y54F should display an enhanced biotin dissociation rate, and we observe this experimentally (Δkoff = 73). W79 is one of four tryptophan residues that form the so-called "tryptophan collar" which locks biotin into position once it enters the binding pocket.15,16 Increased W79 side chain fluctuations would make the tryptophan collar "looser" and less effective at preventing dissociation. The increased mobility of the L5,6 loop also means the binding site entrance opens wider and more frequently, favoring an enhanced biotin dissociation rate.</p><p>In an attempt to provide more quantitative support for our hypothesis, we have used the equilibrium MD trajectories to estimate the relative binding free energy for biotin to WT versus Y54F streptavidins. Previously, we calculated the absolute binding free energy for biotin to WT streptavidin by using a free energy perturbation technique to implement an effective potential of mean force calculation along a biotin dissociation reaction coordinate.46 We computed a binding free energy of -17.0 ± 3.0 kcal/mol, in excellent agreement with the experimental value of -18.3 ± 1.0 kcal/mol.15 While this approach produced quantitatively accurate results, the procedure is extremely computationally expensive, and the expected statistical uncertainty (± 3.0 kcal/mol) is comparable to the relative binding free energy difference we wish to calculate in this current study. For these reasons, we chose instead to use the mixed molecular mechanics/continuum model methods MM-GBSA and MM-PBSA to extract binding free energy estimates for our WT and Y54F complexes from the equilibrium trajectories.47</p><p>We first utilized the MM-GBSA method to estimate biotin binding free energies, since the Generalized Born continuum model is much less expensive computationally than Poisson-Boltzmann methods. As in all other trajectory analyses, we omitted the first 75 ns of each equilibrium trajectory. We calculated the absolute binding free energy for each subunit independently, and also calculated results for fragments, or sub-blocks, of the full trajectories to check for variations in computed binding free energy as a function of total number of configurations included in the calculations. The MM-GBSA analysis yielded absolute binding free energies that do not agree well with the experimental measurements or earlier free energy perturbation calculations, and the standard deviations in the computed results are quite large. For the WT complex, the calculated binding free energy is -64.5 kcal/mol ± ~30.0 kcal/mol, while the result for the Y54F mutant is -84.6 ± ~25.0 kcal/mol. The computed binding free energies were generally consistent for the full trajectories versus trajectory fragments, although the standard deviations are smaller when trajectory fragments containing only 10,000 – 15,000 configurations are used in the calculations. These results suggest that biotin should bind more tightly to the Y54F mutant by ~20 kcal/mol, although the standard deviations are so large that no firm conclusions can be drawn from these results.</p><p>We next utilized the MM-PBSA method to estimate the binding free energies for each complex. This method is considerably more expensive than the MM-GBSA technique (E.g., calculation of the absolute binding free energy for one subunit requires ~fifteen days of CPU time for each trajectory on a 24-processor Intel cluster), so we have computed PBSA results for only two of four subunits in each complex thus far. Given that the MM-GBSA results were comparable for each subunit, we expect the same will be true for the MM-PBSA calculations. Unlike the Generalized Born calculations, the MM-PBSA method yields results that appear much more physically reasonable, and the standard deviations are smaller as well. For the WT complex, the computed binding free energy is -22.3 kcal/mol ± 11.0 kcal/mol, while for the Y54F mutant the result is -22.6 kcal/mol ± 7.3 kcal/mol. As was the case with the MM-GBSA analysis, the computed binding free energies are consistent whether the full trajectories or only fragments consisting of 10,000 – 15,000 configurations are used. However, unlike the MM-PGSA calculations, using fewer configurations does not reduce the standard deviations for the MM-PBSA method. These results suggest that biotin binds essentially equally well to both WT streptavidin and Y54F mutant, although again the standard deviations are too large to draw any meaningful conclusions.</p><p>While it is somewhat disappointing that neither continuum method provided any compelling evidence for biotin binding preference, some interesting data did emerge from these calculations nonetheless. The molecular mechanics component of these calculations (the "gas-phase" contribution) shows consistently for all subunits that the WT complex is preferred by ~ 6-7 kcal/mol, suggesting that the intrinsic biotin-streptavidin interactions are better in the WT complex. This result agrees reasonably well with the experimental relative binding enthalpy difference, ΔΔH = 3.4 kcal/mol, also favoring the WT complex.</p><!><p>Our biophysical experiments show that the Y54F mutation reduces biotin binding free energy significantly (ΔΔG = 2.7 kcal/mol at 37 °C) and leads to a dramatically increased biotin dissociation rate (Δkoff = 73). The crystal structure reveals no differences relative to the wild type streptavidin-biotin complex in the first contact shell: there are no changes in hydrogen bonding distances or side chain conformations for residues immediately surrounding biotin (Table III), and no perturbation of protein structure surrounding the mutation (Figure 1). However, the loss of the Y54 hydroxyl group disrupts a hydrogen bonding network involving protein backbone atoms of N49, W79, and R84 (Figure 2, A). Our MD simulations for the wild type complex suggest that this water-mediated hydrogen bond network stabilizes the backbone structure for the L5,6 loop. When immobilized, this loop provides packing constraints that maintain the W79 side chain in close juxtaposition to biotin, thus supporting or "stabilizing" the important W79-biotin interaction. In the Y54F mutant, the stabilizing hydrogen bond network is disrupted and the loop residues exhibit dramatically increased mobility. This increased loop mobility in turn leads to an increase in W79 side chain fluctuations which diminish the W79-biotin interaction as a function of time over the duration of the MD trajectory. The time-averaged or equilibrium structure from the MD trajectory is consistent with the Y54F crystal structure, and suggests no meaningful change in the W79-biotin interaction. Hence, the diminished W79-biotin interaction we observe is a dynamic effect, and we propose that the modified dynamics are responsible for the reduced biotin binding free energy observed experimentally.</p><p>The MM-GBSA and MM-PBSA results did not provide any additional compelling support for our hypothesis to explain why biotin binds more favorably to WT streptavidin, in part because the standard deviations in our calculations are much larger than the ΔΔG value we were attempting to calculate. The MM-GBSA calculations yielded absolute binding free energy estimates that were physically unreasonable. By contrast, the MM-PBSA results were only ~20-30% too large, when compared to experimental measurements. The superior performance of the MM-PBSA method for absolute free energy estimates is not surprising, and is consistent with previous studies.50,51 We suspect that neither continuum method can represent adequately the detailed hydration structure present in the binding site for the unliganded WT and Y54F streptavidins. We know from our numerous streptavidin crystal structures that there are typically four to six water molecules that form specific hydrogen bonds in the binding pocket when biotin is absent. We also know from our previous MD simulations that this hydrogen bonding network is dynamic and in exchange with bulk water, even though the individual water positions exhibit occupancy values close to one in the crystal structures and in the simulations. We believe that our previous explicit solvent calculation of absolute binding free energy for biotin to WT streptavidin performed well in part because those simulations modeled the binding site "rehydration" accurately upon biotin dissociation, based on comparisons to the resolved water molecules in the corresponding crystal structures. It is possible to perform mixed explicit/implicit solvent calculations, including selected water molecules explicitly in the calculations while treating the remainder of the solvent with a GBSA or PBSA method, and that approach may be necessary to get more reliable results for the complexes studied here. Mixed explicit/implicit solvent model calculations have been shown to provide improved results in some cases, although they can be quite sensitive to exact number and placement of explicit water molecules.52 These calculations would be exceptionally tedious to perform for the biotin-streptavidin complexes, since explicit water molecules would have to be selected carefully from each snapshot of an equilibrium MD trajectory for unliganded streptavidin to create the ensemble of configurations for unliganded protein to use in the subsequent MM-PBSA calculations.</p><p>The continuum model calculations did provide some support for our hypothesis. As noted above, all calculations consistently showed that gas-phase WT complex is favored over the Y54F mutant complex by 6-7 kcal/mol, in respectable agreement with the experimental ΔΔH measurement that favors WT complex by 3.4 kcal/mol. Since we know from our crystal structures that the hydration of WT and Y54F complexes is quite similar, for both the biotin-bound and the unliganded protein, we expect that solvation/desolvation free energies may be comparable for each complex. We also know from the crystal structures that biotin adopts the same conformation in both WT and mutant complexes, so there should be little or no difference in ligand conformational strain energy for the two complexes. Therefore, it is perhaps not too surprising that the gas-phase molecular mechanics contribution from the MM-PBSA/MM-GBSA calculations, when combined with the experimental ΔΔS result, exhibits reasonable agreement with the experimental binding free energy difference.</p><p>The magnitude of the impact of the Y54F mutation on biotin binding seems reasonable in the context of previous mutagenesis involving residue W79.16,17 The increased local dynamics we observe with this point mutant has a more significant impact than the conservative binding site mutation, W79F, but a less significant impact than the W79A mutation, which effectively eliminates the biotin interaction.</p><p>The MD simulations provide a reasonable explanation for the enhanced biotin dissociation rate measured for the Y54F mutant. The increased W79 side chain fluctuations should make the tryptophan collar looser or weaker and facilitate biotin dissociation, since in wild type streptavidin this tryptophan collar represents the primary barrier for biotin dissociation.17, 46 The increased mobility observed for the L5,6 loop also favors an increased dissociation rate since the binding pocket entrance is open wider and more frequently in Y54F than in the wild type complex.</p><p>The large loss of equilibrium binding enthalpy (3.4 kcal/mol at 37 °C) and small gain in binding entropy (0.7 kcal/mol) for Y54F are consistent with a mutation that causes a weakened interaction with biotin and a small gain in configurational entropy in the bound state. Increased fluctuations of residue W79 would reduce contact time for this side chain with biotin, reducing van der Waals interactions and binding enthalpy. A proportionate increase in binding entropy might also have been expected, but only a small increase was observed. However, other factors are likely to be significant. For example, Y54F gains two more bound water molecules during biotin binding in the area immediately surrounding the mutation (Waters 1 and 2 in Figure 2, A), while wild type streptavidin gains none. (These waters are present in liganded Y54F, but absent in unliganded Y54F (not shown), while in the wild type structure, Water 1 is present and Water 2 is absent regardless of the binding state.) Binding additional waters is unfavorable entropically for Y54F and may oppose any gain in configurational entropy; however, a much more detailed biophysical and computational study would be required to estimate the contributions of bound waters to the binding energetics.</p><p>In summary, we have used a combination of biophysical measurements, x-ray crystallography, and MD simulation to characterize a streptavidin point mutation distal from the biotin binding pocket, Y54F. This mutation reduces biotin binding free energy significantly but has no observable effect on equilibrium structure, either in the binding pocket or at the mutation site. Our combined experimental and computational analysis suggests that the reduced biotin binding affinity is a result of increased structural fluctuations of the W79 side chain, an important contact residue for biotin. The increased W79 side chain fluctuations are coupled to increased mobility in the adjacent L5,6 loop residues K80-N82, which in turn is caused by disruption of a hydrogen bonding network involving the Y54 hydroxyl group. These results suggest that dynamical effects can impact ligand binding thermodynamics and dissociation kinetics even in the absence of observable equilibrium structural changes in the first contact shell. These results also demonstrate how distal point mutations can serve as the origin of dynamical changes that are relayed mechanically to effect changes remotely – in this example, altered protein-ligand binding.</p>

PubMed Author Manuscript

A [4Fe-4S]-Fe(CO)(CN)-L-cysteine intermediate is the first organometallic precursor in [FeFe] hydrogenase H-cluster bioassembly

Biosynthesis of the [FeFe] hydrogenase active site (the \xe2\x80\x9cH-cluster\xe2\x80\x9d) requires the interplay of multiple proteins and small molecules. Among them, the radical S-adenosyl-methionine enzyme HydG, a tyrosine lyase, has been proposed to generate an Fe(CO)2(CN) moiety-containing complex that is eventually incorporated into the H-cluster. Here we describe the characterization of an intermediate in the HydG reaction: a [4Fe-4S][(Cys)Fe(CO)(CN)] species, termed \xe2\x80\x9cComplex A\xe2\x80\x9d, in which a CO, a CN\xe2\x88\x92 and a cysteine (Cys) molecule bind to the unique \xe2\x80\x9cdangler\xe2\x80\x9d Fe site of the auxiliary [5Fe-4S] cluster of HydG. Identification of this intermediate\xe2\x80\x94the first organometallic precursor to the H-cluster\xe2\x80\x94validates the previously hypothesized HydG reaction cycle and provides a basis for elucidating the biosynthetic origin of other moieties of the H-cluster.

a_[4fe-4s]-fe(co)(cn)-l-cysteine_intermediate_is_the_first_organometallic_precursor_in_[fefe]_hydrog

<!>Intermediates in the HydG reaction with one equivalent of Tyr.<!>13CO/13C15N in Complex A.<!>Bridging Cys molecule in Complex A.<!>Dangler 57Fe in Complex A.<!>Dicyano analogue of Complex A.<!>Discussion<!>Generation of 3-13C-Cysbridge and 57Fedangler labeled HydG.<!>The \xe2\x80\x9cstandard condition\xe2\x80\x9d to generate HydG reaction mixture containing Complex A<!>Generation of the di-CN analogue of Complex A.<!>Data availability.

<p>[FeFe] hydrogenases catalyze the efficient interconversion between H2 and H+/e− and are involved in many metabolic processes that balance the redox potentials in cells.1 There is also considerable interest in their application in biofuel cells.2,3 The active site of [FeFe] hydrogenases, the H-cluster, consists of a canonical cysteine (Cys)-bound [4Fe-4S]H subcluster linked to a [2Fe]H subcluster in which the two Fe centers are coordinated by CO, CN−, and a bridging azadithiolate ligand (Fig. 1a).4,5 The unique structural features and catalytic activity of the H-cluster have stimulated interest regarding its biosynthesis.2,6–12 Recent studies have demonstrated that assembly of the [2Fe]H subcluster requires several iron-sulfur cluster-containing maturases: HydE, HydF, and HydG (Fig. 1a).13–21 Among them, HydG is a bifunctional radical S-adenosyl-methionine (SAM) enzyme that cleaves its substrate tyrosine (Tyr) to generate CO and CN− (Fig. 1b)16,22–24 and forms an Fe(CO)2(CN) moiety-containing organometallic precursor that is eventually incorporated into the H-cluster (Fig. 1c).25</p><p>These two reactions are carried out by HydG using two Fe-S clusters with distinct roles.8,16,24–30 A SAM-binding [4Fe-4S]RS (RS = radical SAM) cluster bound near the N-terminus initiates the radical chemistry that leads to Tyr cleavage. Upon one-electron reduction of the [4Fe-4S]RS cluster, SAM is cleaved to a 5'-dAdo radical, which likely abstracts an H atom from the amino group31 of Tyr. This induces Cα-Cβ cleavage to give a 4-hydroxybenzyl radical (4-OB•),27 along with dehydroglycine (DHG) which is in turn converted into CO and CN− through mechanisms under investigation (Fig. 1b). Near the C-terminus, HydG harbors a unique [5Fe-4S] auxiliary cluster, recently identified by X-ray crystallographic analysis with support from biochemical and spectroscopic studies.32 The auxiliary cluster contains a high spin "dangler" Fe2+ (S = 2) chelated by a bridging Cys molecule by which it is linked to a conventional [4Fe-4S]+ cluster (S = 1/2) through the cysteine S, forming an S = 5/2 resting state in the reduced form (Fig. 1c).32,33 The dangler Fe and the bridging Cys were shown to be labile towards chelating agents such as EDTA and CN−.32,33 Several species have been identified in the HydG reaction in addition to Tyr cleavage product (p-cresol): stopped-flow Fourier transform infrared (SF-FTIR) spectroscopic studies indicated the sequential formation of two CO/CN−-containing Fe complexes,25 and electron paramagnetic resonance (EPR) spectroscopic studies revealed the presence of a [4Fe-4S]aux-CN species with the synthon built upon the dangler Fe presumably released (Fig. 1c).33</p><p>Taken together, these results lead to a hypothesized mechanistic framework of HydG (Fig. 1c),33 in which CO and CN− bind to the dangler Fe in the auxiliary cluster to generate a discrete [(Cys)Fe(CO)2(CN)]− complex as the reaction product. In this mechanism, it has been suggested that Cys serves, at a minimum, to deliver the [Fe(CO)2(CN)]+ moiety to the next maturase, and that the Cys ligand in this complex may be further processed to install the azadithiolate bridging ligand.33 These proposals were supported by evidences that Cys binds the auxiliary [4Fe-4S] cluster,33 and that Cys is required for generating organometallic Fe(CO)x(CN)y intermediates.34 However, it has not been demonstrated that the Cys-chelated dangler Fe is the site of organometallic complex formation; i.e., there has not been any direct observation of the proposed [4Fe-4S][(Cys)Fe(CO)x(CN)y] species. We reasoned that the first organometallic intermediate proposed—[4Fe–4S][(Cys)Fe(CO)(CN)], termed "Complex A" (Fig. 1c)—could serve this purpose. In particular, we expected such a species to adopt an S = 1/2 ground state in its reduced form, owing to the low spin S = 0 dangler Fe2+ that results from the binding of the strong field π-acid ligands CO and CN−, leaving the S = 1/2 [4Fe–4S]+aux cluster as the sole source of paramagnetism.</p><p>In this paper, we report the spectroscopic characterization of Complex A trapped during HydG turnover. The EPR spectroscopic studies herein establish the connectivity between the C-terminal auxiliary [4Fe-4S] cluster, the bridging Cys molecule, the dangler Fe, and the CO and CN− ligands in Complex A, and thereby provide direct evidence that the first organometallic precursor to the H-cluster is a HydG-bound [4Fe-4S][(Cys)Fe(CO)(CN)] species. In addition, a dicyano analogue of Complex A, [4Fe-4S][(Cys)Fe(CN)2], was generated non-enzymatically, which further demonstrates the formation of Fe(CO)x(CN)y species on the dangler Fe site.</p><!><p>When the HydG reaction was performed by using excess dithionite and SAM, but only one equivalent Tyr, and freeze-quenched at 24 s (the "standard condition", see Methods), the resulting sample exhibited a complex set of signals near g ~ 2 in its continuous wave (CW)-EPR spectrum (Fig. 2a, top black trace). Simulation of this spectrum reveals three species (Fig. 2a, Supplementary Information): species 1 (red trace) with g = [2.058, 1.922, 1.882] is assigned to Complex A for reasons we discuss below; species 2 (blue trace) with g = [2.009, 1.881, 1.842] originates from the SAM-bound [4Fe-4S]+RS cluster, as previously established;26,27 and species 3 (green trace) with g = [2.044, 1.942, 1.904], the origin of which is currently unknown. The spectral composition of the 24 s reaction sample was further confirmed by deconvolution of its Q-band electron spin-echo detected EPR spectrum (Fig. 2b) and the corresponding pseudo-modulated spectrum (Supplementary Fig. 1).</p><p>In determining the identity of species 1, we considered the following observations. First, the intensity of EPR signal from species 1 decreases as the reaction proceeds (Supplementary Fig. 2 and 3), which is consistent with the kinetics of Complex A in our previous SF-FTIR studies.25 Second, the intensity of this signal also decreases with increasing equivalents of Tyr (Supplementary Fig. 4), which explains why it was not as obvious in previous studies in which 10~15 equivalents of Tyr were used.27 Third, the temperature profile of species 1 is similar to that of a typical S = 1/2 [4Fe-4S]+ cluster (Supplementary Fig. 5). Fourth, compared in Supplementary Table 1 are the g tensors for several HydG auxiliary cluster-derived species. The g tensor of species 1 is distinct from that of the resting state of HydG auxiliary cluster that has a ground state of S = 5/2 (geff = [9.5, 4.7, 4.1, 3.7]). It is also different from two other S = 1/2 species observed previously: the cyanide bound HydG auxiliary cluster ([4Fe-4S]+aux-CN, g = [2.09, 1.94, 1.93]) observed in the prolonged reaction (20 min) sample,32,33 and the dangler Fe-deficient HydG auxiliary cluster ([4Fe-4S]+aux-Cys, g = [2.064, 1.895, 1.865]).33 These facts suggest that species 1 could be Complex A. In what follows, we test this hypothesis by employing electron nuclear double resonance (ENDOR) and hyperfine sublevel correlation (HYSCORE) spectroscopy to probe the structure/ligand environment of species 1 using isotopically labeled Tyr (13C and 15N), Cys (3-13C), and Fe (57Fe).</p><!><p>HydG-catalyzed cleavage of isotopically labeled Tyr (13C and 15N) generates 13C/15N labeled Fe(CO)x(CN)y intermediates (Fig. 1c and Fig. 3a).25,33 In Fe(CO)x(CN)y intermediates, the dangler Fe to which CO and CN− are bound is proposed to be a diamagnetic, low spin Fe2+ center; as such, the electron spin density is expected to reside primarily on the [4Fe-4S]+aux cluster, and the 13CO, 13CN, C15N hyperfine coupling interactions (HFI) are expected to be small, making these I = 1/2 nuclei good targets for Mims-ENDOR (Supplementary Methods).</p><p>The Mims-ENDOR spectrum recorded at g = 1.922 of a sample generated under the standard condition (vide supra) using U-13C9-Tyr as the substrate exhibited a sharp doublet centered at the Larmor frequency of 13C, with a splitting of ~0.27 MHz (Fig. 3b). To clarify which EPR species contribute to the 13C ENDOR signals, we collected field-dependent 13C Mims-ENDOR spectra (Fig. 3b) under optimized conditions (Supplementary Fig. 6), at field positions indicated by the arrows shown in Fig. 2b. Mims-ENDOR signals (Fig. 3b) were not observed at g = 1.851 (the rightmost arrow, Fig. 2b) where only species 2 (SAM-bound [4Fe-4S]+RS cluster) is present, but were observed at g = 1.882 (the second rightmost arrow, Fig. 2b) where both species 1 and 2 are present while species 3 is not present (Fig. 3b), clearly indicating that these ENDOR signals arise from species 1. Consistent with this assignment, the 13C Mims-ENDOR signals were observed across the absorption envelope of species 1 (Fig. 2b, g = 2.055 to 1.882; Fig. 3b).</p><p>To clarify which carbon(s) in Tyr give rise to the observed 13C ENDOR signals, we prepared reaction samples under the standard condition using selectively 13C-labeled Tyr, specifically either 1-13C-Tyr which generates 13CO, or 2-13C-Tyr which generates 13CN− (Fig. 3a). Interestingly, 13C Mims-ENDOR spectra of these two samples recorded at g2 (1.922) of species 1 were nearly indistinguishable, and the signals were not observed when SAM is not present (Supplementary Fig. 7), ruling out its origin from any unreacted Tyr species bound to the clusters. Further field-dependent 13C Mims-ENDOR measurements indicate that the HFI from 13CO ([0.10, 0.30, 0.27] MHz, Fig. 3c) and 13CN− ([0.15, 0.30, 0.28] MHz, Fig. 3d) are almost identical, and that they both contribute to the ENDOR spectra in Fig. 3a where both 13CO and 13CN− HFI are detected. Based on these 13C Mims-ENDOR experiments, we conclude that species 1 corresponds to Complex A as observed by FTIR spectroscopic studies.25</p><p>We also looked for the evidence for C15N− binding to the dangler Fe by probing the 15N HFI (generated from [U-13C9, 15N]-Tyr). The small 15N HFI, observed in X-band HYSCORE spectra of the reaction sample (Supplementary Fig. 8), further supports our assignment of species 1 as Complex A.</p><!><p>Having assigned species 1 to Complex A, we then looked for evidence of the bridging Cys in this species. To this end, we selectively installed 3-13C-Cys in the bridging position of the auxiliary [5Fe-4S] cluster of HydG using our previously established protocol (Fig. 3a, see Methods).33 When the reaction was performed using this 3-13C-Cysbridge-labeled HydG and U-13C9-Tyr at standard condition, the 13C Mims-ENDOR spectra of the resulting sample showed the same small 13C HFI (aiso = 0.23 MHz, Fig. 3e, red trace) from 13CO/13CN, as observed with unlabeled HydG (Fig. 3b, vide supra). In addition, a second set of 13C ENDOR signals with a larger hyperfine splitting was observed (Fig. 3e, simulated with blue trace), which must originate from the labeled 13C on the bridging Cys. A similar field-dependent 13C Mims-ENDOR study (vide supra) reveals that this set of signals appears concurrently with the 13CO/13CN− signals (Fig. 3e, from g = 2.055 to g = 1.882), which indicates that it also originates from species 1, i.e., Complex A.</p><p>The ENDOR of 3-13C-Cysbridge in Complex A can be simulated using a 13C hyperfine tensor of A = [1.00, 0.20, 1.00] MHz (Fig. 3e, blue trace), with aiso = 0.73 MHz and T = −0.27 MHz (sign relative to aiso). Importantly, this 13C HFI is distinct from that reported for the Cys-bound HydG auxiliary cluster, [4Fe-4S]+aux-[3-13C-Cys] (A = [0.83, 0.83, 1.03] MHz).33 The dipolar component of this 13C HFI in Complex A (|T| = 0.27 MHz) approximately corresponds to a distance between the labeled carbon and the nearest Fe in the [4Fe-4S]aux cluster of ~4 Å (Supplementary Information). This distance is broadly consistent with the proposed structure of Complex A in which the Cys C3 carbon is separated from the [4Fe-4S] cluster by a bridging thiolate sulfur atom.</p><!><p>We next sought to verify the presence of the dangler Fe in Complex A and to clarify its spin state by measuring the 57Fe HFI. A sample of 57Fedangler-labeled HydG was generated using a previous established protocol33 (Fig. 3a) and was subjected to the standard reaction conditions to generate the 57Fedangler-labeled Complex A (Fig. 4c). The Q-band HYSCORE spectrum of this sample collected at g = 2.05 reveals crosspeaks on the antidiagonal line centered at the 57Fe Larmor frequency (1.63 MHz at 1186 mT, Fig. 4a). These signals are not present in the corresponding natural abundance spectrum recorded under the same conditions (Fig. 4b) and must therefore be attributed to the dangler 57Fe.</p><p>Field-dependent HYSCORE spectra were collected and simulated to extract the 57Fe HFI tensor, which is nearly isotropic with A = [0.45, 0.30, 0.50] MHz, aiso = 0.42 MHz, and a major |T| ~ 0.07 MHz (Fig. 4d). The spin density on the dangler 57Fe is estimated as follows:35 a Fermi contact (aiso) of 0.42 MHz corresponds to 0.056% spin density in the 4s orbital, and the anisotropic contribution (|T| ~ 0.067 MHz) gives a spin density of 0.24% in 3d orbitals, adding up to a total spin density of ~0.3% on the dangler 57Fe. The small 57Fe spin density validates the diamagnetic nature of the dangler Fe upon binding to CO/CN−, and is consistent with the small 13C HFI detected for 13CO/13CN−. In the point dipole approximation, the dipolar part of 57Fe HFI corresponds to a distance between the dangler Fe and the nearest Fe in the [4Fe-4S] cluster of ~4 Å (Supplementary Information). For comparison, the same distance in the X-ray crystal structure of HydG is 4.1 Å (PDB ID code: 4WCX32). Due to the likely structural changes upon CO/CN− binding (and the bridging Cys molecule was absent in the X-ray structure), this HYSCORE-derived distance is reasonably consistent with the X-ray crystallographic structure, and it suggests that the dangler Fe is close to the [4Fe-4S]aux cluster in Complex A.</p><p>The 57Fe we observed in Complex A has the smallest 57Fe HFI that has been reported to date. A few other examples of weak 57Fe HFI are summarized in Table 1: a reaction intermediate of LipA has 57Fe aiso = 1–2 MHz between a formally diamagnetic [4Fe-4S]2+ cluster and an carbon-centered organic radical;36 the distal Fe in the Hox-CO state of [FeFe] hydrogenase H-cluster has 57Fe aiso = 0.8 MHz (from Desulfovibrio desulfuricans DdH37) or 1.3 MHz (from Chlamydomonas reinhardtii HydA138 reconstituted with synthetic [57Fe2(adt)(CO)4(CN)2]2-); the Ni-A, Ni-B and Ni-C states in [NiFe] hydrogenase from Desulfovibrio gigas have 57Fe aiso = 0.8–1.4 MHz between a formally diamagnetic Fe2+ and the Ni center.39,40 In all cases, the 57Fe coupled to paramagnetic centers are either formally diamagnetic, or have very small spin density.</p><p>Taken together, the 13C and 57Fe HFI values demonstrate that Complex A is comprised of a low-spin dangler Fe2+ center bound to a CO and a CN−, and that it is covalently linked to the [4Fe-4S]aux cluster through a bridging Cys molecule. Clarifying the structure of this intermediate provides direct evidence that HydG generates [Cys][Fe(CO)x(CN)y] species as the reaction product(s). Interestingly, Complex A is an unusual S = 1/2 [5Fe-4S] cluster, with the only similar example being the sirohaem-[5Fe-4S] cluster in which the sirohaem iron (with CO or CN− bound) is weakly coupled to a [4Fe-4S] cluster through a bridging Cys residue.41–43 In this regard, Complex A also shares some similarity with the recently identified hydride state (Hhyd) of the [FeFe] hydrogenase in which two low-spin Fe2+ centers are linked to a paramagnetic [4Fe-4S]+ cluster through a bridging Cys residue.44–46</p><!><p>We further tested the possibility of generating Complex A and its analogs in non-enzymatic manners, that is, by incubating HydG with CO and/or CN− and comparing the corresponding EPR signals to species 1. Although such attempts to generate Complex A have not yet proven fruitful, we were able to generate a cyanide adduct of HydG (Fig. 5a). In a cyanide titration experiment, we observed a new S = 1/2 species on CW-EPR upon addition of 12 eq. of K13CN to HydG, and the signal intensity reached a maximum when 50 eq. of K13CN was added (Supplementary Fig. 9). The CW-EPR spectrum of the "HydG + 50 eq. K13CN" sample is simulated with three species, the most abundant new species among which (Fig. 5b, red trace) has g = [2.054, 1.927, 1.879], almost identical to that of Complex A (Supplementary Table 1). Based on this g tensor similarity, we tentatively assigned this EPR species to the dicyano analog of Complex A, [4Fe-4S][(Cys)Fe(CN)2] (Fig. 5a), though a [4Fe-4S][(Cys)Fe(CN)(L)] structure cannot be ruled out (L is a different ligand that completes the octahedral coordination sphere of the dangler Fe). 13CN− binding to dangler Fe was further ascertained by field-dependent 13C Mims-ENDOR studies (Fig. 5c). The observed ENDOR signals, simulated with a 13C HFI of A = [0.20, 0.30, 0.30] MHz (Fig. 5c, red trace), are also similar to the 13CO/13CN− signals in Complex A, but distinct from that in the previously reported [4Fe-4S]aux-13CN species (A = [−5.0 −4.0 0.9] MHz).32 Observation of this cyanide adduct provides further evidence that the dangler Fe site is involved in the formation of Fe(CO)x(CN)y species.</p><!><p>Using insights from previous work and results in this study, the molecular-level mechanism of H-cluster biosynthesis may be summarized as follows. The radical SAM cluster of HydG initiates cleavage of Tyr into p-cresol and DHG—a CO/CN− precursor. At the auxiliary cluster of HydG, CO/CN− add to a dangler Fe to generate a [4Fe-4S][(Cys)Fe(CO)(CN)] intermediate, as demonstrated here. A second Tyr cleavage-generated CO/CN− pair further converts this intermediate into a [Fe(CO)2(CN)]-moiety containing complex, proposed to be [(Cys)Fe(CO)2(CN)]. Two equivalents of the latter Fe complex may be processed by HydE to generate the azadithiolate bridging ligand from the bound Cys fragment, followed by further processing on HydF to form the H-cluster-like 2Fe precursor. Alternatively, the azadithiolate ligand may be sourced from other molecular precursors. At the end of this cascade, HydF is thought to deliver a 2Fe subcluster to apo-hydrogenase to give active hydrogenase.</p><p>Many important questions about this process remain, including to which maturase enzyme the [(Cys)Fe(CO)2(CN)] complex is delivered; how the azadithionate bridge is formed and whether the HydG-derived [(Cys)Fe(CO)2(CN)] is a precursor to it; and at what point the first 2Fe center is formed. The results reported here provide a useful foundation for studying these important problems by definitively showing that HydG generates [Cys][Fe(CO)x(CN)y] species. Further deploying these in vitro selective isotope-labeling strategies and advanced spectroscopic methods should enable additional mechanistic insights to be gleaned.</p><!><p>This procedure is essentially the same to that for labeling the dangler Fe reported previously.33 As-isolated HydG was concentrated to ~300 μM. Sodium dithionite, SAM, 3-13C-Cys was added to 3 mM, and EDTA to 600 μM (in that order). This mixture was incubated for 10 min at room temperature, diluted by 10-fold with a buffer containing 3 mM dithionite, SAM and 3-13C-Cys, and concentrated to the original volume by using 30 kDa cutoff Amicon centrifugal filters. To the dangler Fe-removed protein was then added 57Fe2+ (see Supplementary Methods) to a final concentration of ~1 mM. This labeled protein sample was used immediately for HydG reactions.</p><!><p>As-isolated or labeled HydG was concentrated to ~800 μM. SAM and dithionite was added to a final concentration of ~8 mM. The reaction was initiated by transferring 60 μL of this solution into an EPR sample tube (Ka-band quartz tube with an outer diameter of 2.4 mm and inner diameter of 2.0 mm) and mixing with 10 μL of 5 mM Tyr (labeled as desired) at the bottom of the EPR tube. The final concentrations for HydG and Tyr were ~700 μM. For X-band samples, a 150 μL protein solution was mixed with 25 μL Tyr in an X-band tube. The reaction was mixed quickly by pipetting up and down once, rapidly transferred outside of the glovebox and freeze-quenched at different time points. With this procedure, we were able to quench the reaction as early as 24 s. The EPR samples were stored in liquid nitrogen prior to analysis.</p><!><p>As-isolated HydG was concentrated to ~600 μM. Sodium dithionite and SAM were added to a final concentration of 6 mM. To 50 μL of the enzyme solution was added 10 μL 60 mM K13CN solution. After brief mixing, the solution was transferred to EPR sample tubes mentioned above and frozen in liquid nitrogen prior to analysis.</p><!><p>The data collected in this study are available from the corresponding author upon request (including data presented in the main text and in the Supplementary Information).</p>

PubMed Author Manuscript

Remodeling of the Natural Product Fumagillol Employing a Reaction Discovery Approach

In search for new biologically active molecules, diversity-oriented synthetic (DOS) strategies break through the limitation of traditional library synthesis by sampling new chemical space. Many natural products can be regarded as intriguing starting points for DOS, wherein stereochemically rich core structures may be reorganized into chemotypes which are distinctly different from the parent structure. Ideally such transformations should be general and involve few steps in order to be suited for library applications. With this objective in mind, the highly oxygenated natural product fumagillol has been successfully remodeled in several ways utilizing a reaction discovery-based approach. In reactions with amines, excellent regiocontrol in a bis-epoxide opening/cyclization sequence can be obtained by size-dependent interaction of an appropriate catalyst with the parent molecule, forming either perhydroisoindole or perhydroisoquinoline products. Perhydroisoindoles can be further remodeled by cascade processes to afford either morpholinone or bridged 4,1-benzoxazepine-containing structures.

remodeling_of_the_natural_product_fumagillol_employing_a_reaction_discovery_approach

<!>Results and Discussion

<p>In order to identify pharmacological tools for biological processes, compound discovery must expand beyond the sp2-dominated synthetic libraries common in biological screening.1 Diversity-oriented libraries have been demonstrated to occupy areas of chemical space not normally accessed by more traditional planar, heterocyclic libraries.2–4 One approach to access increasingly diverse libraries would employ natural products as starting scaffolds. A number of studies have exploited natural products as starting materials, including use of α-santonin5–7 and (−)-shikimic acid8,9 to identify biologically active molecules including 5-lipoxygenase inhibitors and aurora A kinase ligands. Other methods have focused on altering the core framework of natural products to create small collections of structurally unique compounds. A method involving catalytic, site-selective derivatization of complex natural products (e.g. erythyromycin) has been demonstrated by the Miller group at Yale.10,11 In another study, the macrocyclic diterpenoid lathyrane was converted into a small collection of polycyclic structures using transannular reactions.12 The Miller group from Notre Dame has incorporated oxazine heterocycles into natural products bearing 1,3-butadiene subunits employing iminonitroso Diels-Alder cycloadditions.13,14</p><p>Further evolution of these ideas would involve the creation of a diverse library of remodelled structures derived from a natural product, each one significantly different from the parent compound. Transformations utilized should allow for the incorporation of new functionality and ideally be carried out in a single-step or tandem processes. We therefore initiated a reaction discovery-based approach15–18 that meets these criteria employing a readily available natural product as a starting point for chemically diverse library synthesis (Figure 1a). The highly oxygenated natural product fumagillol (2) was chosen as the reactivity and proximity of the two epoxides present site(s) for potential chemistry, while the hydroxyl and alkene groups offer additional functionality for further diversification. Crude fumagillin (1), a natural product which is readily available from the fermentation broth of Aspergillus fumigates and can be hydrolyzed to fumagillol (2) (see Supplementary Information),19,20, has generated significant interest as both a synthetic target21 and for its anti-angiogenic properties.22–24 We envisioned a series of tandem processes which could remodel fumagillol into novel chemotypes as dictated by either by catalyst or reaction partner choice. Herein, we report our initial studies aimed at remodeling fumagillol through Lewis acid-promoted addition of amines.</p><!><p>A reaction screen15,16 was first undertaken to explore sequential aminolyses of the 1,4-bis-epoxide. We anticipated that the sequence would be initiated at the spirocyclic epoxide, thereby mimicking the reactivity of fumagillin with aminopeptidase MetAP-2, the putative mode of its antiangiogenic activity.23 An initial reaction screen (see Supporting Figures 1–5) with twelve Lewis acids and four amines resulted in the conversion of the bis-epoxide motif into perhydroisoindole (3)25–27 and/or perhydroisoquinoline (4),28–30 compounds which were identifiable through several characteristic signals in 1H NMR spectra. Best results were obtained using p-anisidine and a metal triflate catalysis. Preliminary optimization of this transformation demonstrated that 2,6-di-tert-butyl-4-methylpyridine (DTBMP) as proton scavenger significantly improved yields, presumably by buffering adventitious triflic acid.31,32 Several metal triflates were subsequently investigated in a second screen and a linear correlation was found between the atomic radius of the metal catalyst and the distribution of isomeric products (Figure 1b). As metal size increased, perhydroisoindole product 3 was increasingly favored. Lanthanum triflate proved to be optimal for production of 3 with >95:5 regioselectivity (entry 2). Conversely, the smaller, bivalent-metal Zn(OTf)2 favored formation of perhydroisoquinoline 4 (entry 8, 13:87), thereby allowing access to either isomer simply by changing the catalyst.33–35 In the absence of a catalyst, no reaction occurred and fumagillol was fully recovered.</p><p>Further optimization using La(OTf)3 and Zn(OTf)2 catalysts was next pursued. The transformations were robust and did not require inert atmosphere, nor special precautions for anhydrous solvent. Other nonpolar solvents provided similar regioselectivity, though toluene proved to be optimal, in which case catalyst loading could be reduced to 10 mol% while maintaining reasonable reaction times. Production of 3 was ultimately optimized using La(OTf)3 to 91% isolated yield (91:3 regioselectivity), while 4 could be obtained with Zn(OTf)2 in 76% yield (9:76 regioselectivity).</p><p>Bis-epoxide opening, and in particular, catalyst-controlled regioselectivity, proved to be quite general (Table 1). A variety of electron-rich and electron-deficient anilines produced either heterocyclic motif (entries 1 – 5), with La(OTf)3 catalysis forming predominantly perhydroisoindoles 5 and Zn(OTf)2 yielding perhydroisoquinolines 6. In the case of the La(III)-promoted reactions, there was a direct correlation of the nucleophilicity of the aniline with the reaction rate, with electron-rich amines reacting faster. The rate of formation for perhydroisoquinoline 6 under Zn(II)-catalysis was largely unaffected by the electronics of the aniline, until a sufficiently electron deficient analogue (entry 5) was used. Thus, with p-trifluoromethylaniline, the reaction was significantly slower than with the more electron-rich anilines, which all proceeded at approximately the same rate. Heteroaryl amines including 2-aminopyridine and 2-aminothiazole failed to react using either La(OTf)3 or Zn(OTf)2, in which case fumagillol was fully recovered.</p><p>More basic amines were also well tolerated in the reaction (entries 7 – 9), with La(III) again proving to be optimal for perhydroisoindole formation. It was necessary, however, to increase the catalyst loading to 50 mol% in order to obtain reasonable reaction times, presumably due to the greater basicity of these amines leading to tighter interaction with the catalyst. An even greater reduction in reaction rate was observed with Zn(II) catalysis, rendering the reaction unacceptably slow (60 h, approx. 5–10% conversion). Use of Mg(OTf)2 (50 mol%) as the catalyst, however, also led predominantly to the desired perhydroisoquinoline products (6f – 6g) in acceptable reaction times (entries 7 – 9, 16 – 36 hrs). Addition of aromatic and aliphatic secondary amines were also carried out providing highly substituted tetrahydrofuran products 7 and 8, both isolated as triflate salts (Table 1, entries 10 and 11).</p><p>Insights into the interaction of metal catalysts with fumagillol were achieved through a series of 13C NMR experiments (Figure 2, Supplementary Figure 6). In the 13C NMR spectra of fumagillol obtained with 2 mol% of paramagnetic catalysts Yb(OTf)3 (r = 175 pm) or Fe(OTf)2 (r = 140 pm),35,36 broadening of the C5 and C6 resonances was observed, indicating that different sized metals are preferentially bound to the pocket formed by the hydroxyl and methoxy groups of fumagillol (7). The same interaction was not observed in the C6-silylated analogue 10, with only modest broadening of the C2, C1′, and C2′ signals observed.</p><p>To probe the importance of the interactions observed by 13C NMR, selectivity of silyl ether 11 under the optimized reaction conditions was evaluated. When 11 was reacted with several anilines, the regioselectivity obtained with La(III) catalysis inverted, leading predominantly to perhydroisoquinoline products as originally found in the Zn(II)-catalyzed reactions (Figure 2b). By comparison, reaction of 11 using Zn(II) became more selective to afford perhydroisoquinolines providing greater than 20:1 selectivity. These results suggest a mechanism wherein coordination of the metal to the C6 hydroxyl group of fumagillol with different sized metals greatly affects the regiochemical outcome, either through multidentate ligand effects and/or conformational control.</p><p>The observed regioselectivity can be rationalized from a model of metal-coordinated amino alcohol intermediate 14, obtained from opening of the more labile spirocyclic epoxide (Figure 2c).37,38 The larger La(III) catalyst may more easily accommodate the C6 hydroxyl group simultaneously with C1′ epoxide activation (cf. 15), thereby leading to tridentate coordination to the substrate wherein the amine is positioned closer and at a more optimal trajectory for addition to C1′. In contrast, smaller metals such as Zn(II) which cannot as easily accommodate the C6 hydroxyl while activating the second epoxide may adopt a looser bidentate coordination which places C2′ closer the amine (cf. 16). In the case of Zn(II) catalysis, activation of the second epoxide through adoption of 16 may be rate determining, as perhydroisoquinoline formation appeared to be largely independent of aniline nucleophilicity. Further studies to understand the precise mechanism leading to selectivity are currently underway.</p><p>The methodology was further extended by the use of L- and D-phenylalanine methyl esters which underwent lactonization after initial epoxide opening (Figure 3a). With Mg(OTf)2 as catalyst, perhydroisoquinoline products 17 and 19, respectively, were produced in approximately 3:1 regioselectivity relative to the perhydroisoindole- derived products (18, and 20/21, respectively) for each amino acid. Further lactonization of the perhydroisoquinoline analogues was not observed. In comparison, the perhydroisoindole formed with D-phenylalanine (catalyzed by La(OTf)3) lactonized in situ to afford the polycyclic morpholinone derivative 18. Reaction with L-phenylalanine, however, formed isoindole 20 and morpholinone 21 in a 4:1 ratio. The observed resistance to lactonization of 21 can be rationalized from steric congestion caused by the additional pseudoaxial prenyl substituent at C2′ which was calculated to be 3.1 kcal/mol higher in energy relative to diastereomer 18 (Supplementary Figure 7). Lactonization of 20 could eventually be accomplished under basic conditions to yield morpholinone 21 (85%).</p><p>Reaction of 1 with 2-ethynylaniline under Mg(OTf)2 catalysis produced perhydroisoquinoline 22, while La(III) afforded the expected perhydroisoindole product 23 (Figure 4a). Upon extended reaction times (48 h) with La(OTf)3 or at elevated temperature (90 °C), the novel 4,1-benzoxazepine39–41 24 bearing a [4.2.1] ring system was formed from 23 in a highly efficient cascade process.42–45 Benzoxazepine 24 is presumably formed by initial hydroalkoxylation of the alkynyl alcohol of 23 to enol ether 2546,47 followed by protonation to oxonium 26 (Figure 4b). Subsequent Prins cyclization48 forms 4,1-benzoxazepine 24, thereby providing a dramatic example of natural product remodeling via an unanticipated cascade sequence.</p><p>In summary, the natural product fumagillol has been selectively remodeled into a series of perhydroisoindoles and perhydroisoquinolines through sequential ring-opening with amines. Regiocontrol was achieved through choice of metal triflate catalysts, with smaller Zn(II) and Mg(II) catalysts leading to perhydroisoquinolines, while the larger La(III) catalyst favored production of perhydroisoindoles. Addition of secondary amines provided highly substituted tetrahydrofurans. Perhydrosoindole products underwent further reactions, including lactonizations employing amino acid esters as epoxide-opening nucleophiles and bridged 4,1-benzoxazepines from an unexpected cascade sequence with 2-ethynylaniline. Remodeled structures produced in this study are currently being examined in a range of biological screens, including those as part of the Molecular Libraries Probe Production Centers Network (MLPCN, http://mli.nih.gov/mli/) and the NIMH Psychoactive Drug Screening Program (PDSP, http://pdsp.med.unc.edu/indexR.html). These studies should pave the way for work to remodel other natural product scaffolds to access novel chemotypes and pharmacological tools.</p>

PubMed Author Manuscript

Metal-free aqueous redox capacitor via proton rocking-chair system in an organic-based couple

Safe and inexpensive energy storage devices with long cycle lifetimes and high power and energy densities are mandatory for the development of electrical power grids that connect with renewable energy sources. In this study, we demonstrated metal-free aqueous redox capacitors using couples comprising low-molecular-weight organic compounds. In addition to the electric double layer formation, proton insertion/extraction reactions between a couple consisting of inexpensive quinones/hydroquinones contributed to the energy storage. This energy storage mechanism, in which protons are shuttled back and forth between two electrodes upon charge and discharge, can be regarded as a proton rocking-chair system. The fabricated capacitor showed a large capacity (.20 Wh/kg), even in the applied potential range between 0-1 V, and high power capability (.5 A/g). The support of the organic compounds in nanoporous carbon facilitated the efficient use of the organic compounds with a lifetime of thousands of cycles. Electrical energy storage systems have numerous applications, including as portable devices, transport vehicles, and stationary energy resources. In particular, interest in electrical energy storage for ''smart grids'' has grown significantly. Smart grids are promising energy distribution systems that enable a balance between power application and power generation, including renewable energy sources. Renewable energy sources, e.g., solar and wind power, inherently exhibit large and rapid variability, with short-term intermittent spikes and drops. Therefore, for the power grid, a large electrical energy storage system with a high power capability must be installed to smooth the variability 1-3 .Another important factor in a large stationary application of an electrical energy storage system is cost, which can be subdivided into installation, operation, and maintenance components. The preferred energy storage devices should be composed of inexpensive, easily acquired materials and fabricated through a relatively simple manufacturing process. High durability and reliability coupled with long lifetime and safety are also desirable. Although battery technologies (e.g., lead-acid, sodium-sulfur, redox-flow, and lithium ion batteries) with high energy densities have attracted much attention for grid-scale energy storage, no existing batteries satisfy the above-mentioned requirements. For example, lithium ion batteries, which are the focus of intensive interest as transportation power sources, have unresolved safety issues due to their flammable organic electrolytes. Moreover, battery technologies generally have problems related to rapid charge/discharge.Capacitor technology can be recognized as the most promising solution for electrical energy storage systems with rapid response [4][5][6][7][8][9] . A general capacitor can store electrical charge in an electrical double layer (EDL) at the electrode-electrolyte interface. The advantages of capacitors are:. High power capability . Excellent lifetimes, exceeding 1000 charge/discharge cycles . Maintenance-free operation for ,10 years ''Aqueous capacitors,'' which work in aqueous electrolytes, have additional advantages:. High safety due to non-flammable aqueous electrolyte . Low manufacturing cost (fabrication under ambient conditions without requiring an oxygen-or humidityfree environment)To maximize the interface area in the electrode for the enhancement of energy density, high-surface-area electrode materials, e.g., activated carbon, are employed. However, the energy density of aqueous capacitors based

metal-free_aqueous_redox_capacitor_via_proton_rocking-chair_system_in_an_organic-based_couple

<!>Results<!>Discussion<!>Methods

<p>only on EDL formation is not high (,5 Wh/kg) due to a limiting operating voltage of about 1.23 V at which water decomposes. Under the developed technologies, their combined use with battery systems having high energy densities is necessary for energy storage systems in the power grid. If the energy density of the capacitor exceeds at least that of lead acid batteries (20-40 Wh/kg), the current combined energy storage system will be replaced by a capacitor-only device.</p><p>To enhance the energy density of energy storage devices, electrodes containing active redox-reactive materials, e.g., metal oxides, have been investigated. However, large-scale installations using current redox materials consisting of rare metals, e.g., LiCoO 2 in lithium ion batteries 10,11 or RuO 2 and MnO 2 in electrochemical capacitors 5,9,12,13 , would leave large ecological footprints and face resource restrictions. Against this background, inexpensive organic compounds have recently attracted much attention as redox-active materials for secondary batteries [14][15][16] and electrochemical capacitors [17][18][19][20] . Unlike the redox-active materials based on metal oxides, organic compounds are free from resource restrictions and potentially environmentally benign. Moreover, many redox-active organic materials possess large ion storage capacities beyond those of conventional redox-active metal oxides, primarily owing to the multielectron reactions in a low-molecular-weight moiety.</p><p>Among the redox-active organic materials, the ketone group is known to show fast redox reactions with protons in aqueous capacitors 21 . Quinones can facilitate two-proton reactions in a molecule. Reports have shown that both the capacitance and energy density were enhanced by the redox reactions of quinones loaded in activated carbon via impregnation 17,20 , grafting 18,19 , or dissolution in the electrolyte 22,23 . However, most such redox capacitors consisted of at least one unmodified or unloaded carbon electrode, and redox capacitors configured from two different electrodes employing a couple based on two organic compounds having different reaction potentials have rarely been investigated, despite their promise for further increases in energy density.</p><p>In this work, we present a metal-free redox capacitor using a couple comprising low-molecular-weight organic compounds in an aqueous electrolyte. In the capacitor, shown schematically in Fig. 1, each electrode consisted of a different organic compound impregnated in nanoporous activated carbon. The quinone/hydroquinone couple employed consisted of anthraquinone (AQ) and tetrachlorohydroquinone (TCHQ) as the active materials with different redox potentials for proton insertion/extraction reactions. Protons in the aqueous electrolyte played the roles of charge carriers. This energy storage mechanism, in which protons are shuttled back and forth between two electrodes upon charge and discharge, can be regarded as a proton rocking-chair system. The nanoporous carbon facilitated the high utilization rate of the redox reactions of the organic compounds simply impregnated in the carbon at high loading. The manufactured capacitor exhibited promising performance in terms of high energy density, high power density, and long cycle lifetime.</p><!><p>We conducted the galvanostatic charge/discharge cycling of the redox capacitor fabricated with an AQ and TCHQ couple in a three-electrode configuration. The electrodes consisted of the organic compound (27 wt%), the nanoporous carbon (63 wt%), and polytetrafluoroethylene (PTFE, 10 wt%). PTFE was employed as a binder to prepare a paste of the organic compound-carbon composite. Both electrode weights were identical. The positive and negative electrodes contained TCHQ and AQ impregnated in the nanoporous carbon, respectively. During the charge process, AQ acted as a proton acceptor and was transformed into anthrahydroquinone, whereas the proton-donor TCHQ was transformed to tetrachlorobenzoquinone. During the discharge process, the roles were reversed, and both materials were restored to their initial states.</p><p>Figure 2 shows the typical galvanostatic cycle and the evolution of the potential corresponding to the positive and negative electrodes in 0.5 M H 2 SO 4 aqueous solution. The plateau potentials of the positive and negative electrodes corresponded to the redox peaks of TCHQ (0.50 V vs. Ag/AgCl) and AQ (20.16 V vs. Ag/AgCl) observed in the cyclic voltammograms (see Supplementary Fig. S1 online), and the plateau potential in the galvanostatic cycle (at ,0.65 V) corresponded to their difference. These behaviors indicated that the proton-extraction/insertion reaction of TCHQ occurred at the positive electrode simultaneously with the proton-insertion/extraction reaction of AQ at the negative electrode, and the energy storage of this capacitor relies on a proton rocking-chair mechanism. The rechargeable energy density of this redox capacitor based on the total weight of two electrodes was approximately 20.3 Wh/kg at a current density of 0.28 A/g. In contrast, the rechargeable energy of an EDL capacitor, which had a symmetric configuration and consisted of nanoporous carbon without the organic compounds, was approximately 6.7 Wh/ kg at the similar current density (0.34 A/g) (see Supplementary Fig. S2 online). Thus, the energy density of the capacitor increased threefold via the impregnation of the organic compounds.</p><p>The calculated energy density of the redox capacitor includes both the contribution from the EDL capacity in addition to that from the redox reactions of the organic compounds. To focus on the capacity of electrical charge derived only from the organic compounds, we estimated the EDL capacity by extrapolating the galvanostatic discharge profile in the range from 0 to 0.4 V, where the redox reactions did not take place. The near-linear profile in this range should be derived from the EDL of the nanoporous carbon support. The capacities of electrical charge derived from the EDL and the redox reactions can be estimated to be 34.8 mAh/g -carbon based on the carbon weight in single electrode and 186 mAh/g -(AQ or TCHQ) based on the organic compound weight in single electrode, respectively. The calculation detail is in the online Supplementary Information (Fig. S3). The capacity derived from the EDL in the redox capacitor was nearly the same as that observed for the pure EDL capacitor (32.6 mAh/ g -carbon ). The ideal proton storage capacities of AQ and TCHQ with two-proton redox reactions were 257 mAh/g -AQ and 216 mAh/ g -TCHQ , respectively. Because the weights of the organic compounds were identical, the capacity derived from the redox reactions should be limited by the TCHQ capacity. Therefore, the utilization rate of the organic compounds in our redox capacitor can be regarded to be approximately 86% based on the weight of TCHQ.</p><p>We examined the rapid response capability of this redox capacitor. As mentioned, a high power capability is essential to level the variability from renewable energy sources. Aqueous capacitors are potentially advantageous for rapid charge/discharge owing to the high ionic conductivity of the aqueous electrolyte. Figure 3a shows the galvanostatic cycles at 0.11 and 5.6 A/g. Both displayed similar profiles, consisting of plateau or plateau-like regions derived from the redox reactions of the organic compounds, and linear regions derived from EDL formation in the charge and discharge conditions. These profiles mean the proton insertion/extraction reaction also occurs even at high current density, and the proton rocking-chair mechanism using the organic-based couple is effective for rapid response. It should be noted that the plateau potentials at the charge and discharge conditions, which were nearly equal at low current density, moved apart from each other as the current density increased. This separation (polarization) is considered to be derived both from the electrolyte resistance and the overpotential of the redox reaction of the organic compounds, and deteriorates the rechargeable energy density of our redox capacitor. Figure 3b shows the rechargeable energy density of the redox capacitor and a pure EDL capacitor consisting of the nanoporous carbon at various current densities. As the current density increased, the rechargeable energy density decreased (21.8 Wh/kg at 0.11 A/g and 7.15 Wh/kg at 5.6 A/g). However, the capability of the redox capacitor for high current density was similar to that of the pure EDL capacitor. The fast redox reactions of quinones should contribute to this outstanding capability, which is comparable to that of EDL capacitors.</p><!><p>Support of the organic compounds in the nanoporous carbon should play a crucial role for the effective use of the capacity of organic compounds. Because of the low-or non-conductivity of AQ and TCHQ, their crystalline bulk could not be utilized for electrochemical redox reactions without a carbon support. However, as indicated above, the organic compounds held in nanoporous carbon contributed to the energy storage. X-Ray diffraction (XRD) and 1 H-NMR studies revealed that the most of the organic compounds were held on the surface of the nanoporous carbon with less-crystalline or nanocrystalline structures (see Supplementary Figs. S4-S7 online). Because the proportion of the organic molecules contacting the carbon surface to the total of the loaded organic molecules increases by downsizing the supporting pores, the interaction between the sp 2 carbon surfaces and the aromatic rings will become apparent and stabilize the absorbed organic compounds in the nanometer-scaled pores. The organic compounds held in carbon materials with lesscrystalline or nanocrystalline structures showed high utilization rate (see Supplementary Figs. S8 and S9 online). It is highly anticipated that such adsorption states of the organic compounds on a carbon material with electrical conductivity contribute to the redox reactions with a high utilization rate.</p><p>We also evaluated the dependence of the cycle lifetime on the kinds of carbon materials for a single electrode system (see Supplementary Fig. S8). In the measurements, AQ impregnated in the carbon materials was employed in the working electrode, where the excess amount of carbon material was employed in the counter electrode. For a carbon material with a larger averaged pore size (w11 nm), the retention rate of capacitance for the working electrode after 1000 charge/discharge cycles at 0.56 A/g was 41%. For the nanoporous carbon material (averaged pore size: w2.2 nm) employed in this study, the retention rate after 1000 cycles was 82%. From this comparison, we found that the support of organic compounds in the nanoporous carbon also contributed to the cycle lifetime enhancement of the redox capacitor.</p><p>Durability becomes a critical issue when using organic compounds as the redox-active materials, because of their solubility. Generally, as the molecular weight decreases, the solubility of quinones in aqueous solution generally increases. For example, the solubilities of benzoquinone and AQ in water at 25uC are approximately 10 g/l and 1.4 mg/l, respectively. Moreover, hydroquinones, which forms by proton insertion reaction with quinones, are more soluble than quinones (the solubility of hydroquinone in water at 25uC are 80 g/l). Therefore, the dissolution of organic compounds during charge/discharge degrades the cycle performance of the aqueous capacitor that uses such compounds. It was supposed that the support by carbon in the nanometer-scaled pores would be effective against dissolution. It is known that the nanoporous carbon employed in this study has a complicated branching pore structure 24 . Support inside this structure would prevent organic compounds from dissolving out of the carbon matrix, and enable their reversible redox reactions with long cycle lifetimes. Figure 4a shows the rechargeable energy density of the redox capacitor device as a function of cycle time. The coulombic efficiency was above 99% for almost all charge-discharge cycle. At a current density of 0.28 A/g, the energy density exhibited an initial growth until about the 100 th cycle, and then gradually decreased. At present, the delay time that it takes for the redox capacitor to show its maximal energy density has not been fully controlled, and remains a challenge. After 1000 cycles, the retention rate based on the maximum energy density was approximately 70%.</p><p>Finally, we suggest a strategy for further enhancement of cycle lifetime. Figure 4b shows the evolution of the potential corresponding to the positive and negative electrodes. The final potential after the discharge process completed at the 1000 th cycle, 0.38 V vs. Ag/ AgCl, was much higher than that at the 100 th cycle, 0.25 V vs. Ag/ AgCl. This increase in the final potential (0.13 V) implies that the decrease in the rechargeable energy density after many cycles results from the degradation of the negative electrode, because the evolution of the potential for the negative electrode caused by its shortened plateau region interrupts the full-capacity charge/discharge of the positive electrode. It was conjectured that the degradation of the negative electrode could be attributed to the dissolution of AQ into the aqueous electrolyte as anthrahydroquinone. Therefore, instead of AQ, we employed 1,5-dichloroanthraquinone (DCAQ) as the redoxactive material in the negative electrode. The attachment of hydrophobic functional groups was effective in minimizing dissolution of the organic compounds into the aqueous solution. For example, the solubility of TCHQ, which has 4 hydrophobic chloro groups, in water at 25uC is quite low, 76 mg/l, compared with that of hydroquinone (80 g/l).</p><p>Figure 5a shows the rechargeable energy density of the redox capacitor using the DCAQ and TCHQ couple as a function of cycle time at 0.26 A/g. The coulombic efficiency was above 99% for almost all charge-discharge cycle. Similar to the redox capacitor that used the AQ and TCHQ couple, the energy density reached its maximum around at the 100 th cycle. On the other hand, compared with the case of AQ/TCHQ, the cycle performance was enhanced by the attachment of the hydrophobic chloro groups, and the degradation of the rechargeable energy density was not observable after 1000 cycles. In addition, the cycle performance of the redox capacitor (TCHQ-DCAQ) at 2.6 A/g, the degradation of the rechargeable energy density was not observable after 10000 cycles (see Supplementary Fig. S10). These results suggest that the attachment of hydrophobic functional groups was effective as a strategy for the further enhancement of cycle lifetime. Figure 5b shows a typical galvanostatic cycle and the evolution of the potential corresponding to the positive and negative electrodes in the redox capacitor. The potentials of the redox reactions with protons were 20.05 V (vs. Ag/AgCl) for DCAQ and 0.50 V (vs. Ag/AgCl) for TCHQ. Therefore, the potential plateaus around 0.55 V corresponded to the difference of the redox reaction potential between DCAQ and TCHQ, and it was confirmed that the energy storage of this capacitor relies on the redox reactions of DCAQ and TCHQ. Compared with the AQ and TCHQ couple, the decrease in the energy density resulted from the decrease in the potential difference. However, in this case, the decrease in the final potential after the discharge process was not observed. This indicated that the suppressed solubility of DCAQ by the attachment of hydrophobic functional group contributed to avoid the degradation of the negative electrode and the interruption of the full-capacity charge/ discharge of the positive electrode. In summary, we demonstrated an aqueous redox capacitor using a couple comprising organic compounds that enabled a proton rocking-chair-type energy storage. Fundamentally unlike conventional batteries and supercapacitors, the employed electrodes, which consisted only of light elements (hydrogen, carbon, oxygen, and chlorine), can be installed with low material and manufacturing costs. By employing nanoporous carbon, a redox capacitor with a long cycle lifetime and a high energy density, derived from the high utilization rate of the redox reaction of the organic compounds, was achieved. Moreover, this capacitor can respond to rapid variability. These features are promising for large stationary applications in electrical energy storage systems for the energy grid.</p><p>Furthermore, we proposed a strategy to further enhance cycle lifetime by controlling quinone/hydroquinone solubility. In addition to hydrophobicity, the required properties of the organic compounds in the capacitor were low molecular weights, ability to participate in fast redox reactions for high energy density, and high power capability. The coupling of different organic compounds having different redox potentials in the limiting operating voltage of an aqueous solution was also important. Further improvements via the choice of the best organic-based couple and the optimized loading of the organic compounds on carbon will accelerate the replacement of current electrical energy storage systems for grid use.</p><!><p>Preparation of organic compounds supported by carbon materials. Nanoporous activated carbon (MaxsorbH MSC-30, Kansai Coke and Chemicals Co., Ltd.) was employed as the support for the organic compounds. N 2 adsorption/desorption behaviors obtained by an automatic adsorption apparatus (BELSORP-18, BEL Japan, Inc.) revealed that the nanoporous activated carbon had a BET surface area of 3070 m 2 /g and a pore volume of 1.70 cm 3 /g. In addition, BJH analysis indicated that the number of pores larger than 10 nm was negligible.</p><p>To prepare the organic compound-carbon composites, the as-received organic compound (AQ, TCHQ, or DCAQ (Tokyo Chemical Industry Co., Ltd.)) was dissolved or dispersed in acetone with the carbon material by sonication. A detailed discussion of the selection of these compounds is included in the online Supplementary Information (Table S1 and Figs. S11 and S12). The weight ratio of the carbon material to the organic material was fixed at 753. By evaporating the acetone at 70uC, the organic compound was supported in the carbon material.</p><p>Electrochemical measurement setup. The electrodes were prepared in the form of pressed pellets (w7 mm) using a mixture of organic compound-carbon composite and polytetrafluoroethylene (PTFE) as a binder. The weight ratio of the composite to PTFE was fixed at 951, and the total weight of an electrode is fixed to be approximately 4 mg. The electrodes were attached onto Au mesh current collectors. Aqueous H 2 SO 4 solution (0.5 M) and a Ag/AgCl electrode were employed as the electrolyte and the reference electrode, respectively. Electrodes are inserted into 70 ml electrolyte in glass tube. After degassing the electrodes and electrolyte under vacuum until the open circuit potential (OCV) became stable, we initiated measurements using a potentiostat/galvanostat (VMP3, Bio-Logic) under ambient conditions. The applied potential range between the positive and negative electrodes was 0-1.0 V. To remove the oxygen dissolved in the aqueous electrolyte, N 2 gas was bubbled through the electrolyte throughout the charge-discharge measurements. The current density calculations were based on one electrode weight. The energy density calculations were based on the total weight of both electrodes.</p>

Scientific Reports - Nature

Discovery and characterization of 2-(cyclopropanesulfonamido)-N-(2-ethoxyphenyl)benzamide, ML382: a potent and selective positive allosteric modulator of MrgX1

Previous studies have shown that activation of mouse MrgC11, a G-protein coupled receptor, by its peptide ligand BAM8-22 can inhibit chronic pain. A large scale screen has been carried out to isolate small molecule allosteric agonists of MrgX1, the human homologue of MrgC11. The goal of this study is to improve the efficacy and potency of the positive allosteric modulators with therapeutic implications of anti-chronic pain. Here, we report an iterative parallel synthesis effort and a structure-activity relationship of a series of arylsulfonamides, which led to the discovery of the first positive allosteric modulator (PAM) of MrgX1, ML382.

discovery_and_characterization_of_2-(cyclopropanesulfonamido)-n-(2-ethoxyphenyl)benzamide,_ml382:_a_

<!>Experimental Section

<p>Chronic pain is difficult to treat and has become a major health and economic burden worldwide reaching nearly epidemic levels with 20–25% of the population affected. Chronic pain is often refractory to current therapies and the many of the major analgesics (e.g., opioids) carry with them dose-limiting adverse events and serious risk of addiction and abuse. These risks of addiction and abuse present substantial barriers to their clinical use and have become a serious health threat. Thus the discovery of novel therapies to treat pain without the addiction potential is a major goal.</p><p>Mrgs (also named Mrgprs/SNSRs) are a family of G protein-coupled receptors (GPCRs) consisting of more than 50 members in the mouse genome.[1] Strikingly, the expression of many Mrgs, including MrgC11, is restricted to subsets of small-diameter sensory neurons in DRG and trigeminal ganglia as these genes have not been detected in the central nervous system or in the rest of the body.[2] Similarly, some of human MrgprXs such as MrgX1 are also selectively expressed in DRG neurons.[1b] BAM8-22, a 15 amino acid endogenous peptide, is a specific agonist for mouse MrgprC11 and human MrgX1.[1a, 2] Based on the expression pattern, ligand specificity, and similarity in signal pathways, MrgprC11 and MrgprX1 are functional orthologues.</p><p>Several complementary lines of evidence suggest that MrgprC11 functions as an endogenous regulator of persistent pain.[3] Particularly, spinal cord application of BAM8-22 and other MrgC11 agonists in mice significantly attenuates inflammatory hyperalgesia and neuropathic pain in an Mrgpr-dependent manner.[3a, 4] It has been shown recently that the activation of MrgC11 leads to inhibition high-voltage activating Ca2+ channels which play essential role in pain signal transmission.[5] Because of the specific expression, agonists and PAMs for MrgX1 may represent a new class of anti-hyperalgesics for chronic pain without side effects in the central nervous system.[6],[7]</p><p>The project commenced with a screen of the >300000 NIH Molecular Library Small Molecule Repository (MLSMR) compound collection using a triple addition protocol using BAM-22 as the as the ligand (AID: 588675) (Figure 1).[8] Using a HEK293 cell line that stably expresses the MrgX1 protein, the first addition is the drug itself (agonist mode), the second addition is an EC10 to EC30 of BAM-22 for identification of positive allosteric modulators, and the third addition uses an EC90 to Emax of BAM-22 for identification of antagonists/negative allosteric modulators. From the initial screen, ~1900 compound were identified as hits and after a medicinal chemistry triage, ~1100 compound were retested and then counter screened against HEK293 parental cells. There were ~150 compounds that showed no activity against the parental cells and these were then evaluated in a 5-point concentration response curve (CRC) leaving 29 active compounds. These compounds were then tested in 10-point CRC format revealing 19 confirmed MrgX1 positive allosteric modulators.</p><p>The results of the HTS screening campaign are summarized in Table 1. From the ~39 identified "hits", there were four compounds from a sulfonamide benzamide series. Combining the activity in the MrgX1 assay and excellent selectivity profile as seen in the PubChem Assay and PubChem Hit rate[9], this series was chosen for further development towards a novel MrgX1 PAM. The initial SAR revealed a couple structural variants worth noting. First, the 2-ester analog was significantly less potent (3.85 μM), and second, the 2-methyl thioether was more potent than the 2-methoxy (~2-fold). The SAR chemical optimization strategy is highlighted in Table 1 wherein we evaluated multiple sites of the ligand in parallel through iterative synthesis.</p><p>The synthesis of the MrgX1 compounds (generically shown as 4) is outlined in Scheme 1. The substituted anthranilic acid, 1, was converted to the acid chloride (SOCl2, DMF) and then coupled with aniline, 2, (pyridine, Et2O) to yield the amide, 3. The sulfonamide compounds, 4, were made by coupling, 3, with an appropriate sulfonyl chloride (RSO2Cl, pyridine, DCM). This synthesis represented a modular approach to the final compounds allowing for easy modification of R, R1, R2 and R3, which allowed for the synthesis of numerous analogs for this project.</p><p>The SAR evaluation started with the re-synthesis and re-confirmation of the four initial hits that were identified in the HTS campaign (Table 2). These compounds, 4a, 4b, 4g, and 4h were all re-confirmed within ~2–5-fold of the initial HTS values (EC50), except for 4h which was inactive (AID: 743016). Those compounds with an Emax>50% were considered "active" and were progressed to an EC50 determination. From the SAR in Table 2, it was shown that ester groups on the R2 phenyl group were not tolerated (4d and 4h). In addition, phenyl sulphonamides were not tolerated (4j and 4k). Simple methyl sulphonamides coupled with either ether or thioether substituents on the R2 phenyl were well tolerated with many of the compounds showing nanomolar EC50's (e.g., 4a, EC50 = 531 nM; 4c, EC50 = 419 nM; 4g, EC50 = 500 nM and 4i, EC50 = 551 nM). The 2-methoxy is less active than the 2-ethoxy or 2-methylthio groups. Also, substitution of the sulphonamide nitrogen was potency neutral compared to the unsubstituted NH sulfonamide (e.g., 4a vs 4i).</p><p>The final round of SAR utilized the 2-ethoxyphenyl group as the right-hand group and evaluated a variety of small, alkyl sulfonamides as these were identified from the previous round of SAR as the optimal substituents (Table 3). This round of SAR was far more productive than the previous round and showed a distinct SAR trend. Cyclopropyl sulfonamide, 4l was the most potent (EC50 = 190 nM; Figure 2) and efficacious (Emax = 195%). For this round of SAR, the Emax values were normalized to the positive control (4i, 100%). There were several compounds that showed >100% Emax. Another sub-micromolar compound was the 2-chloroethyl sulfonamide, 4m; however, as this compound possessed the reactive alkyl halo group, it was not progressed further. The isopropyl sulfonamide, 4o (EC50 = 1.82 μM; Emax = 255%) and the cyclopropyl methyl, 4p (EC50 = 4.47 μM, 267%) were significantly less potent than 4l. Based on the superior potency and efficacy, 4l has been declared the MLPCN probe compound for the MrgX1 allosteric agonist project, ML382.</p><p>The dose responses studies show that ML382 enhances the potency of BAM8-22 on MrgX1 for >7-fold (i.e. EC50 of BAM8-22 from 18.7nM to 2.9nM; Supplemental Figure 1). However, ML382 does not affect the EMax of BAM8-22 when the maximum concentration of BAM8-22 was used (Supplemental Figure 1). Furthermore, in the absence of BAM8-22, ML382 does not activate MrgX1 suggesting ML382 itself does not has agonistic activity on MrgX1 (Figure 2). Pharmacological and selectivity profiling of ML382 were performed against the closely related MrgX2 (Supplemental Figure 2), and against a larger panel of receptors. ML382 was evaluated in HEK293 cells expressing MrgX2 as a selectivity screen. As can be seen in Supplemental Figure 2, ML382 showed no significant effect on activation of MrgX2 in the presence of the specific agonist peptide, PAMP (both EC50 and EMax). Also, ML382 had no effect on MrgprC11, the mouse homolog of MrgprX1 which indicates the compound is species selective (results not shown). In addition, ML382 was evaluated using EuroFin's Lead Profiling Screen which is a binding assay panel of 68 GPCR's, ion channels and transporters screened at 10 μM.[10] ML382 did not inhibit 67 of the 68 targets assayed (inhibition of radio ligand binding >50% at 10 μM); with the only target that was considered active was the serotonin (5-hydroxytryptamine), 5HT2B which showed 63% inhibition at 10 μM. Overall, ML382 displayed a very favourable selectivity profile.</p><p>Lastly, ML382 was further profiled in a battery of Tier 1 in vitro DMPK assays (Table 4). The intrinsic clearance was assessed in hepaticmicrosomes (rat/human) and ML382 was shown to be unstable to oxidative metabolism and predicted to display high clearance in both species (65.7 and 15.7 mL/min/kg, respectively). In addition, using an equilibrium dialysis approach, the protein binding of ML382 was evaluated and it was shown to have moderate free fraction in rat and low free fraction in human (1.7% in rat and 0.4% in human).</p><p>In summary, we have developed 4l (also known as ML382 or VU0485891) as a potent and selective positive allosteric modulator (PAM) of MrgX1. This is the first report of a PAM for MrgX1. In the presence of the known ligand, BAM8-22, ML382 has an EC50 = 190 nM with an Emax = 148% and is inactive against the closely related MrgX2, and against a panel of selected receptors from the EuroFins panel. In addition, ML382 was profiled in our Tier 1 DMPK assay and possesses favourable free fraction in rat plasma protein binding; however, instability in rat liver microsomes indicates the utility of ML382 as an in vivo probe will be limited to non-oral dosing regimens (e.g., intraperitoneal (IP), subcutaneous (SC), or intrathecal). Further use of ML382 in an in vivo animal study will be reported in due course.</p><!><p>Experimental procedures for the medicinal chemistry, pharmacology and drug metabolism studies, as well as compound characterization data are provided in the Supporting Information, available at http://dx.doi.org/XXXX.</p>

PubMed Author Manuscript

Manduca sexta Moricin Promoter Elements can Increase Promoter Activities of Drosophila melanogaster Antimicrobial Peptide Genes

Insects produce a variety of antimicrobial peptides (AMPs). Induction of insect AMP genes is regulated by the Toll and IMD (immune deficiency) pathways via NF-\xce\xbaB and GATA factors. Little is known about species-specific regulation of AMP genes. In this report, we showed that activities of most Manduca sexta and Drosophila melanogaster AMP gene promoters were regulated in a species-specific manner in Drosophila (Dipteran) S2 cells and Spodoptera frugiperda (Lepidopteran) Sf9 cells. A \xce\xbaB-GATA element (22bp) from M. sexta moricin (MsMoricin) promoter could significantly increase activities of Drosophila AMP gene promoters in S2 cells, and an MsMoricin promoter activating element (MPAE) (140bp) could increase activity of drosomycin promoter specifically in Sf9 cells. However, \xce\xbaB and GATA factors alone were not sufficient for MsMoricin gene activation, suggesting that other co-regulators may be required to fully activate AMP genes. Our results suggest that induction of insect AMP genes may require a transcription complex composed of common nuclear factors (such as NF-\xce\xbaB and GATA factors) and species-related co-regulators, and it is the co-regulators that may confer species-specific regulation of AMP genes. In addition, we showed that activity of Drosophila drosomycin promoter could be activated cooperatively by the inserted exogenous \xce\xbaB-GATA element and the endogenous \xce\xbaB element. These findings revealed an approach of engineering AMP genes with enhanced activities, which may lead to broad applications.

manduca_sexta_moricin_promoter_elements_can_increase_promoter_activities_of_drosophila_melanogaster_

1. Introduction<!>2.1 Insects, bacterial peptidoglycan (PG) and insect cell lines<!>2.2 Genomic DNA extraction and genome walking<!>2.3 RNA extraction and 5\xe2\x80\xb2 RACE<!>2.4 Sequence analysis<!>2.5 Construction of luciferase reporter plasmids<!>2.6 Insect cell culture and transfection<!>2.7 Data analysis<!>3.1 Species-specific regulation of AMP gene promoters in S2 and Sf9 cells<!>3.2 Identification of an active \xce\xbaB-GATA element in MsMoricin promoter<!>3.3 Identification of an activating element upstream of the MsMoricin \xce\xbaB-GATA element<!>3.4 The \xce\xbaB-GATA element of MsMoricin can enhance promoter activities of AMP genes<!>3.5 The \xce\xbaB-GATA element and the \xce\xbaB site2 cooperatively increase drosomycin promoter activity<!>3.6 MPAE may contain lepidoptera-related co-regulator binding sites<!>4.1 Species-specific regulation of AMP genes<!>4.2 The \xce\xbaB, GATA and MPAE elements cooperatively induce gene transcription<!>4.3 The \xce\xbaB-GATA and MPAE elements activate D. melanogaster AMP gene promoters in S2 and SF9 cells differently<!>4.4 Effect of the position, direction and consensus sequence of \xce\xbaB and GATA sites on AMP gene promoter activity<!>4.5 Potential applications of our finding

<p>The innate immune system is conserved from insects to humans (Ganesan et al., 2011; Mogensen, 2009). Insect innate immune system relies on various germ line encoded pattern recognition receptors (PRRs) to sense pathogen-associated molecular patterns (PAMPs) and induce cellular and humoral responses (Charroux et al., 2009; Charroux and Royet, 2010; Kanost et al., 2004; Lemaitre and Hoffmann, 2007; Marmaras and Lampropoulou, 2009). An important aspect of insect humoral responses is induced production of a variety of antimicrobial peptides (AMPs) (Diamond et al., 2009). Most AMPs are small cationic peptides with activities against microorganisms and parasites (Brogden, 2005; Imler and Bulet, 2005; Kokoza et al., 2010). In Drosophila melanogaster, induced production of AMPs is regulated by the Toll and IMD (immune deficiency) pathways (Lemaitre et al., 1995; 1996; DeGregorio et al., 2002). The Toll pathway mediates recognition of fungi and Gram-positive bacteria (Ashok, 2009), while the IMD pathway senses infection by most Gram-negative bacteria (Kaneko and Silverman, 2005). The Toll pathway triggers activation of NF-κB factors Dorsal and Dif, while the IMD pathway leads to activation of Relish (Engström et al., 1993; Ganesan et al., 2011; Gross et al., 1996; Hetru and Hoffmann, 2009; Ip et al., 1993; Stöven et al., 2003).</p><p>GATA factors are a family of zinc finger containing transcription factors, which recognize the (A/T)GATA(A/G) consensus sequence, and they are involved in regulation of gene expression and differentiation (Patient and McGhee, 2002). GATA factors have been identified in vertebrates, D. melanogaster, Caenorhabditis elegans, and plants (Patient and McGhee, 2002; Reyes et al., 2004). Vertebrate GATA-1 interacts with many other regulatory partners, such as Friend of GATA (FOG), p300/CBP, PU.1 and c-myb (Lowry and Mackay, 2006). GATA factors interact with different factors to regulate gene transcription (Dai et al., 2002; Eisbacher et al., 2003; Gordon et al., 1997; Zhang et al., 2007). In D. melanogaster, Serpent (dGATAb) and dGATAe regulate induced gene expression in fat body and midgut, respectively (Petersen et al., 1999; Senger et al., 2006). GATA factors are also required for immunity in C. elegans and the silkworm Bombyx mori (Cheng et al., 2006; Kerry et al., 2006). Adjacent κB and GATA sites have been identified in many insect immune gene promoters and both sites are required for gene induction (Harshman and James, 1998; Kadalayil et al., 1997; Senger et al., 2004; Tingvall et al., 2001). Human GATA-3 and/or GATA-2 interact with NF-κB to trigger GlcNac6ST-1 transcription (Chen et al., 2008). In addition, κB and GATA sites are both required for induced expression of Drosophila cecropin A (Kadalayil et al., 1997). Therefore, κB-GATA synergy seems to be a common mechanism for immune gene regulation (Senger et al., 2004). However, little is known about synergistic effect of κB and GATA factors in insects.</p><p>In D. melanogaster, seven groups of AMPs have been identified, some AMP genes (drosomycin, diptericin, metchnikowin) have been identified only in Drosophila, others (attacin, cecropin, drosocin, defensin) are also found in other insect species (Imler and Bulet, 2005; Levashina et al., 1998). Drosomycin is an anti-fungal peptide isolated from immune challenged D. melanogaster (Fehlbaum et al., 1994; Tian et al., 2008). Expression of drosomycin is synergistically regulated by the Toll and IMD pathways (Tanji et al., 2007; Tanji et al., 2010). Diptericin is another species-related AMP gene first identified in Phormia terranovae and later in D. melanogaster (Dimarcq et al., 1988; Wicker et al., 1990). Different groups of AMP genes have also been identified in lepidopteran insects, such as B. mori and Manduca sexta (Kanost et al., 2004). Moricin, gloverin and lebocin genes have been identified only in lepidopteran insects (Axen et al., 1997; Chowdhury et al., 1995; Hara and Yamakawa, 1995; Kanost et al., 2004). Moricin was originally isolated from the hemolymph of B. mori and showed antibacterial activity against several Gram-negative and Gram-positive bacteria (Hara and Yamakawa, 1995, 1996). The N-terminal region of B. mori Moricin adopts an amphipathic alpha-helix structure that may increase permeability of the cytoplasmic membrane (Hemmi et al., 2002). Moricin analogues have been identified in other lepidopteran species, including M. sexta, Galleria mellonella, and Spodoptera litura (Brown et al., 2008; Oizumi et al., 2005; Zhu et al., 2003).</p><p>Our previous research reveals that lipopolysaccharide (LPS) and lipoteichoic acid (LTA) can induce AMP gene expression in M. sexta larvae (Rao and Yu, 2010). In D. melanogaster, peptidoglycan can activation AMP genes (Werner et al., 2000; 2003), but ultrapure LPS molecules do not induce AMP expression in adult flies (Kaneko et al., 2004), indicating that there may be important differences between dipteran and lepidopteran species regarding regulation of AMP genes. It is not known whether expression of AMP genes is regulated in a species-specific manner, and whether different co-regulators are involved in regulating AMP gene expression in lepidopteran and dipteran insects. In this study, we cloned promoters for M. sexta moricin (MsMoricin), cecropin and lysozyme genes and compared activities of the three M. sexta (Lepidopteran) and seven D. melanogaster (Dipteran) AMP gene promoters in D. melanogaster S2 cells and Spodoptera frugiperda (Lepidopteran) Sf9 cells. We found that most AMP gene promoters were regulated in a species-specific manner in the two cell lines in that D. melanogaster AMP gene promoters had no or low activity in Sf9 cells and M. sexta AMP gene promoters had no or low activity in S2 cells. We then showed that κB and GATA factors alone were not sufficient to activate MsMoricin promoter, and a κB-GATA element (22bp) from the MsMoricin promoter could significantly increase activities of D. melanogaster AMP gene promoters when inserted into the promoters. We also showed that the κB-GATA element and the endogenous κB site2 of drosomycin promoter were all required to cooperatively enhance drosomycin promoter activity. More importantly, we identified an activating element, designated as MsMoricin promoter activating element (MPAE) (140bp), which could increase activity of drosomycin promoter specifically in Sf9 cells, thus MPAE may contain co-regulator binding sites for nuclear factors specifically expressed in lepidopteran species. Our results suggest that common factors such as NF-κB and GATA factors are functional in both dipteran and lepidopteran insects, while co-regulators may confer species-specific regulation of AMP genes.</p><!><p>M. sexta eggs were kindly provided by Professor Michael Kanost, Department of Biochemistry at Kansas State University. Larvae were reared on an artificial diet at 25°C (Dunn and Drake, 1983), and the 5th instar larvae were used for hemocytes collection. Ultrapure peptidoglycan from E. coli strain K12 (Cat#: tlrl-pgnek) was purchased from InvivoGen (San Diego, California, USA) and used for activation experiments. Drosophila melanogaster S2 cells were purchased from American Type Culture Collection (ATCC). Spodoptera frugiperda Sf9 cells were purchased from Invitrogen Corporation, USA.</p><!><p>M. sexta genomic DNA was extracted from hemocytes collected from the 5th instar larvae with PureLink™ Genomic DNA Kit (Invitrogen, USA). D. melanogaster genomic DNA was extracted from S2 cells. Genome walking was performed to clone MsMoricin and MsCecropin promoters with GenomeWalker Universal Kit (Clontech, USA) following instructions of the manufacturer. Briefly, 2.5 μg M. sexta genomic DNA was digested with Dra I, EcoR V, Pvu II or Stu I, respectively. Digested fragments were purified and ligated to a synthetic adaptor GWAdaptor. Adaptor primers (GW-AP1 and GW-AP2) and gene specific primers (MsMoricinGSP1-4, MsCecropinGSP1 and 2) (Table S1) were used for PCR reactions.</p><!><p>M. sexta hemocytes were collected from the 5th instar larvae at 6 h after E. coli XL1-blue injection and total RNA was prepared from hemocytes with TRI reagent (Sigma Aldrich, USA). cDNA was prepared with ImProm-II reverse transcriptase (Promega, USA). 5′ RACE was performed to determine transcription start site of M. sexta moricin promoter with SMARTer™ RACE cDNA amplification kit (Clontech, USA).</p><!><p>Transcription factor binding sites were predicted with Alibaba2.1 (http://www.gene-regulation.com/). Other sequences were analyzed with DNAMAN (Lynnon Corporation, Quebec, Canada).</p><!><p>For luciferase reporter plasmids, promoters from antimicrobial peptide (AMP) genes of M. sexta and D. melanogaster were cloned by genome walking or PCR using genomic DNAs as templates. PCR was performed with Taq DNA polymerase using gene specific primers listed in Table S1. For MsMoricin, MsLysozyme and D. melanogaster AMP genes reporters, PCR products were digested and ligated to the Kpn I/Bgl II sites of pGL3Basic vector (Promega, USA). For MsCecropin reporter, PCR product was digested and ligated to the Xho I and Hind III sites of pGL3Basic vector. MsMoricin, MsCecropin, MsLysozyme, diptericin, DmAttacin A, DmDefensin, drosomycin, DmCecropin A1, drosocin and metchnikowin luciferase reporters contained 1456bp, 877bp, 1241bp, 980bp, 977bp, 1651bp, 812bp, 670bp, 660bp, and 1560bp of 5′ upstream sequences, respectively. In these luciferase reporters, +1 indicates the translation start site (ATG) in MsCecropin, while +1 indicates transcription initiation sites for MsMoricin, drosomycin, diptericin and DmAttacin A genes. Deletion and mutation reporters were constructed by overlapping PCR. The first round of overlapping PCR was performed to amplify the 5′ and 3′ end DNA fragments individually with overlapping regions, and the second round of overlapping was done by mixing fragments amplified from the first round PCR as templates with the 5′ and 3′ primers. MsMoricin κB5 (GTAAAGTCCC) was mutated to TTAGAGTTAT, and GATA-1 (TCGTTATCTG) was mutated to TCGCGTATCG. Drosomycin κB site-1 (GGGTTTAACC) was mutated to ATTTTTAACC, κB site-2 (AGTAGTTCCC) was mutated to AGTAGTTAAT, and a predicted κB site-4 (GGACAGTCCA) was mutated to TGAGAGTTAT. MsMoricin κB-GATA element (GTAAAGTCCCTATCGTTATCTG) was mutated to mutκB-GATA (with a mutated κB site) (TTAGAGTTATTATCGTTATCTG), κB-mutGATA (with a mutated GATA site) (GTAAAGTCCCTATCGAAAAACG), or mutκB-mutGATA (with both mutated κB and GATA sites) (TTAGAGTTATTATCGAAAAACG). To make insertion constructs, κB-GATA, mutκB-GATA, κB-mutGATA, mutκB-mutGATA, GATA (TATCGTTATCTGAGAG), MPAE (−242 to −57), MPAE-κB (−242 to −47), and MPAE-κB-GATA (−242 to −35) were inserted into different promoters at positions indicated in figures, respectively. Plasmids for transfection were prepared with PureYield™ Plasmid Miniprep System (Promega, USA). Sequences of the promoters are given in Texts S1–S9, where known or predicted NF-κB elements and GATA elements are noted by single and double underlines, respectively. Transcriptional/Translational start sites are shown in boxes.</p><!><p>Drosophila Schneider 2 (S2) cells and S. frugiperda Sf9 cells were maintained at 27°C in TNM-FH (HyClone, USA) supplemented with 10% fetal bovine serum, 1×L-Glutamine, 50 IU/mL penicillin and 0.05 mg/mL streptomycin (Hyclone, USA). For DNA transfection, 104 cells were plated in each well of a 96-well plate and transfected with 150 ng reporter plasmid and 15 ng pRL-TK renilla luciferase plasmid as an internal control (Promega, USA) for 12 h. Then fresh medium containing 10 μg/mL PG-K12 was used to stimulate cells for 48 h before measuring the luciferase activities. Even though 20-hydroxyecdysone-treated cells are more sensitive to immune challenges (Dimarcq et al., 1997; Silverman et al., 2000), we did not use the hormone to treat cells in our experiments because our purpose was to observe increase in promoter activities activated by PG-K12. Each transfection was performed in three wells independently and transfection experiments were repeated three times. Firefly and renilla luciferase activities were measured using Dual-luciferase Reporter Assay system (Promega, USA). Briefly, S2 cells or Sf9 cells in 96-well plates were washed once with sterile PBS solution (137 mM NaCl, 2.7 mM KCl, 4.3 mM Na2HPO4, 1.47 mM KH2PO4, pH 7.4), and then lysed with 30 μL 1× lysis buffer (Promega, USA) at room temperature for 15 min with shaking. The cell lysate (30 μL) was transferred to a tube containing 100 μL luciferase assay reagent II and mixed well, then reading was recorded immediately, 100 μL stop & glo reagent was added to quench the first reaction and the control renilla luciferase activity was measured using a Liquid Scintillation Counter (Cat #: 425-034, HIDEX, Turku, Finland).</p><!><p>Figures were made with the GraphPad Prism software with one representative set of data. Significance of difference was determined by an unpaired t-test or by one way ANOVA followed by a Tukey's multiple comparison test using the same software (GraphPad, San Diego, CA).</p><!><p>We cloned a 1.4-kb 5′-regulatory fragment of M. sexta moricin (MsMoricin) gene (GenBank accession number: JF309316.1) and constructed a luciferase reporter. Initial activity assay in D. melanogaster S2 cells (a dipteran cell line) and S. frugiperda Sf9 cells (a lepidopteran cell line) showed that MsMoricin promoter did not have activity in S2 cells, but had relatively high activity in Sf9 cells after peptidoglycan (PG) stimulation (Figure 1). This result suggests that insect AMP genes may be regulated in a species-specific manner. We used PG-K12 (from E. coli K12) to directly stimulate S2 and Sf9 cells since we did not overexpress any Rel/NF-κB proteins in these cells, and PG-K12 can bind to cell surface peptidoglycan-recognition proteins (PGRPs) to activate the IMD pathway in Drosophila (Kaneko et al., 2004). Thus, the activity observed is due to activation of promoters by endogenous transcription factors in S2 or Sf9 cells. Activation of the Toll pathway by Lys-type peptidoglycan from Gram-positive bacteria not only requires PGRPs but also involves activation of proteinases and Spätzle (ligand for the Toll receptor) (Ganesan et al., 2011), thus we did not use Gram-positive peptidoglycan in our study. All the following experiments in S2 and Sf9 cells were stimulated with PG-K12.</p><p>To further test whether insect AMP gene promoters indeed are regulated in a species-specific manner, we cloned two more M. sexta AMP promoters (MsCecropin and MsLysozyme) and seven D. melanogaster AMP promoters (drosomycin (Drs), diptericin (Dpt), drosocin, metchnikowin, DmCecropin A1, DmAttacin A, and DmDefensin), and constructed these promoters as luciferase reporters. Results from dual luciferase assays showed that MsMoricin and MsCecropin promoters were active only in Sf9 cells; drosomycin, DmDefensin, DmCecropin A1, metchnikowin and drosocin promoters were either active only in S2 cells or showed significantly higher activities in S2 cells than in Sf9 cells (Figure 1). DmAttacin A and diptericin promoters had lower activities in both S2 and Sf9 cells, and MsLysozyme promoter showed similarly high activities in both cell lines (Figure 1). These results suggest that some AMP gene promoters indeed are regulated in a species-specific manner.</p><!><p>In order to identify promoter regulatory elements, we first studied regulation of MsMoricin gene. Analysis of MsMoricin promoter sequence showed that there are five predicted κB sites and eight GATA sites with one GATA-1 site only 2bp downstream of the κB5 (Figure 2A). Deletion of the κB1, 2, 3 or 4 did not have an effect on MsMoricin promoter activity; however, deletion of the κB5 almost completely abolished promoter activity in Sf9 cells (Figure 2B). Deletion or mutation of either the κB5 or the adjacent GATA-1 site significantly decreased MsMoricin promoter activity in Sf9 cells (Figure 2C), indicating that the κB5 and GATA sites function together as an active element required for activation of MsMoricin promoter.</p><!><p>To further identify active elements in MsMoricin promoter, we made several deletion constructs and found that a short promoter of about 240 bp (Mor0.24) was fully active as 1.4kb MsMoricin promoter (Figure 3). Further deletions to the 240-bp promoter caused gradually loss in promoter activities. Noticeably, Mor0.1 (0.1kb), a construct with a complete κB5-GATA element, was inactive. The region between Mor0.24 and Mor0.1 was obviously critical to promoter activity as Mor0.24 (0.24kb) had similarly high activity as MsMoricin promoter (1.4kb) did, but Mor0.1 (0.1kb) almost had no activity (Figure 3). These results indicated that the κB-GATA element was necessary but not sufficient to activate MsMoricin promoter, and the region between -240bp and −100bp in the promoter (designated as Moricin Promoter Activating Element, MPAE) may contain important co-regulator binding sites, and these co-regulators are also required to fully activate moricin gene.</p><!><p>We showed that the κB-GATA element was essential but not sufficient to activate MsMoricin promoter (Figures 2 and 3). To test whether the κB-GATA element (22bp) from MsMoricin promoter can increase promoter activities of M. sexta and D. melanogaster AMP genes in S2 and/or Sf9 cells, the κB-GATA element was inserted into MsCecropin, diptericin and DmAttacin A promoters. Diptericin and DmAttacin A promoters were selected for the experiments since they had low activities in S2 cells (Figure 1). MsCecropin promoter contains six predicted GATA sites (Figure 5A). Diptericin and DmAttacin A promoters both contain a GATA site adjacent to a κB site (Senger et al., 2004) (Figures 4A and 5A), and the two κB sites in diptericin promoter have identical sequences (Kappler et al., 1993). When an MsMoricin κB-GATA element was inserted 40bp upstream of the endogenous GATA site (30bp downstream of the κB1) in diptericin promoter, a significantly increase in promoter activity was observed in both S2 and Sf9 cells, but insertion of mutant κB-GATA elements (either with a mutated κB or GATA site, or both) did not significantly change promoter activity (Figure 4B and C). In addition, insertion of a κB-GATA element increased diptericin promoter activity to a significantly higher level in S2 cells (~24-fold of diptericin promoter) than in Sf9 cells (~15-fold) (Figure 4B and C). Similarly, insertion of a κB-GATA element into DmAttacin A and MsCecropin promoters also significantly increased activities of the two promoters in S2 cells (Figure 5B), but did not increase promoter activities in Sf9 cells (Figure 5C). Together, these results suggest that the MsMoricin κB-GATA element can increase AMP promoter activity, and this element may not be species-related.</p><!><p>D. melanogaster drosomycin promoter contains three κB sites, the κB site1 and site2 are mediated by the Toll and IMD pathways, respectively, and the two κB sites have synergic effects on drosomycin activation (Tanji et al., 2007; Tanji et al., 2010), but the κB site3 is not essential for promoter activity (Tanji et al., 2007). Drosomycin promoter also contains four GATA binding sites (Senger et al., 2004). Analysis of drosomycin promoter sequence showed another predicted κB site in between the κB site1 and site2, which we named the κB site4 (Figure 6A). In the following experiments, we focused our experiments on drosomycin promoter since its two κB sites are regulated differently. To test whether the predicted κB site4 is essential for drosomycin promoter, the κB site4 was deleted or mutated, and the results showed that deletion and mutation of the κB site4 did not decrease drosomycin promoter activity in S2 and Sf9 cells (Figure S1, panel A), indicating that the predicted κB site4 is not required for drosomycin promoter. We then replaced the κB site4 (10bp) with an MsMoricin κB-GATA element (22bp). Replacement of the κB site4 with a κB-GATA element significantly increased promoter activity in S2 cells (~21 folds, indicated by 4rep(κB-GATA)-Drs), but similar replacement with mutant κB-GATA elements (with a mutated κB or GATA site or both, indicated by 4rep(mutκB-GATA)-Drs, 4rep(κB-mutGATA)-Drs, or 4rep(mutκB-mutGATA)-Drs) did not significantly change activity of these promoters compared to drosomycin promoter (Figure 6B). These results demonstrated that an additional κB-GATA element can increase drosomycin promoter activity, and both the κB and GATA sites are required to increase promoter activity.</p><p>To test whether increase in drosomycin promoter activity by the MsMoricin κB-GATA element requires the endogenous κB site1 or site2, the κB site1 or site2 in the κB-GATA replaced promoter was mutated (indicated by 4rep(κB-GATA)-1mut-Drs or 4rep(κB-GATA)-2mut-Drs). Mutation of the site1 did not impair the κB-GATA element to increase promoter activity; however, mutation of the site2 completely abolished the ability of the κB-GATA element to enhance promoter activity in S2 cells (Figure 6B). These results suggest that the κB-GATA element and the endogenous κB site2 act cooperatively to increase drosomycin promoter activity. In Sf9 cells, all the site4 replacement and mutation constructs showed similar low activities as drosomycin promoter did (Figure S1, panel B), suggesting that the κB-GATA element cannot increase drosomycin promoter activity in Sf9 cells.</p><!><p>We also mutated the κB site1 and site2 in drosomycin promoter and found that mutation of the κB site1 did not cause a loss of promoter activity in S2 cells, and mutation of the κB site2 significantly decreased promoter activity in S2 cells, but both mutations did not have an effect on promoter activity in Sf9 cells (Figure 7). These results in S2 cells were consistent with those reported previously that the κB site1 and site2 are activated by the Toll and IMD pathways, respectively (Tanji et al., 2007; Tanji et al., 2010), as PG-K12 (peptidoglycan from E. coli strain K12) that can activate the IMD pathway in Drosophila was used to stimulate S2 cells in our experiments. To test whether the MPAE (140bp), κB, GATA, or the whole MPAE-κB-GATA element (205bp, from −240 to −35bp) from MsMoricin promoter can increase drosomycin promoter activity in S2 cells and/or Sf9 cells in particular, an MPAE was inserted prior to the κB site1 or site2, or the κB site1 or site2 was replaced with an MPAE, MPAE-κB or MPAE-κB-GATA. Our results showed that insertion of an MPAE before the κB site1 did not significantly change promoter activity in S2 cells, but did significantly increase promoter activity in Sf9 cells (1.8-fold of drosomycin promoter) (Figure 7A and B). Replacement of the κB site1 with an MPAE increased promoter activity significantly in both S2 cells (3.5-fold) and Sf9 cells (1.7-fold) (Figure 7A and B). Replacement of the κB site1 with an MPAE-κB did not further increase promoter activity compared to the MPAE replacement. However, replacement of the κB site1 with a whole MPAE-κB-GATA element increased promoter activity to a significantly higher level in both S2 cells (11.6-fold) and Sf9 cells (3.8-fold) (Figure 7A and B). Since the κB site 1 is not activated by PG and drosomycin promoter did not have activity in Sf9 cells (Figure 1), these results suggest that MPAE could specifically increase drosomycin promoter activity in Sf9 cells.</p><p>Drosomycin κB site2 is activated by Gram-negative peptidoglycan via the IMD pathway (Tanji et al., 2007; Tanji et al., 2010), thus mutation of the κB site2 significantly decreased promoter activity in S2 cells (0.2-fold of drosomycin promoter) (Figure 7C). Insertion of an MPAE prior to the κB site2 and replacement of the κB site2 with an MPAE and MPAE-κB significantly decreased promoter activity in S2 cells (0.6-, 0.25- and 0.3-fold of drosomycin promoter, respectively) (Figure 7C), suggesting that an MPAE or MPAE-κB could not replace the κB site2 to activate drosomycin in S2 cells. But insertion of an MPAE before the κB site2 significantly increased promoter activity in Sf9 cells (5.6-fold), while replacement of the κB site 2 with an MPAE and an MPAE-κB did not have an effect on promoter activity in Sf9 cells (Figure 7D). However, replacement of the κB site2 with a whole MPAE-κB-GATA element increased promoter activity to a significantly higher level in both S2 cells (1.8-fold) and Sf9 cells in particular (14.3-fold) (Figure 7C and D). Since the κB-GATA element could increase drosomycin promoter activity in S2 cells but not in Sf9 cells (Figures 6B and S1), the κB site2 but not the κB site1 is required for peptidoglycan stimulation (Tanji et al., 2007; Tanji et al., 2010), these results indicated that MPAE may contain some binding sites for nuclear factors expressed specifically in Sf9 cells (a lepidopteran cell line), which can increase drosomycin promoter activity in Sf9 cells.</p><p>To further confirm that the MPAE can increase drosomycin promoter activity in Sf9 cells, we replaced the predicted κB site4 (10bp), a non-essential site, with an MsMoricin κB-GATA or an MPAE-κB-GATA (Figure 8A and B). Replacement of the κB site4 with a κB-GATA significantly increased promoter activity in S2 cells (Figure 8A), but did not have an effect on promoter activity in Sf9 cells (Figure 8B). Replacement of the site4 with an MPAE-κB-GATA though also significantly increased promoter activity in S2 cells compared to the control drosomycin promoter, but decreased promoter activity compared to the κB-GATA replacement (Figure 8A), suggesting that an MPAE did not further increase (actually decreased) promoter activity in S2 cells. However, the MPAE-κB-GATA replacement significantly increased promoter activity in Sf9 cells (Figure 8B), indicating that MPAE indeed contains lepidoptera-related co-regulator binding sites that can increase drosomycin promoter activity in Sf9 cells. We also inserted an MsMoricin GATA site just after the κB site1 and the site2 of drosomycin promoter and our results showed that a GATA insertion after the κB site1 significantly increased promoter activity in S2 cells (5.8-fold), but did not increase promoter activity in Sf9 cells, and a GATA insertion after the κB site2 did not have an effect on promoter activity in both S2 and Sf9 cells (Figures 8C and S1).</p><!><p>Most AMP promoters investigated in this study showed species-specific regulation (Figure 1), suggesting that certain components in the transcription complex may account for the species-specific regulation. Since the κB and GATA transcription factors bind to similar consensus sequences across different species, M. sexta κB-GATA element may not be species-related and therefore is functional in D. melanogaster S2 cells. We hypothesize that transcription of insect AMP genes may involve formation of a transcription complex composed of both common factors (NF-κB and GATA) and species-related co-regulators, and it is the co-regulator that confers species-specific regulation. Indeed, some AMPs are species-specific. For example, moricin, gloverin and lebocin genes have been identified only in lepidopteran insects so far (Axen et al., 1997; Chowdhury et al., 1995; Hara and Yamakawa, 1995). D. melanogaster deformed epidermal autoregulatory factor-1 (DEAF-1) has been identified as a new factor that contributes to induced expression of metchnikowin and drosomycin (Gross and McGinnis, 1996; Reed et al., 2008), two species-specific AMP genes in Drosophila. In addition, Dorsal interacting proteins have been identified (Li et al., 2007; Ratnaparkhi et al., 2008). These factors/proteins may function as species-related co-regulators. MsLysozyme promoter showed similarly high activity in both S2 and Sf9 cells (Figure 1). We did not identify κB sites in MsLysozyme promoter by in silico analysis. Moreover, induced expression level of MsLysozyme mRNA by different bacterial components was always significantly lower than that of other AMP genes that are regulated by NF-κB factors (X-J Rao and X-Q Yu, unpublished data). Therefore, MsLysozyme is likely not regulated by NF-κB factors. It is also possible that regulation of AMP promoters may be tissue-specific, since S2 and Sf9 cells were used in this study, and S2 cells were hemocyte origin whereas Sf9 cells were developed from the ovary.</p><!><p>MsMoricin κB-GATA element is necessary but not sufficient to activate MsMoricin promoter induced by E. coli peptidoglycan (Figures 2 and 3). This result indicates that other co-regulators are required to cooperate with NF-κB and GATA factors to activate transcription of MsMoricin. These co-regulators likely bind to the 140bp MPAE (MsMoricin Promoter Activating Element) region (Figures 2A and 3), which contains predicted binding sites for nuclear factors such as YY-1, Pit-1, Oct-1, and C/EBP. However, deletion of a predicted YY-1 (−201 to −192bp) and a Pit-1 (−185 to −176bp) site in MPAE did not have an effect on MsMoricin promoter activity (X-J Rao and X-Q Yu, unpublished data), suggesting that there may be novel co-regulator binding sites in the MPAE element.</p><!><p>It has been reported that when Drosophila Toll and IMD pathways are stimulated simultaneously at a low level, there is a synergic effect on activation of drosomycin gene probably due to formation of Dorsal-Relish and/or Dif-Relish heterodimers that bind to the κB site2 (Tanji et al., 2007; Tanji et al., 2010). However, how Dorsal-Relish and/or Dif-Relish heterodimers synergistically activate drosomycin promoter is not well understood. We showed that the κB-GATA element from MsMoricin promoter could enhance activities of Drosophila AMP promoters (Figures 4–6) in S2 cells. We also showed that drosomycin promoter could be activated cooperatively by the endogenous κB site2 and the exogenous κB-GATA element (Figures 6B). Nonetheless, the same set of reporters consistently showed low activities in Sf9 cells (Figure S1), indicating that the κB-GATA element is not a species-related activating element.</p><p>MPAE specifically increased drosomycin promoter activity in Sf9 cells (Figure 8B), but not in S2 cells (Figure 8A), strongly suggesting that MPAE indeed contains lepidoptera-related co-regulator binding sites. Since the κB site1 is not activated by the IMD pathway (Tanji et al., 2007; Tanji et al., 2010), high activity of the 1rep(MPAE-κB-GATA)-Drs reporter in S2 cells (Figure 7A) may result from synergic effect of the κB site2 and the κB-GATA element. But high activity of the 1rep(MPAE-κB-GATA)-Drs reporter in Sf9 cells (Figure 7B) is due to both the MPAE and the κB-GATA elements. The κB site2 is activated by the IMD pathway (Tanji et al., 2007; Tanji et al., 2010), low activities of the MPAE and MPAE-κB replaced reporters (2rep(MPAE)-Drs and the 2rep(MPAE-κB)-Drs) (Figure 7C) suggest that an MPAE or an exogenous κB cannot substitute for the κB site2 to activate drosomycin in S2 cells. However, a whole MPAE-κB-GATA element is a stronger element than the κB site2, since it could increase drosomycin activity to a significantly higher level in S2 cells (Figure 7A and C). Insertion of an MPAE alone before the site2 already caused a significant increase in promoter activity in Sf9 cells (2ins(MPAE)-Drs reporter in Figure 7D). A whole MPAE-κB-GATA element increased activity in Sf9 cells to the highest level (2rep(MPAE-κB-GATA)-Drs reporter in Figure 7D). These results altogether indicate that MPAE contains lepidoptera-related regulators that can activate drosomycin promoter in Sf9 cells, and the MsMoricin κB-GATA is a stronger element than the endogenous κB site2 in activation of drosomycin promoter.</p><p>The κB-GATA element from MsMoricin increased activity of diptericin promoter in both S2 and Sf9 cells, and increased activity of drosomycin promoter only in S2 cells but not in Sf9 cells (Figures 4, 6 and S1). Diptericin is activated by the IMD pathway, while drosomycin is mainly activated by the Toll pathway. MsMoricin κB-GATA was also activated by the IMD pathway (Figure 3) as we used PG-K12 to stimulate cells. Thus, it is possible that an extra κB-GATA element and the endogenous κB sites in diptericin promoter act cooperatively to increase promoter activity in Sf9 cells.</p><!><p>D. melanogaster Dorsal binds consensus sequence of GGG(A/T)(A/T)(T/A)(A/T/C)(C/A/T)(T/G/C); Relish binds consensus sequence of GGGA(A/T/C)N(C/T)(C/A)(C/T); Dif/Relish heterodimer binds consensus sequence of GGGA(A/T)TC(C/A)C (Busse et al., 2007; Senger et al., 2004). Drosomycin κB site1 (GGGTTTAACC) is consistent with Dorsal binding consensus; the κB site2 (GGGAACTACT) is consistent with Relish binding consensus; MsMoricin κB5 (GGGACTTTAC) is not completely consistent with any of the three consensus sequences. MsMoricin GATA site (CAGATAACGA) is consistent with Drosophila Serpent consensus sequence [(A/T/C)GATA(A/G)(C/T/G)] (Senger et al., 2004). Based on previous reports and our data, we propose that active RelN/RelN homodimer may bind to MsMoricin κB5 site and drosomycin κB site2, and a Drosophila GATA factor (Serpent, for example) may bind to MsMoricin GATA site to achieve maximal synergy in drosomycin promoter in S2 cells (Figure 9).</p><p>The position, direction and distance between the κB and GATA sites are also important for Rel-GATA synergy (Senger et al., 2004; Vardhanabhuti et al., 2007). The features of MsMoricin κB-GATA element used in our experiments are consistent with those reported previously (Figure 2A and Text S1) (Kadalayil et al., 1997; Senger et al., 2004). An MsMoricin GATA element alone could significantly increase drosomycin promoter activity in S2 cells when inserted after the endogenous κB site1 but did not show this effect when inserted after the κB site2 (Figure 8C). This may be due to special needs for the position, orientation and/or spacing of the inserted GATA element.</p><!><p>What we report here may have broad applications in transgenic engineering to increase antibacterial activities in different organisms. A κB-GATA element may be inserted into a promoter to drive expression of a transgene only after induction by bacteria, and a species-specific element, if identified, may be inserted into a promoter to drive expression of an exogenous gene in an organism. Plants do not have NF-κB factors, but they do have GATA factors and AP-1 like WRKY factors (Reyes et al., 2004; Ronald and Beutler, 2010). AP-1 and GATA-2 cooperatively regulate expression of Endothelin-1, a vasoactive peptide from endothelial cells (Kawana et al., 1995). Thus, WRKY-GATA synergy might exist in plants too, although there have been no reports so far. More research is needed to further identify unknown co-regulators in the transcription complex as this will shed light on molecular mechanisms of immune gene regulation.</p>

PubMed Author Manuscript

Cryo-EM structure of acylpeptide hydrolase reveals substrate selection by multimerization and a multi-state serine-protease triad

The first structure of tetrameric mammalian acylaminoacyl peptidase, an enzyme that functions as an upstream regulator of the proteasome through the removal of terminal N-acetylated residues from its protein substrates, was determined by cryo-EM and further elucidated by MD simulations. Selfassociation results in a toroid-shaped quaternary structure, guided by an amyloidogenic b-edge and unique inserts. With a Pro introduced into its central b-sheet, sufficient conformational freedom is awarded to the segment containing the catalytic Ser587 that the serine protease catalytic triad alternates between active and latent states. Active site flexibility suggests that the dual function of catalysis and substrate selection are fulfilled by a novel mechanism: substrate entrance is regulated by flexible loops creating a double-gated channel system, while binding of the substrate to the active site is required for stabilization of the catalytic apparatusas a second filter before hydrolysis. The structure not only underlines that within the family of S9 proteases homo-multimerization acts as a crucial tool for substrate selection, but it will also allow drug design targeting of the ubiquitin-proteasome system.

cryo-em_structure_of_acylpeptide_hydrolase_reveals_substrate_selection_by_multimerization_and_a_mult

Introduction<!>Results and discussion<!>pAAP monomers present an open structure with a conformationally destabilized catalytic apparatus<!>Conclusion<!>Cryo-EM sample preparation and data collection

<p>The mode and purpose of protein multimerization are still under debate. A great number of proteins function as homodimers and a higher degree of homo-and hetero-association is also common, while the active form of only approximately 1/4 of human enzymes is monomeric. 1,2 Multimerization carries numerous advantages derived partly from the decrease of surface over volume ratio; it may restrain denaturation and promiscuous interactions, increase stability of the active sites by reducing internal motion and reinforcing their spatial protection, or decrease aggregation propensity by shieldingor utilizinginteraction-prone hot-spots, as well as allowing for cooperativity and providing a vehicle for introducing allosteric regulation. [3][4][5] The homo-multimer state is a special case of assembly, one that is reached overcoming the entropic disadvantage coupled to self-association of the monomeric units containing interaction prone surfaces, over association with any other protein (or non-protein) partners. 6,7 This requires homomeric contact surfaces to achieve a high degree of selectivity which seems to be ne-tuned and modulated by intrinsic dynamics, 8 insertions and deletions 9 and also by regions remote from the interaction surfaces 2 to achieve or preserve functionality. However, examples of different homo-multimerization forms fullling the same function can also be found and in these cases association surfaces seem not to be conserved or specied, shiing along the outer surface of the monomers. 5 Thus, it is still to be clar-ied whether the complexity of multimeric forms arose to create specic functions or simply through neutral or adaptive evolutionary processes.</p><p>Here we considered acylaminoacyl peptidase (AAP), a member of the S9 serine proteaseor oligopeptidasefamily, which is tetrameric in its mammalian form. [10][11][12] The functional form of S9 proteases ranges from monomeric through dimeric, tetrameric to hexameric. The central locus of multimerization is preserved, but not the sequence of the association motifs or the relative orientation of the interacting monomers. The most important interaction center is a b-edge at the surface of the catalytic domainthe outermost member of an 8-stranded bsheet forming the core of this region. This b-edge carries varying amyloidogenic sequence motifs, and thus it remains highly interaction-prone in all S9 proteases.</p><p>At the core of the physiological functions of AAP lies its ability to cleave the N-terminal acetylated (N-Ac) amino-acid from proteins and peptides. 13 Removal of the N-Ac tail of proteins leads to dramatic changes in their physicochemical properties such as charge, fold, localization, and proteinprotein interactions, but perhaps most importantly, it affects ubiquitination andthrough itthe degradation pathway proteins will enter. 14,15 Although the regulation of protein lifespan through the addition or removal of the terminal N-Ac is far from claried, AAP has been shown to play a signicant role in protein maturation and degradation processes, [16][17][18][19] as well as being a key upstream modulator of proteasomal function. 20,21 Downregulation of AAP leads to the accumulation of misfolded proteins through disruption of the ubiquitin-proteasome system (UPS), [22][23][24] a crucial sustainer of healthy homeostasis. Inhibiting components of the UPS is already in use as an anticancer therapeutic strategy and potential further targets are intensively researcheda eld where thus AAP could also be considered.</p><p>AAP is a serine-protease enzyme, however, its unique structure and multimerization state seem to create such a seclusion of its active site that potentiates interactions very much unlike those of regular serine-proteases, the most striking example of which is its interaction with and inhibition by carbapenem antibiotics. [25][26][27][28] Furthermore, AAP was suggested to be able to catalyze the cleavage of amyloid oligomers formed in the progress of Alzheimer's disease. 29 It was also shown to be a key protein in cellular response to DNA-damage 30 and oxidative stress, 31,32 and was proposed to be a potential tumor suppressor. [33][34][35] AAP seems to contribute to the membrane localization of major oncogenic protein K-Ras which may allow K-Ras induced cancers to be targeted through AAP inhibition. 36 A detailed understanding of AAP function and catalysis has so far been hindered by the lack of its structure. In spite of reports on its production, and preliminary results of crystallographic studies, 13,[37][38][39][40] the structure of mammalian AAP could not be determined over four decades that passed since its rst characterization. Structures of archeal and bacterial AAPs are unt for model building, since they not only share low sequence identity with the mammalian enzyme (13-20%, Table S1 †), but also mostly assume different oligomeric forms. 41 Here we present the cryo-EM determined structure of AAP isolated from porcine liver (pAAP). The tetrameric structure shows a gated and channeled intricate interior. Unique inserts that appear in only the mammalian variants participate in forming the main interaction surfaces of the multimer, therefore dening exibility and architecture of the complex. Our MD simulations show that these large insertions function as shutters for the substrate access channel system, providing the means for substrate screening and selection. Thus, in the case of pAAP, tetramerization seems to be a prerequisite for reaching the catalytically competent form. We found the classical serine protease active site of pAAP in an alternating state via the unprecedented destabilization of the active Ser region, shiing between an active and a latent conformer, as conrmed by both the experimentally determined cryo-EM map and MD simulations. The determined structure will allow, for the rst time, structure-based modeling, drug design and detailed analysis of the catalysis of this many faceted enzyme.</p><!><p>Using cryo-EM, the 3.27 Å resolution structure of the pAAP tetramer was determined (Fig. 1 and S1 †). The overall structure is toroidal shaped, showing D2 point group symmetry, with two large openings on the outer surface and an intricate antechamber formed on the inside, from where the active sites of the monomeric units can be reached. To characterize the robustness and exibility of the determined structure and to model those segments that could not be resolved in the cryo-EM map, MD simulations were also carried out, using the experimentally determined structure as a starting model.</p><!><p>AAP is an oligopeptidase belonging to the S9 family of serine proteases. The monomeric unit of AAPsand of other members of the S9 family such as prolyl-oligopeptidase (POP), dipeptidylpeptidase 4 (DPP4) or oligopeptidase B (OPB)is built from two domains: the better conserved hydrolase domain and the highly variable propeller domain (usually composed of 7 blades, each formed by four antiparallel b-strands) (Fig. 1, S2 and Table S2 †). The active site with the catalytic Ser/His/Asp triad is buried between the two. 41 The pAAP monomers contain a 7-bladed propeller domain (residues 23-457) that is tilted away from the hydrolase (residues 1-22 and 458-732) by approximately 40 , creating a large opening near the active site beneath blade2 (Fig. 1C) that can be accessed from the central cavity of the tetramer. The opening created by domain orientation is similar in extent to that seen in the case of the open form of Aeropyrum pernix AAP (ApAAP). 43 In contrast to ApAAP, where the resting state of the enzyme was shown to be a mixture of inactivated open and catalytically competent closed conformers, in case of pAAP interdomain contacts within the tetramer suggest that the domainorganization is rigid. The entrance of the pAAP monomer is further shaped by the loops of blades 3-5, downscaling the large opening to approximately 20-25 Å diameter. These segments are not present in the open form of ApAAP, which thus presents a decidedly wider entrance inside its monomeric unit.</p><p>Comparing the structure of the monomer of pAAP to those of the S9 family members determined previously reveals three unique insertions on its surface (Table S2 and Fig. S2, S3 †). All four strands of blade1 are elongated, protruding from the outer surface of the propeller domain (Fig. 1: insert1: residues 33-46, 52-61 and 69-80). Blade3 also contains an unusually long insert (Fig. 1: insert2: residues 168-222), forming a long, crossed loop that reaches the surface and contributes to the shaping of the tetramer entrance. A long insertion, containing mainly hydrophobic residues, can also be found in the hydrolase domain (Fig. 1, insert3: residues 636-644), facing blade4 of the propeller. Catalytic Ser587 is placed on a tight turn between a bstrand and an a-helix (Fig. 2)the conformational strain imposed on the backbone contributes to its activation.</p><p>In pAAP, a crucial substitution was found in the middle of the b-strand running parallel and adjacent to the strand leading up to the catalytic Ser587. In place of the Val, Ile, Ala or Gly residues found in most of the S9 proteases in this spot, in pAAP (and also in the human variant) Pro506 appears (Fig. 2A and S4 †). The presence of the Pro breaks the b-strand and abolishes the stabilizing interaction between the neighboring strands, allowing greater exibility for the segment containing the catalytic Ser587. Accordingly, we detected conformational heterogeneity of the 584-591 segmenthosting Ser587in the cryo-EM map, which could be described with two conformers of this segment: one can be considered as active, while the other as a latent state of the catalytic apparatus based on their conformation and the corresponding H-bonding network (Fig. 2B and S5 †). In the latent state, Ca of the catalytic Ser587 is shied by 2.4 Å away from the catalytic His707, which places the Ser-Og at 3.8 Å from the N32 of His (while they are at 2.5 Å from each other in the active conformer). The conformational heterogeneity of the 584-591 segment appears to be localized to the catalytic sites which are distant from each other within the tetramer (the closer pairs are separated by more than 30 Å (measured at Ser587 Cas)) and show no signs of being in (direct) structural communication with each other, and thus it is reasonable to suppose that the different conformational states of the Ser-loop are evenly distributed among the subunits (both in time and space). Since these regions are also buried deep within the tetrameric assembly, they resulted in no global structural differences that could have been detected during data processingat least not at the 4.12 Å resolution of the unsymmetrized C1 mapand thus the nal models were built using the D2 averaged map of signicantly better resolution (3.27 Å). This procedure, however, does not allow differentiation of conformational patterns within the tetramer (resulting in a single, averaged picture of the monomers).</p><p>On the other hand, MD simulation of the pAAP tetramer (using a model based on the experimentally determined structure) also sampled both conformers seen in the cryo-EM results, in addition to a number of intermediate states (Fig. 2C and D). While the distance of His707 Nd1 and the carboxylate oxygens of Asp675 remained within H-bonding distance in over 97% of the snapshots, in only 16.3% of the snapshots was the Ser587-His707 H-bond intact. Inactivation of the catalytic triad was previously reported for the open conformers of monomeric or dimeric S9 proteases 43,44 (and for the truncated version of Deinococcus radiodurans carboxypeptidase, DrCP 45 ), but in all cases this was achieved by the destabilization of the His-loop (containing the catalytic His). A serine protease containing multiple conformers of the Ser-loop is unprecedented, to the best of our knowledge. The cryo-EM and simulation results thus both suggest that the active site of pAAP is conformationally heterogeneous and probably requires the docking of the substrate to reach a fully catalytically competent state.</p><p>The primary substrate specicity pocket is hydrophobic, similarly to that of other AAPs (Table S3 †). However, compared to archeal AAPs, it is narrower with a Trp sidechain lining its lower region (Trp628; Fig. S6 †) possibly playing a role in shiing S1 specicity of pAAP towards smaller P1 residues (Val, Ala 37 ). AAPs are exopeptidases, but were also reported to have endopeptidase properties. 13,46,47 The substrate binding groove of ApAAP and PhAAP is widened near the surface so they can accommodate substrates, where the site of hydrolysis is not the peptide bond following the rst (N-acetylated) residue of the chain, but further along the sequence. 48,49 In contrast, this region in the pAAP structure is restricted by Phe274 and Cys275 (of an insert to blade4) forming a barrier locked in place by hydrogen bonds with His325 (Fig. S6 †). This also forces the conserved Arg677 into a latent conformation forming a salt bridge with Asp624, from which however, Arg677 may be released to form a hydrogen bond with the P2 residue of larger substrates. These features indicate that pAAP and its human variant might be less effective endopeptidases than ApAAP or PhAAP.</p><p>Another notable feature of AAPs is that a cisPro also appears at the active site ensuring the conformation of its AAP-specic HGGP motif that forms the oxyanion site. 50 The corresponding cisPro510 of pAAP restricts the pliability of the 507-519 loop placing the backbone N-H of Gly509 in ideal position for coordinating the tetrahedral transition state of the catalytic reaction. Based on the structure, this loopas an important interaction hubcan be effected by a pathogenic mutation of human AAP, the Thr541Met mutation that causes 50-60% decrease in activity. 51 Thr541 makes contacts with loop 507-519, and the Thr / Met switch may lead to loop restructuring, and through that, distortion of the oxyanion site (Fig. S7 and S8). † 52 The model of the monomeric unit of human AAP (hAAP) can be found in the AlphaFold Database. 53 The model is remarkably close to the experimentally determined structure of pAAP (backbone RMSD for the propeller and hydrolase domains is 1.29 Å and 1.28 Å, respectively, while along the full sequence, 1.48 Å) with a few distinct and functionally important differences. The domain opening was slightly underestimated (Fig. S9 †), the cis conformer of Pro510 was not predicted, neither was the liberation of the Ser-loopor the mode of tetramerization.</p><p>Highly interconnected tetrameric assembly with a doublegated shutter system Formation of the tetrameric structure of mammalian pAAP is guided by two main interaction surfaces and is ne-tuned by the long blade3 insertions (insert2) that wrap around the outer pore of the toroid, creating a highly interconnected structure (Fig. 3). All three unique inserts shown in Fig. 1 participate in forming the key interaction surfaces of the tetramer. The primary interaction surface (A/B and C/D interactions, Fig. 3B) contains, among others, the outermost b-strand of the core b-sheet of the hydrolase domain, the unique (mammalian-specic) insert2 of the propeller and insert3 of the hydrolase domain. Since the segments that participate in forming the interaction surface come from both domains, and numerous propeller-hydrolase contacts are also formed, tetramerization xes the interdomain opening of the monomers. The secondary, orthogonal interaction surface (A/D, B/C, Fig. 3B) is formed by insert1the extra elongation of blade1 and blade2. The topology of this interaction is unusualthere is an opening between blade1 and blade2 of each propeller domain creating a deep crevicethis is where blade1 of the neighboring monomer is docked in a domain-swapping interaction. And since the blade1 extensions only appear in pAAP (and hAAP, according to the sequence t), this type of interunit interaction has thus far been unseen in the S9 family. A third, small interaction surface was also detected between monomers of A/C and B/D between blade2 of one monomer and insert2 of blade3 of the opposing monomer (Fig. 3B) completing the network of interconnections between all monomers. Simply put, this means that insert2 of blade3 is long enough not only to reach the outer surface of the tetramer and wind along the hydrolase domain of the neighboring unit, but also to progress along the pore all the way to the monomer at the opposing end.</p><p>The presence of two orthogonal interaction surfaces that both have a signicant contribution to the stability of the multimer means that pAAP is not merely a loose association of dimers but a genuine tetramer. Since tetramerization is responsible for the emergence of the double-gated shutter system that seems to be able to screen potential substrates of the enzyme, the inserts present that make this specic mode of multimerization possible are of great signicance. The only other member of the S9 family that is known to form a tetrameric structure is DrCP. 45 However, in that case only one type of interunit interaction contributes to the stability of the multimer, one that is similar to the primary interaction surface of pAAP, because insert1 is missing and insert2 is not long enough to reach the surface of the tetramer. Thus, the resultant tetramer is characteristically different from that of mammalian AAP.</p><p>Being an interconnected tetramer, pAAP possesses all the possible substrate routes previously seen in oligopeptidase structures: leading through the propeller channel or through the tetramer pore and monomer side opening (Fig. 1). The propeller channel is quite wide here compared to other oligopeptidases, 55 and therefore it may serve as a potential alternative entry-route towards the catalytic cavity (Fig. S10 †) for smaller substrates. In fact, using the FTMap server 56 smallligand binding sites were located along the propeller channel, indicating its accessibility (Fig. S10 †). To explore the possible roles of all different pores and channels in the substrate entrance, we analyzed the MD trajectories from this respect also. The segments that could not be resolved in the cryo-EM map (residues 1-8, 39, 110-115, 183-198, 496-497) proved also to be the most exible in the MD simulations (Fig. S11 †). Two especially mobile regions were found. One is the outer segment of insert2 that forms the shape and size of the entrance pores on the two opposing surfaces of the tetramer and the other is the 110-117 moiety of blade2 that guards the entrance to the monomeric units in the inner cavity, where the corresponding segments of all four monomers are placed into proximity by tetramerization (Fig. 4). Both may allow or block the further advance of the substrates, and neither would be present in different multimerization states.</p><p>We carried out clustering of the conformations accessed by these most mobile regions during the MD simulation and selected extreme conformations of both. These were added to the cryo-EM determined core structure. Channel analysis 55 of the resultant (hybrid) structural models indicates that the gating loops uctuate between endpoints of an open and a closed structure: a state that provides an approximately 80 Å long channel through the tetramer pore and the central antechamber to the monomer side openings and active sites, and another where the catalytic triad can only be accessed by the signicantly narrower ($60% in diameter) channel (Fig. 4). It is important to note that the width of the long channel from the outer pore to the active site is just wide enough for allowing unstructuredor unfoldedchains or shorter oligopeptides to reach the catalytic apparatus. Opening of the outer pore is a frequent phenomenon during the simulation -25.3% of the structures belong to this class (while 24.9% can be categorized as closed, with 49.8% of the snapshots carrying the outer loop in intermediate conformation) and was also identied by PCA analysis as the most signicant uctuation of the structure (Fig. S11 †). On the other hand, the opening of the inner gate at the monomer side is a rare event (4.2% of the snapshots). The results indicate that the two most exible regions of pAAP together form a double-gated shutter system that larger substrates (i.e. oligopeptides, protein segments) must encounter with while reaching the active sites: rstly at either of the two outer openings of the tetramer toroid, secondly within the interior chamber at the monomer openings.</p><p>The outer gatekeeping loops contain a large number of charged and polar amino acids, in distinct patches: a KKRK and a TSDDE motif followed by a RKK motif. Entering substrates must pass between two such gating-loops. We propose that interacting with these highly charged segments might be sufficient to promote the unfolding of the substrates, by decoupling salt-bridges and H-bonds, destabilizing their secondary and tertiary structure. The exibility of the long loops allows interaction with a variety of targets. Curiously, Ser187 of human AAP (also present in pAAP)located in the central negatively charged motif of insert2was identied as a phosphorylation site. 57 Phosphorylation at this spot could inuence interaction with the substrates and between the two gatekeepers themselves, offering a possible mode of regulating AAP function.</p><p>The interaction hot-spot: b-edge of the hydrolase domain Within the S9 protease family, the histidine residue of the catalytic triad is placed on a loop (His-loop) that is connected to the terminal b-strand of the 8-membered central b-sheet, at the surface of the core region of the hydrolase domain. This terminal b-strandkey member of the primary interaction surface of pAAPis situated as an "ideal" aggregation primer ("sticky b-edge"). The corresponding sequences in various oligopeptidases were recognized by ve different predictors as being (at least partially) aggregation-prone (Table S4 †). In the case of monomeric oligopeptidases, the sticky b-strand is covered by long N-terminal extensions, while in the case of dimeric, tetrameric and hexameric oligopeptidases, where these unprotected interaction-prone b-strands run along the outermost surface of the monomers, they are hidden by the process of multimerization 41,42 (Fig. 3C and S12 †).</p><p>A similar scenario was recently outlined in connection with steroid hormone receptors. The large, hydrophobic surface segment of monomeric ketosteroid receptors is covered by a Cterminal extension, while it is hidden by dimerization in estrogen receptors that do not carry the C-terminal tail. The authors conclusively demonstrated that these mechanisms provide structural stabilization (by covering aggregation-prone surfaces), but do not affect function. A model was proposed, where the evolutionary transition in the composition of the vulnerable interaction hot spotsmaking their protection more efficient while simultaneously destabilizing their unprotected forms -"entrenches" these regions, trapping the achieved multimerization state and interaction topology. 58 Intriguingly, among S9 proteases, it is only the central locus of interaction that perseveredthe outermost sticky b-edge of the hydrolase domain however, the mode of its protection greatly varies. And even more importantly, we believe that, among these enzymes, multimerization is a key component of the substrate selection and preparation apparatus.</p><!><p>Oligopeptidases pre-screen their substrates based on size. Some members of the S9 family were proposed to provide access to the active site through the narrow channel dissecting the propeller domain, 59 but in the case of most S9 proteases, two different substrate selection scenarios could be deciphered based on the determined structures. In one, the loops containing the His and Asp residues of the triad (His-loop and Asploop) are stabilized by inter-domain interactions, which provides the means for disassembly of the catalytic apparatus where the domains shied away from one another. This was shown to be the case for POP, OPB and ApAAP, where the monomers open up in a clamshell-like motion, inactivating the active site, but also making it readily accessible. Only those substrates will be cleaved by these enzymes thatbesides being compatible with the binding surface and substrate specicity pockets of the enzyme interioralso allow the reclosing of the domains, thus resulting in the restoration of the inter-domain interactions that lock the His-and Asp-loops in a conformation that reinstates the functional triad. This requires a build-up that does not restrict domain movement. Accordingly, POP and OPB are monomeric, while ApAAP forms dimers in a way that only involves hydrolase-hydrolase type of interactions, allowing the free movement of the propeller domains. Another strategy for pre-screening of substrates applied by S9 oligopeptidases is providing a permanent entrance to the active site but shielding it by multimerization (Fig. 1A). Dimeric DPP4 and hexameric PhAAP are examples for this latter. 41 Mammalian AAP seems to apply a hybrid of the two strategies observed so far: it possesses an open monomer, the entrance of which is however narrowed by the inner loops of blades 3-5, and its closing is prohibited by the hydrolase-propeller interactions with the neighboring monomers. Thus, in this case, activation and inactivation of the catalytic triad have been decoupled from the opening and closing of the monomers. Instead, with a Pro inserted into the middle of the fourth strand of the central bsheet, sufficient conformational freedom is awarded to the 584-591 segment containing the catalytic Ser587 that the active site alternates between the active and inactive conformation even in the absence of domain movements. This is truly a twist on the classical serine protease setupto the best of our knowledge, no such liberation of the catalytic triad has been previously detected. It is a necessary addition to the S9 oligopeptidase build-up too, since beside functioning as an oligopeptidaseassisting the degradation of smaller peptidesmammalian AAP also removes terminal N-acetylated amino acids from intact proteins, 17,18 processing a considerably greater variety of substrates than the other family members do. The highly charged clusters of the exible gatekeeper loops, lining the outer pore together with the shielded interior of the tetramer, might add a chaperon-like function 60 to the entrance: enabling it to rst strip the substrate protein segments of their solvent shell and then stabilize the exposed hydrophobic residues, promoting the unfolding of the chain that will allow the substrate to reach the buried and gated active sites.</p><p>The structure of pAAP is a splendidly ne-tuned system: selfassociation guided by the interaction prone b-edge and unique inserts leads to self-compartmentalization that equips the enzyme with its "channels-and-shutters" system guaranteeing that only selected substrates can reach the active site. The Hisloop sequentially follows the hot-spot of self-association, and the segments responsible for shaping the outer pore of the tetramer are directly connected to loops that shape the entrance of the monomeric unitsalso linking self-association and catalysis. Therefore, in the case of mammalian AAP, multimerization is a prerequisite for controlled catalytic function.</p><p>The pAAP structure lends further support to the previous hypothesis that among the closely related enzymes of the S9 protease family, the different modes and extents of homomultimerization equip the "all-purpose" serine protease catalytic machinery embedded in all with unique substrate selection mechanisms.</p><p>The determined structure of pAAP thus not only provides a sufficient model of the human enzyme (Fig. S13 and S14 †) that will allow drug design efforts, but also contributes to our understanding of the signicance and mechanisms of protein multimerization.</p><!><p>The preparation and purication of the mammalian AAP sample (from porcine liver) were based on a previous method (Fig. S15 †) 39 with an additional size exclusion chromatographic step using a Superose 6 30/100 column on an AKTA FPLC system (GE Healthcare, 20 mM TRIS, pH ¼ 8, 0.15 M NaCl, 1 mM EDTA, 1 mM DTT). Tetrameric composition was veried by size exclusion chromatography. To monitor that the catalytically competent form of the enzyme was preserved during the puri-cation process, concentrated samples of pAAP were incubated with N-acetyl-alanine p-nitroanilide (AANA, eNovation Chemicals LLC) as a substrate (1.6 mM in 5% DMF/water) in buffer (50 mM phosphate, pH ¼ 8, 0.3 M NaCl, 1 mM EDTA, 5 mM mercaptoethanol) at 37 C (reaction mixture: 10 ml of AANA solution, 985 ml buffer, 5 ml protein sample). The formation of pnitroaniline was measured spectrophotometrically by monitoring the increase in absorbance at 410 nm.</p><p>The puried protein sample (3 ml) of 6 mg ml À1 concentration in 10 mM TRIS (pH ¼ 7.5) buffer was placed on a Quantifoil R1.2/1.3 grid (GIG, 1.0 mm hole size, 200 mesh) and was vitri-ed. Aer 2 and 4 s of blotting time, the grid (4 C, 90% humidity) was plunge-frozen in liquid ethane (Leica EMGP). Cryo-EM single particle data collection was performed using a CRYO ARM 300 microscope operated at 300 kV equipped with a K3 camera (Gatan). Images were recorded at 80 000-fold magnication corresponding to 0.95 Å per pixel using a 20 eV energy lter (Omega Filter) with an exposure time of 4 s and a total electron dose of 37.5 e ÅÀ2 . The spherical aberration coefficient (C s ) was 2.7 mm and the defocus range was 0.5-2.5 mm. A total of 1157 micrographs were collected from a single grid (ESI -Detailed methods, Fig. S1 and Table S5 †).</p>

Royal Society of Chemistry (RSC)

Anomalous Nanoparticle Surface Di usion in Liquid Cell TEM is Revealed by Deep Learning-Assisted Analysis

The motion of nanoparticles near surfaces is of fundamental importance in physics, biology, and chemistry. Liquid cell transmission electron microscopy (LCTEM) is a promising technique for studying motion of nanoparticles with high spatial resolution. Yet, the lack of understanding of how the electron beam of the microscope affects the particle motion has held back advancement in using LCTEM for in situ single nanoparticle and macromolecule tracking at interfaces. Here, we experimentally studied the motion of a model system of gold nanoparticles dispersed in water and moving adjacent to the silicon nitride membrane of a commercial liquid cell in a broad range of electron beam dose rates. We find that the nanoparticles exhibit anomalous diffusive behavior modulated by the electron beam dose rate. We characterized the anomalous diffusion of nanoparticles in LCTEM using a convolutional deep neural network model and canonical statistical tests. The results demonstrate that the nanoparticle motion is governed by fractional Brownian motion at low dose rates, resembling diffusion in a viscoelastic medium, and continuous time random walk at high dose rates, resembling diffusion on an energy landscape with pinning sites. Both behaviors can be explained by the presence of silanol molecular species on the surface of the silicon nitride membrane and the ionic species in solution formed by radiolysis of water in presence of the electron beam.

anomalous_nanoparticle_surface_di_usion_in_liquid_cell_tem_is_revealed_by_deep_learning-assisted_ana

<!>Results and discussions<!>Models of anomalous diffusion.<!>Nanoscopic interpretation.<!>Materials and Methods<!>Liquid cell preparation.<!>Anomalous Diffusion Models Fractional Brownian Motion. A FBM process x(t) is characterized by the following properties<!>Discrete time FBM.<!>p-Variation Test<!>Waiting Time Distribution<!>MotionNet (MoNet) Architecture, Training, and Inference<!>Performance of MoNet in classification andprediction

<p>U nderstanding the motion of nanoparticles in boundary layers is of fundamental importance in scientific fields such as biophysics and colloidal self-assembly, and of practical importance in technological applications such as drug delivery and additive manufacturing. The physics behind the motion of nanoparticles is particularly challenging to understand due to the multitude of e ects including particle-particle interactions, particle-surface interactions, and changes in the rheological properties in boundary layers close to a liquid-solid interface.</p><p>The common technique to study the motion of particles has been optical microscopy, which has limitations in terms of spatial resolution. The advent of in situ liquid cell transmission electron microscopy (LCTEM) has now made it possible to visualize the motion of nanoparticles near a surface with an unprecedented spatial resolution at the nanometer length scale (1)(2)(3). However, the electron beam of a transmission electron microscope (TEM), which is the key acquisition tool to enable nanoscale visualization, can significantly influence both interactions and dynamics of nanoparticles (4)(5)(6). Previous literature has reported that the motion of nanoparticles near the surface of a liquid cell and in the presence of the electron beam is subdi usive (i.e., non-Brownian, or "anomalous") (7)(8)(9)(10)(11)(12)(13)(14)(15). Such subdi usive motion suggests that the nanoparticle motion is significantly influenced by interactions with the nearby substrate or interface, but what precisely is the nature of these interactions and the forces that create them? Are they stable or fluctuating? Do they arise because of the electron beam or are they native to the system? How do the changes in rheology within a few nanometers of the interface figure into the picture? The nature of the observed anomalies are still very much under debate as the new technique of LCTEM continues to be developed (7)(8)(9)(10)(11)(12)(13)(14)(15).</p><p>Two canonical processes that describe anomalous motion are continuous time random walk (CTRW) and fractional Brownian motion (FBM) (16)(17)(18)(19). In the context of particle di usion, each of these types of subdi usive motions imply a distinct physical picture of the environment. CTRW indicates a random energy landscape of potential wells, where the time a particle spends in any well diverges when averaged over all well depths. FBM, on the other hand, indicates a viscoelastic environment such as those found in crowded fluids (20)(21)(22)(23). The goal of this work is to identify the type of anomalous motion of nanoparticles near the surface in LCTEM, elucidate the nanoscopic physical features in the system that give rise to this motion, and understand how the electron beam can influence them.</p><p>A key challenge in studying the motion of nanoparticles under the e ect of the electron beam is that one needs to resort to a limited number of short trajectories from a single in situ LCTEM experiment. This is because achieving high spatial resolution requires a relatively small field of view, which limits the number of nanoparticles accessible (experiments are done in dilute solutions to avoid interactions between nanoparticles). Moreover, state-of-the-art cameras on TEM microscopes are limited by lower bounds on time resolution (hundreds of frames per second) and upper bounds on measurement time (minutes long trajectories) (24). This limitation creates a challenge for canonical methods used to characterize di usive particle dynamics such as the mean-squared displacement (MSD) analysis. These methods often rely on features of the trajectory that converge upon averaging over very long single-particle trajectories (for systems obeying ergodicity) or hundreds of medium-length trajectories collected under the same experimental conditions (19,25,26). Here, we show that computational and theoretical tools can be developed to extract hidden features that exist in short trajectories of single nanoparticles in order to elucidate the type of anomalous di usion.</p><p>In this study, we collected 30 trajectories of a model system of gold nanoparticles dispersed in water and di using near a silicon nitride (SiNx) membrane of a commercial liquid cell irradiated by a broad range of electron beam dose rates. Inspired by the recent advances in using machine learning tools to study the di usion of single microparticles in biological media (25)(26)(27)(28), we developed a convolutional deep neural network (CNN) model, dubbed MotionNet (MoNet), which solves an inverse problem of determining the underlying di usion mechanism behind the anomalous motion of nanoparticles in LCTEM. The architecture of the neural network employed in MoNet is designed based on classical tests in statistics (29) and is trained on thousands of simulated short trajectories from three classes of di usion, i.e. Brownian, FBM, and CTRW. Guided by MoNet, our analysis reveals that at low dose rates the anomalous di usive motion of nanoparticles in LCTEM is governed by viscoelasticity-dominated FBM, while at high dose rates the motion is governed by a pinning site-mediated CTRW process (23). The prediction results were benchmarked against the statistical p-variation test (29) to confirm the behavior in low and high dose rate limits.</p><p>The dose rate-dependent transition can be explained by the existence of silanol molecular groups on the surface of the SiNx membrane, which act as pinning sites and exhibit a broad distribution of restoring forces (14,30). At low dose rates, the binding strength of these pinning sites is high compared to the thermal energy and their e ective restoring force acts similar to the e ect of a viscoelastic environment. This results in nanoparticle motion confined to the local vicinity of a pinning site. Upon increasing the dose rate and thus passivating the charges on the pinning sites, the binding strength decreases, making nanoparticles more mobile, which allows them to diffuse across the SiNx membrane only making intermittent stops on randomly distributed pinning sites. This understanding provides us with important insight into the mechanism of nanoparticle motion near a substrate in LCTEM and opens up the path to use in situ LCTEM as a technique for studying motions of nanoparticles in complex systems at the nanoscale.</p><!><p>Anomalous diffusion of gold nanorods. To study the e ect of electron beam dose rate on the motion of nanoparticles near a surface, we chose a simple and tunable model system of 60 nm long gold nanorods (AuNRs) dispersed in water and probed their dynamics near the SiNx membrane of a commercial TEM's liquid cell (see Methods for details of synthesis). We collected trajectories of 30 AuNRs for electron beam dose rates ranging from 2 to 49 e ≠ /Å 2 s (see Figure S1 for all trajectories collected). Figure 1a shows the first 30 seconds of five representative trajectories. Plotted at the same scale, these trajectories indicate that the e ective di usivity of AuNRs increases with increasing dose rate of the electron beam. Figure 1b shows the same trajectories as Figure 1a in their entirety, magnified to reveal details of the dynamics. In addition to the increase in e ective di usivity, a qualitative change in the dynamics is observed as the dose rate is increased from 2 to 49 e ≠ /Å 2 s. At low dose rates, AuNR dynamics are dominated by motion confined to the vicinity of a local point. This motion is punctuated by infrequent, relatively long-distance jumps. At high dose rates, long-distance jumps between short periods of confinement become the dominant behavior at observation timescales.</p><p>To identify the underlying di usive behavior and to understand how the electron beam changes the local environment and the local interactions, we first calculated the MSD. In the framework of anomalous di usion, the MSD is described by the power law (19):</p><p>Here, brackets denote an ensemble average, and accordingly we refer to this as the ensemble averaged or e-MSD. If -= 1 the process is characterized by Brownian motion, and if -< 1 or -> 1 the process is subdi usive or superdi usive, respectively. The MSD may also be computed by window averaging over a single trajectory, which we refer to as a time averaged or t-MSD, and is defined by</p><p>Here, T is the total measurement time, is the time delay window, and ( • ) indicates an average over time. For ergodic processes, the e-MSD and t-MSD are equal in the long time limit as T ae OE. In case of non-ergodic subdi usive processes, the e-MSD contains more information about the underlying anomaly mechanism; however, it is not practically accessible in many experimental systems, including LCTEM as it is available today. The t-MSD measurements as a function of for all thirty trajectories are presented in Figure S2. We also measured the related time-averaged di usion constant D-= "x 2 ( )/ -, using Eq. [2], which varies between 10 and 10 4 nm 2 /s for values of AE 0.25 s and for all dose rates studied across 30 trajectories in three experiments; see Figure S3. This shows that the motion of AuNRs near the surface is orders of magnitude slower than what is theoretically estimated for a Brownian nanoparticle in bulk water outside of TEM based on the Stokes-Einstein relation (D = kBT /(6fi÷L) ¥ 4 ◊ 10 6 nm 2 /s with ÷ the viscosity of the medium and L the characteristic size of the di using nanoparticle). The slow motion of AuNRs observed here is consistent with previous reports on the suppressed di usive motion of nanoparticles in LCTEM experiments (8,9,11,13,14). Figures S2 and S3 also show that D-increases as the dose rate is increased, consistent with the observations from Figure 1a. However, it is not possible to identify the type of di usion as well as whether it is anomalous or not based solely on the t-MSD curves. This can be explained by a closer look at two common anomalous di usion models, subdi usive CTRW and subdi usive FBM, and their corresponding MSDs.</p><!><p>Di usion processes in which particles move with stop-and-go motion on an energy landscape with heterogeneous pinning sites are well described as a continuous time random walk (CTRW) (19). In a CTRW process, a particle moves by making random jumps in space and time (see SI Appendix for details). The particle remains immobile for a random "waiting time" • , drawn from distribution Â(• ), before jumping in the distance and direction x, drawn from the distribution ⁄( x) (22). If Â(• ) is heavy-tailed, i.e., the asymptotic behavior at large • decays as Â(• ) ≥ 1/• 1+with 0 < -< 1, the mean waiting time È• Í diverges (È• Í ae OE) and the resulting process is subdi usive (21). The diverging È• Í also indicates that ergodicity is broken; no matter how long the measurement time T is, the t-MSD and the ensemble averaged t-MSD (average of t-MSDs over an ensemble of particles, or et-MSD) will not be the same (21). It can be shown that the et-MSD for a CTRW process can be written as (see SI Appendix for mathematical derivation) (31,32) È"x 2 ( )Í ≥ D-T 1≠-.</p><p>[3]</p><p>Eq. [3] shows that for a CTRW process, the et-MSD is a linear function of time delay, . The e-MSD is obtained from the et-MSD in the limit ae T , recovering the anomalous form of Eq. [1]. This property of the subdi usive CTRW process makes it extremely di cult to identify and to estimate its inherentvalue, when only a limited number of short trajectories from an experiment is accessible since no anomaly can be detected by measuring the t-MSD.</p><p>Another canonical model of subdi usion is Fractional Brownian Motion (FBM) (17). Subdi usive FBM can be qualitatively described as a random process in which the direction of each step is anti-correlated with the previous step, resulting in the next step having a higher probability than random to be in the opposite direction (30). This correlation of positions at two di erent points in time, t1 and t2 along the trajectory can be expressed as:</p><p>where as before, -< 1 corresponds to subdi usion. Unlike the CTRW model, a FBM process is ergodic, and thus t-MSD and its ensemble average, i.e. et-MSD, are the same and follow È"x 2 ( )Í = D--(see SI Appendix for mathematical derivation of a FBM process).</p><p>Deep learning analysis. To identify the underlying anomalous di usion process for a limited set of short trajectories in a LCTEM experiment, we developed a convolutional neural network model which we have named MoNet (shown in Figure 2a). We trained MoNet on 10, 000 simulated trajectories from three classes of di usion: Brownian, subdi usive FBM, and subdi usive CTRW. Each simulated trajectory was 300 frames in length; short enough to cover the shortest experimental trajectories collected and long enough to achieve more than 90% validation accuracy (see Figure S6). For consistency, the model was then applied to 300-frame intervals of all nanoparticle trajectories. The final results are reported as the predicted probability for each di usion class, averaged over the entire length of the trajectory (see Figure S6 for validation accuracy on an independent set of test data). As shown in Figure 2a, MoNet receives input data in the form of a matrix comprising the x and y coordinate of the locations of the nanoparticle throughout the trajectory and outputs the probability of the predicted di usion class, i.e. FBM, CTRW, and Brownian. The architecture of MoNet is inspired by previous literature for temporal sequence type data such as particle trajectories (26,33). See the Methods section and SI Appendix for details of the architecture of MoNet. Figure 2b presents the predicted probability of the di usion class for all 30 trajectories as a function of dose rate, increasing from top to bottom in the first column and left to right across the table; see Figure S7 for the probability values associated with each class. Interestingly, there is a crossover from FBM to CTRW as the electron beam dose rate is increased, consistent with the qualitative picture of Figure 1b. The crossover occurs around the dose rate of 15 e ≠ /Å 2 s, where 7 trajectories have been collected. analysis results for all trajectories studied as a function of dose rate (increasing from top to bottom in the first column and left to right across the table). Pie charts show the diffusion class probability where at low dose rates, there is a higher probability associated with a fractional Brownian motion (FBM) (green) and at high dose rates there is a higher probability associated with a continuous time random walk (CTRW) (blue).</p><p>To verify the results we compared MoNet against a statistical method, known as the p-variation test, V p n (see the Methods section for definition). p-variation has been successful in distinguishing FBM from CTRW for medium length trajectories (29). Here, we have analyzed the quadratic variation (p-variation for p = 2), which measures the sum of squares of increments of a trajectory of length T = 2 Nmax , divided into 2 n segments. For a FBM process, the quadratic variation must diverge as n ae OE (i.e., the size of time increment t ae 1 frame), while for a CTRW process, the quadratic variation must stabilize with increasing n (29,34). Comparison of our predictions with the quadratic variation results presented in Figure S8 confirms that there is indeed a crossover from FBM to CTRW while increasing the dose rate. Figure 3 shows the quadratic variation results for two example trajectories of Figure 1b at dose rates 15 and 49 e ≠ /Å 2 s. The unbounded increase in the slope of the quadratic variation vs. measurement time curve as n ae Nmax (i.e., t ae 1 frame) confirms that at dose rate 15 e ≠ /Å 2 s, the trajectory is predominantly characterized by a FBM behavior. However, for a higher dose rate of 49 e ≠ /Å 2 s, the quadratic variation curve does not show any specific dependence as n ae Nmax, suggesting that the anomaly does not stem from a FBM process.</p><p>Another characteristic of FBM and CTRW processes in terms of displacement, "x, is their probability distribution of displacements P ("x) (19,30). Comparison of the distribution of displacements collected over time delays of 0.0125 s in Figure 4a for two example trajectories at dose rates 15 and 49 e ≠ /Å 2 s, (same trajectories as Figure 3), also confirms the presence of a FBM process at low dose rates with a Gaussian distribution and a CTRW process at high dose rates with a power-law tailed distribution. The power-law exponent of this tail is estimated to be about ≠2.0 (see Figure 4a). Figure S9 shows that this power-law value of ≠2.0 is consistent for all high-dose rate trajectories studied here.</p><p>The power-law decay of the probability distribution P ("x) for large values of displacement, "x, at high dose rates does not necessarily mean that the underlying CTRW process is subdi suive (35). However, it suggests that there is a broad distribution of binding sites on the surface of the SiNx membrane. It is known that for harmonic energy potentials with equal binding sti ness k, the resulting probability distribution of displacements must follow a Gaussian form, P ("x) = exp(≠k("x) 2 /kBT ) (30,36). Hence, the non-Gaussian and heavy-tailed probability distribution of displacements observed for all high-dose rate trajectories indicates that binding sites with various binding a nities exist over the surface of the SiNx membrane, suggesting that an underlying CTRW process could be subdi usive. To confirm that the CTRW process observed at high dose rates is subdi usive, we used MoNet trained on three thousand simulated CTRW trajectories withvalues between 0.1 and 0.99, and predicted theexponent for all trajectories collected (see Figure S10). The results show that the underlying mechanism at high dose rates is subdi usive withexponents ranging from 0.7 ≠ 0.8.</p><p>We also did a similar analysis using MoNet to predictexponents of FBM processes (commonly known as the Hurst exponent H = -/2 in the literature (37)); see SI Appendix and Figure S10. Theexponent obtained from this analysis is very similar to the values ofextracted from t-MSD curves (see Figure S10). Figure 4b shows the t-MSD curves calculated for trajectories of dose rates 15 and 49 e ≠ /Å 2 s withexponents of 0.48 and 1, respectively. As shown in Eq. [3], the t-MSD curve of a CTRW process grows linearly in time delay , consistent with our measurements shown in Figure 4b. For subdi usive FBM processes, the t-MSD curves grows sublinearly in time delay . Therefore, the t-MSD curve can only provide us with a value ofat low dose rates, where the process is predominantly characterized by a FBM model and et-MSD measurements are further required to estimate the value of for CTRW processes. Using MoNet predictions for theexponent for both low and high dose rate trajectories, we showed that at all dose rate studied the underlying di usive process is subdi usive.</p><!><p>The physical picture governing the di erent di usive behavior at low and high electron beam dose rates may be explained by the molecular groups existing on the surface of the SiNx membrane of the TEM liquid cell, which in turn are influenced by the electron beam. It has been previously suggested that the surface of the SiNx membrane is decorated with silanol Si≠O molecular groups (14). These silanol groups that are randomly distributed across the membrane create pinning sites and can locally trap nanoparticles, which are positively charged with cetyltrimethylammonium chloride (CTAC) ligands. At low electron beam dose rates (AE 15 e ≠ /Å 2 s), the thermal energy of AuNRs is smaller than the binding strength of these pinning sites resulting in particles being trapped for significant periods in the vicinity of a local pinning site, with membrane restoring forces and solvent interactions acting as a viscoelastic medium. This viscoelastic picture may be explained by the hydrogen bonding of the water molecules with the Si≠O species on the surface of the SiNx that may result in a gel-like viscoelastic water layer next to the membrane at low dose rates, leading to the FBM behavior at this dose rate. We note that a close inspection of the trajectories of Figure 1b and Figure S1 shows that for some of our low dose rate experiments, the immobility in pinning sites is punctuated by a few relatively long-distanced jumps. Yet, these jumps are smaller than 50 nm which is smaller than the body length of the AuNRs studied (≥ 60 nm) and could be explained as the head or tail of the same AuNR being trapped in the same pinning site. Upon increasing the dose rate, radiolysis of water occurs, changing the local pH value of the solution close to the membrane. This change in pH results in free H + ions in solution that passivate the Si≠O groups on the membrane's surface, reducing their binding strength. Therefore, at high dose rates, AuNRs can occasionally de-trap and move with long distance jumps until they get trapped in another pinning site associated with a waiting time • drawn from a heavy tailed distribution function Â(• ).</p><p>While this interpretation could explain this set of observations, we note that alternative scenarios may exist such as coexistence of both FBM and CTRW behavior. A close look at the predictions of Figure 2b shows that for certain low dose rates (see 15 e ≠ /Å 2 s) classified as FBM by MoNet, there is a non-negligible probability associated with the CTRW class. By tracking the predicted di usion class along the entire length of the trajectory for each window of 300 frames separately, we can observe that windows including long distance jumps of > 50 nm are more likely to be classified as CTRW (see Figure S12).</p><p>The presence of these jumps, even at low dose rates, also shows up in the t-MSD curves. The t-MSD curve for the dose rate of 15 e ≠ /Å 2 s presented in Fig. 4b has anexponent of 0.48 at short time delays, while at long time delays the exponent increases to 1. This is in contrast to high dose rates, where throughout the t-MSD curve, theexponent remains constant at a value of 1. The change in theexponent as well as the non-negligible probability associated with the CTRW class at dose rates of 15 e ≠ /Å 2 s suggests that both FBM and CTRW behavior could potentially coexist at this dose rate but at di erent timescales. Furthermore, the scatter in t-MSD curves for all dose rates in Figure S2 shows that that ergodicity might be broken even at low dose rates where the di usion class is predominantly characterized by FBM, which is by definition an ergodic process. This is reminiscent of the subordinated di usion processes reported in biological systems as well as single molecule tracking experiments in water (30,(38)(39)(40). This type of subordinated di usion is complex to capture through canonical methods and indeed requires data spanning multiple timescales both on short and long time delays. Therefore, while the current data is insu cient to support or nullify this hypothesis (especially at high dose rates), our analysis suggests the possibility of such a scenario. Regardless, the presence of predominantly FBM behavior at low dose rates and CTRW at high dose rates supports the interpretation that the di usive motion at low dose rates is mostly influenced by the local viscoelasticity of the fluid next to the surface and at high dose rates the motion is governed by the heterogeneous pinning sites in the timescales studied ( = 0.01 s to 100 s). Therefore, the electron beam dose rate not only increases the di usion coe cient, but also it fundamentally alters the fluid and the dominant di usive behavior of nanoparticles near the membrane.</p><p>This understanding of how the electron beam can a ect the local environment near the membrane, which in turn governs the di usive motion of nanoparticles near the surface, can be used in applications of nanoparticles in LCTEM as nanoscale probes to study the local material properties of the fluid near the surface. Similar analysis can also be performed on undamped motion of nanoparticles in bulk in LCTEM, (41,42) to investigate the e ect of electron beam on the bulk material properties. Furthermore, the change in the local material properties of the fluid next to the surface in presence of the electron beam may play a role in other LCTEM studies such as in situ growth of nanocrystals (43)(44)(45). The knowledge base developed here can be also extended to study the motion of nanoparticles in LCTEM near surfaces with various combinations of nanoparticles, fluids, and surfaces with high spatial resolution.</p><!><p>Chemicals and materials. Hexadecyltrimethylammonium bromide (CTAB, > 98.0%), hexadecyltrimethylammonium chloride (CTAC, > 95.0%) and sodium oleate (NaOL, > 97.0%) were purchased from TCI America. Acetone (99.5%) was purchased from Fisher Scientific. Hydrogen tetrachloroaurate trihydrate (HAuCl 4 •3H 2 O, Ø 99.9%), L-ascorbic acid (BioXtra, Ø 99.0%), silver nitrate (AgNO 3 , Ø 99.0%), sodium borohydride (NaBH 4 , 99.99%), and hydrochloric acid (36.5%≠38.0% wt.%) were obtained from Sigma Aldrich (USA). NaBH 4 powder was stored in an argon glovebox. HAuCl 4 • 3H 2 O, L-ascorbic acid, and AgNO 3 were stored in a vacuum desiccator at room temperature. Deionized water (DI-water, Milipore, Milford, MA, USA) was used for all aqueous solutions. All the glassware was thoroughly cleaned using freshly prepared aqua regia (3 : 1 volume ratio of HCl and HNO 3 , respectively) followed by fully rinsing with copious amounts of DI-water. All chemicals were of reagent grade and used without further purification.</p><!><p>Commercially available silicon nitride liquid cell top (EPT≠52W≠10) and bottom (EPB≠52DNS) microchips (Protochips Inc.) with electron transparent membranes and a 150 nm static spacer, were cleaned by being immersed in a clean petri dish filled with acetone to remove the protective resist coating and immediately transferred to a second petri dish filled with high purity ethanol. The microchips were then dried by blotting them on a filter paper to remove the excess ethanol. The microchips were fully dried by blowing gently nitrogen gas parallel to their surface. Following that they were plasma-treated for 3 minutes to remove any residual organic material and to improve their surface hydrophilicity. The microchips were then assembled in a Poseidon 200 holder according to the Protochips Inc. protocols with 0.75 µL of the AuNR solution containing an extra 5 mM of CTAC ligands.</p><p>TEM imaging. in situ experiments were performed on a FEI Tecnai T20 S-TWIN TEM operating at 200 KV with a LaB6 filament. Time series of images were collected using a Gatan Rio 16 IS camera in Digital Micrograph format at nominal magnifications of 25.3 kx and 38.1 kx with various exposures of 0.1, 0.05, 0.0125, and 0.00625 seconds corresponding to frame rates of 10, 20, 80, and 160 frames per seconds with 4096 ◊ 4096, 2048 ◊ 2048, and 1024 ◊ 1024 pixels by pixels readout, resulting in 0.355208, 0.710415, and 1.42083 nm/pixels resolutions, respectively. Prior to imaging, the electron beam dose rate was calibrated at each magnification using a custom digital micrograph script as described in the previous literature by converting counts to electrons with a conversion value of 124 (46). The range of dose rates accessible after calibration at this magnification spans from 2 to 49 e ≠ /Å 2 s. Data were collected in three sets of experiments using the same dose rates to assure the consistency of the outcomes. Furthermore, dose rates were increased and decreased to verify the reversibility of the process. Time series of high dimensional images were processed in MATLAB using custom scripts to obtain trajectories of nanoparticles presented in Figure S1 by tracking the centroid of AuNRs in each frame.</p><p>Synthesis of gold nanorods. Homogenous AuNRs were synthesized by a facile seed-mediated growth involving a binary surfactant mixture (47). The seed solution was prepared as follows: 10 mL of 0.1 M CTAB solution was mixed with 100 µL of 25 mM HAuCl 4 in a 20 mL scintillation vial under vigorous stirring. 600 µL of ice cooled 10 mM NaBH 4 was rapidly injected into the Au-CTAB solution and stirred for 2 minutes. Upon the addition of NaBH 4 , the color of the seed solution turned yellow-brownish. Afterward, the seed solution was left undisturbed at 28 ¶ C for 30 minutes prior to use in the following step.</p><p>The growth solution was obtained by first mixing 3.6 g of CTAB and 0.4936 g of NaOL in 196 mL of DI-water in a 500 mL Erlenmeyer flask. The solution was heated with occasional agitation until all the CTAB was dissolved. The mixture was allowed to cool down to 30 ¶ C and 1.45 mL of 10 mM AgNO 3 was then added under stir at 700 rpm for 15 min. Afterward, 4 mL of 25 mM HAuCl 4 was added to the mixture and kept undisturbed at 28 ¶ C for 90 min. The yellowish color of growth solution turned to colorless. 840 µL of HCl was added to the solution and the mixture was stirred at 400 rpm for 15 min. Finally, 500 µL of 0.064 M ascorbic acid was injected into the growth solution, and the mixture was vigorously stirred at 1200 rpm for 30 s. 80 µL of the seed solution was then injected, and the solution was stirred for 30 s before left undisturbed at 28 ¶ C for 12 hr to complete the growth process. 40 mL of the final products were isolated by centrifugation at 8, 000 rpm for 15 min followed by careful removal of the supernatant. 50 mL of DI-water was added to the pellet and the mixture was sonicated briefly to disperse the pellet for long-term storage. For the sample preparation of the liquid cell experiment, a second centrifugation step was performed at 5, 500 rpm for 10 min followed by removal of the supernatant and adding 50 mL of DI-water. 1 mL of the stock solution was centrifuged at 5, 500 rpm for 8 min and the supernatant was carefully removed. 1 mL of 50 mM CTAC solution was added and sonicated for 10 min. The solution was centrifuged again at 5, 500 rpm for 8 min followed by removal of the supernatant and adding 1 mL of the DI-water.</p><p>Deep learning. MoNet architecture consists of 6 convolution layers (including 5 dilated convolution layers) followed by 3 dense layers. The dilated layers have 32 filters of sizes k = 2, 3, 4, 10, and 20 with a combination of dilation factors of 2 n for n = 0, 1, 2, and 3 (inspired by p-variation method) to capture long distance correlations existing in increment of 2 n along the trajectory. See Figures S4 and S5 for the schematic of the neural net architecture. The validation accuracy of MoNet has been tested on simulated trajectories of di erent length (Figure S6). For a 300-frame long trajectory the prediction accuracy of the di usion class is 90%. The mean squared error associated with the task ofprediction in CTRW and FBM models are 0.02 and 0.003, respectively. See SI Appendix for more details.</p><p>p-variation test. To distinguish between subdi usive FBM and CTRW dynamics, Magdiarz et al. proposed the p-variation test (29,34). This test generalizes the concept of the total variation V , in which the increments (i.e., particle displacements) are summed over the entire trajectory. The p-variation V (p) n (t) generalizes the concept of total variation by exponentiating each increment by p before summing (48)</p><p>Given a trajectory with a length of 2 N , in case of p = 2 (quadratic variation) V 2 n (t), we sum up the square of the increments which are spaced 2 N ≠n in time. See SI Appendix and Figure S4a for more details.</p><!><p>• is a zero mean process Èx(t)Í = 0.</p><p>• starts at x(0) = 0.</p><p>• has stationary increments</p><p>' , [1] where d = denotes equality in distribution. A consequence is that the expectation of any function f of an increment is invariant to time translation of that increment; that is,</p><p>Together with the previous property, this implies</p><p>an identity which will be used shortly.</p><p>• has the probability density function (PDF) of the form ( 1)</p><p>with Èx(t)Í = 0, and Èx(t</p><p>Here, H is known as the Hurst exponent that is related to the anomalous di usion exponentas H = -/2. If 0 < H < 1/2 the process is subdi usive, if H = 1/2 the process is fully Brownian, and if 1/2 < H < 1 the process is super-di usive. The second moment or the ensemble-averaged mean-squared displacement (e-MSD) of the FBM process is then</p><p>With this definition and using a binomial expansion and using stationarity and zero mean properties of the last term on the second line (Èx 2 (t1) ≠ x 2 (t2)Í = Èx 2 (t1 ≠ t2)Í = 2DH (t1 ≠ t2) 2H ), and finally using Eq. ( 5) for each term, the FBM process x(t) has a covariance of the form ( 2)</p><p>It can be concluded from the covariance of equation Eq. ( 6) that the FBM process is self-similar</p><p>Note that H is also known as the self-similarity parameter. The FBM process (of which ordinary Brownian motion can be considered a subset with H = 1/2) is the only Gaussian process that is both self-similar and stationary. The time evolution of x(t) can be assumed to have the general form</p><p>where ›(t Õ ) is called fractional Gaussian noise. Equivalently, in di erential form</p><p>This implies that the time correlation of the fractional Gaussian noise can be obtained by di erentiating equation Eq. ( 6) with respect to each of the time variables</p><p>where "(x) is the Dirac delta function and sgn(x) is the sign function. Here we have used the properties</p><p>Finally, note that in the final line of equation Eq. ( 6) we have assumed that 2H " = 1 to eliminate the second term in parentheses of the previous line. If instead we examine the case 2H = 1, only the second term remains, leaving [14] which is the expected delta-correlated noise characterizing Brownian motion.</p><!><p>Here we reconsider the above analysis from the perspective of a discretized time variable, as will be made use of in the following section. From equation Eq. ( 6) the covariance for discrete-time increments of</p><p>Hence, assuming that particle is at x = 0 at time zero, the covariance of increments</p><p>). [16] Note that for H = 1/2 (i.e., Brownian case) these increments are non correlated and the process is not self-similar as we expect. The increments of the FBM process are also called fractional Gaussian noise ›, where ›(k + 1) = x k+1 ≠ x k defined on increment of ˆt = 1 ( ˆx ˆt = ›(t)). Therefore, Eq. ( 16) is indeed the covariance of the fractional Gaussian noise È›(k + 1)›(1)Í. One can rewrite Eq. ( 16) by factoring the k 2H term</p><p>where</p><p>Using the Taylor expansion of f (x) at the origin (x = 1/k ae 0) the covariance of the fractional Gaussian noise is</p><p>with k oe {0, ..., N ≠ 1} and "(0) = 1. This is true only if t1 " = t2 (4).</p><p>Simulating a FBM process. A FBM process can be simulated using a circulant matrix embedding algorithm and using fractional Gaussian noise › = (›1, ›2, ..., ›N ) T and its covariance matrix:</p><p>.</p><p>In order to simulate a FBM process x(t), we need to find the square root of the matrix. Finding square roots of this matrix is hard. Hence, a more convenient method often used is to embed this matrix in a larger circulant matrix called C of size 2M ◊ 2M with M = 2N : 2) "( 1) "( 1</p><p>, where the red box indicates the matrix. Since the matrix C is circulant, it can be decomposed into C = F F ú using Fourier transform, where F is a unitary matrix and is a diagonal matrix of eigenvalues of matrix C. F ú denotes the conjugate transpose of F and FF ú = I. Therefore, FCF ú = . We can generate the matrix using the eigenvalues (i.e., FFT coe cients of C) (4):</p><p>with Cj the (j + 1)th elements of the first row if C matrix, i = Ô ≠1, and F defined as:</p><p>To find the square roots of matrix C, we can write C = SS ú with S = F 1/2 F ú and 1/2 = diag(⁄</p><p>. The last step to simulate a FBM process is to multiply matrix S with a vector V with i.i.d. standard normal elements and take the first N elements corresponding to the fractional Gaussian noise vector ›.</p><p>Continuous Time Random Walk. A continuous time random walk (CTRW) process is a class of anomalous di usion with a combination of random walks in space and time. Consider a test particle di using with a CTRW behavior where x(t) denotes the position of the particle at time t. The particle will make a random jump of distance xi = x(ti) ≠ x(ti≠1) after a waiting time of •i = ti ≠ ti≠1 in its previous site. After the jump, the process is renewed. For a CTRW process, we assume (6)</p><p>• The spatial step length xi, i = 1, 2, • • • are i.i.d. random variables drawn from the PDF ⁄( x)</p><p>Therefore, the joint probability distribution function Ï( x, • ) (known as the jump PDF) can be written as (6), where the distribution of the spatial jump and waiting times are (7):</p><p>We will now focus only on the subdi usive CTRW process which is more relevant to the anomalous di usion of gold nanorods in the liquid cell environment. For a subdi usive CTRW process, the waiting times • are drawn from a heavy-tailed power-law distribution with the asymptotic behavior lim</p><p>Here, •0 is a scaling factor with the dimension of time. The average waiting time in the subdi usive case (-< 1) diverges; that is</p><p>The power-law distributed waiting times can be thought of as a physical picture where tracer particles are continually caught in potential wells with various depths (8,9). The spatial step lengths are assumed here only to have zero mean and finite variance.</p><p>As mentioned in the main text, for a Brownian particle, the ensemble-averaged MSD (e-MSD) Èx 2 (t)Í grows linearly in time. However, for a subdi usive CTRW process of total duration T , the e-MSD is (10)</p><p>To obtain this form, we begin by considering the ensemble average of time averaged of MSD (et-MSD), È"x 2 Í, over an ensemble of independent trajectories of duration T È"x 2 ( ; T )Í = 1</p><p>The integrand can be expressed in terms of the variance of the jump length È"x 2 Í and the average number of jumps n(t, t + ) in the time span of (t, t + ) as (10,11):</p><p>For a subdi usive CTRW process, the average number of jumps for a specified time interval corresponds to a fractional Poisson process with Èn(0, t)Í ≥ t -. Therefore,</p><p>) .</p><p>[27]</p><p>In the limit π T :</p><p>which shows a linear dependence on time delay despite the nonlinear anomalous di usive behavior with the measurement time T . The e-MSD of eq. Eq. ( 24) corresponds to the limit ae T . The fact that the measurement time T shows up in the eq. ( 28) shows the aging behavior of the subdi usive CTRW process. This suggests that as the CTRW process goes on in time, the t-MSD becomes smaller, meaning that it is more likely that longer trapping times would happen, which stalls the progress of x(t) (1). Moreover, we observe a drastic di erence between the subdi usive CTRW and Brownian motion: that the t-MSD "x 2 ( ; T ) and e-MSD Èx 2 (T )Í do not converge towards agreement even in the limit of infinite sampling, a condition known as weak ergodicity breaking Èx 2 (T )Í " = lim T aeOE "x 2 ( ; T ).</p><p>[29]</p><p>This ergodicity-breaking nature of the CTRW process results in scatter in t-MSD " 2 ( ) vs. time delay curves.</p><!><p>The p-variation test introduced in the Methods section generalizes the concept of the total variation V , in which the increments (i.e., particle displacements) are summed over the entire trajectory</p><p>[30]</p><p>Here we have expressed the total variation as a functional of the trajectory x(t) and rescaled the duration of this trajectory to the interval t oe [0, 1]. The total variation V [x(t)] measures the total length of the path traced out by x(t). It is defined in the limit of n ae OE. In this limit, the total variation of Brownian motion is infinite as will be shown momentarily. This is a simple example of the "coastline paradox" described by Benoit Mandelbrot in the context of self-similarity and fractal dimension: the total length of a continent's coastline depends on the size of the ruler used to measure it and, in principle, can be infinite for an infinitesimal ruler (12). The p-variation V (p) n (t) generalizes the concept of total variation by exponentiating each increment (see Fig. S4 for the increments size at each n) by p before summing (13)</p><p>. [31] Note that V (p) n (t) is defined for finite n and on any interval of the trajectory [0, t]. We consider now the p-variation of fractional Brownian motion (FBM). The variance of FBM in Eq. ( 5) can be rewritten as</p><p>where the symbol "≥" indicates expectational proportionality of the Euclidean norm. Together with the the stationarity property of FBM, this allows equation Eq. ( 31) to be evaluated to</p><p>Thus, in the limit of n ae OE the p-variation falls into three regimes depending on the choice of p.</p><p>. [34] Earlier it was stated that the total variation of Brownian motion is infinite. This can be identified with the first case above, in which H = 1/2 for Brownian motion and p = 1 for the total variation. We can also see that the quadratic variation (p = 2) of Brownian motion is finite and proportional to t; that is, limnaeOE V</p><p>(2)</p><!><p>Subdi usive behavior in the context of a CTRW process arises as a consequence of a heavy-tailed waiting time distribution, characterized by the asymptotic behavior described in Equation Eq. ( 23). In Fig. S11 we have plotted the distribution of waiting times for one of the trajectories at a dose rate of 49 e ≠ /Å 2 s, counting the time required for displacements larger than a radial threshold, where displacements below this radius are considered immobile. This figure shows the waiting time distributions for radial thresholds of 20 and 100 nm. The choice of the cut-o radius has a significant e ect in the power-law exponent of the waiting time distribution. For small values (AE 20 nm) the distribution has a power law tail of ≥ ≠2.0, suggesting that the ae 1 corresponding to a Brownian case. However, as discussed in the text, displacements smaller than the length of the nanorods (AE 60 nm) could also mean that the nanorods got trapped with the head or tail on the same pinning site. Therefore, we have also plotted the distribution for a radial threshold value of 100 nm. However, the number of data points are insu cient to draw any firm conclusions, but the apparent asymptotic behavior in the 100 nm case may be an indication of subdi usive behavior.</p><!><p>Input. For di usion classification, MoNet is trained on 10, 000 simulated trajectories from three classes of Brownian, subdi usive FBM, and subdi usive CTRW. The steps on how to sample trajectories from these processes have been discussed in the previous sections. For FBM, the range ofconsidered was 0.2 AE -AE 0.96. For CTRW the range ofconsidered was 0.1 AE -AE 0.99.</p><p>Forprediction for both FBM and CTRW processes, MoNet is trained on 3, 000 simulated trajectories for each task.</p><p>For the task of classifying the trajectories into their di usion class and predicting theexponent for CTRW process, the input to MoNet is the vector of discrete-time increments of the simulated trajectories. Given a batch of N simulated trajectories {x1, x2,</p><p>, the vector of discrete-time increments is defined as</p><p>). It has been reported previously that for theprediction task (in case of FBM processes) learning the velocity autocorrelation of a trajectory is more e ective that the trajectory increments (14). Hence, we followed the same procedure and used the velocity autocorrelation of the discrete-time increments vector as the input for the MoNet with autocorrelation defined as dx i ú dx i T , where ú denotes convolution and dx T i is the transpose of vector dxi.</p><p>Architecture. Fig. S4 and S5 show the architecture of MoNet, adapted from Granik et. al. (14) and modified based on the p-variation method introduced in the previous section. We use the same architecture universally regardless of the task (regression/classification). The architecture of MoNet comprises of 4 layers where the first layer consists of 6 convolutional sublayers (f11, f12, f13, f14, f15, f16) that are applied on the input data in parallel. The first 5 convolutuonal sublayers are three layers deep with relu activation units (relu(•) = max(•, 0) for rectification of the feature map), batch normalization (normalizing the responses across features map), and max pooling (finding the maximum over a local neighborhood). The number of filters applied in all of these sublayers are set to 32. After training, each of these filters capture a certain distinct pattern along the trajectory (e.g., descending, ascending patterns). The diversity among the filters are typically ensured via random initialization of the filters and regularization techniques such as batch normalization and drop out. The filter sizes are k = 3, 4, 2, 10, and 20 respectively for the five convolutional sublayers to capture the local dynamics of trajectories in several spacial resolutions. The convolutional sublayers also di er in their dilation factor (i.e., the number of steps that filters skip). Following p-variation we chose dilation factors that span the trajectory via steps of size 2 n . The last convolutional sublayer, f16 augments the model using large filter sizes of length 20 without any dilation. The output of the convolutional sublayers are fed into two fully connected layers of size 512 and 128 (f2 and f3, respectively). The final layer of MoNet (f4) is set based on the prediction task. For the anomalous classification task, the last layer is a dense layer of size 3 (corresponding to the three classes of di usion) with a Softmax activation. Softmax function maps the output of the layer 3 after applying the linear transformation g(dxi; W) = [g1(dxi; W), g2(dxi; W), g3(dxi; W)], where W denotes all the parameters in MoNet, to the predicted probability of output classes P defined as: P(dxi; W) = e g(dx i ;W) q C c=1 e gc(dx i ;W) , [35] where C = 3 is the number of classes and P(dxi; W) = [P1(dxi; W), P2(dxi; W), P3(dxi; W)].</p><p>For the regression task of finding theexponent, a dense layer of size 1 with a Sigmoid activation is used in the last layer to capture the output. Sigmoid function maps the output g(dxi; W) to a variable between 0 and 1 (i.e., the predicted value of -), and is defined as: S p (dxi; W) = 1 1 + e ≠g(dx i ;W) .</p><p>[36]</p><p>The overall architecture of the neural net shown in fig. S4 can be written as F (dxi; W) = f4 ¶ f3 ¶ f2 ¶ f1(dxi; W) where f1 = [f11, f12, f13, f14, f15, f16] is the concatenation of the output of all the 6 convolutional sublayers applied in the first layer (Fig. S5).</p><p>Loss Function. For classification task, the loss function is a categorical cross-entropy loss function, L, defined as:</p><p>qi,c log Pc(dxi; W), [37] where Qi = [qi,1, qi,2, qi,3] is the ground truth probability of each class for a trajectory xi. Note that qi,c is 1 if the sample i is in class c and 0 otherwise. Pc(dxi; W) is the output predicted probability that sample i is in class c. DKL is the Kullback-Leibler divergence between two distributions Pi and Qi.</p><p>Forprediction, the loss function is a mean squared error (MSE) L defined as:</p><p>with S p i , the predicted value ofby MoNet (the output) and Si, the ground truth value offor sample i.</p><p>Training. All the parameters of the network including the filters in the first layer and the weight matrices in the following layers were trained by back-propagating the derivative of the loss function with respect to the parameters W using a stochastic gradient descent (15). MoNet is trained using the ADAM optimizer with an adaptive learning rate that starts from 10 ≠5 .</p><p>Validation. The validation accuracy and validation MSE are evaluated on a set of hold-out unseen simulated data with the same size as the training data (i.e., 10, 000 for classification and 3, 000 forprediction).</p><p>Inference. For testing our experimental data (30 trajectories shown in Fig. S1), we treated x and y coordinates independently. For all trajectories (xi,1, xi,2, • • • , xi,T i )| 30 i=1 , we tested each 300-frame intervals separately by dividing the trajectory into m = ÂT /300Ê segments. The final results where then reported as the mean value of the output (probability in case of classification andvalue in case ofprediction) averaged over all 300-frame segments and x and and y coordinates. See Fig. S9 and S12 for the prediction outcomes for the 30 trajectories presented in Fig. S1.</p><!><p>In order to show the e ect of trajectory length on the performance of MoNet in both classification andprediction tasks, we have plotted Fig. S6 where we report validation accuracies and MES's averaged over 320 hold-out simulated trajectories. Fig. S6 shows that the accuracy increases and the MSE decreases, with increasing the trajectory length. However, the validation accuracy of MoNet for classification saturates around 88.5% ± 2.3 and validation MSE saturates over 0.02 ± 0.002 for CTRW, and 0.002 ± 0.0002 for FBMprediction, for trajectories longer than 300</p><p>In case ofprediction, estimatingbased on a single trajectory and without having an ensemble average is a challenging task for CTRW processes. Therefore, as expected, the error associated withprediction for CTRW processes is higher than the case of FBM processes (Fig. S6a).</p>

ChemRxiv

The Oriented and Flux-Weighted Current Density Stagnation Graph of LiH

An alternative, natural scheme is introduced to quantitatively analyze the magnetically induced molecular current density vector field, J. The set of zero points of J, which is called its stagnation graph (SG), has been previously used to study the topological features of the current density of various molecules. Here, the line integrals• dℓ along all edges ℓ i of the connected subset of the SG are determined. The edges ℓ i are oriented such that all Φ ℓ i are non-negative and they are weighted with Φ ℓ i . An oriented flux-weighted (current density) stagnation graph (OFW-SG) is obtained. Since J is, in the exact theoretical limit, divergence-free and due to the topological characteristics of such vector fields, the flux of all separate vortices and neighbouring vortex combinations can be determined exactly by adding the weights of cyclic subsets of edges of the OFW-SG. The procedure is exemplified by the case of LiH for a perpendicular and weak homogeneous external magnetic field, B. The method provides the means of accurately determining the strength of the current density even in molecules with a complicated set of distinct vortices.

the_oriented_and_flux-weighted_current_density_stagnation_graph_of_lih

I. INTRODUCTION<!>II. RESULTS AND DISCUSSION<!>III. CONCLUSION AND OUTLOOK

<p>All physical changes in molecules are ultimately triggered by the electromagnetic force. Molecules are restricted to interact with their environment via (not necessarily small) electronic or magnetic perturbations. Therefore, studies of the molecular magnetic response are highly relevant not only from a theoretical perspective. In molecular magnetic response theory, the induced electronic current density, J, is a key quantity from which all other magnetic response properties can be calculated in a quasi-classical fashion by evaluating the expectation value integrals. Hirschfelder has coined the term subobservable for such quantities for which it is possible to formally define a quantum mechanical operator. 1 A series of reviews on the current state of research on magnetically induced molecular current densities is available. [2][3][4] One branch of this research field is concerned with both the topological and the quantitative characterization of J, an undertaking that can be seen in analogy with the topology analysis of the electron density based on the concept of the quantum theory of atoms in molecules (QTAIM). 5,6 The J field typically consists of multiple vortices, some embedded in each other, however, their boundary surfaces may never cross due to charge conservation. These surfaces are also known as separatrix surfaces.</p><p>We have recently reported some progress on the quantitative characterization of the current-density field 7 from the secondary magnetic field that it induces which is better known as nucleus-independent chemical shifts (NICS) or also as induced magnetic field B ind (the latter with an opposite sign as compared to NICS). 8 The obtained data can be used in arguments about the aromatic or antiaromatic nature of a compound. For the quantification of the molecular current density, one is typically interested in its flux, Φ, through a particular surface, S , chosen according to chemical intuition or other demands or model ideas, such as the "ring-current" model for molecular systems consisting of one or more rings. 2 The idea underlying our recent work 7 was to apply the integral form of the Ampère-Maxwell law instead of evaluating a surface integral of J. The line integral of B ind (the induced magnetic field), over the boundary line ∂ S of the surface S is defined by</p><p>In this work, we employ the simple example of the lithium hydride molecule to illustrate how this method can be naturally extended by applying it to the so-called stagnation graph (i.e., the set of zero points) of J, such that the quantification of separate current vortices becomes exact, simple and possible to automate. Analysis of the magnetically induced current density is done routinely for various molecules, including compounds of potential practical application such as porphyrinoids. There are multiple pathways, the strength of which is difficult to determine using the standard approach of placing an integration plane through the molecule and calculating the strength of the flux across the plane. In our scheme, one first needs to obtain the stagnation graph of the molecule and then perform the integration along the points of the SG as described in Section II. Hence, it becomes possible to precisely characterize the magnetically induced current density in molecules where the vortices are hard to identify and discern.</p><!><p>We exemplify our proposed method on the lithium hydride molecule (LiH) since this gives, to the best of our knowledge, the simplest case of a non-trivial stagnation graph of J and thus also the simplest case of a molecule that shows two physically distinct (separable) vortices. The topology of the magnetically induced current density field, J, of LiH has been discussed previously in great detail by Stevens and Lipscomb (1964) 9 , Keith and Bader (1993), 6 and later by Pelloni, Lazzeretti and Zanasi (2009) 10 , so we will only give a summary here. General procedures to compute stagnation points and stagnation lines have been published, as well, and are freely available. 11,12 Placing B parallel to the z axis and perpendicular to the Li -H bond (lying along the y direction) results in a J field that is composed of exactly two separate current vortices, each of which possessing a central stagnation line (see Fig. 1). Both vortices are separated from each other by a single closed surface K of the exact topology and the approximate geometry of a sphere. The spherical vortex domain (inner vortex) is completely enclosed in the other domain, extending over the remaining full molecular space (outer vortex). The domains are illustrated as streamlines in Fig. 3. The outer vortex has an open stagnation line, ℓ 1 , extending from z = −∞ to z = ∞ and lying in the (y, z) plane. Line ℓ 1 is passing the H atom at a short distance, only a fraction of an atomic unit, and is bending towards it. Above and below the (x, y) plane, ℓ 1 is bending slightly towards the center of the LiH molecule but straightens out at larger ±z heights. The current vortex around this stagnation line ℓ 1 is diatropic, thus, according to our convention, the vortical flow is clockwise if the external field B is pointing upwards (= z direction) and the reader looks towards negative z. The approximately spherical inner vortex domain can be imagined as inserted in-between the streamlines of the outer vortex which are bypassing this domain similarly to how a laminar-flowing fluid would pass around a ball fully submerged into it. This results in two isolated "toroidal" saddle stagnation points p on the separatrix sphere K where the outer flow diverges/converges in/from all directions on K .</p><p>The inner domain encloses the Li atom and its radius extends roughly to the midpoint between the two atoms. The geometrical center of K does not lie on the Li atom, but instead, it is significantly shifted away from the H atom. The stagnation line, ℓ 2 , of the current vortex is closed, which gives rise to a toroidal flow with a reflux region in the middle. On the side towards the H atom, the flux resembles that of a paratropic vortex, while on the remote side of the Li atom, the flux acts as if it were diatropic. The respective parts of the stagnation line are illustrated in red and green in Fig. 2 and Fig. 4. The streamline representations show that the current-density flow in the middle region is similar to the water circulating in a waterspout fountain with an inner reflux tube. The inner reflux stream passes close to the Li atom, roughly perpendicular to both B and the Li -H bond, lying in the (x, y) plane. It has the shape of a double S and contains a single separatrix line which connects the source critical saddle point p The stagnation line, ℓ 2 , of the inner vortex is a topological circle (a closed loop) that we split into two branches, ℓ 2,d (with diatropic current vortices in the vicinity) and ℓ 2,p (with para- tropic vortices in the vicinity), connected by the "branching points" p 0 and p ′ 0 illustrated in orange. Strictly speaking, this division is arbitrary, and ℓ 2 is actually one unbranched topological domain. Nevertheless, this scheme provides a simple example of the general method of current flux determination we are outlining here. We also note that this type of toroidal current vortex is, obviously, neither diatropic nor paratropic over its whole domain.</p><p>Due to charge conservation, no currents can pass from one vortex (domain) into or out of another one. Hence, a total current flux can be assigned to each individual vortex. As a chemical interpretation, one would assign the diamagnetic and spatially unconfined current vortex around ℓ 1 to the hydride anion, which dominates the response and encloses a spatially confined toroidal vortex arising from the Li + cation.</p><p>As we have shown previously, 7 and since B ind is vanishing at infinity in z and y direction, the total current flux, Φ ℓ 1 , of the outer vortex can be obtained from the line integral</p><p>Here we have chosen the direction of stagnation line ℓ 1 such that Phi ℓ 1 becomes positive. In this way, every current-density stagnation graph can be uniquely oriented to yield a directed graph. Furthermore, to each "edge" (line segment) of a stagnation graph, a flux integral can be assigned, such that an oriented edge-weighted graph with strictly positive weights is obtained.</p><p>To obtain the current flux of the toroidal current inside K , one can make use of the notion that the full flux is passing through the D-shaped closed stagnation line ℓ 2 (composed of ℓ 2,d and ℓ 2,p ). Thus,</p><p>where again an orientation is obtained for each integral associated with ℓ 2,d and ℓ 2,p . A simple numerical integration scheme has been applied (see Appendix B for details), through which we have obtained We conjecture that any field J partitioned into vortices separated by asymptotic surfaces (separatrices), i.e., the field is decomposable into pairwise disjunct sets and where the union is the complete set, has cyclic stagnation subgraphs, defining the vortex, like the pair of stagnation lines in the example of the toroidal vortex in LiH. Setting this cyclic subgraph to ∂ S in the line integral in equation ( 1) then gives the current flux, Φ, in this vortex. Then in order to obtain the flux of a separate vortex one only has to add the weights of the corresponding cyclic subgraph in the correct orientation (i. e. sign).</p><p>Regarding the efficiency and feasibility of the proposed method, there are the following advantages • Instead of surface integrals, only line integrals have to be calculated, which roughly leads to scaling of order O(N) versus O(N 2</p><p>• For integrals of the current flux for which the stagnation lines are (approximately) known, e.g., the central axis in case of planar symmetric rings, and which yield the total "ring" current, 7 no current density has to be calculated directly but rather only the typically simpler chemical shielding tensors have to be computed.</p><p>• With this method, there is no ambiguity concerning the precise definition and localization of the vortices under consideration, which is in stark contrast to the commonly used schemes which rely on explicit or implicit topological knowledge that very often was present only in rudimentary form. So it seems highly likely that many of the previous studies where surface integrals of the current density have been calculated on the basis of a casual inspection of streamline graphs or various adhoc integration schemes are flawed due to ill-defined current domains.</p><p>On the contrary one has two face two methodological disadvantages using this method • For all but the simplest cases (total current in symmetric ring systems), the SG has to be determined; as outlined above, we suggest that this is only an ostensible disadvantage since integration without knowledge of the topology is in the risk of being errant.</p><p>• For open stagnation line vortices (like the hydride vortex in our example), one has to integrate relatively far out (i.e., approximately 100 or even 200 atomic units) into the shielding cone of the molecule in order to achieve convergence. This is a consequence of the nonlocal nature of the shielding contributions with respect to the underlying J field.</p><!><p>The presented approach effectively gives a complete analysis of a molecule's magnetically induced current flux. It can be achieved by evaluating the line integral from Eq. (1) for each "edge" of the stagnation graph (and thereby, orienting all edges). The results of an OFW-SG can be condensed into a schematic diagram as shown in Fig. 1. The strategy is completely generalizable to non-planar and non-cyclic molecules, as well. Hence, it can be carried out for any molecule with a non-trivial connected stagnation graph. It has been described in the literature that large molecules without symmetry elements may not possess a non-trivial stagnation graph, meaning that J cannot be partitioned into smaller non-trivial simple vortices, or at least only barely so. This is usually the case if there are significant current contributions parallel to the external field B which cause the vortices to show a helical component, necessarily leading to non-zero current transfer in between otherwise separate vortices. For such cases, instead of J, a pseudo-J field can be investigated where the parallel component is projected out. One then obtains a pseudostagnation graph 13,14 for which the same procedure as above can be applied. We are currently investigating this possibility.</p><p>dimensions and positioned such that the Li nucleus is centered on the right z edge between two grid points at a distance 0.005 a 0 . The 2 • (100 + 50) − 4 = 296 boundary points {r b i } i∈{1,...,296} were used for the numerical line integration,</p>

ChemRxiv

Self-assembly of tetrareduced corannulene with mixed Li–Rb clusters: dynamic transformations, unique structures and record⁷Li NMR shifts

Self-assembly processes of the highly reduced bowl-shaped corannulene generated by the chemical reduction with a binary combination of alkali metals, namely Li-Rb, have been investigated by variabletemperature 1 H and 7 Li NMR spectroscopy. The formation of several unique mixed metal sandwich products based on tetrareduced corannulene, C 20 H 10 4À (1 4À ), has been revealed followed by investigation of their dynamic transformations in solutions. Analysis of NMR data allowed to propose the mechanism of stepwise alkali metal substitution as well as to identify experimental conditions for the isolation of intermediate and final supramolecular products. As a result, two new triple-decker aggregates with a mixed Li-Rb core, [{Rb(THF) 2 } 2 ]//[Li 3 Rb 2 (C 20 H 10 ) 2 {Li + (THF)}] (2) and [{Rb(diglyme)} 2 ]// [Li 3 Rb 3 (C 20 H 10 ) 2 (diglyme) 2 ]$0.5THF (3$0.5THF), have been crystallized and structurally characterized. The Li 3 Rb 2 -product has an open coordination site at the sandwich periphery and thus is considered transient on the way to the Li 3 Rb 3 -sandwich having the maximized intercalated alkali metal content. Next, the formation of the LiRb 5 self-assembly with 1 4À has been identified by 7 Li NMR as the final step in a series of dynamic transformations in this system. This product was also isolated and crystallographically characterized to confirm the LiRb 5 core. Notably, all sandwiches have their central cavities, located in between the hub-sites of two C 20 H 10 4À decks, occupied by an internal Li + ion which exhibits the record high negative shift (ranging from À21 to À25 ppm) in 7 Li NMR spectra. The isolation of three novel aggregates having different Li-Rb core compositions allowed us to look into the origin of the unusual 7 Li NMR shifts at the molecular level. The discussion of formation mechanisms, dynamic transformations as well as unique electronic structures of these remarkable mixed alkali metal organometallic selfassemblies is provided and supported by DFT calculations.

self-assembly_of_tetrareduced_corannulene_with_mixed_li–rb_clusters:_dynamic_transformations,_unique

Introduction<!>Results and discussion<!>X-ray crystallographic studies<!>Li NMR shis and structural correlations<!>View Article Online<!>DFT computational studies<!>Conclusions

<p>Non-planar carbonaceous aromatic compounds, such as fullerenes and carbon nanotubes, have recently emerged as promising light-weight materials for electronics and energyrelated applications. 1 Their potential application in lithium-ion battery technology has also attracted special attention. 2 Moreover, the unique thermoelectric and superconducting properties of the reduced forms of curved carbon frameworks has triggered a renewed interest in these materials. 3 All these recent advances stimulated a rapid expansion of the family of curved carbon-rich compounds which now includes a great variety of carbon bowls, 4 warped nanographene sheets, 5 as well as nanobelts and nanoribbons of different sizes. 6 Bowl-shaped polyaromatic hydrocarbons (also referred to as carbon or p-bowls) have been broadly investigated over the last two decades to reveal their unique coordination and redox properties. 4,7 For example, they can readily accept multiple electrons in stepwise reduction reactions to form the sets of the consequently reduced non-planar polyaromatic carbanions. The latter have been the subjects of numerous investigations 8 due to their unique interplay of strain and conjugation. It was demonstrated that the smallest bowl-shaped polyarene, corannulene (C 20 H 10 , 1, Scheme 1a) is able to undergo four reduction steps to form a set of corannulene anions, C 20 H 10 nÀ (n ¼ 1-4). 9 The nal fourfold reduction is accompanied by a signicant core rearrangement of the corannulene tetraanion which can be considered to have an annulene-within-an-annulene electronic structure (Scheme 1b). 10 Notably, C 20 H 10 4À bears one electron per ve carbon atoms and is more electron rich than the fullerene-hexaanion (one electron per ten carbon atoms in C 60 6À ). This makes it a very interesting and unique carbanion for investigation of metal coordination and self-assembly reactions.</p><p>The highly reduced corannulene anions exhibit remarkable ability to form unique supramolecular aggregates in solutions, as shown by extensive 1 H and 7 Li NMR spectroscopy studies. 10 However, structural investigations of the charged p-bowls were lagging behind their solution spectroscopic studies until we have accomplished the rst X-ray diffraction characterization of the product formed by the tetrareduced corannulene with lithium counterions. 11 The formation of a remarkable aggregate with a Li 5 -core sandwiched between two C 20 H 10 4À decks has been established (Scheme 2a). This supramolecular aggregation with lithium ions allows to achieving a tetrareduced state of corannulene, as the electrochemical generation of C 20 H 10 4À cannot be accomplished due to a very large negative standard potential located outside of the current experimental window. 12 We have also demonstrated that the triple-decker supramolecular aggregate, [Li 5 (C 20 H 10 4À</p><p>) 2 ] 3À , can be crystallized in different external coordination environments, showing no signicant effect on the geometry of the sandwich core. 11,13 In addition, the formation of the [Li 5 (C 20 H 10 4À</p><p>) 2 ] 3À product (abbreviated as Li 5 below) is conrmed in solution based on the observed shis for sandwiched Li + ions (À11.70 ppm) and a proper 3 : 5 integration of external vs. internal ions in the 7 Li NMR spectra. Importantly, these results illustrated the ability of tetrareduced corannulene to engage all of its adjacent sixmembered rings in alkali metal binding and thus to encapsulate a large amount of Li + ions, which can be related to the high charge capacity of the corannulene-based electrodes in Li-ion batteries. 14 Remarkably, the coordination limit of the highly electron rich corannulene tetraanion can be further extended through the synergistic use of two alkali metals as the reducing agents. The concomitant reduction of C 20 H 10 using a Li-K mixture resulted in the recent discovery of a novel class of mixed metal supramolecular products, in which tetrareduced corannulene exhibited its new coordination record. 15 The C 20 H 10 4À anion is able to engage all its sites, ve benzene rings along with a central ve-membered ring, for binding of six alkali metal ions in the resulting mixed metal core, Li 3 K 3 (Scheme 2b) or LiK 5 , triple-decker sandwiches. Notably, the previously unseen involvement of the hub-site of corannulene tetraanion in lithium ion binding is accompanied by unprecedented negative shis in 7 Li NMR spectra (up to ca. À25 ppm). 15 These recent results established a new paradigm for curved polyaromatic ligands in alkali metal binding and opened new opportunities for design and synthesis of novel mixed metal organometallic supramolecular products. However, the origin of the observed high negative shis in 7 Li NMR spectra was not understood and required additional attention. Herein, we set to investigate the reduction reactions of corannulene using a binary alkali metal combination comprised of Li and Rb, the heavier congener of K. For the Li-Rb combination, we expect that a greater size mismatch of two alkali metals could not only facilitate the transformations observed in the Li-K-C 20 H 10 system but also afford novel supramolecular products and open new reaction pathways that have not been seen before. We also expect that isolation and analysis of novel mixed metal organometallic products formed by the highly charged corannulene could help to shed light on the record negative shis observed in 7 Li NMR spectra. We therefore used a combination of NMR spectroscopy, X-ray crystallography and DFT theoretical methods for thorough investigation of these complex systems. As a result, we have been able to reveal a direct relation between the structures of supramolecular assemblies and coupling effects of the highly charged polyaromatic bowls at the molecular level and to correlate those with the observed 7 Li NMR shis.</p><!><p>Corannulene reacts with alkali metals, Li through Cs, to initially produce intense green solutions characteristic for the corannulene monoanion. The excess of alkali metal quickly reduces the resulting C 20 H 10 c À monoanion to the purple C 20 H 10 2À dianion. In the case of Li metal, dianions can then be further reduced to tetraanions that form supramolecular [Li 5 (C 20 H 10 4À ) 2 ] 3À aggregates having characteristic 7 Li NMR shis for sandwiched Li + ions (À11.70 ppm). 11,13 While monoand dianions of corannulene do not form such supramolecular species, they exhibit a variety of coordination modes that depend on the size of alkali metal ions and other experimental variables. 16,17 The corannulene bowl is not attened upon acquisition of the rst electron but more pronounced changes can be seen upon addition of the second electron. In contrast, a signicant bowl depth decrease and C-C bond length alteration pattern are observed for tetrareduced corannulene. 11,13,15,18 Scheme 1 Corannulene (a) along with the electronic structure of C 20 H 10 4À (b).</p><p>Scheme 2 Supramolecular aggregates with (a) Li 5 -and (b) Li 3 K 3 -cores sandwiched between two C 20 H 10 4À decks.</p><p>These experimental observations are in line with multiple theoretical predictions. 11,19,20 In Li 5 -sandwiches all ve benzene rings of C 20 H 10 4À are engaged in lithium ion binding leaving an internal space between two corannulene hub-sites empty. Notably, the distance between the centroids of 5-membered rings in the triple-decker Li 5 -products is 3.5 Å 11 and this may not be sufficient to accommodate an additional lithium ion in that space.</p><p>For comparison, the separation between 5-membered rings in lithium cyclopentadienide, [(h 5 -Cp) 2 Li] À , is ca. 4.0 Å, 21 and the Li-Cp distances range from 3.8 to 4.1 Å in some other cyclopentadienide Li-organometallic compounds. 22,23 We have recently demonstrated 15 that in order to get access to the central cavity larger alkali metals, such as K, should be introduced into the reaction along with Li. The initially formed Li 5 -sandwich having both corannulene decks parallel to each other starts to open up upon stepwise substitution of small Li + ions by larger K + ions, leading to the products having angled corannulene decks (Scheme 3). This results in the opening of a channel which allows the insertion of Li + ion from periphery into the previously inaccessible inner cavity of the sandwich (Scheme 3, pathways A and B). It is necessary to have two large alkali metal ions at the rim of the sandwich for the lithium insertion process to take place, as the rst substitution step (formation of Li 4 K) does not provide a sufficient opening for such insertion. For the second step, there are two possible pathways, A and B (through the formation of a and b isomers), where rings 2 and 5 or 2 and 3 are occupied by larger alkali metals (Scheme 3). Our DFT calculations showed that in the Li-K systems the insertion occurs through the b-isomer only (pathway A). 15 Monitoring the reduction reaction of 1 with a mixture of Li and Rb metals in THF-d 8 by NMR spectroscopy conrms the formation of supramolecular products formed by C 20 H 10 4À . The reaction seems to proceed faster than in the case of Li-K and substantial amounts of Li-Rb-sandwiches are quickly observed within a few hours (Fig. 1a).</p><p>At the initial stage of reaction, which is determined by its characteristic brown-red color, two major broad peaks at À23.92 and À24.46 ppm and one very small peak at À24.86 ppm are observed in 7 Li NMR spectrum (Fig. 1a). The signicant upeld-shi of these peaks is associated with the Li + ion squeezed between two 5-membered rings of tetrareduced corannulene, revealing the formation of several mixed metal Li-Rb sandwichtype assemblies. The latter peaks are assigned to the Li 3 Rb 3 and Li 4 Rb 2 products (a-and b-isomers, see Fig. 2 for schematic representation of the sandwich cores) based on integration of individual peaks and the possibility to crystallize these major products in the single-crystalline form, as reported below. It can be mentioned here that all NMR peaks coalesce into a single broad signal at temperatures ranging from +20 to À20 C, showing the rapid lithium ion exchange between all molecular structures existing in solution (ESI, Fig. S1 †).</p><p>The 7 Li signal at À11.70 ppm in Fig. 1a belongs to the monometallic Li 5 -product in which ve Li + ions are occupying the sites in between the benzene rings of C 20 H 10 4À (Scheme 2a).</p><p>Additional peaks between À6.2 and À8.3 ppm are the signals of lithium ions located between the same 6-membered ring sites but belonging to the mixed metal Li-Rb sandwiches (similar to the reported Li-K aggregates). 15 In contrast, the 7 Li NMR signals of . Note that the pathways A and B have different energetic requirements. Li is shown as a sphere with small radius in contrast to a larger sphere that is used to represent larger alkali metals (M). ) 2 ] 3À to the mixture of all Li-Rb sandwiches is 0.38 to 1. This is consistent with the 1 H NMR spectra which also show a 0.38 ratio of the representative peaks at 6.86 ppm (the signal of C 20 H 10 4À in Li 5 ) and 6.69 ppm (a broader peak associated with all mixed Li-Rb products). It should be mentioned here that the precise time assignment to a particular point in the series of dynamic NMR measurements is problematic due to very strong dependence of the system reactivity on very minor and hard-to-control experimental variations. For example, we observed that diminutive changes in purity of commercial metals (from the same supplier) play a drastic role in the development. The NMR spectra represented in Fig. 1 show the reaction progression in the time frame of ca. 30 hours. These processes may be much slower (up to 3-5 days), if the quality of reagents is just slightly less perfect. In some cases, the reaction could be fully decomposed at an early stage not allowing for the nal product to be obtained, indicating poor quality of the reagents used.</p><p>The next measurement (Fig. 1b) shows a smaller amount of the initially formed Li 5 -product and subsequent growth of the Li-Rb sandwiches, as illustrated by a better dened shape of the corresponding NMR signals. Therefore, a more detailed discussion of this spectrum is carried out below based on Fig. 2. It should be noted that the precise peak integration is still slightly hindered by peak overlaps and the elevated background downeld from ca. À10 ppm. The signal of external lithium ions of the Li 5 -sandwich, appearing as a very broad peak between À4 to À6 ppm, and the corresponding signals of external Li + ions bound to other sandwich products existing in the mixed metal system all contribute to the elevated background.</p><p>Since the 7 Li signal at À11.70 ppm belongs to the Li 5 -sandwich, containing lithium ions only, 11 the close small peak at À9.62 ppm is assigned to the product resulting from the rst substitution of one sandwiched Li + ion by a Rb + ion (to form a Li 4 Rb core). No peak correlates with this one in the region of À20 to À30 ppm in the 7 Li NMR spectrum and, as expected, there is no internal lithium insertion at this stage. This transient species undergoes further substitution and thus does not accumulate in any signicant amount in solution (Scheme 3).</p><p>The second substitution of Li by Rb leads to two Li 3 Rb 2 isomers that open a path for the internal lithium insertion (pathways A and B, Scheme 3). Aer insertion, the resulting aand b-isomers of Li 3 Rb 2 have an open coordination site between the two benzene rings of corannulene decks and can be abbreviated for clarity as a-Li 3 Rb 2 , and b-Li 3 Rb 2 ,. Subsequent addition of Li + to this open site provides a-Li 4 Rb 2 and b-Li 4 Rb 2 isomers, respectively. Final substitution of this loose Li + ion by Rb + affords the Li 3 Rb 3 -product (Fig. 2). Since the Li 3 Rb 3 sandwich is a single product of both insertion pathways, the corresponding 7 Li NMR signal at À23.92 ppm is notably growing with time (Fig. 1a-c). Three peaks at À6.21, À7.61, and À8.32 ppm can be identied in the 7 Li NMR spectrum with a related triad of peaks having the same relative intensities being observed at the highly-negative region at À23.92, À24.44, and À24.86 ppm. Notably, the signals from these two triads correlate really well and therefore should belong to the same type of supramolecular sandwiches, as assigned in Fig. 2 with arrows. The triad on the right is associated with the Li + ion internally located between the central 5-membered rings of C 20 H 10 4À , while the signals on the le stem from the Li + ions sandwiched between the peripheral 6-membered rings of tetrareduced corannulene. The relative integration of peaks is ca. 2.6, 3.5, and 3.0 to 1 in pairs of À6.21/À23.92, À7.61/À24.44, and À8.32/ À24.86 ppm, respectively. Slight overestimation of the integrated intensity may suggest that Li + ions located at the sandwich periphery exchange with external environment. A small shoulder of the peak at À7.61 ppm may be indicative of differentiation between two peripheral Li + ion sites observed in b-Li 4 Rb 2 .</p><p>It should be mentioned here that the insertion mechanism through the a-isomer of Li 3 Rb 2 (pathway B, Scheme 3) is much less energetically favourable. The internal lithium migration for b-isomer was found to be barrierless, whereas for a-isomer the barrier was calculated to be +15.48 kcal mol À1 . 24 As a result, the corresponding NMR signals of a-isomer are always much less intense, being at about 10-20% relative to the b-isomer. In the case of Li-K-aggregates, the formation of a-isomer of the Li 3 K 2 composition was not observed at all, thus showing that a larger Rb + ion opens a new reaction pathway in comparison with K + .</p><p>Aer the Li 5 -sandwich is fully consumed in the stepwise substitution reactions (as seen by the disappearance of the 6.85 ppm peak in 1 H and À11.70 ppm peak in 7 Li NMR spectra, Fig. 1c) the major product in solution is the Li 3 Rb 3 sandwich, with the b-isomer of Li 4 Rb 2 being a minor component of the mixture. Interestingly, there is a new peak appearing at À21.46 ppm which is assigned to the fully substituted product having ve rubidium ions at the rim (LiRb 5 ). Similar to the Li-K system ( 7 Li NMR signal of LiK 5 appears at ca. À22.40 ppm), this is the only mixed metal aggregate that can be observed at elevated temperatures (already clearly visible at ca. 0 C, Fig. S4 †) due to negligible exchange of the central Li + ion trapped inside the sandwich.</p><p>In the corresponding 1 H NMR spectra, the protons of C 20 H 10 4À in Li 3 Rb 3 and Li 4 Rb 2 aggregates are observed at 6.68 ppm, while a sharp peak, corresponding to the LiRb 5 sandwich, appears at 6.47 ppm. As concluded from the relative peak integration, LiRb 5 constitutes approximately 20% of the reaction mixture in solution. At this point, the amount of solventseparated Li(THF) x + species (broad peak at ca. 0 ppm) and nonsandwiched lithium ions bound to the external surface of C 20 H 10 4À (very broad, 0 to À4 ppm) is quite signicant. This is also accompanied by the appearance of some precipitationdecomposition products in the NMR ampoule. Notably, NMR monitoring shows that the a-isomer of Li 4 Rb 2 can no longer be seen and the amount of the corresponding bisomer is very small. At the nal stage, LiRb 5 is revealed as the major component in the mixture (Fig. 1d), and it nally becomes the only product observed in solution (Fig. 1e). The formation of LiRb 5 is accompanied by the appearance of a single peak at À21.46 ppm in 7 Li NMR and at 6.49 ppm in 1 H NMR spectra. It should be mentioned here that a substantial amount of unidentied precipitate is observed in the NMR ampoule at this stage.</p><!><p>As revealed by NMR studies (Fig. 1a-c S2 †).</p><p>It should be emphasized here that the fact that the transient b-Li 3 Rb 2 , sandwich can be isolated (Fig. 3a and c) provides compelling experimental evidence for the proposed Li-insertion mechanism. The sandwich opening (measured as an angle between the mean planes passing through the rim of two pbowls) in the case of b-Li 3 Rb 2 , is greater than in the Li 3 Rb 3 or Li 3 K 3 sandwiches (27 vs. 21 and 20 , respectively). The distance between the rim C-atoms at the sandwich opening side is about 6.6 Å, which is substantially longer than the corresponding separations in Li 3 Rb 3 (Fig. 3b) or Li 3 K 3 cases (6.2 and 6.1 Å, respectively).</p><p>In b-Li 3 Rb 2 ,, the fourth Li + ion (also having a coordinated THF molecule) is closely approaching the sandwich (to show this side binding, the sandwich is abbreviated as b-Li 3 Rb 2 ,-Li) from the Li 3 -triangular side (Fig. 3c), thus forming a Li 4 parallelogram with the Li/Li contacts of 2.86 Å inside the sandwich and of 2.92 Å outside the sandwich. This additional lithium cation brings both corannulene bowls to a very short distance of 3.8 Å (compared with 4.1 Å in Li 3 Rb 3 ). In Li-Rb sandwiches in 2 and 3, both external concave cavities of C 20 H 10 4À are lled with rubidium ions that participate in the formation of a 1D network through the shared external solvent molecules (Fig. S5 †). In 2, these 1D chains are further packed into 2D sheets via additional C-H/p interactions (Fig. S6 †). Crystallization of the nal product in this series of dynamic transformations, namely of LiRb 5 , has been very problematic. By the time this sandwich appears as the only product in solution, which is determined by its characteristic red color, there is a signicant amount of precipitation-decomposition observed in the system. Removal of this unidentied solid affords rather dilute solutions, which do not allow efficient crystal growth of LiRb 5 . Aer numerous attempts we were able to isolate the single crystals of LiRb 5 sandwich, but even the best crystals show diffraction only up to ca. 1 Å resolution. Nevertheless, these X-ray diffraction data allowed us to conrm the LiRb 5 core structure of this product (Fig. 4), as predicted based on the stepwise alkali metal substitution mechanism and NMR data.</p><!><p>The record negative 7 Li NMR shis characteristic of supramolecular Li-K and Li-Rb sandwich products formed by C 20 H 10 4À anions are worthy of special discussion. Such unprecedented shis of the central Li + ion occupying the cavity between the two hub-sites of corannulene tetraanions (signals range from À21 ppm to À25 ppm) are due to strong shielding of this internal cation from the external magnetic eld. This implies that the centrally located Li + ion is surrounded by some electron density, which clearly separates and distinguishes it from other alkali metal ions (Li, K, or Rb) sandwiched in-between the peripheral 6-membered rings of</p><!><p>C 20 H 10 4À anions. Importantly, this provides the rst indication that some electronic communication should exist between the negatively charged corannulene decks separated by an alkali metal layer in the triple-decker organometallic products. The most plausible way for such electronic exchange should be through the orbital interactions between the two tetrareduced p-bowls. This is clearly conrmed by the observation that the shi of 7 Li NMR signal of the central Li + cation correlates with the distance between two corannulene decks in the isolated sandwich structures.</p><p>In Li 3 K 3 , the central Li + ion resonates at À24.48 ppm but the signal is shied to ca. À22.40 ppm in LiK 5 (the separations between the hub-sites of two C 20 H 10 4À decks are 3.8 and 4.0 Å, respectively). 15 The centrally located Li + ions in Li 3 Rb 3 and LiRb 5 products appear to be deshielded (À23.92 and À21.46 ppm) in comparison with their Li 3 K 3 and LiK 5 analogues, which is consistent with the size of Rb + ions requiring more space than K + ions. As a result, the corresponding separations between two corannulene bowls in Li 3 Rb 3 and LiRb 5 are 4.1 and 4.2 Å, respectively. Clearly, Rb + ions move the charged decks farther apart and thus weaken communication between two p-bowls in the triple-decker aggregates. Consequently, the central Li + cation becomes less shielded and the corresponding 7 Li NMR signals shi downeld for sandwich structures having higher amount of heavier alkali metals.</p><p>The close analogy can be drawn with the strong shielding effect of the C 60 -fullerene hexaanion. The 3 He atom inside the C 60 6À cage was found to be signicantly more strongly shielded (by ca. 20 ppm) than any other previously reported encapsulated 3 He atom, suggesting the ability of electrons to move freely about the surface of a charged spheroidal p-system. 25 In the case of mixed alkali metal sandwich-type assemblies formed by highly charged corannulene, electrons can also move freely about the p-bowl surface of 1 4À , but in addition to that, they are able to move from one surface to the other, as we found here for the rst time.</p><!><p>To get further theoretical insights into the experimentally observed electronic communication between the highly charged corannulene anions, we carried out DFT calculations at the PBE0/def2-TZVPP(Li, Rb)//cc-pVDZ(C, H) level of theory. The choice of the model system was discussed elsewhere. 15 Indeed, theoretical calculations revealed 24 that p-systems of two tetrareduced corannulene bowls are coupled with each other, forming shared electronic density that is mainly concentrated in the region between two 5-membered rings. The central lithium cation is thus wrapped into a negatively charged cocoon, which seems responsible for a record high shielding effect observed in 7 Li NMR. The corresponding orthogonal molecular orbitals representing delocalization of electronic density between the C 20 H 10 4À bowls are shown in Fig. 5.</p><p>Theoretical analysis of components of shielding tensor revealed that the main difference between 7 Li NMR signals of the lithium ions placed in-between six-membered rings and of the central Li + ion comes from the paramagnetic term of the Ramsey formula (s tot ¼ s d + s p ). The paramagnetic term involves mixing of ground and excited states of the molecule due to the magnetic eld, and it is rather sensitive to the molecular electronic structure. Thus, the main reason for such a large shielding of the central Li + ion, in comparison with those located between six-membered rings of C 20 H 10 4À , should arrive from the electronic structure of the sandwich and, more precisely, from the local electronic environment of the cation. MOs presented in Fig. 5 are found to provide the largest contributions to the s p component of the shielding tensor. These MOs clearly show extended delocalization around the central Li + ion and no delocalization for the lithium(s) sitting between peripheral 6-membered rings. At the same time, the calculated charges of lithium cations of both types indicate essentially no charge transfer from the highly negatively charged corannulene tetraanions to Li + ions. The NBO charges are very similar (+0.88 and +0.91, respectively, in Li 3 Rb 3 and +0.87 in LiRb 5 ). The larger distance between two bowls makes the coupling between them weaker, as suggested from experimental data. Consequently, this results in a downeld-shi of the 7 Li NMR signal from the central Li + ion. The corresponding component of the shielding tensor was found to show the trend that is opposite to the distance between two C 20 H 10 4À anions.</p><p>Thus, the signal from the central lithium cation could be used for qualitative evaluation of the strength of coupling between the charged bowls in sandwich-like aggregates.</p><!><p>Several remarkable triple-decker organometallic aggregates having a mixed metal core (Li 3 Rb 2 ,, Li 3 Rb 3 , and LiRb 5 ) sandwiched between two tetrareduced corannulene decks have been isolated in this work, following the NMR studies of their dynamic transformations in solutions. These sandwiches have the central cavity located in between of the hub-sites of two C 20 H 10 4À decks occupied by an internal Li + ion that exhibits the record high negative shi (up to À25 ppm) in 7 Li NMR spectra. Theoretical investigation of these unique systems revealed that coupling of two highly-charged corannulene bowls results in a shared region of high negative electron density around the central lithium ion that is responsible for a record shielding effect observed in 7 Li NMR spectra. Analysis of three new sandwich structures allowed us to identify the trend: the larger separation between two p-bowls makes their coupling to be weaker, which is accompanied by the down-eld shi of the corresponding 7 Li NMR signal from the internally encapsulated Li + ion. Consequently, these 7 Li NMR signals can be used for qualitative evaluation of the strength of coupling between the charged carbon bowls in sandwich-like supramolecular aggregates. The higher the negative chemical shi is the stronger coupling should be expected.</p>

Royal Society of Chemistry (RSC)

Probing the Diacylglycerol Binding Site of Presynaptic Munc13-1

Munc13-1 is a presynaptic active zone protein that acts as a master regulator of synaptic vesicle priming and neurotransmitter release in the brain. It has been implicated in the pathophysiology of several neurodegenerative diseases. Diacylglycerol and phorbol ester activate Munc13-1 by binding to its C1 domain. The objective of this study is to identify the structural determinants of ligand binding activity of the Munc13-1 C1 domain. Molecular docking suggested that residues Trp-588, Ile-590, and Arg-592 of Munc13-1 are involved in ligand interactions. To elucidate the role of these three residues in ligand binding, we generated W588A, I590A, and R592A mutants in full-length Munc13-1, expressed them as GFP-tagged proteins in HT22 cells, and measured their ligand-induced membrane translocation by confocal microscopy and immunoblotting. The extent of 1,2-dioctanoyl-sn-glycerol (DOG)- and phorbol ester-induced membrane translocation decreased in the following order: wild type > I590A > W588A > R592A and wild type > W588A > I590A > R592A, respectively. To understand the effect of the mutations on ligand binding, we also measured the DOG binding affinity of the isolated wild-type C1 domain and its mutants in membrane-mimicking micelles using nuclear magnetic resonance methods. The DOG binding affinity decreased in the following order: wild type > I590A > R592A. No binding was detected for W588A with DOG in micelles. This study shows that Trp-588, Ile-590, and Arg-592 are essential determinants for the activity of Munc13-1 and the effects of the three residues on the activity are ligand-dependent. This study bears significance for the development of selective modulators of Munc13-1.

probing_the_diacylglycerol_binding_site_of_presynaptic_munc13-1

<!>MATERIALS AND METHODS<!>Site-Directed Mutagenesis of Full-Length Munc13-1 and the C1 Domain.<!>Cell Culture.<!>Cell Fractionation and Immunoblotting.<!>Confocal Microscopy.<!>Expression and Purification of N15-Labeled Munc13-1 C1 and Its Mutants for the NMR Study.<!>NMR Sample Preparation.<!>NMR Spectroscopy.<!>Molecular Docking.<!>Statistical Analysis.<!>The Ligand Binding Site of Munc13-1 C1 Is Different from That of PKC\xce\xb4 C1B.<!>Effect of Mutations of Trp-588, Ile-590, and Arg-592 on the Membrane Translocation of Full-Length Munc13-1.<!>Munc13-1 C1 Partitions into a Hydrophobic Environment in a Ligand-Independent Manner.<!>Munc13-1 C1 Binds Endogenous Activator DAG and Tumor-Promoting Phorbol Ester PDBu.<!>Factors That Influence Munc13-1 C1\xe2\x88\x92Diacylglycerol Recognition.<!>DISCUSSION

<p>Munc13-1 is a presynaptic protein that regulates vesicle exocytosis by bridging the vesicle and the plasma membrane.1–4 It facilitates the formation of the SNARE [soluble NSF (N-ethylmaleimide-sensitive factor) attachment protein receptor] complex by interacting with proteins, such as syntaxin-1 and Munc18, and regulating vesicle priming5,6 and neurotransmitter release.2,7 Multiple studies indicated Munc13-1's role in short-term presynaptic plasticity8–13 and long-term potentiation.14 Munc13-2, Munc13-3, and Munc13-4 are the other members of the Munc13 protein family.15 Complete abolishment of neurotransmitter release was observed in the double-knockout mice of Munc13-1 and Munc13-2.3,16–18 Munc13-1 has been implicated in the Aβ-induced neurotoxicity in an Alzheimer' s disease model.11,12,19,20 Clinical evidence suggested a strong correlation of single-nucleotide polymorphism in the Munc13-1 gene with amyotrophic lateral sclerosis and frontotemporal dementia.21–24 Munc13-1's role has been implicated in neuron cells in amyotrophic lateral sclerosis25 and motor neuron degeneration.26 Additionally, Munc13-1 is involved in insulin release,27 and recent studies highlighted the role of the Munc13 protein in alcoholism.28–30</p><p>Munc13-1 is an ~200 kDa protein consisting of three C2 domains (C2A, C2B, and C2C), one C1 domain, and a MUN domain. The C2A domain is at the N-terminus followed by the diacylglycerol (DAG)/phorbol ester binding C1 domain, the Ca2+ binding C2B domain, and the MUN domain, and the C2C domain is at the C-terminus.17,31 The C2A domain forms a homodimer/heterodimer with the Rab3-interacting molecules (RIMs)32–34 and regulates neurotransmitter release and Rab3- and RIM-dependent presynaptic plasticity.35,36 The C2B domain binds to the plasma membrane and regulates Ca2+- and PIP2-dependent short-term plasticity.37 The C1 domain binds to DAG/phorbol ester, and this binding lowers the energy barrier for vesicle fusion and neurotransmitter release.38,39 Phorbol ester binds to the C1 domain with an affinity higher than that of DAG.40 The MUN domain is a rodlike module critical for opening syntaxin-1.31,41,42 A recent study proposed a model in which Munc13-1 bridges the plasma membrane and synaptic vesicle membrane by binding the plasma membrane with its C1 and C2B domains and binding the vesicle membrane with its C2C domain while MUN acts as a tether between these domains.43 There is a small region between the C2A and C1 domains, called the calmodulin-binging (CaMb) region, that is responsible for Ca2+-dependent short-term plasticity.44</p><p>The C1 domain of Munc13-1 is structurally homologous to the C1 domain of protein kinase C (PKC). The phorbol 13- acetate-bound PKCδ C1B domain is the only ligand-bound C1 domain structure known to date.45 Both PKCδ C1B [Protein Data Bank (PDB) entry 1PTR] and Munc13-1 C1 (PDB entry 1Y8F) have two membrane binding loops, two β-sheets, a short α-helix at the C-terminus, and two Zn2+ binding sites.45,46 Although the ligand-bound structure of the Munc13-1 C1 domain is not known, on the basis of its structural similarity with PKCδ C1B, one would expect the DAG/phorbol ester to bind inside a groove between the two loops as seen in PKCδ C1B. A comparison of the residues of the ligand binding sites reveals that there are several differences in the residues and the orientation of their side chains. Leucine and lysine in PKCδ are replaced by isoleucine and arginine in Munc13-1 at homologous sites 590 and 592, respectively. The orientations of side chains of these residues are different in Munc13-1 and PKCδ (Figure 1). While in Munc13-1 the side chain of Trp-588 occludes the DAG/phorbol ester site, it does not do so in PKCδ. Mutation of this Trp-588 to other residues reduced its phorbol ester binding and phorbol ester-induced membrane translocation properties.47 Homologous tryptophan in PKCδ C1B contributes to partitioning of the C1 domain into the membrane48 and binding to DAG rather than phorbol ester.49</p><p>Binding of DAG/phorbol ester to C1 domains and subsequent activation of the protein depend on several factors, such as the lipophilicity of the ligand, the nature of the residues that interact with the ligand, the lipid composition, and other lipid-interacting domains.50 In our previous study, we reported that the C1 domain of Munc13-1 with a net negative charge showed maximum phorbol ester binding at 20% phosphatidylserine (PS), unlike PKCδ C1B and PKCθ C1B with a net positive charge that showed maximum phorbol ester binding at 100% PS.47,51</p><p>In this study, we investigated the role of Trp-588, Ile-590, and Arg-592 in DAG- and phorbol ester-induced membrane translocation of Munc13-1 in HT22 cells using immunoblotting and confocal microscopy. Solution nuclear magnetic resonance (NMR) spectroscopy was used to monitor the interactions of the C1 domain with ligands in a residue-specific manner. Our results highlight the importance of Trp-588 in ligand-independent membrane partitioning and suggest electrostatic interactions as an additional important factor in the recognition of the membrane-embedded ligand.</p><!><p>1,2-Dioctanoyl-sn-glycerol (DOG) and 2-dihexanoyl-sn-glycero-3-phospho-l-serine (DPS) were obtained from Avanti Polar Lipids (Alabaster, AL). Phorbol 12-myristate 13-acetate (PMA) was purchased from LC Laboratories (Woburn, MA). Phorbol-12,13-dibutyrate (PDBu) was obtained from Sigma-Aldrich (St. Louis, MO). Deuterated dodecylphosphocholine (d38-DPC), DMSO (d6-DPC), and 15NH4Cl were obtained from Cambridge Isotope Laboratories, Inc. (Tewksbury, MA). Protein quantification was conducted using the Bradford protein assay from Bio-Rad (Hercules, CA) and the BCA (bicinchoninic acid) protein assay kit from Thermo Fisher Scientific (Grand Island, NY). Glutathione sepharose 4B was obtained from GE Healthcare Life Sciences (Piscataway, NJ). HT22 cells were purchased from ATCC (Manassas, VA). Fetal bovine serum (FBS) was from ZenBio (Research Triangle Park, NC). Rattus norvegicus Munc13-1 conjugated with green fluorescent protein (GFP) was a generous gift from N. Brose (Max Planck Institute for Experimental Medicine, Gottingen, Germany). The rabbit anti-GFP antibody, rabbit anti-Na/K-ATPase antibody, rabbit anti-β-actin, and HRP (horseradish peroxidase)-conjugated rabbit anti-IgG antibody used for the Western blot analysis were obtained from Cell Signaling (Danvers, MA). All other reagents were purchased from either MilliporeSigma (Burlington, MA) or Thermo Fisher Scientific.</p><!><p>The W588A, I590A, and R592A single-point mutations in Munc13-1 (full-length) were generated by site-directed mutagenesis at Epoch Life Science (Sugar Land, TX) using the Munc13-1-pEGFP-N1 vector as the template. The corresponding mutations in the isolated C1 domain were introduced using the pGEX-2TK vector for expression as glutathione S-transferase (GST)-tagged proteins. The mutations were verified for their correct sequences using the sequencing facility at Epoch Life Science. The gene sequencing chromatograms of the mutants were analyzed using SnapGene Viewer (GSL Biotech LLC, San Diego, CA).</p><!><p>The HT22 hippocampal neuronal cell line was used for protein expression and membrane translocation studies. The HT22 cells were cultured in Dulbecco's modified Eagle's medium (DMEM) containing 10% FBS, l-glutamine (2 mM), streptomycin (100 μg/mL), and penicillin (100 units/mL) at 37 °C in a humidified atmosphere supplemented with 5% CO2. Cells were plated on six-well plates for Western blotting (1.0 × 106 cells per well) or 12-well plates with 12 mm glass coverslips (VWE, Atlanta, GA) for confocal analysis (2.0 × 105 cells per well). Once they were 75−85% confluent, the cells were transfected with either pEGFPN1-Munc13-1WT, pEFGPN1-Munc13-1W588A, pEFGPN1-Munc13-1I590A, or pEFGPN1-Munc13-1R592A using Lipofectamine 3000 and P3000 reagent as per the manufacturer's instructions. The optimized ratio of the reagents was 1 μg of DNA to 1 μL of Lipofectamine 3000 to 1 μL of P3000 reagent. The growth medium deficient in antibiotics was used during transfection.</p><!><p>Proliferative HT22 cells expressing wild-type or mutant GFP-tagged Munc13-1 were treated with DOG (100 and 250 μM) for 15 min or PMA (1.0 and 5.0 μM) for 5 min. Subcellular fractionation was carried out using the subcellular fractionation kit (catalog no. 78840, Thermo Scientific Inc., Rockford, IL) following the manufacturer's recommendations. The kit extracts cytoplasmic and membrane fractions efficiently for the studies of the localization and distribution of proteins. Briefly, treated cells were lysed in cytosolic extraction buffer (CEB) for 10 min at 4 °C. The lysate in CEB was the cytosolic fraction. The cytosolic fraction was collected, and the cells were washed and harvested in ice-cold phosphate-buffered saline (PBS) and then centrifuged at 800g for 5 min. The pellet was mixed with membrane extraction buffer (MEB) and then incubated at 4 °C for 15 min. The sample in MEB was centrifuged at 3000g for 5 min, and the supernatant was collected as the membrane fraction. The concentration of proteins in each isolated fraction was measured by using the BCA kit, and 8 μg of protein was used for sodium dodecyl sulfate−polyacrylamide gel electrophoresis (SDS−PAGE) and Western blot analysis. The samples were prepared by mixing with a Laemmli sample buffer (LSB). The membrane fraction samples were prepared without heating. The dilutions of the antibody were as follows: 1:1000 anti-rabbit GFP, 1:1000 anti-rabbit Na,K-ATPase, and 1:2000 anti-rabbit β-actin (Cell Signaling). The protein bands were observed using the ChemiDoc Touch Imaging System (Bio-Rad) using the goat anti-rabbit HRP-linked secondary antibody. The bands were detected and acquired in the linear intensity range. The band intensity was quantitated using ImageLab 6.0.1 (Bio-Rad).</p><!><p>HT22 cells were grown and transfected with the plasmid on 12 mm coverslips. After being treated with DOG (100 and 250 μM) or PMA (1.0 and 5.0 μM), the cells were washed with PBS and then fixed with 4% paraformaldehyde (PFA) for 10 min. The coverslips containing cells were mounted on microscope slides with mounting medium. The cells (GFP-tagged Munc13) were excited with an argon laser at 488 nm, and images were captured using a confocal microscope (100×, Leica SP8, Leica Microsystems). The distribution of Munc13-1 in the cytosol and the plasma membrane was quantified from confocal microscope images using ImageJ (http://rsb.info.nih.gov/ij/). The mean fluorescence intensities of the whole cell and plasma membrane of individual images were measured. The size of the plasma membrane was defined as 300 nm from the outer edge of the cell. The ratios of the mean intensity of the plasma membrane to the whole cells are presented as described previously.52,53</p><!><p>The pGEX-2TK vector (GE Healthcare Bio-Sciences, Pittsburgh, PA) was used to express the GST-conjugated C1 domain of Munc13-1. Munc13-1 C1 WT and its mutant, W588A, I590A, and R592A, plasmids were transformed into Escherichia coli BL-21(DE3) competent cells. The transformants were grown in LB medium at 37 °C for expression of the desired proteins until the optical density of the cell suspension reached 0.5−0.6. Isotope labeling of Munc13-1 WT and its mutants was carried out by using the method of Marley et al.54 Briefly, the cells grown on LB medium were resuspended and grown in M9 minimal medium, supplemented with 1 g/L 15NH4Cl. Protein expression in M9 minimal medium was induced with 0.5 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) and 10 μM ZnSO4 at an optical density of 1.5−1.7, and the cells were grown at 15 °C for an additional 15−16 h. The cells were harvested and lysed in lysis buffer (PBS, 1 mg/mL lysozyme, and 0.1% Triton X-100) with sonication (15 times, 5 s, 30% amplitude) using a digital sonifier (model 250, Branson, Danbury, CT). Soluble protein was recovered by centrifugation at 7000g and 4 °C for 10 min, and the soluble extract was further treated with polyethylenimine (PEI) (0.1%) to remove nucleic acids. Extracted lysates were loaded onto a glutathione-sepharose column pre-equilibrated with PBS, and the column was washed with PBS until maximum amounts of impurity was eliminated. Protein was cleaved from its conjugated GST with thrombin on the column, and the protein was eluted in PBS. Eluted protein was further purified using a size exclusion column (Sephadex 75 10/300 GL, GE Life Science, Madison, CA) with gel filtration buffer [40 mM HEPES, 150 mM NaCl, and 50 μM ZnCl2 (pH 7.0)]. The purity of the protein was examined by SDS−PAGE analysis followed by Coomassie staining (Instant Blue, Expedeon, San Diego, CA).</p><!><p>After purification, Munc13-1 C1 and its variants were concentrated and exchanged into an "NMR buffer" containing 40 mM HEPES (pH 7.0), 150 mM NaCl, 50 μM ZnCl2, and 8% D2O. A stock solution of mixed micelles was prepared by combining chloroform solutions of DPS and d38-DPC at a molar ratio of 3:7. Chloroform was removed under a mild stream of nitrogen gas followed by 2 h under vacuum. The DPS/DPC detergent film was resuspended in NMR buffer and vortexed for 1 min to form a clear micellar solution. Stock solutions of DOG and PDBu were prepared in d6-DMSO. The concentration of DMSO at the titration end point was <8% (v/v).</p><!><p>The NMR experiments were carried out at 25 °C on either an 11.7 T AVANCE III HD or a 14.1 T (both AVANCE III HD and NEO) NMR instrument equipped with room-temperature (11.7 T) and cryogenically cooled (14.1 T) probes. The temperature was calibrated with methanol. The NMR-detected binding experiments were carried out by adding stock solutions of micelles to 100 μM [U-15N]C1 or stock solutions of an agonist (DOG or PDBu) to the mixture of 100 μM [U-15N]C1 and 20 mM DPS/DPC micelles. The binding was monitored as a change in protein cross-peak position in a series of two-dimensional 15N−1H HSQC/SOFAST-HMQC spectra at different ligand concentrations. All NMR spectra were processed with NMRPipe55 and analyzed with SPARKY (https://www.cgl.ucsf.edu/home/sparky/). Residue-specific binding curves were constructed by plotting the combined chemical shift perturbation Δδ=ΔδH   2+(0.152ΔδN)2 as a function of total ligand concentration L0. The apparent dissociation constant (Kd) was determined by fitting the binding curves globally to the single-site binding equation:56 Δδ=Δδmax{Kd+P0+L0−[(Kd+P0+L0)2−4P0L0]1/2}/2P0 where ΔδH and ΔδN are the chemical shift changes of the 1H and 15N nuclei, respectively, Δδmax is the maximum chemical shift change upon saturation, and P0 is the total concentration of C1.</p><!><p>Phorbol 13-acetate was docked into the WT C1 domain of Munc13-1 using AUTODOCK 4.2. The molecular docking was prepared, run, and analyzed using AutoDockTools (ADT). The structure of phorbol 13-acetate was prepared using ChemBioDraw version 12.0, and then energy minimization was performed for the ligand with a 0.1 root-mean-square kcal mol−1 Å−2 gradient level using MOE 2018 (MOE, Chemical Computing Group, Montreal, QC). The NMR structure of the Munc13-1 C1 domain (PDB entey 1Y8F) was used for the molecular docking.46 Gasteiger charges and Kollman charges were given to the ligand and the proteins, respectively. For the docking space of the ligand, a grid box was generated and centered on the geometry of the phorbol 13-acetate in phorbol 13-acetate-bound PKCδ C1B (PDB entry 1PTR), which is the homologue of Munc13-1 C1.45 The grid box included the residues of the active site between the two loops of C1 within a box size set at x = 40 Å, y = 55 Å, and z = 40 Å. The best 20 conformers were determined using the Lamarckian genetic algorithm (LGA) during the molecular docking. The volume of the binding pocket of PKCδ C1B and Munc13-1 C1 was measured using MOE. The phorbol 13-acetate-docked structures were visualized using Discovery Studio Visualizer 4.5 (DS, Biovia Inc., San Diego, CA).</p><!><p>Statistical analysis of data was conducted using Prism 5.0 software (GraphPad Software, Inc., San Diego, CA). At least three independent experiments were used for all of the statistical analysis. Raw data from DOG and PMA treatments of HT22 cells were first analyzed using two-way analysis of variance (ANOVA), and then Bonferroni's multiple-comparison post hoc tests were performed to compare all treatment groups. Differences at the P < 0.05 level were considered significant.</p><!><p>Munc13-1's presynaptic function is dependent on its activation by DAG, which binds to its C1 domain. Although the structure of Munc13-1 C1 has been elucidated,46,57 the activator-bound structure is not known at present. To identify the residues of the C1 domain in activator binding, we used the activator-bound structure of PKCδ C1B (PDB entry 1PTR), a structural homologue of Munc13-1 C1, as a template. A comparison of the sequence and structures of ligand binding sites of the Munc13-1 C1 domain (PDB entry 1Y8F) and PKCδ C1B (PDB entry 1PTR) suggests that there are several differences in the residues and the orientation of their side chains. This includes residues Trp-588, Ile-590, and Arg-592 (Figure 1B) and the orientation of their side chains (Figure 2).</p><p>Molecular docking results suggest phorbol 13-acetate docked into the active site of Munc13-1 between the two loops and Trp-588, Ile-590, and Arg-592 form the phorbol ester binding site (Figure 2A). The phorbol ester formed three hydrogen bonds, one each with Trp-588, Ile-590, and Arg-592. The docking site is similar to those of our previous docking studies for resveratrol and bryostatin-1.58,59 However, in PKCδ C1B, the phorbol 13-acetate formed three hydrogen bonds with Thr-242 and Leu-251 that are located deep inside the binding pocket (Figure 2B). The phorbol 13-acetate binding sites of Munc13-1 C1 and PKCδ C1B (PDB entry 1PTQ) are slightly different because of the different orientation of Trp-588. While Trp-588 of Munc13-1 is oriented between the two loops, in PKCδ C1B the homologous Trp-252 resides outside of the two loops. This particular orientation of Trp-588 formed an alternative shallow pocket in Munc13-1 C1 instead of a deep pocket present in PKCδ C1B (Figure 2 and Figure S1). Indeed, the phorbol 13-acetate could adhere closely to PKCδ C1B rather than to Munc13-1 C1. This is because PKCδ C1B has a pocket volume of 88.4 Å3, while the volume of the pocket in Munc13-1 C1 is 33.4 Å3 (Figure S1). The pocket surface area of PKCδ C1B (111.3 Å2) is also ~2 times larger than that of Munc13-1 C1 (50.3 Å2). The ligand binding pocket of Munc13-1 C1 is surrounded by Trp-588, Ile-590, and Arg-592 (Figure 2A). To understand the role of these residues in the activity and ligand binding, we generated I590A, R592A, and W588A mutants in full-length Munc13-1 and in the isolated C1 domain. For the full-length protein, we performed functional assays using immunoblotting and confocal microscopy, and for the isolated C1 domain, we conducted NMR-detected binding studies.</p><!><p>On the basis of our docking results, we identified three residues, Trp-588, Ile-590, and Arg-592, in the C1 domain of Munc13-1 that could potentially interact with the ligand. To understand the role of these three residues in ligand-induced membrane translocation, three mutants, W588A, I590A, and R592A, were generated and expressed as GFP-conjugated full-length Munc13-1 in HT22 cells. The cells were treated with either DOG for 15 min or PMA for 5 min and imaged by confocal microscopy (Figure 3). DOG induced the translocation of wild-type Munc13-1 from the cytosol to the plasma membrane (Figure 3A). However, the extent of translocation of the mutants was significantly lower than that of wild-type Munc13-1 at both 100 and 250 μM doses. The extent of translocation decreases in the following order: WT > I590A > W588A > R592A. The difference between W588A and R592A was not significant. PMA also induced the translocation of wild-type Munc13-1 to the plasma membrane (Figure 3B). The 1.0 and 5.0 μM PMA treatment groups showed 5 and 7 times more translocation, respectively, than the no treatment group. The extent of translocation decreases in the following order for either PMA concentration: WT > W588A > I590A > R592A. The order of translocation for the wild type and the mutants, however, was slightly different than what was found for DOG. The reduced level of translocation of W588A compared to that of the wild type is similar to the result using living cell images reported in our previous study.47 For the wild type, there was a difference in translocation at two different concentrations of PMA, but this difference was not prominent for the mutants. For the mutants, there were significant differences in translocation between the PMA treatment and control groups, but there was not much difference in the mutants between the DOG treatment and control groups.</p><p>The relative membrane translocation of the wild type and the mutants was also analyzed by cell fractionation and immunoblotting. Like the confocal analysis results, the results of immunoblotting analysis also showed a decrease in the extent of translocation of mutants compared to that of wild-type Munc13-1 for either DOG or PMA (Figure 4). For DOG, the mutants showed significantly less translocation at both 100 and 250 μM compared to the wild type (Figure 4A). At 250 μM, the extents of translocation of W588A, I590A, and R592A were 43%, 57%, and 39% of that of the wild type, respectively. The extent of translocation decreased in the following order: WT > I590A > W588A > R592A. This order is similar to the order obtained in the confocal analysis for DOG treatment groups. At 5.0 μM PMA, the translocation levels of W588A, I590A, and R592A were 61%, 40%, and 43% of that of the wild type, respectively. Overall, the extent of translocation decreases in the following order: WT > W588A > I590A > R592A. This order of translocation is different than the order found in the case of the DOG treatment group but follows the same order as determined with the confocal analysis of the PMA treatment group. Only at 1.0 μM was there no significant difference in the extent of translocation between W588A and WT.</p><p>In summary, both DOG and PMA induced translocation of Munc13-1 from the cytosol to the plasma membrane and PMA was more potent than DOG. Mutation of Trp-588, Ile-590, and Arg-592 to alanine caused a decrease in the extent of translocation, and the extent of this reduction was dependent on the ligand.</p><!><p>The translocation studies described above were conducted with full-length Munc13-1 in a cellular context. We used solution NMR spectroscopy of the isolated C1 domain to gain residue-specific insights into the activator−protein interactions.</p><p>We first sought to evaluate the propensity of the Munc13-1 C1 domain to partition into a hydrophobic environment in a ligand-independent manner. This experiment mimics a scenario in which C1 is initially recruited to the membranes but is not yet ligand-bound. The binding of the U-15N-enriched Munc13-1 C1 domain to DPC/DPS (70:30) mixed micelles was characterized by two-dimensional (2D) solution NMR spectroscopy. The suitability of mixed micelles as a membrane mimic and successful application of NMR spectroscopy to probe C1 interactions have been demonstrated by us previously for the C1B domains of PKCα and PKCδ.48,60 The partitioning of C1 domains from the aqueous state into a hydrophobic environment causes chemical shift perturbations (CSPs) of the amide (NH) cross-peaks in the 2D 15N−1H HSQC spectra. This provides us with the means to conduct residue-specific analysis of these interactions. The binding of the Munc13-1 C1 domain to micelles falls into the "fast" exchange regime on the NMR chemical shift time scale (Figure 5A). The binding curves were constructed by plotting the CSP values as a function of detergent concentration. The data were fitted with the single-site binding equation to obtain an apparent dissociation constant (Kd) of 104 μM (Figure 5B). These results indicate that the Munc13-1 C1 domain possesses a high degree of amphiphilicity and can partition into the hydrophobic environment in a ligand-independent manner. The residues that experienced the most chemical shift perturbations belong to the loop regions that form the binding groove (Figure 6A).</p><!><p>The next step was to characterize the interactions of the C1 domain with diacylglycerol. The NMR structural ensemble of the Munc13-1 C1 domain (PDB entry 1Y8F) determined by Rizo's laboratory shows that the binding groove is occluded by the Trp-588 side chain.46 This led to the hypothesis that considerable conformational rearrangement (necessary to remove the occlusion) would decrease the affinity of C1 for diacylglycerol.46 To quantify the diacylglycerol affinity, we conducted NMR-detected DOG binding experiments in the presence of DPC/DPS (70:30) micelles. The binding of DOG to C1 was in the fast-to-intermediate exchange regime (Figure 5C). Nearly all loop residues were in the intermediate regime and hence broadened beyond detection (Figure 6B). The CSPs of the fast-exchanging residues were used to construct the binding curves, whose analysis produced an apparent dissociation constant of 118 μM for the C1−DOG interactions (Figure 5D). This affinity is ∼600-fold weaker than that of PKCδ C1B that was measured under similar experimental conditions.48 The low affinity of the Munc13-1 C1 domain for diacylglycerol supports the Trp-588 occlusion model proposed by Rizo's group.46</p><p>Diacylglycerol is a weaker C1 ligand compared to non-endogenous phorbol ester PDBu, a known tumor-promoting agent. To determine how the W588 occlusion influences C1−PDBu interactions, we first added a stoichiometric amount of PDBu to U-15N-enriched Munc13-1 C1 in the presence of 20 mM DPC/DPS (70:30) mixed micelles. Large CSPs were observed in the spectra for many cross-peaks (Figure 7). Addition of PDBu to a 1:2 ratio did not cause any further perturbations in the spectra, suggesting that the formation of the high-affinity C1−PDBu complex takes place at stoichiometric protein and ligand concentrations. This is also consistent with the notion that the Munc13-1 C1 domain is a cellular target for high-affinity phorbol ester binding.40 The stoichiometric binding of PDBu to C1 precludes the assignment of resonance identities in the PDBu-bound state based on the 2D spectra. However, we were able to identify several residues of the loop region that were perturbed significantly (Figure 7A, Trp-572/588, Thr-576, and Gly-585). The effect of PDBu binding was particularly prominent for the Nε−Hε resonance of the Trp-588 side chain (Figure 7B). In addition, two residues that reside on the α-helical region of the domain, His-605 and Glu-606, showed significant perturbations. Overall, the CSP data show a major change in the electronic environment of the ligand binding loops and Trp-588 side chain. These findings are consistent with Rizo's model46 that involves a conformational change due to the displacement of the Trp-588 side chain from the binding groove.</p><!><p>It is proposed that Munc13 exists in the autoinhibited state akin to the DAG-sensitive PKC isoforms.38,61 Both C1 and C2B domains play a role in the inhibition of synaptic transmission by Munc13.61 Interactions of C1 with the membrane-embedded DAG are believed to facilitate the release of such an inhibitory state. Therefore, understanding how the Munc13-1 C1 ligand binding region partitions into the hydrophobic environment, where it recognizes and captures its ligands, is essential. At the atomistic level, this process could be driven by two types of conserved C1 residues45,62,63 that have (1) aromatic and hydrophobic side chains that engage in the interactions with apolar lipid moieties and (2) charged side chains that guide the domain to anionic lipid headgroups. These factors were investigated sequentially with NMR using three Munc13-C1 variants: W588A, I590A, and R592A.</p><p>The W588A 2D 15N−1H SOFAST-HMQC spectrum revealed significant perturbations of the loop residues compared with the spectrum of the wild-type Munc13-1 C1 domain (WTC1) (Figure 8A). This is expected due to the occluding orientation of the Trp-588 side chain in the apo state, and the contacts made with the other loop residues: Ala-574, Thr-575, Thr-576, and Arg-592. The perturbation of Glu-606, a residue at the hinge of the helix away from the loops, suggests that the W588A mutation may have an allosteric effect on the other regions of the domain. The overall chemical shift dispersion of the W588A resonances indicates that the mutation does not affect the domain fold.</p><p>The tryptophan residue at the analogous position of the PKCδ C1B domain (Trp-252) is essential for the high-affinity PKCδ−DAG interactions48,49,64 and has been described as the "DAG toggle". In Munc13-1 C1, Trp-588 is also essential for DAG binding.46 To dissect the specific role of Trp-588 in membrane partitioning and DAG recognition steps, we conducted NMR-detected micelle and DOG binding experiments. In sharp contrast to WTC1, addition of DPC/DPS mixed micelles to the W588A variant caused no detectable CSPs (Figures 8B and 5C), indicating that the mutation abrogates the C1−micelle interactions. Inclusion of DOG in the micelles produced only minor perturbations indicating negligible binding (Figure 8B). We conclude that the extremely low affinity of W588A for DAG, observed in CHAPS46 and in our membrane-mimicking system, is due to the inability of this variant to partition into the hydrophobic environment. As a result, membrane-embedded DAG becomes inaccessible to the protein and no productive protein−ligand complex is formed. Our NMR data clearly establish the pivotal role of Trp-588 in the membrane recruitment of the Munc13-1 C1 domain.</p><p>Next, we acquired the 2D 15N−1H SOFAST-HMQC spectra of the I590A and R592A variants and compared them with that of WTC1 (Figure S2). A large chemical shift dispersion of resonances for both variants indicates the preservation of the three-dimensional fold (Figure S2A,B). The extent of CSPs was less prominent for these mutants than for W588A, allowing a reliable transfer of resonance assignments from WTC1. The CSP analysis using WTC1 as a reference shows that both mutations perturb the loop regions (Figure S2C,D). The nature of these perturbations could be structural and/or dynamical in origin. In the I590A variant, the side chain of Trp-588, which belongs to loop 2, is relatively unperturbed. However, the backbone N−H resonances of the loop 2 region show significant perturbations. The isoleucine to alanine mutation preserves the hydrophobic nature of the side chain, and the Ile-590 side chain does not appear to interact with any residues in the NMR ensemble of WTC1. It is therefore possible that small changes in the hydrophobicity of this region have allosteric effects on the overall interaction network of the loops. Further studies are needed to explore these possibilities.</p><p>In the R592A variant, the side chain Nε−Hε resonance of Trp-588 is broadened beyond detection. Given the proximity of the Arg-592 side chain to the indole ring of Trp-588 in the NMR ensemble (Figure S2, inset),46 it is plausible that they can be engaged in cation−π interaction. The disruption of this interaction due to alanine mutation could alter the loop dynamics and thereby result in the broadening of the Trp-588 Nε−Hε resonance.</p><p>We tested the effect of these mutations on interactions of C1 with micelles and DOG (Figure 9). Unlike W588A, both I590A and R592A variants partition readily into the DPC/DPS (70:30) mixed micelles as reported by the CSPs (Figure 9A,C). Changes induced by the additions of DOG were monitored by collecting a series of 2D 15N−1H SOFAST-HMQC spectra at each concentration point. Similar to WTC1, the loop resonances showed chemical shift changes (Figure 9A,C) in the intermediate-to-fast exchange regime for both variants. Plotting these changes as a function of DOG concentration (Figure 9B,D) produced apparent dissociation constants of 200 μM (I590A) and 390 μM (R592A). Compared to that of WTC1, these affinities are ∼1.5-fold (I590A) and 3-fold (R592A) weaker. Because the Ile to Ala mutation preserves the hydrophobic nature of the position, the effect on DOG binding is expectedly smaller. In contrast, the R592A mutation neutralizes the positive charge, while concomitantly increasing the hydrophobicity of the loop region. A more significant effect of the R592A mutation on the DOG affinity suggests that electrostatics may substantially contribute to the formation of the ternary Munc13-1 C1−ligand−membrane complex.</p><p>Taken together, our experiments with Munc13-1 C1 mutants establish the significance of Trp-588 in ligand-independent membrane partitioning and suggest electrostatic interactions as an additional important factor in the recognition of the membrane-embedded ligand. Because membrane partitioning and ligand recognition in the C1 domains are influenced by the concerted actions of multiple residues, loop region dynamics may also play an important role in the mechanism of DAG sensing.</p><!><p>Munc13-1 acts as a master regulator of neurotransmitter release machinery in the brain7,14,65 and has been implicated in the pathophysiology of Alzheimer's disease66,67 and alcohol addiction.28,30 Knowledge of the activator binding site of Munc13-1 is expected to aid in developing Munc13-1 modulators for intervening in these disease states. Toward this end, we identified the core residues in the activator binding site of the Munc13-1 C1 domain and characterized their contribution to Munc13-1 activation and interactions with ligands using ligand-induced membrane translocation assays and NMR-detected binding experiments, respectively. Using molecular docking, we first identified Trp-588, Ile-590, and Arg-592 as potential ligand-interacting residues (Figure 2). To understand the role of these residues in the activity of Munc13-1, we generated the corresponding alanine mutants and measured their ligand-induced membrane translocation. The efficiency of DOG-induced membrane translocation decreased in the following order: WT > I590A > W588A > R592A. The efficiency of PMA-induced membrane translocation decreased in the following order: WT > W588A > I590A > R592A. This suggests that the translocation is dependent on the type of ligand.</p><p>In the isolated C1 domain, the apparent binding affinity of DOG for the purified C1 domain wild type and mutants in the DPS/DPC micelles decreased in the following order: WT > I590A > R592A (Figures 5 and 6). We did not detect any binding of W588A either with micelles or with DOG in the presence of micelles. This suggests that Trp-588 plays a more critical role in the binding of Munc13-1 C1 to the lipid membrane than that of Ile-590 and Arg-592. Trp-588 is known play a role in the interaction between the C1 domain and membrane through its hydrophobic side chain.47 Tryptophan can potentially engage in several types of interactions, such as cation−π bonds between its fused aromatic ring system and positively charged choline groups of the lipids and detergents,68 hydrogen bonds, N−H…π bonds, π-stacking, and C−H···π bonds.69 Substitution of Trp with Ala in W588A is expected to show weakened or no DOG binding.</p><p>As the most hydrophobic residue of the rim of the two loops in Munc13-1 C1, Ile-590 may also play a role in the hydrophobic interaction between the C1 domain and membrane (Figures S3 and S4). Thus, the DOG binding affinity of I590A was decreased 1.7-fold compared to the affinity in the wild type (Figures 5D and 9B). However, I590A showed a higher affinity for DOG than either W588A or R592A, because isoleucine and alanine do not differ significantly in chemical composition, polarity, or molecular volume according to Grantham's distance.70 The physico-chemical distances (calculated using the chemical composition, polarity, and molecular volume in an equation) between Ile and Ala, Trp and Ala, and Arg and Ala are 94, 148, and 112, respectively, suggesting that the Ile to Ala substitution will have a weaker influence on binding and/or activity as compared to Trp to Ala or Arg to Ala substitution.</p><p>Among the residues in the DOG binding site, Arg-592 is the most hydrophilic residue, and our NMR titration (Figures 6 and 9) and modeling (Figure S4 showing how arginine can be embedded into the lipid headgroup region in the membrane) results suggest that Arg-592 interacts with the lipid membrane and ligand. The guanidinium group of arginine can form multiple hydrogen bonds with the polar group of the lipid head and also can be located in the membrane at the depth of the hydrocarbon core with its hydrophobic aliphatic hydrocarbon chain.71 These properties of arginine may explain why it can penetrate into the lipid membrane.72 The positive charge of Arg-592 was neutralized in R592A, and therefore, its affinity for DOG in micelles was lower than those of the wild type and I590A (Figures 3, 4, and 9). Also in the membrane translocation assay, R592A showed the lowest activity because Arg-592 could strongly interact with the negatively charged phosphatidylserine in the cellular context, but Ala-592 could not. These results highlight the importance of electrostatic interactions in the ternary complex (Munc13-1 C1−ligand−lipid membrane) and the function of Mun13-1 as reported previously.47</p><p>The crystal structure of PKCδ C1B complexed with phorbol 13-acetate is the only ligand-bound C1 domain structure known to date.45 How different is the Munc13-1 activator binding site compared to PKCδ? In PKCδ C1B, several hydrophobic residues lining the two activator binding loops facilitate its binding to the plasma membrane.48,60,73 Once the phorbol ester or DAG binds to the rim of the two loops, it covers the hydrophilic region and enhances the interaction between the C1 domain and the plasma membrane.48,73 It was found previously that PKCδ C1B did partition into the DPS/DPC micellar system with a Kd value of 9 μM in the absence of any ligand.48 In the study presented here, Munc13-1 C1 bound to the lipid micelles and partitioned into the micelles with a Kd value of 104 μM in the absence of any ligand (Figure 5A,B). This ∼10-fold higher value of Kd for Munc13-1 could be due to the presence of fewer hydrophobic residues on the rim of two loops as compared to that in PKCδ C1B, judging the amino acid properties by the Kyte−Doolittle hydrophobicity scale (Figure S3).74 In PKCδ C1B, Met-239 forms a hydrophobic surface, but the homologous residue of Munc13-1 C1 (Thr-575) forms a hydrophilic surface. More-over, the Trp-588 residing inside the two loops causes a decrease in the hydrophobic surface area of the rim of the two loops. Furthermore, the positively charged Arg-592 of Munc13-1 C1 is exposed to the plasma membrane unlike the homologous Lys-256 of PKCδ C1B, which is oriented away from the membrane. The apparent Kd of DOG with respect to Munc13-1 C1 (118 μM) is ∼600-fold weaker than that of PKCδ C1B (<0.2 μM).48 Collectively, these data suggest that differences in both ligand-independent membrane partitioning and intrinsic DOG affinities contribute to the differential DOG responses of Munc13-1 and PKCδ.</p><p>Another major difference between Munc13-1 C1 and PKCδ C1B is the orientation of Trp-588. In Munc13-1, Trp-588 occludes the ligand binding site as evidenced by both the solution46 and the crystal structure,57 while in PKCδ C1B, the homologous Trp-252 resides outside the ligand binding loop. The Rizo group suggested46 that this particular orientation could be the reason why Munc13-1 C1 has a lower affinity for DOG compared to that of PKCδ C1B. Now, does Trp-588 change its orientation when the ligand binds to Munc13-1 C1? A structural analysis of the solution structure of Munc13-1 (PDB entry 1Y8F) suggests that there are amide-π stacked hydrophobic interactions between Trp-588 and Thr-576 and π-alkyl hydrophobic interactions between Trp-588 and Pro-577, between Trp-588 and Ala-574, and between Trp-588 and Arg-592 (Figure S5). A molecular dynamics simulation (80 ns) of Munc13-1 C1 in the absence of a ligand and the membrane showed minimal movement of Trp-588, unlike the homologous tryptophan in PKCδ C1B that undergoes substantial rotameric flips.47 In the C1B domains of PKCα and PKCδ, the loop regions are highly dynamic as revealed by the NMR studies.60,75 The lack of Trp-588 displacement in the MD data might indicate that the dynamics of the Munc13-1 C1 loop region is on a slower time scale. The docked C1−phorbol 13-acetate complex may therefore represent initial interaction of the C1 domain with the ligand. A rotation of Trp-588 out of the loop region likely follows, allowing full accommodation of the ligand in the binding groove. Large chemical shift perturbations of the Nε−Hε resonance observed in the NMR spectra of the PDBu-complexed C1 domain, and the intermediate exchange regime of the Nε−Hε resonance in the DOG-complexed spectra (Figure 7B), suggest that significant changes in the electronic environment of the Trp side chain take place as a result of ligand interactions. This is consistent with the Trp side chain movement upon ligand binding.</p><p>The exact orientation of the Trp side chain in the complex may also depend on the chemical identity of the ligand. The C1 domain-interacting moiety of phorbol esters is ∼3 times larger than that of DOG (excluding the amphiphilic chains that interact with hydrophobic lipid tails of the plasma membrane) and contains four rings that are conformationally restrained. The number of potential hydrogen bond donors and acceptors is also different for the two ligands. The differences in the protein−ligand van der Waals interactions and hydrogen-bonding patterns may therefore dictate the Trp side chain orientation and underlie the functional response (Figures 3 and 4). High-resolution structures of agonist-bound Munc13-1 C1 domain complexes are required to determine the extent of the binding site arrangement that takes place in the C1 domain of Munc13-1. If such information is obtained, the distinctive structural features of the Munc13-1 activator binding site could be exploited in designing a selective modulator of Munc13-1, one of the 70 different C1 domain-containing proteins.</p><p>Our results showing that the DOG-induced activity of the full-length Munc13-1 mutants, particularly W588A, does not exactly correlate with the binding affinity of the isolated C1 domain suggest that domain−domain interaction may play a role in the DOG-induced activity of Munc13-1. A recent structure of the C1, C2B, and MUN domains of Munc13-1 (PDB entry 5UE8) revealed that the C1 and C2B domains are in the proximity of each other and the surface residues, Tyr-581, Glu-582, and Glu-584, of the C1 domain interact with the surface residues, Gln-717, Arg-750, Lys-752, and Arg-754, of the C2B domain through hydrogen bonds and electrostatic interactions. The C2B domain also binds to the plasma membrane following Ca2+ binding.76 Therefore, Ca2+-bound C2B could influence the C1−ligand−plasma membrane interactions in the full-length protein, but not in the isolated C1 domain.</p><p>In summary, our study characterized Trp-588, Ile-590, and Arg-592 as the ligand binding residues of Munc13-1 and found that DAG and phorbol ester could interact differently with these residues. Further time-resolved structural studies are required to understand the relative orientation of these residues in the membrane in the presence of various ligands.</p>

PubMed Author Manuscript

Endothelial MAP2K1 Mutations in Arteriovenous Malformation Activate the RAS/MAPK Pathway

Arteriovenous malformation (AVM) is a locally destructive congenital vascular anomaly caused by somatic mutations in MAP2K1. The mutation is isolated to endothelial cells (ECs). The purpose of this study was to determine the effects of mutant MAP2K1 on EC signaling and vascular network formation. Pathway effects were studied using both mutant MAP2K1 (K57N) human AVM tissue and human umbilical vein endothelial cells (HUVECs) engineered to overexpress the MAP2K1 (K57N) mutation. Western blot was used to determine cell signaling along the RAS/MAPK pathway. Geltrex tube formation assays were performed to assess EC vascular network formation. Cells were treated with a MAP2K1 inhibitor (Trametinib) to determine its effect on signaling and vascular tube formation. Human mutant MAP2K1-AVM ECs had similar baseline MEK1 and ERK1/2 expression with controls; however, mutant MAP2K1-AVM ECs produced significantly more phosphorylated ERK1/2 than wild-type ECs. Mutant MAP2K1 HUVECs demonstrated significantly more phosphorylated ERK1/2 than control HUVECs. Trametinib reduced the phosphorylation of ERK1/2 in mutant cells and prevented the ability of ECs to form vascular networks. AVM MAP2K1 mutations activate RAS/MAPK signaling in ECs. ERK activation and vascular network formation are reduced with Trametinib. Pharmacotherapy using MAP2K1 inhibitors may prevent the formation or progression of AVMs.

endothelial_map2k1_mutations_in_arteriovenous_malformation_activate_the_ras/mapk_pathway

INTRODUCTION<!>Specimen collection<!>Plasmids<!>Cell Lines<!>Trametinib Treatment<!>Western Blot Analysis<!>Endothelial Cell Tube Formation Assay<!>Statistical analysis<!>RESULTS<!>DISCUSSION

<p>Arteriovenous malformation (AVM) is a congenital vascular anomaly consisting of abnormal connections between arteries and veins through either a fistula or nidus instead of a normal capillary bed. Extracranial lesions enlarge and cause deformity, pain, ulceration, bleeding, and occasionally heart failure. We previously showed that most sporadic extracranial AVMs contain a mutation in MAP2K1 that is isolated to the endothelial cell (EC) [1]. MAP2K1 mutations in non-ECs may cause neoplasms (melanoma, lung, hematopoietic) and can increase MEK1 activity [2-11]. The effects of MAP2K1 mutations on EC function and AVM formation, however, are unknown. The purpose of this study was to determine if the most common AVM MAP2K1 mutation, MAP2K1-K57N, influences EC signaling and whether or not a MEK inhibitor affects mutant cells. Understanding how MAP2K1 mutations alter EC biology will provide insight into mechanisms by which AVMs form and enlarge.</p><!><p>The Committee on Clinical Investigation approved this study and informed consent was obtained. AVM tissue was collected during a clinically-indicated procedure and processed to separate endothelial cells (ECs) from non-ECs as we have previously described [1]. AVM specimen was washed in PBS to remove blood cell contaminants, digested with collagenase A (2.5 mg/mL) (Roche) for 1 hour at 37°C, then filtered through a 100 μm strainer to produce a single cell suspension. Cells were placed on fibronectin-coated (1 μg/cm2) tissue culture plates (Olympus Plastics) in endothelial growth medium-2 (EGM-2, PromoCell) supplemented with 10% fetal bovine serum (FBS, Gibco, Life Technologies). After 5-7 days of expansion, cells were fractionated into 2 populations (endothelial and non-endothelial) using anti-human CD31 (endothelial cell marker) magnetic beads (DynaBeads™, Life Technologies). DNA was extracted from each cell population using the DNeasy Blood & Tissue kit (Qiagen) and the mutant allele frequency (MAF) was determined using ddPCR as previously described [1]. CD31+ ECs and CD31- non-ECs were grown in endothelial cell growth medium (PromoCell) and mesenchymal stem cell growth medium (Lonza), respectively.</p><!><p>A pReceiver-M14 plasmid containing the human MAP2K1 ORF was obtained from Genecopoeia (Cat# EX-A0826-M14). The K57N mutation was introduced into the MAP2K1 ORF using the Agilent QuikChange II XL site directed mutagenesis kit (Cat# 200521) (FP: 5'-ctttcttacccagaatcagaaggtgggagaac RP: 5'-gttctcccaccttctgattctgggtaagaaag-3'). pLVX-puro-IreszsGreen1 lentiviral overexpression plasmids for wild-type and K57N MAP2K1 were generated using Infusion HD (Takara) in combination with a high fidelity DNA polymerase (Clone Amp, Takara). 3 PCR fragments were generated: a pLVX-Puro backbone (FP: 5'-gtcgacggtaccgcgggcccgggatc-3'; RP: 5'-tgcagaattcgaagcttgagctcg-3'; template pLVX-Puro (Takara), a MAP2K1 ORF (FP: 5'-gcttcgaattctgcatccaaaatgcccaagaagaagccgacgcccatc-3'; RP: 5'gagaggggttagacgccagcagcatgggttg; template: pReceiver-M14-MAP2K1 (wild-type or K57N)) and an IRES-zsGreen1 reporter ORF (FP: 5'-cgtctaacccctctccctcccccccccctaac-3'; RP: cgcggtaccgtcgactcagggcaaggcggagccggag-3'; template: pLVX-IRES-zsGreen1 (Takara)). The pLVX-puro-zsGreen1 empty vector control was generated with Infusion HD merging 2 PCR fragments: a pLVX-Puro backbone (FP 5'- gaattctgcagtcgacggtaccgcg-3' RP: 5'-gaattcgaagcttgagctcgagatc-3'; template pLVX-Puro) and an IRES-zsGreen1 reporter ORF (FP: 5'- ctcaagcttcgaattccccctctccctcccccccccctaac −3'; RP: 5'-gtcgactgcagaattctcagggcaaggcggagccggag; template: pLVX-IRES-zsGreen1 (Takara). Lentiviral stocks were generated by the Massachusetts General Hospital Viral Vector Core.</p><!><p>Endothelial colony forming cells (ECFCs) were isolated from human white adipose tissue as previously described [12]. ECFCs were cultured in fibronectin-coated flasks or plates maintained in EGM2 complete medium (EGM2 + endothelial growth supplement (PromoCell) + 20% FBS). Human umbilical vein endothelial cells (HUVECs) were obtained from ThermoFisher Scientific (Cat # C01510C) and were cultured in fibronectin-coated flasks or plates and maintained in EGM2 complete medium. The HUVEC-pLVX, HUVEC-pLVX-WT and HUVEC-pLVX-K57N cell lines were generated using lentiviral infection. 300,000 HUVECs were plated in one well of a 6-well plate. The next day the medium was replaced with 3 ml of EGM2 complete medium containing 8 μg/ml Hexadimethine Bromide (Sigma). 2.4 μl of lentiviral preparation [2.0 ×109 IU (Infectious Units)/mL] was added to the wells. 20 hrs later the lentivirus containing medium was replaced and cells were cultured for another 72 hours. Selection of infected cells was then performed using EGM2 complete medium containing 2 μg/mL puromycin (Invitrogen). On reaching confluency, HUVECs were passaged to a 25 cm2 flask, kept under puromycin selection, and allowed to reach confluency. Cells then were collected using 0.25% Trypsine-EDTA and sorted for zsGreen1 expression (Green fluorescent). Cells expressing zsGreen1 were plated in a 25 cm2 flask, kept in culture under puromycin selection, and expanded to a 75 cm2 flask. A fraction of the cells then was used to analyze the overexpression of wild type and K57N mutant MAP2K1, while the remainder of the cells were stored in liquid nitrogen. These cells were designated passage 1 and were not used later than passage 5 for our experiments.</p><!><p>Patient-derived AVM MAP2K1 ECs (MAF of 39%) and MAP2K1 engineered HUVECs were seeded on fibronectin coated dishes at 10,000 cells/cm2 in complete growth medium. After 24 hours, cells were incubated for 18 hours with DMSO (vehicle, 1:1000 in complete growth medium) or Trametinib (SelleckChem) at concentrations of 0.1μM, 1μM, or 10μM. Cells were lysed in the culture dish on ice using mammalian lysis buffer (Promega) containing protease and phosphatase inhibitor (Roche) for 10 minutes. Experiments were repeated 3 times.</p><!><p>Protein concentration was determined using a Pierce BCA Protein Assay Kit (Thermo Scientific). 15 μg of protein was separated on 4-20% gradient SDS-PAGE gels (Biorad; cat#: 456-1093). Because the protein components of the MAP2K1 signaling pathway (MEK1, ERK, P-ERK) and the loading controls (GAPDH and ACTIN) all have a similar molecular weight, a master loading mix was made for each sample. 20 μl (containing 15 μg protein) of this mastermix was loaded on 4 separate gels. The gels were transferred to a PVDF membrane (Invitrolon 0.45 μm pore; Life Technologies) and each membrane was detected using an antibody against one specific protein. Membranes were blocked in TBST containing 5% non-fat dry milk for 30 minutes at room temperature. Primary antibodies were diluted in blocking buffer and incubated with the membranes for 1 hour at room temperature. Membranes were washed 3 times for 5 minutes each with TBST buffer. Alkaline phosphate-coupled secondary antibodies were diluted in blocking buffer and incubated with the membranes for 30 minutes at room temperature followed by 3 washes with TBST. Membranes then were rinsed two times with water and incubated for 5 minutes with Tropix CDP star substrate (Applied Biosystems). Immuno-reactive bands were visualized using Hyblot CL autoradiography film (Denville). In order to determine the MAP2K1 mutation effects on cell signaling, the downstream target of MAP2K1 (ERK) was targeted. Activation of ERK was determined by its phosphorylation (PERK). Primary antibodies used were: anti-p44/42-ERK1/2 (Cell Signaling: #9102; 1/1000); anti-Phospho-p44/42 ERK1/2 (Cell Signaling: #9101; 1/1000); anti-MEK1 (MAP2K1) (Cell Signaling: 9124; 1/1000); anti-GAPDH (Cell signaling: #5174; 1/1000), and anti-beta-ACTIN (Sigma Aldrich: #A1978; 1/15,000). Secondary antibodies included: Goat anti-Rabbit IgG (H+L)-AP conjugated (Invitrogen: #31340; 1/10,000) and Goat anti-Mouse IgG (H+L)-AP conjugated (Invitrogen: #31320; 1/10,000). Western blots were repeated a minimum of 3 times.</p><!><p>Tube formation assays were performed using GelTrex (ThermoFisher Scientific, cat#: A1413202). GelTrex was thawed on ice and wells of a 24 well plate were coated with 300 μl GelTrex matrix and heated to 37°C for 30 minutes to allow the GelTrex to solidify. HUVECs then were plated in EGM2 complete medium on the surface of the GelTrex matrix at a density of 9000 cells/well (500 μl of an 18,000 cells/ml stock). Cells were placed in a 37°C CO2 incubator and the formation of a cell tube network was analyzed after 16 hours using an inverted microscope. GelTrex assays were done in triplicate.</p><!><p>Computation was performed used VassarStats [13]. Mean and standard deviation for pERK densitometry data was calculated. Groups were compared using the Mann-Whitney U-test. Statistical significance was defined as a p-value < 0.05.</p><!><p>Human AVM ECs were obtained from a patient who underwent resection of an AVM of the hand which had a confirmed MAP2K1-K57N mutation in ECs (MAF = 39%, passage 5). We found that MAP2K1-K57N mutant ECs do not have a proliferative advantage because continuous cultivation revealed a loss of mutant ECs, and by passage 10 mutant ECs were lost from the cultures. Thus, we only used cells up to passage 5 for our experiments.</p><p>Because it is not possible to obtain ECs from a non-affected part of a patient undergoing resection of an AVM, we used human white adipose tissue (HWAT) extracted ECFCs as a control. These cells display a stable endothelial phenotype and have robust in vivo blood vessel-forming capacity [12]. Western blot analysis showed that MAP2K1-K57N-AVM ECs had similar baseline MEK1 and ERK1/2 expression as compared to control HWAT-ECFCs. However, MAP2K1-K57N AVM ECs had a significant increase in the levels of phosphorylated and thus active ERK1/2 protein compared to ECFCs (mean increase 622% ±244%, p<0.05) (Figure 1).</p><p>To independently verify that the higher P-ERK levels in mutant AVM-ECs were the result of the MAP2K1-K57N mutation, we used lentiviral infection to overexpress either wild-type MAP2K1 or K57N mutant MAP2K1 in HUVECs. HUVECs infected with the empty lentiviral vector were used as control cells. Western analysis showed a moderate increase in MAP2K1 protein levels in cells infected with either wild-type or mutant MAP2K1 overexpressing vectors. While P-ERK levels in cells overexpressing wild type MAP2K1 was only slightly upregulated, a strong increase in P-ERK proteins levels was found in HUVECs overexpressing MAP2K1-K57N compared to both empty vector (mean increase 426%±164%, p<0.05) and overexpressing wild-type (mean increase 345%±135%, p<0.05) confirming the results obtained in patient derived AVM-ECs.</p><p>The FDA approved MEK1(MAP2K1) inhibitor Trametinib has been used to treat neoplasms that contain activating MAP2K1 mutations [14] and has been used off-label to treat AVMs [15]. To investigate whether Trametinib was able to counteract the increased phosphorylated ERK1/2 protein level in mutant cells, we exposed AVM derived ECs to increasing concentrations of Trametinib (Figure 2). A Trametinib concentration of 0.1 μM was able to reduce the P-ERK level in AVM ECs to the level of the control ECFCs. Finally, we tested whether Trametinib could influence vessel formation. Exposure of both AVM-derived ECs and MAP2K1-K57N overexpressing HUVECs to Trametinib reduced the ability of mutant ECs to form vascular networks (Figure 3).</p><!><p>Our data show that the most common AVM mutation, MAP2K1-K57N, over-activates the RAS/MAPK signaling pathway in ECs. This finding is consistent with previous reports that the MAP2K1 mutation is activating in other cell types that cause cancer [8,9], and when overexpressed in HEK293T embryonic kidney cells [11]. MAPK signaling is enhanced by receptor tyrosine kinases, integrins, and G-protein coupled receptors; MAP2K1 phosphorylates ERK1 and ERK2 [16].</p><p>In mammals, this cascade plays a crucial role in development, including fate determination, differentiation, proliferation, survival, migration, growth and apoptosis [17-19]. Interestingly, we have consistently found that prolonged cultivation of AVM derived ECs lowers the MAF in the cultures. This suggests that mutant ECs do not have a proliferative advantage despite their increased RAS/MAP2K1 signaling. We hypothesize that stimulation of RAS/MAPK signaling by mutant ECs might lead to abnormal coordination of artery-capillary-vein formation. The fundamental pathological finding in AVMs is the connection of arteries to veins through a nidus or fistula instead of a normal capillary bed [20]. MAP2K1 mutant ECs affecting normal vascular development might be the initiating stimulus that causes the pathological connection of arteries to veins. Absence of capillaries in AVMs reduces oxygen delivery to tissues leading to ischemia, ulceration, bleeding, and pain. Reactive neovascularization then contributes to enlargement of the AVM.</p><p>Our finding that Trametinib blocks mutant MAP2K1 upregulated signaling suggests that it might prove effective for AVMs, similar to its role in treating MAP2K1 dependent neoplasms (i.e., melanoma, lung adenocarcinoma) [21,22]. Trametinib also stopped vascular network formation. MAP2K1 inhibition of upregulated EC signaling might prevent the formation and progression of AVMs; regression of lesions also might occur. This hypothesis is supported by a recent case report showing reduction in the size of an AVM treated with Trametinib [15].</p>

PubMed Author Manuscript

Reassessment of the Transport Mechanism of the Human Zinc Transporter SLC39A2

The human zinc transporter SLC39A2, also known as ZIP2, was shown to mediate zinc transport that could be inhibited at pH <7.0 and stimulated by HCO3\xe2\x88\x92, suggesting a Zn2+/HCO3\xe2\x88\x92 cotransport mechanism [Gaither, L. A., and Eide, D. J. (2000) J. Biol. Chem. 275, 5560\xe2\x80\x935564]. In contrast, recent experiments in our laboratory indicated that the functional activity of ZIP2 increases at acidic pH [Franz, M. C., et al. (2014) J. Biomol. Screening 19, 909\xe2\x80\x93916]. The study presented here was therefore designed to reexamine the findings about the pH dependence and to extend the functional characterization of ZIP2. Our current results show that ZIP2-mediated transport is modulated by extracellular pH but independent of the H+ driving force. Also, in our experiments, ZIP2-mediated transport is not modulated by extracellular HCO3\xe2\x88\x92. Moreover, a high extracellular [K+], which induces depolarization, inhibited ZIP2-mediated transport, indicating that the transport mechanism is voltage-dependent. We also show that ZIP2 mediates the uptake of Cd2+ (Km ~ 1.57 \xce\xbcM) in a pH-dependent manner (KH+ ~ 66 nM). Cd2+ transport is inhibited by extracellular [Zn2+] (IC50 ~ 0.32 \xce\xbcM), [Cu2+] (IC50 ~ 1.81 \xce\xbcM), and to a lesser extent [Co2+], but not by [Mn2+] or [Ba2+]. Fe2+ is not transported by ZIP2. Accordingly, the substrate selectivity of ZIP2 decreases in the following order: Zn2+ > Cd2+ \xe2\x89\xa5 Cu2+ > Co2+. Altogether, we propose that ZIP2 is a facilitated divalent metal ion transporter that can be modulated by extracellular pH and membrane potential. Given that ZIP2 expression has been reported in acidic environments [Desouki, M. M., et al. (2007) Mol. Cancer 6, 37; Inoue, Y., et al. (2014) J. Biol. Chem. 289, 21451\xe2\x80\x9321462; Tao, Y. T., et al. (2013) Mol. Biol. Rep. 40, 4979\xe2\x80\x934984], we suggest that the herein described H+-mediated regulatory mechanism might be important for determining the velocity and direction of the transport process.

reassessment_of_the_transport_mechanism_of_the_human_zinc_transporter_slc39a2

<!>Chemicals and Reagents<!>Cell Culture Methods<!>Cd2+-Flux Measurements Using the FLIPR Tetra<!>Oocyte Isolation and Injection<!>Radioactive Uptake<!>Two-Electrode Voltage Clamp<!>pH Microelectrodes<!>Confocal Cd2+-Flux Imaging: A Single Cell with Clamped Membrane Potential (patch clamping)<!>Confocal Cd2+ Imaging: Multicell Recording<!>Statistical Analyses<!>Dependence of ZIP2 Functional Activity on Extracellular pH<!>Zn2+ Transport Mediated by ZIP2 Is Not Coupled to HCO3\xe2\x88\x92<!>ZIP2 Does Not Transport H+<!>ZIP2 Is Not Electrogenic<!>Transport Mediated by ZIP2 Is Not Dependent on Na+ or Cl\xe2\x88\x92 Gradients but Is Inhibited by K+<!>Transport Mediated by ZIP2 Is Not Coupled to K+ but Is Voltage-Dependent<!>Transport Kinetics and pH Dependence of ZIP2<!>Cation Selectivity of ZIP2<!>DISCUSSION<!>CONCLUSION

<p>Zinc is an essential trace element for human nutrition and the second most abundant transition element after iron in living organisms. The importance of zinc becomes evident when looking at a recent bioinformatics analysis indicating that as much as 10% of the human proteome is potentially capable of binding zinc.6 More than 3000 different types of proteins require zinc as a key structural or catalytic component. Among them are transcription factors, signaling proteins, transport/storage proteins, zinc finger proteins, and proteins involved in DNA repair, replication, and translation.7 Whole body and cellular zinc homeostasis is being thoroughly regulated. Whereas systemic zinc intoxication is relatively rare, zinc deficiency is a widespread problem leading to growth retardation, cognitive impairment, and immune dysfunction.8 The maintenance of mammalian zinc homeostasis is achieved by high-affinity zinc transport systems that are regulated by metal sensors.7 There are at least two different solute carrier (SLC) families of zinc transporters that control the movement of Zn2+ across membranes: (1) the SLC30 zinc transporter family, also known as the ZnT family, that facilitates cellular efflux or uptake into intracellular compartments and (2) the SLC39 family, also known as the ZIP family, that facilitates cellular uptake or efflux from intracellular compartments.9–12</p><p>The mammalian SLC39/ZIP family consists of 14 members, which can be divided into four subfamilies based on sequence similarity. SLC39A2 (ZIP2) together with SLC39A1 (ZIP1) and SLC39A3 (ZIP3) comprises subfamily II. ZIP2 was originally cloned and characterized by Gaither and Eide in 2000.1 In this study, they showed that more 65Zn2+ accumulated in ZIP2-expressing K562 cells than in parental cells, in a time-, temperature-, and concentration-dependent manner. They found that ZIP2-mediated zinc transport was not dependent on ATP hydrolysis or on Na+ or K+ gradients. In their assay, ZIP2 was inhibited at acidic pH (<7.0) and stimulated by 0.5 mM HCO3−. Thus, they proposed a Zn2+/HCO3− cotransport mechanism. In the same study, the expression level of ZIP2 mRNA was found to be generally low or negligible in human tissues and cultured cell lines, except in prostate and uterus,1 where it indeed could also be detected at the protein level.3 Cao et al. found ZIP2 mRNA in peripheral blood mononuclear cells (PBMCs) and monocytes.13 In their study, zinc depletion in both cell lines triggered upregulation of ZIP2 with concomitant downregulation of zinc-binding metallothioneins. Recently, Inoue et al. detected ZIP2 in the epidermis of healthy human frozen skin samples.4 They discovered that ZIP2 was upregulated by differentiation induction of cultured keratinocytes. Interestingly, ZIP2 knockdown inhibited the differentiation of keratinocytes and consequently the formation of a three-dimensional cultured epidermis. Several studies have linked the downregulation of ZIP2 in prostatic tissue to decreased zinc levels in prostatic epithelial cells and to prostate cancer.3,14–18 Studies with ZIP2-KO mice did not reveal any specific phenotype. However, these mice were more susceptible to abnormal embryonic development because of zinc deficiency during pregnancy.19</p><p>Recently, we published a screening assay that was established using the FLIPR Tetra high-throughput microplate reader to identify specific modulators of ZIP2 as potential therapeutic hit or lead compounds.2 This assay is based on the use of a Ca2+-sensitive dye, Calcium 5 (Molecular Devices), which, in addition to Ca2+, binds Cd2+ with high affinity. Binding of either Ca2+ or Cd2+ to the dye induces emission of a fluorescent signal that can be monitored, allowing quantification of the transport activity of proteins that mediate Cd2+ influx, such as ZIP2, which transports Cd2+ efficiently.1,2 In our laboratory, this assay has been successfully used to monitor the activity of a variety of Cd2+-transporting proteins such as the human divalent metal transporter DMT1 (SLC11A2)20 or the epithelial calcium channel TRPV6.21 Interestingly, during the development of the assay, we discovered discrepancies between our results on ZIP2 transport characteristics and the originally reported functional characterization of ZIP2.1 Therefore, the goal of the current work was to reexamine and extend the functional characterization of ZIP2.</p><!><p>Unless mentioned, all the chemicals and reagents were purchased from Sigma-Aldrich.</p><!><p>HEK293 cells were grown in complete Dulbecco's modified Eagle's medium (Gibco) supplemented with 10% fetal bovine serum (FBS), 10 mM HEPES, 100 μM minimal essential medium non-essential amino acids, and 1 mM sodium pyruvate (Gibco). Cells were cultured at 37 °C and 5% CO2 and subcultivated when confluency reached 90%.</p><p>Cells were transfected 24 h after being plated, following the manufacturer's protocol for the Lipofectamine 2000 (Invitrogen) reagent, using 50% of the recommended amount of human SLC39A2/ZIP2 (UniProt entry Q9NP94) encoding DNA and Lipofectamine 2000. The transfection medium was changed after 4 h. The transfection efficiency was estimated to be at least 70% using fluorescence microscopy.</p><!><p>Cells were plated in 96-well, clear-bottom, black-well plates coated with poly-D-lysine at a density of 2 × 104 cells/well. The next day, cells were transfected with ZIP2-pIRES2 DsRed-Express2 or mock-transfected using lipotransfection. On the experimental day, the cell culture medium was replaced with 100 μL of loading buffer [modified Krebs buffer containing 117 mM NaCl, 4.8 mM KCl, 1 mM MgCl2, 10 mM d-glucose, 5 mM HEPES, 5 mM MES, and Calcium 5 fluorescence dye (Molecular Devices) (pH 6.5)]. Cells were then incubated in the loading buffer at 37 °C for 1 h. Fluorescence measurements were taken using a FLIPR-Tetra high-throughput fluorescence microplate reader. Cells were excited using the 470–495 nm LED module, and the emitted fluorescence signal was filtered with a 515–575 nm emission filter. Cd2+ and the other tested solutes were prepared in assay buffer as 2× concentrated solutions in a separate 96-well plate. Establishment of a stable baseline was followed by addition of 100 μL of the indicated [Cd2+] and measurements of fluorescence for 15 min. Measurements were taken at 1 s intervals. In the negative control group, no substrate was administered. Results were exported from the FLIPR raw data as the "area under the curve" (AUC) of the fluorescence signal intensity in the interval after the addition of substrates (460–750 s). In sodium replacement experiments, 120 msssM NaCl was replaced with 120 mM choline chloride, NMDG, or KCl. In chloride replacement assays, all the chloride-containing salts were replaced by their corresponding equimolar gluconate salts. In pH dependence experiments, the different pH values were adjusted with 1 N HCl/NaOH.</p><!><p>Capped cRNA was synthesized using a linearized cDNA template and the T7 mMessage mMachine kit (Ambion). Xenopus laevis oocytes were isolated and dissociated using collagenase as described previously22 followed by injection with 50 nL of water or cRNA at 0.4 ng/nL (20 ng/oocyte), using a Nanoject-II injector (Drummond Scientific, Broomall, PA). Oocytes were maintained at 16 °C in OR3 medium22 and studied 3–6 days after injection.</p><!><p>The uptake of zinc by oocytes was measured by incubating groups of 8–10 X. laevis oocytes for 30 min in 1 mL of uptake buffer [ND96: 96 mM NaCl, 2 mM KCl, 1 mM MgCl2, 1.8 mM CaCl2, 5 mM MES, 1 mM HEPES, and 1 mM Tris (pH 6.0)] with 100 μM ZnCl2 including 0.05 μCi of 63Zn2+ (prepared as reported in ref 23). Oocytes were washed three times in uptake buffer with 1 mM ZnCl2 before measurement of incorporation of 63Zn2+ into single oocytes as γ emission with a WIZARD2 gamma counter (PerkinElmer). For 63Zn2+ uptake at different pH values, the amounts of MES, HEPES, and Tris were varied. To determine the effect of HCO3− on ZIP2 function, 96 mM NaCl was replaced with 96 mM NaHCO3 in the standard ND96. In sodium replacement assays, 96 mM NaCl was replaced with 96 mM choline chloride or 96 mM KCl.</p><p>Radioactive iron uptake experiments in HEK293 cells were performed as described previously.20 Briefly, cells were plated in 96-well, clear-bottom, white-well plates coated with poly-Dlysine at a density of 2 × 104 cells/well. The next day, cells were transfected with ZIP2-pIRES2 DsRed-Express2 or hDMT1-pIRES2 DsRed-Express2 using lipotransfection. On the experimental day, the growth medium was aspirated and cells were washed three times with Krebs−Ringer buffer [140 mM NaCl, 2.5 mM KCl, 1.2 mM MgCl2, 1.2 mM CaCl2, 10 mM d-glucose, 5 mM HEPES, and 5 mM MES (pH 7.4)]. To measure iron uptake, Krebs−Ringer buffer (pH 5.5) was supplemented with 1 mM ascorbic acid, 1 μM FeCl2, and 5 μCi/mL radioactive 55Fe (American Radiolabeled Chemicals). The assay was terminated after 15 min when the plates were washed four times with ice-cold Krebs−Ringer buffer. Subsequently, 100 μL of MicroScint-20 (PerkinElmer) was dispensed into each well and incubated at RT for 1 h under constant agitation. Radioactive 55Fe uptake was measured using a TopCount Microplate Scintillation and Luminescence Counter (PerkinElmer).</p><!><p>Two-microelectrode voltage clamping (TEVC) was used to measure steady-state currents in control oocytes and oocytes injected with ZIP2 mRNA, 4–7 days after injection. Oocyte membrane currents were recorded using an OC-725C voltage clamp (Warner Instruments, Hamden, CT), filtered at 2–5 kHz, digitized at 10 kHz, and recorded with Pulse software, and data were analyzed using the PulseFit program (HEKA), as previously described.22 Voltage microelectrodes (resistance of 0.5–5 MΩ) were made from fiber-capillary borosilicate and filled with 3 M KCl. Oocytes were perfused at room temperature in ND96 buffer. For periods when current−voltage (I-V) protocols were not being run, oocytes were clamped at a holding potential (Vh) of −60 mV. I-V protocols consisted of 100 ms step changes in membrane potential from −120 to 40 mV in 20 mV increments before and after the addition of the test substrate. The resulting data were filtered at 5 kHz (eight-pole Bessel filter, Frequency Devices) and sampled at 1 kHz. The I-V relationship was determined by plotting the mean steady-state current against the voltage for a given set of experiments. Additionally, the current was monitored continuously in oocytes clamped at a Vh of −60 mV. Test solutions were perfused at room temperature for several minutes until a steady-state current was observed.</p><!><p>Ion-selective microelectrodes were used to monitor the intracellular pH (pHi) of ZIP2- and water-injected oocytes as previously described.22 Ion-selective electrodes were pulled like those used for TEVC and silanized with bis(dimethylamino)-dimethylsilane (Fluka Chemical Corp., Ronkonkoma, NY). Electrode tips were filled with hydrogen ionophore 1-cocktail B (Fluka) and backfilled with phosphate buffer at pH 7.0. The intracellular pH was measured as the difference between the pH electrode and a KCl voltage electrode impaled into the oocyte, and the membrane potential (Vm) was the potential difference between the KCl and an extracellular calomel microelectrode. Electrodes were calibrated using pH 6.0 and 8.0 (Fisher), followed by point calibration in ND96 (pH 7.50). All pH microelectrodes used had slopes of at least −54 mV/pH unit.</p><!><p>The membrane potential was clamped to −60 mV using the patchclamp technique in the whole-cell configuration. The pipet solution was composed of 120 mM Cs-Asp, 20 mM TEA-Cl, 5 mM K2ATP, 8 mM NaCl, 5.6 mM MgCl2 (0.75 mM free Mg2+), 20 mM HEPES, and 50 μM Fluo 3 pentapotassium salt. Note that, for this experimental set, to monitor Cd2+ fluxes, the Calcium 5 dye was replaced with Fluo 3, another calcium indicator that is sensitive to Cd2+,24 because a membrane impermeable dye was required to avoid unspecific signal loss after cell loading. The pH was adjusted to 7.2 with CsOH, and the osmolality was 285 mosmol/kg. The external solution was composed of 117 mM NaCl, 4.8 mM KCl, 1 mM MgCl2, 5 mM d-glucose, 5 mM HEPES, and 5 mM MES. The pH was adjusted to 6.5 with 1 N HCl. Under high-potassium conditions, NaCl was replaced with equimolar KCl. All experiments were performed at room temperature.</p><!><p>Calcium 5 fluorescent dye (Molecular Devices) was used to record Cd2+ uptake. Prior to the measurements, cells were incubated for 1 h at 37 °C in the following external solution: 117 mM NaCl, 4.8 mM KCl, 1 mM MgCl2, 10 mM d-glucose, 5 mM HEPES, 5 mM MES, and Calcium 5 dye. The pH was adjusted to 6.5 with 1 N HCl. Under high-potassium conditions, NaCl was replaced with equimolar KCl. All experiments were performed at room temperature.</p><p>Cd2+ images were acquired with a FluoView 1000 (Olympus) confocal laser-scanning microscope. Fluo 3 and Calcium 5 dye were excited at 473 nm with a solid-state laser, and fluorescence was detected between 515 and 585 nm. To control cell transfection, DsRed-Express2 was excited at 561 or 488 nm with a solid-state laser and fluorescence was detected at >585 nm. Images were processed and analyzed using the software ImageJ. The measurements are expressed as ΔF/F0, where F0 is the fluorescence recorded before the application of Cd2+.</p><!><p>The normal distribution of the experimental groups was determined by Kolmogorov−Smirnov (N > 50) and Shapiro−Wilk (N < 50) tests. Normally distributed independent experimental groups were compared with an unpaired Student's t test. When the data sets were not normally distributed, a Mann−Whitney U test was used to assess differences. Statistical tests were performed using the IBM statistics 20 software. P values of <0.05 are considered statistically significant.</p><!><p>On the basis of the studies of Gaither et al., the functional activity of ZIP2 was found to be inhibited at pH <7.0.1 On the other hand, our fluorescence-based transport assay using transiently transfected HEK293 cells revealed that, at acidic pH (6.5), the level of ZIP2-mediated Cd2+ transport was greatly increased compared to that at pH 7.5.2 What could be the reason for this discrepancy? In our assay, pH changes may have influenced the binding affinity of the fluorescent dye used (Calcium 5) to measure Cd2+. Also, Cd2+ rather than Zn2+ transport was measured, which might account for the reversed pH dependence. We therefore decided to perform additional experiments to evaluate this incongruity. To this end, we used the standard radioactive tracer method using X. laevis oocytes as an expression system, thus ensuring the same functional readout that was used by Gaither et al. Using this methodology, the effect of variable extracellular pH values (pH 5.0–8.2) on ZIP2-mediated 63Zn2+ uptake was determined (Figure 1A). Whereas the water-injected oocytes showed only a slight pH dependence of endogenous Zn2+ transport activity, ZIP2 transport activity was maximal at an acidic pH of <6.0 and almost negligible at pH >7.5. These results confirmed our previous findings2 and further validate our fluorescence assay that was used as a screening assay to identify ZIP2 modulators.</p><!><p>It was previously proposed that ZIP2 operates as a Zn2+/HCO3− cotransporter.1 We aimed to confirm this hypothesis by measuring the uptake of 63Zn2+ by ZIP2 cRNA-microinjected X. laevis oocytes in the presence and absence of HCO3−. To this end, we replaced the [NaCl] of the uptake solution by an equimolar [NaHCO3], which resulted in a pH of 8.2. Because adjusting the pH of the solution would alter [HCO3−], we compared the 63Zn2+ uptake in the HCO3−-containing solution with that in the normal uptake solution, both at pH 8.2 (Figure 1B). We did not observe a difference in the transport activity of ZIP2, whereas the H2O-injected oocytes showed a higher activity at increased HCO3− concentrations. Additionally, to confirm these findings, we used pH-sensitive microelectrodes to measure intracellular pH (pHi) changes due to HCO3−-coupled Zn2+ transport via ZIP2expressed X. laevis oocytes (Figure 2A). The uptake buffer was equilibrated by addition of 5% CO2 and 33 mM HCO3−. The CO2 caused an acidification in ZIP2-injected oocytes (Figure 2A,C) as well as in H2O-injected oocytes (data not shown). Upon perfusion of Zn2+, the pHi did not change, contrary to what would be expected if the transport was coupled to HCO3− (Figure 2A,C).</p><!><p>Our experiments suggest that H+ may be involved in the ZIP2-mediated transport process. Thus, we also used the pH-sensitive microelectrodes in X. laevis oocytes to investigate whether protons are coupled to ZIP2-mediated Zn2+ transport (Figure 2B). Almost no change was observed when the pH 7.5 uptake solution was replaced by the pH 6 solution, indicating that there is no H+ permeation via ZIP2. Also, addition of Zn2+ did not cause a significant change in pHi in ZIP2-injected oocytes (Figure 2B,C). These results indicate that ZIP2 does not facilitate transport of H+, neither alone nor coupled to Zn2+ transport.</p><!><p>To test whether ZIP2-mediated Zn2+ transport is electrogenic, functional experiments were performed using two-electrode voltage clamping (TEVC). We did not observe any discernible change in the current−voltage (I-V) relationship following step changes in membrane potential (Vm) of oocytes expressing ZIP2 at extracellular pH 7.5 or 6.0 (Figure 3A,B). Moreover, perfusion of Zn2+ (100 μM) did not evoke any appreciable change in the I-V relationship (Figure 3A,B). Similar results were obtained when the current was monitored continuously in oocytes clamped (Vh) at −60 mV (data not shown). Hence, functional experiments performed with TEVC indicate that there are no measurable currents associated with ZIP2-mediated Zn2+ transport, which is surprising, given the positive charge of the divalent metal ion Zn2+ and the highly significant transport activity observed for ZIP2 cRNA-injected oocytes when measuring 63Zn2+ accumulation under similar experimental conditions (Figure 1).</p><!><p>We investigated whether ZIP2-mediated transport is coupled to Na+, K+, or Cl− by isosmotic replacement of these ions in the standard uptake solution. To address this question, we used both a fluorescentbased Cd2+ influx assay (Figure 4A) in transiently transfected HEK293 cells and a 63Zn2+ uptake assay in ZIP2 cRNA-injected X. laevis oocytes (Figure 4B). Na+ was replaced by equimolar NMDG, choline, or K+, whereas Cl− was replaced by the corresponding gluconate salts. Replacement of Na+ with NMDG or choline had no effect on ZIP2 activity, whereas replacing it by K+ reduced the rate of transport by ≈60–40%. Replacement of chloride by gluconate salts did not have any effect on ZIP2 activity.</p><!><p>To clarify whether the decrease in the rate of ZIP2-mediated Cd2+ or Zn2+ transport when Na+ is replaced with K+ is due to direct K+-coupled metal ion transport or merely a result of membrane depolarization generated by an increased extracellular [K+], fluorometric analysis under voltage-clamp conditions was conducted. First, fluorometric measurements were taken without voltage clamping, within an open field of ZIP2-transfected cells (Figure 5A–C). In line with our previous observations, replacement of Na+ with equimolar K+ induced ≈40% inhibition of the ZIP2-mediated influx of Cd2+. Next, the same procedure was performed in individual cells under voltage-clamp conditions (Vh = −60 mV) (Figure 5D–F). Interestingly, under voltage-clamp conditions, the inhibition was lost and the ZIP2-mediated influx of Cd2+ was similar in the presence and absence of a high extracellular [K+]. These results demonstrate that K+ is not part of the translocation mechanism of ZIP2 and that transport is voltage-dependent.</p><!><p>The kinetics of ZIP2-mediated transport was studied using our Cd2+-flux fluorescence-based assay in transiently transfected HEK293 cells. Fluorescence intensity changes increased with extracellular [Cd2+] (Figure 6A, top panel). Measuring Cd2+ flux through ZIP2 gave a dose−response curve that reached saturation at 5 μM Cd2+ (Figure 6C). The calculated apparent affinity constant (Km) for Cd2+ was ~1.57 ± 018 μM. Empty vector-transfected cells did not show any change in fluorescence intensity within the tested range of Cd2+ concentrations (Figure 6A, bottom panel).</p><p>Using the same methodology, the pH dependence of the ZIP2-mediated transport was studied. In line with our previous findings, fluorescence intensity changes after Cd2+ (10 μM) perfusion increased with extracellular [H+] (Figure 6B, top panel). Transport was completely saturated at extracellular pH 6.5, and the calculated apparent affinity (KH+) was ~66 ± 16 nM, corresponding to pH ~7.2 (Figure 6C). Again, no effect was observed over the empty vector-transfected cells (Figure 6B, bottom panel).</p><!><p>To investigate the cationic selectivity of ZIP2, Cd2+ flux was measured in the presence of high extracellular concentrations (50 μM) of different divalent cations such as Ba2+, Mn2+, Co2+, Zn2+, and Cu2+ (Figure 7A). Note that none of these metal ions showed significant interactions with the Calcium 5 dye when high concentrations of them (100 μM) were perfused individually into ZIP2-overexpressing cells (data not shown). As expected, in the presence of Zn2+, Cd2+ flux was completely inhibited. Interestingly, Cu2+ and Co2+ inhibited 75 and 25%, respectively, of the Cd2+ flux, while no significant inhibition by Ba2+ or Mn2+ was observed. Fe2+ is another putative substrate of ZIP2. However, because it also interacts with the Calcium 5 dye, it was not included in this set of experiments. To overcome this issue, we measured directly radiolabeled iron (55Fe2+) uptake (Figure 7B). As a positive control for this assay, we used human divalent metal transporter 1 (hDMT1, SLC11A2).25 Uptake of 55Fe2+ by ZIP2 was not significantly different from that of empty vector-transfected cells and 7-fold lower than the uptake mediated by hDMT1, demonstrating that Fe2+ is not a substrate of ZIP2.</p><p>To determine the apparent affinity of ZIP2 for Zn2+ and Cu2+, Cd2+ flux was measured in the presence of a range of extracellular concentrations of Zn2+ or Cu2+. In both cases, inhibition of Cd2+ flux gave dose−response sigmoidal curves. The calculated IC50 values were ~0.52 ± 1.7 μM for Zn2+ (Figure 7C) and ~2.98 ± 1.3 μM for Cu2+ (Figure 7D). Given that the Km of ZIP2 for Cd2+ is 1.57 μM (Figure 6C), according to the Cheng−Prusoff equation,26 the IC50 values for Zn2+ and Cu2+ are 0.32 and 1.81 μM, respectively. Hence, the cationic selectivity for ZIP2 decreases in the following order: Zn2+ > Cd2+ ≥ Cu2+ > Co2+. Fe2+, Mn2+, and Ba2+ are not transport substrates.</p><!><p>In our functional experiments using different approaches, the level of ZIP2-mediated transport was increased at acidic pH, even though no cotransport with H+ was observed. We therefore conclude that the ZIP2 transport process is modulated by extracellular pH, independent of the H+ driving force. Transport was not stimulated by the presence of HCO3−, as previously reported.1 Thus, we conclude that ZIP2-mediated transport is not coupled to bicarbonate. Also, in contrast to the previous observations,1 our experiments revealed that an increasing extracellular [K+] inhibits ZIP2-mediated metal ion uptake under non-voltage-clamp conditions. However, when the K+ inhibitory effect was measured under voltage-clamp conditions, it was abolished. This indicates that the inhibitory effect was due do the depolarization caused by increasing the extracellular K+ concentration to 140 mM. Therefore, we concluded that ZIP2-mediated metal ion transport is voltagedependent, and given the positive charge of Zn2+, we expected that transport would be electrogenic.</p><p>Paradoxically, our electrophysiological analysis revealed that ZIP2-mediated Zn2+ transport is electroneutral. The electrophysiological experiments were performed with ZIP2 overexpressed in X. laevis oocytes and in the presence of a robust inwardly directed electrochemical Zn2+ gradient, favoring transmembrane influx (i.e., the membrane voltage was kept constant at −60 mV and 100 μM ZnCl2 was perfused). Using the same experimental approach, our group observed prominent transmembrane inward currents for Fe2+ transport via DMT1 expressed in X. laevis oocytes, even at neutral pH (that is in the absence of an inwardly directed H+ gradient).27 Given that ZIP2-injected oocytes exhibited a high level of 63Zn accumulation as shown in Figure 1, this rules out any issue related to plasma membrane expression. To explain the lack of electrogenicity, we propose the following: (1) Transport is still electrogenic, but the turnover rate of the transport process is too slow to allow any detection of transport-associated currents. (2) Transport is electroneutral because there is an as yet unidentified coupling ion (via cotransport or exchange) that balances the positive charges of Zn2+. In this case, given the voltage dependence of the transport process, the transport cycle must contain steps that are limited by the membrane potential.</p><p>Interestingly, the transport features of ZIP2 resemble those of a ZIP transporter from the Gram-negative, rod-shaped bacterium Bordetella bronchiseptica (ZIPB).28 In that work, ZIPB was described as a selective electrodiffusional channel, in which Zn2+ uptake is driven only down its concentration gradient. Remarkably, Zn2+ transport by ZIPB was modulated by the effect of K+ on the resting membrane potential, indicating that ZIPB is also voltage-dependent. Furthermore, ZIPB-mediated Zn2+ flux was modulated by pH and not stimulated by HCO3−. Also in line with our findings, Fugu pufferfish ZIP2, sharing 30 and 60% sequence identity with human ZIP2 and ZIP3, respectively, exhibited, when expressed in MDCK cells, Zn2+-mediated transport in a pH-dependent manner. Transport was stimulated by acidic pH medium (pH 5.5–6.5) but was not enhanced (but rather slightly inhibited) by the presence of extracellular HCO3−.29</p><p>Altogether, given that ZIP2-mediated transport is ATP-independent1 and not coupled to Na+, H+, K+, HCO3−, or Cl−, we propose that Zn2+ uptake occurs via simple passive transport. Given that Zn2+ is a trace element essential for most mammalian cells, efficient uptake mechanisms must exist to allow Zn2+ to accumulate within cells. Because intracellular Zn2+ is complexed with specific binding proteins, cytoplasmic Zn2+ concentrations are kept at very low (femtomolar to picomolar) levels. Consequently, the inwardly directed electrochemical Zn2+ gradient is expected to be sufficient to facilitate cellular Zn2+ uptake, supporting this concept of passive ZIP2-mediated transport.28</p><p>The only ion showing interaction with the transport process mediated by ZIP2 is H+. Our results show that, at low extracellular pH values, the rate of transport of Zn2+ is increased. However, H+ was not cotransported with Zn2+, and there was transport at pH >7.5, indicating that ZIP2-mediated transport is modulated by pH, rather than H+ acting as a coupling ion. In addition to the aforementioned ZIPB and Fugu pufferfish ZIP2, there are many examples of ion channels that can be modulated by external pH, including Cl− channels,30 Na+ channels,31 and aquaporins,32 among others.33 In these channels, the protonation state of specific titratable residues affects voltage dependence or gate opening, leading to modulation of channel permeation. Future structure−function studies of ZIP2 to identify amino acid residues responsible for H+ sensitivity will shed further light on this pH modulatory mechanism.</p><p>As described previously, ZIP2 expression has been found in prostate epithelial cells,34 peripheral blood mononuclear cells of patients with tuberculosis and asthma,5 and epidermal keratinocytes.4 Interestingly, these tissues and/or cell types are involved in physiological processes occurring in an acidic environment. The main function of the prostate is to secrete prostatic fluid, which is acidic (pH 6.5–6.7).34,35 On the basis of its apical membrane localization, ZIP2 is hypothesized to help maintain prostate Zn2+ homeostasis by reabsorbing Zn2+ from the prostatic fluid.3 Similarly, acidification of the airways linked to different pathological processes, including inflammation, ischemia or aspiration of refluxing gastric contents, and obstructive airway diseases such as asthma, may lead to an increased rate of ZIP2-mediated transport.33 ZIP2 expression has been described in the epidermis, and moreover, transportermediated Zn2+ uptake is necessary for the differentiation of keratinocytes.4 The surface of healthy skin has a pH oscillating between 4.0 and 6.0.36 Altogether, these findings further support the role of H+ in the transport processes mediated by ZIP2 because, as our functional experiments point out, the functional activity of ZIP2 will be increased in these acidic environments. In turn, it seems counterintuitive to use HCO3− as a driving force for the transport of Zn2+ under such physiological conditions as, at reduced pH, a significant part of bicarbonate will be in the conjugated acid form carbonic acid (H2CO3) (pKa = 6.1 at 37 °C). In this regard, upregulation of ZIP2 expression in peripheral blood mononuclear cells of patients with tuberculosis was accompanied by downregulation of the expression of SLC39A8 (ZIP8), and the authors proposed that this could be a consequence of changes in the pH and Zn2+ concentrations.5 These findings suggest a complementary function of ZIP2 and ZIP8. In line with this, our preliminary experiments using the fluorescence-based assay described herein revealed opposite pH modulation for ZIP8 and ZIP2 (data not shown).</p><p>With respect to the kinetic properties of ZIP2, our experiments indicate that the Km for ZIP2-mediated Cd2+ flux was ~1.6 μM, similar to that reported by Gaither and Eide for Zn2+ (Km ~ 3 μM).1 On the other hand, on the basis of our assay, ZIP2-mediated transport reached Vmax already at 5 μM Cd2+, whereas in the study by Gaither et al., the Vmax for Zn2+ transport was 20–40 μM.1 Another difference between the two studies exists upon comparison of the divalent metal competition experiments. According to our experiments, ZIP2 can transport Zn2+ and Cd2+ but not Fe2+ and Cu2+ and Co2+ are likely also substrates, while Ba2+ and Mn2+ were not transported by ZIP2. In contrast, Gaither et al. proposed that all of these metals ions could serve as substrates of ZIP2.1 In line with our findings, studies with HEK293 cells overexpressing mouse ZIP2 showed similar Michaelis−Menten kinetics for Zn2+ (Km ~ 1.6 μM).37 Moreover, Zn2+ transport was inhibited in this study by excesses of Cu2+, Cd2+, and Co2+ but not Fe2+ or Mn2+. In contrast, the previously mentioned pufferfish ZIP2 exhibited a 10-fold lower affinity for Zn2+ (Km ~ 13 μM), while Zn2+ transport was inhibited by Cu2+, Cd2+, Co2+, Fe3+, and to a lesser extent Fe2+.29 This variability among competition experiments highlights the importance of determining the substrate selectivity of transporters by direct measurements of each putative substrate. In this regard, direct measurements of 63Zn (Figures 1 and 4B), Cd2+(Figure 6), and Fe uptake (Figure 7B) confirm that Zn2+ and Cd2+ are real substrates of ZIP2, while iron was not found to be a substrate. With regard to the other proposed ZIP2 substrates (i.e., Cu2+ and Co2+), direct measurements will also be required to verify they are transport substrates.</p><p>The incongruities between this work and that of Gaither et al. are likely due to the use of different expression systems. Gaither et al. used the chronic myeloid leukemia cell line K562 for their radiolabeled Zn2+ uptake experiments. These cells endogenously express Zn2+ transporters, as well as the sodium− proton exchanger NHE1 (SLC9A1)38 and the chloride/ bicarbonate anion exchanger AE2 (SLC4A2).39 Thus, the reported Zn2+ transport activities and divalent metal ion specificities in K652 cells represent the sum of both endogenous and expressed ZIP2 transporters. In addition, the inverse pH sensitivity and role of bicarbonate in ZIP2-mediated transport could be related to interfering activities of NHE1 and AE2. Our functional experiments were conducted in HEK293 cells, which express endogenous NHE3 (SLC9A1)40 but not AE2.41 Also, experiments were performed in Xenopus oocytes, which express an NHE exchanger homologue but not any endogenous anionic exchangers.42 As functional readouts, we used a combination of different methods, including electrophysiological measurements, radiolabeled Zn2+ uptake experiments in X. laevis oocytes, and a Cd2+-flux-based fluorescent assay in HEK293 cells. Importantly, in non-injected control oocytes or in empty vector-transfected HEK cells, endogenous Zn2+ transport (Figure 1) or Cd2+ transport (Figure 6A) was negligible compared to that of ZIP2-expressing oocytes or cells. Hence, our experimental approaches guarantee an optimal signal-to-noise ratio for studying different aspects of the ZIP2 transport mechanism. This allowed us to validate our observations using different techniques, thereby generating data with great consistency, as demonstrated, for example, for the pH dependence (Figures 1A and 6B) or the effect of K+ (Figures 4 and 5).</p><p>We anticipate that the data reported herein are valuable for predicting the putative roles of ZIP2 in pathological situations or during physiological challenges. Indeed, human genetic studies revealed that ZIP2 polymorphisms constitute a risk factor for a wide variety of human diseases, including carotid artery disease in aging,43 arsenic-related bladder cancer,44 and cystic fibrosis.45 In addition, ZIP2 activity is important for prostate function and related to prostate cancer development,3,34 keratinocyte differentiation,4 and macrophage46 and monocyte function.13 Also, ZIP2 knockout mice studies revealed increased susceptibility to Zn deficiency during pregnancy.19 As a follow-up, specific experiments are needed to identify the particular roles of ZIP2 in physiologically relevant environments. For example, functional studies with the aforementioned genetic variants will be required to reveal the precise molecular mechanisms leading to the associated disease conditions. Also, tissue specific knockout studies in cellular or animal models are required to describe the physiological and pathological impacts of ZIP2 dysfunction. Such studies may in turn accelerate the discovery of therapeutic applications targeting ZIP2.</p><!><p>Our data show that ZIP2-mediated transport is modulated by extracellular pH, in an H+ driving force-independent and voltage-dependent manner. Accordingly, we propose that ZIP2 is a facilitative transporter that mediates transport of Zn2+ down its concentration gradient, which can be modulated by interaction of H+ with titratable acidic amino acid residues within the ZIP2 protein. Specifically, we propose that protonation of such a titratable amino acid stabilizes the ZIP2 protein in a conformation in which substrate transport is more favorable. This would explain why the transport rate is increased in the presence of H+. ZIP2 is expressed in acidic environments, where this regulatory mechanism is expected to be important to accelerate and determine the direction of the transport process.</p><p>The herein proposed transport mechanism is consistent with those of ZIPs from lower organisms (i.e., ZIPB and pufferfish ZIP2).28,29 Nevertheless, this does not necessarily hold true for all the ZIP members, as some of them have been postulated to possess different transport mechanisms. For example, ZIP8 and ZIP14 are described as metal/bicarbonate symporters.47,48 In this regard, as mentioned previously, a preliminary experiment from our laboratory, monitoring Cd2+ fluxes through human ZIP8-overexpressing cells, showed that the activity of this transporter is not stimulated by extracellular H+, indicating a transport mechanism that is different from that for ZIP2 described herein. This highlights the need for future studies for each ZIP family member individually, to reveal their particular transport mechanisms and to understand their distinctive contributions to body Zn2+ homeostasis.</p>

PubMed Author Manuscript

Synthesis and Reactivity of Alkyl-1,1,1-trisphosphonate Esters

The \xce\xb1\xe2\x80\x93trisphosphonic acid esters provide a unique spatial arrangement of three phosphonate groups, and may represent an attractive motif for inhibitors of enzymes that utilize di- or triphosphate substrates. To advance studies of this unique functionality, a general route to alkyl derivatives of the parent system (R = H) has been developed. A set of new \xce\xb1-alkyl-1,1,1-trisphosphonate esters has been prepared through phosphinylation and subsequent oxidation of tetraethyl alkylbisphosphonates, and the reactivity of these new compounds has been studied in representative reactions that afford additional examples of this functionality.

synthesis_and_reactivity_of_alkyl-1,1,1-trisphosphonate_esters

Introduction<!>Results and Discussion<!>General Procedures<!>Methylidynetrisphosphonic acid, hexaethyl ester (7)<!>2-Phenylethylidynetrisphosphonic acid, hexaethyl ester (10)<!>3-Butenylidynetrisphosphonic acid, hexaethyl ester (11)<!>4-Methyl-3-pentenylidynetrisphosphonic acid, hexaethyl ester (12)<!>(3E)-4,9-Dimethyl-3-nonadienylidynetrisphosphonic acid, hexaethyl ester (13)<!>Tetraethyl 6-hepten-1,1-bisphosphonate<!>6-Heptenylidynetrisphosphonic acid, hexaethyl ester (14)<!>Butylidynetrisphosphonic acid, hexaethyl ester (15)<!>3-Butyn-1-ylidynetrisphosphonic acid, hexaethyl ester (16)<!>Trisphosphonate 15 via catalytic hydrogenation of compound 11<!>4-Hydroxybutylidynetrisphosphonic acid, hexaethyl ester (19)<!>3-Pentenylidynetrisphosphonic acid, hexaethyl ester (20) and compound 12<!>Compound 20 via metathesis with 2-butene<!>7-Methyl-6-octenylidynetrisphosphonic acid, hexaethyl ester (21)<!>7-Hydroxyheptylidynetrisphosphonic acid, hexaethyl ester (22)<!>1-Benzyl-4-[2,2,2-tris(diethyoxyphosphinyl)ethyl-1H-1,2,3-triazole (23)<!>3-Butenylidynetrisphosphonic acid, pentasodium, 2,4,6-trimethylpyridinium salt (24)

<p>The geminal bisphosphonate moiety is found in a number of drugs that are in widespread clinical use, including alendronate (1, Fosamax®), risedronate (2, Actonel®), and zoledronate (3, Zometa®).1 The clinical applications of these compounds in treatment of various diseases of the bone,1 together with the prevalence of di- and triphosphate intermediates in metabolism, have encouraged studies of many other bisphosphonates, and there are numerous reports on their chemical synthesis2 and biological activity.3 In sharp contrast to the extensive work with geminal bisphosphonates, there are very few reported studies of aryl- or alkyl-1,1,1-trisphosphonate esters (4). Phosphonate esters often are prepared through reaction of a trialkyl phosphite with an alkyl halide, but simple alkyl halides are not very reactive in this classical Michaelis-Arbuzov synthesis and chloroform does not react with triethyl phosphite even under forcing conditions.4 However, trichloromethylamine is known to react with triethyl phosphite to afford the aminotrisphosphonate 55 through a reaction sequence now assumed to be based on elimination-addition reactions.6 A similar strategy with a quinone methide was used to prepare the aryl trisphosphonate 6,7 and ultimately the parent compound 7 was prepared through addition to tetraethyl diazo-bisphosphonate.6 The parent trisphosphonate 7 also has been prepared via C-P bond formation. In this case, the anionic bisphosphonate proved unreactive with diethyl chlorophosphate but phosphinylation of the bisphosphonate anion followed by oxidation to the phosphonate was successful,6 a strategy which already had been applied to preparation of β-keto phosphonates from ketone and ester enolates.8 However, apart from some intriguing studies by Blackburn et al., who prepared adenosine esters derived from compound 7 and its halogenated analogues,9 little has been done with trisphosphonates for some time. The limited information available on trisphosphonates esters and our longstanding interest in C-P bond formation10 led us to investigate the synthesis and reactivity of this functionality.</p><!><p>One might reasonably assume that the shortest route from tetraethyl methylenebisphosphonate (8) to a family of alkyl trisphosphonates would involve preparation of the parent trisphosphonate 7 followed by alkylation. To explore this possibility, compound 7 was prepared starting with the literature approach that employed phosphinylation of tetraethyl methylenebisphosphonate, but followed by oxidation of the presumed phosphinate intermediate with hydrogen peroxide8b,10b rather than air (Scheme 1).6 This modified procedure gave an improved yield (48% vs. the 32% in the original report6) when the methylenebisphosphonate is thoroughly dried (vide infra), and the 1H, 13C and 31P spectra of the material prepared this way matched literature data.6 Attempted reaction of the methylenebisphosphonate anion with diethyl chlorophosphate was not successful under the same reaction conditions.</p><p>Upon treatment of compound 7 with NaH, formation of the anion was strongly suggested by a downfield shift in the 31P NMR spectrum (from 14 to 32 ppm). However, addition of benzyl bromide did not induce any further change in the 31P NMR resonance, nor did addition of the less sterically encumbered alkylating agent allyl bromide. Presumably the limited reactivity of the anion under these conditions is a consequence of the fact that the carbanionic center is both well stabilized and considerably hindered. Several experiments were conducted to determine the approximate acidity of compound 7. For example, the 31P NMR resonance observed for the anion at 32 ppm persisted even after addition of water or saturated NH4Cl to the NMR sample. Only after addition of acetic acid was a resonance representing the neutral compound 7 again observed at 14 ppm. Titration of the trisphosphonate ester 7 with NaOH gave a pKa of ~6.5, suggesting that the negative charge is highly delocalized and that this ester should be viewed as a strong carbon acid.</p><p>To explore alternate approaches to these compounds, tetraethyl benzylbisphosphonate 9 was prepared by alkylation of bisphosphonate 8.2c, 2d No evidence for C-P bond formation could be detected by 31P NMR upon treatment of compound 9 with NaH and diethyl chlorophosphate, but reactions that employed diethyl chlorophosphite as the electrophile6,8 were more encouraging. After treatment of bisphosphonate 9 with NaH and diethyl chlorophosphite, exposure to air under standard conditions afforded just trace amounts of the desired trisphosphonate. However, when reaction of the bisphosphonate 9 with NaHMDS and diethyl chlorophosphite at 0 °C was followed by treatment with H2O2, an exothermic reaction ensued. A new product was detected by TLC analysis, and analysis of the reaction mixture by 31P NMR revealed a new resonance at 18 ppm. After isolation of this product via column chromatography, the 1H NMR spectrum displayed a notable phosphorus coupling to the benzylic hydrogens (q, J = 15.8 Hz). The 13C NMR spectrum was even more striking, with observable couplings to 31P throughout the spectrum and a resonance for the quaternary carbon that appeared as a clear quartet (JCP = 118 Hz). Based on these data, as well as a consistent elemental analysis, the product was assigned the structure of trisphosphonate 10.</p><p>The three-step protocol of alkylation, phosphinylation, and oxidation proved to be successful with several other alkyl halides but the isolated yields initially were modest (<20%). Addition of excess base did not result in increased yields. Elemental analyses of alkylbisphosphonates have consistently revealed the presence of water, suggesting that hydroxide generated in situ might complicate formation of the desired intermediate in this case. Thorough drying of the alkylated bisphosphonates via azeotopic distillation with either benzene or toluene prior to deprotonation and phosphinylation resulted in a dramatic increase in reaction yields. Ultimately this strategy resulted in conversions ranging from 64 to 86% by 31P NMR, with isolated yields typically just slightly lower. This methodology was then applied to the synthesis of a variety of alkyl trisphosphonates with good isolated yields (Table 1). Furthermore, there is at least the potential to recover the alkylbisphosphonate in the cases of lower conversion, and that decision can be based on inspection of the 31P NMR spectrum of the reaction mixture (~18 ppm for trisphosphonate 10 versus ~23 ppm for the bisphosphonate 9).</p><p>Because so few alkyl-1,1,1-trisphosphonates are known, it was unclear whether it would be possible to carry out various functional group transformations in the presence of this group. Some reactions proved to be routine while others were not. For example, treatment of allyltrisphosphonate 11 with hydrogen over Pd/C resulted in selective reduction of the olefin to afford the saturated compound 15 in good yield (Scheme 2). However, some oxidative transformations proved to be more problematic. After treatment of the allyl trisphosphonate 11 with ozone under typical conditions11 some evidence for formation of the desired aldehyde 17 was obtained, including a resonance appropriate for the aldehyde hydrogen in the 1H NMR spectrum. However, this material was obtained in low yield and there was detectable dephosphorylation to a bisphosphonate. Attempts to avoid decomposition by treatment with a limited amount of ozone under Rubin conditions12 also resulted in decomposition of the allyltrisphosphonate. Similar observations were found upon attempted epoxidation. Treatment of allyl trisphosphonate 11 with m-CPBA resulted in the disappearance of the resonances for the olefinic hydrogens, but again only low amounts of material that appeared to be the epoxide 18 were obtained and decomposition to a bisphosphonate may be competitive.</p><p>As one might expect, the trisphosphonate functionality is of considerable size, and several attempted reactions appeared to be difficult due to steric hindrance. For example, the allyl trisphosphonate 11 did not undergo hydroboration readily upon treatment with 9-BBN, but treatment with borane in THF resulted in conversion to the primary alcohol 19 in reasonable yield. The trisphosphonate group was not significantly impacted by this oxidative work-up with H2O2, as might be expected given that hydrogen peroxide was used during the trisphosphonate synthesis. Cross metathesis reactions of trisphosphonate 11 also may be affected by the size and proximity of the trisphosphonate moiety. Attempted cross metathesis with 2-methyl-2-butene and the Grubbs second generation catalyst gave the unexpected cis and trans 1,2-disubstituted olefins 20 as the major product, and only a small amount of the expected trisubstituted alkene 12.13 The identity of the olefins 20 was established unequivocally when the cross metathesis reaction of trisphosphonate 11 with 2-butene also gave a mixture of the same cis and trans olefins 20.</p><p>The importance of steric factors in the reactivity of compound 11 may be clarified by consideration of the reactivity of trisphosphonate 14, which can be viewed as a less sterically congested analogue where the number of methylene carbons between the double bond and the trisphosphonate group has been increased. In this case, attempted cross metathesis of compound 14 with 2-methyl-2-butene proceeded smoothly and gave the expected trisubstituted olefin 21 in high yield (87%, Scheme 3). In a similar sense, treatment of olefin 14 with 9-BBN followed by standard oxidative work-up gave the primary alcohol 22 in reasonable yield (57%).</p><p>While metathesis reactions of the trisphosphonate 11 may be limited in their ability to afford a diverse array of new trisphosphonates, this is clearly a promising approach with more distal olefins such as compound 14. Another approach to facile preparation of compound libraries is based on the 1,3-dipolar cycloaddition of azides with acetylenes (or click chemistry).14 Somewhat to our surprise given the results with the metathesis reaction, the copper-catalyzed reaction of the acetylene trisphosphonate 16 with benzyl azide proceeded smoothly to give the triazole 23 in 85% yield (Scheme 4). This reaction clearly demonstrates that the trisphosphonate group will tolerate standard conditions for this cycloaddition, and strongly suggests that more distal acetylenes would react at least as well.</p><p>Hydrolysis of these trisphosphonate esters could provide a variety of salts depending upon the extent of ester hydrolysis. Initial attempts to bring about complete hydrolysis of benzyl trisphosphonate 10 by treatment with HCl under reflux resulted in decomposition.15 Even though the corresponding benzylbisphosphonate 9 undergoes complete hydrolysis under parallel conditions,2d the more relevant comparison may be with the parent trisphosphonate 7 which also was reported to undergo decomposition when subjected to acid hydrolysis.6 Treatment of trisphosphonate 11 with TMSBr and collidine (Scheme 5)16 resulted in the formation of the TMS esters, as monitored by 31P NMR spectroscopy. Addition of 1N NaOH to a solution of the TMS ester led to complete hydrolysis and formation of the mixed sodium and collidinium salt17 (24) which could be isolated by standard work-up.</p><p>In conclusion, these studies have led to a synthesis of hexaethyl methanetrisphosphonate more efficient than the original report,6 and determined that this compound should be viewed as a strong carbon acid. They also have established a general strategy for preparation of alkyl-1,1,1-trisphosphonates from the corresponding alkyl-1,1-bisphosphonates through phosphinylation and oxidation with hydrogen peroxide. As long as steric factors from the bulky trisphosphonate group are considered, alkyl-1,1,1-trisphosphonates can undergo a variety of functional group transformations although they are sensitive to some oxidative conditions. In particular, steric factors already have led to an interesting variation on the Grubbs metathesis where a disubstituted olefin was observed as the major product from metathesis with 2-methyl-2-butene rather than the expected trisubstituted alkene. Furthermore, the ability of an acetylene trisphosphonate to undergo click chemistry suggests that libraries of trisphosphonates should be readily available. Thus it appears likely that further studies of alkyl-1,1,1-trisphosphonates will unveil other new chemistry, and that screening of trisphosphonate libraries could be used to identify biologically active compounds of this general structure. Investigations along these lines are continuing, and will be reported in due course.</p><!><p>Both THF and Et2O were distilled from sodium and benzophenone immediately prior to use. All non-aqueous reactions were performed with either oven-dried or flame-dried glassware under an argon atmosphere. Flash chromatography was performed on silica gel with an average particle size of 40-63 μm. The 1H NMR spectra were recorded at 300 MHz (75 MHz for 13C) with CDCl3 as solvent and (CH3)4Si as internal standard unless otherwise noted. The 1H NMR spectra recorded in D2O used residual H2O (4.80 ppm) as a reference, while 1,4-dioxane (67.0 ppm) was used as a reference for these 13C NMR spectra. Chemical shifts of 31P NMR spectra are reported in ppm relative to H3PO4 as an external standard. Elemental analyses were performed at a commercial facility. High resolution mass spectral analysis was performed with a quadrupole time of flight hybrid mass spectrometer with the capacity for positive and negative ionization modes. Electrospray ionization was employed with acetonitrile or aqueous (24) solutions.</p><!><p>Tetraethyl methylenebisphosphonate (251 mg, 0.87 mmol) was dissolved in benzene (5 mL) and then concentrated in vacuo to remove traces of water. After three such cycles, the residue was dissolved in THF (10 mL) and cooled to 0 °C in an ice bath. A solution of NaHMDS in THF (1.0 M, 1.3 mL, 1.3 mmol) was added, and the mixture was allowed to stir at 0 °C for 30 min after which ClP(OEt)2 (340 mg, 2.17 mmol) was added. After an additional 30 min, H2O2 (2.0 mL, 17.6 mmol) was very slowly added dropwise to the vessel. The reaction mixture was allowed to stir for one h, then diluted with brine, and extracted with CH2Cl2. The organic portions were combined, dried (MgSO4), and filtered, and the filtrate was concentrated in vacuo. Final purification via flash chromatography (silica gel, 0 to 50% EtOH in EtOAc) gave the desired trisphosphonate 7 as a light yellow oil (176 mg, 48%). Both 31P and 1H NMR data were consistent with previously reported values.5b,6</p><!><p>General procedure for the synthesis of alkyl trisphosphonates. A sample of tetraethyl benzylbisphosphonate (9)18, 19 (518 mg, 1.37 mmol) was dissolved in benzene (5 mL) and then concentrated in vacuo. After three such cycles, the residue was dissolved in THF (6.4 mL) and cooled to 0 °C in an ice bath. A solution of NaHMDS in THF (1.0 M, 2.1 mL, 2.1 mmol) was added, and the mixture was allowed to stir at 0 °C for 30 min, after which ClP(OEt)2 (437 mg, 2.74 mmol) was added. After an additional 30 min, excess H2O2 (2.0 mL, ~30% by titration) was slowly added dropwise (5 – 10 min) to the vessel. The reaction mixture was allowed to stir for one h, then diluted with brine, and extracted with CH2Cl2. The organic portions were combined, dried (MgSO4), and filtered, and the filtrate was concentrated in vacuo. Final purification via flash chromatography (silica gel, 0 to 30% EtOH in EtOAc) gave the desired trisphosphonate 10 as a clear oil (432 mg, 61%): 1H NMR δ 7.60–7.57 (m, 2H), 7.23–7.20 (m, 3H), 4.24–4.13 (m, 12H), 3.59 (q, JPH = 15.8 Hz, 2H), 1.27 (t, J = 8.9 Hz, 18H); 13C NMR (100 MHz) δ 136.3 (q, JPC = 5.9 Hz), 132.5 (2C), 126.9 (2C), 126.6, 63.4 (m, 6C), 52.8 (q, JPC = 117.8 Hz), 36.1 (q, JPC = 5.1 Hz), 16.3 (m, 6C); 31P NMR (121 MHz, CDCl3) +17.7 ppm; HRMS calcd for C20H37O9NaP3 (M+Na)+ 537.1548, found 537.1553. Anal. Calcd for C20H37O9P3·H2O: C, 45.12; H, 7.38. Found: C, 45.26; H, 7.57.</p><!><p>According to the general procedure for synthesis of alkylated trisphosphonates, allylbisphosphonate20 (985 mg, 3.0 mmol) was treated with NaHMDS (4.5 mL, 4.5 mmol) and ClP(OEt)2 (1.29 g, 8.2 mmol), and then after 30 min H2O2 (3.50 mL, 31 mmol) was added to the reaction mixture. After standard workup the product was purified via column chromatography on silica gel (0 to 40% EtOH in EtOAc) and compound 11 was isolated as a clear oil (987 mg, 71%): 1H NMR δ 6.32–6.18 (m, 1H), 5.15–5.05 (m, 2H), 4.34–4.14 (m, 12H), 2.92 (qd, JPH = 9.0 Hz, J = 6.6 Hz, 2H), 1.34 (t, J = 7.2 Hz, 18H); 13C NMR δ 134.4 (q, JPC = 6.3 Hz), 117.1, 63.4 (m, 6C), 50.5 (q, JPC = 119.8 Hz), 35.0 (q, JPC = 5.5 Hz), 16.5 (m, 6C); 31P NMR +18.0 ppm; HRMS calcd for C16H35O9NaP3 (M+Na)+, 487.1392, found 487.1407. Anal. Calcd for C16H35O9P3·H2O: C, 39.84; H, 7.73. Found: C, 40.19; H, 7.83.</p><!><p>According to the general procedure for synthesis of alkylated trisphosphonates, prenylbisphosphonate2e (494 mg, 1.39 mmol) was treated with NaHMDS (2.10 mL, 2.1 mmol) and ClP(OEt)2 (472 mg, 2.77 mmol). After 30 min H2O2 (2.00 mL, 17.6 mmol) was added to the reaction mixture. The product 12 was purified by column chromatography on silica gel (0 to 30% EtOH in EtOAc) and was isolated as a faint yellow oil (461 mg, 68%): 1H NMR δ 5.67 (t, J = 6.6 Hz, 1H), 4.29–4.17 (m, 12H), 2.86 (qd, JPH = 15.9 Hz, J = 6.6 Hz, 2H), 1.72 (s, 3H), 1.64 (s, 3H), 1.36–1.30 (m, 18H); 13C NMR δ 132.5, 120.2 (q, JPC = 6.1 Hz), 63.4–63.3 (m, 6C), 50.4 (q, JPC = 117.6 Hz), 29.5, (q, JPC = 5.4 Hz), 26.0, 17.9, 16.5–16.3 (m, 6C); 31P NMR (121 MHz, CDCl3) +18.6 ppm; HRMS calcd for C18H40O9P3 (M+H)+, 493.1885, found 493.1883.</p><!><p>According to the general procedure for synthesis of alkylated trisphosphonates, geranylbisphosphonate3e (249 mg, 0.6 mmol) was treated with NaHMDS (1.00 mL, 1.0 mmol) and ClP(OEt)2 (125 mg, 0.8mmol) and then after 30 min H2O2 (1.00 mL, 8.8 mmol) was added to the reaction mixture. After standard workup, the product 13 was purified by column chromatography on silica gel (0 to 30% EtOH in EtOAc) and was isolated as a faintly yellow oil (208 mg, 63%): 1H NMR δ 5.72 (t, J = 6.3 Hz, 1H), 5.13 (t, J = 6.2 Hz, 1H), 4.25 (q, J = 7.1 Hz, 12H), 2.87 (qd, JPH = 16.2, J = 7.1 Hz, 2H), 2.09–2.03 (m, 4H), 1.68 (s, 3H), 1.63 (s, 3H), 1.60 (s, 3H), 1.33 (t, J = 6.6 Hz, 18H);13C NMR δ 135.8, 131.1, 124.1, 119.8 (q, JPC = 6.2 Hz), 63.1 (m, 6C), 50.2 (q, JPC = 119.6 Hz), 39.9, 29.2 (q, JPC = 5.6 Hz), 26.5, 25.5, 17.4, 16.3–16.2 (m, 6C), 16.1; 31P NMR +18.6 ppm; HRMS calcd for C23H48O9P3 (M+H)+, 561.2511, found 561.2529. Anal. Calcd for C23H47O9P3·H2O: C, 47.75; H, 8.54. Found: C, 47.98; H, 8.44.</p><!><p>Tetraethyl methylenebisphosphonate (5.31 g, 18.4 mmol) was added dropwise to a stirring suspension of NaH (810 mg, 20.2 mmol) in THF (10 mL). After 30 min, 6-bromo-1-hexene (3.00 g, 18.4 mmol) was added and the mixture was heated at reflux overnight. After the reaction mixture had cooled to room temperature, saturated NH4Cl was added and the organic and aqueous portions were separated. The aqueous portion was extracted with Et2O and the organic layers were combined, dried (MgSO4), and concentrated in vacuo. The resulting oil was purified via flash chromatography (silica gel, 10% EtOH in hexanes) and the desired bisphosphonate was isolated in 46% yield (3.10 g): 1H NMR δ 5.84–5.73 (m, 1H), 5.03–4.92 (m, 2H), 4.23–4.12 (m, 8H), 2.27 (tt, JPH = 24.3 Hz, J = 6.3 Hz, 1H), 2.01–1.83 (m, 4H) 1.64–1.54 (m, 2H), 1.45–1.32 (m, 14H); 13C NMR δ 138.6, 114.3, 62.4–62.1 (m, 4C) 36.6 (t, JPC = 132.5 Hz), 28.5 (2C), 25.3, 16.3–16.2 (m, 4C); 31P NMR +23.9 ppm; HRMS calcd for C15H33O6P2 (M+H)+, 371.1752, found 371.1745.</p><!><p>According to the general procedure for synthesis of alkylated trisphosphonates, tetraethyl 6-hepten-1,1-bisphosphonate (508 mg, 1.37 mmol) was treated with NaHMDS (2.06 mL, 2.06 mmol) and ClP(OEt)2 (438 mg, 2.75 mmol), and then H2O2 (2.00 mL, 17.6 mmol) was added. After standard workup the product 14 was purified by column chromatography on silica gel (0 to 30% EtOH in EtOAc) and was isolated as a faintly yellow oil (435 mg, 78%): 1H NMR (CDCl3) δ 5.89–5.75 (m, 1H), 5.04–4.92 (m, 2H), 4.29–4.19 (m, 12H), 2.11–2.00 (m, 4H; 2 exchange with D2O) 1.91–1.80, (m, 2H) 1.43–1.13 (m, 22H); 13C NMR (CDCl3) δ 138.8, 114.2, 63.4–63.1 (m, 6C), 50.6 (q, JPC = 119.3 Hz), 33.4, 30.8 (q, JPC = 5.3 Hz), 29.3, 25.2 (q, JPC = 5.0 Hz), 16.5–16.2 (m, 6C); 31P NMR (121 MHz, CDCl3) +18.7 ppm; HRMS calcd for C19H41O9NaP3 (M+Na)+, 529.1861, found 529.1867.</p><!><p>According to the general procedure for synthesis of alkylated trisphosphonates, propylbisphosphonate19, 21 (292 mg, 0.9 mmol) was treated with NaHMDS (1.30 mL, 1.3 mmol) and ClP(OEt)2 (346 mg, 2.2 mmol). After 30 min H2O2 (2.00 mL, 17.6 mmol) was added to the reaction mixture. Standard workup and purification via column chromatography on silica gel (0 to 35% EtOH in EtOAc) gave compound 15 as a clear oil (324 mg, 79%). Both the 31P and 1H NMR spectra are consistent with material prepared via hydrogenation of compound 11 (vide infra).</p><!><p>According to the general procedure for synthesis of alkylated trisphosphonates, propargylbisphosphonate22 (291 mg, 0.9 mmol) was treated with NaHMDS (0.9 mL, 0.9 mmol) and ClP(OEt)2 (280 mg, 1.8 mmol) and then after 30 min H2O2 (2.00 mL, 17.6 mmol) was added to the reaction mixture. After standard workup the product was purified via column chromatography on silica gel (0 to 30% EtOH in EtOAc) and compound 16 was isolated as a clear oil (243 mg, 59%): 1H NMR δ 4.33–4.21 (m, 12H), 3.03 (qd, JPH = 14.7 Hz, J = 3.3 Hz 2H), 2.09–2.07 (m, 1H), 1.34–1.32 (m, 18H); 13C NMR δ 79.9 (q, JPC = 9.1 Hz), 70.5, 63.9–63.6 (m, 6C), 49.7 (q, JPC = 120.0 Hz), 21.0 (q, JPC = 5.6 Hz), 16.4–16.3 (m, 6C); 31P NMR +16.7 ppm; HRMS calcd for C16H34O9P3 (M+H)+, 463.1416, found 463.1420.</p><!><p>Trisphosphonate 11 (96 mg, 0.2 mmol) in EtOH (5 mL) was treated with Pd/C (23 mg, 0.2 mmol) under an H2 atmosphere. After 12 h the reaction mixture was filtered through celite, and the filtrate was collected and concentrated in vacuo. The resulting oil was purified using flash chromatography (silica gel, 0 to 25% EtOH in EtOAc) to obtain compound 15 as a clear oil (81 mg, 84%): 1H NMR δ 4.30–4.18 (m, 12H), 2.18–1.77 (m, 4H), 1.34 (t, J = 6.6 Hz, 18H), 0.91 (t, J = 6.9 Hz, 3H); 13C NMR δ 63.4 (m, 6C), 51.6 (q, JPC = 119.8 Hz), 33.0 (q, JPC = 5.5 Hz), 19.2 (q, JPC = 5.2 Hz) 16.4 (m, 6C), 15.0; 31P NMR +18.8 ppm; HRMS calcd for C16H37O9NaP3 (M+Na)+, 489.1548, found 489.1564. Anal. Calcd for C16H37O9P3·H2O: C, 39.67; H, 8.11. Found: C, 39.66; H, 8.02.</p><!><p>Trisphosphonate 11 (102 mg, 0.22 mmol) was dried under vacuum in the presence of P2O5 overnight. The remaining oil was dissolved in THF (5 mL) and placed into an ice bath. To the reaction flask, BH3·THF (1M in THF, 0.45 mL, 0.45 mmol) was added and the mixture was allowed to stir. After 1.5 h, MeOH (2 mL) was added to the flask, followed by NaOH (3M, 0.5 mL, 1.5 mmol) and then H2O2 (0.3 mL, 2.7 mmol), and the resulting mixture was heated at 50 °C for 1 h. The reaction mixture was washed with saturated NaCl and the aqueous portions were retained and extracted with CH2Cl2. The organic portions were combined, dried (MgSO4), and concentrated in vacuo. The resulting oil was purified via flash chromatography (silica gel, 0 to 45% EtOH in EtOAc) to obtain compound 19 as a clear oil (66 mg, 62%): 1H NMR δ 4.33–4.20 (m, 12H), 3.64 (t, J = 5.7 Hz, 2H), 2.30–2.05 (m, 7H; 2 exchange with D2O) 1.35 (t, J = 6.6 Hz, 18H); 13C NMR δ 63.5–63.4 (m, 6C), 63.0, 51.6 (q, JPC = 119.8 Hz), 33.0 (q, JPC = 5.5 Hz), 19.2 (q, JPC = 5.2 Hz) 16.4 (m, 6C); 31P NMR +18.8 ppm; HRMS calcd for C16H37O10NaP3 (M+Na)+, 505.1497, found 505.1503.</p><!><p>Grubbs second generation catalyst (4.9 mg, 3 mol %) was dissolved in 2-methyl-2-butene (1 mL) and placed in a 1-dram vile. The trisphosphonate 11 (88.3 mg, 0.2 mmol) was added to this mixture, along with an additional 1 mL of 2-methyl 2-butene. The vile was sealed and the reaction was allowed to stir at 40 °C overnight. After the solvent was removed in vacuo, the resulting oil was purified via flash chromatography (silica gel, 0 to 30% EtOH in EtOAc). The reaction products (76 mg, 83% total) were isolated as an inseparable mixture of cis and trans isomers of compound 20 (68%, 1.2:4.3 isomer ratio) and prenyl trisphosphonate 12 (15%, 1:5.5 ratio with respect to olefins 20). The 31P, 1H and 13C NMR spectra were consistent with a mixture of compounds 20 and 12, both of which had been prepared independently.</p><!><p>Grubbs second generation catalyst (2.3 mg, 6 mol %) was dissolved in CH2Cl2 (0.5 mL) and placed in a 1-dram vile, and trisphosphonate 11 (24 mg, 0.1 mmol) was added to this mixture. After 2-butene was added to the vessel via balloon, the vessel was sealed and the reaction was allowed to stir at 40 °C overnight. The volatile materials were removed in vacuo and the resulting oil was purified via flash chromatography (silica gel, 0 to 30% EtOH in EtOAc). The olefins 20 were isolated as a mixture of trans and cis isomers (19 mg, 78%) in a 2.8:1 ratio. For the trans isomer: 1H NMR (500 MHz, CDCl3) δ 5.85 (dt, J = 14.0, 7.0 Hz, 1H), 5.57–5.50 (m, 1H), 4.29–4.18 (m, 12H), 2.89–2.84 (m, 2H), 1.67 (dd, J = 7.0, 1.5 Hz, 3H), 1.35–1.32 (m, 18H); 13C NMR (125 MHz, CDCl3) δ 128.0, 126.7 (q, JPC = 6.3 Hz), 63.4–63.3 (6C), 50.7 (q, JPC = 119.5 Hz), 33.9 (q, JPC = 5.3 Hz), 17.9, 16.5–16.3 (6C); 31P NMR (121 MHz, CDCl3), +18.4 ppm. For the cis isomer: 1H NMR (500 MHz, CDCl3), δ 5.94–5.92 (m, 1H), 5.57–5.50 (m, 1H), 4.29–4.18 (m, 12H), 2.89–2.84 (m, 2H), 1.64 (dd, J = 7.0, 1.0 Hz, 3H), 1.35–1.32 (m, 18H); 13C NMR (125 MHz, CDCl3) 126.1 (q, JPC = 6.0 Hz), 125.0, 63.5–63.4 (6C), 50.1 (q, JPC = 119.6 Hz), 28.3 (q, JPC = 8.0 Hz), 16.5–16.3 (6C), 12.9; 31P NMR (121 MHz, CDCl3) +18.5 ppm; HRMS calcd for C17H37O9NaP3 (M+Na)+, 501.1548, found 501.1554. Anal. Calcd for C17H37O9P3·H2O: C, 41.13; H, 7.92. Found: C, 41.35; H, 7.91.</p><!><p>Grubbs second generation catalyst (3.1 mg, 3 mol %) was dissolved in 2-methyl-2-butene, placed in a 1-dram vile, and trisphosphonate 14 (62 mg, 0.12 mmol) was added along with an additional 1 mL of 2-methyl-2-butene. The vile was sealed and the reaction was allowed to stir at 40 °C overnight. After concentration in vacuo, the resulting oil was purified via flash chromatography (silica gel, 0 to 30% EtOH in EtOAc), and the desired product 21 was isolated as an oil (57 mg, 87%): 1H NMR δ 5.12 (t, J = 6.0 Hz, 1H), 4.19–4.31 (m, 12H), 1.96–2.24 (m, 6H), 1.77–1.88 (m, 2H), 1.68 (3H), 1.57 (3H), 1.34 (t, J = 6.6 Hz, 18H); 13C NMR δ 131.2, 124.7, 63.5–63.2 (m, 6C), 50.7 (q, JPC = 119.4 Hz), 31.0–30.9 (m), 30.9, 27.8, 25.7, 25.4 (q, JPC = 5.2 Hz) 17.6, 16.4–16.2 (m, 6C); 31P NMR +18.8 ppm; HRMS calcd for C21H46O9P3 (M+H)+, 535.2355, found 535.2357.</p><!><p>Trisphosphonate 14 (109 mg, 0.2 mmol) was dried overnight under vacuum in the presence of P2O5. The remaining oil was dissolved in THF (5 mL) and placed into an ice bath. To the reaction flask, 9-BBN (0.5 M in THF, 1.0 mL, 0.5 mmol) was added and the mixture was allowed to stir. After 1.5 h, MeOH (2 mL) was added to the flask, followed by NaOH (3 M, 0.5 mL, 1.5 mmol) and then H2O2 (0.5 mL, 4.4 mmol), and the resulting mixture was heated at 50 °C for 1 h. The reaction mixture was washed with saturated NaCl and the aqueous portions were retained and extracted with CH2Cl2. The organic portions were combined, dried (MgSO4), and concentrated in vacuo. The resulting oil was purified via flash chromatography (silica gel, 0 to 40% EtOH in EtOAc) to obtain compound 22 as a clear oil (64 mg, 57%): 1H NMR δ 4.30–4.17 (m, 12H), 3.63 (t, J = 6.3 Hz, 2H), 2.18–2.03 (m, 3H), 1.90–1.82 (m, 2H), 1.60–1.53 (m, 2H), 1.44–1.28 (m, 22); 13C NMR δ 63.5–63.2 (m, 6C), 62.8, 50.6 (q, JPC = 119.5 Hz), 32.7, 30.8 (q, JPC = 5.3 Hz), 30.3, 25.5 (q, JPC= 5.3 Hz), 25.3, 16.5–16.2 (6C); 31P NMR +18.8 ppm; HRMS calcd for C19H43O10NaP3 (M+Na)+, 547.1967, found 547.1991.</p><!><p>Benzyl bromide (182 mg, 1.1 mmol) was added to a suspension of sodium azide (83 mg, 1.3 mmol) in DMF (5 mL) and the resulting mixture was allowed to stir. After 10 min, trisphosphonate 16 (164 mg, 0.4 mmol) was added along with 0.1 mL CuSO4 (5M), sodium ascorbate (43 mg, 0.2 mmol), and a solution of tBuOH in water (1:4 ratio, 5 mL), and the reaction mixture was allowed to stir at room temperature. After 24 h EDTA and 1M NH4OH were added, the resulting solution was placed in a continuous liquid-liquid extractor and extracted for 4 h with EtOAc. The organic portion was retained and concentrated in vacuo. The resulting oil was purified via flash chromatography (silica gel, 0 to 50% EtOH in EtOAc) to provide the desired triazole 23 (179 mg, 85%): 1H NMR δ 7.95 (s, 1H), 7.35–7.31 (m, 5H), 5.47 (s, 2H), 4.22–4.07 (m, 12H), 3.66 (q, JPH = 15.9 Hz, 2H), 1.26–1.21(m, 18H); 13C NMR δ 143.3 (q, JPC = 7.4 Hz), 135.2, 128.9 (2C), 128.4, 128.0 (2C), 124.7, 63.7–63.4 (m, 6C), 53.9, 50.6 (q, JPC = 119.2 Hz), 27.9, (q, JPC = 5.5 Hz), 16.4–16.1 (m, 6C); 31P NMR +17.6 ppm; HRMS calcd for C23H40N3O9NaP3 (M+Na)+, 618.1875, found 618.1893.</p><!><p>A solution of 2,4,6-collidine (524 mg, 4.3 mmol) and TMSBr (568 mg, 4.3 mmol) was allowed to stir in an ice bath. After 20 min, trisphosphonate 11 (75mg, 0.2 mmol) was added and the reaction was allowed to stir for 24 h with periodic monitoring by 31P NMR spectroscopy. Once the reaction was complete, it was diluted by addition of toluene, the solvent was removed in vacuo, and aqueous sodium hydroxide (1.5 mmol, 9 eq) was added. The mixture was allowed to stir overnight and again was monitored by 31P NMR. The reaction mixture then was lyophilized, the resulting solid was dissolved in a minimum amount of water, then slowly poured into cold acetone and kept at 40 °C overnight. The resulting precipitate was filtered and washed with cold acetone. The remaining residue was dissolved in water and lyophilized to afford compound 24 as a flocculent white residue (51 mg, 60%): 1H NMR (D2O) δ 7.43 (s, 2H), 6.19–6.10 (m, 1H), 5.23–5.05 (m, 2H), 2.92– 2.85 (m, 2H), 2.93 (s, 6H), 2.50 (s, 3H); 13C NMR (D2O) δ 164.1 (2C), 155.8, 139.1–139.0 (m), 129.2, 121.6 (2C), 52.3 (q, JPC = 103.6 Hz), 38.6. 25.3 (2C), 22.5; 31P NMR (121 MHz, D2O) +17.5 ppm; HRMS calcd for C4H10O9P3 (M–H)−, 294.9538, found 294.9542.</p>

PubMed Author Manuscript

The lysosomal protein ABCD4 can transport vitamin B12 across liposomal membranes in vitro

Vitamin B12 (cobalamin) is an essential micronutrient for human health, and mutation and dysregulation of cobalamin metabolism are associated with serious diseases, such as methylmalonic aciduria and homocystinuria. Mutations in ABCD4 or LMBRD1, which encode the ABC transporter ABCD4 and lysosomal membrane protein LMBD1, respectively, lead to errors in cobalamin metabolism, with the phenotype of a failure to release cobalamin from lysosomes. However, the mechanism of transport of cobalamin across the lysosomal membrane remains unknown. We previously demonstrated that LMBD1 is required for the translocation of ABCD4 from the endoplasmic reticulum to lysosomes. This suggests that ABCD4 performs an important function in lysosomal membrane cobalamin transport. In this study, we expressed human ABCD4 and LMBD1 in methylotrophic yeast and purified them. We prepared ABCD4 and/or LMBD1 containing liposomes loaded with cobalamin and then quantified the release of cobalamin from the liposomes by reverse-phase HPLC. We observed that ABCD4 was able to transport cobalamin from the inside to the outside of liposomes dependent on its ATPase activity and that LMBD1 exhibited no cobalamin transport activity. These results suggest that ABCD4 may be capable of transporting cobalamin from the lysosomal lumen to the cytosol. Furthermore, we examined a series of ABCD4 missense mutations to understand how these alterations impair cobalamin transport. Our findings give insight into the molecular mechanism of cobalamin transport by which ABCD4 involves and its importance in cobalamin deficiency.

the_lysosomal_protein_abcd4_can_transport_vitamin_b12_across_liposomal_membranes_in_vitro

<!>Reconstitution of purified ABCD4 in liposomes<!><!>ABCD4 transports cobalamin as substrate<!><!>Effect of LMBD1 on the function of ABCD4<!>Dysfunction of ABCD4 bearing disease-related missense mutations<!><!>Dysfunction of ABCD4 bearing disease-related missense mutations<!>Discussion<!>Yeast strains and media<!>Purification of His-ABCD4 and LMBD1–GST<!>Preparation of proteoliposomes<!>ATPase activity<!>Photocrosslinking<!>Cobalamin transport assay<!>Pull-down assay<!>Data availability<!>Supporting information<!>Conflict of interest<!>Supporting information

<p>Edited by Phyllis Hanson</p><p>The ABC transporters comprise a superfamily of membrane-bound proteins that exist in almost all organisms from eubacteria to mammals. In humans, there are 48 members classified into the seven subfamilies A to G, based on structural organization and amino acid homology. These proteins catalyze the ATP-dependent transmembrane transport of a wide variety of substrates in order to sustain cellular homeostasis. Defect in their functions results in various inherited metabolic diseases (1, 2).</p><p>To date, four ABCD proteins, ABCD1‒4, have been identified in humans (3, 4, 5, 6). They are half-sized ABC proteins and mainly exist in the form of a homodimer (7). ABCD1‒3 are known to be peroxisomal proteins and involved in the transport of CoA derivatives of long, very long, or branched-chain fatty acids (8, 9). On the other hand, ABCD4 is located on lysosomal membrane (10). ABCD4 only slightly interacts with peroxisomal biogenesis factor Pex19p because of lack of the NH2-terminal hydrophobic region responsible for peroxisomal targeting. As a result, ABCD4 is recognized by signal recognition particles and integrated into the endoplasmic reticulum (ER) membrane (11). The ABCD4 dimer then forms a complex with lysosomal membrane protein LMBD1, and this complex is translocated from the ER to lysosomes (12).</p><p>ABCD4 is reportedly involved in the transport of vitamin B12 (cobalamin) from the lysosomal lumen to the cytosol, since the dysfunction of ABCD4 results in the failure of lysosomal cobalamin efflux (10). In humans, cobalamin binds with transcobalamin in the blood stream, and this complex is taken up into lysosomes by transcobalamin receptor–mediated endocytosis. Subsequently, cobalamin is released into the cytosol and converted into two active cofactors: methylcobalamin and adenosylcobalamin, which are required by the cytosolic enzyme methionine synthase and the mitochondrial enzyme methylmalonyl-CoA mutase, respectively. These cofactors are indispensable for the homeostasis of homocysteine and methylmalonic acid (13). As mentioned previously, mutations of ABCD4 result in the failure of the release of cobalamin from lysosomes into the cytosol. A similar phenotype is caused by mutations of LMBRD1, which encodes the lysosomal membrane protein LMBD1 (14). LMBD1 shares a high degree of homology with the limb region protein (15) and lipocalin-1 interacting membrane receptor (16). It is reported that a small portion of LMBD1 exists on plasma membranes and is involved in the clathrin-mediated endocytosis of the insulin receptor as a specific adaptor (17). Since mutations of ABCD4 and LMBD1 result in a quite similar phenotype, these two proteins are thought to function as a complex during cobalamin metabolism.</p><p>We previously demonstrated that ABCD4 and LMBD1 form a complex on the ER membrane, and then this complex is translocated from the ER to lysosomes. In addition, this targeting ability of LMBD1 is indispensable for the targeting of ABCD4 to lysosomes (12). Thus, LMBD1 seems necessary to translocate ABCD4 to lysosomes from the ER, and ABCD4 plays a key role in the transport of cobalamin from the lysosomal lumen to the cytosol. Almost all mammalian ABC transporters, with some notable exception, localized on plasma and organelle membranes, transport the substrate from the cytosolic side to the extracellular space or the organelle lumen, which is considered to be the outside domain of cells; that is, they function as exporters (18). If ABCD4 actually transports cobalamin in this direction, ABCD4 would be the first example of a mammalian ABC importer that transports a soluble compound from the outside of cells.</p><p>In this study, to clarify the role of ABCD4 and LMBD1 in the transport of cobalamin, we expressed human ABCD4 and LMBD1 using the methylotrophic yeast Komagataella phaffii (formerly called Pichia pastoris) and reconstituted purified ABCD4 and/or LMBD1 in liposomes, as reported previously (19). Furthermore, we established a cobalamin transport assay employing ABCD4- and/or LMBD1-containing liposomes and characterized the transport mechanism. These results show that ABCD4 by itself transports cobalamin from the inside to the outside of liposomes and that LMBD1 is not required for the transport of cobalamin. These results indicate that ABCD4 functions as an importer on lysosomal membranes. Moreover, insights into the molecular mechanisms of the effect of mutant ABCD4 reported in cobalamin deficiency are provided.</p><p>ATPase activity of the reconstituted ABCD4.A, purified His-ABCD4 (Eluate) and reconstituted His-ABCD4 (liposomes) were subjected to SDS-PAGE, and the gel was stained with Coomassie brilliant blue. The arrow head and asterisk indicate His-ABCD4 and a nonspecific protein, respectively. B, photocrosslinking of His-ABCD4 was performed with succinimidyl 4,4'-azipentanoate (SDA) and UV irradiation. His-ABCD4 was detected by immunoblot analysis with an anti-His antibody. C, ATPase activity of reconstituted ABCD4 was measured. Proteoliposomes containing ABCD4 (6.96 μg) or negative control liposomes containing nonspecific protein (5.39 μg) were incubated with 5 mM ATP at 37 °C, and the phosphate that was released was measured. The ATPase activity of ABCD4 (•) and the negative control (○) are shown. Error bars indicate the standard deviation (n = 3). Differences between the ABCD4 and negative control were considered significant when p < 0.05 or p < 0.01 based on Student's t test (∗p < 0.05; ∗∗p < 0.01). D, different amounts of ABCD4-liposomes or negative control liposomes were incubated with 5 mM ATP at 37 °C for 30 min, and the phosphate that was released was measured. The ATPase activities of ABCD4 (open bar) and a nonspecific protein (gray bar) are shown. Error bars indicate the standard deviation (n = 3). Differences between ABCD4 and the negative control were considered significant when p < 0.05 or p < 0.01 based on Student's t test (∗p < 0.05; ∗∗p < 0.01).</p><!><p>To examine proper insertion of ABCD4 on liposomes, we evaluated the proportion of the right-side-out orientation. ABCD4-liposomes were treated with trypsin and subjected to immunoblot analysis using anti-ABCD4 antibody recognizing C-terminal 149 amino acid of ABCD4. As shown in Fig. S2A, approximately 40 kDa band emerged. This band seems to be derived from ABCD4 existing in inside-out orientation. Approximately 70% of ABCD4 reconstituted into liposomes existed in right-side-out orientation calculated from the signal intensities (Fig. S2B). The 110 kDa band is a nonspecific protein derived from the yeast cells during His-tag affinity chromatography. This protein is not an ABCD4-associated protein because it was not coimmunoprecipitated with ABCD4 (Fig. S3A). Since this protein is also reconstituted in liposomes, the liposomes containing only this protein were used as a negative control (Fig. S3B).</p><p>As mentioned previously, ABCD4 functions as a homodimer in mammalian cells (9, 20). We examined whether ABCD4 is reconstituted in liposomes as a dimer or not. ABCD4-liposomes were incubated with succinimidyl 4,4'-azipentanoate (SDA), a cross-linking reagent. As shown in Figure 1B, ABCD4 displayed a ∼140 kDa band consistent in mass with dimeric ABCD4 when SDA was photoactivated by UV irradiation. Subsequently, the ATPase activity of liposome-reconstituted ABCD4 was assayed based on the release of organic phosphate from ATP. The reconstituted ABCD4 exhibited ATPase activity, and the activity was increased in a time- and dose-dependent manner (Fig. 1, C and D). These results indicate that ABCD4 was successfully reconstituted in liposomes in the active form.</p><!><p>Effect of cobalamin on the ATPase activity of ABCD4.A, ATPase activity of reconstituted ABCD4 was measured in the presence of 1 mM cobalamin inside or outside of the liposomes. Proteoliposomes with or without cobalamin containing ABCD4 (6.76 μg or 5.08 μg, respectively) were incubated with ATP at 37 °C for 30 min. Error bars indicate the standard deviation (n = 3). Differences among the conditions were considered significant when p < 0.05 based on Student's t test (∗p < 0.05). B, ATPase activity of reconstituted ABCD4 was measured in the presence of various concentrations of cobalamin in the liposome lumen. Error bars indicate the standard deviation (n = 3), when not shown, fall within the symbol.</p><p>Cobalamin transport activity of ABCD4.A, ABCD4-liposomes containing 100 μM of cobalamin were prepared. Proteoliposomes containing ABCD4 wildtype or K427A (5.54 μg or 4.95 μg, respectively) were incubated with ATP at 37 °C. The amount of cobalamin transported from inside of the liposomes at each time point was calculated by subtraction of the amount of cobalamin in the liposomes in each incubation from that without any incubation. The cobalamin transport activity of wildtype ABCD4 (•), ABCD4 (K427A) (▲), and the negative control (○) are shown. Error bars indicate the standard deviation (n = 3), when not shown, fall within the symbol. B, cobalamin containing ABCD4-liposomes with different internal pH conditions was incubated with ATP, and then the amount of cobalamin inside of the liposomes was quantified as in (A). Proteoliposomes at pH 5.5 or pH 7.5 containing ABCD4 (10.1 μg or 10.2 μg, respectively) were used for this assay. Error bars indicate the standard deviation (n = 3), when not shown, fall within the symbol.</p><!><p>Next, to test whether cobalamin transport depends on ATPase activity, we prepared ABCD4 (K427A), a Walker A lysine mutant deficient in ATPase activity. We also expressed His-ABCD4 (K427A) under the control of the AOX1 promoter in K. phaffii. His-ABCD4 (K427A) was purified and reconstituted in liposomes using the same procedure as used for His-ABCD4 (Fig. S7A). It was confirmed that this mutant had lost ATPase activity (Fig. S7B). In the transport assay, cobalamin was not released from ABCD4 (K427A)-liposomes loaded with cobalamin at any time point (Fig. 3A). This indicates that ABCD4 (K427A) had lost transport activity in addition to ATPase activity. These results demonstrated that ABCD4 transports cobalamin from the inside to the outside of liposomes in a manner that is dependent on ATPase activity.</p><!><p>Effect of LMBD1 on the enzyme activities of ABCD4.A, interaction analysis between purified ABCD4 and LMBD1. The eluates of His-ABCD4 and LMBD1–glutathione-S-transferase (GST) were mixed and rotated at 4 °C for 15 min. Then the mixture was incubated with His-tag affinity resin overnight at 4 °C. Coprecipitated proteins were analyzed by SDS-PAGE followed by immunoblotting using an anti-His or an anti-GST antibody. The ATPase activity (B) and cobalamin transport activity (C) of ABCD4 and/or LMBD1 liposomes were evaluated. Proteoliposomes containing ABCD4 (3.91 μg), LMBD1 (3.75 μg), or ABCD4 and LMBD1 (2.34 and 2.54 μg, respectively). Error bars indicate the standard deviation (n = 3). Differences among the liposomes were considered significant when p < 0.01 based on Student's t test (∗∗p < 0.01).</p><!><p>Subsequently, we prepared ABCD4/LMBD1-liposomes. It was confirmed that ABCD4 and LMBD1 maintained the same binding stoichiometry in ABCD4/LMBD1-liposomes (Fig. S9C). ABCD4/LMBD1-liposomes also exhibited both ATPase and cobalamin transport activity. There was no difference in either the ATPase or the cobalamin transport activities compared with the ABCD4-liposomes (Fig. 4, B and C). LMBD1 itself had neither ATPase nor cobalamin transport activities (Fig. 4, B and C).</p><!><p>To date, eight clinical mutations in the ABCD4 gene that result in cobalamin deficiency have been reported (21). Among them, there are three missense mutations, N141K, Y319C, and R432Q. Arg432 is located on or very close to the Walker A motif, which is indispensable for ATPase activity, and it is reported that the R432Q mutant exhibits much lower ATPase activity (22). It is also reported that Asn141 and Tyr319 are located on the cytosolic side of transmembrane helix 3 (TM3) and the lysosomal side of transmembrane helix 6 (TM6), respectively, as shown by cryo-EM (22). We focused on the role of these amino acid residues during the transport of cobalamin mediated by ABCD4.</p><p>We initially examined the subcellular localization of ABCD4 (N141K) and ABCD4 (Y319C) in mammalian cells. We previously established CHO cells stably expressing LMBD1–GFP and demonstrated that the ABCD4–hemagglutinin (HA) transiently expressed in these cells is localized to lysosomes (12). ABCD4 (N141K)–HA or ABCD4 (Y319C)–HA was also transiently expressed in these CHO cells. The distribution of ABCD4 (N141K)–HA or ABCD4 (Y319C)–HA partially coincided with that of LMBD1–GFP (Fig. S10), indicating that a portion of both ABCD4 (N141K) and ABCD4 (Y319C) is localized in lysosomes, as in the case of wildtype ABCD4.</p><!><p>ATPase and cobalamin transport activities of mutant ABCD4. The ATPase activity (A) and cobalamin transport activity (B) of mutant ABCD4 were evaluated. Error bars indicate the standard deviation (n = 3). Proteoliposomes containing ABCD4 wildtype, N141A, N141D, N141K, Y319A, Y319C, or Y319F (6.62, 6.79, 7.63, 6.49, 5.35, 6.38, or 9.48 μg, respectively) were used for these assays. Differences between the wildtype and mutant were considered significant when p < 0.05 or p < 0.01 based on Student's t test (∗p < 0.05; ∗∗p < 0.01).</p><!><p>Asn141 is located on the cytosolic side of TM3 and faces the transmembrane cavity and is deduced to form a hydrogen bond with Asp221, which is located on the cytosolic side of TM4 (22). The hydrophobic or ionic environment of this area might be important for the transport of cobalamin through the conformational change of the TMs associated with the ATP binding and hydrolysis at the nucleotide-binding domain (NBD). We changed the mutated Lys141 to Ala141 or Asp141 and measured the ATPase and cobalamin transport activities. As shown in Figure 5A, ABCD4 (N141A) and ABCD4 (N141D) also exhibited ATPase activity equal to that of the wildtype. Interestingly, the cobalamin transport activity was recovered to the same level as the wildtype in both ABCD4 (N141A) and ABCD4 (N141D) (Fig. 5B). These results indicate that an increase in the anionic charge in this area by substitution of Asn141 to Lys141 might disturb the cobalamin transport of ABCD4.</p><p>In the case of Tyr319, substitution to Cys resulted in the loss of the ATPase activity as well as cobalamin transport activity. Surprisingly, Tyr319 is located on the lumen side of TM6 far away from the NBD that is indispensable for ATPase activity (22). To evaluate the role of Tyr319, we first tested whether ATP can access the NBD in ABCD4 (Y319C) using an ATP probe for photoaffinity labeling (Fig. S12A), since ABCD4 (Y319C) lacks ATPase activity. As a result, ABCD4 (Y319C) was labeled by the ATP probe under UV irradiation, and this labeling was inhibited by the presence of adenylyl imidodiphosphate, a nonhydrolyzable analog of ATP, in a dose-dependent manner (Fig. S12B). These results suggest that the conformational change coupled to the hydrolysis of ATP was disturbed in the Y319C mutant. Recently, in Cyanidioschyzon merolae ABCB1, the van der Waals contacts and hydrogen-bonding network in the cluster of aromatic hydrophobic side chains in the upper part of the transmembrane helices were shown to play an important role in the interactions among the transmembrane helices (23, 24). The Tyr319 residue in ABCD4 is deduced to be involved in the corresponding cluster. Therefore, we prepared ABCD4 (Y319A) and ABCD4 (Y319F) and evaluated both the ATPase and cobalamin transport activities. As shown in Figure 5, A and B, neither the ATPase activity nor the transport activity was recovered in ABCD4 (Y319A). In contrast, ABCD4 (Y319F) recovered both the ATPase and transport activities (Fig. 5, A and B). These results indicate that Tyr319 in ABCD4 plays an important role in the conformational change related to cobalamin transport.</p><!><p>Since mutations in the genes ABCD4 and LMBRD1, which encode ABCD4 and LMBD1, respectively, result in a failure to release cobalamin from lysosomes (10, 14), ABCD4 and LMBD1 are considered to be coordinatively involved in the transport of cobalamin across the lysosomal membrane. We previously demonstrated that translocation of ABCD4 from the ER to lysosomes is assisted by LMBD1 since ABCD4 itself only targets the ER (12). However, ABCD4 exists in a complex with LMBD1 on the lysosomal membrane, and the molecular mechanism of cobalamin transport mediated by ABCD4/LMBD1 has yet to be elucidated including whether ABCD4 is by itself sufficient for cobalamin transport. Previously, we reconstituted purified ABCD4 in liposomes and showed ABCD4 possesses ATPase activity. Here, we characterized transport mechanism of cobalamin by ABCD4 with or without LMBD1 and how the transport is impaired by disease-related mutations of ABCD4.</p><p>In our assay system, the ATPase activity of ABCD4 was stimulated only when cobalamin is present inside the liposomes and ATP was added to the outside (Fig. 2). In addition, ABCD4 was able to transport cobalamin from the inside to the outside of liposomes in a manner that was dependent on its ATPase activity (Fig. 3A). It is predicted that the COOH-terminal half of ABCD4, including the NBD, is exposed to the cytosolic surface (10, 11). Therefore, this system reproduces the transport of cobalamin from the lysosomal lumen to the cytosol. The transport of one molecule of cobalamin requires the hydrolysis of approximately 50 molecules of ATP. This estimate is comparable to that of the transport of several different substances by the multidrug resistance protein 1/ABCC1 reconstituted in liposomes (25, 26). ABCC1 is known to export cobalamin into the extracellular space from the cytosol (27). Recently, as ABCD4 and LMBD1 have been shown to interact with the cobalamin processing proteins, methylmalonic aciduria and homocystinuria type C protein and methylmalonic aciduria and homocystinuria type cblD (20), it has been hypothesized that cobalamin might be transported more efficiently by ABCD4 from lysosomes in vivo.</p><p>Concerning the direction of the transport carried out by the ABCD4 family of transporters, ABCD1‒3 transport fatty acid CoAs into peroxisomes from the cytosol. ABCD4 transports cobalamin from the lysosome lumen to opposite side. ABCD1‒4 are half-sized transporters and exist as a homodimer on the membrane. They exhibit the same orientation as that of the NBD, which is exposed to the cytosolic surface (10, 11). It is interesting that ABCD1‒3 and ABCD4 recognize substrate on the cytosolic or lumenal face of lysosomes, respectively, and transport these substrates in different directions via conformational changes of the transmembrane domain coupled with ATP binding and hydrolysis of the NBD. ABC transporters are categorized into two types: importers and exporters. ABC importers, which transport substrates into the cytosol from the extracellular space, are restricted to prokaryotes, with only a few exceptions in plants and mammals. Recently, ABCA4, cystic fibrosis transmembrane conductance regulator (ABCC7), SUR1 (ABCC8), and SUR2 (ABCC9) are categorized as importers (18). ABCA4 functions as flippase that translocates N-retinylidene-phosphatidylethanolamine from the lumen to the cytoplasmic leaflet of the disc membrane in retinal photoreceptor cells (28). Cystic fibrosis transmembrane conductance regulator functions as an ATP-gated chloride channel (29), and SUR1 and SUR2 are regulatory elements of the hetero-octameric ATP-sensitive potassium (KATP) channels built from four SUR subunits (30). Therefore, ABCD4 is the first mammalian ABC importer shown to transport a soluble compound.</p><p>LMBD1 regulates the lysosomal translocation of ABCD4 and forms a complex with ABCD4 on the lysosomal membrane (12, 21). However, LMBD1 itself did not exhibit any cobalamin transport activity and also did not assist either the ATPase or the cobalamin transport activity of ABCD4 (Fig. 4, B and C). These results indicate that LMBD1 is not directly involved in cobalamin transport on the lysosomal membrane, although the dysfunction of LMBD1 did result in a failure in lysosomal cobalamin release. Lysosomes contain more than 60 hydrolases to degrade extracellular materials as well as intracellular components. Many of the lysosomal membrane proteins are highly glycosylated within the lysosomal lumen in order to let them escape lysosomal proteolysis (31). LMBD1 possesses six putative glycosylation sites and is, in fact, glycosylated (14, 20). In contrast, ABCD4 possesses two potential glycosylation sites but is not glycosylated in mammalian cells (11). It is likely that LMBD1 contributes to the protection of ABCD4 from lysosomal proteolysis by forming a complex with ABCD4.</p><p>There are eight clinically relevant mutations in ABCD4, including two in-frame deletions, three flame shifts, and three point mutations, N141K, Y319C, and R432Q. Arg432 is located next to the Walker A motif, and ABCD4 (R432Q) exhibits much lower ATPase activity than the wildtype (22). Since it was demonstrated that ABCD4 transports cobalamin in a manner that is dependent on ATPase activity (Fig. 3A), the defect of cobalamin metabolism in the R432Q mutant is almost certainly because of the loss of ATPase activity.</p><p>On the other hand, Asn141 and Tyr319 are located on the cytosolic side of TM3 and the lysosomal side of TM6, respectively. ABCD4 (N141K) lacks only cobalamin transport activity, whereas ABCD4 (Y319C) lacks both ATPase and cobalamin transport activity (Fig. 5, A and B). From the amino acid coordinates of ABCD4 deposited in the Protein Data Bank (6JBJ), Asn141 faces the transmembrane cavity (22). The hydrophobic or ionic environment of the area might be important for the transport of cobalamin via the conformational change that is associated with the ATP binding and hydrolysis of the NBD. Since transport activity was recovered in ABCD4 (N141A) and ABCD4 (N141D) (Fig. 5B), an increase in the cationic charge around Asn141 by substitution to Lys141 might disturb the cobalamin transport of ABCD4. Asn141 is believed to form a hydrogen bond with Asp221 located on the cytosolic side of TM4. However, the substitution of Asn141 to Ala141, which does not form a hydrogen bond with Asp221, did not influence the cobalamin transport activity of ABCD4. It is thus suggested that the hydrogen bond between Asn141 and Asp221 is not important for cobalamin transport. From the cryo-EM structure, the distance between Asn141 and Asp221 is very close (2.5 Å) (Fig. S13). Since the Lys residue possesses a large side chain with a positive charge, the local environment around Lys141 might be altered and TM3 becomes strained. Therefore, dimerization of the NBD accompanied by ATP hydrolysis might be linked with the conformational change in the transmembrane domain required to transport cobalamin.</p><p>Tyr319 is located on the lysosomal side of TM6 and deduced to be involved in the hydrophobic entrance of the transmembrane cavity (22). Since the Y319C mutant lacks the ATPase as well as transport activity (Fig. 5, A and B), disturbance in the accessibility of cobalamin to the cavity does not seem to be the exclusive reason for the impaired transport of cobalamin. It is speculated that in the Y319C mutant, the two Cys319 residues have a tendency to form a disulfide bond in a dimeric conformation (22) and, as a result, there is a certain impairment in the conformational change that might impact transport. However, disulfide bond formation is not the reason, since ABCD4 (Y319A) also lost ATPase as well as cobalamin transport activity (Fig. 5, A and B). ABCD4 (Y319C) was able to bind but not hydrolyze ATP (Fig. 5A and Fig. S12B). These results indicate that the Y319C mutant has an additional conformational change impairment. Recently, the inward- and outward-facing X-ray crystal structure of C. merolae ABCB1 was reported, and the structural features during conformational realignment revealed based on high-resolution crystallographic data (24). The aromatic amino acid residues in the upper part of the transmembrane helices of TM4 and TM6 form van der Waals contacts and hydrogen-bonding networks as a hydrophobic cluster and are involved in the extracellular gating that occurs during the conformational changes that take place along with ATP binding. ABCD4 possesses a similar cluster as TM4 and TM6. The Tyr319 residue in TM6 is deduced to form the cluster and play an important role during the conformational change associated with ATP binding and hydrolysis because the substitution of Tyr319 to Phe with an aromatic ring restored the ATPase as well as cobalamin transport activities of ABCD4 (Fig. 5, A and B). Further crystallographic study of ABCD4 will be needed.</p><p>In this study, it was demonstrated for the first time that ABCD4 transports cobalamin from the lysosomal lumen to the cytosol as an "importer." In contrast, other members of the ABC transporter subfamily D, ABCD1‒3, are involved in the transport of fatty acyl-CoAs from the cytosol to the peroxisomal lumen. Furthermore, it is recently reported that some bacterial ABC transporters, which possess the same transmembrane folding with mammalian ABC exporters, function as an importer (32, 33). Elucidation of the mechanism underlying the direction of substrate transport is a subject for future investigation.</p><!><p>K. phaffii strain SMD1168 his4 was used as the host strain to express the human ABCD4 and LMBD1. The strains were grown in 1% yeast extract, 2% peptone, and 2% glucose or buffered minimal (0.5% yeast extract and 0.5% methanol) medium.</p><!><p>The expressed proteins were purified according to the procedure previously described (19) with some modifications. Yeast cells expressing His-ABCD4 were grown to midlog phase on 1% yeast extract, 2% peptone, and 2% glucose medium. Subsequently, the cells were transferred to buffered minimal medium and incubated for 18 h at 30 °C. The cells were resuspended with Tris buffer (50 mM Tris–HCl at pH 7.5, 300 mM NaCl, and 5 mM DTT) and then disrupted with 0.3-mm zirconia beads in a Multi-Beads Shocker (YASUI KIKAI Co, Ltd). All purification steps were conducted at 4 °C. Undisrupted cells, nuclei, and other cell debris were removed by centrifugation at 1500g for 10 min. The resulting supernatant (cell-free extract) was subjected to centrifugation at 14,000g for 30 min to obtain an organelle pellet. Membranes were solubilized by 0.5% β-DDM for 3 h on an end-over-end rotator. Insoluble material was removed by centrifugation at 100,000g for 30 min. The supernatant was incubated with cOmplete His-Tag Purification Resin (Roche) with 5 mM imidazole on an end-over-end rotator for 16 h. Subsequently, the resin was washed twice with Tris buffer containing 0.1% β-DDM and 50 mM imidazole, and His-ABCD4 was eluted with Tris buffer containing 0.1% β-DDM and 500 mM imidazole.</p><p>To purify LMBD1–GST, solubilization of LMBD1–GST was performed by the same procedure as used for His-ABCD4. The supernatant after β-DDM treatment was incubated with COSMOGEL GST-Accept agarose (Nacalai Tesque, Inc). The resin was washed with Tris buffer and eluted with Tris buffer containing 50 mM glutathione.</p><p>Protein concentration in eluate fraction was determined by the method of Bradford. As His-ABCD4 and the nonspecific protein existed in the eluate fraction, the amount of His-ABCD4 was calculated from the ratio of the intensity of His-ABCD4 and the nonspecific protein in acrylamide gel after SDS-PAGE and Coomassie brilliant blue staining.</p><!><p>Liposomes were prepared as previously described (34). Soybean l-α-phosphatidylcholine (10 mg/ml, Type II-S; Sigma) was suspended in the suspension buffer (20 mM Tris–HCl at pH 7.5). The mixture was sonicated until clear in a bath-type sonicator and then frozen and thawed five times. Liposomes were stored at −80 °C until use. Aliquot of 50 μg of protein in eluate fraction was mixed with 500 μg of liposomes in a 1:10 ratio (W/W) and then frozen and thawed twice. The mixture was diluted 30-fold with the reconstitution buffer (20 mM Tris–HCl at pH 7.5 and 0.5 mM DTT). Reconstituted proteoliposomes were pelleted via centrifugation at 150,000g for 1 h at 4 °C and then suspended in 200 μl of 20 mM Tris–HCl at pH 7.5. To prepare proteoliposomes containing both ABCD4 and LMBD1, protein solutions of His-ABCD4 and LMBD1–GST were incubated for 1 h at 4 °C before being mixed with the liposomes. To prepare proteoliposomes containing cobalamin, we added cobalamin to both the suspension and the reconstitution buffer. The amount of ABCD4 incorporated into liposomes was calculated by immunoblot analysis of His-ABCD4 using His-ABCD4 in eluate fraction as standard. The signal strength of His-ABCD4 was quantified by the ImageJ image analysis software (NIH).</p><!><p>The ATPase activity of the reconstituted proteins in the liposomes was determined by measuring the release of inorganic phosphate from ATP, as previously described (35). Proteoliposomes were mixed with reaction buffer (50 mM Tris–HCl at pH 7.6, 300 mM NaCl, 11 mM MgCl2, 1.1 mM EGTA, 2.2 mM DTT, 10 mM sodium azide, and 2 mM ouabain). The reaction was started by the addition of ATP (final concentration of 5 mM) and incubated at 37 °C. This reaction was stopped by the addition of 1 N HCl and supplemented with malachite green solution (0.034% malachite green, 1.05% ammonium molybdate, and 0.1% TritonX-100). Absorption at 620 nm was measured after incubation at 30 °C for 30 min.</p><!><p>ABCD4-liposomes were incubated with SDA (Thermo Fisher Scientific) (final concentration of 2 mM) at room temperature for 30 min. The reaction was stopped by the addition of Tris–HCl at pH 8.8 at a final concentration of 200 mM and then incubated on ice for 15 min. SDA-labeled proteins were photoactivated by UV irradiation at 365 nm for 1 min. Samples were subjected to immunoblot analysis. To evaluate binding stoichiometry between ABCD4 and LMBD1, m-maleimidobenzoyl-N-hydroxysuccinimide ester (Thermo Fisher Scientific) (final concentration of 2 mM) was used in addition to SDA.</p><!><p>Proteoliposomes containing 100 μM of cobalamin were incubated with the reaction buffer (50 mM Tris–HCl at pH 7.6, 100 mM KCl, 20 mM MgCl2, 4 mM DTT, and 2 mM ATP) at 37 °C. Aliquots were removed at various time points and applied to a Sephadex G-50 column (GE Healthcare) to collect the proteoliposomes. The proteoliposomes were then disrupted with 1% Triton X-100 and applied on a COSMOSIL 5C18-MS-II column (4.6 × 25 cm; Nacalai Tesque, Inc) at a flow rate of 1 ml/min equilibrated with 20 mM ammonium acetate buffer. Elution was performed with a linear gradient of an increasing acetonitrile concentration (0‒5 min, 5%; 2‒25 min, 5‒40%; and 25‒30 min, 40%) in 20 mM ammonium acetate. Cobalamin was detected by UV 361 nm. The amount of cobalamin transported from inside of the liposomes at each time point was calculated by subtraction of the amount of cobalamin in the liposomes at each incubation from that without any incubation.</p><!><p>The eluates of each protein (15 μg each) were mixed and rotated at 4 °C for 15 min. The mixture was then incubated with cOmplete His-tagged Resin or COSMOGEL GST-Accept agarose overnight at 4 °C. The resin or agarose was collected by centrifugation and washed twice with wash buffer (50 mM Tris–HCl at pH 7.5, 300 mM NaCl, and 0.1% β-DDM). The bound proteins were eluted in SDS contained in the sample buffer and subjected to immunoblot analysis.</p><!><p>All data are given in the main article or supporting information.</p><!><p>This article contains supporting information (11, 12, 19, 36).</p><!><p>The authors declare that they have no conflicts of interest with the contents of this article.</p><!><p>Supplemental Figures S1–S13</p>

PubMed Open Access

Towards visible light driven photoelectrocatalysis for water treatment: Application of a FTO/BiVO4/Ag2S heterojunction anode for the removal of emerging pharmaceutical pollutants

The resulting BiVO 4 /FTO electrode was thoroughly washed with deionized water and dried at room temperature. In order to obtain BiVO 4 / Ag 2 S electrode, the prepared FTO/BiVO 4 was dipped in a 0.3 M AgNO 3 solution for 10 s and followed by immersion in 0.3 M Na 2 S for 10 s. The cycle was repeated ten times and the obtained electrode was rinsed with deionized water and air dried at room temperature for 24 h.

towards_visible_light_driven_photoelectrocatalysis_for_water_treatment:_application_of_a_fto/bivo4/a

<!>Results and discussion<!>conclusion

<p>Water pollution is a global challenge with a lot of negative environmental and health implications. Pollution has the likelihood of increasing owing to increase in industrial activities, improper discharge of household effluents, inefficient wastewater treatment of polluted water and so on. Recalcitrant organic compounds such as pharmaceuticals constitute a major class of emerging water pollutants. Pharmaceuticals, particularly antibiotics have been reportedly found in wastewater and groundwater 1 . When such antibiotics are present in water, they pose serious danger to aquatic organisms and continuous consumption of such water by human can also result in chronic health issues such as the development of strains of bacteria that are resistant to antibiotics 2 . Over the past decades much attention has been directed to developing water treatment methods that are efficient and environmentally friendly since methods based on conventional wastewater treatments often lead to secondary pollutions and incomplete removal of target pollutants in water 3,4 . A recent approach for removing these recalcitrant organics from wastewater is photoelectrocatalytic (PEC) oxidation using suitable semiconducting materials as photoanodes 5,6 .</p><p>Photoelectrocatalytic oxidation is an environmentally friendly approach that uses both photon and electric energy to generate powerful oxidants (such as hydroxyl radical) that attack and destroy organic molecules that are present in aqueous solution. When this technique is used to treat water contaminated with recalcitrant organic molecules such as pharmaceuticals, total mineralization to water and carbon dioxide can be achieve over a period of time or the molecules can be broken down to non-toxic organic molecules within a short period of time 7 . Another interesting advantage of this approach is that the use of bias potential results in significant reduction in the challenge of rapid and spontaneous recombination of charge carriers that is peculiar to photocatalysis 8 . Titanium dioxide (TiO 2 ) and zinc oxide (ZnO) remained the most applied semiconducting photocatalyst as anodic material for photoelectrocatalytic degradation of organics [9][10][11] . Owing to the wide band gaps of TiO 2 (3.2 eV) and ZnO (3.5 eV), they perform best with the application of UV light but the UV region accounts for less than 5% of the solar spectrum 12 . Therefore, other sources of UV light which are expensive are often needed when using TiO 2 and ZnO. In order to cut down the cost associated with the operation of photoelectrocatalytic degradation process, solar light has been considered as a source of photon energy but this requires that the anodic material be made up of visible light active photocatalyst. In this line, semiconductor photocatalysts such as WO 3 13 , CuI 14 , Ag 3 VO 4 15 , Cu 2 O 16 , BiVO 4 17,18 , Fe 2 O 3 19 , CuS 20 , Ag 3 PO 4 21 , WS 2 22 and C 3 N 4 23,24 have been studied for PEC processes.</p><p>In the large pool of visible light active semiconductors, monoclinic sheelite bismuth vanadate (m-BiVO 4 ) has proven to be a choice material for PEC applications. As an n-type semiconductor, with narrow band gap (2.4 eV), BiVO 4 possesses impressive photocatalytic activity under the application of solar light, it is is non-toxic and has good stability. m-BiVO 4 has been employed as anodic material for the degradation of pollutants in wastewater 6 . It has also found application in PEC water splitting for hydrogen evolution 25 . Unfortunately, the use of unmodified BiVO 4 is faced with the problem of poor transport of charge carriers as well as relatively fast recombination of photo-excited charge carriers. Over the years, researchers have employed several techniques to counter this problem which include doping with metallic and/or non-metallic impurities, preparation nanosized BiVO 4 with well-defined morphology, loading of catalyst and formation of heterojunctions with other semiconductors 26,27 . Among these approaches, the formation of heterojunction with other semiconductors have proven to be the most effective.</p><p>Basically, heterojunction is formed when two semiconductors of unequal band gap combined in such a way that it results in band alignment 28 . It has been observed that the formation of heterojunctions between p-type and n-type semiconductor can improve PEC activity through improved light harvesting, effective separation of photogenerated electron-hole pairs and thus increased the lifespan of the charge carriers. For instance, Soltani et al. 29 , prepared BiFeO 3 /BiVO 4 with p-n heterojunction through facile ultrasonic/hydrothermal route and they observed improved charge separation in the composite as shown in the current density of 0.23 mA/cm −2 achieved on BiFeO 3 /BiVO 4 which was three times higher than that of pristine BiVO 4 . Additionally, higher percentage degradation of tetracycline was reported with the application of the prepared BiFeO . The selection of an appropriate p-type semiconductor is a critical step to achieve p-n heterojunction of BiVO 4 with improved performance. Recently, attention has been given to silver sulfide (Ag 2 S) as a suitable semiconductor to form p-n heterojunction with BiVO 4 . As a chalcogenide based p-type semiconductor, Ag 2 S has good optical properties and photocatalytic activity owing to its small band gap (between 0.9-1.1 eV) 40 . BiVO 4 /Ag 2 S p-n heterostructure prepared through hydrothermal routes have shown enhanced photocatalytic performance for the degradation of dyes and pharmaceuticals 41 . Guan et al. 42 , have also demonstrated the photoelectrochemical performance of BiVO 4 /Ag 2 S in water splitting and achieved a high photocurrent density of 1.91 mA/cm 2 . To the best of our knowledge, the performance of BiVO 4 /Ag 2 S in photoelectrochemical oxidation of pharmaceuticals in aqueous solution have not been reported.</p><p>Herein, we report for the first time the photoelectrocatalytic degradation of pharmaceuticals in water using a p-n heterostructure of BiVO 4 /Ag 2 S prepared on FTO glass as anode. The BiVO 4 /Ag 2 S photoanode with improved PEC performance was prepared on FTO glass using two-step electrodeposition and successive ionic layer adsorption/reaction (SILAR) methods. The optical property of the electrode was studied using UV diffusive reflectance spectroscopy (UV-DRS) while structural and morphological studies were carried out with XRD, SEM and EDS. Chronoamperometry and linear sweep voltammetry were used to confirm improved photocurrent response of the material. Electrochemical impedance spectroscopy and Mott Schottky plot were also used to establish the formation of heterojunction between the two semiconductors. Ciprofloxacin and sulfamethoxazole were selected as pollutant of interest for the photelectrocatalytic degradation experiments. experimental Materials and reagents. All the chemicals used were purchased from Sigma Aldrich (South Africa. These include bismuth nitrate pentahydrate (Bi(NO 3 ) 3 . 5 H 2 O), potassium iodide, vanadylacetylacetonate, silver nitrate, sodium sulfide, sodium hydroxide pellets, p-benzoquinone, sodium sulfate, potassium hexacyanoferrate (II), potassium hexacyanoferrate (III), ciprofloxacin and sulfamethoxazole.</p><p>preparation of BiVo 4 /Ag 2 S photoanodes. The binary photoanode with p-n heterojunction was fabricated through a two-step electrodeposition and Successive Ion Layer Adsorption/Reaction methods. First, BiVO 4 were electrodeposited on a FTO glass (5 cm × 1.3 cm × 0.22 cm, surface resistivity of ~7 Ω/sq) using a modified previously documented electrodeposition technique 34,43 . Summarily, from a well sonicated precursor solution containing 0.49 g Bi(NO 3 ) 3 •5H 2 O, 1.66 g KI in 25 mL and 0.23 M p-benzoquinone maintained at pH 4.3, films of BiOI were first potentiostatically electrodeposited onto a clean FTO glass at −0.13 V for 720 s. FTO glass, platinum wire and Ag/AgCl (3.0 M KCl) electrode were employed as the working electrode, counter electrode and reference electrode respectively. After rinsing the obtained BiOI electrode with water several times and drying at room temperature, 100 μL of 0.20 M vanadylacetylacetonate (dissolved in DMSO) was drop-cast evenly onto Structural and morphology characterisation of the prepared electrodes. X-ray diffractometer (Rigaku Ultima IV, Japan) using Cu Kα radiation (k = 0.15406) with K-beta filter at 30 mA and 40 kV was used to identify the phase, degree of crystallinity and purity of the prepared semiconductor photoanodes. TESCAN Vega 3(Czech Republic) scanning electron microscope was employed to determine surface morphology of the material. Energy-dispersive spectrometer (EDS) attached to the SEM instrument was used to confirm the presence of expected elements in the prepared materials in the appropriate ratio. The light absorption properties of the materials were analyzed using UV/Visible-Diffuse Reflectance Spectroscopy. A similar characterisation methods has been previously reported 13 .</p><p>electrochemical and photoelectrochemical experiments. The electrochemical and photoelectrochemical experimental protocols are similar to that described in our previous reports 8,43 . Photocurrent measurements, linear sweep voltammetry (LSV) and electrochemical impedance spectroscopy (EIS) were performed on an Autolab PGSTAT204 (Netherlands) potentiostat/galvanostat. The working electrodes were the prepared BiVO 4 , Ag 2 S and BiVO 4 /Ag 2 S electrodes. Platinum sheet with equal dimension as the FTO glass was employed as counter electrode while the reference electrode was Ag/AgCl (3.0 M KCl). Chronoamperometry and LSV were carried out in a 0.1 M Na 2 SO 4 solution. EIS was done in a 5 mM solution of [Fe(CN) 6 ] 3−/4− (prepared in a 0.1 M KCl solution). Data for Mott Schottky plots were obtained under dark condition at room temperature. For photoelectrochemical experiments, a solar simulator equipped with a 100 W xenon lamp was used as the light source. The prepared electrode was fixed vertically facing the incident light of the simulator and the distance between the photoelectrochemical cell and the light source was 10 cm and the glass were illuminated from the rear. The experiments were performed in a 70 mL capacity reactor made of quartz glass. For the degradation experiments, the working solution was 50 mL solution of 0.1 M Na 2 SO 4 (supporting electrolyte) and 10 mgL −1 of the pharmaceuticals. Aliquots of the solution taken from the reactor at predefined time intervals using a disposable syringe were analyzed using UV-Visible spectrophotometer to obtain the concentration decay pattern. The total organic carbon was also measured using TOC analyser (Teledyne Tekmar TOC fusion). The applied bias potential was optimized by performing the PEC experiments at different bias potential.</p><!><p>Structural and morphology characterization of the electrodes. The X-ray diffractograms of the prepared photoanodes are presented in Fig. 1. All the peaks correspond to those of monoclinic scheelite BiVO 4 (JCPDS no. 75-1866) in the XRD pattern of the BiVO 4 . The main peaks at 18.8°, 28.73°, 30.64°, 34.01°, 35.06°, 39.95° and 42.35°can be indexed as (110, 011), (121), (040), ( 200), (002), ( 211) and (150) crystal planes respectively 44 . In the XRD pattern of BiVO 4 /Ag 2 S, the diffraction peaks of Ag 2 S are not pronounced and well visible which could suggest that the particles are well dispersed on the surface of the BiVO 4 or probably due to relatively lower content and low crystallinity of Ag 2 S loading 45 . Nonetheless, the presence of Ag 2 S in the sample is evident in the peaks at 32° and 35° which appeared to be superimposed on those of BiVO 4 and therefore changing their intensities 46 . Expectedly, all the characteristic peaks BiVO 4 were still observed in the XRD pattern of the BiVO 4 / Ag 2 S electrode. In order to further confirm the presence of Ag 2 S on the binary electrode, other morphological studies were carried out.</p><p>The surface morphology of the photoanodes prepared on FTO glass are shown in Fig. 2(a-c). The prepared Ag 2 S appears as fine particles dispersed on the FTO glass (2a) while the particles of BiVO 4 are agglomerated on the FTO glass forming film like structure (2b). The incorporation of Ag 2 S particles onto the BiVO 4 on FTO glass www.nature.com/scientificreports www.nature.com/scientificreports/ resulted in agglomerated globules with openings which could serve as active sites for capturing of target analytes as shown the SEM image of FTO-BiVO 4 /Ag 2 S (Fig. 2c) and this further established that Ag 2 S were successfully coupled with the electrodeposited BiVO 4 . HR-TEM was further used to evaluate the nanostructure and heterostructure interface property between BiVO 4 and Ag 2 S in the BiVO 4 /Ag 2 S composite and the from the results it can be seen that BiVO 4 materials appeared as nanorods of different sizes (Fig. 2d) while Ag 2 S were nanoparticles which were well dispersed on the surface of BiVO 4 nanorods (Fig. 2e) suggesting the successful formation of appropriate heterostructure interface. As shown in Fig. 2f, the EDS spectrum revealed that only Bi, V, O, Ag and S were present in the composite electrode suggesting that the material is reasonably pure since no unwanted element was observed in the spectrum. Additionally, the percentage composition of each element obtained from the result also agreed with theoretical calculation of the elemental composition of BiVO 4 /Ag 2 S composite. It is also interesting to note that distribution of the elements on the electrode surface is uniform as revealed in the EDS mapping (Fig. 2g) and this further confirmed the evenly spread of Ag 2 S particles on the electrodeposited BiVO 4 . 3a. All the electrodes absorb photons in the visible light region and the absorption edges can be traced to 540 nm and 620 nm for BiVO 4 and BiVO 4 /Ag 2 S respectively while the absorption edge of Ag 2 S tends towards the near infrared. The shift of the absorption edge of BiVO 4 /Ag 2 S and increase in absorption can rightly be attributed to the enhancement of BiVO 4 optical ability through the addition of Ag 2 S. In order determine the band gap energy of the two semiconductors, the data obtained from the UV-DRS analyses were fit into Tauc equation which established that the band gap energy of a semiconductor can be determine using Eq. ( 1).</p><p>where α, h, A, E g and v are the absorption coefficient, Planck's constant, constant, band gap energy and incident light frequency respectively; 'n' is a constant that depends solely on the optical transition characteristics of the semiconductors under consideration. For direct transition semiconductors the value of 'n' is 1 47 . Since both BiVO 4 and Ag 2 S are direct semiconductors, a plot of (αhv) 2 against hv was made from which the value of E g for BiVO 4 and Ag 2 S were estimated to be 2.36 eV and 0.97 eV respectively (Fig. 3b). The values for the band gap energies obtained for both BiVO 4 and Ag 2 S are in agreement with previously reported values for the semiconductors 48,49 . These results further confirm the successful preparation of visible light active semiconductor photocatalysts. The improved photoabsorption in BiVO 4 /Ag 2 S suggested enhanced charge separation through band alignment when BiVO 4 and Ag 2 S combined together and this was further established through series of photoelectrochemical experiments.</p><p>electrochemical and photoelectrochemical analysis. Linear sweep voltammetry (LSV) of the photoanodes were carried out in a solution of 0.1 M Na 2 SO 4 (pH 7) at a scan rate of 20 mVs −1 . The linear voltammograms (Fig. 4a) were recorded in both the presence and absence of visible light illumination. All the electrodes showed higher current responses with illumination than without illumination which could be attributed to that fact that when the materials were irradiated, there is instantaneous excitation of electrons from the valence band to the conduction band and this enhanced better conductivity. Ag 2 S shows improved responses at 0.28 V and 0.85 V while BiVO 4 show a continuous increase in photocurrent response with increase in potential. Interestingly, though the binary electrode of BiVO 4 /Ag 2 S showed the characteristics features of both the voltammogram obtained with Ag 2 S and BiVO 4 (Fig. 4a) its overall increase in photocurrent response with increase in potential was higher than both the pristine electrode suggesting a good improvement in charge separation resulting in higher light responsiveness through the construction heterointerface. The anodic peak at ca 250 mV (Fig. 4a) is due to the oxidation of silver. This peak occurs only in Ag containing electrodes and thus a further confirmation of the presence of Ag 2 S in the heterojunction electrode BiVO 4 /Ag 2 S. It has been also been established that there is a linear correlation between the transient photocurrent of a semiconducting material and the charge separation process taking place within the material 13 . Therefore, with applied www.nature.com/scientificreports www.nature.com/scientificreports/ external potential of +0.8 V (selected based on the LSV performance of the electrodes), the transient photocurrent responses of the electrodes were recorded using chronoamperometry method (Fig. 4b). As expected, the highest photocurrent (1.194 mA/cm −2 ) was attained with BiVO 4 /Ag 2 S electrode which was significantly higher than that of pristine BiVO 4 (0.802 mA/cm −2 ) and almost ten times greater than that of Ag 2 S (0.165 mA/cm −2 ). Therefore, it was clear that the construction of p-n heterojunction between BiVO 4 and Ag 2 S promotes charge transfer between the interfaces of the two semiconductors which greatly inhibit the rapid recombination of photogenerated electron -hole pairs in the BiVO 4 /Ag 2 S electrode.</p><p>The results obtained with electrochemical impedance spectroscopy further corroborate the improved performance of BiVO 4 /Ag 2 S heterojunction through synergistic effect of both BiVO 4 and Ag 2 S. The experiments were performed in an electrolytic solution of 5 mM [Fe(CN) 6 ] 3−/4− in 0.1 M KCl (pH 7) with external application of +0.2 V. The obtained Nyquist plots for the fabricated photoanodes are displayed in Fig. 4c. For all the electrodes single characteristic semicircles were obtained in the EIS spectra which signifies the charge transfer process happening at the solution-electrode interface. The size of the semi-circular arc in the spectra is a function of the charge-transfer resistance (R ct ) at the interface of the heterojunction and studies have shown that the smaller the arc radius the better the charge transfer efficiency 50,51 . Accordingly, the lowest R ct was obtained from the Nyquist plot of BiVO 4 /Ag 2 S. This further affirmed that the formation of heterojunction between BiVO 4 and Ag 2 S resulted in better charge mobility and lowered rate of instantaneous recombination of photogenerated electron-hole pairs. Furthermore, the impedance data were analyzed with a bode phase angle plot (Fig. 4d) d to determine the electrons lifetime and charge transfer resistance in the BiVO 4 /Ag 2 S electrode. As shown in Fig. 4d, the maximum phase angle of the heterojunction electrode shifts to lowest frequency as compared to both Ag 2 S and BiVO 4 . This confirms the rapid electron transport process happening in the heterojunction. The life time of electrons is related to the frequency as given in Eq. ( 2) 52 .</p><p>max where f max is the frequency at the maximum phase angle the bode plot. Using Eq. ( 2), the electron lifetime of BiVO 4 /Ag 2 S was calculated to be 0.40 ms which was longer than those of BiVO 4 (0.32 ms) and Ag 2 S (0.31 ms). This value of life time calculated further confirms that the fabrication of the heterojunction helped in minimizing rapid recombination of electron -hole pairs in the two semiconductors through a fast charge transfer process 53 .</p><p>The flat band potential (E FB ) and charge carrier density (N D ) of a semiconductor can also be used as a measure of improved charge separation in semiconductor -semiconductor heterojunction interfaces = These values can be obtained from potential scan measurements and fitting of data to obtain Mott Schottky plot. Mott Schottky equation is given in Eq. 3.</p><p>C, e, ɛ, ɛ 0 , E app , T, N D , E FB and k represent the capacitance at the semiconductor/electrolyte interface (Fcm −2 ), elementary charge (1.60 × 10 −19 C), dielectric constant (68 for BiVO 4 54 ), permittivity of vacuum, external applied potential, absolute temperature, donor density, flat band potential and Boltzmann constant respectively.</p><p>From Eq. (3), a plot of 1/C 2 against E app was constructed and Donor density (N D ) was calculated from the slope while the approximately value of flat band potential was extrapolated from the intercept (Fig. 4e). As an n-type semiconductor, the MS plot of BiVO 4 gave a positive slope value. A negative shift in the flat band potential from −0.512 V in BiVO 4 to −0.548 V in BiVO 4 /Ag 2 S was also observed and this suggested that the rate of rapid recombination of charge carriers in the constructed BiVO 4 /Ag 2 S heterojunction was greatly reduced. This observation was further justified the carrier density of BiVO 4 /Ag 2 S (3.87 × 10 22 cm −3 ) which was significantly larger than that of pristine BiVO 4 (8.07 × 10 21 cm −3 ).</p><p>photoelectrocatalytic degradation of pollutants. The photoelectrocatalytic degradation of organics on the prepared BiVO 4 /Ag 2 S electrode was evaluated by using ciprofloxacin and sulfamethoxazole as target water contaminants. The degradation was achieved with an applied bias potential of 1.2 V, pH 7 and simulated sunlight was used as light source. The degradation processes of the pharmaceuticals were followed using UV-Visible spectrophotometer and evidence of reduction in the concentrations of ciprofloxacin and sulfamethoxazole was seen by the decrease in the intensity of the peaks at 276 nm and 265 nm for ciprofloxacin and sulfamethoxazole respectively. Within 120 min, a percentage removal of 80% and 86% for ciprofloxacin and sulfamethoxazole was achieved (Fig. 5a,b). The breaking down of ciprofloxacin molecules was further confirmed through the percentage total organic carbon removal (TOC) which was 69%. In the absence of light, the percentage anodic electrochemical degradation of ciprofloxacin and sulfamethoxazole were 59% and 61% respectively while percentage degradation achieved with photocatalysis alone was 35% and 40% respectively. The highest degradation achieved with photoelectrocatalytic degradation showed that the application of bias potential in conjunction with photocatalysis facilitated the breaking down of the organic molecules as the bias potential helps in driving away photoexcited electrons from the surface of the photoanode and thereby reducing the occurrence of recombination of the electron -hole pairs.</p><p>The kinetics studies also revealed that the degradation processes of both ciprofloxacin and sulfamethoxazole were fastest with the application of photoelectrocatalytic oxidation (Figs. S1 and S2). It is also interesting to note that the degradation of sulfamethoxazole on the BiVO 4 /Ag 2 S electrode appeared to be more favourable than that of ciprofloxacin as evident with the higher percentage degradation (86%) and supported with the apparent rate constant (0.0147 min −1 ) which was higher than that of ciprofloxacin (80% and 0.0137 min −1 ). This could be attributed partly to the larger molecular mass and more complex structure of ciprofloxacin. www.nature.com/scientificreports www.nature.com/scientificreports/ role in separation electron -hole pair by driving the photogenerated electrons away from the anode towards the cathode. As shown in the result, the higher the applied potential, the higher the driving force for the electron. Therefore, the degradation increases because there was reduced recombination of photoinduced electron -hole pairs at higher potential. When higher potential of 1.5 V was applied, the difference in the percentage degradation with that obtained with 1.2 V was approximately 1% which was relatively insignificant when compared with the trend. This revealed that beyond the optimal potential, higher applied bias potential could yield insignificant improvement in the percentage degradation which could be due to side reaction of evolved oxygen at higher potential 55 . Based on the result obtained, 1.2 V was selected as the optimal bias potential for the photoelectrocatalytic degradation of pharmaceuticals on the BiVO 4 /Ag 2 S electrode.</p><p>The improved charge separation in the binary electrode through the construction of p-n heterojunction was also confirmed by comparing its performance in the photoelectrocatalytic degradation of pharmaceuticals with that of the pristine electrodes of BiVO 4 and Ag 2 S. As shown in Fig. 5d, BiVO 4 and Ag 2 S electrodes gave a percentage removal of 63% and 50% respectively which were lower than 80% ciprofloxacin removal achieved on the BiVO 4 /Ag 2 S photoanode. The better performance of the binary electrode suggested that p-n heterojunction constructed facilitated the migration of electrons from the conduction band of Ag 2 S to BiVO 4 while holes from BiVO 4 moves to the valence band of Ag 2 S yielding enhanced photogenerated charge carriers separation resulting in better photoelectrocatalytic performance 28 .</p><p>One of the advantages of photoelectrocatalytic degradation over convention photocatalysis is the ease of reusability of the material. The BiVO 4 /Ag 2 S electrode also showed impressive stability and reusability as seen from the cycling experiments (Fig. 5e). The results were obtained by using the same BiVO 4 /Ag 2 S three different times. After each cycle, the electrode was purged with deionized water and air dried at room temperature. After the third application the percentage removal of ciprofloxacin was approximately 79% suggesting that there was no remarkable change in the performance of the electrode after using it three times showing that the electrode is relatively stable and can be reused.</p><p>proposed mechanism of degradation and scavenger studies. The degradation of organic molecules during photoelectrocatalytic processes happens when generated reactive species attack and oxidize the organic molecules. The photogenerated holes, hydroxyl and superoxide radicals usually play the predominant roles in photoelectrocatalytic degradation experiments. Equations 4-9 give the mechanism of formation of these reactive species and their oxidation reactions with the pharmaceutical molecules for total mineralization. The contribution of individual reactive specie in the PEC degradation of the ciprofloxacin molecule was determined by trapping experiments which was conducted by inhibiting the effects of holes, hydroxyl radicals and superoxide radicals through the introductions of ethylenediaminetetraacetate salt (EDTA), t-butanol (t-BuOH) and p-benzoquinone (p-BZQ) respectively 56,57 in the reaction medium. As seen in Fig. 6a, photogenerated holes play a crucial role in the breaking down of the pharmaceutical molecules since the percentage removal dropped to almost 10% when holes were masked through the addition of EDTA. The effect of hydroxyl radicals on the degradation of the pharmaceutical molecules cannot also be overlooked process because the degradation efficiency dropped to 52% with the addition of t-butanol. The hydroxyl radicals were produced in the reaction system through the oxidation reactions of water molecules by photogenerated holes. Unlike the holes and hydroxyl radicals, superoxide radicals performed a seemingly insignificant role in the oxidation of the pharmaceutical molecules because percentage removal of 76% was still achieved when the superoxide radicals were trapped by the addition of p-BZQ. Literatures have shown that hydroxyl radicals are not produced in detectable amount when BiVO 4 is illuminated due to rapid recombination because of rapid recombination with photogenerated electrons 48,58 . But in this work, the fact that holes and hydroxyl radicals play predominant roles the breaking down of the pharmaceuticals confirms that better charge separation can be achieved with BiVO 4 through the formation of heterojunction with Ag 2 S.</p><p>The possible mechanism of the spontaneous mobility of photogenerated electron-hole pairs between the interface the two semiconductors is proposed by obtaining the relative band edge potential of the conduction band and valence band of both semiconductors using Eqs. 9 and 10. E VB and E CB stand for the valence and conduction band edge potentials respectively. X represent the electronegativity of the semiconductor usually calculated as the geometric mean of the absolute electronegativities of the constituent atoms in the semiconductor (X = 6.04 for BiVO 4 and 4.96 for Ag 2 S). E C is the energy of the free electrons on hydrogen scale which is approximately 4.50 eV (vs NHE). Energy band gap (E g ) has been estimated to be 2.36 eV for BiVO 4 and 0.97 eV for Ag 2 S respectively using Tauc equation (Fig. 3b). Therefore, the E CB and E VB of BiVO 4 were calculated to be 0.360 eV and 2.720 eV respectively while for Ag 2 S, the values obtained were −0.025 eV and 0.945 eV for E CB and E VB respectively. The values of E CB and E VB for Ag 2 S are lower than the corresponding values for BiVO 4 indicating that the formation of type II heterojunction is possible when the two semiconductors aligned. As a p-type semiconductor, the Fermi energy level of Ag 2 S is located slightly above the valence band while that of BiVO 4 , n-type, is slightly below its conduction band. As shown in Fig. 6b, when the two semiconductors are in contact, the Fermi energy level aligned and internal electric field is established in such a way that the holes can be effectively separated into the valence band of Ag 2 S and be available to oxidize directly the pharmaceutical molecules or produce hydroxyl radicals from water molecules while the electrons migrate to the conduction band of BiVO 4 in opposite direction of the internal electric field resulting in efficient charge separation 28 . The electrons can also react with oxygen molecules to produced superoxide radicals 59 but the effect superoxide in the degradation process in this study is limited.</p><!><p>A photoanode of BiVO 4 /Ag 2 S with p-n heterojunction was successful prepared through electrodeposition and successive ionic layer adsorption/reaction method on FTO glass. The construction of p-n heterojunction reduced the common problem of rapid recombination of photogenerated electron-hole pairs. This was confirmed through the photocurrent response of BiVO 4 /Ag 2 S (1.194 mA/cm −2 ) which was higher than both pristine BiVO 4 (0.802 mA/cm −2 ) and Ag 2 S (0.165 mA/cm −2 ). When applied for the photoelectrocatalytic degradation of pharmaceuticals, the percentage removal of 80% and 86% were recorded for ciprofloxacin and sulfamethoxazole respectively. Overall, the reports from this research revealed that BiVO 4 /Ag 2 S electrode can be applied to oxidize recalcitrant pharmaceuticals in aqueous medium and pre-treated real pharmaceutical effluents and this can be achieved a reduced cost through the use of lower bias potential. In future works, enhancement of BiVO 4 /Ag 2 S performance for PEC water treatment will be studied using photoactive cathode.</p>

Scientific Reports - Nature

Single-stranded DNA as a Cleavable Linker for Bioorthogonal Click Chemistry-based Proteomics

In this report, we present a new class of cleavable linker based on automatically synthesized, single-stranded DNAs. We incorporated a DNA oligo into an azide-functionalized biotin (biotin-DNA-N3) and used the probe to enrich alkyne-tagged glycoproteins from mammalian cell lysates. Highly efficient and selective release of the captured proteins from streptavidin agarose resins was achieved using DNase treatment under very mild conditions. A total of 36 sialylated glycoproteins were identified from the lysates of HL60 cells, an acute human promyeloid leukemia cell line. These sialylated glycoproteins were involved in many different biological processes ranging from glycan biosynthesis to cell adhesion events.

single-stranded_dna_as_a_cleavable_linker_for_bioorthogonal_click_chemistry-based_proteomics

<p>Since the discovery of the Cu(I)-catalyzed azide-alkyne cycloaddition (CuAAC)1,2 and the strain-promoted copper-free click chemistry,3,4 bioorthogonal click chemical reactions have become indispensable tools for the identification of posttranslationally modified proteins and for activity-based protein profiling (ABPP).5 To identify proteins that are posttranslationally modified, azide or alkyne tags are introduced into the target protein pools either metabolically6,7,8 or via enzymatic approaches.9,10 In the second step, affinity probes, e.g. biotin, functionalized in a complementary fashion, are ligated, allowing for affinity capture, e.g. by immobilized streptavidin, and subsequent mass spectrometry analysis. This method has been successfully applied to the identification of proteins that are glycosylated,11 lipidated12 and methylated13 among many other modifications. In the scenario of ABPP, an activity-based probe functionalized with the azide or the alkyne is used to label target proteins covalently either in living animals or in crude cell lysates.14 The labeled proteins are then enriched and identified in an analogous manner as in the former case.15,16</p><p>Although a number of elegant strategies have been developed to introduce azide and alkyne tags into target proteins for bioorthogonal ligation with biotin probes and affinity capture with streptavidin agarose resins, relatively less effort has been devoted to the development of methods for the selective elution of the captured proteins from the resins. Due to the tight binding affinity of biotin to streptavidin (Kd∼10-15 M), the release of biotinylated species from streptavidin agarose resins requires harsh, denaturing conditions, i.e. boiling the resins at 100 °C in sodium dodecyl sulfate (SDS) buffer. Under these conditions, contaminated proteins binding to immobilized streptavidin through nonspecific, hydrophobic interactions and proteins that are endogenous biotinylated are co-eluted, which complicates the downstream mass-spectrometry analysis. To address this issue, new probes have been designed that incorporate a cleavable linker into the azide-or the alkyne-functionalized biotin in the format of an "azide/alkyne–cleavable linker–biotin". Applying these probes to the click chemistry-based proteomics, target proteins can be selectively released from the immobilized streptavidin by cleaving the linker. A few cleavable linkers that have been developed exploit their sensitivities toward protease-,17 pH-,18 redox18,19 and photo-mediated cleavage conditions18,20, respectively. However, most of these strategies are complicated either by stringent cleavage conditions or multistep chemical synthesis. For example, the reduction of azobenzene-based linkers with sodium dithionite is both pH- and air-sensitive, affording low cleavage efficiency under typical reaction conditions.18 Also, the construction of ortho-nitrobenzene-based photo sensitive probes requires nine linear step synthesis, and the nitrosoaldehyde byproduct formed upon photolysis is reactive toward polypeptides.21 Due to these complications, it can be a challenge to adopt these methods for routine proteomic analysis in biology-oriented laboratories.</p><p>Here we report a new type of cleavable linker based on automatically synthesized, single-stranded DNAs and its application to bioorthogonal chemical proteomics (Scheme 1). We incorporated a DNA oligo into an azide-functionalized biotin (biotin-DNA-N3) and used the probe to enrich for alkyne-tagged glycoproteins from mammalian cell lysates. Highly efficient and selective release of the captured proteins from streptavidin agarose resins was achieved following DNase treatment at 37 °C.</p><p>To integrate a single-stranded DNA into biotin-N3 to form a cleavable probe, we used biotin-modified controlled pore glasses (CPG) as the solid support to introduce nucleotide building blocks sequentially via a DNA oligonucleotide synthesizer (Scheme 2). The DNA oligo was designed to bear no predictable secondary structure and to have a length of 20 nucleotides, ensuring efficient cleavage by DNase when the biotin probe was captured on streptavidin agarose resins. The oligonucleotide was then capped by reacting with 5′-bromohexyl phosphoramidite. Substitution of the bromide with an azide was realized by reacting with NaN3. The resulting biotin-DNA-N3 was then eluted from the CPG solid support and purified with a NAP-10 column (Figure S1 in the Supporting Information).</p><p>With the cleavable biotin-DNA-N3 probe in hand, we first tested its compatibility with CuAAC coupling conditions. It is well known that reactive oxygen and nitrogen species generated from the canonical CuAAC system, i.e. CuSO4·5H2O + tris[(1-benzyl-1H-1,2,3-triazol-4-yl)methyl]amine (TBTA) + sodium ascorbate, could induce the breakdown of polynucleotides.22,23 Therefore, we adopted the new CuAAC catalyst system, i.e. CuSO4·5H2O + 3-(4-((bis((1-tert-butyl)-1H-1,2,3-triazol-4-yl)methyl)amino)methyl)-1H-1,2,3-triazol-1-yl)propan-1 ol (BTTP) ([BTTP]:[Cu]=4:1) + sodium ascorbate, recently developed in our laboratory.24 The new catalyst system showed excellent activity in protein labeling studies, especially in conjugate reactions with negatively charged species (e.g FLAG peptide–DYKDDDDK). Based on this precedent, we reacted the biotin-DNA-N3 probe with an alkyne-modified Alex Fluor 488 dye (488-alkyne) in an eight to one ratio in the presence of the BTTP-Cu(I) catalyst, mimicking the condition of labeling biomolecules with limited quantities. After one hour, the reaction mixtures were divided into three portions. Portion one was subjected to HPLC analysis, which showed near-quantitative dye-conjugation without breakdown of the DNA strand (Figure S2 in the Supporting Information). Subsequently, portion two was incubated with streptavidin agarose resins. Following centrifugation to precipitate the resins, the supernatant was transferred to a 96-well microtiter plate to measure the residual fluorescence of the uncaptured dye. As shown in Figure 1a, the reaction mixture that was treated with streptavidin agarose resins only showed ∼10% fluorescence intensity as compared to that of the control (portion three), suggesting that 90% of 488-alkyne was successfully reacted with biotin-DNA-N3 and captured by streptavidin agarose resins. Therefore, the CuAAC condition has no apparent adverse effects on the DNA linker.</p><p>We then analyzed the site-specific cleavage of the DNA linker by DNase. We captured Biotin-DNA-488, the reaction product of the biotin-DNA-N3 with 488-alkyne, on streptavidin agarose resins, and performed on-bead digestion with benzonase, a commercially available recombinant endonuclease. The cleaved product of biotin-DNA-488 was subjected to LC/MS analysis. Two major peaks (a1 and a2) were observed in the LC trace showing absorbance at 488 nm corresponding to the two regioisomers of the 488 dye (Figure 2a). Their masses were consistent with 488-5′GT-3′MS, suggesting that the majority of the cleaved product of biotin-DNA-488 by benzonase bears two nucleotides "GT" (Figure 2b and 2c). Since the digestion of DNAs by benzonase is not sequence-dependent, a small fraction of the cleaved product contains nucleotide fragments other than "GT", as shown in Figure 2a (a3 488-5′GTAA-3′ and a4 488-5′GTA-3′). Because digestion of DNAs by a restriction endonuclease would guarantee a higher specificity of cleavage, we then performed on-bead digestion of biotin-DNA-488 with the restriction endonuclease PvuII that recognizes the CAGCTG sequence in the DNA linker after annealing biotin-DNA-488 with its complementary DNA strand. LC/MS analysis showed the site-specific cleavage and the formation of two regioisomers of 488-5′GTAACGATCCAG-3′ (Figure 2d-f).</p><p>Next, we designed a proof-of-principle experiment to test the efficiency of the biotin-DNA-N3 probe to capture a single alkyne-tagged protein from a complex protein mixture. We introduced a single terminal alkyne group into cysteine34 of bovine serum albumin (BSA)25 and mixed it with the lysates of Jurkat cells, a human T lymphocyte cell line, in a 1:6 weight ratio as a model system. The protein mixture was reacted with the biotin-DNA-N3 probe in the presence of the BTTP/Cu(I) catalyst, followed by the enrichment with streptavidin agarose resins. The captured BSA was then released from resins by incubation with benzonase in Tris buffer (pH 7.5) for two hours at 37 °C. After three washes, the resins were further boiled in SDS buffer to release residual proteins that were resistant to the benzonase cleavage. In parallel, the same protein mixture was reacted with biotin-N3 and then incubated with streptavidin agarose resins. The captured BSA was released from resins by boiling in SDS buffer. The samples released from both experiments were collected and resolved by SDS polyacrylamide gel electrophoresis (SDS-PAGE). As shown in Figure 1b, comparable levels of BSA were detected in eluents obtained from benzonase cleavage and boiling-only treatment with significantly more contaminant proteins detected in the boiling-only eluents.26 Notably, no residual BSA was detectable in the fraction obtained from benzonase-followed-by-boiling double treatment, indicating the elution with benzonase was complete (Figure 1b and Figure S3 in the Supporting Information). Taken together, this experiment demonstrated that biotin-DNA-N3 is an efficient and selective probe to enrich alkyne-functionalized proteins from complex mixtures and the cleavage of this probe can be achieved under mild conditions.</p><p>After verifying the specificity and the compatibility of the biotin-DNA-N3 probe with the optimized CuAAC conditions, we then applied the probe to the enrichment of sialylated glycoproteins from HL60 cells for their identification. The cultured HL-60 cell is an acute human promyeloid leukemia cell line that has been widely used for the study of the molecular events of myeloid differentiation and the impact of pharmacologic and virologic agents on this process.27 Due to well-documented roles of the negatively charged sialic acid in cellular adhesion and the regulation of the growth and differentiation of hematopoietic progenitor cells, HL60 cells have also been used extensively for the study of sialylated glycoconjugates in these processes.28 To our surprise, however, the proteomic-wide profiling of sialylated glycoproteins in this cell line has never been pursued.</p><p>To profile the sialylated glycoproteins in HL60 cells, we cultured the cells in media supplemented with peracetylated N-(4-pentynoyl) mannosamine (Ac4ManNAl), an alkyne-tagged metabolic precursor of sialic acid (Figure 3a).7 The unnatural monosaccharide allows the metabolic incorporation of the alkyne tag into membrane sialylated glycoproteins (Figure S4 and S5 in the Supporting Informaton). After three days, cells were lyzed. Cell lysates were reacted with the biotin-DNA-N3 probe before incubation with streptavidin agarose resins for affinity enrichment (Scheme 1). To elute the captured glycoproteins, the resins were incubated with benzonase in Tris buffer (pH 7.5) at 37 °C for eight hours. The eluted proteins were resolved using SDS-PAGE. Following in-gel trypsin digestion, the resulting peptide fragments were analyzed using LC-MS/MS.</p><p>Data Analysis from two sets of repeated experiments revealed a total of 36 hits from lysates obtained from the cells treated with Ac4ManNAl but not those from untreated counterparts (Table S1 in the Supporting Information). Among these proteins, many are known to be sialylated, including integrin, intercellular adhesion molecule 3, leukosialin among a few others. We discovered that eight identified proteins are known to reside in the membrane of the endoplasmic reticulum or Golgi apparatus and are involved in the biosynthesis of N-linked glycans or in glycan metabolism (Table 1). Most other proteins are located in the plasma membrane; they either serve as cell-surface receptors or transporters or participate in metabolic processes (Figure 3b).</p><p>In summary, we discovered here that single-stranded DNAs serve as a new type of cleavable linkers for the CuAAC-based proteomic analysis. When integrated into biotin probes, linker cleavage and release of captured proteins was accomplished with high efficiency in very mild conditions. Using this new probe, we were able to identify 36 sialylated glycoproteins from the lysates of HL60 cells that are involved in many different biological processes ranging from glycan synthesis to adhesive events. By replacing benzonase with a restriction endonuclease, site-specific cleavage of the probe was realized, which is crucial for the integration this new technique into proteomics studies that require the identification of posttranslational-modification sites. Taking advantage of the automatic and commercially available synthesis of single-stranded DNAs, the integration of this cleavable linker into other affinity probes for proteomics, e.g. FLAG probes, would be a straightforward application.</p>

PubMed Author Manuscript

New paradigm for configurational entropy in glass-forming systems

We show that on cooling towards glass transition configurational entropy exhibits more significant changes than predicted by classic relation. A universal formula according to Kauzmann temperature T K is given: S = S 0 t n , where t = (T − T K )/T . The exponent n is hypothetically linked to dominated local symmetry. Such a behaviour is coupled to previtreous evolution of heat capacity �C config. P (T) = (nC/T)(1 − T K /T) n−1 associated with finite temperature singularity. These lead to generalised VFT relation, for which the basic equation is retrieved. For many glass-formers, basic VFT equation may have only an effective meaning. A universal-like reliability of the Stickel operator analysis for detecting dynamic crossover phenomenon is also questioned. Notably, distortionssensitive and derivative-based analysis focused on previtreous changes of configurational entropy and heat capacity for glycerol, ethanol and liquid crystal is applied.Glass transition has remained a grand cognitive challenge of solid-state physics, chemical physics and material engineering for decades 1,2 . The hallmark feature is Super-Arrhenius (SA) previtreous behaviour of such dynamic properties as the primary relaxation time τ (T) or viscosity η(T) 2,3 :

new_paradigm_for_configurational_entropy_in_glass-forming_systems

<!>OPEN<!>Results and discussion<!>Liquids Abr

<p>where T > T g , and E a (T) is the apparent activation energy. Basic Arrhenius behaviour is retrieved for E a (T) = E a = const in the given temperature domain. T g denotes glass temperature, which is empirically linked to τ T g = 100 s, and η T g = 10 13 P 4,5 .</p><p>General SA portrayal of previtreous dynamics described by Eq. ( 1) has a rational meaning and cannot be used to parameterize experimental data, due to unknown form of activation energy E a (T) 3 . Consequently, replacement relations must be applied. The dominant one is the Vogel-Fulcher-Tammann (VFT) dependence 2,6 : where T > T g , the amplitude A VFT = D T T 0 = const , D T is fragility strength coefficient, T 0 denotes extrapolated singular temperature T 0 < T g . The fragility m = dlog 10 τ (T)/d T g /T T=T g is the key metric of the SA dynamics, indicating a deviation from the Arrhenius behaviour related to m min . = log 10 τ T g − log 10 τ ∞ = 2 − log 10 τ ∞ . It is often estimated by the use of the fragility strength coefficient, namely: m = D T T 0 T g / T g − T 0 2 ln10 , and m = m(1 + ln10/D T ) min 2,4,5 . The enormous popularity of the VFT relation, illustrated in Fig. 1, causes that it is often indicated as an empirical 'universal' scaling pattern for previtreous dynamics. Consequently, its derivations are often treated as a checkpoint for glass transition models [6][7][8][9][10][11][12] .</p><p>The emergence of previtreous dynamics is associated with passing a melting temperature without crystallization and entering a metastable, supercooled domain 2,11,12 . In many 'predominantly' glass-forming systems, being of a particular interest of glass transition physics, supercooling is possible at any practical cooling rate, facilitating broadband dielectric spectroscopy (BDS) studies. In the previtreous domain, BDS requires frequency scans of electric impedance ranging from seconds to hours near T g . BDS studies deliver high-resolution estimations of primary (α, structural) relaxation time from loss curve peak frequency τ = 1/2πf peak . Previtreous changes of τ (T) are recognised as a basic characterization of previtreous SA dynamics [2][3][4][5]11,12 .</p><p>Configurational entropy ( S C ) is an essential thermodynamic characteristic of previtreous domain [2][3][4][5]8,9,[11][12][13][14][15][16][17][18][19][20][21][22] . It describes a non-equilibrium entropy excess, taking entropy of equilibrium crystalline state as a reference. In 1948 Walter Kauzmann indicated that for some extrapolated temperature, hidden in a solid amorphous glass state one should expect S C (T → T K ) → 0 , usually 20-50 K below T g 13 . The challenge associated with configurational entropy and the Kauzmann temperature T K explains the recent resume-report 20 : 'The configurational</p><!><p>Institute of High Pressure Physics of the Polish Academy of Sciences, Warsaw, Poland. * email: starzoneks@unipress. waw.pl entropy is one of the most important thermodynamic quantities characterizing supercooled liquids approaching the glass transition. Despite decades of experimental, theoretical, and computational investigation, a widely accepted definition of the configurational entropy is missing, its quantitative characterization remains fraught with difficulties, misconceptions, and paradoxes, …practical measurements necessarily require approximations that make its physical interpretation delicate… the Kauzmann transition remains a valid and useful hypothesis to interpret glass formation. We also insisted that this is still a hypothesis but in no way a proven or necessary fact…' . Following above, for an ultimate cognitive insight into glass transition phenomenon, crucial may be reliable experimental evidence for S C (T) behaviour, matched to clearly non-biased estimation of T K , and a nonambiguous link to dynamics.</p><p>On the other hand, Berther et al. 20 claimed, that: 'there is no, and that there cannot be any, unique definition of S c ′. However, based on author's as well as other researchers' best knowledge and experience, we decided to find a universality in configurational entropy behaviour. In the next part of the Report, we present a conventional definition of configurational entropy and its new critical-like description.</p><p>Experimentally, the configurational entropy may be estimated from an evolution of a heat capacity �C P (T) 2,12,15,16,20,21 : where �C P (T) = C SL P − C glass P = �C config. P</p><p>, with the heat capacity of glass instead of hardly detectable for 'predominant' glass formers, solid crystal entropy changes.</p><p>Assuming:</p><p>with C P = const , one obtains from Eq. (3) the 'classic' , dependence for the configurational entropy 2,15,16 :</p><p>where t = (T − T K )/T. It is commonly used for describing changes of the configurational entropy in previtreous domain and an estimation of T K 2,4,14-22 . One of the most inspiring models for glass transition was proposed by Adam and Gibbs (AG), five decades ago 8 . It links previtreous slowing-down to cooperatively rearranged regions (CRR), which influence configurational entropy, leading to following relation for previtreous changes of relaxation time 8 :</p><p>where A AG = const is the AG model amplitude.</p><p>(3) 57 , ethanol 58 , cyclooctanol 61 and 5 * CB liquid crystal 62 . Red and blue straight lines denote Eq. ( 5) and Eq. ( 10) respectively. Dashed arrows present glass transition temperatures T g . The insert shows configurational entropy as a function of reciprocal temperature S c (1/T) for all studied systems. (B) Configurational entropy normalised to the Kauzman temperature T K for all samples basing on generalised Eq. ( 10). Impact of different n parameter on Eq. ( 10) is shown as an insert. Limit values n = 0.1 and n = 2 as well as classical case for n = 1 are marked by bold lines. Fitting parameters may be found in Table 1 www.nature.com/scientificreports/ Substitution of Eq. ( 5) into Eq. ( 6) yields the VFT relation, if 8,12 . Numerous reports empirically support such a coincidence between a 'dynamic' and 'thermodynamic' singular temperatures for glass-forming systems 2,3,[7][8][9][10][11][12][19][20][21] . Such an agreement also constitutes an essential reference for a set of theoretical models which link a finite temperature singularity in dynamics to a 'hidden' phase transition 2,3,[7][8][9][10][11][12][19][20][21] . These empirical and theoretical correlations between 'thermodynamic' and 'dynamic' characterisations of previtreous domain, matched to enormous popularity of the VFT Eq. ( 2), significantly support Eq. ( 5) for describing configurational entropy and its usage as a tool for determining T K . However, there are blots and non-coherences on the above landscape. Equation ( 5) poorly reproduce a variety of observed patterns for the heat capacity for T → T g (see Fig. 2). As an empirical solution of this problem a relation �C conf . P (T) = �C P /T ϑ , with power exponent 0 < ϑ < 2 adjusted to a given glass former, was introduced 23 . However, it does not yield a coherent relation for configurational entropy and its model-basis is not clear. In 2003, Tanaka 24 carried out validation tests of the VFT equation for 52 glass-forming systems and showed that 0.8 < T 0 /T K < 2.2 , i.e., the correlation T 0 ≈ T K appears only for a limited number of glass formers. There is also growing evidence questioning the omnipotence and a fundamental reliability of the VFT relation. It bases mainly on a comparison between experimental data and their scaling via VFT and other model relations. Subsequently, using visual or analytic-residual assessment of fitting quality, the VFT or other relations' prevalence is tested. Nevertheless, observed discrepancies are subtle, occurring only in some temperature domains and they are close to an experimental error limit 2,11,12,[25][26][27][28][29] . Consequently, such tests cannot yield decisive conclusions. Another type of validation of scaling relations is based on a superposition of τ (T) or η(T) experimental data for a dozen glass-forming systems, using model-related parameters with individually selected (fitted) values for each tested system 2,11,12,[30][31][32][33] . In the authors' opinion, such a model-dependent scaling approach has tautological features and cannot lead to a breakthrough model-validation.</p><p>The recalled above record of puzzling results focused on confirming or rejecting the fundamental validity of the VFT relation had to be carried out for T > T g , i.e., 20-50 K above singular temperatures ( T K , T 0 ). However, remote from singular temperatures, only subtle discrepancies between experimental data and model relations may be expected. An experimental error notably amplifies such a problem. Relatively strong discrepancies between experimental data and scaling relations can be expected only near hypothetical singular temperatures, i.e., in experimentally non-accessible domain.</p><p>To address mentioned inherent features of previtreous domain, an analysis concentrated exclusively on subtle distortions between a hypothetical scaling relation and experimental data may be decisive. In Refs. [34][35][36] . linearised derivative-based analysis focused on a portrayal via VFT 5,6,29,34,37,38 , MYEGA 27,35,36 , Avramov-Milchev 36,39 and critical-like [40][41][42] scaling relations were developed. For instance, the VFT parameterisation may validate a linear domain appearing in a plot based on the following transformation of τ (T) experimental data 34 : Equation (7), in the form of the plot ϕ T = lnτ (T)/d(1/T) vs. 1/T , often named 'Stickel operator' analysis 43 , was used earlier for detecting a dynamic crossover temperature T B , i.e., the crossover between ergodic and non-ergodic previtreous dynamical domains. The appearance of two lines in such a plot and their intersection related to T B are indicated as a 'universal' feature of previtreous domain 43-46 . Novikov and Sokolov strengthen this 'universality' , suggesting a 'magic' time scale τ (T B ) = 10 −7±1 s, estimated empirically by the 'Stickel-operator' analysis of 30 glass-formers, including low-molecular-weight liquids, polymers, ionic systems, covalent systems and plastic crystals 47 . However, some criticism regarding this finding appeared, due to glass formers with strongly different τ (T B ) values 48 . Later, Roland showed a pressure-temperature invariance of τ (T B , P B ) 49 . It is worth nothing, that the linearised distortions-sensitive analysis showed that for glass-forming liquid crystals, plastic crystals and low-molecular-weight liquids with uniaxial molecules as well as a critical-like description are more reliable than the 'classic' VFT description 41,42 .</p><p>Hecksher et al. 50 proposed to analyse previtreous dynamics using activation energy index I DO (T) = −dlnE a (T)/dlnT = (dE a /E a )/(dT/T) , i.e., to transform experimental data τ (T) → I DO (T) . The required apparent activation energy was calculated using the general Super-Arrhenius Eq. ( 1), E a (T) = RTln(τ (T)/τ ∞ ) , assuming a 'universal' value for pre-exponential factor τ ∞ = 10 −14 s. In Ref. 50 the analysis for 42 low-molecular-weight glass formers led to the conclusion: '…there is no compelling evidence for the Vogel-Fulcher-Tammann (VFT) prediction that the relaxation time diverges at a finite temperature. We conclude that theories with a dynamic divergence of the VFT form lack a direct experimental basis. ' However, results from Ref. 50 might be biased by assuming a 'universal' value for the pre-factor, whereas experimental evidence suggests 10 −16 s < τ ∞ < 10 −10 s 34,39 . In Ref. 51 , apparent activation energy was determined using a protocol avoiding this problem. It is based on a numerical solution of a differential equation directly resulted from the Super-Arrhenius Eq. ( 1) and applied for a given set of τ (T) experimental data 51 :</p><p>The analysis of 26 glass-formers, including low-molecular-weight liquids, polymers, liquid crystals, colloids and even plastic crystals, revealed a common empirical pattern 51 : This result led to a general 'empirical' relation for the index 44,45 : 1/I DO (T) = nT 0 /(T − T 0 ) , where T 0 is singular temperature determined from the condition 1/I DO (T 0 ) = 0 and the parameter n = −1/a . It was found ( 7) www.nature.com/scientificreports/ that for tested systems 0.18 < n < 1.6 , and limits were related to domination of translational and orientational symmetries, respectively [51][52][53] . The previtreous dynamics described by the VFT relation is linked to n = 1 . Following mentioned results, a new relation for the configurational entropy was derived 51 :</p><p>The 'classic' Eq. ( 5) is retrieved for n = 1.</p><p>Problems of the VFT relation inspired the development of new scaling dependences for the previtreous dynamics. The leading position has gained Mauro-Yue-Ellison-Gupta-Allan (MYEGA) relation, which avoids the finite temperature singularity 27,35 :</p><p>Notably, it can be approximated by the VFT relation at 'high-temperature' domain 54 :</p><p>where K ≈ T 0 , and C ≈ D T T 0 , if comparing with VFT Eq. ( 2).</p><!><p>When discussing previtreous behaviour, one may consider substitution of Eq. ( 10) to the AG model relation Eq. ( 6). This yields a 'generalised' VFT relation:</p><p>where t = (T − T 0 )/T . The 'classic' VFT formula (Eq. ( 2)) is retrieved for n = 1.</p><p>Equation ( 13) has already been used for describing dynamics in glass-forming polyvinylidene difluoride (PVDF), PVDF + Barium-Strontium-Titanate (BST) microparticles composite 55 , and in its parallel form for describing relaxation time in relaxor ceramics 56 . Nevertheless, these tests cannot be considered as a crucial validation of Eq. ( 13) if recalling the above discussion. The milestone meaning could have derivative-based and distortions-sensitive tests focused directly on S C (T) experimental data. To fill such a cognitive gap a new solution is proposed in given report.</p><p>The analysis presented below explores state-of-the-art experimental results for the configurational entropy for 8 glass-forming liquids: glycerol 57 , ethanol 58 , sorbitol 59 , diethyl phthalate 60 , cycloheptanol 61 , cyclooctanol 61 as well as liquid crystals 62,63 (5*CB, 8*OCB). Basing on Eq. ( 10) one can propose the following distortions-sensitive transformation of experimental data: Consequently:</p><p>Temperature dependence of the configurational entropy S C (T) of experimental data expressed by Eq. (15) should follow a linear behaviour, yielding optimal values for the reference Eq. ( 10): n = 1/B and T K = B/A.</p><p>The characteristic feature of 'generalised' VFT Eq. ( 13) is power exponent n, influencing a distance from singular temperature distance T 0 . Notably, a similar correction was advised in 1984 by Bengtzelius, Götze and Sjölander (BGS) 64 , basing on the mode-coupling theory, in 1988 by Bendler and Shlezinger (BS) 65 , using the mobile defects ('random walk') approach, as well as Hall and Wolyness 66 for randomly packed spheres (HW):</p><p>where α ≈ 1.76 for BGS, α = 3/2 for BS, and α = 2 for HW models.</p><p>More recently, the random first-order transition (RFOT) model resulted in a similar dependence with an exponent α = ψ/(d − θ) 2 , where the exponent d is the spatial dimension, θ is for free energy surface cost on linear size of interface between two amorphous states and the exponent ψ is a free energy barrier that must be overcome to rearrange a correlated volume. It is worth stressing that exponent α value, for mentioned models, is located within frames empirically indicated for the exponent n [51][52][53] .</p><p>Returning to the generalised Eq. (10) for configurational entropy, one can derive the relation for previtreous changes of the heat capacity, namely: Heat capacity changes resulted from Eq. ( 17) are presented in Fig. 2, for the selected terminal, values of parameter n. Except the 'classic' case n = 1 , they show previtreous changes linked to a finite temperature singularity at (10) www.nature.com/scientificreports/ T K , which has been not expected for heat capacity so far. The insert in Fig. 2 recalls different heat capacity change patterns in a normalised scale for T → T g . To follow this issue, see also Refs. 67,68 .</p><p>One of glass transition experimental features is approaching the hypothetical Kauzmann temperature closer in heat capacity studies by increasing a cooling rate than in BDS tests for which the cooling rate factor is not important. Shifting below the standard T g value in DTA (differential thermal analysis) studies is often too strong 'anomalous' heat capacity changes. Such a behaviour via singularities appearing in Eq. ( 17). The description introduced by Eqs. ( 10) and ( 17) also correlates with recent indications for more pronounced changes of the configurational entropy than predicted by the classic Eq. ( 4) or indication for decoupling between VFT based estimations of the fragility (see comments below Eq. ( 2) and the real value of the fragility determined from the Angell plot (Fig. 3) 4,5,11,[14][15][16] .</p><p>Notably, hypothetical validity of Eq. ( 17) opens a new possibility for distortions-sensitive tests directly exploring previtreous changes of the heat capacity:</p><p>The linear regression fit for a plot based on Eq. ( 18) may yield A and B coefficients, what gives consequently</p><p>Figure 1 presents the configurational entropy evolution for supercooled glycerol, ethanol, cycloheptanol, cyclooctanol, diethyl phthalate, 5*CB and 8*OCB. Curves in the part A of Fig. 1 portraying experimental data, for selecting liquids, are related to the 'classic' Eq. ( 5) (in red) and the 'generalised' Eq. (10) (in blue). The Fig. 1A insert shows experimental data presentation based on a hardly explored scale S C vs. 1/T , directly resulted from the Eq. ( 5). Figure 1B portrays configurational entropy normalised to the Kauzmann temperature T K calculated from Eq. (10). The insert presents a behaviour of the Eq. ( 10) with different parameter n, i.e., 0.1 < n < 2.</p><p>Parameters calculated from the distortions-sensitive analysis. The T K was calculated directly from the Eq. ( 10), when the condition S C (T) → 0 ⇒ T| S C (T)=0 = T K is fulfilled. a Glass temperature calculated for the relaxation time τ = 100 s. 15). All calculated parameters n corresponds well with ones obtained using Eq. ( 10) (see Table 1).</p><!><p>Figure 4 presents results of the distortions-sensitive analysis of S C (T) experimental data based on Eq. ( 15). The linear behavior suggested by Eq. ( 15) appears, but with different slopes ( B ∼ 1/n ). Obtained parameters for studied glass-forming liquids are collected in Table 1. These values are, within the limits of the experimental errors, the same as in Ref. 43 e.g., n = 1.04 for glycerol and n = 1.28 for ethanol, which were obtained from the analysis of 'dynamic' experimental data τ (T) → I DO (T).</p><p>These results indicate that for glycerol and diethyl phthalate one can assume n = 1 , what leads to the VFT relation for relaxation time and the 'classic' expression for configurational entropy (Eq. ( 5)). On the other hand, for ethanol, sorbitol, 5*CB and 8*OCB the parameter n > 1 , what in Ref. 43 was linked to glass former consisted of molecules with the uniaxial symmetry. One can expect that in such a case, the generalised VFT Eq. ( 13) may offer much more.</p><p>The main part of Fig. 3 presents previtreous behaviour of primary relaxation time in glycerol and ethanol using Angell plot 4,5 . Figure 3 shows the linearised distortions-distortions sensitive analysis of data from the central part of the plot, based on Eq. (7). Linear domains indicate the preference for describing τ (T) changes by the VFT relation (Eq. ( 2)). Such a behaviour is evidenced for glycerol but absent for ethanol.</p><p>Results related to Figs. 3 and 5 may be considered as the argument against the 'universal' validity of the 'Stickel operator' analysis used for testing dynamic crossover phenomenon [43][44][45][46][47][48][49] , due to inherently coupling to pre-assumption of an omnipotent validity of the basic VFT relation. The question also raised for general validity 2) (in red) and the generalised VFT Eq. ( 13) (in blue), basing on the fitting domain 0.16 < T g /T < 1 . Data obtained from Ref. [51][52][53] . Molecular structures taking from Wikimedia Commons.</p>

Scientific Reports - Nature

Medium-Length Lipids Facilitate Cell-Permeability and Bioactivity

The majority of bioactive molecules act on membrane proteins or intracellular targets and therefore needs to partition into or cross biological membranes. Natural products often exhibit lipid modifications to facilitate critical molecule-membrane interactions and in many cases their bioactivity is markedly reduced upon removal of a lipid group. However, despite its importance in nature, lipid-conjugation of small molecules is not commonly used in chemical biology and medicinal chemistry, and the effect of such conjugation has not been systematically studied. To understand the composition of lipids found in natural products, we carried out a chemoinformatic characterization of the 'natural product lipidome'. According to this analysis, lipidated natural products predominantly contain saturated linear medium-length lipids, which are significantly shorter than those found in membranes and lipidated proteins. To study the usefulness of such modifications in probe design, we systematically explored the effect of lipid conjugation on five different small molecule chemotypes and find that permeability, cellular retention, subcellular localization, and bioactivity can be significantly modulated depending on the type of lipid tail used.We demonstrate that medium-length lipid tails can render impermeable molecules cell-permeable and switch on their bioactivity. Saturated medium-length lipids (e.g. C10) are found to be ideal for the bioactivity of small molecules in mammalian cells, while saturated long-chain lipids (e.g. C18) often significantly reduce bioavailability and activity. Together, our findings suggest that conjugation of small molecules with medium-length lipids could be a powerful strategy for the design of probes and drugs.

medium-length_lipids_facilitate_cell-permeability_and_bioactivity

Introduction<!>Chemoinformatic Characterization of the 'Natural Product Lipidome'<!>Cellular Uptake of Fluorophore-Lipid Conjugates<!>Discussion<!>Experimental Section<!>General procedure for synthesis of lipid-conjugated fluoresceins<!>General procedure for synthesis of lipid-conjugated rhodamines<!>General procedure for synthesis of lipid-conjugated NBDs<!>Live cell imaging of lipid-fluorophore conjugates<!>Dynamic Light Scattering<!>Parallel artificial membrane permeability assay (PAMPA)<!>Cell Viability Assay<!>COCONUT database preprocessing<!>Lipidated natural products identification<!>Lipidated natural products characterization<!>TMAP generation

<p>To function within a membrane or cell, small molecules need to be actively or passively transported into or across biological membranes. The physicochemical properties allowing for both membrane permeability and aqueous solubility are thought to be tightly confined. [1][2][3] Strikingly, nature and medicinal chemistry vary in their approaches for the synthetic design of highly bioactive molecules (Figure 1). With a few notable exceptions of amphiphilic probes and pharmacophores, including fingolimod (FTY720), 4 eliglustat, 5 palmostatin B, 6 salmeterol, 7 and orlistat, 8 synthetic small molecule probes and pharmacophores rarely exhibit lipid tails. Natural products, on the other hand commonly exhibit lipid functionalizations that fine-tune their pharmacokinetic and/or pharmacodynamic properties. Well known examples include bryostatin, 9,10 thapsigargin 11 , daptomycin, 12 and phorbol ester 13 . In addition to membrane permeability, direct drug-membrane interactions play key roles in targeting proteins that are embedded in or associated with biological membranes. [14][15][16] These include G protein-coupled receptors (GPCRs), receptor-linked enzymes, ion channels, transporters, pumps, and many transiently membrane localized proteins (e.g. RAS, PI3K, PKC), which together represent the majority of known drug targets. 17 In the case of transmembrane proteins, lipophilic drugs can enter a target directly through lateral diffusion from the intermembrane space of the lipid bilayer, as has been recently established for a number of endogenous and synthetic GPCR and ion channel ligands. 16,18,[18][19][20][21][22][23] The prevalence of lipidated natural products, and some notable examples of amphiphilic endogenous ligands, synthetic probes, and pharmacophores, as well as the fundamental tenet that the vast majority of drugs interact in critical ways with biological membranes poses many pressing questions. While the lipid composition of biological membranes 24,25 and the posttranslational modifications of proteins with lipid groups 26,27 have been major fields of research, the systematic conjugation of lipids to small molecules and the composition of lipidated natural product are largely unexplored. 28 Here, we present a systematic characterization of the lipid composition of lipidated natural products through 4 chemoinformatics, as well as a survey of the effects of lipid conjugation on several classes of small 60 molecules. 61</p><!><p>Lipid-functionalization of small molecules occurs very commonly in nature. Many natural products are lipidated, and their bioactivity is often significantly reduced when the lipid tail is altered or removed. 9,11,13 To systematically study how widespread this modification is and what type of lipids are most prevalent, we sought to characterize lipidated natural products (Figure 2A).</p><p>For our search, we turned to the COlleCtion of Open Natural prodUcTs (COCONUT) 31 , which is an open-source database of over 400,000 elucidated and predicted natural products. Our analysis focused only on the 67,656 COCONUT entries annotated with a taxonomical origin and a publication DOI. Of these, the majority are of plant origin (50%), followed by fungal (23%), bacterial (16%), homo sapiens (2.5%), animal (2%), and marine (1.5%) origin. 32 The remaining 5% lack a superclass annotation, and it was annotated as "other". To define a lipidated subset among the 67,656 unique natural products, we selected those with an uninterrupted hydrocarbon chain of at least eight atoms. Then, in order to exclude classical lipids (e.g. glycolipids, sphingolipids, and sterols), we further confined our search to molecules exhibiting at least one closed ring that is not a sugar and excluded molecules that have a molecular weight lower than 200 Da without the lipid tail, or exhibit a sterol substructure. Through such filters and manual curation, we identified 1,308 lipidated natural products.</p><p>To allow an overview of the 67,656 selected COCONUT entries and of the lipidated subset, all structures were encoded in the MAP4 (MinHashed atom pair fingerprint of diameter 4) 33 feature space and visualized in an interactive manner using TMAP (TreeMap) 34 . The MAP4 fingerprint combines the characteristics of atom pair fingerprints, which are well suited for large molecules and peptides, and substructure fingerprints, which are instead preferred for small molecules, and it is therefore suitable for both. TMAP is a dimensionality reduction method suitable for analyzing vast high-dimensional datasets used in combination with the 3D visualization tool Faerun. 35 The obtained map is available at https://tm.gdb.tools/map4/lipidatedNP_tmap/, and it can be used to visualize the lipidated NPs subset among the 67,656 COCONUT entries (Figure 2B). Additionally, the map can be navigated using Molecular weight (MW), fraction of sp3 C (Fsp3), hydrogen bond donors (HBD) and acceptors (HBA) count, the estimated octanol:water partition coefficient (AlogP), the presence/absence of peptide and sugar moieties, compliance with Lipinski's rules of five, 36 based on origin of the NPs, and the presence or absence of glycoside and peptide substructures (Figure S1). To characterize the natural product lipidome, we conducted several analyses of the identified hits. These include the distribution of lipid chain length (Figure 2B-D), correlation with hydrogen bond acceptor count (HAC) (Figure 2E), the distribution of connecting functional groups (Figure 2F-G), and the occurrence of varying degrees of saturation (Figure 2H-I). Since the lipidated NPs subset contained only 20 human NPs, 20 marine NPs, and two animal NPs, we have grouped these three classes with the 84 lipidated NPs missing a superclass annotation under the label "other". Notably, we found that the lipid tail lengths in natural products across different origins (animal, bacteria, fungi, and plants) exhibit a similar trend in length distribution in being predominantly significantly shorter than lipid groups used in the functionalization of proteins or as constitutional components of biological membranes. 37,38 Correlation with the functional group connector of the lipid tails shows that the chain length is not strongly correlated with a particular type of connecting group. Finally, our analysis shows that unsaturation is not uncommon, but most lipid chains are fully saturated. These finding suggested that medium-length lipid tails were well suited for bioactive molecules. To investigate this point closer, we set out to prepare a series of fluorophore-lipid conjugates and study their cellular uptake and distribution.</p><p>10</p><!><p>To study the capability of lipids to mediate cellular uptake of non-permeable small molecules, we synthesized a series of lipidated fluorescein thiourea conjugates from fluorescein isothyiocyanate (FITC) through treatment with various aminolipids (Figures 3A). Fluorescein exhibits low intrinsic cell permeability and has been used extensively to assess new mechanisms for cellular uptake, including thiol-mediated uptake 39,40 or cell-penetrating peptide mediated uptake. [41][42][43] Cellular uptake of lipidated fluorescein derivatives was assessed in HeLa cells through live cell fluorescent imaging (Figure 3B). Strikingly, large differences were found in intracellular fluorescence after a short (15 min) treatment of live cells with the respective lipid-fluorescein conjugate (Figure 3C and D). Very short tails in 1 (C2) and 2 (C6) did not mediate significant cellular uptake, while conjugates with longer saturated tails showed a >100 times increase in cellular fluorescence. 3 (C10), 4 (C14), and 5 (C18) showed strong plasma membrane localized fluorescence, but notably 3 and 4 also exhibit fluorescence localized to endomembranes indicative of cellular uptake. It is unclear if 5 was bound to the inner leaflet of the plasma membrane or associated with cells on the outer leaflet of the cell membrane. Conjugate 10 also exhibited strong plasma membrane localized fluorescence. Strikingly, all other lipid structures explored did not mediate strong association with the plasma membrane. These data suggests that the capacity to associate with the plasma membrane is correlated with the conformational flexibility and not molecular weight or overall lipophilicity. Other lipid structures also enabled significant cellular uptake (e.g. 6, 8, 9, and 12), while highly structurally confined types of lipids (7 and 11) including a sterol-derived structure appeared unable to mediate rapid plasma membrane association or cellular uptake. To provide further evidence that these effects are not primarily based on lipophilicity, but the ability of a lipid structure to mediate interactions with cellular membranes, we calculated the logP values and found that they did not correlate with uptake efficiency (Figure 3E). Noticeable differences were also found in the localization of lipid-fluorescein conjugates to different cellular membranes. While the overall fluorescence of 3 was stronger than 6, the latter was found to exclusively localize to ER membranes, while 3 was predominantly localized to the 151 plasma membrane (Figure 3F and G). 152 We next studied the effect of lipid conjugation on molecules that exhibit good cellular permeability to assess if lipid-conjugation could also be detrimental to cellular uptake. Rhodaminederivatives are thought to exhibit good cellular permeability and rapid cellular accumulation. 44,45 Lipid-rhodamine conjugates were prepared from rhodamine B isothiocyanate (RITC) and aminolipids (Figure 4A). We decided to test a medium-length lipid tail (C10-Rhodamine) and a long lipid tail (C18-Rhodamine) and a derivative without lipid tail for comparison (C2-Rhodamine). Both the non-lipidated analog C2-Rhodamine and C10-Rhodamine exhibited comparable cellular uptake (Figure 4B and C) indicating that a medium-length lipid tail does not interfere notably with the cellular uptake. Interestingly, C18-Rhodamine exhibited markedly reduced cellular uptake suggesting that long lipid tails inhibit the bioavailability of this small molecule scaffold. To test if the rapid uptake within 15 min is following an active or a passive transport mechanism, we used a series of inhibitors of active transport in combination with C10-Rhodamine (Figure 4D). Wortmannin is a non-specific PI3K inhibitor that inhibits clathrinmediated endocytosis, 46 cytochalasin D (CytoD) inhibits actin polymerization affecting many different types of active transport, 47 and chlorpromazine inhibits clathrin-mediated endocytosis through inhibiting the adapter complex AP2. 48 None of these drugs reduced the amount of C10-Rhodamine taken up suggesting that uptake may be mediated either through solute carriers (SLCs) or passive diffusion. Unlike the corresponding FITC-conjugate, C18-Rhodamine does not associate strongly with the plasma membrane and overall cellular fluorescence is decreased. This points to a separate mechanism that reduces bioavailability and we reasoned that this could be due to a higher propensity to form aggregates. A dynamic light scattering measurement confirmed a lower critical aggregation concentration for C18-Rhodamine compared to the other analogs (Figure 4E). This could also potentially contribute to reduced bioactivity of longer lipid-small molecule conjugates.</p><p>To expand on the number of small molecule scaffolds, we turned to another fluorophore, which is widely used in live cell imaging, nitrobenzoxadiazole (NBD). 49 NBDs are commonly used 15 for the development of fluorescent lipids and considered membrane-permeable. 50 Lipidation of NBD was achieved through nucleophilic aromatic substitution (SNAr) of 4-chloro-7-nitrobenzo-2oxa-1,3-diazole with aminolipids (Figure 4F). Analogously to the lipid-rhodamine series, we decided to conjugate NBD with C2 (C2-NBD), C10 (C10-NBD), and C18 (C18-NBD) saturated lipid tails. Notably, only the medium-length lipid conjugate C10-NBD resulted in accumulation in HeLa cells (Figure 4G and H). Both C2-NBD and C18-NBD were not found to show significant cellular uptake. While the increased uptake of C10-NBD compared to C18-NBD is consistent with our previous findings, it was surprising that C2-NBD was not found to accumulate in cells. To assess the capacity of these conjugates to permeate passively through a phospholipid bilayer, we conducted a parallel artificial membrane permeability assay (PAMPA) experiment (Figure 4I).</p><p>Interestingly, C2-NBD was found to permeate an artificial membrane much more rapidly than both C10-NBD and C18-NBD, indicating that poor retention could be the primary reason why C2-NBD does not accumulate in cells. This highlights another potential advantage of medium-chain lipid tail conjugation, namely the increased cellular retention due to the capacity of lipids to interact with proteins and membranes. Consistent with our previous findings, we found that none of the inhibitors of active cellular transport inhibited the uptake of C10-NBD (Figure 4J). turned our attention to two small molecules with well-known biological effects. Taxol and its derivatives are microtubule stabilizing molecules that have been approved for the treatment of various types of cancer. 51 Due to P-glycoprotein-mediated drug efflux, the delivery of taxol can present a challenge, and various methods have been developed to ensure lasting accumulation in cells. 52 To test if lipid-conjugation could be a suitable platform to increase bioavailability of taxol derivatives, we conjugated docetaxel (DTX) with various lipids. Lipidation could alter the distribution of taxol-conjugates in ways that either attenuate or facilitate drug efflux. Synthesis of lipid conjugates was achieved through Boc-deprotection and subsequent amide-coupling with docetaxel (Figure 5A). We synthesized docetaxel derivatives with C5 (C5-DTX), C11 (C11-DTX), and C18 (C18-DTX) saturated lipid tails and a terminal alkyne. The terminal alkyne could be used for indirect detection of the respective derivatives using copper-catalyzed azide-alkyne cycloaddition (CuAAC) (Figure 5B). We expected that the wash-out of lipid-DTX-fluorophore conjugates occurs more slowly with longer lipids due to increased cellular retention; therefore, the lower fluorescence intensity of C18-DTX compared to C11-DTX likely reflects reduced cellular uptake of C18-DTX. This result is consistent with previous findings that long-chain lipids can present a barrier to the cellular permeability of small molecules. To compare the bioactivity of Lipid-DTX conjugates, we conducted cell viability experiments in two different cancer cell lines, HeLa and MCF7 (Figure 5C) cells. Notably, we found that C11-DTX exhibits the greatest cytotoxicity of all three conjugates. While it was found to be slightly more toxic than C5-DTX, the toxicity of C18-DTX is markedly reduced. It also should be noted that non-conjugated DTX was found to be significantly more toxic than all lipid conjugates. This could be due to increased membrane affinity of these conjugates either leading to drug efflux or reduced availability for microtubule binding, which is one of the few non-membrane bound cellular drug-targets.</p><p>We next decided to explore molecules targeting membrane proteins. The quaternary ammonium-based sodium channel blockers QX222 and QX314 are commonly used in neuroscience and lack cellular permeability. [53][54][55] When applied internally (e.g. through a patch pipette), channel blocking can be observed, while external application to cells does not show any effect. 56 We decided to synthesize a series of analogs without lipid tail (C0-QX), with a mediumlength lipid tail (C10-QX), and with a long lipid tail (C16-QX). Lipidated QX-analogs were synthesized through activation of betaine as the acyl chloride and subsequent addition of alkylanilines (Figure 5D). We tested this series of QX analogs in NG108.15 cells which express Nav1.7 channels. Currents were recorded in whole cell patch clamp mode upon external application.</p><p>Notably, we only recorded external Na + current blockage with the medium-length lipid conjugate, C10-QX (Figure 5E). This suggests that the medium-length tail enables membrane entry leading to the observed bioactivity. 21</p><!><p>How small molecules interplay with biological membranes is central to their bioactivity.</p><p>Molecule-membrane interactions determine cellular uptake, retention, partioning, and accumulation in intracellular compartments. In natural products, bioactive molecules are often conjugated with lipid groups to fine-tune these interactions. Herein, we systematically characterized the types of lipid groups used in natural products, and outlined a strategy to study how different synthetic chemotypes could benefit from such conjugation.</p><p>Our chemoinformatic characterization revealed that lipids in secondary metabolites are significantly shorter than those used in the functionalization of proteins or as components of biological membranes. Systematic synthetic lipidation of fluorophores and bioactive small molecules demonstrated how permeability, retention, and bioactivity of small molecules can benefit from this conjugation. Strikingly, we found that saturated medium-length lipids facilitated cellular accumulation and bioactivity of the chemotypes explored. This was especially pronounced for non-permeable molecules like fluorescein or QX-derivatives, where lipid conjugation effectively switched on uptake and bioactivity. This also extended to other chemotypes, where this conjugation was best tolerated compared to longer lipids and even shorter lipids. In many cases, long-chain lipid tails were found to exhibit the reverse effect and hamper cellular uptake and bioactivity. In the case of fluorescein, long-chain conjugates associate readily with the plasma membrane but are less prone to cellular uptake than their medium-length analogs. Other longchain conjugates do not readily associate with the plasma membrane, which could reflect a higher propensity of longer lipids to self-aggregate. This presents another reason why longer chain lipids are generally less suitable for small molecule conjugation. Our findings on chain-length dependence of cellular uptake and bioactivity are consistent with a recent computational study on the energetic barrier for the translocation of lipidated quorum sensing modulators, finding an energetic minimum for the translocation of medium-length lipids. 57 22</p><p>Another important finding of this study is the dependence of the described effects on lipid structure. A large screen of lipid fluorescein conjugates revealed that structural differences (e.g. between a floppy saturated lipid and more rigid variants) can be highly consequential for both uptake and even affinities to different types of cellular membranes, while the overall lipophilicity (logP) is comparable. The different preferences for membrane composition could be an interesting mechanism for subcellular targeting which warrants further investigation.</p><p>Our systematic analysis of lipidated natural products and functionalization of small molecule scaffolds suggests that conjugation of lipid tails could be a useful design principle for probes and pharmacophores. Nature has used this modification extensively to fine-tune the bioactivity of natural products and it has chosen saturated medium-length lipids for this purpose.</p><p>Our data suggests that this modification should be broadly applied to improve the bioavailability of pharmacophores and probes.</p><!><p>General Methods. All reagents and solvents were purchased from commercial sources (Sigma-Aldrich, TCI Europe N.V., Strem Chemicals, etc.) and were used without further purification.</p><!><p>Fluorescein 5-isothiocyanate (15.0 mg, 38.5 mol, 1.0 equiv.) was dissolved in THF (1 mL) in a 20 mL glass vial. Aminolipid (46.2 mol, 1.2 equiv.) was added and stirred for 16 h at room temperature. Solvents were removed under reduced pressure and the residue was purified by silica flash column chromatography (CH2Cl2 MeOH:CH2Cl2 1:4) to yield the product as orange solid.</p><!><p>Rhodamine 5(6)-isothiocyanate (20.7 mg, 38.5 mol, 1.0 equiv.) was dissolved in THF (1 mL) in a 20 mL glass vial. Aminolipid (46.2 mol, 1.2 equiv.) was added and stirred for 16 h at room temperature. Solvents were removed under reduced pressure and the residue was purified by silica flash column chromatography (CH2Cl2 MeOH:CH2Cl2 1:4) to yield the product as pink solid.</p><!><p>Aminolipid (150 mol, 1.0 equiv.) was dissolved in MeOH (1 mL) in a 20 mL glass vial. 4-chloro-7-nitrobenzo-2-oxa-1,3-diazole (30.0 mg, 150 mol, 1.0 equiv.) and NaHCO3 (37.9 mg, 451 mol, 3.0 equiv.) were added and stirred at 50 °C for 3 h. MeOH was removed under reduced pressure, water was added, and acidified with 1 M HCl. The aqueous phase was extracted with EtOAc and the organic phase was dried over Na2SO4, filtered, and concentrated under reduced pressure. The residue was purified by silica flash column chromatography (hexanes EtOAc:hexanes 1:1) to yield the product as red solid.</p><p>General procedure for synthesis of lipid-conjugated taxols Docetaxel (33.0 mg, 40.8 mol, 1.0 equiv.) was dissolved in CH2Cl2 (1 mL) in a 20 mL glass vial.</p><p>TFA (1 mL) was added at 0 °C and stirred for 1 h. Sat. aqueous NaHCO3 was added to neutralize and extracted with CH2Cl2, washed with brine, dried over Na2SO4, filtered, and concentrated under reduced pressure. Lipid was dissolved in DMF (0.5 mL) in a separate 20 mL glass vial. HOBt (15.6 mg, 102 mol, 2.5 equiv.) and EDC (17.6 mg, 91.9 mol, 2.2 equiv.) were added. A solution of DIPEA (21.1 mg, 163 mol, 4.0 equiv.) in DMF (0.5 mL) was added. Crude mixture of deprotected docetaxel was added in DMF (0.5 mL) and stirred for 16 h at room temperature. Volatiles were removed under reduced pressure and the residue was purified by silica flash column chromatography (CH2Cl2 MeOH:CH2Cl2 1:9) to yield the product as colorless solid.</p><p>General procedure for synthesis of lipid-conjugated QX Betaine (20.0 mg, 169 mol, 1.0 equiv.) was suspended in DMF (1 mL) in a 20 mL glass vial.</p><p>Aniline (203 mol, 1.2 equiv.), TBTU (65.2 mg, 203 mol, 1.2 equiv.), and DIPEA (52.5 mg, 406 25 mol, 2.4 equiv.) were added and stirred for 16 h at room temperature. The solution was filtered and purified by HPLC to yield the product as colorless solid.</p><!><p>HeLa cells were plated on a poly-lysine pre-coated 8-well imaging dish at a density of 25k cells per well. After overnight incubation at 37 °C and 5% CO2, the medium was removed and lipid fluorophore conjugates were added as 10 M solution in PBS at a final concentration of 0.2 % DMSO for 15 min. The cells were washed twice with PBS and directly imaged on a Leica DMI6000B inverted confocal microscope equipped with a Leica HC PL APO 63x/1.30 Flyc CORR CS2 immersion objective and a Leica TCS SP8 X white laser.</p><!><p>Measurements were performed using a DynaPro MS/X (Wyatt Technology) with a 55 mW laser at 826.6 nm, using a detector angle of 90°. Histograms represent the average of three data sets.</p><!><p>PAMPA assays were measured with a bioassay systems kit (PAMPA-096) following the vendor's protocol.</p><!><p>Hela</p><!><p>The coconut database was downloaded, and its 400,837 entries were filtered down to the 67,730 structures having taxonomical annotation and a DOI annotation not shorter than 10 characters.</p><p>MW, Fsp3, HBD, and HBA count, and the LogP calculated following the Crippen 58 method (AlogP) were calculated using RDKit. 59 Molecules breaking more than one Lipinski's rule 36 were labeled as non-Lipinski. The presence/absence of a peptide or a glycoside moiety was evaluated using Daylight 60 SMILES arbitrary target specification (SMARTS) language and RDKit.</p><!><p>Within this analysis, lipidated natural products were selected following four criteria: (i) the presence of a terminal eight carbons long aliphatic chain, (ii) the presence of at least one non-carbohydrate ring, (iii) the presence of a non-lipidic and non-carbohydrate core of at least 200 Da (iv), and the absence of a sterol substructure. The presence of a terminal eight carbons long aliphatic chain was determined with RDKit using the Daylight SMARTS language. To assess if terminal, the lipidic chain substructure was identified through SMARTS and removed, and the length of the SMILES of the remaining fragments was calculated. When only one of the remaining fragments had a SMILES length of more than ten characters, the chain was considered terminal. Noncarbohydrates rings were counted using RDKit and the "sugar free SMILES" annotated in COCONUT. To assess the MW of the non-lipidic and non-carbohydrate core, the lipid chain substructure was identified through SMARTS and removed from the COCONUT "sugar free" SMILES, and the MW of the remaining largest fragment was calculated. The absence of sterol substructure was evaluated using Daylight SMARTS language with RDKit. The selection led to 1,390 lipidated natural products, which were further filtered down manually to 1,308 structures.</p><!><p>The 1,308 structures were characterized based on their origin, lipidic linker, and the length and unsaturation level of their longest lipidic chain. The lipidic linker was classified as amide, esters, ether, or amine using Daylight SMARTS language and RDKit. The origin of the natural products was obtained as described above. The unsaturation number was calculated by counting the number of double and triple bonds in the longest lipidic chain. The fraction of unsaturation was calculated by doubling the number of unsaturation and dividing it by the number of atoms present in the longest lipidic chain. The lipidic chains were identified as described above, and the number of unsaturation and atoms was calculated using RDKit.</p><!><p>TMAP was used to visualize the 67,730 entries COCONUT subset. MAP4 was calculated using the related open-source code. The indices generated by the MinHash procedure of the MAP4 calculation were used to create a 32 trees locality-sensitive hashing (LSH) forest. 61 Then, for each structure to display, the 20 approximate nearest neighbors (NNs) were obtained from the LSH forest, and the TMAP minimum spanning tree layout was calculated. The LSH forest and the minimum spanning tree layout were calculated using the TMAP open-source code. Finally, Fearun 35 was used to display the obtained tree layout interactively. The MW, Fsp3, HBD and HBA count, AlogP, taxonomical origin, presence/absence of a peptidic substructure, presence/absence of a glycoside substructure, the Lipinski classification, and the lipidation of the displayed structures were used to color code the interactive TMAP.</p>

ChemRxiv

Cytochrome P450-catalysed L-tryptophan nitration in thaxtomin phytotoxin biosynthesis

Thaxtomin phytotoxins produced by plant-pathogenic Streptomyces species contain a nitro group that is essential for phytotoxicity. The N,N\xe2\x80\x99-dimethyldiketopiperazine core of thaxtomins is assembled from L-phenylalanine and L-4-nitrotryptophan by a nonribosomal peptide synthetase and nitric oxide synthase-generated NO is incorporated into the nitro group, but the biosynthesis of the non-proteinogenic amino acid L-4-nitrotryptophan is unclear. Here we report that TxtE, a unique cytochrome P450, catalyzes L-tryptophan nitration using NO and O2.

cytochrome_p450-catalysed_l-tryptophan_nitration_in_thaxtomin_phytotoxin_biosynthesis

<p>Thaxtomin A exerts its phytotoxicity by inhibiting cellulose biosynthesis in plants1. Conserved pathogenicity islands on the chromosomes of Streptomyces turgidiscabies, Streptomyces scabies and Streptomyces acidiscabies contain the thaxtomin biosynthetic gene cluster (Fig. 1a)2. The txtD gene within this gene cluster encodes a nitric oxide synthase (NOS) that generates NO from L-arginine (Fig. 1b)3. Deletion of txtD in S. turgidiscabies severely attenuates thaxtomin A production3. A 15N label from the guanidino group of L-arginine is specifically incorporated into the nitro group of thaxtomin A and NO donors boost thaxtomin production in the txtD mutant, indicating that part of the nitro group is derived from NO3,4. The txtA and txtB genes encode nonribosomal peptide synthetases that are implicated in the assembly of the thaxtomin N,N'-dimethyldiketopiperazine core 1 from the unique non-proteinogenic amino acid L-4-nitrotryptophan 2, L-phenylalanine and S-adenosyl-L-methionine, using ATP and Mg2+ as cofactors5. However, the mechanism for biosynthesis of L-4-nitrotryptophan remains to be elucidated. Nitro groups in natural products usually result from oxidation of an amino group6. While biomimetic syntheses and feeding experiments with labeled precursors implicate direct nitration in the biosynthesis of two metabolites6, a natural product biosynthetic enzyme that is capable of direct nitration has never been reported. Peroxidases and globins can catalyze nitration of L-tryptophan in the presence of NO −2/H2O2 and NO/O2, respectively, but the reactions are unselective, resulting in several nitrotryptophan regioisomers7. Similarly, in Deinococcus radiodurans a NOS in partnership with an adjacently-encoded t-RNA synthetase can catalyze regioselective nitration of L-tryptophan. However, the catalytic efficiency of this transformation is very low8. In both cases, the biological significance of the reactions is unknown; it is unclear whether nitration is anything more than an adventitious side reaction. Here we report that txtE encodes a unique cytochrome P450 (CYP) that catalyzes direct regiospecific 4-nitration of L-tryptophan, the first committed step in thaxtomin A biosynthesis, using NO and O2 as substrates and spinach ferredoxin (Fd) and ferredoxin reductase (Fr) as surrogate electron donors.</p><p>CYPs are ubiquitous heme-dependent enzymes that activate molecular oxygen to catalyse a vast array of oxidative transformations including epoxidation, C- and N-hydroxylation and oxidative dealkylation9,10. Coordination of NO to the heme iron atom causes reversible inhibition of CYPs. Subsequent reaction with O2 forms reactive nitrogen species that nitrate amino acid residues such as tyrosine, irreversibly inhibiting enzyme activity7. The inhibition of CYP monooxygenases by NO significantly affects signal transduction, cellular inflammation, neurodegeneration, and drug metabolism11,12. P450nor (CYP55A1) from Fusarium oxysporum is an interesting exception, because it reduces bound NO to N2O using electrons from NADH 13.</p><p>We hypothesized that TxtE and TxtD are together responsible for 4-nitration of L-tryptophan in thaxtomin A biosynthesis. Thus, we expected the genes encoding these enzymes to be co-transcribed. S1 nuclease protection analysis confirmed co-transcription of txtE and txtD (Supplementary Methods; Supplementary Fig. 12). No additional protected RNA transcripts were observed, suggesting that txtD transcription is solely dependent on the txtE promoter. The transcriptional start site of txtE is located 232 bp upstream of the translational start site. The predicted DNA binding regions for the transcription factor are both shifted upstream of the −10 and −35 positions. A comparison of the txtE promoter region with other previously-characterized Streptomyces promoters revealed a potential core promoter motif CANNAT and a putative −35 core promoter motif, GNTTNC14. A 9 bp inverted repeat in the vicinity of the −10 and −35 regions in S. turgidiscabies (Supplementary Fig. 12) may indicate an additional level of transcriptional regulation of txtE.</p><p>To determine if TxtE is required for thaxtomin biosynthesis, txtE was deleted from S. turgidiscabies Car8 using marker-exchange mutagenesis. Thaxtomin A production was abolished in the ΔtxtE strain and partially restored by complementation with txtE expressed under the control of the thiostrepton-inducible tipA promoter in plasmid pIJ8600txtE (Supplementary Results; Supplementary Table 1). Partial restoration of thaxtomin A production in the ΔtxtEpIJ8600txtE strain and improved complementation with addition of NO donors is consistent with a polar effect on txtD (Table 1). Reverse-transcriptase PCR was used to confirm the absence of txtD expression in S. turgidiscabies ΔtxtE. The low level of thaxtomin A production in the strain lacking txtD is likely due to the production of NO by nitrate/nitrite reductases in S. turgidiscabies. In order to assess the individual contribution of TxtE to thaxtomin A biosynthesis, txtD was introduced into S. turgidiscabies ΔtxtE using pIJ8600txtD3. Ectopic expression of txtD did not restore thaxtomin A production nor did addition of NO donors, indicating that deletion of txtE alone is responsible for the loss of thaxtomin A production in this strain (Table 1 and Supplementary Table 1).</p><p>To further analyze the respective roles of TxtD and TxtE in thaxtomin A production, thiostrepton-inducible txtE and txtD were introduced into S. turgidiscabies Car8 and the ability of these constructs to modulate thaxtomin A production was examined. Overexpression of txtE in the wild type strain resulted in a 34% increase in thaxtomin A production (Supplementary Table 1), suggesting that TxtE is a rate-limiting enzyme in its biosynthesis. In contrast, overexpression of txtD in wild type resulted in a 23% decrease in thaxtomin A production, consistent with NO cytotoxicity.</p><p>Since cultures of the ΔtxtE strain did not accumulate non-nitrated thaxtomin derivatives, we hypothesized that TxtE-catalyzed 4-nitration of l-tryptophan is the first committed step in thaxtomin A biosynthesis. Addition of L-4-nitrotryptophan to cultures of the ΔtxtE strain resulted in the restoration of thaxtomin A production (Table 1). This result is consistent with the hypothesis that L-4-nitrotryptophan is a biosynthetic intermediate utilized by the nonribosomal peptide synthetase TxtB for thaxtomin A biosynthesis (Fig. 1b)5.</p><p>To directly examine the role of TxtE in L-tryptophan nitration we cloned and overexpressed txtE in E. coli. Attempted overproduction of TxtE from S. turgidiscabies was not successful. Thus, we chose to work with the ortholog from S. scabies 87-22 (94% similarity, 87% identity). Recombinant TxtE was produced as an N-terminal His6-tagged fusion protein, which was purified to homogeneity by nickel affinity and gel filtration chromatography (Supplementary Fig. 3). The latter indicated that the protein exists as a monomer in solution. Purified TxtE was analysed by UV-Vis spectroscopy (Supplementary Fig. 4). The Soret band at 447 nm in the ferrous minus ferrous-CO difference spectrum of TxtE confirmed that it is a CYP (Supplementary Fig. 4)15.</p><p>The substrate specificity of TxtE was examined by investigating changes in the UV-Vis spectrum upon titration with different amino acids. Characteristic type I binding spectra were observed (Supplementary Fig. 5) in the presence of l-tryptophan, arising from conversion of the heme iron atom from low to high spin upon substrate binding and resulting in a λmax shift from 417 nm to 390 nm16. We estimated that the dissociation constant for the L-tryptophan-TxtE complex is 60 ±6 μM from a plot of the difference in absorbance at 390 nm and 420 nm with increasing L-tryptophan concentrations (Supplementary Fig. 5). No shift in λmax was observed with d-tryptophan suggesting that TxtE is stereospecific.</p><p>Assuming that, like many other bacterial CYPs, TxtE requires the electron transfer enzymes Fd and Fr for activity, we incubated TxtE with spinach Fd and Fr, NADPH, the NO donor 2-(N,N-diethylamino)-diazenolate 2-oxide (DEANO), and L-tryptophan. The reaction was carried out in air at room temperature for 90 minutes, over which time the solution became yellow. LC-MS analyses identified a product, absent from control reactions, with m/z 250 and a λmax at 400 nm, consistent with L-4-nitrotryptophan (Fig. 2a and Supplementary Fig. 6)8. The reaction was scaled up and the product was purified from the reaction by HPLC. The retention time, the ESI-MS/MS spectrum and the aromatic region of the 1H NMR spectrum of this compound were identical to those for synthetic DL-4-nitrotryptophan (Supplementary Methods; Supplementary Figs. 7 and 8). No products with m/z corresponding to hydroxylated tryptophan derivatives were observed in either the enzymatic reaction or the control reaction lacking DEANO, suggesting that TxtE is not capable of hydroxylating L-tryptophan. Furthermore, compounds with m/z corresponding to 4-nitrosotryptophan, a potential intermediate in the nitration of L-tryptophan, were not detected in the enzymatic reaction.</p><p>CYPs usually utilize O2 as a co-substrate. O2 binds to the iron atom of the ferrous heme resulting from reduction of the ferric enzyme-substrate complex by Fd/Fr (using electrons from NADPH). The resulting ferrous-O2/ferric-superoxide complex undergoes further reduction and dehydration to form a ferryl complex (Fe(IV)=O porphyrin cation radical; often referred to as "compound I") that oxidizes the substrate10. To examine the involvement of O2 in the TxtE-catalyzed nitration reaction, we incubated the enzyme with L-tryptophan, NADPH, DEANO and spinach Fd/Fr under an 18O2 atmosphere. High resolution MS analysis of the resulting L-4-nitrotryptophan showed that it was >90% labeled with a single 18O atom (Fig. 2b). No labeling of L-4-nitrotryptophan was observed in a control reaction in which 18O2 was replaced with air (Fig. 2b), or H218O (Supplementary Fig. 9). Moreover, no turnover of L-tryptophan to L-4-nitrotryptophan was observed when air was excluded from the reaction (Supplementary Fig. 10). These data are consistent with the utilization of O2 as a co-substrate by TxtE.</p><p>Taken together, our results prompt us to propose a catalytic mechanism for the formation of potential nitrating species, based on the well-established catalytic mechanism of CYP monooxygenases (Figure 2c)9,10. L-tryptophan binds to the active site of TxtE, triggering loss of the water ligand from the ferric heme, resulting in the observed spin state change. This promotes reduction of the heme to its ferrous form by Fr/Fd and O2 binds to the heme iron atom, forming a ferric-superoxide complex. Reaction of NO with the ferric-superoxide complex affords a ferric-peroxynitrite complex, which can undergo homolytic cleavage to yield NO2 and an Fe(IV)=O species (often referred to as "compound II"). Nitration of the bound L-tryptophan could then occur via NO2 addition and compound II-mediated hydrogen atom abstraction, resulting in formation of an Fe(III)-OH species (or vice versa). An alternative mechanism for nitration involving protonation-triggered heterolytic cleavage of the ferric-peroxynitrite complex to yield NO2+ and an Fe(III)-OH species, followed by classical electrophilic aromatic substitution, is also possible. Irrespective of the nitration mechanism, protonation of the Fe(III)-OH species formed is required to regenerate the Fe(III)-OH2 resting state of the enzyme. The observed incorporation of predominantly a single oxygen atom from 18O2 into L-4-nitrotryptophan is consistent with all of the mechanistic scenarios outlined above and further experiments will be required to discriminate between them.</p><p>In conclusion, TxtE is a unique new member of the CYP superfamily that catalyzes regiospecific 4-nitration of L-tryptophan using NO and O2. Sequence alignments with several structurally-characterized CYPs indicate differences in several functionally-important regions of TxtE (Supplementary Fig. 2), providing clues to the origin of its dramatically different catalytic activity. TxtE may prove to be a useful nitration biocatalyst, given that in organic synthesis direct nitration of indoles is often carried out under harsh conditions and lacks regioselectivity17-19.</p>

PubMed Author Manuscript

Polymer immobilized Cu(I) formation and azide-alkyne cycloaddition: A one potreaction

During the polymerization of aniline using copper sulphate, act as an oxidizing agent, the in-situ synthesized Cu(I) ion catalyzed the cyclo-addition between azides and alkynes. This work represents the merging of two steps, synthesis of the catalyst and application of the catalyst, in a one pot reaction. The elimination of the separate catalyst synthesis step is economic in terms of cost and time. As aniline was used as one of the reactant components so there is no requirement to use additional base for this reaction that further eliminates the cost of the process. Again, the catalyst can be readily recovered by filtration and efficiently used for the several sets of reactions without any significant loss of catalytic activity.

polymer_immobilized_cu(i)_formation_and_azide-alkyne_cycloaddition:_a_one_potreaction

<!>Result and discussion<!>Conclusion<!>Methods

<p>T he cycloaddition between an azide and a terminal alkyne produce 1,2,3-triazoles are typical nitrogencontaining heterocyclic molecules that have attracted enormous interest due to their wide range of applications in biology 1,2 , medicinal chemistry 3,4 , design of new catalysts 5 and also found wide industrial applications such as corrosion inhibitors, agrochemicals, optical brighteners, and photographic materials 6 . The cycloaddition process is based on a copper-catalyzed reaction protocol, which is highly regioselective to produce a 1, 4-disubstituted triazoles.</p><p>The azide-alkyne cycloaddition between an azide and a terminal or internal alkyne to give a 1,4-or 1,5disubstituted 1,2,3-triazole, was developed by Rolf Huisgen 7 . The drawbacks of the Huisgen cycloaddition reaction are the requirement of high reaction temperatures and a lack of regioselectivity. Later, Sharpless 8 and Meldal 9 independently discovered that Cu(I) catalysts could facilitate the azide-alkyne cycloaddition in a regiospecific manner to give only 1,4-disubstituted triazoles.</p><p>Cycloaddition protocol was catalyzed with a Cu(I) source by using a Cu(I) salt 10 , CuSO 4 -ascorbate system 11 and stabilized Cu(I) onto polymers 12 or zeolite 13 . Copper nanoparticles 14 , metallic copper turnings 15 and CuO nanoparticles 16 have also successfully demonstrated activity for the title reaction. Cu 2 O is also a source of catalytic Cu (I) for azide-alkyne cycloaddition reactions. Applying Cu 2 O powder directly in a title reaction usually results an incomplete conversion and also require long reaction time 16 . Efforts have also been made to enhance the catalytic efficiency of Cu 2 O [17][18][19][20] . It is reported that 17 polyvinylpyrrolidone-coated Cu 2 O nanoparticles can act as an efficient catalyst for cycloaddition reactions in water at physiological temperature. The results in this paper indicated that Cu 2 O-NPs were less toxic than the commonly used CuSO 4 -reductant catalyst systems 21,22 . Polymers have the potential benefit as a support of the catalyst for a wide range of applications [23][24][25][26] due to the combination of both robust and flexible nature 27 .</p><p>Scientists has given attention to develop the catalysts for the synthesis of 1,2,3-triazoles in such a way so that Cu (I) efficiently catalyzed the reaction under mild conditions to give 1,4-disubstituted 1,2,3-triazoles. In connection with our on-going research on the development of effective catalysts for synthetic organic transformations [27][28][29][30][31][32] , we have found that a polyaniline supported Cu(I) supramolecular composite system can be used for the azidealkyne cycloaddition reaction where heterogeneous catalyst could be easily separated from the crude reaction mixture and recycled in a given process.</p><p>In recent years, the environmental aspects such as atom efficiency, waste production and energy consumption are very important issues for consideration of a chemical reaction. In this regard, the combination of two or more synthetic steps into one operation is a very appealing methodology since time, energy and resources consuming workup and purification steps can be minimized. Considering the above facts, in this present communication we report a convenient one pot method for the synthesis of polymer stabilized Cu(I) catalyst and Cu(I) catalyzed azide-alkyne cycloaddition reaction under ambient condition. In the reaction pot, polymer stabilized Cu(I) catalyst was formed due to the 'in-situ polymerization and composite formation' (IPCF) reaction [33][34][35][36][37][38] .</p><!><p>Polymer immobilized Cu(I) formation (Figure 1): Proof of evidence. In a typical experiment, aniline monomer (5.0 mM) was diluted in methanol in a conical flask and an aqueous solution of CuSO 4 , 5H 2 O (10 22 M) was added drop-wise (1:2 molar ratio of copper sulphate to aniline) to it under stirring condition. During the addition, the solution took on a green colourization and at the end a parrot green precipitation was formed at the bottom of the conical flask. The entire reaction was performed at room temperature and under open atmosphere. Here, IPCF synthesis technique has been followed for the preparation of a Cu(I)polyaniline supramolecular composite material using copper (II) sulphate as an oxidizing agent for polymerizing aniline. During the polymerization process each step is associated with a release of electron and that electron reduces the Cu 21 ion to form Cu 1 ion. The Cu 1 ion binds with the chain nitrogen of the polyaniline to form an NRCu(I) type of bond, where polymer acts as a micro ligand. The SEM image (Figure 2A) illustrates the fiber-like morphology of the Cu(I)-polyaniline complex. The TEM image (Figure 2B) shows the surface morphology and internal microstructure of the polymer. A thin area of the sample was selected for viewing and acquiring the TEM images. It is clear from the TEM image that the surface is very smooth as well as transparent and has no evidence for the presence of copper nanoparticles. Figure 2C represents the colour of the resultant dried sample. The sample was also characterized with X-ray diffraction (XRD) analysis (Figure 2D). The XRD pattern confirms the crystalline character of the polyaniline and there is no indication for the formation of the metallic copper. To confirm the valence state of copper present in the sample X-ray photoelectron spectroscopy (XPS) analysis was done. A high intensity peak at 932.5 eV could be assigned to the binding energies of Cu (I) (Figure 1D, in-set). No characteristic peaks are identified for Cu (II) and Cu (0), suggesting that copper (II) precursor is converted to Cu (I).</p><p>Figure 3 shows the optical characterization of the resultant Cu(I)polyaniline composite. The IR analysis of the fingerprint region is useful for examining the resonance modes of the benzenoid and quinoid units of polyaniline. In the IR spectra (Figure 3A), the peak at 1638 cm 21 corresponds to the group N5Q5N (where Q represents a quinoid ring), while the N-B-N group (where B represents a benzenoid ring) absorbs at 1496 cm 21 . The N-H stretching mode at 3400 cm 21 has been identified for the Cu(I)-polyaniline sample. These results are in good agreement with previously reported spectroscopic characterizations data of the polyaniline 39 . The intensity of the peak for quinoid ring structure is higher indicates that the polymers are higher in oxidation state. The UV-vis spectrum (Figure 3B) of Cu-polyaniline show a shoulder-like appearance at about 330 nm corresponds to p-p* transition of benzenoid rings (inter-band transition) and at about 400 nm a prominent broad peak represents polaron/bipolaron transition. A weak absorption band with a curvilinear behaviour has been observed within the range of 500-700 nm indicates the benzenoid to quinoid excitonic transition in both the polymers 40 . All the above microscopic and spectroscopic characterization techniques proved the formation of Cu(I)-polyaniline during the reaction between aniline and copper sulphate. Polyaniline supported Cu(I) formation and azide-alkyne cycloaddition. After confirmation of the formation of the Cu(I) species we have followed the procedure mentioned in 'Method: 1' for the cycloaddition reaction between azide and alkyne (Figure 4A).</p><p>The 1,3-dipolar cycloaddition reaction has been tested using benzyl azide, 1a, with phenyl acetylene, 2a, for the synthesis of di-substituted 1,2,3-triazoles, 3a, at room temperature under different solvent conditions such as dichloromethane, chloroform, toluene, ethanol, methanol, water and methanol : water (1:1) mixture in the presence of copper sulphate and aniline. Among the above solvents, methanol and the combination of methanol-water system gave the highest product conversion, product yield 99% for the period of 7h (Table 1). Considering the above results we have decided to use methanol as a solvent for the rest of the study to ease the work-up procedure. Due to the basic nature of aniline, in this study we did not add any external base as per recommendation for the 1,3-dipolar cycloaddition reaction 41 . The best result was achieved when the catalyst concentration was 3.0 mol% Cu (on the basis of the amount of aniline present in the reaction mixture and also considering all the aniline to be converted to polyaniline as a support). By increasing the amount for Cu concentration, no further improvement of the reaction has been identified in terms of time (Table 1, entry 6). Besides that, the reaction between benzyl azide with acetylene was also carried out in the presence of Et 3 N under the same reaction condition to find out the significance of Et 3 N in the reaction. We have observed the presence of Et 3 N delayed the reaction significantly may be due to the coordination between Et 3 N and copper sulphate forms relatively stable intermediate complex, [Cu(NEt 3 ) 4 ] 21 , which require more energy to breakup and for the participation of the reaction 42 . The product, 1-benzyl-4-phenyl-1H-1,2,3-triazole (3a), was characterised by spectroscopic method and found to be identical with the previously reported one 43 .</p><p>Based on the above optimized reaction condition, we have explored the versatility of the in-situ generated catalyst for the 1,3dipolar cycloaddition of various azides and alkynes and the results are summarized in Tables 2. In this study, we also have used structurally diverse azides and alkynes. All the substrates produced the expected cycloaddition product with very good to excellent yields and selectivity. Phenylacetylene and its derivatives (Table 2, entries www.nature.com/scientificreports 1-3) gave a higher isolated yield when coupled with azides. It was found that the yield was as high as 99% for the coupling of benzyl azide with phenylacetylene (Table 2, entry 1). When benzyl azide coupled with phenylacetylene with electron withdrawing and donating groups no such noticeable difference has been observed in terms of yield for the cycloaddition product (Table 2, entries 2 and 3 respectively). Alkyne attached with heteroaromatic molecule afforded the product 1-benzyl-4-(thiophen-3-yl)-1H-1,2,3-triazole when coupled with benzyl azide and a decrease of yield has been observed in comparison with the aromatic substituted molecules (Table 2, entry 4). Cycloaddition between aliphatic alkynes and benzyl azide (Table 2, entries 5-8) is comparatively less efficient than alkynes attached with aromatic and heteroaromatic molecules. The cycloaddition of 2-bromobenzyl azides (bromine substituted benzyl azide) with different alkynes (Table 2, entries 9-16) shows an identical reactivity trend that found for the benzyl azide (Table 2, entries 1-8). All the above products have been achieved over the period of 7 h under the ambient atmospheric condition.</p><p>Performance of the recovered catalyst. In heterogeneous catalysis, the durability of the catalyst is an important issue from the economic and sustainability point of view.</p><p>To study the performance of the recovered catalyst, for the reaction mentioned in Table 2, entry 1, we have increased the amount of the reactants by a factor of 10 (for convenience, the concentration of the copper sulphate has been changed to 0.1 mol dm 23 ) and monitored the reaction using thin layer chromatography technique. After completion of the reaction, which took about 7 h, the product (3a, as confirmed by spectroscopic analysis and with a yield of ,98%) was extracted and the other product, Cu-polyaniline, was separated. The stability and recyclability performance of the in-situ synthesized, Cupolyaniline, was tested as a catalyst for the above cycloaddition reaction using the following procedure, Figure 4B. Alkyne (1a) and azide (2a) were mixed in the presence of methanol and to this solution triethylamine and recovered Cu-polyaniline catalyst were added. In the cycloaddition reaction, the role of triethylamine is to activate the acetylenic proton to form the phenyl acetylide which further react with the copper catalyst to form copper acetylide. Copper acetylide then reacted with azide to form trizole derivative. Whereas, in one pot reaction aniline performed the role of base and no need to use an external base like triethylamine. The recovered catalyst (Cu-polyaniline) was also characterized by TEM. The presence of the copper nanoparticles was clearly noted with a wide range of size distribution (10-40 nm) on the polymer matrix (Figure 5). So far as the nanoparticles are concerned, the surface of the particles is considered to be more reactive as a catalyst and the present study revealed the similar experience during the reaction process. A yield of 98% of the coupled product (3a) has been achieved for the reaction between 1a and 2a and that took about 5 h, which is two hours less than the original single pot reaction, indicates the catalytic effect of the nanoparticles. At the end of the fifth cycle, a yield of 87% of cycloaddition product was achieved at about 5 h. The recyclability study has also been performed using the recovered catalyst in the absence of base (NEt 3 ) and only 53% of the product has been achieved under the same reaction condition for 7h.</p><p>We have also performed the kinetic studies of the cycloaddition reaction (Table 2, entry 1) for the (1) in-situ reaction, (2) reaction where the recovered Cu-polyaniline was used as a catalyst in presence of base and also (3) for the reaction using recovered Cu-polyaniline as a catalyst in absence of the base. The results are shown in the graph (Figure 6). From the graph it is clear that the recovered catalyst is more active in presence of a base than the in-situ synthesized catalyst but for the first 30 min of the reaction an identical amount of product (,5% of the yield) has been achieved for the first two reactions. So, from the kinetic study it is confirmed that Cu(I) and Cu(0) are the catalyst species, for the cycloaddition reaction between organic azides and terminal alkynes, for the reaction ( 1) and ( 2), respectively, and it is also evident from the recyclability study that the catalytic activity of copper nanoparticles are higher than copper (I). The results are also supported by the previously reported literature 44 . For the reaction using preformed Cu(0)-polymer as a catalyst in absence of base (3), the reaction was slow, only ,5% product has been formed in the first 60 min of the reaction and total 53% product has been achieved at the end of the reaction.</p><p>Various sources of the active Cu(I) catalyst for the alkyne-azide cycloaddition has been reported. Cu(II) sulphate has also been successfully used as a catalytic precursor in the presence of sodium ascorbate to generate the catalytically active Cu(I) species 45 . The Cu-carbon catalyst using charcoal and Cu(NO 3 ) 2 as the precursor in presence of water as a solvent works very efficiently for the title reaction 46 . Both Cu (I) and Cu(II) oxide show the catalytic activity for the synthesis of 1,2,3-triazole products in the multicomponent click synthesis under ambient conditions 43 . There is also an evidence of direct participation of Cu(II) for the synthesis of 1,2,3-triazoles using high catalyst loading in aqueous solutions for 20 h 47 , indicates Cu(II) may not be an efficient solution for alkyne-azide cycloaddition reaction. We found that the use of only CuSO 4 , 5H 2 O as a catalyst need more than 24 h to achieve a 55% yield of the cycloaddition product between azide and alkyl in presence of excess base.</p><p>For the synthesis of the desired compound, metal contamination in the product is a matter of serious concern 48 . Leaching of the catalyst into the product would implicate a time-consuming and costly process, which would make the whole process more expensive. Several methods have been developed to distinguish between soluble and insoluble catalysts 49 and some of these methods were also used for the current study in order to investigate whether the solid catalyst is heterogeneous or not.</p><p>As our study was carried out at ambient temperature so room temperature filtration test was performed. During this test, the catalytically active species were removed from the reaction mixture by filtration and the filtrate was monitored for catalytic activity. It was observed that after removal of the catalyst; the reaction did not proceed, indicating that no catalytically active copper remained in the filtrate. However, the filtration test alone cannot prove the heterogeneous nature of the reaction as the leached metal species may not be sufficient enough to show the catalytic performance. To confirm that, the reaction supernatant was analysed by ICP-MS (Inductively coupled plasma mass spectrometry) technique, a type of mass spectrometry which is capable of detecting metals at concentrations as low as one part in 10 12 (part per trillion) level, and no detectable amount of copper species was found in the solution suggest a heterogeneous mechanism for the cycloaddition reaction using Cu(I)-polyaniline as a catalyst.</p><p>A single pot multicomponent reaction both for Cu(I) catalyst formation and azide-alkyne cycloaddition. Most of the copper catalysed azide-alkyne cycloaddition reports are on two component (organic azide and alkyne) reaction systems. In the two component synthesis method, the organic azides need to be synthesized in advance and the isolation process can be problematic. It is thus desirable to develop an efficient one-pot methodology that uses alkyl halides and sodium azide for direct cycloaddition with alkynes in the presence of suitable catalyst. Multicomponent reactions have many advantages in comparison with multi-step reactions according to environmental and economic considerations. Therefore, the design of novel multicomponent system has attracted a lot of attention from research groups working in various areas of organic synthesis. In the present work, we also turned our attention towards the one-pot, three-component Click reaction (Table 3) in which the azidealkyne cycloaddition products were generated in-situ from their precursor, aryl bromides, sodium azide and alkyne, by minimising one step. The presence of aniline and copper sulphate in the multicomponent system acts as the precursor of Cu(I)-polyaniline catalyst in presence of methanol as a solvent for the period of 9 h to give the desired products (Table 3, entries 1-6) with the isolated yields ranging from 81-92% (Method 2). To perform the recyclability test of the catalyst for the single pot multicomponent reaction (Table 3, entry 1), we have increased the amount of all the reactants by a factor of 10 and achieved about 92% of the cycloaddition product, 1-benzyl-4-phenyl-1H-1,2,3-triazole (3a), in 9h. After the first run, we have recovered the copper-polymer composite and used for the recyclability test to find out the performance of the reused catalyst. At the end of first cycle a yield of 92% of the coupled product (3a) has been achieved and that took about 8 h, which is one hour less than the original single pot multicomponent reaction. The reason for the improved performance can be addressed in terms of nanoparticle formation (as discussed before). At the end of the fifth cycle, a yield of 76% of cycloaddition product was achieved at 8 h (Figure 7).</p><p>The in situ generated Cu(I) plays the catalytic role for the title reaction. Polyaniline acts as a ligand to coordinate to the Cu(I) species which involves the formation of a Cu(I)-acetylidine complex through the coordination with alkyne followed by the addition with the azide group to give 1,2,3-triazole. It is also important to mention that in the present study we found that all reactions were highly regioselective towards the formation of 1,4-disubstituted triazoles with a wide range of diversely substituted terminal alkynes and azides under the optimized conditions.</p><!><p>In this report, we have presented an interesting method where the catalyst formation occurs in the reaction medium that prevents the catalyst from the environmental degradation. The elimination of the separate catalyst synthesis step may be economical by saving the time as well as the solvents. Aniline was used as one of the reactant components so there was no requirement of adding additional base for this reaction as recommended by the original protocol of the azidealkyne cycloaddition (Click) reaction. Furthermore, the catalyst can be readily recovered by filtration and efficiently used for the similar reaction without any significant loss of catalytic activity. The operational simplicity and the purity (regeioslectivity) of the products make this method attractive for wide range of applications.</p><!><p>General procedure for azide and alkyne cycloaddition reaction. In a 25 mL round bottom flask, alkyne (1 equiv.) and azide (1 equiv., benzyl azide/o-bromo benzyl azide) were taken and dissolved in 5 ml methanol. To this reaction mixture 1 ml of 0.1 M of aniline in methanol was added and stirred at room temperature. To this solution 5 ml of 0.01 M solution of CuSO 4 , 5H 2 O (in water) was added drop wise. A green colourization was appeared during the addition of the CuSO 4 , 5H 2 O. The reaction mixture was stirred for 7 h at room temperature and progress of the reaction was monitored using thin layer chromatography technique. After completion, the reaction mixture was filtered and the residue was dissolved with methanol. The remaining solid catalyst was recovered, dried and reused for the recyclability experiment. The methanol was evaporated from the filtrate and extracted with ethyl acetate, washed with water and dried over anhydrous sodium sulphate. Combined organic layer was concentrated in vacuum to give the corresponding triazoles which was pure enough or was purified by column chromatography technique. The products were characterised by spectroscopic analysis or by comparison of the spectroscopic data with those described in the literature.</p><p>General procedure for multicomponent azide-alkyne cycloaddition. The above mentioned procedure was followed in a 25 mL round bottom flask using alkyl halide (1 equiv.), NaN 3 (1 equiv.) and an alkyne (1 equiv.) in methanol (5.0 mL) in the presence of 1 ml of 0.1 M of aniline. To this solution 5 ml of 0.01 M solution of CuSO 4 , 5H 2 O (in water) was added drop wise for the cycloaddition reaction. 3, entry 1.</p>

Scientific Reports - Nature

An UPLC-MS/MS Method for Determination of Osimertinib in Rat Plasma: Application to Investigating the Effect of Ginsenoside Rg3 on the Pharmacokinetics of Osimertinib

Osimertinib is a novel oral, potent, and irreversible epidermal growth factor receptor tyrosine kinase inhibitor (EGFR-TKI) for treatment of advanced T790M mutation-positive advanced non-small cell lung cancer, which is commonly combined with ginsenoside Rg3 in clinic to enhance the efficacy and minimize adverse reactions. In the present study, a highly sensitive UPLC-MS/MS method was established and validated for analysis of osimertinib in rat plasma according to US FDA guideline. Separation was performed on a C18 (2.1 × 50 mm, 2.6 μm) column using a gradient elution of ammonium formate (10 mM) with 0.1% formic acid buffer (A) and ACN (B) at a flow rate of 0.2 mL/min. Detection was carried out on a triple quadrupole tandem mass spectrometer equipped with electrospray ionization in the MRM mode. The method was validated over a concentration range of 1–400 ng/mL for osimertinib. The intra- and interday accuracy and precision values were within ±15%. No significant degradation occurred under the experimental conditions in stability assays. There was a further investigation on the effects of multiple doses of ginsenoside Rg3 on the pharmacokinetics of osimertinib in rats for the first time. The results implied that osimertinib exhibited a slow absorption and moderate-rate elimination in rats following oral administration. Coadministeration with ginsenoside Rg3 (5 mg/kg, 7 days, i.g.) may have no effect on the pharmacokinetics of osimertinib in rats. The results provide a reference for the clinical concomitant medications of Rg3 and osimertinib.

an_uplc-ms/ms_method_for_determination_of_osimertinib_in_rat_plasma:_application_to_investigating_th

1. Introduction<!>2.1. Chemical and Reagents<!>2.2. Animals<!>2.3. Preparation of Stock Solution, Calibration Standards, and Quality Control Samples<!>2.4. Sample Preparation<!>2.5. Chromatographic Separation and MS/MS Conditions<!>2.6. Analytical Method Validation<!>2.6.1. Selectivity and Specificity<!>2.6.2. Linearity and LLOQ<!>2.6.3. Precision and Accuracy<!>2.6.4. Matrix Effect and Recovery<!>2.6.5. Stability<!>2.7. Application to Pharmacokinetic Study in Rats<!>2.8. Data Analysis<!>3.1. Optimization of Mass Spectrometric Parameters<!>3.2. Optimization of Chromatographic Conditions<!>3.3. Sample Preparation<!>3.4.1. Selectivity and Specificity<!>3.4.2. Linearity and LLOQ<!>3.4.3. Accuracy and Precision<!>3.4.4. Matrix Effect and Extraction Recovery<!>3.4.5. Stability<!>3.4.6. Effect of Rg3 on the PK Profiles of Osimertinib in Rats<!>4. Conclusions

<p>Osimertinib (AZD9291), the first third-generation oral potent and irreversible epidermal growth factor receptor tyrosine kinase inhibitor (EGFR-TKI), was approved by US-FDA and NMPA for first-line treatment of advanced non-small cell lung cancer (NSCLC) patients with acquired metastatic EGFR T790M mutation [1, 2]. Pharmacokinetic results shows that the absorption of osimertinib in rats was slow with the Tmax values about 4.48 h, and the elimination might be slow (t1/2 and MRT values were estimated to be 14.96 and 14.51 h) [3]. Preclinical studies demonstrate that osimertinib is principally metabolized by cytochrome P450 (CYP3A), and AZ5104 and AZ7550 are the two circulating active metabolites. In vitro studies also describe osimertinib as the substrate of P-gp and BCRP [3]. Hence, the pharmacokinetic profiles and therapy effect of osimertinib could be significantly altered when coadministered with other medications. A phase I study found that the exposure (AUC) of osimertinib increased by 24% when coadministered with itraconazole (strong CYP3A4 inhibitor) and decreased by 78% when coadministered with rifampicin (CYP3A4 inducer) in patients with advanced NSCLC [4].</p><p>Ginsenoside Rg3, the compound extracted from the Chinese herb Panax ginseng, proved to inhibit the growth, invasion, and metastasis of several kinds of malignancies, including lung cancer, breast cancer, ovarian cancer, glioma, and leukemia, and enhance the efficacy of chemotherapy. Kim et al. reported that Shenyi capsule (the main active ingredient is ginsenoside Rg3) could improve the life span of NSCLC patients by improving the immune function and antitumor angiogenesis [5]. Meta-analyses demonstrated that Shenyi capsule plus chemotherapy could increase incidence of short-term efficacy and improve the quality of life and survival rate of NSCLC patients compared with chemotherapy alone [6]. However, many literatures reported that ginsenoside could modulate CYP450s and transporters [7–10]. Malati et al. found that the exposure of midazolam when orally administered via the gastrointestinal tract for 28 days in twelve healthy volunteers was significantly reduced, and the reason may be that the gastrointestinal tract induced CYP3A activity [9]. Lei-Qiong Yang reported that ginsenoside Rg3 significantly enhanced the oral bioavailability of paclitaxel in rats and improved antitumor activity in nude mice via inhibiting P-glycoprotein (P-gp) [11]. Dou et al. demonstrated that Rg3 could induce MRP1 to attenuate NAPQI-induced toxicity by activating Nrf2 [12]. In case of adverse reactions and inadequate dosages, drug interactions should be evaluated, when Rg3 and TKIs are comedicated.</p><p>An appropriate analytical method should be established for evaluating the drug interaction between Rg3 and osimertinib. However, to the best of our knowledge, there were only two published articles about liquid chromatography-tandem mass spectrometry (LC-MS/MS) methods for determination of osimertinib in rat plasma [13, 14]. There were some drawbacks in these methods such as being time-consuming, tedious sample preparation, and high sample volume (100 μL), which was not adequate for pharmacokinetic studies. Therefore, this paper describes the development and validation of UPLC-MS/MS method for the quantitative analysis of osimertinib in rat plasma, using nilotinib as an internal standard. Samples were pretreated by an optimized protein precipitation method with acetonitrile (ACN) as the precipitant. This method provides a simple, sensitive, fast, and accurate method for the quantitative analysis of osimertinib. Meanwhile, this method was successfully applied for the drug-interaction pharmacokinetic study of osimertinib in rats after oral administration.</p><!><p>Osimertinib (purity ≥ 99.95%) was purchased from MedChemExpress Co., Ltd. (Shanghai, China). Nilotinib (purity ≥ 99%) was procured from Aladdin Co., Ltd. (Shanghai). Shenyi capsule (mainly included 10 mg ginsenoside Rg3 in one capsule) was supplied from Yatai Pharmaceuticals Co., Ltd. (Jiutai, China). The chemical structures of the analytes are depicted in Figure 1. Methanol (MeOH, HPLC grade) and acetonitrile (ACN, HPLC grade) was obtained from Thermo Fisher Scientific (Fairlawn, NJ, USA). Ammonium acetate was obtained from Zhiyuan Chemical Co., Ltd. (Tianjin, China). Formic acid (HPLC grade) was provided from Aladdin Co., Ltd. (Shanghai). Ultrapure water was acquired from the Milli-Q water purification system (Millipore, Bedford, MA, USA).</p><!><p>Male Sprague-Dawley rats, weighing 200–220 g, were obtained from Beijing Vital River Laboratory Animal Technology (Beijing, China). The rats were housed in an environmentally controlled breeding room (12 h light/dark cycle) for three days before the experiment, and the standard laboratory food and water were available ad libitum. All experimental protocols were approved by the First Affiliated Hospital of Zhengzhou University Animal Ethics Committee. Animal facilities and welfares were performed strictly in accordance with the National Institutes of Health Guidelines regarding the principles of animal care.</p><!><p>Dissolving accurately weighed amounts of standard osimertinib and nilotinib in methanol to obtain the stock solutions in a final concentration of 2 mg/mL. To prepare the working solutions, stock solutions were diluted with ACN/water (1/1, V/V). The calibration standard samples were prepared by spiking the corresponding working solutions with blank rat plasma. Final calibration standard concentrations for osimertinib were 400, 300, 200, 100, 50, 20, 5, and 1 ng/mL, respectively. In a similar way, LLOQ and QC samples were prepared in diluting working solutions with rat plasma at four concentration levels: 350 ng/mL (high quality control (HQC)), 30 ng/mL (medium quality control (MQC)), 3 ng/mL (low quality control (LQC)), and 1 ng/mL (lower limit of quantification (LLOQ)). All stock and working solutions were stored at −80°C until further analysis.</p><!><p>Samples were extracted from rat plasma by protein precipitation. For this method, a total of 50 μL of plasma sample (blank, calibration standards, LLOQ, and QCs) was mixed with 5 μL of IS (600 ng/mL) solution. Then, 150 μL ACN was added for protein precipitation. The tubes were closed and vortexed and mixed for 1 min. The samples were centrifuged at 14,000 rpm for 10 min at 4°C. Then, 100 μL of supernatant was separated and transferred to fresh EP tubes and kept in the refrigerator prior to analysis. Finally, 5 μL of the solution was injected into the UPLC-MS/MS system.</p><!><p>Osimertinib and IS were analyzed by an Exion LC Analytical System (AB Sciex, USA) coupled with a Qtrap 4500 mass spectrometer equipped with Turbo Ion Spray Interface (AB Sciex, USA). Separation of osimertinib and IS was achieved on a Phenomenex HPLC Kinetex C18 column (2.1 × 50 mm, 2.6 μm) with a column heater kept at 40°C. Gradient elution was applied at a flow of 0.2 mL/min and performed by varying the proportion of solvent A (10 mM ammonium formate with 0.1% formic acid buffer) and solvent B (ACN) as follows: 0–0.8 min (5% B); 0.8–1 min (5–60% B); 1–5.5 min (60%B); 5.5–6 min (60–5%B), 6–8 min (5%B). The temperature of the autosampler was optimized at 4°C, and the injection volume was 5 μL.</p><p>The Qtrap 4500 mass spectrometer equipped with Turbo Ion Spray interface operating in positive ESI mode (AB Sciex, USA) was selected for mass spectrometric detection. The multiple reaction monitoring (MRM) mode was acquired, and the MS spectrometry parameters were defined as follows: source temperature 500°C; ion spray voltage 4500 V; nebulizer gas (gas1) 50 psi; heater gas (gas2) 50 psi; curtain gas 40 psi; a low collision gas. The dwell time for osimertinib and nilotinib was 100 ms. The mass-dependent parameters are summarized in Table 1, and the data acquisition was processed with Analyst™ software (AB Sciex, version 1.6.3, USA).</p><!><p>According to the guidelines of the US Food and Drug Administration (FDA) [15], European Medicines Agency (EMA) [16], CFDA guidelines for the validation of bioanalytical methods [17], and Pharmacopoeia of the People's Republic of China [18], the present UPLC-MS/MS method was validated by the following parameters: selectivity, linearity and LLOQ, precision and accuracy, matrix effect and recovery, and stability.</p><!><p>To explore the selectivity and specificity of this method, six individual blank rat plasma samples, blank rat plasma samples spiked with osimertinib and IS at LLOQ level, and actual rat plasma samples after oral administration of osimertinib were analyzed.</p><!><p>The linearity, expressed by calibration curves with the correlation coefficient (R2) and weighed 1/x2 quadratic least-squares regression analysis, was evaluated by analyzing the standard rat plasma samples over a range of 1–400 ng/mL (at least eight concentration levels) in three separate days. We defined the LLOQ as the lowest validated concentration on the calibration curve with a signal-to-noise ratio (S/N) of more than 10. The back-calculated concentration on this calibration curve was less than 15% of the nominal concentration and less than 20% at the LLOQ level.</p><!><p>Four different concentration levels of QC samples (1, 3, 30, and 350 ng/mL for LLOQ, LQC, MQC, and HQC, resp.) with six replicates for osimertinib were determined to evaluate the precision (intraday and interday) and accuracy of this method. The intraday experiment was assayed in one run within one day, and the interday experiment was assessed in three analytical runs on three separate days. The intraday and interday precision were assessed by calculating the relative standard deviation (RSD) and accuracies, which was required to be within ±15% of the nominal concentrations for LQC, MQC, and HQC samples and should be less than 20% RSD for LLOQ. Accuracy was calculated by analyzing the averaged measurements to normal values, which was expressed in relative error percentage: [(calculated concentration − true concentration/true concentration) ∗ 100].</p><!><p>The matrix effect of osimertinib was investigated by matrix factor, a peak area ratio of the analyte/IS with three QC concentrations in the presence of matrix ions (rat plasma) to those in the absence of matrix at equivalent concentrations. The RSD% of the matrix effect should less than 15%. The recovery was assessed by comparing the peak areas obtained from extracted QC samples with those in the mobile phase at the same concentrations and expressed as percentage. The matrix effect and recovery of IS were evaluated in the similar method at a concentration of 600 ng/mL in plasma.</p><!><p>(1) Short-Term Stability. QC samples (low, medium, and high levels) were assessed by analyzing QC samples kept at room temperature for six hours.</p><p>(2) Autosampler Stability. QC samples (low, medium, and high levels) were evaluated after the processed QC samples were placed in an autosampler (4°C) for 24 hours.</p><p>(3) Freeze-Thaw Stability. QC samples (low, medium, and high levels) were investigated after storage at room temperature for 8 h, followed by three freeze-thaw cycles (thawing at room temperature during 2 h and freezing again at −80°C for at least 12 h).</p><p>(4) Long-Term Stability. QC samples (low, medium, and high levels) were determined by analyzing after storage at −80°C for 28 days.</p><!><p>The rats were randomly divided into two groups (six rats for each group). Osimertinib was dissolved and suspended by 1% carboxymethylcellulose sodium solution. The Rg3 powder in Shenyi capsule was dissolved in saline. In the first period, ginsenoside Rg3 (5 mg/kg) was orally administered to the rats in group I for seven days, and the rats in the other group were orally treated with normal saline for seven days. In the second period, the rats were fasted overnight with free access to water before the experiment. After overnight fasting, ginsenoside Rg3 (5 mg/kg) plus osimertinib (10 mg/kg) was orally administered to the rats in group I, and the rats received oral ginsenoside Rg3 30 min prior to osimertinib at the eighth day. Rats in group II were only orally treated with osimertinib (10 mg/kg) at the eighth day. Blood samples (each 100 μL) were collected via the oculi chorioideae vein at 0.5, 1, 2, 3, 4, 5, 6, 7, 8, 10, 12, 15, and 24 h after treatment in heparinized tubes. The blood samples were immediately centrifuged at 3000 rpm for 10 min. The plasma was separated from blood samples and transferred to clean tubes and immediately stored at −80°C until further analysis. The plasma concentrations of osimertinib were determined according to the aforementioned developed UPLC-MS/MS method.</p><!><p>The pharmacokinetic parameters, including the maximal plasma concentration (Cmax), area under the plasma concentration-time curve (AUC), time for the maximal plasma concentration (Tmax), and mean residence time (MRT), were calculated through noncompartmental analysis by Phoenix WinNonlin (Pharsight Inc., USA, version 1.1) software.</p><p>All data were presented as the mean ± SD. Statistical differences analyses between the mean values were performed in SPSS software version 11.5 (SPSS, Chicago, IL, USA) and analyzed for significance by a nonpaired two-tailed Student's t-test. p values less than 0.05 were considered statistically significant.</p><!><p>In order to achieve high sensitivity and better response, the positive ionization and multiple reaction monitoring scan mode was applied and optimized by a systematic approach. The precursor and product ions were determined by directly injecting standard solution of osimertinib and IS to the mass spectrometer (Figure 2). It was observed that the high-intensity peaks of the analytes were conducted in these optimized mass spectrometric conditions. The main mass spectrometer parameters of osimertinib in our study were similar to previous studies [19, 20].</p><!><p>To optimize the appropriate chromatographic conditions for separation, commercially available columns and various mobile phases (various proportions and gradients) were evaluated for their chromatographic behavior and the ionization response.</p><p>The results indicated that Phenomenex HPLC Kinetex C18 column (2.1 × 50 mm, 2.6 μm) achieved reasonable run time, symmetric peak shape, and good resolution. The different proportions of ACN were chosen on account of its low background noise, strong elution effect, and narrower peaks. Various proportions and gradients of water, 0.1% formic acid, and ACN and water, 1, 2, 5, and 10 ammonium formate, and ACN were tested during optimization. The results indicated that ammonium formate (10 mM) with 0.1% formic acid in gradient elution achieved desired separation, good peak symmetry, retention time, and appropriate MS responses. The retention time for osimertinib and nilotinib (IS) was 2.51 and 2.90 min, respectively, in a runtime of 8.00 min.</p><!><p>Various extraction was evaluated to achieve a maximum extraction efficiency of osimertinib and IS in rat plasma. The solid-phase extraction method was unnecessary for its complication. Poor recovery was observed in liquid-liquid extraction method with n-hexane and ethyl acetate as extraction solvent. As previously mentioned, the protein precipitation method was applied for its simple steps and sufficient recovery. The methanol and ACN were tested for protein precipitation solvents, and the results implied that cold ACN could achieve high protein precipitation efficiency and better mass spectrometric response. A sample: ACN ratio of 1 : 4 resulted in a promising recovery of osimertinib and IS. One hundred μL of rat plasma was prepared for sample preparation as Dong described [14]. Many blood samples from each rat were collected for the pharmacokinetic study, which may influence the physiological and functional characteristics of rats. In addition, the supernatant was evaporated to dryness through nitrogen gas for sample preparation which was time-consuming and tedious [14]. Protein precipitation method was chosen in our experiment, and the method was simple and fast.</p><!><p>There was no significant chromatographic interference peaks from endogenous substances at the retention time with the osimertinib and IS in rat plasma. The responses of osimertinib in blank rat plasma samples were all <20% of the LLOQ response, and the blank IS responses were below 1% of the normal response. The typical chromatograms of blank rat plasma, blank plasma sample spiked with osimertinib and IS at LLOQ level, and rat plasma sample after administration were shown in Figure 3.</p><!><p>Good linearity of the calibration curve of osimertinib, generated by linear regression of peak area ratios against concentrations with a weighting factor of 1/x2 at eight concentrations over the range of 1 to 400 ng/mL in rat plasma, was observed in this method. The regression coefficients of all the calibration curves were greater than 0.99, and the typical equation was y = 0.04111x  + 0.02013 (r2 = 0.99862) (y represents the peak area ratios of osimertinib to the IS; x stands for the plasma concentrations of osimertinib). Back-calculated concentrations of calibration standards for osimertinib in rat plasma are listed in Table 2. LLOQ for determination of osimertinib in rat plasma was 1 ng/mL, and the ratio of signal-to-noise was much higher than five.</p><!><p>The intra- and interday precision and accuracy for the determination of osimertinib at LLOQ and three QC levels in rat plasmas on three consecutive days are summarized in Table 3. Compared with nominal concentration, RSD of LLOQ and QC concentration were less than 15%, and the accuracy was within 85%–115%. The intra- and interday precision and accuracy assays only investigated LQC, MQC, and HQC concentrations as Dong described [14]. According to the Bioanalytical Method Validation Guidance of FDA, the precision and accuracy assays should be validated with four QC levels per run (LLOQ, LQC, MQC, and HQC) and should be ±20% of nominal concentration at LLOQ [15]. The four QC levels (LLOQ, LQC, MQC, and HQC) were validated for precision and accuracy assay in our experiment. Overall, the accuracy and precision values were within the validation criterion, and the method was highly reliable and reproducible for the determination of osimertinib in rat plasma.</p><!><p>The ratio of the peak area in the presence of matrix to the peak area in absence of matrix was determined as matrix factor. The IS-normalized osimertinib was calculated by dividing the osimertinib of the analyte by the osimertinib of the IS. As shown in Table 4, the RSD of the IS-normalized matrix factor of osimertinib and IS in rat plasma at three QC levels was in the range of 0.71% to 6.97%, which demonstrates that no endogenous substance could cause a significant effect on ionization. The overall mean extraction recovery of osimertinib in rat plasma was 99.72 ± 7.11% to 109.75 ± 8.38%, and the recovery of IS was 98.99 ± 7.90%, indicating no significant loss during the extraction process. As previously mentioned, the results indicated that high recovery and no significant matrix effect aided the successful validation of the method for osimertinib and IS in rat plasma.</p><!><p>The stability of osimertinib at three QC concentrations under four different storage conditions is described in Table 5. According to the results, the measured concentrations, kept at room temperature for 6 h, placed in an autosampler for 24 h, freeze-thawed (−80°C) for three cycles, and stored at −80°C for 28 days, were all within ±15% of the norminal concentrations, which indicated that osimertinib was stable under the conditions investigated in this study.</p><!><p>The current fully validated UPLC-MS/MS method was successfully applied for determining the concentration of osimertinib in rat plasma after oral administration of osimertinib (10 mg/kg) to male rats in the presence or absence of Rg3 (5 mg/kg) for seven days. The mean plasma concentration–time curves of osimertinib in rats are shown in Figure 4, and the major pharmacokinetic parameters are summarized in Table 6.</p><p>The pharmacokinetic data showed that osimertinib was slowly absorbed at the highest plasma concentration in a long time (Tmax was 3.33 ± 0.82 h) after oral administration (10 mg/kg) in rats. The peak plasma concentration (Cmax) was 0.317 ± 0.138 μg/mL achieved with the area under the curve (AUC0-t) that was 2.823 ± 1.206 h·μg/mL. The mean terminal half-life (t1/2) was 4.46 ± 0.94 h, and the mean residence time (MRT0-t) was estimated to be 6.88 ± 0.62 h. Additionally, the plasma clearance (CL/F) and distribution volume (Vz/F) values were 19.821 ± 8.132 L/h/kg and 129.695 ± 58.112 L/kg, respectively. All the calculated pharmacokinetic parameters revealed that osimertinib gets absorbed slowly into blood and exerts a moderate elimination rate in rats. The obtained pharmacokinetic parameters were in agreement with the previous studies [13, 14].</p><p>Small-molecule tyrosine kinase inhibitors (TKIs) that target the EGFR are commonly used in combination with Chinese herbal medicines (e.g., ginseng, astragalus, and Scutellaria barbata) to enhance the efficacy and overcoming the drug resistance in NSCLC patients [21, 22]. Osimertinib was given frequently coadministered with many compound preparations including active ingredient Rg3 for exerting antihepatitis, anti-inflammatory, and antiviral effects. It is worth noting that the potential for clinical DDIs with these EGFR-TKIs could provide recommendations for managing, minimizing, or avoiding DDIs with the different agents [3, 23]. A clear and detailed understanding of their propensity for drug–drug interactions (DDIs) is required before their concomitant use in NSCLC patients. However, there was no report regarding the effect of Rg3 on the pharmacokinetics profile of osimertinib. As Rg3 and osimertinib were usually taken long term in clinic, we have investigated the effect of Rg3 on the pharmacokinetics profile of osimertinib for the first time.</p><p>On coadministering with Rg3 (5 mg/kg) for seven days, Cmax, AUC0-t,t1/2 values of osimertinib in rats were 0.297 ± 0.122 μg/mL, 2.597 ± 0.839 h·μg/mL, 5.08 ± 1.03 h, respectively. There were no significant changes observed in Cmax, AUC0-t, t1/2, Tmax, CL/F, and the other main pharmacokinetic parameters of osimertinib in coadministered and alone group. The results implied that there might have been no significant effect on the main pharmacokinetic parameters of osimertinib in rats when coadministered with Rg3 (5 mg/kg) for seven days. It is reasonable to hypothesize that Rg3 may have no potential effect on CYP3A in rats because of CYP3A mainly mediated the metabolism of osimertinib.</p><p>This validated UPLC-MS/MS method and pharmacokinetic study may prove to be of great significance for future investigations when conducting human clinical drug interaction trials on Rg3 and osimertinib.</p><!><p>In conclusion, a sensitive, rapid, and selective UPLC-MS/MS method was developed and validated for quantification of osimertinib in rat plasma, within a linear range of 1–400 ng/mL. This method was successfully applied to the pharmacokinetic study of osimertinib in rats across different administrations. This is the first study that discloses the effect of Rg3 on the pharmacokinetic profiles of osimertinib in rats at a clinical dosage level. Pharmacokinetic properties demonstrated that osimertinib exhibited a slow absorption and moderate-rate elimination in rats following oral administration of osimertinib (10 mg/kg). Coadministration with Rg3 (5 mg/kg) for seven days may have no obvious effects on the pharmacokinetics of osimertinib in rats. This study provides the theoretical foundation for the concomitant medications of Rg3 and osimertinib such that Rg3 could improve the life span, quality of life, and survival rate of NSCLC patients.</p>

PubMed Open Access

Boosting the Anticancer Activity of Aspergillus flavus “endophyte of Jojoba” Taxol via Conjugation with Gold Nanoparticles Mediated by γ-Irradiation

Taxol production by fungi is one of the promising alternative approaches, regarding to the natural and semisynthetic sources; however, the lower yield and rapid loss of Taxol productivity by fungi are the major challenges that halt their further industrial implementation. Thus, searching for fungal isolates with affordable Taxol-production stability, in addition to enhance its anticancer activity via conjugation with gold nanoparticles, is the main objectives of this study. Twenty-four endophytic fungal isolates were recovered from the barks, twigs, and leaves of jojoba plant, among these fungi, Aspergillus flavus MW485934.1 was the most potent Taxol producer (88.6 µg/l). The chemical identity of the extracted Taxol of A. flavus was verified by the TLC, HPLC, HNMR, and FTIR analyses. The yield of Taxol produced by A. flavus was optimized by the response surface methodology (RSM) using Plackett–Burman (PBD) and faced central composite designs (FCCD). The yield of Taxol by A. flavus was increased by about 3.2 folds comparing to the control cultures (from 96.5 into 302.7 µg/l). The highest Taxol yield by was obtained growing A. flavus on a modified malt extract medium (g/l) (malt extract 20.0, peptone 2.0, sucrose 20.0, soytone 2.0, cysteine 0.5, glutamine 0.5, and beef extract 1.0 adjusted to pH 6.0) and incubated at 30 °C for 16 days. From the FCCD design, the significant variables affecting Taxol production by A. flavus were cysteine, pH, and incubation time. Upon A. flavus γ-irradiation at 1.0 kGy, the Taxol yield was increased by about 1.25 fold (375.9 µg/l). To boost its anticancer activity, the purified Taxol was conjugated with gold nanoparticles (AuNPs) mediated by γ-rays irradiation (0.5 kGy), and the physicochemical properties of Taxol-AuNPs composite were evaluated by UV–Vis, DLS, XRD, and TEM analyses. The IC50 values of the native-Taxol and Taxol-AuNPs conjugates towards HEPG-2 cells were 4.06 and 2.1 µg/ml, while the IC50 values against MCF-7 were 6.07 and 3.3 µg/ml, respectively. Thus, the anticancer activity of Taxol-AuNPs composite was increased by 2 folds comparing to the native Taxol towards HEPG-2 and MCF-7 cell lines. Also, the antimicrobial activity of Taxol against the multidrug resistant bacteria was dramatically increased upon conjugation with AuNPs comparing to authentic AuNPs and Taxol, ensuring the higher solubility, targetability, and efficiency of Taxol upon AuNPs conjugation.Supplementary InformationThe online version contains supplementary material available at 10.1007/s12010-022-03906-8.

boosting_the_anticancer_activity_of_aspergillus_flavus_“endophyte_of_jojoba”_taxol_via_conjugation_w

<!>Introduction<!>Isolation and Culturing of the Endophytic Fungi<!>Screening, Extraction, and Quantification of Taxol from the Endophytic Fungi<!>Morphological and Molecular Identification of the Recovered Endophytic Fungi<!>Chemical Structure of the Extracted Taxol<!>Effect of Different Types of Media on Taxol Production<!>Bioprocess Optimization of the Nutritional Conditions to Maximize the Taxol Yield<!>Placket-Burman Design<!>Central Composite Design and Interactions Between Factors Affecting Taxol Production<!>Effect of Gamma Irradiation on Taxol Yield<!>Synthesis and Characterization of Gold Nanoparticles (AuNPs); Conjugation with Taxol<!>Anticancer Activity of Taxol<!>Antimicrobial Activity of Taxol and Taxol-AuNPs Conjugates<!>Statistical Analyses<!>Fungal Deposition<!><!>Impact of Gamma Radiation on A. flavus on the Taxol Productivity<!><!>Synthesis of AuNPs and Conjugation with A. flavus Taxol Mediated by γ- Radiation<!><!>Discussion<!>

<p>Open access funding provided by The Science, Technology & Innovation Funding Authority (STDF) in cooperation with The Egyptian Knowledge Bank (EKB).</p><!><p>Taxol is one of the most commercialized broad spectrum anticancer drugs [1]. The activity of Taxol elaborates from its unique specificity for binding with the cellular tubulin β-subunits heterodimer, promoting tubulin polymerization, thus disrupting the mitotic division of tumor cells [2]. Taxol displayed a strong activity against breast, lung, head and neck, uterine cancers, and advanced forms of Kaposi's sarcoma [3]. Taxol was firstly produced from the bark of yew trees Taxus brevifolia "family Taxaceae" [4, 5]; however, the lower yield of Taxol that being < 0.001%, i.e., to produce 1 g Taxol, it requires ~ 10 kg of plant bark that collected from 3 to 5 trees [6], is the main challenge. In addition, the vulnerability of this plant to unpredicted fluctuations with the environmental conditions strongly influences the Taxol yield, heterogeneity, and reproducibility [7–9]. Exploring the Taxol producing potency of the endophytic fungi inhabiting medicinal plants raises the hope for overcoming the low yield by the above-mentioned method [10, 11], due to their fast growth, cost-effectiveness, independence on climatic changes, and feasibility for genetic manipulation [12, 13]. Subsequently, a plethora of endophytic fungi with metabolic potency to produce Taxol has been reported as reviewed [1, 14–25]. However, the anticipation of these fungi for industrial production of Taxol has been challenged by the attenuation and loss of Taxol productivity by the fungal storage and multiple subculturing [21, 22, 26–28].</p><p>Thus, searching for a novel fungal isolate with affordable metabolic stability and sustainability for Taxol production is the challenge. Medicinal plants of well-known ethnopharmacological relevance and traditional pharmaceutical applications could be the repertoire of novel fungal isolates with unique features of metabolic stability for Taxol biosynthesis. Among the most common medicinal plants, jojoba "Simmondsia chinensis" is a monogenetic dioecious grey-green shrub belonging to Simmondsiaceae family. Jojoba seeds contain up to 65% of a light golden and odorless high-viscosity oily metabolites [29]. Jojoba oil has been frequently used for the relief of headaches, throat inflammation, and wounds treatment [30, 31]. As well as Jojoba oil has been used as anti-inflammatory and antimicrobial agents [30, 31]. The leaves of jojoba are rich with antioxidant flavonoid compounds that traditionally used for treating of various disorders such as asthma, inflammation, and cancer [32]. Thus, the main objective of this work was to explore a new fungal isolate from jojoba plant with unique metabolic stability for Taxol production, to evaluate the different approaches to maximize their Taxol yield, as well as, to enhance the antiproliferative activity of extracted Taxol compounds via conjugation with gold nanoparticles, mediated by gamma irradiation.</p><!><p>Different parts of jojoba (Simmondsia chinensis) as leaves, barks, twigs, and buds were collected from Faculty of Agriculture, Cairo University and used as the source for endophytic fungi. The plant parts were collected and washed under running tap water, surface sterilized with 70% ethanol for 1 min and then rinsed with sterile water [28]. The surface-sterilized plant parts were cut into small pieces under sterile condition and placed on plates of potato dextrose agar (PDA) medium, Czapek's-Dox, and malt extract agar medium [33–36], and the plates were incubated at 30 °C for 10 days. The effectiveness of surface sterilization of the plant parts was assessed by centrifuging the rinsing water, then 500 μl sterile water was added to the precipitate and plated into PDA medium [37]. The purified endophytic fungal isolates were inoculated on PDA slants for 7 days and stored at 4 °C.</p><!><p>The recovered endophytic fungi inhabiting jojoba were screened for Taxol production by growing on potato dextrose broth (PDB) [38]. One plug of each of the 7 days old fungal isolates was inoculated into 100 ml of PDB/250 ml Erlenmeyer flasks, incubated for 15 days at 30 ± 1 °C, under shaking conditions (120 rpm). After incubation, the cultures were filtered, and the filtrates were amended with 0.2% sodium bicarbonate to precipitate fatty acids. Taxol has been extracted with dichloromethane, and the organic phase was collected and evaporated to dryness, and the residues were re-dissolved in methanol [17, 39]. Taxol was separated and identified by TLC using Merck 1 mm (20 × 20 cm) pre-coated silica gel plates (TLC Silica gel 60 F254, Darmstadt, Germany), detected by UV illumination at 254 nm [39]. The putative spots of Taxol were scraped-off from the TLC silica gel plates and dissolved in methanol, vortexed vigorously for 10 min, and centrifuged at 1000 rpm for 5 min. The precipitated silica particles were removed, and the supernatant was taken for Taxol quantification and purity checking by HPLC (YOUNG In, Chromass, 9110 + Quaternary Pump, Korea) of C18 reverse phase column (Eclipse Plus C18 4.6 × 150 mm, 3.5 μm, Cat. # 959,963–902). The mobile phase used was methanol/acetonitrile/water (25:35:40, v/v/v) at a flow rate of 1.0 ml/min for 20 min [40], and Taxol fractions were measured at 227 nm, and their chemical identity and concentrations were confirmed from the retention time and absorption peak area comparing to authentic sample.</p><!><p>The endophytic fungal isolates were identified to their species levels based on their macro and micro-morphological features by growing on PDA, Czapek's-Dox, and malt extract media according to the reference's keys [33–36]. The identity of the most potent Taxol producing fungal isolates were further molecularly confirmed based on the sequence of internal transcribed spacer (ITS) [41, 42]. Fungal genomic DNA (gDNA) was extracted by pulverizing the mycelia (~ 0.2 g) in liquid nitrogen, then dispensing in 1 ml CTAB extraction buffer (2% CTAB, 2% PVP40, 0.2% 2-mercaptoethanol, 20 mM EDTA, 1.4 M NaCl in 100 mM Tris − HCl, pH 8.0). The PCR primer sets were ITS4 5′-GGAAGTAAAAGTCGTAACAAGG-3′ and ITS5 5′-TCCTCCGCTTATTGATATGC-3′. The PCR reaction contains 10 μl of 2 × PCR master mixture (i-Taq™, Cat. No. 25027), 2 μl of gDNA, 1 μl of each primer (10 pmol/μl), and completed to 20 μl with sterile distilled water. The PCR was programed to initial denaturation at 94 °C for 2 min, denaturation at 94 °C for 30 s, annealing at 55 °C for 10 s, extension at 72 °C for 30 s for 35 cycles, and final extension at 72 °C for 2 min. The PCR amplicons were analyzed by 1.5% agarose gel in 1 × TBE buffer (Ambion Cat# AM9864), using 1 kb DNA ladder (Cat. # PG010-55DI) and visualized by gel documentation system. The amplicons were purified and sequenced by Applied Biosystems Sequencer, HiSQV Bases, Version 6.0 with the same primers sets. The obtained sequences were BLAST searched non-redundantly on the NCBI database, imported into MEGA 6.0 software and aligned with Clustal W muscle algorithm [43] and the phylogenetic tree was constructed with neighbor-joining method of MEGA 6.0 [44].</p><!><p>The putative spots of Taxol were scraped-off from the TLC silica gel plates, purified, and the purity and concentration were determined by the UV–Vis analyses at λ 227 nm (RIGOL, Ultra-3000 Series) comparing to authentic Taxol [39]. Blank media under the same conditions were used as negative baseline for the spectrophotometric analyses. FT-IR spectrum of the purified Taxol samples was analyzed by JASCO FT-IR 3600 spectrophotometer. The Taxol sample was grinded with KBr pellets, pressed into discs under vacuum, and the absorption was measured in the region 400 to 4000 cm−1 [3], comparing to authentic one. The chemical structure of extracted Taxol was confirmed from the HNMR spectroscopy (JEOL, ECA-500II, 500 MHz NMR) comparing to authentic Taxol. The samples were dissolved in CDCl3, chemical shifts are given in ppm (δ-scale), and the coupling constants are expressed in hertz (Hz).</p><!><p>Two agar plugs (9 mm) from 7 days old cultures of each fungal isolate were inoculated in triplicate into 100 ml medium/250 ml Erlenmeyer flask of potato dextrose (PDB), Czapek's-Dox (CZD), M1D, and malt extract (ME) broth media. Uninoculated controls from each media free of fungal spores were used as negative control, incubated at 30 °C for 15 days under the same conditions. After incubation, fungal cultures were filtered, and Taxol was extracted and determined as mentioned above.</p><!><p>Optimization of the medium composition for maximizing the Taxol yield by the potent fungal isolate was conducted by response surface methodology using Placket-Burman design followed by central composite design [17–20, 45]. From the RSM designs, the positive and significant variables affecting Taxol production by the potent fungal isolate were assessed using the statistical software package by Design-Expert 7.0 (Stat Ease Inc., Minneapolis, USA). Each experiment was run in three biological replicates and the mean values were considered. After incubation at the desired conditions, fungal biomass was filtrated, and Taxol was extracted, and quantified by TLC and HPLC as described above.</p><!><p>Placket-Burman design has been frequently used for optimization of the media component for fungal growth and production of bioactive secondary metabolites, evaluating the significant variables affecting Taxol production [18, 20, 46]. Choice of factor was based on media used in qualitative and quantitative screening. Eleven factors have been included; malt extract, peptone, sucrose, soytone, glutamine, beef extract, and temperature, pH, incubation time, and shaking speed values and factors were varied over two levels, and the minimum and maximum levels ranges were selected. The statistical Design-Expert 7.0 was used to generate a set of 12 experiments. For each experiment, Taxol production was determined in three biological replicates, and the average of Taxol yield was considered. Regression analysis of the data was conducted using statistical software. The effect of each variable was calculated (Biometrika, 2020), using the following equation:\documentclass[12pt]{minimal} sepackage{amsmath} sepackage{wasysym} sepackage{amsfonts} sepackage{amssymb} sepackage{amsbsy} sepackage{mathrsfs} sepackage{upgreek} \setlength{\oddsidemargin}{-69pt} egin{document}$$E=(\sum {M}_{+}-\sum rac{{M}_{-}}{N})$$\end{document}E=(∑M+-∑M-N)</p><p>where, E is the effect of a testing variable, M+ and M− are Taxol concentration of trials at that the parameter was at its higher and lower levels respectively, and N is the number of experiments that was carried out. The effect of each variable on the production was determining by calculating their respective E-values.\documentclass[12pt]{minimal} sepackage{amsmath} sepackage{wasysym} sepackage{amsfonts} sepackage{amssymb} sepackage{amsbsy} sepackage{mathrsfs} sepackage{upgreek} \setlength{\oddsidemargin}{-69pt} egin{document}$$E=rac{{Tot}_{hi\mathrm{g}h}-{Tot}_{low}}{No}$$\end{document}E=Tothigh-TotlowNo</p><p>Where Tot high is the total responses at the high level, Tot low is the total responses at the low level, and No is the number of trials.</p><!><p>The most significant positive factors affecting Taxol production by the selected fungal isolate were optimized using a response surface type CCD model experimental design [47]. By using CCD, the concentrations of the medium components were optimized, and their studied interactions were used to generate a total of 20 experiments for the three variables.</p><p>To determine the optimal levels of the variables for Taxol production from the potent fungal isolate, three-dimensional (3D) response surface curves were plotted to study the interaction between the various factors, and to determine the variable condition of each factor affecting Taxol production. The 3D graphs were carried out by holding three factors' constants in an ideal level and plotting the obtained response of Taxol yield for varying levels of the other two factors.</p><!><p>The potent endophytic isolates producing Taxol were exposed to γ-irradiation with 60Cobalt source (Gamma cell 4000-A-India) at different gamma radiation doses (0.25–3.0 kGy) compared to the non-irradiated cultures control; a dose rate 1.2 kGy/h at the time of experiments. The optimized media were inoculated by the irradiated cultural under standard cultural conditions, compared to the non-irradiated spore's inoculum as control. The cultures were incubated at 30 ± 2 °C for 15 days on a rotary shaker (120 rpm). After incubation, the cultures were filtered and Taxol was extracted, purified, and quantified by TLC and HPLC as described above.</p><!><p>Polyvinylpyrrolidone (PVP)-capped gold nanoparticles (AuNPs) were synthesized by mixing 1 mM PVP (dissolved in distilled water) with 0.5 mM gold (III) chloride hydrate magnetically stirred, and the solution was irradiated by gamma rays at different doses (0.25–10.0 kGy). The obtained PVP-Au3+ solution has been amended with 1 ml sodium borohydride (1 mM) as reducing agent. Taxol (100 µg/ml) was mixed with PVP-AuNPS at ratio 1:2 (v/v), and the obtained Taxol-PVP-AuNPs conjugate was characterized by the UV–Vis analysis.</p><p>The size distribution and average particle size of the Taxol-PVP-AuNPs conjugate was measured by dynamic light scattering (DLS) (PSS-NICOMP 380-ZLS particle sizing system St. Barbara, CA, USA, at NCRRT). FTIR measurements were carried out to obtain information about chemical groups of the Taxol-PVP-AuNP conjugates in relation to their structural stability, comparing to the native Taxol (JASCO FT-IR 3600 infra-red spectrometer). The size and morphology of the synthesized AuNPs were recorded by using high-resolution transmission electron microscope (HRTEM), and drop coating AuNPs prepared TEM studies onto carbon-coated TEM grids. The X-ray diffraction (XRD) patterns were obtained with the XRD-6000 series, including residual austenite quantitation, stress analysis, crystallinity calculation, and crystallite size/lattice strain materials analysis by overlaying X-ray diffraction patterns (Shimadzu apparatus with Cu-Kα target, and nickel filter Shimadzu Scientific Instruments (SSI), NCRRT).</p><!><p>The activity of the purified Taxol and Taxol-PVP-AuNPs conjugates against liver carcinoma (HPG2), and breast carcinoma (MCF7) was determined by 3- (4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide (MTT) assay [48]. The 96-well plate was seeded with 103 cells per well, incubated overnight at 37 °C, then different concentrations of the drug were added, and the plates were re-incubated for 48 h. The MTT reagent (25 μl) was added, incubated for 2 h, and the purple color of the developed formazan complex was measured at λ570 nm. The IC50 value was expressed by amount of drug reducing the growth of 50% of initial number of tumor cells normalizing to positive control.</p><!><p>The antimicrobial activity of Taxol and Taxol-AuNPs conjugates was assessed against different bacterial isolates; Bacillus subtilis ATCC 6633 and Staphylococcus epidermidis, Pseudomonas aeruginosa, Escherichia coli, and Enterobacter agglomerans, in addition to Candida albicans. The tested bacterial cells were suspended in sterile peptone water to obtained standard inoculum of ~ 0.5 McFarland (1–1.5) × 108 CFU/ ml at λ600 nm. The growth inhibition (mm) of microbial pathogens growth was assessed by agar disc diffusion method. Sterile standard antibiotic disks with the diameter of 6.0 mm were used as positive controls. Sterile antibiotic discs (6.0 mm) were loaded with 20 μl of methanol and amoxicillin clavulanic acid (AMC) as negative and positive control. Discs were loaded with the same concentration of Taxol, Taxol-PVP-AuNPs, and AuNPs (1.0 μg/ml). Three biological replicates were prepared. The plates were incubated at 37 °C for 24 h, and the zones of inhibition were measured. Amoxicillin clavulanic acid (AMC) and nystatin were used to normalize the antimicrobial activity of Taxol. The inhibition zone of growth was determined by a vernier caliper (mm).</p><!><p>The experiments were conducted in three biological replicates, and the results were expressed by mean ± STDV. The significance was calculated by one-way ANOVA with Fisher's least significant difference of post hoc test (https://www.easycalculation.com/statistics/fishers-lsd-calculator.php).</p><!><p>The isolate A. flavus Bd was deposited at genbank under accession #MW485934.1 as well as at Assiut University Mycological Center (AUMC), Egypt, with deposition #AUMC13892.</p><!><p>Screening for Taxol producing endophytic fungi of jojoba</p><p>A Morphological views of jojoba plant. B Plate cultures of the potent Taxol producing endophytic fungi; A. flavus Bd1 (13), A. niger Lv1 (21), Penicillium polonicum (23), and A. oryzae Bd (25) on PDA after 8 days o incubation at 30 °C. The fungal isolates were grown on PDB, incubated at the standard conditions, and Taxol was extracted and checked by TLC (C). D HPLC chromatogram of Taxol from the potent fungal isolates. E Yield of Taxol as quantified from HPLC. F, UV–Vis spectral analysis of extracted Taxol from the fungal isolates. G FT-IR analysis of extracted Taxol comparing to authentic one</p><p>A Macromorphological features of A. flavus an endophyte of jojoba after 3, 5, and 8 days of growth on PDA. Micro-morphological features, conidial head of A. flavus by 400X magnification. C PCR amplicon of A. flavus ITS region of 500 bp, normalizing to 1 kb ladder (Cat.#. SM0312). D Phylogenetic analysis of ITS A. flavus by maximum likelihood method [44]</p><p>Chemical analysis of extracted Taxol from A. flavus. Taxol was extracted, fractionated by TLC. The putative spots of Taxol were scraped-off from the TLC plates and analyzed. HPLC chromatogram of authentic Taxol (A), A. flavus Taxol (B). C Spectra of 1HNMR of A. flavus Taxol sample. D FT-IR spectra of extracted Taxol of A. flavus</p><p>Nutritional optimization of Taxol production from A. flavus using Plackett–Burman design. A and B Pareto charts showing the effect of individual factors on Taxol production (A malt extract, B peptone, C sucrose, D soytone, E cysteine, F glutamine, G beef extract, H temperature, J pH, K incubation time, and L shaking speed. C Normal probability plots of the variables for Taxol production by A. flavus from the first order polynomial equation. D Plot of correlation between predicted and actual Taxol yield by A. flavus. Three-dimensional response surface curves showing the effect of interactions of pH and cysteine (E), incubation time and cysteine (F), and pH and incubation time (G)</p><p>Minimum and maximum ranges of the parameters selected in P-BD for optimization of Taxol production</p><p>Analysis of variance (ANOVA) for Taxol production concentration</p><p>Plackett–Burman design to evaluate factors affecting Taxol production by A. flavus</p><p>FCCD optimization design for the significant variables affecting Taxol production by A. flavus</p><p>ANOVA analysis for response surface linear model for the experiments with CCD</p><!><p>The effect of gamma irradiation on A. flavus and on their Taxol yield was estimated by irradiating the fungal spores at different doses of γ-rays, then growing the spores on the modified malt extract medium obtained from the RSM by Plackett–Burman design. After cultural incubation under standard conditions, Taxol was extracted and quantified by TLC and HPLC. From the obtained results (Fig. S1), there is no significant effect on Taxol productivity upon γ-irradiation at the different doses.</p><!><p>Physicochemical properties AuNPs-Taxol conjugates in response to different doses of γ-irradiation. A Scheme of Taxol conjugation with polyvinylrrolidone-gold nanoparticles, mediated by γ-irradiation at 1.0 kGy. B High resolution transmission electron microscope (HRTEM) of AuNPs. C Dynamic light scattering (DLS) pattern of the synthesized AuNPs. D UV–Vis spectrum of AuNPs at different doses of γ-irradiation. E FTIR spectra of Taxol and AuNPs-Taxol conjugates at 1.0 kGy. F X-ray diffraction (XRD) analysis of AuNPs and Taxol-AuNPs conjugates at 1.0 kG</p><!><p>The chemical conjugation of Taxol and PVP-AuNPs was checked from the FT-IR spectral analysis, as revealed from the slightly shift of the intensity and transmission ratio of the functional groups of Taxol-PVP-AuNPs conjugates comparing to Taxol as control. The intensity of the peak at 3393.3 cm−1 assigned for the hydroxyl (OH) in Taxol-PVP-AuNPs consortium was increased by about 3 folds comparing native Taxol, revealing the chemical stretch on the hydroxyl groups (Fig. 5). As well as the intensity of the peaks 1661.0 cm−1 referring to C = O stretching frequency was strongly increased by about 10 folds upon conjugation with PVP-AuNPs, ensuring the stretching on the C = O bonds. The intensity of the peak 2923.5 corresponding to aliphatic CH stretching was quite stable. A slight shifting on the intensity of observed peaks at 1452.0 cm−1 and 1404.0 cm−1 has been observed due to the NH stretching frequency. The intensity of the peak 1109.0 cm−1 of carbonyl group-oxygen stretching frequency was slightly reduced upon conjugation with PVP-AuNPs.</p><p>The crystal/physical structure of the Taxol upon conjugation with PVP-AuNPs were resolved from the XRD analysis (Fig. 5). From the XRD, the Taxol-PVP-AuNPs consortium had the same structural configuration, crystal orientation, and size of the native-Taxol, ensuring the lack of negative effect on the crystal structure of native Taxol. From the results, the amorphous and crystal structure of the Taxol-PVP-AuNPs comparing to the native Taxol has been confirmed. From the XRD analysis, conjugation of Taxol and PVP-AuNPs has been resolved as revealed from the diffraction properties concerning 2ɵ = 38.18°, 44.01°, 64.57°, 77.67°, and 81.74° which described the Bragg's observations at (111), (200), (220), (311), and (222), respectively. All the peaks were related to the ideal card of Joint Committee on Powder Diffraction Standards (JCPDS) of AuNPs (JCPDS card No. 04–0784) [51]. So, the strength of crystals of the synthesized AuNPs was shown, providing the face-centered cubic (fcc) crystalline structure. In addition, there is a simply amorphous peak at 19.25° for Taxol that is included in the organization and permanence of AuNPs. The XRD pattern confirmed the successful conjugation of Taxol and PVP-AuNPs. The mean crystallite size of the incorporated AuNPs was determined from the Scherrer's equation [52], and it was 20.2 nm for Taxol-PVP-AuNPs consortium mediated by gamma rays as mentioned in equation:\documentclass[12pt]{minimal} sepackage{amsmath} sepackage{wasysym} sepackage{amsfonts} sepackage{amssymb} sepackage{amsbsy} sepackage{mathrsfs} sepackage{upgreek} \setlength{\oddsidemargin}{-69pt} egin{document}$$\mathrm{D}=rac{k\lambda }{eta \mathrm{cos} heta }$$\end{document}D=kλβcosθ</p><p>where D is the average crystallite size, β is the full-width at half maximum, λ is the X-ray wavelength, and θ is the Bragg's angle, and k is a constant.</p><!><p>Antimicrobial and anticancer activities of the purified Taxol samples and Taxol-AuNPs conjugates. A Antimicrobial of native and Taxol-AuNPs conjugates towards E. coli, E. agglomerans, and C. albicans s activity, as revealed from the inhibition zones (A is the authentic Taxol, B is the A. flavus Taxol, C is the Taxol-AuNPs conjugates, D methanol, and E is the amoxicillin calvulanic acid). B The IC50 values of the authentic Taxol, native A. flavus Taxol, and Taxol-AuNPs conjugates. C Antiproliferative activity of native Taxol and Taxol-AuNPs conjugates towards HEPG-2 and MCF-7 cells as revealed from the cell viabilities. D The IC50 values of Taxol conjugates against the HEPG-2 and MCF-7 normalizing to authentic Taxol and AuNPs</p><p>Antimicrobial activity of Taxol against different pathogenic microorganisms represented by diameter of inhibition zone (mm)</p><p>The concentration of authentic Taxol and A. flavus Taxol was 20 µM</p><!><p>Fungal endophytes with Taxol producing potency raised the hope for mass production of Taxol due to their fast growth, cost effective fermentation process, independence on climatic changes, and feasibility of genetic manipulation. However, the anticipation of fungi for industrial production of Taxol has been challenged by their lower reproducible yield and loss of Taxol productivity with the subculturing [14, 16, 26, 39, 45, 53, 54]. Most of the endophytic Taxol-producing fungi were isolated from Taxus sp. and Podocarpus sp. which are belonging to family Taxaceae [18, 39]. Exploring of the Taxol producing by endophytic fungi from plants outside Taxaceae family with probable Taxol productivity is the main objective by biotechnologists. Jojoba plant is one of the most traditionally recognized medicinal plants for its ethnopharmacological relevance such as antimicrobial activity, anti-inflammatory, and anticancer activities [24, 55]. Thus, the objective of this study was to isolate and estimate the Taxol producing potency of the endophytic fungi from the jojoba plant. Among the recovered fungi, the endophytic isolate A. flavus gave the highest Taxol yield (88.65 μg/l), as authenticated from the TLC, UV- absorption, and HPLC analysis. Similar screening paradigm for Taxol production has been reported for endophytes from Ginko biloba [28] and other plants [15, 18, 20, 39, 56]. The Taxol yield by A. flavus has been agreed with P. polonicum, an endophytes of Ginkgo biloba [28] and A. flavipes [15, 39], A. terreus [15, 39], an endophytes of P. gracilior, as well as for endophytes of Taxus spp. such as A. candidus, Fusarium solani [57], A. niger [58], and A. fumigatus. The identity of A. flavus, the potent Taxol producing fungal isolate, was confirmed from the molecular analysis of the ITS region, and the sequence was deposited on Genbank with accession #MW485934.1, as well as at Assiut University Mycological Center (AUMC), Egypt with deposition # AUMC13892. Similarly, the identity of A. flavus was confirmed based on the sequence of ITS region [49, 59], 2018, 2019, [1, 11, 36, 43, 60–68]. The chemical structure of extracted Taxol from A. flavus was 1HNMR, and FT-IR analyses. The resolved signals of HNMR for A. flavus Taxol were identical to the standard Taxol, which distributed between 1.0 and 8.0 ppm. Three proton signals were resolved at 1.0–3.0 ppm corresponding to methyl, acetate, and acetylene groups, whereas the signals for aromatic moieties were resolved at 6.5–9.0 ppm [17, 45]. Consistently, for all Taxane scaffolds, signals for their side chains protons were resolved at 2.0–7.0 ppm, while those for benzoate (C2), phenyl (C3), and benzamide (C3) groups were resolved at 7.0 and 8.4 ppm [69] ( [70]. The FT-IR spectra of A. flavus Taxol were like authentic Taxol, as coincident with Taxol from other fungal isolates [3], Visalakchi et al. 2010). The yield of Taxol from A. flavus was optimized by response surface methodology using Plackett–Burman Design [17–20, 45, 46, 49, 71]. The yield of Taxol was increased by about 1.8 folds upon on response to factorial design nutritional optimization process [17, 45]. The highly significant variables affecting Taxol production by A. flavus was further optimized with the CCD design, giving the maximum yield (302.72 μg/l) with cysteine (0.5 g/L) at pH 6.0 and incubated for 7 days. In an endeavor to enhance the production of Taxol by A. flavus, the fungal spores were exposed to ionizing γ- radiation, and the spores were grown on the modified malt extract broth medium, and Taxol was extracted and quantified. Upon γ-rays irradiation, the Taxol yield by A. flavus was not significantly increased comparing to the control cultures, suggesting the lack of induction of biosynthetic gene cluster of Taxol. However, the Taxol yield was slightly increased by Fusarium maire and Nodulisporium sylviforme in response to gamma-irradiation [1, 72], El-Sayed et al., 2020). Taxol productivity by A. flavus in response to different media has been studied. A. flavus gave the highest Taxol yield by growing on Czapek's-Dox media and malt extract, as consistent with results for Taxol production by A. terreus [16, 53] and P. polonicum [28]. However, preference of PDB for Taxol production was reported for A. candidus and F. solani [73].</p><p>To increase the targetability, solubility, and efficacy of Taxol to bind with the target protein in vivo, several successful trails have been motivated. Conjugation of AuNPs with the less soluble chemotherapeutic drugs is one of the recent magnificent technologies to increase the solubility and targetability of several chemotherapeutic drugs. The purified Taxol from A. flavus was conjugated with AuNPs, in the presence of PVP as capping and stabilizing agent, mediated by irradiation by γ-rays. The AuNP was prepared by γ-irradiation-based reduction in the presence of PVP as stabilizing agents to prevent the metal colloids from the rapid aggregation, as revealed from the color change and a strong absorption band at λ540 nm. The color change is usually attributed to the surface plasmon resonance (SPR) [74], and the sharp and high intensity peaks ensure the highest yield and size distribution of the synthesized AuNPs. Gamma-irradiation has been authenticated as one of the most desirable methods for synthesis of metallic nanoparticles due to their highly reducing radicals and generating free electron without byproduct formation [50]. The particle size obtained from DLS measurements 47.58 nm was larger than the TEM results (13.0–21.0 nm) because DLS analysis measures the hydrodynamic radius. From the XRD analysis, the development of crystal structure of Taxol-PVP-AuNPs consortium has been as revealed from the diffraction peaks at 2ɵ = 38.18°, 44.01°, 64.57°, 77.67°, and 81.74° which described the Bragg's observations at (111), (200), (220), (311), and (222), respectively [51]. The synthesized AuNPs showed crystal strength and provided the face-centered cubic (fcc) crystalline structure, with one amorphous peak at 19.25° for Taxol that is included in the organization and permanence of AuNPs [52].</p><p>The antiproliferative activity of Taxol-PVP-AuNPs conjugates was assessed against HEPG-2 and MCF-7 cell lines, normalizing to native Taxol and AuNPs, separately. The bioactivity of Taxol was dramatically increased upon conjugation with AuNPs, comparing to native Taxol as control.</p><p>The enhanced anticancer activity of Taxol by AuNPs could be due to the decreasing on the lymph drainage, increasing the blood circulation and Taxol solubility upon conjugation with AuNPs, and comparing to native Taxol [75]. AuNPs receive a great attention due to their low toxicity, higher versatility to bind to with different molecules, and biocompatibility [76]. From the IC50 values, Taxol-PVP-AuNPs consortium displayed the significant activity against HEPG-2 and MCF-7, comparing to native Taxol and AuNPs separately. Regarding to the bioactivity of AuNPs, the antiproliferative activity of Taxol-PVP-AuNPs consortium was increased by about 4 folds comparing to AuNPs as control. Interestingly, the activity of Taxol-AuNPs consortium was plausibly consistent with the authentic Taxol. The antimicrobial activity of A. flavus Taxol-AuNPs consortium was evaluated against various pathogenic multidrug resistant microorganisms. Upon conjugation with AuNPs, the antimicrobial activity of Taxol was strongly increased comparing to native Taxol and AuNPs, separately. Taxol-AuNPS consortium displayed the highest activity against E. agglomeranus, C. albicans, and E. coli. The higher activity of Taxol upon conjugation with AuNPs authenticates the bioavailability, solubility, and efficacy of Taxol to bind with the target tubulin protein in vivo.</p><p>In conclusion, Aspergillus flavus MW485934.1, an endophyte of jojoba, has been isolated and identified as potent Taxol producer, based on the metabolic and chromatographic analysis. The chemical identity of extracted Taxol from A. flavus was verified from the TLC, HPLC, NMR, and FTIR analyses. The yield of Taxol by A. flavus was maximized by the response surface methodology with the Plackett–Burman and faced central composite designs. Upon using factorial designs RSM, the yield of Taxol by A. flavus was increased about 3.2 folds (302.7 µg/l), comparing to control cultures (96.5 µg/l). In addition, conjugates of Taxol-gold nanoparticles (AuNPs) mediated by γ-rays were prepared. The physical and spectroscopic properties of the Taxol-AuNPs conjugates were determined by UV–Vis, dynamic light scattering (DLS), X-ray diffractometer (XRD), and transmission electron microscope (TEM) analyses. With AuNPs conjugation, the anticancer activity towards different tumor cell lines was dramatically increased. As well as the antimicrobial activity of Taxol-AuNPs conjugates towards the different multidrug resistant bacteria was strongly increased comparing to native Taxol compounds.</p><!><p>Supplementary file1 (TIF 61 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

IR-Monitored Photolysis of CO-Inhibited Nitrogenase: A Major EPR-Silent Species with Coupled Terminal CO Ligands

We have used Fourier transform infrared spectroscopy (FT-IR) to observe the photolysis and recombination of a novel EPR-silent CO-inhibited form of \xce\xb1-H195Q nitrogenase from Azotobacter vinelandii. Photolysis at 4 K yields a strong negative IR difference band at 1938 cm\xe2\x88\x921, along with a weaker negative feature at 1911 cm\xe2\x88\x921. These bands and the associated chemical species have both been assigned the label \xe2\x80\x98Hi-3\xe2\x80\x99. A positive band at 1921 cm\xe2\x88\x921 is assigned to the \xe2\x80\x98Lo-3\xe2\x80\x99 photoproduct. By using an isotopic mixture of 12C16O and 13C18O, we show that the Hi-3 bands arise from coupling of two similar CO oscillators with one uncoupled frequency at ~1917 cm\xe2\x88\x921. Although in previous studies Lo-3 was not observed to recombine, by extending the observation range to 200\xe2\x80\x93240 K we found that recombination to Hi-3 does indeed occur, with an activation energy of ~6.5 kJ mol\xe2\x88\x921. The frequencies of the Hi-3 bands suggest terminal CO ligation. We tested this hypothesis with DFT calculations on models with terminal CO ligands on Fe2 and Fe6 of the FeMo-cofactor. An S = 0 model with both CO ligands in exo positions predicts symmetric and asymmetric stretches at 1938 and 1909 cm\xe2\x88\x921 respectively, with relative band intensities of ~3.5:1, in good agreement with experiment. From the observed IR intensities, we find that Hi-3 is present at a concentration about equal to that of the EPR-active Hi-1 species. The relevance of Hi-3 to the nitrogenase catalytic mechanism and its recently discovered Fischer-Tropsch chemistry is discussed.

ir-monitored_photolysis_of_co-inhibited_nitrogenase:_a_major_epr-silent_species_with_coupled_termina

Introduction<!>Photolysis<!>Recombination<!>Mixed Isotope Studies<!>Wavelength-Dependent Photolysis<!>DFT Calculations<!>Discussion<!>Sample Preparation<!>Photolysis and FT-IR spectroscopy<!>DFT

<p>The enzymatic nitrogen fixation system known as nitrogenase (N2ase) catalyzes the reduction of dinitrogen to ammonia at ambient temperatures and atmospheric pressure.[1] For the Mo-dependent N2ase, binding of substrates and inhibitors involves a [Mo-7Fe-9S-X]-homocitrate-containing prosthetic group, called the FeMo-cofactor, of the larger MoFe protein (Av1). A recent crystal structure at 1.0 Å resolution[2] and an X-ray emission study[3] favor assigning C4− carbide as the much-debated atom X at the center of the cofactor. The mechanism by which this biological nanoparticle lowers the activation energy for reduction of the N≡N triple bond remains incompletely understood.</p><p>Carbon monoxide inhibits the reduction of N2 reversibly and non-competitively, and CO has been an important probe molecule for characterizing the active site.[4] Recently, Ribbe and coworkers have shown that CO can actually be a substrate for the V version of N2ase,[5] and with less efficiency, even for the wild-type Mo-dependent N2ase.[6] Seefeldt and coworkers have also demonstrated CO fixation, using the α-V70A and other variants of the MoFe protein.[7] Thus, a second front of interest in N2ase as a Fischer-Tropsch catalyst has emerged.[8]</p><p>Exposure of N2ase to CO during turnover elicits species with a variety of EPR signals, depending on the partial pressure of CO ([CO]), and each of these species and its characteristic EPR signal is described by the [CO] required for its formation.[9] The chemical species and its associated axial EPR signal (g-values near 2.17 and 2.06) are both called hi-CO and are formed under ~101 kPa [CO]. This hi-CO N2ase species is proposed to contain two CO molecules bound to the FeMo-cofactor.[9–10] A species with a rhombic EPR spectrum (g-values of 2.09, 1.98, and 1.93) called lo-CO forms under a much lower [CO][11] and is proposed to contain only one bound CO.[9–10] A third species and EPR signal, both called hi(5)-CO, have also been identified.[11c, 11d] Not all of these EPR signals are generated when variant Mo N2ases are turned over under CO,[12] and the integrated EPR intensities rarely exceed 40% of the active site species.[11d] Presumably, EPR-silent species with bound CO must co-exist with the EPR-active species.</p><p>Another spectroscopic tool that complements EPR for these studies has been FT-IR spectroscopy.[4c, 13] In stopped flow FT-IR experiments, under low [CO], a single ν(CO) band appeared at 1904 cm−1 and peaked in intensity within ~7s before decaying. Under high [CO], a new ν(CO) band was observed at 1936 cm−1 together with a pair of weaker, possibly coupled bands at 1958 and 1880 cm−1. All of these features were much longer lived than the 1904 cm−1 band. All three species were suggested to arise from terminally bound CO. After relatively long times (>10s), under low [CO] the 1904 cm−1 peak is succeeded by a new ν(CO) band at 1715 cm−1. Recently, the rates and relative intensities of these ν(CO) bands were shown to be sensitive to variation of the side-chain at the α-70 position.[13c] Other CO-related IR bands have been observed during spectro-electrochemical studies of FeMoco (the solvent-extracted version of the FeMo-cofactor),[14] where bands at 1808 and 1835 cm−1 were proposed to arise from bridging CO, whereas features at 1885 and 1920 cm−1 were assigned to terminally bound CO species.</p><p>In a recent study of CO-inhibited Azotobacter vinelandii Mo nitrogenase, we showed that FT-IR can be used to monitor CO photolysis by visible light at cryogenic temperatures.[15] Three distinct types of photolyzable CO complexes were found under hi-CO conditions. We labeled these stable inhibited forms 'Hi-1', 'Hi-2', and 'Hi-3'. The photolyses of Hi-1 and Hi-2 were found to be reversible at around 80 K, with activation energies on the order of 3–4 kJ mol−1. However, the Hi-3 photoproduct, labeled 'Lo-3', was stable with respect to recombination up to 110 K. The Hi-3 species was most abundant in the photolysis spectra of N2ase with the variant α-H195Q MoFe protein. In the current work, we have used this variant and an extended temperature range (~200–250 K) to discover that Hi-3 photolysis is also reversible. We have used mixed CO isotopes to demonstrate vibrational coupling in the Hi-3 species, and wavelength-dependent photolysis to determine the spectral region with the greatest photochemical activity. We have also estimated the relative amounts of the Hi-1 and Hi-3 species in our samples. Finally, we have employed DFT calculations to evaluate plausible structures for Hi-3 and Lo-3 CO-bound FeMo-cofactor species.</p><!><p>Figure 1 shows the time courses of changes in the IR spectra of CO-inhibited α-H195Q N2ase samples illuminated with visible light at ~4 K under either 12C16O or 13C18O. Photolysis spectra under 13C16O are included in Figure 2. At long times, the largest spectral changes come from the Hi-3 species,[15] with a strong negative band at 1938 (1894, 1849) cm−1 and a weaker negative band at 1911 (1867, 1824) cm−1. A positive product band from Lo-3 is seen in between these two bands at 1921 (1877, 1833) cm−1. (In these and later descriptions, the first value refers to results with 12C16O whereas the second and third values in parentheses, if present, refer to results with 13C16O and 13C18O, respectively.) As noted before[15], for α-H195Q N2ase the spectra are complicated by the presence of other photolyzable species, including bands for Hi-1 at 1969 (1925, 1879) cm−1 and for Hi-2 at 1932 (1888, 1844) cm−1.</p><p>The spectra can be simplified by employing the fact that the photolysis product Lo-3 is relatively stable at temperatures where the photolysis products from Hi-1 and Hi-2 rapidly recombine.[15] Accordingly, we warmed the photolyzed samples to 120 K for 10–20 min and then photolyzed a second time. As shown in Figure 1, there was almost no trace of the Hi-3 or Lo-3 features in the second photolysis spectrum, consistent with the absence of Hi-3. By taking the difference between the two photolysis results, we obtained a relatively pure Hi-3→Lo-3 photolysis signal in the double difference spectrum (Figure 2). The extracted spectrum exhibited almost pure Hi-3 features at the same frequencies as noted in the mixed spectra. Note that in the third photolysis spectrum, which followed warming the sample from the second photolysis to above 230 K, the Hi-3→Lo-3 signals reappeared.</p><p>An alternative method of cleanly separating the Hi-3 features is to photolyze at higher temperatures, where the recombination of the other photolysis products is fast. As shown in Figure 2, photolysis of a 13C16O sample at 240 K yielded a spectrum with a pair of negative bands at 1891 and 1864 cm−1; nearly the same as the frequencies (1894 and 1867 cm−1) seen in the low-temperature photolysis. In fact, the 3 cm−1 downshifts are similar to those seen in myoglobin-CO upon raising the temperature from 4 K to room temperature.[16] This observation suggests that we should anticipate comparable shifts between our liquid helium-temperature photolysis results and the room-temperature stopped-flow FT-IR studies. In these higher temperature photolyses, there was less evidence of the Lo-3 photolysis product. We cannot exclude the possibility that additional processes involving CO migration and rearrangement and/or possible Lo-3 photolysis might be occurring at higher temperatures; these issues call for additional study.</p><!><p>As seen in Figure 2, Hi-3 returns on the 1h time scale upon warming the sample to ~210 K. From monitoring this process at various temperatures, Arrhenius plot analysis using the initial rates yields an activation energy of ~6.5 kJ mol−1 (Figure 2), compared to the ~4 kJ mol−1 values previously seen for Hi-1 and Hi-2 recombination.[15] (In more recent studies we have found that with brief irradiation, the Hi-2 recombination can occur with a much smaller activation energy).[17] The Hi-3 value is similar to the barriers of 8–10 kJ mol−1 seen for Ni-SIa→NiSCO CO recombination in [NiFe] hydrogenases[18] and for photolyzed MbCO.[19]</p><!><p>The relative intensities of the negative Hi-3→Lo-3 features involve a higher 1938 (1894, 1849) cm−1 frequency band that is about four-times stronger than its lower 1911 (1867, 1824) cm−1 frequency partner. One explanation for the different intensities of the high and low frequency bands is that they represent the symmetric (with Isym intensity) and antisymmetric (Iasym) combinations of stretches for two distinct terminally bound CO molecules. Assuming that the symmetric combination has the higher frequency, the equation Isym/Iasym = cot2θ where θ is half the angle between the CO molecular axes, can be used to calculate a rough approximation of the relative orientations of the two molecules[20]. For an intensity ratio of ~4, the angle 2θ between CO axes is predicted to be ~50°. Such an angle is too small for two CO molecules bound at the same metal ion. Furthermore, coupling of vicinal 12C16O molecules usually produces splittings of about ~40–60 cm−1,[20] approximately twice as large as that observed for Hi-3. Both observations support the notion that there are two vibrationally coupled CO ligands, each of which is bound to a different FeMo-cofactor metal ion.</p><p>To confirm the assignment of the Hi-3 signals to coupled oscillators, we recorded photolysis spectra for α-H195Q N2ase inhibited by a mixture of ~25% 12C16O and ~75% 13C18O. In the absence of coupling, such a sample would produce a spectrum that is the 25:75 = 1:3 weighted average of the two pure 100% 12C16O and 100% 13C18O spectra, respectively. However, as seen in Figure 3, the mixed-CO spectrum has a new negative band at 1917 cm−1. This band does not appear in the 2nd photolysis spectrum after recombination at 120 K (see Figure 1 for the reference), so we can rule out assigning it to either Hi-1 or Hi-2. As expected for a Hi-3 assignment, the 1917 cm−1 band does reappear in the 3rd time photolysis, after recombination overnight at 233 K. We interpret this new band as resulting from photolysis of a Hi-3 species involving uncoupled CO oscillators, one with 12C16O and the other with 13C18O. Thus, at least one of the CO ligands is predicted to have an uncoupled frequency of 1917 cm−1.</p><!><p>So far, we have identified three forms of CO-inhibited N2ase that are produced under high [CO] conditions and are photolyzed by exposure to white light. To better distinguish these species, we investigated the wavelength dependence of the photolysis rates, and the results of these studies are shown in Figure 3. The Hi-3 photolysis rate rises from near zero at 600 nm to maximum efficiency at 340 nm, the shortest wavelength investigated. There appear to be minor features at ~550 and ~400 nm. By comparison, the Hi-1 species, which we associate with the hi-CO EPR species previously examined by Maskos and Hales, exhibits a structureless rise in photolysis efficiency towards shorter wavelengths. For Hi-1, we did not see the minor peak in photolysis yield at ~550 nm that was observed in the previous EPR experiments.[21]</p><!><p>To put speculation about the structure of the Hi-3 species on firmer ground, we used DFT calculations to investigate models that involve pairs of terminal CO ligands on different FeMo-cofactor metal sites. With 8 metal ions, there are 8(8−1)/2 = 28 possible pairs of binding sites on this prosthetic group. For the current study, we limited the calculations to binding at the most commonly proposed sites, the Fe2 and Fe6 ions on the face adjacent to the α-V70 residue.[13c, 22], see Figure 4. The atom and residue numbering here are as in the crystal structure.[23].As done in a recent study by Dance,[24] we considered exo (CO trans to the central X atom, now known to be carbon) and endo (CO trans to μ3 sulfurs, S1A/S1B when coordinated at Fe2/Fe6, respectively) terminal binding modes, which are subsequently referred as "Fe2/Fe6 exo/endo" geometries. A less popular bridging binding mode,[25] where the CO carbon occupies a position similar to that of μ2 S2B sulfur, was also considered. Below, we will discuss mostly those CO coordinations that were found relevant in context of the Hi-3 and Lo-3 photolytic species; a detailed description of the binding alternatives will be covered separately.</p><p>Our favored model for Hi-3 is shown in Figure 4. In the DFT calculation we assumed an EPR silent state for the transition metal ion core of the FeMo-cofactor, 1e− reduced from the S = 3/2 resting state. Other details on the metal oxidation states and spin coupling are given in the Discussion section. In line with earlier modelling by us and many others,[26] the central ligand X was initially assigned as N3−. Subsequently, as described below, we considered the X = C4− alternative which has recently received significant support.[2–3, 27] The μ2 S2B sulfur that bridges Fe2 and Fe6 and forms the hydrogen bond to α-H195 in the wild-type enzyme was chosen as the protonation site for all the models. Terminal CO ligands are at Fe2 and Fe6, both in the exo geometry, and the Fe-C≡O bond angles are essentially 180° (Figure 4). This model predicts two coupled 12C16O stretching modes, a symmetric stretch at 1938 cm−1 and an asymmetric stretch at 1909 cm−1, with relative band intensities of ~3.5:1, in good agreement with the experimental frequencies of 1938 and 1911 cm−1 and amplitude ratio of ~3.7:1. (Here and below, the broadened peak center heights are compared relative to the FT-IR, see Figure 5, in contrast to the raw 'stick' DFT mode intensities in Table S2 of the Supporting Information.) The optimized angle between the bound CO molecular axes is 37°, in line with the above ~50° rough estimate based on the relative FT-IR Hi-3 intensities. The basic FeMo-cofactor framework remained intact. However, we note a significant lengthening of the Fe6-X bond to 2.17 Å, compared to an average of 2.00 Å for the remaining optimized central Fe-X distances. The plasticity of the FeMo-cofactor core has been observed in other calculations; where even a complete loss of Fe coordination to X (Fe-X > 3.0 Å) upon ligand binding has been predicted.[24, 26c]</p><p>For the Lo-3 photolysis product, the structure with the best match to the FT-IR Lo-3 band at 1921 cm−1 involved Fe2 exo binding (Figure 4), producing a 12C16O frequency at 1923 cm−1. The calculated band intensity of this ν(CO) stretch is ~2.1:1 relative to the asymmetric Hi-3 mode at 1909 cm−1, in reasonable agreement with the ~1.7:1 ratio seen in the experiment (Figure 2). For photolysis conducted using the pure 13C18O isotope, the DFT results are essentially of the same quality as those described above for 12C16O.</p><p>The calculations also shed light on our mixed isotope experiments. For models with 12C16O at Fe2 exo or Fe6 exo positions (and the other Fe site populated by 13C18O), they predict essentially 'uncoupled' pairs of ν(CO) frequencies of 1934/1825 or 1916/1842 cm−1 respectively, see the Supporting Information for the animated vibrational modes. We also modeled the actual Hi-3→Lo-3 photolysis IR spectra for 100% 12C16O, 100% 13C18O, and mixed isotope ~25% 12C16O / ~75% 13C18O experiments using our DFT frequency and intensity predictions, see Table S1 in the Supporting Material for the details. Figure 5 shows that in the low frequency 1810–1860 cm−1 region, the mixed isotope spectrum is dominated by the 13C18O/13C18O contributions, and the decoupled 13C18O bands are obscured by these features. However, in the higher 1900–1950 cm−1 frequency region, the mixed isotope species provides most of the intensity, and the DFT peaks at 1934 and 1916 cm−1 show noticeable shifts (of −4 and +7 cm−1) with respect to the pure 12C16O isotope Hi-3 features. The FT-IR bands seen at 1934 and 1917 cm−1 can now be assigned to essentially uncoupled stretching modes of the 12C16O at Fe2 and Fe6 respectively, when 13C18O binds at the other Fe site. The kinetic energy distribution (KED) factors in Table S2 of the Supporting Material indicate that less than 2% energy is accumulated in 13C18O stretch in these modes, and 12C16O stretch accounts for the rest.</p><p>As mentioned above, we also considered models with the same CO binding modes and with the now-favored X = C4− as the central ligand of the FeMo-cofactor. With the same μ2 S2B-H protonation scheme as for the X = N3− case, the calculations yielded bound ν(CO) frequencies >30 cm−1 red-shifted compared to the observed FT-IR values and optimized structures qualitatively identical to those presented in Figure 4. An extra proton addition at the μ2 S5A FeMo-cofactor sulfur places our X = C4− model at the same charge level (−3 units) as that of the X = N3− model, and it significantly improves the correspondence with the FT-IR experiment. For the calculated Hi-3 (12C16O and 13C18O) ν(CO) modes, the consistent negative deviation from the FT-IR peaks is 6–8 cm−1, while for the Lo-3 modes it is positive 2–3 cm−1, see Table S2 and S3 of the Supporting Material.</p><!><p>Studies of the interaction of CO with N2ase have a long history dating back to at least 1941.[4a] The subject has taken on added significance with the discovery of N2ase Fischer-Tropsch-like CO reduction and condensation chemistry.[5–8] To date, nearly all the discussion about the structure and reactivity of CO intermediates has centered about two EPR-observable species, hi-CO and lo-CO.[9–10, 28] However, since integration of these EPR signals usually accounts for less than 50% of the enzyme, it is obvious that a good deal (50% or more) of N2ase-CO chemistry has not been accounted for.</p><p>To access these EPR/ENDOR-silent species, we have employed an FT-IR-monitored photolysis technique in much the same way as used in myoglobin-CO studies.[29] Our initial results[15] under high [CO] conditions indicated the presence of a previously unrecognized species that we labelled Hi-3, which was EPR-silent and so did not correlate with any of the known EPR-active species. In this current study, we have elucidated the spectroscopic properties, photochemistry, and likely structures of Hi-3 and its photolysis product Lo-3. As part of assessing the significance of these new EPR-silent species, an obvious question is: what fraction of our N2ase samples is present in these forms?</p><p>We can estimate Hi-3 abundance relative to Hi-1, by assuming (i) that Hi-1 contains a single terminal CO (as well as a second either bridging and/or more reduced CO species) bound at the FeMo-cofactor and (ii) that the sum of the oscillator strengths of the Hi-3 signals at 1938 and 1911 cm−1 (from two coupled CO oscillators) corresponds to twice the strength of the Hi-1 signal at 1969 cm−1 (from a single CO). From integrating the intensities of bands shown in Figure 1, the Hi-3 population is estimated as ~95% that of Hi-1. Although this is only an estimate, it is clear that the Hi-3 species is a major component of the reaction mixture. From the same assumptions, Hi-1 and Hi-3 signal concentrations are both significantly higher than that of Hi-2. Previous integrations of the EPR spectra of the α-H195Q MoFe protein under high [CO] conditions found 26% hi-CO, 6% hi(5)-CO, and 8% resting state enzyme.[11d] In some of our samples the hi-CO signal approaches 50% of the total MoFe protein. Thus, if we make the further assumption (iii) that the Hi-1 IR species is the same as the hi-CO EPR species,[15] then we conclude that at least a quarter and sometimes approaching half of our samples is present as Hi-3.</p><p>The combination of IR-detected photolysis experiments and DFT calculations has allowed us to propose a model for the Hi-3 species that involves two terminal exo CO ligands, one bound to each of the adjacent Fe2 and Fe6 sites of the FeMo-cofactor that are bridged by the protonated μ2 S2B sulfur (see Figure 4). Vibrational coupling between these two terminal CO molecules is proposed to contribute to the ~27 cm−1 frequency splitting and helps explain relative ~3.7:1 intensities of the Hi-3 bands at 1938 and 1911 cm−1. In the present study, such coupling was confirmed by the appearance of a new band at 1917 cm−1 in samples inhibited by a mixture of 12C16O and 13C18O isotopes. Our DFT model for Hi-3 lends support to this proposal. The model yields a 29 cm−1 splitting and relative intensities of ~3.5:1 for the coupled ν(CO) modes, symmetric at 1938 cm−1 vs. asymmetric at 1909 cm−1. The long-range vibrational coupling between CO molecules bound at the Fe sites ~2.7 Å apart can be rationalized by (i) approximate symmetry of the proposed Hi-3 structure with respect to the plane formed by the three μ2 S sulfurs (and the central X ligand), and (ii) electronic delocalization inherent to the iron-sulfur clusters such as FeMo-cofactor. Notably, this vibrational coupling may have implications to Hi-3 reactivity. Although our structural model is similar to one proposed by Hoffman and coworkers to account for properties of the hi-CO EPR signal,[10, 28b] we instead are using it to account for the properties of an EPR-silent species. Our prior results for the Hi-1 IR species, which we equate with the hi-CO EPR species,[15] indicate that this form instead has a single terminal CO with perhaps either a bridging CO or formyl ligand.</p><p>Dance has also conducted DFT calculations on CO-inhibited FeMo-cofactor with adjacent terminal CO ligands.[24] Two of his models (designated 10 and 11) with Fe2 exo and Fe6 exo CO ligands (for the resting state S = 1/2 cofactor, no protons added, and using X = N3−) most closely correspond to our current Hi-3 DFT proposal and carry the same total negative charge of 3e−. These models, which differ only by 'coordinative allosterism' in bonding with the central atom X, are predicted to have somewhat lower ν(CO) frequencies of 1914/1886 and 1904/1893 cm−1 for 10 and 11, respectively, than the 1938/1911 cm−1 pair observed in our experiments, and the intensity ratios were less than 2:1. From the present experience, these shifts could possibly be brought in line with our IR and DFT results via addition of one electron, and one proton at μ2 S sulfur. The importance of the electron and proton count for the bound ν(CO) frequencies and other FeMo-cofactor properties is also outlined below.</p><p>For the Lo-3 photolysis product, our favored DFT-based model places a single CO ligand in the Fe2 exo position (see Figure 4), implying that the net result of photolysis is displacement of the CO ligand from Fe6. We cannot exclude the possibility that photolysis of CO from the Fe2 position also occurs, with subsequent relaxation to the preferred structure. The DFT model yields a calculated ν(CO) band at 1923 cm−1, which compares well with the Lo-3 FT-IR signature at 1921 cm−1. The calculated intensity of this band is ~2.1:1 relative to the lower frequency Hi-3 mode, compared to the ~1.7:1 ratio seen in the FT-IR experiment. Dance also modeled FeMo-cofactor structures with only one CO ligand.[24] Most relevant to our proposed Lo-3 structure is his model 8, which places CO at Fe2 endo and yields a calculated ν(CO) frequency of 1904 cm−1. Other models with Fe2-H and/or Fe6-H hydrides in addition to a terminal CO (such as model 12), led to vibrational coupling of the C≡O and Fe-H stretches and introduced ν(CO) splittings that do not correspond to our Lo-3 experimental observations.</p><p>With the current DFT calculations, the FeMo-cofactor protonation scheme and the identity of the interstitial X (= C4−/N3−/O2−) ligand were both found to be important for reproducing the Hi-3 and Lo-3 FT-IR spectra. The interplay of these alternatives in defining the charge state of the FeMo-cofactor model was stressed earlier, for example in connection to Mössbauer isomer shifts[27, 30] and redox potential[27, 30b, 31] calculations. For our 1e− reduced Hi-3 and Lo-3 FeMo-cofactor states, we found single protonation of the cofactor at the μ2 S2B sulfur optimal for the X = N3− alternative, while double protonation at μ2 S2B and μ2 S5A gave the best results for X = C4−. The X = N3− option provides very good correspondence (≤2 cm−1 offset) to the Hi-3 and Lo-3 signatures and allows to interpret the minor peaks of the mixed isotope experiment (see Figure 5), while the X = C4− model results in somewhat larger frequency shifts of up to 8 cm−1 with respect to the FT-IR data (see Table S3 of the Supporting Material). However, since observed vs. computed frequency deviations of up to ~1% are inherent to the present DFT setup, our calculations cannot really discern whether X = N3− or C4−. We have no reason to doubt the X = C4− assignment that derives from other methods.[2–3]</p><p>Before discussing how or whether the Hi-3/Lo-3 couple has a role in the inhibition of N2 and other substrate reduction or perhaps in Fischer-Tropsch-like chemistry, we need to determine, in the nomenclature used by Hoffmann and coworkers,[28b, 32] its 'electron inventory'. This refers to the number of electrons delivered from the Fe protein (called n) to form the Hi-3 species, starting from the resting state called E0. The equation n = s + m – p takes into account the number of electrons transferred to substrate (s), the number delivered to the FeMo-cofactor metals (m), and the number of electrons transferred from the P-cluster (p). The EPR silence of the Hi-3 species implies that n is odd, in order to produce an even-electron (most likely S = 0) Hi-3 species from the S = 3/2 EPR-active odd-electron resting (n = 0) state. The fact that Hi-3 is stable for days or longer argues against a highly reduced species with n ≥ 3, because such a species would presumably relax by H2 evolution. Given the preference of CO for reduced metal ions, chemical intuition also argues against a more oxidized form of the FeMo-cofactor (hence negative values for n). We are left with n = 1, a 1e− reduced FeMo-cofactor compared to resting MoFe protein. There is no evidence for P-cluster oxidation under these conditions, thus p = 0. Because the bound CO has not been reduced, s = 0. We thus get m = n = 1 for the FeMo-cofactor metal ion core electron count beyond the resting state.</p><p>A recent proposal suggests that the metal core of the FeMo-cofactor has only two accessible oxidation levels, the EPR-active resting state (MN) with m = 0 and the one-electron-reduced EPR-silent state (MR) with m = 1.[32] If so, then our Hi-3 species would retain the single electron on the FeMo-cofactor and be EPR-silent, whereas the lo-CO, hi-CO, and hi(5)-CO species, being more reduced than the resting state by an even number of electrons, would be EPR-active and these 'extra' electrons would be forced to reside on 'substrates'. Thus, the lo-CO, hi-CO and hi(5)-CO EPR signals would come from forms of N2ase with m = 0 and s = 2 for a total of n = 2. The substrate-localized electrons could reside on either a formyl species or an as yet undetected hydride species. Partially reduced substrates are of course critical intermediates for C-C coupling reactions[33] as demonstrated previously by the formation of a similar range of hydrocarbons from N2ase-catalyzed isocyanide reduction.[34]</p><p>Is the Hi-3 (or Lo-3) species relevant to N2ase Fischer-Tropsch chemistry? Most mechanisms for Fe-catalyzed Fischer-Tropsch chemistry posit the presence of CO-derived reactants on adjacent Fe atoms,[35] and for N2ase, intermediates with adjacent CO or CHxO0,1 species[7, 33] have also been proposed. Evidence for multiple binding sites in N2ase dates back 40 years or more,[4b] but it has always been difficult to separate physically distinct binding sites on the same species from those produced by different redox levels and/or protonation states of the FeMo-cofactor. The current study is the first to demonstrate simultaneous terminal CO binding to distinct and adjacent Fe sites on the FeMo-cofactor. It thus demonstrates that two CO molecules can bind to Fe atoms only ~2.7 Å apart.</p><p>Yang et al. observed that the α-H195Q. α-V70A Av1 double mutant was much less efficient at hydrocarbon production than the α-V70A single mutant, and they suggested that α-H195 functions to deliver protons for the reduction of CO.[7, 36] Disruption of this proton transfer may also explain why we see a larger fraction of the Hi-3 species in the α-H195Q Av1 compared to wild-type enzyme.[15] Additional work is of course needed to determine whether Hi-3 represents a requisite intermediate or a side reaction resulting from a 'proton bottleneck'.</p><p>Finally, the Hi-3 species is only 1-electron reduced from the resting state of the enzyme. Since the reduction of two CO molecules to C2H4 and 2 H2O requires 8 electrons and 8 protons, there is a good deal of chemistry required between Hi-3 and the major product observed during CO reduction. Efforts to identify and characterize those additional species are of course in progress.</p><!><p>Purified α-H195Q MoFe protein and Fe protein had specific activities of 2000–2800 nmol of H2 (min/mg protein) −1 and 2000–2300 nmol of H2 (min/mg protein) −1 respectively. The purified component proteins were exchanged into D2O buffer containing 25 mM HEPES, pH 7.4, 10 mM MgCl2, 250 mM NaCl. For preparation of CO-inhibited samples, low electron-flux conditions were obtained using a 1:4 molar ratio of Fe protein:MoFe protein. The reaction-mixture components, which consisted of 2.5 mM ATP, 5 mM MgCl2, 30 mM creatine phosphate, 25 units/mL of creatine phosphokinase in 25 mM HEPES, pH 7.4 and 20 mM sodium dithionite were prepared in D2O. Turnover was accomplished under 101 kPa CO. The reaction was quenched by the addition of ethylene glycol to a final concentration of 40% after 15 minutes. The resulting turnover product was concentrated in an Amicon microfiltration pressure concentrator using a regenerated cellulose PLHK ultrafiltration membrane with a 100,000 molecular weight cut off under 202 kPa CO of the same composition used for the turnover reaction.</p><!><p>The sample photolysis was conducted in an Oxford liquid He flow cryostat using a Sutter Instruments 300 W Lambda LS xenon arc lamp. Spectra were recorded at 4 cm−1 resolution with a Bruker V-70v FT-IR spectrometer and a MCT detector. For photolysis, the lamp was shone through the side of the cryostat oriented 90° to the IR light path. The sample was held at 45° to both beams. This allowed the use of quartz windows for the visible light. For wavelength dependent photolysis, we used a set of VersaChrome® tunable bandpass filters (Semrock) with central wavelengths at 340, 380, 410, 440, 470, 510, 550, 610, 700 and 800 nm.</p><!><p>The DFT calculations were done using the PBE functional and the LACV3P** basis set, as implemented in JAGUAR 7.7 software. For the first- and second-row elements, LACV3P** implies a 6-311G** triple-zeta basis set including polarization function. For the Fe and Mo atoms, LACV3P** uses the Los Alamos effective core potential (ECP), and the valence part is essentially of triple-zeta quality. The geometries optimized at the PBE/LACV3P** level were used for the analytic Hessian calculations, resulting in the harmonic frequencies and IR intensities discussed in the text. The analysis of the computed vibrational normal modes has been facilitated using an in-house Q-SPECTOR Python tool, applied to model the FT-IR spectra and assess the FeMo-cofactor bound CO modes via kinetic energy distribution (KED) approach. Further important details are given in the Supporting Information sections.</p>

PubMed Author Manuscript

A reactive oxygen species-generating, cancer stem cell-potent manganese(<scp>ii</scp>) complex and its encapsulation into polymeric nanoparticles

Intracellular redox modulation offers a viable approach to effectively remove cancer stem cells (CSCs), a subpopulation of tumour cells thought to be responsible for cancer recurrence and metastasis. Here we report the breast CSC potency of reactive oxygen species (ROS)-generating manganese(II)-and copper(II)-4,7-diphenyl-1,10-phenanthroline complexes bearing diclofenac, a nonsteriodial antiinflammatory drug (NSAID), 1 and 3. Notably, the manganese(II) complex, 1, exhibits 9-fold, 31-fold, and 40-fold greater potency towards breast CSCs than 3, salinomycin (an established breast CSC-potent agent), and cisplatin (a clinically approved anticancer drug) respectively. Encouragingly, 1 displays 61-fold higher potency toward breast CSCs than normal skin fibroblast cells. Clinically relevant epithelial spheroid studies show that 1 is able to selectively inhibit breast CSC-enriched HMLER-shEcad mammosphere formation and viability (one order of magnitude) over non-tumorigenic breast MCF10A spheroids. Mechanistic studies show that 1 prompts breast CSC death by generating intracellular ROS and inhibiting cyclooxygenase-2 (COX-2) activity. The manganese(II) complex, 1, induces a greater degree of intracellular ROS in CSCs than the corresponding copper(II) complex, 3, highlighting the ROSgenerating superiority of manganese(II)-over copper(II)-phenanthroline complexes. Encapsulation of 1 by biodegradable methoxy poly(ethylene glycol)-b-poly(D,L-lactic-co-glycolic) acid (PEG-PLGA) copolymers at the appropriate feed (5%, 1 NP 5 ) enhances breast CSC uptake and greatly reduces overall toxicity. The nanoparticle formulation 1 NP 5 indiscriminately kills breast CSCs and bulk breast cancer cells, and evokes a similar cellular response to the payload, 1. To the best of our knowledge, this is the first study to investigate the anti-CSC properties of managense complexes and to demonstrate that polymeric nanoparticles can be used to effectively deliver managense complexes into CSCs.

a_reactive_oxygen_species-generating,_cancer_stem_cell-potent_manganese(<scp>ii</scp>)_complex_and_i

Introduction<!>Synthesis and characterisation<!>Lipophilicity and stability in biological solutions<!>Cytotoxicity toward CSCs, bulk cancer cells, and normal cells in monolayer systems<!>Epithelial spheroid inhibitory studies<!>Cellular uptake and fractionation in breast CSCs<!>Mechanism of CSC toxicity of 1<!>Encapsulation of 1 into polymeric nanoparticles<!>Delivery of 1 into CSCs using polymeric nanoparticles<!>CSC potency and mechanism of action of the nanoparticle formation, 1 NP 5<!>Conclusions<!>Conflicts of interest

<p>A small subset of tumour cells, termed cancer stem cells (CSCs), are thought to be one of the underlying causes of metastasis and relapse. 1,2 CSCs, akin to normal stem cells, replicate slowly and thus evade conventional cancer treatments (such as platinum chemotherapy and radiotherapy), which primarily target proliferating cells. [3][4][5] Having bypassed conventional treatments, CSCs can instigate metastasis and relapse due to their cell motility prociency and high secondary tumour seeding frequency, respectively. 6,7 Therefore it is imperative for new generations of anticancer agents to remove entire populations of tumour cells, including CSCs, in order to provide a durable clinical response. Several CSC functional targets have been identied, however, there is still no clinically approved chemical or biological agent that can remove both bulk cancer cells and CSCs at non-lethal doses. 8,9 All of the chemical agents currently undergoing clinical trials as prospective anti-CSC agents are completely organic in nature. 8 Recently, we and others have shown that metal complexes can exhibit promising anti-CSC activities and deserve further investigation in this context. 10 Copper-and iron-containing complexes have emerged as promising anti-CSC agents owing to their ability to partake in Fenton-type reactions and produce lethal levels of reactive oxygen species (ROS) inside CSCs. [11][12][13] Notably, copper(II)-phenanthroline and -Schiff base complexes bearing non-steroidal anti-inammatory drugs (NSAIDs) kill breast CSCs and bulk breast cancer cells, both selectivity and indiscriminately at micromolar and sub-micromolar concentrations. 12,14,15 Mechanistic studies show that these complexes induce CSC death by elevating intracellular ROS levels and inhibiting cyclooxygenase isoenzyme-2 (COX-2). This approach exploits the vulnerability of CSCs and bulk cancer cells to changes in their intracellular redox state and the overexpression of COX-2, a major player in regulation, in certain CSCs and bulk cancer cells. 16,17 Here we have sought to improve CSC potency by developing related manganese(II)-phenanthroline complexes bearing diclofenac (a potent COX-2 inhibitor with anti-metastatic potential) 18,19 with higher ROS-generating power. Analogous copper(II)-phenanthroline complexes bearing diclofenac were also prepared and used as controls to showcase the superiority of manganese(II) over copper(II) (within this system) to elevate intracellular ROS levels.</p><p>Manganese is an essential trace element for animals and plants, and plays an important role in many biological processes including cell protection, metabolism, bone mineralisation, and reproductive function. 20 Although the anticancer properties of manganese complexes are underexplored, a number of bulk cancer cell-potent manganese(II/III) complexes containing Schiff base, porphyrin, avonoids, and polypyridyl ligands have been identied. [21][22][23][24] The cellular mechanism of action of the manganese(II/III) complexes is highly dependent on the nature of the coordinated ligand. 25 Notably, a number of studies have shown that manganese metal and its coordination complexes increase markers of oxidative stress in various cell types due to intracellular ROS generation. 20,[26][27][28][29] By utilising manganese over other more common endogenous metals such as copper and iron, we hope to take advantage of the many physiologically accessible oxidation states of manganese and its ability to undergo Fentontype reactions (whereby the manganese(II) form can effectively generate hydroxide radicals from endogenous hydrogen peroxide), to potently kill CSCs via efficient ROS production. Indeed, a series of di-nuclear manganese(II)-phenanthroline complexes were previously reported to signicantly increase intracellular ROS levels and thus cytotoxicity in human-derived colorectal cancer lines, relative to equivalent copper(II) complexes. 27 The impact of manganese-mediated ROS generation in CSCs, however, is completely unknown.</p><p>Therapeutic agents (especially metal complexes) can suffer from systematic toxicity due to non-specicity, resulting from their degradation in biological solutions. Nanoscale technologies offer a method to precisely deliver anticancer therapeutics to tumour microenvironments. 30 Further, nano-systems increase drug solubility, bioavailability, and extend drug halflife. 31,32 Nano-sized particles passively target certain tumours by taking advantage of the enhanced permeability and retention (EPR) effect in tumour tissues. 33,34 From recent clinical studies, it is clear that the EPR effect is highly dependent on tumour histology. 35,36 For instance, human carcinomas are very vascular and have many porous blood vessels and thus more susceptible to targeting by the EPR effect. 37 Slow-growing human sarcomas on the other hand are not extremely vascular and thus less able to accumulate nano-sized entities via the EPR effect. The work presented in this manuscript involves breast carcinomas (bulk and CSCs). Several nanoparticle formulations exist, including those based on iron oxide, carbon, gold, hydrogels, liposomes, and polymers. 38 A number of these formulations are currently used in the clinic to deliver chemotherapies to tumours. 39 Polymeric nanoparticles are of particular interest due to their biocompatibility, synthetic versatility, and tuneable properties. 40 We recently reported a proof-of-concept study where the biodegradable, amphiphilic copolymer, methoxy poly(ethylene glycol)-b-poly(D,L-lactic-co-glycolic) acid (PEG-PLGA) was used to encapsulate and deliver a copper(II)-NSAID complex into (carcinoma-derived) breast CSCs. 41 In addition to improved cellular uptake, cytotoxicity toward breast CSCs over bulk breast cancer cells, was enhanced compared to the payload. Here we use similar polymeric nanoparticles to encapsulate and deliver the most CSC-potent manganese(II)-phenanthroline complex identied in this study into breast CSCs.</p><!><p>The manganese(II)-and copper(II)-phenanthroline complexes, 1-4 investigated in this study are shown in Fig. 1. The manganese(II) complexes, 1 and 2 were prepared by reacting MnCl 2 $4H 2 O with 3,4,7,8-tetramethyl-1,10-phenanthroline or 4,7-diphenyl-1,10-phenanthroline and two equivalents of diclofenac sodium salt in methanol, under basic conditions. The corresponding copper(II) complexes, 3 and 4 were synthesised via the same procedure, with CuCl 2 $2H 2 O used as the metal source. The manganese(II) complexes, 1 and 2 were isolated as yellow solids whereas the copper(II) complexes, 3 and 4 were obtained as green solids. All of the complexes, 1-4 were characterised by infrared spectroscopy and elemental analysis (Fig. S1 †). The difference between the vibrational stretching frequencies between the asymmetric, n asym (CO 2 ) and symmetric, n sym (CO 2 ) carbonyl peaks gives an indication into the binding mode of the associated carboxylic acid group to a given metal centre. 42,43 According to the IR spectra, the difference (D) between n asym (CO 2 ) and n sym (CO 2 ) stretching bands for 1 and 2 varied between 125-128 cm À1 (Fig. S1A and B †), suggestive of a bidentate binding mode for the carboxylate group on diclofenac to the manganese(II) centre (as depicted in Fig. 1). Similarly, the difference (D) between n asym (CO 2 ) and n sym (CO 2 ) stretching bands for 3 and 4 was 107-125 cm À1 (Fig. S1C and D †), indicative of a bidentate binding mode between diclofenac and the copper(II) centre (as depicted in Fig. 1). The purity of 1-4 was established by elemental analysis (see Experimental details section in the ESI †). Mn(3,4,7,8-tetramethyl-1,10-phenanthroline)Cl 2 was also prepared to serve as a control compounda manganese(II)-phenanthroline complex without diclofenac. The synthetic protocol and characterisation of Mn(3,4,7,8-tetramethyl-1,10-phenanthroline)Cl 2 is reported in the Experimental details section in the ESI. †</p><!><p>The lipophilicity of 1-4 was determined by measuring the extent to which it partitioned between octanol and water, P. The experimentally determined log P values varied from 0.9 to 1.1 for the manganese(II) complexes, 1 and 2, and 0.2 to 0.8 for the copper(II) complexes, 3 and 4 (Table S1 †). The hydrophobic nature of 1-4 suggests that the manganese(II) and copper(II) complexes should be readily internalized by cells. UV-vis spectroscopy studies were carried out to assess the stability of 1-4, in biological relevant solutions. The UV-vis metal-perturbed p-p* absorption band of 1 (25 mM) in PBS : DMSO (200 : 1) with and without 10 equivalents of glutathione (a cellular reductant) remained constant over the course of 24 h at 37 C (Fig. S2 and S3 †), indicative of stability. In contrast, the 4,7-diphenyl-1,10phenanthroline-containing manganese(II) complex, 2 was relatively unstable in PBS : DMSO (200 : 1) with and without 10 equivalents of glutathione (Fig. S4 and S5 †). Like the trend observed for the manganese(II) complexes, the copper(II) complex bearing 3,4,7,8-tetramethyl-1,10-phenanthroline, 3 was comparatively more stable in PBS : DMSO (200 : 1) with and without 10 equivalents of glutathione than the 4,7-diphenyl-1,10-phenanthroline-bearing complex, 4 (Fig. S6-S9 †). The UVvis bands associated to 1 and 3 (25 mM) in mammary epithelial cell growth medium (MEGM)/DMSO (200 : 1) remained unaltered over the course of 24 h at 37 C, suggestive of stability in conditions required for cellular studies (Fig. S10 and S11 †). Under the same conditions, 2 and 4 were relatively unstable (Fig. S12 and S13 †). Collectively, the UV-vis spectroscopy studies suggest that the stability of the complexes in biologically relevant solutions is dependent on the metal and the polypyridyl ligand, with the manganese(II)-3,4,7,8-tetramethyl-1,10phenanthroline complex, 1 being the most stable. Highresolution ESI mass spectrometry studies were also performed to conrm the stability of 1 in biologically-relevant conditions. Upon incubation of 1 (100 mM) in PBS : DMSO (200 : 1) for 72 h at 37 C, a peak corresponding to [1-2H + 2Na] + (925.1902 m/z) with the appropriate isotopic distribution was observed in the positive mode of the ESI mass spectrum (Fig. S14 †), further suggesting that 1 is stable under these conditions.</p><!><p>Given the instability of the 4,7-diphenyl-1,10-phenanthrolinecontaining complexes, 2 and 4 in MEGM (Fig. S12 and S13 †), cellular studies were not performed with these complexes. The cytotoxicity of 1 and 3 against bulk bone (U2OS), liver (HepG2), and breast (HMLER) cancer cells, and breast CSC-enriched cells (HMLER-shEcad) was determined using the MTT assay. The IC 50 values were determined from dose-response curves (Fig. S15 and S16 †) and are summarised in Tables 1 and S2. † Both complexes displayed submicro-or micro-molar potency towards the cell lines tested. The manganese(II) complex, 1 killed CSC-enriched HMLER-shEcad cells and CSC-depleted HMLER cells similarly whereas the copper(II) complex, 3 killed CSC-depleted HMLER cells preferentially (9-fold selectivity) (Table 1). Notably, 1 displayed 9-fold, 31-fold, and 40-fold greater potency (p < 0.05, n ¼ 18) for HMLER-shEcad cells than 3, salinomycin (an established breast CSC-potent agent), 12 and cisplatin (a clinically approved anticancer drug) respectively (Fig. S17, † Table 1). As a measure of therapeutic potential, the cytotoxicity of 1 and 3 towards normal skin broblast GM07575 cells was determined. The complexes, 1 and 3 were 61-fold and 11-fold less potent toward GM07575 cells (IC 50 value ¼ 8.5 AE 1.5 mM for 1 and 13.9 AE 0.1 mM for 3, Fig. S18 and S19 †) than HMLER-shEcad cells respectively, indicating selective toxicity for breast CSCs over non-tumorigenic cells. Taken together, the cytotoxicity studies showed that the manganese(II) complex, 1 is signicantly (p < 0.05) more cytotoxic towards breast CSCs than its copper(II) analogous, 3, and displays reduced toxicity towards normal cells.</p><p>Control cytotoxicity studies showed that the potency of MnCl 2 $4H 2 O, CuCl 2 $2H 2 O, diclofenac, and the polypyridyl ligand, 3,4,7,8-tetramethyl-1,10-phenanthroline, towards HMLER and HMLER-shEcad cells was signicantly lower (p < 0.05, n ¼ 18) than 1 and 3 (Table 1 and Fig. S20-S23 †). This suggests that the cytotoxicity of 1 and 3 towards bulk breast cancer cells and breast CSCs is likely to result from the intact manganese(II) and copper(II) complexes rather than their individual components (manganese, copper, diclofenac, or the free polypyridyl ligand). Interestingly the potency of Mn(3,4,7,8-tetramethyl-1,10-phenanthroline)Cl 2 towards HMLER and HMLER-shEcad cells was similar to 1 (Table 1 and Fig. S24 †). This suggests that the Mn-polypyridyl unit is largely responsible for the bulk breast cancer cell and breast CSC toxicity of 1.</p><!><p>Epithelial breast cells (cancer and non-tumorigenic), when grown in serum-free media under low-attachment conditions, are capable of forming three-dimensional spheroids called mammospheres. 44 The ability of a given compound to inhibit mammosphere formation from single cell suspensions (with respect to number, size, and viability) is oen used as a marker for in vivo potency, given that three-dimensional systems are more representative of organs and tumours than monolayer cell cultures. The ability of 1 and 3 to inhibit breast CSC-enriched HMLER-shEcad and non-tumorigenic breast epithelial MCF10A mammosphere formation (at a non-lethal dose, IC 20 value for 5 days) was assessed using an inverted microscope.</p><p>The addition of 1 (IC 20 value for 5 days) to single cell suspensions of HMLER-shEcad cells signicantly (p < 0.05) decreased the number and size of HMLER-shEcad mammospheres formed (Fig. 2A and B). Notably, the HMLER-shEcad mammosphere inhibitory effect of 1 (51% decrease in number of HMLER-shEcad mammospheres formed compared to the untreated control) was comparable to that of salinomycin under identical conditions (53% decrease in number of HMLER-shEcad mammospheres formed compared to the untreated control, Fig. 2A and B). The copper(II) complex, 3 (IC 20 value for 5 days) did not signicantly alter the number or size of HMLER-shEcad mammospheres formed (Fig. S25 and S26 †). We have previously shown that diclofenac does not signicantly affect the number or size of HMLER-shEcad mammospheres formed. 15 This suggests the HMLER-shEcad mammosphere inhibitory effect of 1 is a due to the manganese(II)-polypyridyl component. To gauge the ability of 1 to decrease HMLER-shEcad mammosphere viability, the colorimetric resazurinbased reagent, TOX8 was employed. Encouragingly, the IC 50 value of 1 (3.0 AE 0.1 mM) (Fig. S27 †) was 6-fold lower than that reported for salinomycin (18.5 AE 1.5 mM) 15 under identical conditions. As diclofenac is non-toxic towards HMLER-shEcad mammospheres (IC 50 > 133 mM), 15 we believe that the manganese(II)-polypyridyl component in 1 is the major contributor for HMLER-shEcad mammosphere toxicity.</p><p>The addition of 1 (IC 20 value for 5 days) to single cell suspensions of non-tumorigenic MCF10A cells did not signicantly (p ¼ 0.25) change the number or size of MCF10A spheroids formed (Fig. 2A and B). In contrast, treatment with salinomycin under the same conditions resulted in a signicant (p < 0.05) decrease in the number (40% decrease) and size of MCF10A spheroids formed (Fig. 2A and B). This is similar to the result obtained with HMLER-shEcad mammospheres (Fig. 2A and B). Spheroid viability studies showed that 1 killed nontumorigenic MCF10A spheroids (IC 50 ¼ 36.7 AE 1.9 mM) with 12-fold lower potency than CSC-enriched HMLER-shEcad mammospheres (Fig. S28 †). Salinomycin on the other hand killed MCF10A (IC 50 ¼ 14.9 AE 0.5 mM) and HMLER-shEcad mammospheres equipotently (Fig. S29 †). Overall the epithelial spheroid studies, show that 1 is able to selectively inhibit CSCenriched HMLER-shEcad mammospheres formation and viability over non-tumorigenic MCF10A spheroids. These properties are highly desirable for selecting CSC drug candidates in preclinical studies.</p><!><p>Cellular uptake studies were carried out to determine the breast CSC permeability of 1 and 3. CSC-enriched HMLER-shEcad cells were incubated with 1 or 3 (0.25 mM for 12 h) and the intracellular manganese or copper concentration was determined by inductively coupled plasma mass spectrometry (ICP-MS). The complex, 1 was modestly taken up by HMLER-shEcad cells, with whole cell uptake totalling 5.4 AE 0.1 ppb of Mn/million cells (Fig. S30 †). Under identical conditions, 3 (0.25 mM for 12 h) entered HMLER-shEcad cells to a higher extent (17.5 AE 0.5 ppb of Cu/million cells) than 1. Although the amount of manganese and copper found inside HMLER-shEcad and HMLER cells is relatively low, it is reasonable considering the low administration concentration (0.25 mM). Fractionation studies were conducted to determine the cell localisation of 1 in HMLER-shEcad cells (Fig. S30 †). In terms of the cell distribution pattern, 6% of internalised 1 was found in the nucleus, 39% in the cytoplasm, and 10% in the membrane (Fig. S30 †). The ability of 1 to access the nucleus could be attributed to the presence of the diclofenac ligand, which is known to interact with COX-2 located on the nuclear envelope. 45 Since a relatively large amount of internalised 1 was also found in the cytoplasm and membrane, the fractionation studies suggest that 1-induced cell toxicity could be related to a mixture of genomic DNA-dependent or -independent mechanisms.</p><!><p>The manganese(II) complex, 1 was envisaged to kill CSCs by increasing intracellular ROS levels to lethal doses. To test this hypothesis, the ability of 1 to produce ROS in HMLER-shEcad cells was determined using 6-carboxy-2 0 ,7 0 -dichlorodihydro-uorescein diacetate (DCFH-DA), a well-established ROS indicator. HMLER-shEcad cells treated with 1 (0.25 mM) exhibited a substantial increase in intracellular ROS levels over the course of 48 h, peaking aer 6 h exposure (3.4-fold increase) (Fig. 3A). Under identical conditions, the corresponding copper(II) complex, 3 (0.25 mM) displayed a general increase in ROS levels in HMLER-shEcad cells over the course of 48 h, however, the extent was signicantly lower than 1 (Fig. S31 †). Cytotoxicity studies in the presence of N-acetylcysteine (NAC, 2 mM, 72 h), a ROS scavenger showed that the potency of 1 and 3 towards HMLER-shEcad cells decreased signicantly (IC 50 value increased from 137 AE 15 nM to 1100 AE 89 nM, 8-fold, p < 0.05 for 1 and 1.3 AE 0.3 mM to 3.6 AE 0.1 mM, 3-fold, p < 0.05 for 3) (Fig. S32 and S33 †) suggesting that 1 and 3-induced CSC death is related to intracellular ROS generation. The differential increases in IC 50 values of 1 and 3 in the presence of NAC indicates that 1-induced CSC death is more reliant on intracellular ROS production than 3-induced CSC death. Therefore, the greater intracellular ROS-generating power of 1 over 3 could explain the 9-fold greater potency of 1 over 3 against HMLER-shEcad cells (Table 1).</p><p>Intracellular ROS production can activate Jun-aminoterminal kinase (JNK) and/or p38 MAP kinase (MAPK) pathways. 46 To provide further insight into the mechanism of cytotoxicity of the manganese(II) complex, 1 we conducted immunoblotting studies to monitor changes in the expression of biomarkers related to these pathways. HMLER-shEcad cells dosed with 1 (0.25-1 mM for 72 h) exhibited enhanced phosphorylation of JNK and its downstream effector, c-Jun indicative of JNK pathway activation (Fig. 3B). Interestingly, p38 phosphorylation was not markedly effected although its downstream effector, MAP kinase-activated protein kinase 2 (MAPKAPK-2) was greatly phosphorylated in the presence of 1 (Fig. 3B). Constant activation of JNK or p38 pathways can induce apoptosis. 47 Immunoblotting studies showed that HMLER-shEcad cells treated with 1 (0.25-1 mM for 72 h) displayed increased expression of cleaved caspase-3 and -7 compared to untreated cells, indicative of caspase-dependent apoptosis (Fig. 3B). Interestingly, 1 induced a greater level of caspase-3 and -7 cleavage upon treatment of HMLER-shEcad cells with 0.25 mM of 1 than 0.5 mM of 1. Cytotoxicity studies in the presence of z-VAD-FMK (5 mM), a peptide-based caspasedependent apoptosis inhibitor showed that the potency of 1 towards HMLER-shEcad cells decreased signicantly (p < 0.05, IC 50 value ¼ 220 AE 14 nM) (Fig. 3C). This conrms that 1 induces caspase-dependent CSC death. As expected the potency of cisplatin, a well-known apoptosis-inducer, towards HMLER-shEcad cells decreased signicantly (p < 0.05, IC 50 value ¼ 10.21 AE 0.78 mM) in the presence of z-VAD-FMK (5 mM) (Fig. S34 †). Collectively, the ROS, immunoblotting, and cytotoxicity studies show that 1 elevates intracellular ROS levels, activates the JNK pathway, and triggers apoptotic CSC death.</p><p>COX-2 is overexpressed in certain bulk breast cancer cells and breast CSCs and is thought to play a meaningful role in CSC maintenance and regulation, as well as chemotherapeutic resistance. 17,48,49 COX-2 is widely regarded as a molecular target for CSC-directed therapy. Given that the manganese(II) complex, 1 contains two diclofenac ligands, ow cytometric studies were performed to determine if the mechanism of action of 1 involved COX-2 downregulation. HMLER-shEcad cells were pretreated with lipopolysaccharide (LPS) (2.5 mg L À1 for 24 h), to increase basal COX-2 levels, and then treated with 1 (0.25 mM), diclofenac (20 mM) or Mn(3,4,7,8-tetramethyl-1,10-phenanthroline)Cl 2 (0.25 mM) for 48 h and the COX-2 expression was determined. A noticeable decrease in COX-2 expression was also observed in HMLER-shEcad cells treated with 1 and diclofenac compared to untreated cells (Fig. S35 †). No appreciable change in COX-2 expression was detected in the presence of Mn(3,4,7,8tetramethyl-1,10-phenanthroline)Cl 2 suggesting that the COX-2 downregulation observed in the presence of the 1 is most likely mediated by the diclofenac moieties (Fig. S36 †). Overall this suggests that the cytotoxic mechanism of action of 1 may involve COX-2 downregulation. To probe this further, cytotoxicity studies were carried out with HMLER-shEcad cells in the presence of prostaglandin E2 (PGE2) (20 mM, 72 h), the functional product of COX-2-catalysed arachidonic acid metabolism. The IC 50 value for 1 against HMLER-shEcad cells in the presence of PGE2 was signicantly (p < 0.05, IC 50 value ¼ 255 AE 50 nM) higher than for 1 alone (Fig. S37 †), implying that 1 induces COX-2-dependent CSC death.</p><!><p>In order to improve the anti-CSC properties of 1 further (especially the cellular uptake), biodegradable PEG-PLGA polymeric nanoparticles were employed. PEG-PLGA copolymers are amphiphilic and tend to self-assemble in water to yield spherelike nanoparticles with a hydrophobic PLGA inner core and a hydrophilic PEG outer shell. 50 The hydrophobic nature of 1 (log P ¼ 1.1) enabled encapsulation into the lipophilic core of PEG-PLGA (5000 : 30 000 Da, 1 : 1 LA : GA) nanoparticles using the nanoprecipitation method. Various nanoparticle formulations (1 NP 2.5-60 ) were prepared using a range of feeds (2.5-60%), where feed refers to the percentage (w/w) of 1 to polymer. To determine the most suitable formulation for biological studies, the loading and encapsulation efficiency of 1 was calculated by measuring the manganese concentration using ICP-MS. The loading and encapsulation efficiency was highly dependent on the feed (Fig. 4). Optimal encapsulation conditions were achieved at 5% feed (1 NP 5 ), where the encapsulation efficiency ¼ 5.1% and loading efficiency ¼ 0.25%. The diameter of 1 NP 5 was 114.7 nm and the polydispersity was 0.116, according to dynamic light scattering (DLS) studies (Fig. S38 †). These values are comparable to those reported for other polymeric nanoparticles with internalised metal complexes. 51,52 For reference, the diameter of empty PEG-PLGA nanoparticles was 94.5 nm and the polydispersity was 0.086 (Fig. S39 †). As expected the zeta-potential of 1 NP 5 was negative (À14.4 mV) owing to the PEG outer shell (Fig. S40 †). The zeta-potential of empty PEG-PLGA nanoparticles was À33.3 mV (Fig. S41 †). The surface morphology and size distribution of 1 NP 5 was examined further by scanning electron microscopy (SEM). SEM images showed that 1 NP 5 does indeed adopt relatively uniform spherical structures with an average size of 93 AE 6 nm (Fig. S42 †). The latter is in good agreement with the DLS measurements (Fig. S38 †). The nanoparticle formulation, 1 NP 5 was stable in water, PBS, and PBS with 10% FBS (all at pH 7.4) over the course of 72 h at 37 C (Fig. S43 †). Furthermore, 1 NP 5 effectively releases reasonable amounts of its payload over the course of 72 h at 37 C in PBS (pH 7.4) (Fig. S44 †). The stability of 1 NP 5 and its ability to release sufficient amount of the payload bode well for the effective delivery of 1 into breast CSCs.</p><!><p>To gauge the ability of the nanoparticle formulation, 1 NP 5 to deliver its payload, 1, into CSCs, the manganese content in HMLER-shEcad cells treated with 1 NP 5 (1 mM, 12 h) was measured by ICP-MS. The intracellular manganese concentration (297.1 AE 1.8 ppb of Mn/million cells) was 15-fold higher than that observed for HMLER-shEcad cells treated with 1 under identical conditions (18.6 AE 0.7 ppb of Mn/million cells, Fig. 5A). This shows that 1 NP 5 is able to markedly improve the internalisation of 1 into breast CSCs. Polymeric nanoparticles such as PEG-PLGA tend to undergo cell uptake via energydependent endocytosis. 53 To probe the mechanism of cellular uptake of 1 NP 5 , studies were conducted in the presence of endocytosis inhibitors. HMLER-shEcad cells treated with 1 NP 5 (1 mM, 12 h) in the presence of chloroquine (100 mM) or ammonium chloride (50 mM) displayed signicantly (p < 0.01) lower levels of manganese than HMLER-shEcad cells treated with 1 NP 5 alone (Fig. 5A). This suggests that the mechanism of CSC uptake of 1 NP 5 is likely to be endocytosis. Nanoparticles undergoing endocytosis oen end up in acidic endosomes. Therefore we investigated the ability of 1 NP 5 to release its payload under acidic conditions (sodium acetate buffer, pH 5.2 at 37 C over 72 h). The nanoparticle formulation, 1 NP 5 released 1 to a better extent than in PBS (pH 7.4), under similar conditions (Fig. S44 †), indicating that it is capable of releasing its payload in CSCs upon endocytic uptake.</p><!><p>The cytotoxicity of 1 NP 5 against breast CSC-enriched (HMLER-shEcad) and breast CSC-depleted (HMLER) cells was determined using the MTT assay (Fig. 5B). The nanoparticle formulation displayed micromolar toxicity toward both cell lines (IC 50 value ¼ 1.3 AE 0.1 mM for HMLER-shEcad cells and IC 50 value ¼ 1.8 AE 0.1 mM for HMLER cells, Fig. 5B). The nanoparticle formulation, 1 NP 5 was 10-fold less toxic than the payload, 1 for both HMLER-shEcad and HMLER cells. This result was somewhat expected, as polymeric nanoparticles are well known to reduce the toxicities of incorporated compounds. 54 Importantly, 1 NP 5 killed HMLER-shEcad and HMLER cells with equal potency (like 1), showing that the encapsulation of 1 by PEG-PLGA polymeric nanoparticles does not affect its spectrum of Collectively, the monolayer and three-dimensional toxicity studies show that encapsulation of 1 into PEG-PLGA polymeric nanoparticles (1 NP 5 ) does not detrimentally affect its potency towards CSCs. Further cell studies were conducted to determine the mechanism of action of the nanoparticle formulation, 1 NP 5 . Given that the payload, 1 induced CSC death by elevating intracellular ROS levels and inhibiting COX-2, studies were carried to determine if 1 NP 5 was able to retain the mechanism of action of the payload. HMLER-shEcad cells treated with 1 NP 5 (1.3 mM) displayed a massive increase in intracellular ROS levels over the course of 48 h, peaking aer 3 h exposure (9-fold increase) (Fig. S51 †). Under identical conditions, aer 3 h exposure, the payload, 1 increased intracellular ROS levels by only 2-fold (Fig. 3A). The markedly different ROS generating power of the nanoparticle formulation, 1 NP 5 and the payload, 1 is likely to be due to the better internalisation of the nanoparticle formulation (Fig. 5A). Cell viability studies in the presence of NAC (2 mM, 72 h) showed that the potency of 1 NP 5 towards HMLER-shEcad cells decreased signicantly (IC 50 value increased from 1.3 AE 0.1 mM to 4.8 AE 0.1 mM, 3.7-fold, p < 0.05) (Fig. S52 †) suggesting that 1 NP 5 -induced CSC death, like 1induced CSC death, is related to intracellular ROS generation. Immunoblotting studies showed that like 1, 1 NP 5 is able to activate the JNK and p38 pathways (probably due to ROS formation) and induce caspase-dependent apoptosis (Fig. S53 †). Cytotoxicity studies in the presence of z-VAD-FMK (5 mM), showed that the potency of 1 NP 5 towards HMLER-shEcad cells decreased signicantly (p < 0.05, IC 50 value ¼ 4.4 AE 0.1 mM) (Fig. S54 †), conrming that 1 NP 5 induces caspase-dependent CSC apoptosis.</p><p>HMLER-shEcad cells pre-treated with LPS (2.5 mg L À1 for 24 h) and incubated with 1 NP 5 (0.6-1.3 mM for 48 h) displayed a noticeable decrease in COX-2 levels according to ow cytometric studies (Fig. S55 †). This is comparable to the COX-2 downregulation induced by the payload, 1 (Fig. S35 †). Cytotoxicity studies in the presence of PGE2 (20 mM, 72 h) showed that the potency of 1 NP 5 towards HMLER-shEcad cells decreased signicantly (IC 50 value increased from 1.3 AE 0.1 mM to 2.7 AE 0.02 mM, 2.1-fold, p < 0.05) (Fig. S56 †) suggesting that 1 NP 5 , like 1, induces COX-2-dependent CSC death. Collectively, the mechanistic studies shows that encapsulation of 1 by PEG-PLGA polymeric nanoparticles (1 NP 5 ) does not alter its cellular properties, which augers well for further pre-clinical development.</p><!><p>In summary we report the synthesis, characterisation, anti-CSC properties, and encapsulation of a manganese(II)-3,4,7,8tetramethyl-1,10-phenanthroline complex, 1 bearing a NSAID moiety, namely diclofenac. The manganese(II) complex, 1 is stable in biologically relevant solutions, including PBS with and without glutathione (a cellular reductant) and cell culture media, under physiological conditions. The manganese(II) complex, 1 displays equally toxicity against bulk breast cancer cells (HMLER) and breast CSCs (HMLER-shEcad), in the submicromolar range, suggesting that it has the potential to eliminate heterogeneous breast tumour populations (made up of bulk cancer cells and CSCs) with a single dose. The analogous copper(II) complex, 3 exhibited preferential potency towards bulk breast cancer cells (HMLER) over breast CSCs (HMLER-shEcad), implying that the metal (within this system) plays a determinant role in bulk cancer versus CSC toxicity. Extraordinarily, the manganese(II) complex, 1 inhibited HMLER-shEcad mammospheres formation and viability favourably over non-tumorigenic MCF10A spheroids (12-fold selectivity), indicating that 1 can potentially remove breast CSCs with reduced toxicity towards normal breast epithelial cells. In contrast, salinomycin, one of the leading anti-breast CSC agents identied thus far, killed HMLER-shEcad mammospheres and MCF10A spheroids equipotently. Detailed mechanistic studies showed that the cytotoxic mechanism of action of 1 involved intracellular ROS generation and COX-2 inhibition. The manganese(II) complex, 1 generated signicantly higher levels of ROS inside CSCs than the analogous copper(II) complex, 3 proving that manganese(II)-phenanthroline complexes are, in general, better intracellular ROS generators than their copper(II) counterparts.</p><p>In an attempt to improve CSC delivery, the manganese(II) complex, 1 was encapsulated into PEG-PLGA nanoparticles. The optimal loading conditions were obtained when the feed was set at 5%, 1 NP 5 . This yielded predominantly spherical nanoparticles with a diameter of ca. 100 nm. Importantly, 1 NP 5 was stable and able to release the payload, 1 in biologically relevant solutions over the course of 72 h. The nanoparticle formulation delivered 15-fold higher levels of manganese into CSCs (via an endocytotic pathway) than the free payload. Encapsulation of 1 into PEG-PLGA nanoparticles signicantly reduced overall toxicity towards bulk breast cancer cells (HMLER) and breast CSCs (HMLER-shEcad). The nanoparticle formulation retained the ability of the payload to indiscriminately kill bulk breast cancer cells and breast CSCs and thus its potential to remove whole breast cancer cell populations (including bulk cancer cells and CSCs) with a single dose. Mechanistic studies proved that 1 NP 5 exhibited a similar mechanism of action as 1 -CSC death by intracellular ROS elevation and COX-2 inhibition. This is a highly desirable trait, as one of the major drawbacks associated to nanoparticle encapsulation as a strategy for drug delivery is the potential discrepancy in cellular mechanism of action of the nanoparticle formulation and its payload.</p><p>Overall this study highlights the expanding potential of redox modulating metal complexes as anti-CSC agents and opens the door for the development of other manganese complexes as CSC-potent agents. Furthermore, in light of the ndings reported in this manuscript, the anti-CSC potential of ROS-generating manganese complexes previously reported in the literature should be determined as they could provide promising anti-CSC leads. As well as presenting the rst manganese complex to show therapeutically relevant CSC potency, we demonstrate that polymeric nanoparticles can be used to effectively deliver manganese complexes into CSCs.</p><!><p>There are no conicts to declare.</p>

Royal Society of Chemistry (RSC)

Biomimetic Total Synthesis of (\xc2\xb1)-Griffipavixanthone via a Cationic Cycloaddition\xe2\x80\x93Cyclization Cascade

We report the concise, biomimetic total synthesis of the dimeric, Diels\xe2\x80\x93Alder natural product griffipavixanthone from a readily accessible prenylated xanthone monomer. The key step utilizes a novel intermolecular [4+2] cycloaddition\xe2\x80\x93cyclization cascade between a vinyl p-quinone methide and an in situ generated isomeric diene promoted by either Lewis or Br\xc3\xb8nsted acids. Experimental and computational studies of the reaction pathway suggest that a stepwise, cationic Diels\xe2\x80\x93Alder cycloaddition is operative.

biomimetic_total_synthesis_of_(\xc2\xb1)-griffipavixanthone_via_a_cationic_cycloaddition\xe2\x80\x93

<p>Dimeric xanthones are a class of natural products with a wide range of structural diversity and biological activity, which underscores these compounds as attractive synthetic targets.1 Among these compounds, griffipavixanthone (1) possesses a unique structure wherein the monomeric tetrahydroxyxanthone units are linked via an apparent Diels–Alder-derived bicyclic framework rather than the commonly encountered aryl–aryl or C–O bond (Figure 1).2 Several related dimeric Diels–Alder xanthones, namely garciobioxanthone (2),3 bigarcinenone B,4 and garcilivins A and C (not shown),5 wherein the monomeric xanthone units are linked by a single cyclohexene ring, have also been isolated and characterized.</p><p>Although biological activities of these dimeric xanthones have not been extensively reported, 1 is reported to possess antioxidant properties,6 highlighted by in vitro inhibition observed against xanthine oxidase.7 In addition, 1 is active against several cancer cell lines,2,8,9 and has recently been shown to downregulate the RAF pathway in esophageal cancer cells.10 Herein, we report the first chemical synthesis of 1 utilizing a biomimetic [4+2] cycloaddition–cyclization cascade between a monomeric, vinyl p-quinone methide (p-QM) dienophile and an in situ generated, isomeric diene promoted by either zinc(II) iodide (ZnI2) or trifluoroacetic acid (TFA).</p><p>Biosynthetically, the cyclohexene core of 1 is likely formed via Diels–Alder cycloaddition between a prenylated tetrahydroxyxanthone and a related, formally dehydrogenated 1,3-butadiene,2 both of which may be derived in nature from the monomer garcinexanthone C.11 With our continuing interest in biomimetic syntheses of Diels–Alder natural products, this insight inspired us to consider development of a novel reaction cascade wherein both diene and dienophile could be generated from a single prenylated monomer.12 Retrosynthetically, we envisioned that 1 may be accessed via intramolecular arylation between the p-QM and xanthone moieties of cycloadduct 3 (Scheme 1).13,14 p-QM 3 may be derived from oxidation of monomer 6 and subsequent [4+2] cycloaddition of the resulting vinyl p-QM 4 with the corresponding isomeric diene 5 generated in situ (vide inf ra).15</p><p>To test our biomimetic proposal, we first prepared the requisite prenylated xanthone monomer (Scheme 2). Condensation of 1,3,5-trimethoxybenzene with 2,3-dihydroxy-4-methoxybenzoic acid in the presence of Eaton's reagent at elevated temperature readily provided the desired xanthone as a mixture of 5-hydroxy- and 5-mesyloxyxanthones.16 Treatment of this crude mixture with methanolic potassium hydroxide allowed for smooth demesylation and isolation of the xanthone as the corresponding potassium salt, which was subsequently O-alkylated with prenyl bromide to afford prenyloxyxanthone 7 (53% yield, 3 steps) on a decagram scale. Thermal rearrangement of 7 in refluxing N,N-dimethylaniline (DMA) provided the prenylxanthone monomer 8 in 47% yield.17</p><p>Our preliminary efforts toward assessing oxidation and isomerization of the key monomeric vinyl p-QM intermediate first focused on trimethoxyxanthone substrate 8 in order to eliminate potential side reactions resulting from multiple unprotected phenols. We found that silver(I) oxide (Ag2O) cleanly oxidized 8 to the desired p-QM 9 at ambient temperature in CH2Cl2 as solvent (Scheme 3).18 While treatment of 9 with silica gel, conditions known to isomerize prenylated o-quinones to p-QMs,19 typically led to complex product mixtures, exposure of 9 to catalytic amounts ofN,N-diisopropylethylamine (DIPEA) promoted smooth conversion to 1,3-butadiene 10.20 To the best of our knowledge, this represents the first example of isomerization of a vinyl p-QM to the corresponding diene; however, analogous formal dehydrogenation of prenylated derivatives to 1,3-butadienes has previously been reported.21</p><p>The isomerization of 9 to 10 permitted exploration of the proposed [4+2] cycloaddition. During isomerization experiments, only trace cycloaddition products were observed, even at elevated temperatures. Furthermore, conversion of 9 to 10 did not appear to be reversible under the isomerization conditions. These observations led us to conclude that activation of 9 was necessary for cycloaddition and the rate of cycloaddition must be faster than the conversion of 9 to 10.</p><p>To address these requirements, Lewis22 or Brønsted23 acid promoters appeared to be required, as they have been shown to activate p-QMs toward subsequent reactivity. Indeed, exposure of 9 to catalytic amounts of a number of Lewis acids led to complete consumption of starting material (CH2Cl2 as solvent) as determined by 1H NMR analysis.24 Although conversion of 9 led to numerous reaction products under many of these conditions, the unequal formation of two products with doublets at δ 8.74 and 8.46 ppm (1H NMR) suggested formation of p-QM-containing cycloadducts with varying levels of diastereoselectivity.24 Full spectroscopic characterization revealed that these species were indeed the desired [4+2] cycloadducts. After screening both Lewis acid and solvent, we were able to generate isolable quantities of each cycloadduct (Scheme 4). For example, use of lanthanum(III) triflate (La(OTf)3) as catalyst in a 2:1 mixture of CH2Cl2 and hexafluoroisopropanol (HFIP) with DIPEA as additive at 30 °C selectively provided cycloadduct 11a possessing an anti stereochemical relationship of the aromatic ring systems in 25% yield. The structure of 11a was confirmed via single-crystal X-ray analysis (Figure 2) and reveals a significant deviation from planarity (21.3°) between the exocyclic alkene of the p-QM and the neighboring carbonyl. Use of ZnI2 resulted in the highest selectivity and yield for the diastereomeric syn-cycloadduct 11b (crude ratio 11b:11a = 1.0:0.96) providing 11b and 11a in 17 and 18% yields, respectively. Furthermore, cycloadduct 11a was found to be unreactive to treatment with ZnI2, with no evidence of cycloadduct reversibility or epimerization, suggesting metal-ion-dependent stereoselectivity to access 11a/b from 9.24</p><p>While numerous Lewis acids generated 11b to varying degrees, subsequent arylation to yield the griffipavixanthone core was not readily observed beyond trace quantities. We therefore aimed to improve the Zn(II)-catalyzed conditions which provided the highest observed selectivity for 11b, and develop a one-pot cycloaddition–cyclization cascade. After examining numerous conditions, we determined 9 to be a poor substrate for the desired transformation, which required 30 mol% of ZnI2 and elevated (50 °C) temperature in 1,2-dichloroethane (DCE) to provide griffipavixanthone hexamethyl ether 12 in 4% yield (Scheme 5). We considered that the low yields may in part be due to intermolecular arylation of the electron-rich ring of 9 and the quinone methide moiety of a second molecule of 9, as intermolecular arylations of p-QMs with electron-rich aromatics in the presence of Lewis acids have been reported.25,26 Indeed, reaction of 9 with excess 1,3,5-trimethoxybenzene in the presence of ZnI2 cleanly afforded 1,6-arylation product 13 in 68% yield (eq 1). (1)</p><p>To address this issue, we hypothesized that demethylation of 9 ortho to the xanthone carbonyl would remove electron density from the phloroglucinol-type ring through hydrogen bonding, thereby reducing this decomposition pathway. Indeed, we found that the di-O-methyl-protected vinyl p-QM 1424 was superior to 9, requiring only 15 mol% of ZnI2 at 40 °C for 16 h in DCE (0.05 M) to provide griffipavixanthone tetramethyl ether 15 in 15% isolated yield. With increased amounts of precursor 15 in hand, full demethylation to (±)-1 was achieved in 32% yield using a modified potassium carbonate/p-thiocresol protocol in N,N-dimethylacetamide (N,N-DMA) as solvent (Scheme 5).27 Spectral data for synthetic (±)-1 were in agreement with data previously reported for the natural product.2</p><p>In an effort to understand the direct conversion of 14 to 15, experiments designed to interrogate the reactivity of cycloadduct intermediates leading to 15 were performed. We found that treatment of monomer 14 with 15 mol% of LaCl3·2LiCl provided cycloadducts 16a/b in 32% combined yield (crude ratio of 16a:16b = 1.5:1.0) with minimal conversion to 15 observed. Spectral characterization and comparison with 11a/b revealed that anti-cycloadduct 16a was the major product formed upon reaction of 14 with both LaCl3·2LiCl and ZnI2, in contrast with that of 9. Furthermore, cycloadducts 16a/b were found to be substantially more reactive than 11a/b, as treatment of both 16a and 16b with 30 mol% ZnI2 led to 15 in 73% (brsm) and 80% isolated yield, respectively (Scheme 6). Similar results were also observed in the presence of 30 mol % of TFA, as both 16a and 16b yielded 15 in 78% and 82% yield, respectively.28 In light of these results, we anticipated that 14 may also yield 15 under similar conditions. Indeed, treatment of 14 with 30 mol% of TFA at 35 °C for 18 h provided 15 in an improved 21% isolated yield, demonstrating that both Lewis and Brønsted acids are capable of promoting the observed cycloaddition–cyclization cascade to yield the griffipavixanthone core from monomer 14.</p><p>The formation of griffipavixanthone precursor 15 from either 14 or a 16a/b mixture (cf. Scheme 6) supports an ionic Diels–Alder reaction cascade in which six-membered ring formation is reversible.29,30 Figure 3 outlines a plausible mechanism which is supported by DFT computations.24,31,32 After acid-catalyzed isomerization of 14 to 10a, the intermediate cation 17 may combine with 10a to afford allylic cations, represented as 18a–c. Cation 18a was the lowest energy structure located; this should interconvert with an energetically favorable xanthonium ion 19 which arises from capture of the benzyl cation by the nucleophilic carbonyl of the pendant xanthone. Isomeric allyl cations 18b and 18c were found to possess the correct stereochemistry and conformation required to reversibly progress to stereoisomeric benzyl cation intermediates 20 and 21 with barriers of 9.2 and 17.5 kcal/mol, respectively. The cascade is completed by intramolecular arylation to yield arenium ions 22 or 23 through TS4 or TS5, respectively. Selective formation of griffipavixanthone precursor 22 vs epimer 23 is consistent with the lower predicted barrier (16.5 vs 21.0 kcal/mol) for the irreversible arylation step. This difference is ascribed in part to the lower strain energy required to yield the cis- vs trans-fused polycyclic ring systems of 22 and 23. We note that while these computational results are in agreement with our experimental findings, we cannot at this time rule out alternative stereochemical interconversion of 16a and 16b via elimination followed by regioselective protonation or 1,2-hydride shift33 pathways of intermediates 20 and 21.</p><p>In summary, we have developed the first synthesis of the Diels–Alder natural product griffipavixanthone (1) in racemic form. The unique polycyclic core of 1 was prepared under mild Lewis or Brønsted acidic activation via a multistep, one-pot [4+2] cycloaddition–cyclization cascade involving (1) isomerization of an isolable vinyl p-QM monomer to the corresponding 1,3-butadiene in situ; (2) stepwise, reversible cationic [4+2] cycloaddition between the vinyl p-QM monomer and newly generated 1,3-butadiene; and (3) intramolecular arylation of a p-QM-containing cycloadduct. Experimental and computational investigations of the reaction mechanism support the proposed cationic reaction cascade. Further studies related to the asymmetric synthesis of griffipavixanthone are under investigation and will be reported in due course.</p><p> ASSOCIATED CONTENT </p><p> Supporting Information </p><p>Experimental details, computational data, and characterization data for new compounds (PDF)</p><p>X-ray crystallographic data for 11a (CIF)</p><p>The authors declare no competing financial interest.</p>

PubMed Author Manuscript

Peptide–polymer ligands for a tandem WW-domain, an adaptive multivalent protein–protein interaction: lessons on the thermodynamic fitness of flexible ligands

Three polymers, poly(N-(2-hydroxypropyl)methacrylamide) (pHPMA), hyperbranched polyglycerol (hPG), and dextran were investigated as carriers for multivalent ligands targeting the adaptive tandem WW-domain of formin-binding protein (FBP21). Polymer carriers were conjugated with 3–9 copies of the proline-rich decapeptide GPPPRGPPPR-NH2 (P1). Binding of the obtained peptide–polymer conjugates to the tandem WW-domain was investigated employing isothermal titration calorimetry (ITC) to determine the binding affinity, the enthalpic and entropic contributions to free binding energy, and the stoichiometry of binding for all peptide–polymer conjugates. Binding affinities of all multivalent ligands were in the µM range, strongly amplified compared to the monovalent ligand P1 with a KD > 1 mM. In addition, concise differences were observed, pHPMA and hPG carriers showed moderate affinity and bound 2.3–2.8 peptides per protein binding site resulting in the formation of aggregates. Dextran-based conjugates displayed affinities down to 1.2 µM, forming complexes with low stoichiometry, and no precipitation. Experimental results were compared with parameters obtained from molecular dynamics simulations in order to understand the observed differences between the three carrier materials. In summary, the more rigid and condensed peptide–polymer conjugates based on the dextran scaffold seem to be superior to induce multivalent binding and to increase affinity, while the more flexible and dendritic polymers, pHPMA and hPG are suitable to induce crosslinking upon binding.

peptide–polymer_ligands_for_a_tandem_ww-domain,_an_adaptive_multivalent_protein–protein_interaction:

<!>Introduction<!><!>Introduction<!>Selection of a bivalent protein receptor as a target<!>Selection of polymer carriers and synthesis of multivalent ligands<!><!>Selection of polymer carriers and synthesis of multivalent ligands<!>Binding of multivalent peptide–polymer conjugate to the tandem WW domain<!><!>Molecular dynamics simulations of multivalent ligands<!><!>Molecular dynamics simulations of multivalent ligands<!><!>Molecular dynamics simulations of multivalent ligands<!><!>Conclusion<!>

<p>This article is part of the Thematic Series "Multivalency as a chemical organization and action principle".</p><!><p>Multivalency is a general principle in nature for increasing the affinity and specificity of ligand–receptor interactions [1]. Multivalent binding is characterized by the cooperative, over-additive enhancement of binding affinities of ligands and receptors in a defined spatial arrangement. The strongest affinity enhancement can be expected in the case of a perfectly fitting, rigid arrangement of ligands and receptors (Figure 1A). In such cases the affinity of the multivalent ligand can be potentiated by the degree of multivalency. Prominent examples for this perfect fit have been reported reaching an exponential binding increase [2]. Rigid scaffolds can be used to present ligands in defined spatial arrangements and thus can be exploited to investigate the distances between receptor sites as "molecular ruler" [3–4].</p><!><p>Comparing the entropy loss during ligand–receptor interactions in dependence of the rigidity of the backbone.</p><!><p>Many multivalent receptors in nature, however, are characterized by the flexible arrangement of receptor sites and the resulting relative mobility of binding domains seems to have a significant impact on the proper functioning of these proteins [5]. Flexible arrangements of receptor sites can result from different scenarios. In many proteins flexibility is introduced by regions of inherent structural mobility, e.g., by so-called unstructured regions inserted between the receptor domains of a multireceptor protein. Alternatively, the relative mobility of binding sites is realized by their embedding into membranes giving them a certain degree of freedom to move in the plane of the membrane, or by incorporation into dynamic multiprotein complexes.</p><p>Design of potent multivalent ligands for flexible receptor arrangements is a considerable challenge, as the flexibility of multivalent ligands and the flexibility of receptors have to be matched in order to balance enthalpic gain with entropic loss of the system. In such a setting, a rigid multivalent ligand binding to a flexible receptor can be expected to reduce the entropy of the system upon binding, and thus will result in a partial or complete loss of the multivalent affinity enhancement. For example, the targeting of flexible protein receptors with ligands attached to a rigid DNA-backbone has been reported to be unsuccessful and no preferred ligand distance was found for this "molecular ruler" for flexible divalent protein targets [4].</p><p>Recently, we have introduced multivalent peptide–polymer conjugates as a chemical tool to inhibit protein–protein interactions in living cells [6]. As demonstrated for the pro-apoptotic BH3-peptides, multivalent presentation of monovalent ligand peptides can potentiate the activity of the peptide at identical overall peptide concentrations. Moreover, attachment of bioactive peptides to polymers strongly enhanced their stability and protected them from proteolysis [7–8]. The construction of peptide-polymer conjugates with additional cell-penetrating peptides attached [9] enabled the smooth intracellular delivery of the conjugated polymer; as a third component fluorescent dyes [10] were coupled to the polymers simultaneously with the bioactive and the cell-penetrating peptides in order to enable the monitoring of cellular uptake and intracellular distribution of the peptide–polymer conjugate.</p><p>Until now, various polymer carriers have been used for the construction of peptide–polymer conjugates [11–12], however, a systematic comparison of the different polymeric materials with respect to the structure–activity relationships is missing so far. The goal of this contribution is to synthesize and compare flexible multivalent ligands for an adaptive, divalent receptor as a protein target. As a model protein the tandem-WW-domain of the pre-mRNA splicing factor formin binding protein 21 (FBP21) was selected [13–15]. Considering the importance of FBP21 in the activation of RNA splicing, successful ligands should be valuable tools to interfere with FBP21-dependent splicing events. Several multivalent ligands were synthesized on the basis of various polymer supports differing in their chemical structure, backbone flexibility, morphology, and ligand loading. The obtained materials were then investigated in order to contribute to the understanding of structure–activity relationships of polymeric ligands. For this purpose, the thermodynamics and the stoichiometry of protein binding events were determined experimentally for all multivalent ligands. Finally, atomistic molecular dynamics simulations were conducted in order to rationalize the observed differences on a microscopic level and to derive general principles for the design of optimized multivalent ligands of flexible protein targets.</p><!><p>As a representative example for a protein containing a bivalent domain architecture connected with a flexible linker the tandem WW-domains of the protein FBP21 were selected. FBP21 is a protein component of the spliceosome, the multiprotein complex in the nucleus of cells responsible for the processing of primary RNA-transcripts. The two WW domains of FBP21 bind to proline-rich sequences contained in numerous proteins including the core splicing protein SmB/B´and several splicing factors including splicing factor 3B4 (SF3B4) [16–17]. Recently, the enhanced binding affinity of bivalent and tetravalent peptide ligands to this protein was described suggesting that multivalent ligands may play a significant role also in living cells. In addition, several interaction partners of FBP21 have been profiled by SILAC/MS [18]. As monovalent peptide ligands for each of the two WW domains proline-rich sequences (PRS) of the group Rb have been identified, in which the proline residues are flanked by arginine (R in one-letter-code) [16,19]. Multivalent arrangements of these monovalent ligands therefore could serve as potent inhibitors of FBP21-interactions and could be used for the inhibition of FBP21 function. As a monovalent peptide ligand the decapeptide amide GPPPRGPPPR-NH2 (P1) was selected and synthesized on Rink amide polystyrene resin. For attachment to the polymer carriers the N-cysteinylated peptide CGPPPRGPPPR-NH2 (P2) was prepared, containing a free N-terminus in order to enable the attachment to polymers via native chemical ligation or Michael addition to maleimide residues.</p><!><p>Three biocompatible polymers with different chemical structure, backbone flexibility and polymer morphology were selected as multivalent ligand carriers, two linear polymers and one dendritic polymer (Scheme 1). Linear poly(N-(2-hydroxypropyl)methacrylamide) (pHPMA) possesses a C2 repeating unit with three fully rotatable bonds, which should convey – compared to the other polymers employed in this study – high backbone flexibility to this carrier. Reactive pHPMA was prepared in a copolymerization of HPMA and the thioester-containing building block N-methacryloyl-β-alaninyl-S-benzyl thioester under reversible addition–fragmentation chain-transfer (RAFT) conditions yielding a thioester-containing copolymer with 13.3 kDa and polydispersity of 1.2, which we denominated as NCL-polymer [10]. NCL-polymer was converted into multivalent peptide–polymer conjugates pHPMA-1 and pHPMA-2 via native chemical ligation with the N-cysteinylated peptide CGPPPRGPPPR-NH2 (P2). In contrast, the second carrier molecule, hyperbranched polyglycerol (hPG) was selected as a dendritic polymer. While the backbone of PG is relatively flexible by itself, the dendritic structure of hPG can be expected to limit the flexibility of attached ligands compared to a linear polymer and might induce a more globular arrangement of the ligands. The hPG polymer carrier was synthesized via an anionic ring-opening polymerization of glycidol [20] and also modified with maleimido groups by reaction with N-3-chloropropyl maleimide for ligand attachment.</p><!><p>Selection of three polymer carriers differing with respect to backbone flexibility, and morphology and used for the construction of peptide–polymer conjugates.</p><!><p>Finally, dextran, a polysaccharide containing α-1,6-linked D-glucose as repeating unit, was selected as the second linear carrier. The D-glucose units in the polysaccharide are fixed in the 1C4 chair conformation and thus can be expected to rigidify the polymer backbone compared to the other two polymers, leaving only two freely rotatable bonds per building block. Structural studies with dextran suggested a helical structure as the lowest energy conformations of this polymer [21]. Dextran was used as a linear polymer with an average MW of either 10 kDa (for Dex-1 and Dex-2) or 50 kDa (for Dex-3), both with a polydispersity index of 1.5. Under basic conditions the linear polysaccharide was alkylated with acrylamide selectively in the 2-positions of the sugars. The resulting 2-O-carboxyethyl dextran (2-O-CE-dextran) was further converted by condensation with 2-N-maleimido-ethylamine and N-ethyl-N´-dimethylaminopropylcarbodiimide (EDC) [6]. The monovalent ligand peptide 2 was attached to the dextran carriers by nucleophilic addition of the thiols to the maleimide double bond furnishing peptide–polymer conjugates Dex-1, Dex-2, and Dex-3.</p><p>Peptide loadings of all obtained peptide–polymer conjugates were determined by quantitative amino acid analysis and ranged from 3 to 9 peptides per polymer corresponding to peptide loading densities (percentage of ligand-carrying monomers) between 3 and 10%.</p><!><p>Binding studies with peptide–polymer conjugates were conducted employing isothermal titration calorimetry (ITC). This method enables the determination of the binding affinity of the multivalent ligands and elucidates the composition of the free energy of binding from the enthalpic and entropic contributions. In addition, the method can be used to determine the stoichiometry of the formed protein–ligand complex indicating the ratio of peptide ligand molecules relative to each protein binding site thereby giving valuable insights into the multivalency of binding and/or the degree of crosslinking. Thus, the method enables the identification of polymer–protein aggregates containing several polymers and proteins in a complex. No precipitation of the multicomponent aggregates that interfered with ITC measurements was observed during the experiments.</p><p>ITC-analysis (Figure 2) of the binding of multivalent peptide–polymer conjugates yielded KD values either corresponding to the polymer concentration or relative to the overall peptide concentration (N*KD). A comparison of the binding affinity of the monovalent peptide ligand P1 and its N-acetylated derivative Ac-P1 with seven multivalent peptide ligands to the tandem WW-domain revealed a strong enhancement of the binding through multivalency (Table 1, Figure 3). While the peptide alone bound with a dissociation constant (KD) of larger than 1 mM [16], all multivalent peptide-polymer conjugates possessed KD values below 10 µM. Though all KD values of multivalent ligands were in the same concentration range (i.e., between 1.2 and 7 µM), concise differences were revealed for the thermodynamic composition of KD values (Figure 2). While the ligands based on polymethacrylamide displayed moderate enthalpic and almost negligible entropic contributions , all polyhydroxy-based peptide–polymer conjugates showed significantly stronger generation of heat through binding (enthalpy) together with a pronounced loss in entropy. Binding in all cases was driven mainly by enthalpy, which clearly outweighed the observed entropy loss. In the seven peptide–polymer conjugates investigated, increased loading density of ligands led consistently to increased affinity of the multivalent ligand (Table 1). The most significant difference between dextran and the two other polymer carriers was the stoichiometry of the formed peptide-polymer–protein complex. Inspection of the test solution revealed the formation of a colloidal suspension/dispersion both for pHPMA and for hPG-based peptide conjugates indicating the formation of insoluble aggregates possibly generated through crosslinking. Corresponding to the observed colloidal suspension/dispersion the stoichiometry of peptide ligands per protein receptor resulting from the ITC experiments was >2 for each of either pHPMA or hPG-based material, most pronounced for pHPMA with n = 2.6–2.8. Dextran-based conjugates displayed a ligand stoichiometry of 1.4 for the most potent multivalent ligand with a KD of 1.2 µM, Dex-2. No correlation between ligand density and stoichiometry became evident from the recorded data, however, the observed correlation between low binding stoichiometry, increased binding affinity, and increased binding enthalpy seems to suggest the prevalence of a bivalent binding mode for the complex of Dex-2 and tandem-WW-FBP21, which is supported also by the solubility of the non-crosslinked peptide-polymer–protein complex.</p><!><p>Representative ITC-measurements conducted at 8 °C with the peptide–polymer conjugates A) pHPMA-1 and B) Dex-2 showing an increase in affinity for the interaction of Dex-2 with the FBP21 tandem WW domains. The upper part shows differential heating power (Δp) changes upon injection of peptide–polymer conjugates into the protein; bottom part shows integrated and normalized heat of reaction plotted against peptide/protein molar ratio; binding isotherms are fitted with a 1:1 binding model.</p><p>ITC measurements of peptide–polymer conjugates with tandem WW domain of FBP21.</p><p>aDextran, hyperbranched PG and poly(HPMA) coupled with the N-cysteinylated peptide CGPPPRGPPPR (P2); bN: number of ligands (number of repeating units in the polymeric scaffolds); cbinding affinities of peptide–polymer conjugates; dbinding affinities measured by ITC related to overall peptide concentrations.</p><p>Enthalpic and entropic contributions to the free energy of binding processes of multivalent peptide-polymer conjugates and the tandem WW domain of protein FBP21 determined at 8 °C by ITC measurements.</p><!><p>In order to better understand our experimental observations regarding binding affinities, enthalpic/entropic energy contributions, and binding stoichiometries from a molecular point of view, the three polymer carriers were investigated using atomistic molecular dynamics simulations. Each polymer was represented by one model parameterized in accordance with the AMBER force field [22]. The concentration ratios of peptide ligands and monomeric units were fit to lab conditions such that each polymer was carrying three ligands. In contrast to the linear polymer models of dextran and pHPMA with 10 and 12 units between any two successive ligands, respectively, the hPG configuration was generated randomly with the aid of a probabilistic hPG building algorithm as described previously [23]. After some preparatory relaxation steps, each of the three polymers underwent three explicit solvent molecular dynamics (MD) simulations of 100 ns length serving as production runs. The first 30% of the time steps were considered as an unrestricted equilibration phase and consequently omitted whereas from the remaining time series several promising structural and physical descriptors were determined. For all simulations and analytical calculations the Gromacs software suite was utilized [24]. Table 2 and Figure 4 show these theoretical results averaged over time as well as the three runs per polymer.</p><!><p>Molecular dynamics simulations of the protein target and the multivalent polymeric ligands.</p><p>aExpected mean distance values (calculated by a radial distribution function); mean distance between two peptide ligands on a polymer chain measured between bthe N-terminal sulfur atoms of the Cys-residues at their linking site and cthe C-terminal nitrogen atoms of the Arg residue as the farthest distance between peptide and polymer backbone; average potential energy regarding dthe affinity of the peptide to the polymer and ethe solvation energy of the peptide; fratio of the peptide-polymer conjugates volume and the appropriate sphere.</p><p>MD simulations over time (0–100 ns) yielding A) the mean sulfur distance between two peptides at their linking site, B) the mean nitrogen distance between two peptides at the farthest distance between peptide and polymer chain C) the frequency of observed peptide–polymer distances in dependence of the polymer backbone pHPMA, hPG and dextran, respectively.</p><!><p>Structural properties and descriptors. Dividing the Euklidean distance between two successive peptide attachment points by the number of bonds in between (i.e., between the N-terminal nitrogen atoms of the cysteinylated peptide P2 in the case of pHPMA, and the Cys-sulfur in the cases of both hPG and dextran) yields relative distances which indicate that the peptide ligands in pHPMA are further apart than in dextran and hPG, while the variance of the peptide positions in pHPMA is higher than in the two hydroxyl polymers (Table 2, Figure 4A). Next, we were interested in the distances between the C-terminal positions of the peptide ligands measured between the C-terminal amide nitrogens of the peptides (Table 2, Figure 4B). Here, the peptides on dextran were found to be closer (2.9 nm) to each other than in pHPMA (3.4) and hPG (3.7 nm). The larger distance in hPG might be related to the hypervalent morphology of this carrier, which possibly limits the proximity of attached ligands. Expected values of averaged (over time and atoms) radial distributions (correlating with normalized mean distances) of polymer atoms around peptide atoms clearly reveal a higher polymer-peptide proximity for the dextran system (1.23 nm) than for pHPMA (1.41 nm) and hPG (1.56 nm). Considering the statistical character of the underlying molecular ensemble, the time-averaged radial distribution function (rdf) values indicate a smaller ratio of the fraction of time steps with outstretched peptides (which are more accessible for binding with the tWW domain) and the fraction of time steps characterized by a contracted structure in case of peptides associated with the dextran polymer (Figure 4C). Thus, ligands attached to pHPMA or hPG are more often available for protein binding than those linked to dextran. As a consequence, multiple simultaneously outstretched peptides are even less likely to emerge in case of dextran in comparison with the other polymers. Moreover, after having bound the first protein and due to substantially smaller peptide end-to-end distances given with dextran, its next outstreched peptide will rather bind a free tWW domain of the same protein than of another one which clearly confirms the stoichiometric results. This binding mode is illustrated in Figure 5.</p><!><p>MD simulation image showing the interaction of two dextran–peptide conjugates with three tandem WW domains of FBP21 illustrating the intramolecular mode of binding.</p><!><p>Another descriptor for the spatial arrangement that we denote as the peptide polymer's globularity was defined as the quotient of the volume under the multivalent ligand's solvent-accessible surface area (SASA) and the volume of the minimal sphere incorporating the entire molecule (Table 2). Not unexpectedly, the conformation of the peptide conjugate with the dendritic polymer hPG yields a significantly higher globularity (0.1) compared to those associated with pHPMA (0.04) or dextran (0.07). Regarding these two linear carriers only, the higher globularity of the dextran-based ligand is in good agreement with that material's peptide–polymer distance.</p><p>Thermodynamic properties. From a physical point of view, the significantly varying mean peptide–peptide and peptide–polymer distances are mainly attributed to molecular interactions between the involved atoms. For this reason we calculated non-bonded interaction energies between peptide atoms and both polymer and solvent atoms as the sum of van-der-Waals and electronic contributions (Table 2) While the interaction energies between peptides and solvents are, as expected, nearly identical for all three systems, the interaction of peptide atoms regarding polymer atoms amounts to substantially different values for the three carrier materials. With −913 kJ/mol dextran yielded the by far lowest energy compared with those peptides attached to the two high-stoichiometry polymers (−515 kJ/mol and −783 kJ/mol). Since lower energies correspond to preferential states, the interaction energy can be interpreted as a measure for a state's preference. In general, preferential states are characterized by (negative-signed) attractive forces dominating over (positive-signed) repulsive forces. Hence, according to these results, the peptide is more attracted by the dextran carrier than by the two others most likely causing the small expected polymer–peptide distance and possibly the small peptide end-to-end distances in case of dextran.</p><p>Finally, the molecular dynamics simulations of the peptide–polymer conjugates were compared with those of dimeric complexes with a bivalent binding mode in order to calculate the entropic loss of both the protein and of the peptide–polymer conjugates themselves (Table 3). Interestingly, in all three cases the major contribution to the entropic loss was on the side of the protein, the decrease in entropy on the polymer side was comparably small. Though bivalent binding modes are strongly favoured through enthalpic gain, the free energy gain is limited by the entropy loss, most likely caused by the flexibility of the linker and thus a larger number of alternative conformational states of the protein receptor.</p><!><p>Calculated changes in entropy during binding of the multivalent polymeric ligands to the bivalent receptor by molecular dynamics simulations.</p><!><p>All three investigated biocompatible polymers, namely linear poly(N-2-hydroxypropyl)methacrylamide (pHPMA), hyperbranched polyglycerol (hPG), and linear 2-carboxyethyldextran are suited for the construction of peptide–polymer conjugates, which can be used as potent multivalent ligands for a flexible protein–protein interaction site here exemplified by the tandem WW-domains of FBP-21. 2-Carboxyethyldextran furnished peptide–polymer conjugates with significantly higher binding affinity than the two other carriers. The observed binding modes of the three carriers were distinct. Dextran-based conjugates formed preferably bivalent, soluble complexes with a stoichiometry of <2 peptide ligands per protein binding site, while pHPMA and hPG formed colloidal suspensions/dispersions with stoichiometries >2 ligands per binding site. Molecular dynamics calculations suggested that conjugates with multivalently presented peptides on dextran occupy conformations in which two conjugated peptides are closer to each other and to the polymer backbone, corresponding to the calculated stronger peptide-polymer interaction. From the study it can be supposed that the simulated conformational space of the investigated peptide–polymer conjugates indeed correlates with the experimentally observed binding properties of the multivalent ligands. The construction and experimental investigation of further peptide–polymer conjugates will show, whether the results reported here will be helpful for the construction of even more potent multivalent and/or crosslinking ligands for protein–protein interaction sites and whether the ligands active in the protein binding assay can be further developed toward intracellularly delivered and intracellularly active PPI-inhibitors of the tandem WW-domain.</p><!><p>Experimental.</p>

PubMed Open Access

Disturbances in PP2A methylation and one-carbon metabolism compromise Fyn distribution, neuritogenesis, and APP regulation

The nonreceptor protein tyrosine kinase Fyn and protein Ser/Thr phosphatase 2A (PP2A) are major multifunctional signaling molecules. Deregulation of Fyn and altered PP2A methylation are implicated in cancer and Alzheimer's disease (AD). Here, we tested the hypothesis that the methylation state of PP2A catalytic subunit, which influences PP2A subunit composition and substrate specificity, can affect Fyn regulation and function. Using Neuro-2a (N2a) neuroblastoma cell models, we first show that methylated PP2A holoenzymes containing the Bα subunit coimmunoprecipitate and copurify with Fyn in membrane rafts. PP2A methylation status regulates Fyn distribution and Fyn-dependent neuritogenesis, likely in part by affecting actin dynamics. A methylation-incompetent PP2A mutant fails to interact with Fyn. It perturbs the normal partitioning of Fyn and amyloid precursor protein (APP) in membrane microdomains, which governs Fyn function and APP processing. This correlates with enhanced amyloidogenic cleavage of APP, a hallmark of AD pathogenesis. Conversely, enhanced PP2A methylation promotes the nonamyloidogenic cleavage of APP in a Fyn-dependent manner. Disturbances in one-carbon metabolic pathways that control cellular methylation are associated with AD and cancer. Notably, they induce a parallel loss of membrane-associated methylated PP2A and Fyn enzymes in N2a cells and acute mouse brain slices. One-carbon metabolism also modulates Fyn-dependent process outgrowth in N2a cells. Thus, our findings identify a novel methylation-dependent PP2A/Fyn signaling module. They highlight the underestimated importance of cross talks between essential metabolic pathways and signaling scaffolds that are involved in normal cell homeostasis and currently being targeted for cancer and AD treatment.

disturbances_in_pp2a_methylation_and_one-carbon_metabolism_compromise_fyn_distribution,_neuritogenes

<!>PP2A coimmunoprecipitates with Fyn in a methylation-dependent manner<!><!>PP2A methylation state affects the levels of membrane-associated Fyn in N2a cells<!><!>Changes in PP2A methylation induce profound concomitant changes in the distribution of Fyn and F-actin in N2a cells<!><!>PP2A methylation state affects the partitioning of APP in membrane microdomains and Fyn-dependent APP processing<!><!>Disturbances in one-carbon metabolism that downregulate PP2A methylation alter the distribution of Fyn and Fyn-dependent process outgrowth in N2a cells<!><!>Discussion<!>Materials and reagents<!>Cell culture and transfection<!>Cell treatment and differentiation<!>Confocal microscopy<!>Cell lysis and subcellular fractionation<!>Immunoprecipitation<!>Mouse brain tissue analyses<!>Gel electrophoresis and Western blotting<!>Statistics<!>Data availability<!>Conflicts of interest<!>Supporting information

<p>Edited by Roger Colbran</p><p>Fyn is a member of the Src family kinases (SFKs) of nonreceptor protein tyrosine kinases that modulate a plethora of key cellular functions, including growth, survival, adhesion, migration, and differentiation (1). Expectedly, deregulation of these important signaling enzymes is associated with numerous pathological conditions, including cancer (2). Of particular interest, deregulation of Fyn also participates in Alzheimer's disease (AD) pathogenesis. In AD, the abnormal accumulation of amyloid-β (Aβ) peptides derived from enhanced β- and γ-secretase cleavage of amyloid precursor protein (APP) is believed to be a key event in initiating hallmark neurodegenerative cascades (3). Pathological levels of oligomeric Aβ species are linked to aberrant overstimulation of postsynaptic Fyn-dependent signaling pathways, ultimately leading to impaired synaptic and cognitive functions, and neurotoxicity in several animal models of AD (4). Overactivation of Fyn also induces an abnormal elevation of APP phosphorylated at Tyr682, causing its mistrafficking and missorting in neurons, with important consequences for Aβ production (5). Inhibition of Fyn activity can counteract memory and synaptic deficits in AD mice (6), further demonstrating the essential link between Fyn deregulation and AD pathogenic pathways.</p><p>SFKs transduce signals from a variety of receptors via their ability to form complexes with numerous cytoskeletal and signaling proteins at the plasma membrane (1, 2). The spatial localization and signaling activity of SFKs is tightly controlled by endocytic trafficking (7, 8). In neuronal cells, the myristoylated and palmitoylated Fyn kinase is preferentially enriched and activated in sphingolipid- and cholesterol-enriched plasma membrane microdomains traditionally referred to as lipid or membrane rafts (9, 10, 11). These specialized microdomains serve a key role in cell signaling and function by compartmentalizing and regulating interactions of key membrane proteins (12). For instance, activation of raft-associated Fyn stimulates neurite outgrowth (13, 14, 15) and regulates the targeting of APP to lipid rafts (11). Notably, membrane microdomain switching is a key determinant of APP processing and function (16, 17, 18, 19). Under normal physiological conditions, a majority of APP undergoes proteolytic processing by α-secretase, which precludes Aβ formation and generates neurotrophic-secreted soluble amyloid precursor protein α (sAPPα) fragments. There is strong support that the α-secretase cleavage of APP preferentially occurs in nonraft membrane microdomains, while its amyloidogenic processing primarily takes place in lipid rafts (17, 18, 19).</p><p>Another major signaling molecule deregulated in cancer (20) and AD (21) is protein Ser/Thr phosphatase 2A (PP2A). The "PP2A family" encompasses multimeric enzymes with the typical mammalian holoenzyme being composed of a catalytic "C" subunit (PP2Ac) associated with a scaffolding "A" subunit and a variable regulatory "B" subunit. PP2A biogenesis, stability, and substrate specificity can be modulated by leucine carboxyl methyltransferase 1 (LCMT1)-dependent methylation of PP2Ac on the Leu309 residue; conversely, PP2Ac is demethylated by the methylesterase, PME1 (22). We have previously reported that reduced PP2A methylation is associated with a loss of PP2A/Bα holoenzymes that contain the regulatory Bα (or PPP2R2A) subunit and altered dephosphorylation of PP2A/Bα substrates, including APP phosphorylated at the Thr668 site, in Neuro-2a (N2a) cells and in vivo (23, 24). Phosphorylation of APP at Thr668 and Tyr682 regulates APP interactions (25), subcellular localization, processing, and function, so that abnormally enhanced phosphorylation of APP in AD likely contributes to APP dysfunction (3, 5). PP2A methylation becomes downregulated in AD and after alterations in one-carbon metabolism in cells and in vivo (21). Disturbances in one-carbon metabolism that promote toxic elevation of plasma homocysteine (Hcy) and its oxidized derivatives and inhibition of cellular methylation are strongly associated with AD (26, 27) and cancer (28).</p><p>In this study, using neuroblastoma N2a cell models, we show that intact PP2A methylation is essential for the formation of PP2A/Bα-Fyn protein complexes and their codistribution in membrane rafts. Altered PP2A methylation promotes a redistribution of Fyn and inhibits Fyn-dependent neuritogenesis. It affects the compartmentalization of Fyn and APP in membrane microdomains, which regulates APP processing. Manipulations of one-carbon metabolism that modulates PP2A methylation state also affect Fyn distribution. Our findings identify a novel mechanism of regulation of Fyn at the crossroads of metabolism and signaling.</p><p>Methylated PP2A/Bα enzymes coimmunoprecipitate with Fyn.A, coimmunoprecipitation of endogenous PP2Ac and Fyn from total mouse cortical lysates. B, HA immunoprecipitates and total lysates from N2a cells stably expressing HA-tagged PP2Ac (WT), the methylation-incompetent L309Δ PP2Ac mutant (L309Δ), or empty vector (EV) were immunoblotted for Fyn, HA, or PP2Ac. C, flag immunoprecipitates and total lysates from N2a cells transfected with the indicated plasmids were immunoblotted for Fyn, PP2Ac, and PP2A-Bα. D, the whole NP-40 detergent-insoluble fraction prepared from GFP–Fyn–expressing N2a cells was divided into two equal parts. GFP or PP2Ac immunoprecipitates were prepared from each aliquot and analyzed for the presence of Fyn and PP2Ac (top panel). No Fyn or PP2Ac was found in control GFP immunoprecipitates carried out in EV-transfected, compared with, GFP–Fyn–expressing N2a cells (bottom panel). Representative blots from 3 separate experiments are shown in panels A–D. HA, hemagglutinin; IgG, immunoglobulin; PP2A, PP2A, phosphatase 2A; PP2Ac, catalytic "C" subunit of PP2A; N2a, Neuro-2a.</p><!><p>In agreement with the preferential targeting of Fyn to membrane rafts, Fyn is enriched in detergent-insoluble fractions obtained after lysing tissue or cells with the nonionic detergent, NP-40 (32). Typically, NP-40 detergent-insoluble fractions contain highly insoluble lipid rafts and cytoskeletal components, whereas the detergent-soluble fraction largely consists of cytosolic proteins and extracted proteins "less tightly" attached to membranes. Endogenous PP2Ac coimmunoprecipitated with Fyn in NP-40 detergent-insoluble fractions prepared from GFP–Fyn–transfected but not EV-transfected N2a cells (Fig. 1D), supporting the existence of membrane-associated PP2A–Fyn protein complexes. Together, these findings indicate that methylation-dependent PP2A/Bα holoenzymes are important for the formation of PP2A–Fyn protein complexes.</p><!><p>Changes in PP2A methylation influence the discrete membrane distribution of Fyn in N2a cells.A, representative immunoblots of pY416-SFK (pSFK), Fyn, and actin in total lysates and NP-40 detergent–insoluble fractions prepared from N2a cells stably expressing WT PP2Ac, the L309Δ PP2Ac mutant, LCMT1, PME1, or empty vector (control). Panels in detergent-insoluble fractions originated from the same blot. B, quantification of Fyn levels in NP-40 detergent–insoluble fractions from these cells. Data (mean ± SEM from n = 3–4 independent experiments) were appraised using one-way ANOVA (F (4, 14) = 33.44; p < 0.0001) with Dunnett's post hoc test. ∗∗p < 0.01, ∗∗∗p < 0.001, versus control. C, total lysates and NP-40 detergent–insoluble fractions purified from N2a cells transfected with a validated siRNA targeted to LCMT1 (siLCMT1) or a mismatch siRNA control (siControl) were analyzed by Western blotting for the presence of Fyn and LCMT1. D, Fyn and LCMT1 protein levels were decreased in detergent-insoluble fractions from siLCMT1 relative to siControl-transfected N2a cells. Data (mean ± SEM; n = 3 separate experiments) were analyzed using a student t-test. ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001. E, representative immunoblots of PP2ABα and PP2Ac subunits, Fyn, flotillin-1, and transferrin receptor (TfR) in raft and nonraft membrane fractions purified from N2a cells. Similar results were obtained in three separate purifications. F, representative distribution of PP2Ac and Fyn in aliquots (15 μg) of raft fractions purified from N2a cells that were incubated for 15 min in a serum-deficient medium in the absence (−) or presence (+) of the cholesterol depletion agent MβCD or cholesterol (Chol). Total membrane fractions (input) from these cells were probed with an antibody against the membrane marker, sodium potassium adenosine triphosphatase (Na+/K+ ATPase). G, representative Western blot analysis of total membrane fractions (input) and rafts purified from EV-, WT-, or L309Δ-transfected N2a cells. H, relative levels of raft-associated Fyn were quantified in EV-, WT-, and L309Δ-expressing N2a cells. Data (mean ± SEM from n = 3 separate purifications) were analyzed using a student t-test. ∗p < 0.05; ∗∗p < 0.01, versus control. EV, empty vector; LCMT1, leucine carboxyl methyltransferase 1; N2a, Neuro-2a; MβCD, methyl-β-cyclodextrin; PP2A, PP2A, protein phosphatase 2A; PP2Ac, catalytic "C" subunit of PP2A.</p><!><p>Parallel immunoblotting of total cell lysates confirmed that the differential enrichment of Fyn in NP-40 detergent-insoluble fractions in our cell models was not related to changes in total Fyn protein expression levels (Fig. 2A, Fig. S1A). To assess whether PP2A-mediated alterations in detergent-insoluble Fyn levels are associated with modulation of Fyn activity state, cell fractions were also probed with validated antibodies recognizing SFKs phosphorylated (pSFK) at the conserved regulatory Tyr416 (numbering depending on species), a readout of SFK activity (1). Basal levels of active pSFK were observed in control N2a cells cultured in a "normal" serum-containing medium (Fig. 2A). Relative to controls, there was a marked increase in the pSFK signal in insoluble fractions that closely mirrored the increase in insoluble Fyn protein expression levels in WT- and LCMT1-expressing cells. Conversely, a similar reduction in the levels of pSFK and Fyn was observed in insoluble fractions from L309Δ- and PME1-expressing N2a cells. Yet, after normalizing the pSFK signal (apportioned to Fyn) for Fyn protein expression levels, we found no changes in the net phosphorylation of Fyn in any of the cell lines examined, relative to controls (Fig. S1B). Thus, under our experimental conditions, PP2A methylation influenced steady-state levels of active Fyn in detergent-insoluble fractions via mechanisms that do not implicate overall changes in Fyn activity or protein turnover.</p><p>Because neuronal Fyn is primarily concentrated in lipid rafts that resist extraction by nonionic detergents (9, 10, 11), perturbing PP2A methylation could more specifically alter the targeting of Fyn to these membrane microdomains. In this context, we have previously shown that pools of LCMT1 and methylated PP2Ac and PP2A/Bα holoenzymes are concentrated in lipid rafts, whereas demethylated PP2Ac is preferentially distributed in nonraft membrane microdomains purified from N2a cells (33). These observations suggest that Fyn and methylated PP2A/Bα enzymes are present in the same lipid raft compartment in N2a cells. To confirm this hypothesis, we reanalyzed fully characterized lipid raft and nonraft fractions obtained after membrane fractionation of N2a cells in an earlier study (33). Indeed, Fyn copurified with endogenous PP2Ac and Bα subunits in these flotillin-1–positive membrane rafts isolated from N2a cells (Fig. 2E). As expected, membrane cholesterol depletion by methyl-β-cyclodextrin (MβCD) induced the loss of both PP2A (33) and Fyn from these fractions (Fig. 2F). This effect was reversed by subsequent cholesterol replenishment using a preformed cholesterol–MβCD complex, demonstrating the cholesterol-dependent microdomain association of PP2A and Fyn. We next assessed how disrupting the integrity of PP2A methylation affects the levels of Fyn in N2a cell membrane rafts. We have previously reported that raft-bound PP2Ac levels are increased by ∼30% in WT-expressing N2a cells, compared with controls (33). We observed a similar pattern for Fyn. Expression of the WT enhanced the relative levels of raft-associated Fyn (Fig. 2, G and H), in agreement with the increase in Fyn amounts found in detergent-insoluble cell fractions (Fig. 2, A and B). In contrast, there was a marked reduction in raft-associated Fyn levels after expression of the methylation-incompetent L309Δ mutant. Unlike its WT counterpart, the L309Δ mutant is excluded from rafts, and raft-associated pools of PP2A become downregulated in L309Δ-expressing N2a cells (33). The decrease in raft-associated Fyn was also reminiscent of the loss of Fyn in detergent-insoluble fractions from L309Δ-expressing cells (Fig. 2, A and B). These findings indicate that altering PP2A methylation can negatively influence the targeting of Fyn to membrane rafts in N2a cells. The parallel loss of raft-associated PP2A (33) and Fyn (Fig. 2G) in L309Δ-expressing cells likely point to close spatial and functional interrelationships between the phosphatase and the kinase.</p><!><p>Altering PP2A methylation induces defects in Fyn localization and F-actin organization in N2a cells.A, representative confocal images of the distribution of GFP–Fyn and F-actin in EV (control), WT-, L309Δ-, LCMT1-, or PME1-expressing N2a cells cotransfected with GFP–Fyn. B, Pearson's correlation coefficients (mean ± SD, n = 12 cells/transfection from three separate experiments) showing colocalization of Fyn and F-actin in these cells. Data were analyzed using one-way ANOVA (F (4, 55) = 18.62, p < 0.0001) with post hoc Dunnett's test; ∗∗∗p < 0.001, versus EV. C, cells were also analyzed for the length of actin-positive protrusions. Data (mean ± SD) were appraised using one-way ANOVA (F (4, 928) = 785; p < 0.0001) with post hoc Dunnett's test. ∗∗∗∗p < 0.0001, versus control. D, F-actin distribution in the indicated N2a cell lines in the absence of GFP–Fyn. Images in panels A and D are representative of three separate experiments. Scale bars, 5 μm. EV, empty vector; N2a, Neuro-2a; LCMT1, leucine carboxyl methyltransferase 1; PP2A, protein phosphatase 2A.</p><!><p>Thus, our findings uncover a concurrent reorganization of GFP–Fyn and the F-actin cytoskeleton after alterations in PP2A methylation. Because intact F-actin is critically required for proper peripheral membrane targeting of SFKs (36), we hypothesized that upsetting PP2A methylation homeostasis can influence the subcellular distribution of Fyn, at least in part by inducing a remodeling of the actin cytoskeleton. In support of this hypothesis, alterations in PP2A methylation state in our stable cell models were also associated with a profound rearrangement of the F-actin cytoskeleton and cell shape changes, in the absence of expressed GFP–Fyn (Fig. 3D).</p><!><p>PP2A methylation state affects Fyn-dependent process outgrowth.A, distribution of GFP–Fyn in control or WT-, L309Δ-, LCMT1-, or PME1-expressing N2a cells cotransfected with GFP–Fyn. Cells were incubated for ∼18 h in a low-serum medium to initiate differentiation before fixation. A subset of control cells was incubated in differentiation medium containing 1-μM AMZ-30. B, βIII-tubulin staining was used to assess neurite outgrowth in control or WT- or LCMT1-expressing N2a cells that were transfected with either EV or GFP–Fyn and incubated for ∼18 h in a low-serum medium in the absence or presence of PP2. Subsets of control cells were also incubated in the differentiation medium containing 1-μM AMZ-30. Confocal images shown in panels A–B are representative of three separate experiments. Scale bars, 10 μm. C, cells were comparatively analyzed for neuritic process length. Data are mean ± SD from n = 3 separate experiments and were appraised using two-way ANOVA (effect of PP2A methylation: F = 333.5, p < 0.0001; effect of Fyn–PP2: F = 2506, p < 0.0001; interaction: F = 113.8, p < 0.0001) with Tukey's post hoc multiple comparisons test. ∗∗∗∗p < 0.0001, versus control; ####p < 0.0001. N2a, Neuro-2a; PP2A, protein phosphatase 2A.</p><p>PP2A methylation state affects APP membrane distribution and Fyn-dependent APP processing in N2a cells.A, representative immunoblots of endogenous APP distribution in purified membrane fractions from control or WT-, L309Δ-, LCMT1-, or PME1-expressing N2a cells. Quantitative analyses of the immunoblots from n = 4 separate experiments confirmed that there were no statistically significant changes in APP expression levels (p > 0.05) in total lysates from these cells. B, relative levels of membrane-associated APP in these cells. Data are mean ± SEM from n = 3 separate purifications and were analyzed using one-way ANOVA (F (4, 10) = 99.2, p < 0.0001) with post hoc Dunnett's test. ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001, versus control. C, representative immunoblots of APP distribution in aliquots of the same purified N2a cell raft and nonraft fractions analyzed in Figure 2E. Levels of expressed HA-tagged proteins and APP in corresponding total cell lysates are shown for reference. D, relative levels of raft-associated APP were quantified in EV-, WT-, and L309Δ-expressing N2a cells and are expressed as the percent of total membrane-associated APP. Data (mean ± SEM from n = 3–5 separate purifications) were appraised using one-way ANOVA (F (2, 9) = 382.5, p < 0.0001) with post hoc Dunnett's test. ∗∗p < 0.01, ∗∗∗∗p < 0.0001, versus control. E, levels of secreted sAPPα species and corresponding cellular APP levels were comparatively analyzed by Western blot in WT- and LCMT1-expressing N2a cells, relative to control N2a cells, after incubation for 4 h in a conditioned media in the absence or presence of 5-μM PP2 or PP3. F, the release of sAPPα was quantified in these cells after normalization for total cellular APP levels. Data shown are the mean ± SEM from 3 separate assays and were analyzed using two-way ANOVA (cell line: F = 73.57, p < 0.0001; treatment: F = 112.7, p < 0.0001; interaction: F = 6.46, p = 0.01) using Tukey's post hoc multiple comparisons test. ##p < 0.01, ####p < 0.0001, versus control N2a; ∗∗∗∗p < 0.0001; ns, not significant. APP, amyloid precursor protein; HA, hemagglutinin; LCMT1, leucine carboxyl methyltransferase 1; N2a, Neuro-2a; PP2Ac, catalytic "C" subunit of PP2A.</p><!><p>Changes in the compartmentalization of APP in membrane microdomains critically affect APP processing (16, 17, 18, 19). Indeed, L309Δ-mediated redistribution of APP in rafts (Fig. 5, C and D) correlates with enhanced amyloidogenic processing of APP in L309Δ-expressing N2a cells (23). Conversely, we have previously shown that enhanced PP2A methylation in N2a cells stimulates the nonamyloidogenic processing of APP (23). Because the α-secretase cleavage of APP is also regulated by Fyn (39), we further examined whether Fyn inhibition can affect PP2A-mediated sAPPα secretion. Expression of either WT or LCMT1 in N2a cells boosted the release of sAPPα species (Fig. 5, E and F), in agreement with our earlier studies (23). Notably, these stimulatory effects were inhibited when cells were incubated in the presence of the PP2 inhibitor, but not PP3 (drug control). These findings suggest that PP2A methylation–dependent APP cleavage is dependent on Fyn.</p><!><p>Fyn distribution and Fyn-dependent process outgrowth are dependent on one-carbon metabolism in N2a cells.A, immunoblot analysis of Fyn and pY416-SFK (pSFK) in total lysates (total) and detergent-insoluble (insoluble) fractions from N2a cells that were incubated for ∼16 h with 100-μM SAM, 100-μM Hcy or vehicle (control). B, representative immunoblots of Fyn and pY416–SFK (pSFK) in total lysates and detergent-insoluble fractions from N2a cells that were incubated for ∼16 h with 50-μM 3-deazaadenosine (3-DZA), 100-μM SAH, or vehicle (control). A subset of cells was incubated for 4 h in a folate-deficient (FD) medium. Quantitative analyses of the immunoblots from n = 3 to 4 separate experiments revealed that incubation with either SAM, Hcy, SAH, 3-DZA, or FD did not induce any statistically significant changes (p > 0.05) in total Fyn protein expression levels, or total or detergent-insoluble Fyn phosphorylation levels, relative to vehicle-treated N2a cells. C, detergent-insoluble Fyn levels were quantified in these cells. Data shown are mean ± SEM from n = 3 to 4 independent experiments and were analyzed using one-way ANOVA (F (5, 17) = 69.37, p < 0.0001) with post hoc Dunnett's test. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗∗p < 0.0001, versus control. D, time-dependent changes in detergent-insoluble levels of Fyn, methylated PP2Ac, and LCMT1 enzymes in N2a cells switched for the indicated time from normal folate-containing medium to FD medium. E, immunoblot analysis of detergent-insoluble Fyn levels in control or WT-expressing N2a cells incubated for 16 h with 100-μM SAM, 100-μM SAH, or a combination of 100-μM SAM and 5-nM okadaic acid (OA). F, immunoblot analysis of detergent-insoluble Fyn and actin levels in L309Δ- and PME1-expressing N2a cells incubated for ∼16 h with 100-μM SAM or vehicle. For panels D–F, similar results were observed in three separate experiments. G, representative confocal images of GFP–Fyn in transfected N2a cells that were incubated for ∼18 h in a low-serum medium in the presence of 100-μM SAM, 50-μM 3-DZA, or vehicle (control) before fixation. Scale bars, 10 μm. H, N2a cells transfected with either EV or GFP–Fyn were incubated for ∼18 h in the differentiation medium in the absence or presence of 5-μM PP2 and labeled with anti-βIII-tubulin antibodies. Scale bars, 10 μm. I, quantification of the neurite length in these cells. Data shown are mean ± SD from cells from 3 separate experiments and were analyzed with two-way ANOVA (effect of SAM: F = 1152, p < 0.0001; effect of Fyn–PP2: F =1600, p < 0.0001; interaction: F = 342.1, p < 0.0001) with post hoc Tukey's test. ∗∗∗∗p < 0.0001, versus control + EV or control + Fyn; ####p < 0.0001. EV, empty vector; FD, folate-deficient; Hcy, homocysteine; LCMT1, leucine carboxyl methyltransferase 1; N2a, Neuro-2a; PP2Ac, catalytic "C" subunit of PP2A.</p><!><p>Notably, incubation of N2a cells in the FD medium caused a time-dependent parallel loss of LCMT1, methylated PP2A, and Fyn enzymes from detergent-insoluble fractions (Fig. 6D). Moreover, treatment of N2a cells with SAM further synergized with WT expression to enhance insoluble Fyn levels, whereas SAH blocked the ability of WT to increase insoluble Fyn levels (Fig. 6E). SAM-mediated increase in insoluble Fyn levels was also abolished when cells were coincubated with the PP2A inhibitor, okadaic acid (OA). However, incubation with SAM was unable to rescue the loss of insoluble Fyn levels induced by L309Δ or PME1 expression in N2a cells (Fig. 6F). Together, these findings indicate that the status of one-carbon metabolism can concurrently influence the membrane distribution of PP2A and Fyn. They support the hypothesis that metabolic-induced deregulation of PP2A methylation is a major contributor to Fyn deregulation. In this context, it has been reported in non-neuronal COS-1 cells that Fyn is trimethylated at Lys7/9 (42). This observation raised the possibility that altering one-carbon metabolism could also impact Fyn regulation in a PP2A-independent manner, by directly affecting its methylation state. To address this hypothesis, GFP–Fyn immunoprecipitates were prepared from transfected N2a cells treated with SAM (to enhance methylation), 3-DZA (to inhibit methylation), or vehicle and then analyzed by Western blot with validated anti-methyl-lysine antibodies (Fig. S2). Although methyl-sensitive proteins were clearly present in the GFP–Fyn immunoprecipitates, we were unable to detect any immunoreactivity of GFP–Fyn with these antibodies under our experimental conditions.</p><p>We next investigated whether Fyn-dependent process outgrowth was susceptible to changes in the SAM/SAH ratio. Relative to vehicle-treated controls, incubation with SAM enhanced GFP–Fyn–mediated differentiation of N2a cells (Fig. 6, G–I). This stimulating effect and the pattern of GFP–Fyn distribution in SAM-treated cells were highly reminiscent of those induced by AMZ-30, WT, and LCMT1 (Figs. 3 and 4). Notably, SAM-mediated N2a cell differentiation was abolished by PP2. Fyn-dependent process outgrowth was also inhibited by 3-DZA. The distribution of GFP–Fyn in 3-DZA–treated cells was strikingly similar to that observed in L309Δ-expressing cells (Figs. 3A and 4A). Thus, the integrity of one-carbon metabolism is essential for Fyn-dependent process formation in N2a cells.</p><!><p>Elevated levels of Hcy or its thiolactone derivative induce concomitant PP2A demethylation and alterations in Fyn distribution in acute mouse brain slices.A, representative immunoblots of Fyn and pSFK in total extracts (total) and detergent-insoluble (insoluble) fractions prepared from acute mouse cortical slices incubated for 30 or 60 min with 200-μM Hcy-thiolactone (HTL). B, representative immunoblots of pSFK and Fyn in acute brain slices incubated for 2 h with either 200-μM Hcy or HTL. Two separate slices treated with Hcy are shown. Quantitative analyses of the immunoblots from n = 4 separate experiments revealed that total Fyn protein expression levels, or total or detergent-insoluble Fyn phosphorylation levels were not statistically significant different (p > 0.05) in Hcy or HTL-treated, relative to, vehicle-treated slices. C, quantification of detergent-insoluble Fyn levels in Hcy- or HTL-treated slices, expressed as percent of vehicle-treated, control slices. Data shown are mean ± SEM from 4 separate mouse brain tissue and were analyzed using one-way ANOVA (F (2, 9) = 17.88, p = 0.0007) with post hoc Dunnett's test; ∗∗p <0.01, ∗∗∗p <0.001, relative to vehicle-treated slices. D, comparative immunoblot analysis of demethylated PP2Ac, Fyn, and actin in total extracts and detergent-insoluble fractions prepared from acute mouse cortical slices incubated for 2 h with 200-μM Hcy. Duplicate blots were probed for total PP2Ac levels and actin. Quantitative analyses of the immunoblots showed a ∼58 ± 9% increase (n = 3 separate experiments; p < 0.0001; student t-test) in demethylated PP2A levels in detergent-insoluble fractions from Hcy-, relative to, vehicle-treated slices. Hcy, homocysteine; HTL, Hcy thiolactone; PP2Ac, catalytic "C" subunit of PP2A; pSFK, phosphorylated SFK.</p><!><p>Methylation is a key regulatory post-translational mechanism that controls biogenesis of PP2A/Bα holoenzymes and PP2A subunit composition, thereby influencing PP2A targeting, substrate specificity, and interactions with numerous proteins and regulators (22). Here, using N2a cells, we first show that manipulating PP2A methylation affects the distribution and function of Fyn, a major signaling enzyme deregulated in AD and cancer. The subcellular distribution of Fyn is regulated by trafficking to and internalization from the plasma membrane; owing to N-terminal lipid modifications, Fyn is preferentially targeted to membrane rafts, regardless of its activity (7, 8, 43, 44, 45). We found that enhanced cellular PP2A methylation was associated with increased levels of Fyn in detergent-insoluble N2a cell fractions and membrane rafts and enhanced clustering of Fyn along peripheral actin-rich filopodia. Conversely, the accumulation of demethylated PP2A promoted a steady-state loss of membrane- and raft-associated Fyn and its mislocalization. These effects were not associated with changes in net Fyn activity or expression levels; rather, they closely correlated with a reorganization of the F-actin cytoskeleton. Although general PP2A activity has been implicated in the complex regulation of actin dynamics, underlying mechanisms and contribution of specific PP2A isoforms remain poorly characterized (46). Nevertheless, an intact actin cytoskeleton is critically required for proper peripheral membrane targeting of Fyn (36). Our findings suggest that altering PP2A methylation in N2a cells interferes with the normal distribution of Fyn, likely in part by deregulating F-actin dynamics. This hypothesis is further supported by the strong functional link between cytoskeletal reorganization and clustering of proteins in membrane rafts, which serve as signaling platforms regulating adhesion, differentiation, and polarity (47). These membrane microdomains are also involved in protein sorting, endocytosis, and recycling from/to the plasma membrane (48). Moreover, deregulation of PP2A promotes the endocytosis of E-cadherin by inducing F-actin disassembly (49). The pattern of Fyn distribution in L309Δ- and PME1-expressing N2a cells suggests that enhanced PP2A demethylation could similarly promote Fyn internalization via actin-dependent mechanisms. Yet, defects in Fyn trafficking could also occur by other mechanisms, based on the role of PP2A in dephosphorylating adaptor proteins regulating clathrin-mediated endocytosis (50).</p><p>We also observed that increasing PP2A methylation stimulated Fyn-dependent process outgrowth, whereas altering PP2A methylation abolished it. Because activation of raft-associated Fyn (13, 14, 15) and actin remodeling (35) are intimately linked with neuritogenesis, it is likely that PP2A influences Fyn-dependent N2a cell differentiation by affecting F-actin dynamics and Fyn targeting to rafts. For instance, the reorganization of cortical actin into aggregates and filopodia, which is required for neurite initiation (35), was prevalent in WT- and LCMT1- but absent in PME1-expressing N2a cells. However, additional direct and indirect mechanisms, such as PP2A-induced changes in microtubule stability (46) and Fyn-mediated F-actin dynamics (1), could also be involved.</p><p>Our data also indicate that deregulation of PP2A in N2a cells influences the clustering of APP in membrane microdomains, which governs APP processing (17, 18, 19). Accordingly, we found that enhanced association of APP with rafts (Fig. 5D) coincided with enhanced β-secretase cleavage of APP (23) in L309Δ-expressing N2a cells. Conversely, expression of the WT in N2a cells increased the relative levels of endogenous APP in nonraft plasma membrane microdomains and sAPPα secretion (Fig. 5). In agreement with decreased sAPPα release in Fyn KO mice (39), inhibiting Fyn abolished both WT- and LCMT1-mediated sAPPα secretion in N2a cells. These data further cement the existence of a close methylated PP2A/Fyn functional inter-relationship in regulating APP.</p><p>Phosphorylation-dependent protein–protein interactions also shape APP localization and processing (3, 25). Fyn-mediated Tyr phosphorylation regulates the association of APP with adaptor proteins and promotes the sorting of APP to lipid rafts (11). Aberrant Fyn activation in AD has been linked to enhanced Fyn–APP interactions, deficits in APP sorting and trafficking, and amyloidogenesis (5, 51). We found that PP2A/Bα holoenzymes copurified in membrane rafts and coimmunoprecipitated with detergent-insoluble Fyn. This indicates the existence of membrane-bound Fyn–PP2A complexes; however, these protein–protein interactions may be indirect because purified Fyn and PP2A/Bα holoenzymes do not associate in vitro (52). Relative to control N2a cells, expression of the WT enhanced Fyn levels in rafts while increasing the ratio of APP partitioning in nonraft membrane microdomains, suggesting that WT promotes the segregation of Fyn from its substrate, APP. In contrast to WT, the methylation-incompetent and B binding–incompetent L309Δ mutant failed to associate with Fyn and promoted the sorting of APP into rafts. Whereas the expression of L309Δ decreased total Fyn levels in rafts, enhanced targeting of APP to rafts may increase the potential for functional interactions of APP with the active kinase still present in these microdomains. Enhanced PP2A demethylation in N2a cells (23) also promotes phosphorylation of APP at Thr668, which impacts the APP interactome (25) and APP distribution (3). In neurons, pThr668–APP species are concentrated in endosomes, favoring APP β-secretase cleavage (3). Based on these findings, it is tempting to speculate that PP2A methylation regulates the formation of localized protein scaffolds that play a crucial role in directing the trafficking and processing of APP toward the competing amyloidogenic or nonamyloidogenic pathways. In this context, it is noteworthy that the L309Δ mutant fails to associate not only with Fyn (Fig. 1B) but also with tau proteins (23). Enhanced formation of Fyn–tau scaffolds plays a key role in mediating Aβ-induced synaptic dysfunction and excitotoxicity in AD (4). Because PP2A/Bα and Fyn compete for tau binding, disruption of normal PP2A–Fyn and PP2A–tau protein–protein interactions as a result of PP2A demethylation would enhance the potential for neurotoxic Fyn–tau interactions (53). Thus, interfering with homeostatic PP2A methylation has the potential, via several intricate mechanisms, to deregulate the function of key players in AD pathogenesis.</p><p>Using N2a cells and mouse brain slices, we also established a link between one-carbon metabolism and the regulation of Fyn distribution. Metabolic disturbances that lead to elevated Hcy levels and altered cellular methylation potential induced a concomitant loss of methylated PP2A and Fyn enzymes from detergent-insoluble fractions; conversely, boosting cellular methylation led to their coenrichment. In an earlier mass spectrometry study, Fyn was reported to be trimethylated on Lys residues within the SH4 domain; a regulatory link between methylation and Fyn targeting and function was further proposed based on the use of Lys mutants in COS-1 cells (42). These observations could provide a plausible mechanism by which altering one-carbon metabolism can directly affect Fyn methylation state and thereby affect its distribution. However, to the best of our knowledge, Fyn methylation has never been confirmed in any follow-up studies, and the identity of the Fyn methyltransferase remains unknown to date. Proper validation of protein Lys methylation is challenging, requiring several approaches to avoid pitfalls (54). Assigning effects of Lys mutants to changes in Fyn methylation (53) may be confounded by the fact that Fyn also undergoes acylation in the same N-terminal domain, which controls Fyn association with membrane rafts (44). Under our experimental conditions, we were unable to detect Fyn methylated on Lys residues, arguing against a prevalent direct role of Lys methylation in regulating Fyn distribution. Yet, we do not exclude the possibility that Fyn undergoes methylation on other yet unidentified amino acids, which could render Fyn directly susceptible to alterations in one-carbon metabolism. Nevertheless, our findings in N2a cells (Fig. 6, E and F) and brain slices (Fig. 7) strongly support the hypothesis that altered PP2A methylation actively contributes to deregulation of Fyn in response to disturbances in the methylation cycle. Yet, PP2A- and methylation-independent mechanisms, such as oxidative stress (26), could also participate in Fyn dysregulation in response to altered folate and Hcy metabolism.</p><p>The crosstalk between Hcy metabolism and the major signaling molecules, PP2A and Fyn, is of particular importance for the AD and cancer fields. Hyperhomocysteinemia is an established risk factor for AD (27) and is experimentally associated with the development of hallmark pathological features of AD, including tau and APP phosphorylation, and amyloidogenesis (26). Disturbed Hcy metabolism is strongly associated with cancer (28). Hyperhomocysteinemia promotes PP2A demethylation in vivo (23, 55). Alterations in PP2A methylation are found in patients with AD (21) and cancer (56). In our neuroblastoma model, they promote a loss of Fyn from membrane rafts. Because the confinement of Fyn in membrane rafts limits its ability to promote cell transformation (44, 57), it is possible that deregulation of Fyn contributes to the role of PP2A in cancer.</p><p>Collectively, our findings demonstrate that the integrity of one-carbon metabolism and PP2A methylation are essential for proper regulation of Fyn and APP. Our study identifies an important link between metabolic pathways and multifunctional signaling molecules currently being targeted for AD and cancer therapies.</p><!><p>Unless indicated, all chemicals and drugs were from Sigma-Aldrich/Merck Millipore. Primary antibodies used in this study included the following: Rabbit anti-HA clone C29F4, anti-pSrc Tyr416 clones 100F9 and D49G4, anti-Na+/K+ ATPase #3010, and anti-GFP clone D5.1 (Cell Signaling Technology); rabbit Anti-APP clone Y188 (Abcam); mouse anti-HA clone 16B12 (Covance); rabbit anti-actin #AAN01 (Cytoskeleton Inc); anti-transferrin receptor clone H68.4 (Thermo Fisher Scientific); rabbit anti-methylated Lysine (Enzo Life Sciences); mouse anti-Fyn clone 25 #610163, anti-flotillin-1 clone 18, and anti-PP2Acα clone 46 (BD Transduction Laboratories); rabbit anti-Fyn clone EPR5500, mouse anti-Bα clone 2G9, anti-actin clone C4, anti-LCMT1 clone 4A4, anti-APP clone 22C11 (MAB348), and anti-demethyl PP2Ac clone 1D6 (Merck Millipore).</p><!><p>Mouse N2a neuroblastoma cells were obtained from the American Type Culture Collection. N2a cells stably expressing myc-tagged PME-1, HA-tagged LCMT1, HA-tagged WT PP2Ac, or the HA-tagged methylation site L309Δ C subunit mutant have been extensively characterized in previous studies (23, 31, 33, 34). Control and stable cell lines were maintained in Dulbecco's modified Eagle's medium (DMEM, Thermo Fisher Scientific) containing 2.5-mM Hepes, pH 7.4, 10% fetal bovine serum (FBS, Bovogen Biologicals), and 10-μg/ml gentamycin (Thermo Fisher Scientific). In some experiments, control and stable cells lines were transiently transfected with the indicated plasmids using METAFECTENE PRO reagent, following the manufacturer's instructions (Biontex laboratories, Germany). Plasmids used in this study included the following: pAcCMV Fyn–GFP plasmid encoding GFP-tagged human Fyn (OriGene); peCFP-APP plasmid encoding human WT APP695 (gift from Dr Ottavio Arancio, Columbia University, New York, NY); and Bα/pcDNA5/TO plasmid encoding Flag-tagged PP2A Bα (PPP2R2A) subunit (58) (gift from Dr Brian Wadzinski, Vanderbilt University, Nashville, TN). All plasmids were verified by sequencing. Cells mock-transfected with EVs were used as "controls." Partial knockdown of endogenous LCMT1 in N2a cells was performed using transient transfection with small interfering RNA (siLCMT1) shown to specifically target mouse LCMT1; cells transfected with mismatch siRNA (siRNA control) were used as controls. Experimental conditions were optimized to prevent cell death ultimately caused by the complete or prolonged loss of LCMT1 (33, 34).</p><!><p>Unless otherwise indicated, all experiments and incubation with compounds were performed in ∼80% confluent cells cultured in a regular cell culture medium. To assess the role of one-carbon metabolism, cells were incubated for ∼16 h with 100-μM SAM, 100-μM SAM + 5-nM 100 OA, 100-μM SAH, 50-μM 3-DZA, 1 μM AMZ-30, or vehicle (33). Folate deficiency was induced by switching N2a cells cultured in normal folate-containing medium to folate-free RPMI-1640 medium supplemented with 2% dialyzed FBS (Thermo Fisher Scientific) (34). To assess the role of cholesterol, subsets of cells were incubated at 37 °C for 15 min in a serum-free medium with 1% MβCD premixed or not with 100 μg/ml cholesterol, before harvesting for membrane raft purification (33). To assess the role of Fyn, cells were incubated for the indicated time in the presence of 5-μM PP2 or PP3 (control drug for the SFK inhibitor, PP2). sAPPα secretion was analyzed 4 h after incubation of N2a cells in the conditioned media, exactly as described previously (23). To study neurite outgrowth, cells were plated in a regular medium 24 h after transfection onto poly-L-lysine–coated glass coverslips. Five hours after plating, cells were switched to DMEM containing 0.5% FBS in the absence or presence of the indicated drugs and incubated for ∼18 h before fixation.</p><!><p>To visualize F-actin, cells were fixed for 20 min with 4% paraformaldehyde, permeabilized for 5 min with PBS containing 0.1% Triton X-100, washed, and incubated for 1 h in PBS containing 3% bovine serum albumin. Cells were labeled with Alexa Fluor594 conjugated phalloidin to detect F-actin (Thermo Fisher Scientific). GFP–Fyn was directly visualized in fixed cells. To label neurites, N2a cell lines were fixed for 5 min at −20 °C with absolute methanol before staining with rabbit anti-βIII-tubulin antibodies (Abcam #18207) followed by Alexa Fluor594–conjugated secondary antibody (Thermo Fisher Scientific #A27034) (37). After washing in PBS, all samples were mounted with Fluoromount (ProSciTech) and examined on a Nikon Eclipse 80i confocal microscope using a 60x objective. Captured images (z-stacks) were exported to NIH ImageJ/Fiji for analyses of either the protrusion length or protein colocalization. Images were transferred to Adobe Photoshop/Illustrator 2020 (Adobe Systems Incorporated) for figure preparation.</p><!><p>After washing with PBS, total N2a cell homogenates (100-mm dishes) were prepared in 400-μl buffer 1 [10-mM Tris, pH 7.4, 150-mM NaCl, 1-mM dithiothreitol, 0.5-μM OA, 5-mM PMSF, 1% NP-40, Sigma Protease Inhibitor Cocktail, and Sigma Phosphatase Inhibitor Cocktail] using a mortar and pestle. In some experiments, total cell lysates were further centrifuged for 90 min at 20,000g to generate NP-40 detergent–soluble (supernatant) and NP-40 detergent–insoluble (pellet) fractions. The detergent-insoluble cell pellet was resuspended in 200 μl of buffer 2 (buffer 1 + 0.5% sodium deoxycholate) and carefully homogenized for 80 s using a mortar and pestle. For immunoprecipitation assays, homogenized total and insoluble fractions were cleared by centrifugation at 13,000g for 3 min at 4 °C. For Western blot analyses, total homogenates and detergent-insoluble fractions were further sonicated before clarification. The protein concentration was determined in diluted aliquots of homogenates using the Bradford protein assay kit (Bio-Rad). Previously purified NP-40–insoluble fractions from N2a cells transiently transfected with validated siLCMT1 or siRNA control (33) were also reanalyzed here by immunoblotting. Purification of the plasma membrane from N2a cells was carried out by ultracentrifugation (33). Validated detergent-free procedures based on fractionation of the plasma membrane by centrifugation on an OptiPrep gradient were used to purify raft and nonraft membrane microdomains from N2a cells (33). Aliquots of the same N2a cell membrane microdomain preparations characterized in a previous study (33) were reanalyzed here by Western blot for the presence of Fyn and APP.</p><!><p>Immunoprecipitates were prepared from total homogenates or detergent-insoluble fractions from N2a cells or mouse cortical tissue (∼500 μg proteins/assay). Immunoprecipitation of transfected proteins was performed by incubating samples overnight at 4 °C with either anti-Flag–coupled (clone M2, Sigma #M8823), anti-GFP–coupled (clone RQ2, MBL International #D153–9), or anti-HA–coupled (clone C29F4; Cell Signaling Technology #11846) magnetic beads. When immunoprecipitating endogenous proteins, homogenates were precleared for 1 h at 4 °C before overnight incubation with the indicated antibodies. Samples were then incubated for 1 h at 4 °C with PureProteome Protein A/G mix magnetic beads (Merck Millipore). Magnetic beads were washed 5 times in buffer 2 before being resuspended in a gel loading buffer. Input fractions (∼50 μg proteins) and corresponding immunoprecipitates were analyzed by Western blotting.</p><!><p>Brains were rapidly removed from 8- to 11-month-old female C57/BL6 mice that were sacrificed for another project approved by the Animal Care and Ethics Committee of the University of Newcastle. For the preparation of acute slices, brains were immediately immersed in ice-cold, oxygenated, sucrose-substituted artificial cerebrospinal fluid (250-mM sucrose, 25-mM NaHCO3, 10-mM glucose, 2.5-mM KCl, 1-mM NaH2PO4, 1-mM MgCl2, and 2.5-mM CaCl2) (59). Cortical coronal slices (∼400 μm thick) were obtained using a vibratome (Leica VT-1200S, Heidelberg, Germany) and transferred to an interface storage chamber containing oxygenated artificial cerebrospinal fluid (118-mM NaCl substituted for sucrose in sucrose-substituted artificial cerebrospinal fluid). Slices were allowed to recover for 1 h at 22 to 24 °C before incubation for the indicated time into RT oxygenated artificial cerebrospinal fluid containing the indicated compounds or vehicle. Slices were then harvested for further Western blot analyses. Total homogenates and NP-40 detergent-insoluble fractions were prepared from either acute slices or fresh mouse cortical tissue as described above for N2a cells.</p><!><p>Protein samples (∼50-μg proteins/lane) were resolved on NuPAGE 4 to 12% Bis-Tris gels (Thermo Fisher Scientific). Prestained Protein Standards (Bio-Rad) were used as molecular weight markers. Membrane raft fractions were analyzed by immunoblotting using chemiluminescence as described previously (33). Other Western blot analyses were performed using the indicated primary antibodies, followed by Infrared IRDye-labeled secondary antibodies, and visualized using the Odyssey Infrared imaging system (LI-COR Biosciences). In most cases, blots were cut between molecular weight markers to allow simultaneous immunostaining and reprobing of the top and bottom parts with distinct antibody species. Band intensity was determined using the associated Image Studio Lite, version 5.0, Software (LI-COR Biosciences) to accurately quantify protein expression levels. Anti-actin antibodies were used to normalize for protein loading. Anti-pTyr416 SFK antibodies were used to assess the phosphorylation state of Fyn, which was determined after normalizing the pSFK signal corresponding to Fyn (determined as being the fastest migrating band) for total Fyn expression levels and protein loading. PP2A demethylation was assessed as described previously (33).</p><!><p>Data were analyzed for normal distribution and statistical significance using GraphPad Prism 9.</p><!><p>All the data supporting our conclusions are presented in this article. All materials are available upon request.</p><!><p>The authors declare that they have no conflicts of interest with the contents of this article.</p><!><p>Supplementary Figures</p>

PubMed Open Access

Phosphorylation Dynamics Dominate the Regulated Proteome during Early Xenopus Development

The earliest stages of animal development are largely controlled by changes in protein phosphorylation mediated by signaling pathways and cyclin-dependent kinases. In order to decipher these complex networks and to discover new aspects of regulation by this post-translational modification, we undertook an analysis of the X. laevis phosphoproteome at seven developmental stages beginning with stage VI oocytes and ending with two-cell embryos. Concurrent measurement of the proteome and phosphoproteome enabled measurement of phosphosite occupancy as a function of developmental stage. We observed little change in protein expression levels during this period. We detected the expected phosphorylation of MAP kinases, translational regulatory proteins, and subunits of APC/C that validate the accuracy of our measurements. We find that more than half the identified proteins possess multiple sites of phosphorylation that are often clustered, where kinases work together in a hierarchical manner to create stretches of phosphorylated residues, which may be a means to amplify signals or stabilize a particular protein conformation. Conversely, other proteins have opposing sites of phosphorylation that seemingly reflect distinct changes in activity during this developmental timeline.

phosphorylation_dynamics_dominate_the_regulated_proteome_during_early_xenopus_development

<!>Results and Discussion<!>Quantitative changes in protein expression.<!>Dynamics of the phosphoproteome during early development.<!>Proline-directed kinases account for the majority of changes in phosphorylation.<!>Conclusions<!>Methods<!>Protein preparation.<!>ITRAQ labeling.<!>Proteomic UPLC-ESI-MS/MS analysis.<!>Phosphoproteomics UPLC-ESI-MS/MS analysis.

<p>The path from a fully-grown oocyte arrested in meiosis I, to a fertilizable egg arrested in meiosis II, and onto a diploid zygote, occurs in distinct steps initiated by a steroid hormone in the first instance and sperm entry in the latter. Amphibians have been especially valuable for the identification and characterization of the signaling pathways that mediate these transitions, because the individual steps are easily manipulated. Xenopus oocyte maturation and egg fertilization can be performed in vitro, generating a large population of synchronized cells amenable to temporal studies. For these reasons, the activities that control these processes such as maturation promoting factor (MPF) and cytostatic factor (CSF) were first identified in frog 1 and much of our understanding of early animal development comes from work using Xenopus as a model organism [2][3][4][5] .</p><p>Stable isotope labeling has greatly advanced large-scale quantitative proteomic analysis by mass spectrometry (MS) 6,7 . As a result, deep proteome analyses of a number of model organisms of animal development have been reported [8][9][10] . Several proteomic studies have focused on Xenopus eggs, embryos, and blastomeres and have provided unprecedented information regarding the means by which protein expression is regulated during development, and how this regulation impacts processes such as the length of the cell cycle and partitioning of proteins between the nucleus and cytoplasm [11][12][13][14][15][16][17] .</p><p>While cellular regulation can be achieved through simple control of protein expression, it is clear that post-translational modifications are widely used to regulate protein activity and stability. In particular, signaling pathways involving a panoply of kinases orchestrate the maturation of an animal oocyte to an egg and control the cell cycle after fertilization. For example, phosphorylation coordinates the unmasking of maternal mRNAs, the activation of the anaphase promoting complex (APC/C), and the assembly/disassembly of the nuclear membrane and mitotic spindle. Because a considerable amount of information has been derived from studies using Xenopus oocytes and eggs, they provide a logical target for a deeper examination of the vertebrate phosphoproteome during early development. The first examination of the Xenopus phosphoproteome identified 1441 phosphorylation sites on 654 proteins 18 and a subsequent study identified 1738 sites 19 . Together, these two investigations tallied 2636 unique sites.</p><p>We have analyzed the Xenopus phosphoproteome at seven time points beginning with fully grown stage VI oocytes, through oocyte maturation, and intervals following fertilization. iTRAQ labeling combined with Ti-IMAC enrichment for phosphopeptides enabled us to identify 8974 phosphorylation sites on 5169 different proteins. We combined measurements of the proteome and the phosphoproteome to determine the occupancy change at any individual amino acid position. The concurrent measurement of protein expression and phosphopeptide sites revealed the kinetics of occupancy of 4679 phosphorylation sites across the seven developmental time points. We have captured many of the well-documented phosphorylation events that occur during these developmental stages, which lends confidence in the reliability of this data set. In addition, the data also provide new insights into phosphoproteomics in general.</p><!><p>Deep proteomic and phosphoproteomic analysis of X. laevis oocyte maturation and fertilization.</p><p>Four independent 8-plex iTRAQ experiments were performed to determine quantitative changes in the proteome and phosphoproteome of Xenopus laevis at seven developmental stages (Fig. 1A). The sequence begins with fully grown, stage VI oocytes arrested in prophase of meiosis I (PI, experimental time point 1), followed by Heatmap showing the change in protein expression across the developmental time points. The heatmap was generated from the log 2 normalized data using the default parameters in Matlab. Six groups representing significant trends of protein expression were manually selected for analysis. Proteins with the 15% greatest quantitative change are presented and are considered significant at a FDR of 0.20 using a Benjamini-Hochberg test. Neither clustering nor dendrogram generation was performed along the development stage axis of the heatmaps. These points are defined by the biology of development, and their differences are fixed by that biology.</p><p>two time points after progesterone-induced maturation to a fertilizable egg. One taken 45 minutes after exposure to hormone (TP2) and the next taken approximately six hours later (TP3) following germinal vesicle breakdown (GVBD). The four remaining time points are subsequent to fertilization: the first (TP4) at cortical rotation (to insure successful fertilization) and then at 30 minutes (TP5). The final two samples were taken just as the cleavage furrow became visible (TP6) and, finally, the fully formed two-cell embryo (TP7). Biological duplicates were analyzed for all time points in technical duplicate. Protein identification and quantitation followed procedures described earlier 17,20 .</p><p>Protein samples were digested with trypsin, iTRAQ labeled, and collectively pooled; phosphoproteomic digests were further enriched with Ti-IMAC. Both the proteomic and enriched phosphoproteomic digests were fractionated using high pH reverse phase chromatography and analyzed using nano-UPLC-mass spectrometry. Roughly 2.65 million MS/MS spectra from 30 fractions were acquired for the proteome and 1.21 million MS/ MS spectra from eight fractions for the phosphoproteome, totaling over 200 instrument hours between the two experimental conditions. The .raw files were analyze using MaxQuant and the Genome 9.1 database available from Xenbase. Data were filtered with a peptide identification rate of ~99% (peptide-level FDR set at 0.01).</p><p>Biological duplicates of the proteome exhibit strong intensity correlation (>0.95 R 2 value) at equivalent time points. Of the 6428 identified proteins, 4938 were found in both biological samples. A total of 61,041 peptides were identified (Supplementary Table S1). When compared to published experimental data for the stage 1 embryo (TP5 of this study), the Pearson coefficient was >0.79 for five separate experiments 16,17 . Protein copy numbers and protein amounts are highly correlated (Pearson coefficient > 0.95) with the data of Smits et al. that used dimethyl labeling 16 .</p><p>For the phosphoproteome experiments, biological duplicates and technical duplicates provided reasonable correlation (Pearson coefficient > 0.7) for all time points and when comparing the corrected intensities for both of the biological replicates at identical time points. Of the 5169 identified proteins, 3583 were found in both biological samples (Supplementary Table S2). A total of 15,597 peptides were identified. The phosphoamino acid distribution was 76% serine, 21% threonine, and 3% tyrosine.</p><p>Changes in the abundance of a phosphopeptide can have two causes. Protein expression is constant, but the degree of phosphorylation changes. Alternatively, protein expression changes at a constant level of phosphorylation. Of course, both protein abundance and the extent of phosphorylation can simultaneously change. To separate these phenomena, we quantified changes in both the proteome and phosphoproteome 21 . The concurrent measurement of protein abundance and phosphopeptide levels enabled us to determine the absolute occupancy of 4679 phosphorylation sites across the seven developmental time points (Supplementary Table S3).</p><!><p>Over the developmental period from oocyte maturation through the first zygotic cleavage, we find that there is not a substantial change in the Xenopus proteome (Supplementary Fig. S1). Figure 1B presents a spaghetti plot of the normalized log 2 intensities for proteins that were observed at all time points; very few proteins showed large deviation from the mean. The relative standard deviation (RSD) in normalized (log 2 ) intensity for the population of proteins at each time point is presented in Fig. 1C. Only 486 of the 6428 proteins exhibited a protein expression change greater than 5%. While many of these proteins are involved in general metabolic processes, there are some noteworthy patterns of protein expression that correlate with the biological events that occur during these stages of early development (Fig. 1D, Supplementary Table S4).</p><p>Pioneering studies by Smith and coworkers determined that there is an approximate two-fold increase in the rate of protein synthesis following progesterone-induced oocyte maturation that, nonetheless, did not detectably change the overall pattern of expression 22 . The expression levels measured here by quantitative mass spectrometry are in agreement with this earlier work. Group A (Fig. 1D) contains proteins that increase during oocyte maturation, but then decline after fertilization. An important member of this group is Mos, the kinase that initiates the MAP kinase pathway that leads to activated M-phase promoting factor (MPF). Mos, which is encoded by masked maternal mRNA, is not detected in stage VI oocytes, but appears 45 min after progesterone treatment, and remains reasonably constant until fertilization when it declines to an undetectable level (Fig. 2). Interestingly, Cdk1 also exhibits an immediate increase of approximately 60% that would enable the formation of Cdk1/ Ringo, which is responsible for the initial progesterone-dependent inactivation of Myt1 kinase 23 . In addition, this increase in Cdk1 mirrors that of cyclin B 24 , indicating a coordinated expression of the two MPF subunits (Fig. 2).</p><p>MetaCore GO analysis of proteins in Group B, which increase following progesterone treatment, detected enrichment for processes involved in fertilization that can be attributed to the appearance of proteins such as ZP2, ZP3, ZP4, and ZPAX, glycoproteins that confer fertilization competency on the egg, and uroplakin 1b, a component of the complex that is believed to act as the sperm receptor 25,26 . Enriched molecular functions for this group include calcium transport 27,28 , phosphatidylinositol signaling 29,30 , and adenylate cyclase activity 31,32 that, likewise, are all essential activities for egg activation at fertilization and exit from meiotic arrest. Thus, many of the quantitative increases in the proteome during oocyte maturation manifest the cell's preparation for the next developmental event, fertilization.</p><p>Conversely, the proteins in Group F decline immediately after progesterone treatment (TP2). This group includes proteins involved in the storage and activation of maternal RNA such as ZAR1, a repressor of translation, that regulates the early expression of Mos (prior to GVBD) and late expression of Wee1 (post GVBD). The temporal difference in the translational activity of these mRNAs may result from diminishing levels of ZAR1 and its lower affinity for Mos relative to Wee1 mRNAs 33 . Another member of this group, cytoplasmic polyadenylation element binding protein 1 (CPEB1), acts as a translation repressor and activator. Earlier analyses by western blotting showed an approximate 70% decrease of CPEB1 during oocyte maturation that enables the expression of "late class" mRNAs (e.g., cyclin B) at or following GVBD (TP3) 34,35 . Over the interval from 45 minutes (TP2) to six hours post-progesterone treatment (TP3), we measure an 80% decrease in CPEB1. CPEB1 rises at fertilization only to drop again at cytokinesis (Fig. 2). Thus, the programmed degradation of CPEB1 needed for oocyte maturation appears to repeat itself during the first cell division cycle. Unexpectedly, Cdc20 (Fizzy), an activator of the APC/C complex, is also found Group F. This decline at GVBD may represent its partial degradation as part of a mechanism that fine tunes APC/C activity that enables release from meiosis I and arrest in metaphase II 5,36 .</p><p>Group C contains proteins whose expression increases modestly at fertilization and then remains constant. The proteins in this group are mostly associated with general metabolism and the cytoskeleton.</p><p>A second increase in expression occurs upon completion of the first cell division (Group E). Proteins involved in DNA replication, cell-cell contact, and ubiquination are predictably present in this group. Oocytes arrested in PI have lost the ability to enter S-phase, which has been attributed to the absence of Cdc6 in the oocyte that prevents formation of replication initiation complexes, although other activities associated with GVBD could not be excluded 37,38 . In support of this mechanism, we detect a distinct drop in the amount of Cdc6 at GVBD (Fig. 2).</p><p>The GO analysis of group E unexpectedly showed functional enrichment of proteins involved in mRNA binding and process enrichment in RNA transport and mRNA processing that can be attributed to the appearance of several splicing factors and RNA helicases. This increased synthesis of proteins involved in mRNA metabolism would support recent evidence for limited, but essential, transcriptional activity prior to the midblastula transition 39 .</p><!><p>The results of the quantitative proteomic analysis demonstrate that progression from a mature oocyte to an egg and subsequent fertilization to produce a zygote entails the differential expression of a limited number of critical proteins encoded mostly by maternal RNAs, but not a large-scale reprogramming of the proteome. Therefore, the precise temporal regulation of protein activities that underlie the transitions between these developmental stages must rely on well-documented post-translational modifications, especially protein phosphorylation. A galaxy plot of changes in the amount of individual proteins versus changes in the amount of phosphorylation at individual sites at each time point relative to stage VI oocytes corroborates this point (Fig. 3 and Supplementary Fig. S2).</p><p>The elliptical profiles demonstrate that, in general, phosphorylation is changing to a much greater degree than protein levels. ANOVA analysis 40 suggests that this differential distribution is significant (α value of 0.05). There are two particularly notable features of this plot. Transition to a mature oocyte exhibits the greatest increase in phosphorylation, yet is the only time point with an appreciable decrease in the amount of several proteins. This result is consistent with an increase in the activities of several kinases, triggered by progesterone, that leads to targeted destruction of proteins involved in masking mRNA as well as ubiquitin-dependent processes needed for release from meiotic arrest. The opposite behavior is seen with the stage 2 (two-cell) embryo, where there is an increase in the amount of protein, but relatively smaller changes in phosphorylation. This behavior is also seen in the hierarchical clustering of changes in phosphoproteins (Fig. 4, Supplemental Fig. S3, Supplemental Table S5). While there are appreciable changes in phosphorylation during oocyte maturation and fertilization, all subsequent time points for the zygote show a much smaller dynamic range.</p><p>Oocyte maturation. Progression from a stage VI oocyte (TP1) arrested in PI to a mature egg arrested in MII (TP3) is orchestrated by sequential unmasking of maternal mRNAs and the resulting activation of multiple kinases, producing an exceptionally dynamic phosphoproteome that coordinates this complex transition 41 . By the end of oogenesis, a small amount of inactive MPF has accumulated. PI arrest is maintained by phosphorylation of the Cdk1 subunit of MPF by Myt1 kinase. Progesterone stimulation ultimately leads to the dephosphorylation of Cdk1 by Cdc25 phosphatase to produce active MPF and progression to metaphase II arrest (Fig. 5A). The earliest steps in the activation of MPF can be traced to the translational derepression of Ringo mRNA. The expression of RINGO/CDK acts on two levels: direct phosphorylation and inactivation of Myt1 23 and phosphorylation of Musashi that, in turn, allows expression of Mos and activation of the MAP kinase pathway that also targets Myt1 42 . In parallel, Plk1 activates Cdc25 by phosphorylation. The combined inhibition of Myt1 and activation of Cdc25 leads to the burst of MPF activity that is then maintained by various amplification loops as well as sustained expression of Mos 41 .</p><p>Our analysis has captured several of the temporal changes in phosphorylation that control cell cycle progression following progesterone stimulation (Fig. 5B). Groups pE and pH contain proteins that exhibit a marked increase in phosphorylation at GVBD and include: Plk1, Mek1, p42MAPK, Rsk1/2, Cdc25, MELK (maternal embryonic leucine zipper kinase), Cdc20, and Apc1 (a subunit of the anaphase-promoting complex/cyclosome). The phosphorylation of Plk1 results from the decrease in intracellular cAMP upon progesterone stimulation; the activated Plk1 then phosphorylates Cdc25. In parallel, de novo synthesis of Mos initiates the MAP kinase pathway made up of Mek1, p42MAPK, Rsk1/2, and Myt1. We detect highly phosphorylated forms of all these proteins at GVBD (TP3) except for Myt1. Inactivation of Myt1 is the result of two kinase activities with initial phosphorylation by RINGO/CDK that facilitates subsequent recruitment and phosphorylation by Rsk1/2 43 . We did not detect phosphopeptides corresponding to the target sites of either kinase, accounting for the apparent absence of phosphorylation of Myt1 at GVBD. However, we did detect phosphorylation of the protein after fertilization at sites in the N-terminal kinase domain and the C-terminal Cdk1/cyclin B interaction domain; none of these sites is predicted to be a target of cyclin-dependent kinases. Apparently, phosphorylation of Myt1 in the zygote is used for a distinctly different purpose compared with its role in regulating progression through meiosis I.</p><p>Entry into anaphase requires the ubiquitin-dependent destruction of the cyclins and securin that is mediated by APC/C. Cdc20 is a positive activator of the APC/C that is negatively regulated through phosphorylation by Bub1. The single Cdc20 phosphopeptide that we detect at TP3 does not correspond to any of the reported sites phosphorylated by Bub1 44 , but instead has a putative consensus sequence for Cdk1 embedded with one for CK2, strongly suggesting that phosphorylation at this site is not involved in negative regulation of Cdc20, but rather its activation. We also detect phosphorylation of the Apc1 subunit of APC/C at TP3. It was recently shown that phosphorylation at multiple CDK sites in the loop domain of Apc1 is required for binding of Cdc20 36 . We detect phosphorylation at several of these sites as well as at other positions. We pose that phosphorylation of Cdc20 and APC/C by Cdk1 is required for activation of the complex and exit from meiosis I. As expected, groups pE and pH also include proteins involved in the regulated translation of masked transcripts (eIF4G, cytoplasmic poly(A) binding protein, eIF3a, eIF4E binding protein, ELAV1). We also find an appreciable number of nucleoporins in these groups. Phosphorylation is required for the disassembly of the nuclear pore complex and, hence, GVBD. Many of these proteins are targeted by kinases that are activated upon oocyte maturation, especially Cdk1, but also PKC, Plk1, and Aurora A (Eg2) 45 . Proteins in these groups are associated with cell cycle processes, chromosome organization, nuclear envelope organization, RNA/nucleic acid binding, cytoskeletal interactions, and nuclear pore structure. GO results also show enrichment for proteins involved in cell-cell adhesion; indeed, expression of some cadherins from maternal mRNA during Xenopus oocyte maturation has been reported 46,47 .</p><p>Group pF contains proteins that exhibit an increase in phosphorylation at 45 min (TP2) and includes CPEB1, an early target of Aurora A. Phosphorylation of S174 enables the interaction of CPEB1 with CPSF, which triggers cytoplasmic polyadenylation of early maternal mRNA 48 . Subsequent phosphorylation by Cdk1 and Plk1 marks CPEB1 for proteolytic degradation 35,49 . We did not detect a phosphopeptide corresponding to S174, but did identify four other sites of phosphorylation. Two residues, S210 and S223, show some phosphorylation at 45 minutes that reach a maximum after GVBD, while T246 and S248 exhibit a marked increase only after GVBD. It has been demonstrated that Cdk1 targets S210 and S248, with phosphorylation of the former site required for high affinity binding of the prolyl isomerase, Pin1, whose action is required for the ubiquitin-dependent destruction of CPEB1 50 . While phosphorylation at S223 and T246 has not been previously reported, deletion of the region encompassing residues 211 to 290 prevents association of Pin1 with CPEB1 51 . Pin1 (group pD) is found in an inactive form associated with CPEB1 in oocytes. Upon progesterone stimulation, Pin1 is rapidly converted into an enzymatically active form by dephosphorylation at S68 (S71 in human Pin1) 51,52 . We detect a marked dephosphorylation of S68 at 45 min that continues through GVBD (Fig. 5B). Phosphorylation of Pin1 is restored following GVBD (TP4) and then declines after fertilization as the egg progresses through its first mitotic cycle. Finally, phosphorylation is reestablished immediately after cell division (TP7). This striking cyclic behavior suggests that Pin1 phosphorylation/dephosphorylation is synchronized with the cell cycle, with its activity fully expressed during M phase. In addition, we have detected unreported phosphorylation at the adjacent serine residue (S67) that exhibits the same cyclical behavior.</p><p>GO analysis of Group pF detected enrichment for miRNA biogenesis and regulation by miRNA, which is particularly notable in light of recent studies that determined CPEB1 interacts with miRNA complexes through Ago2 to temporally regulate the translation of cyclin E1 mRNA during oocyte maturation 53 . In addition, miRNA complexes are required to maintain the level of oocyte Myt1 and, thus, arrest in PI 54 . The GO analysis is not only consistent with emerging evidence for translational control by miRNA complexes during oocyte maturation, but also indicates that this mechanism, which may be more widespread than previously realized, is regulated by phosphorylation.</p><p>Groups pG and pI also contain proteins whose phosphorylation increases immediately after progesterone treatment, but with somewhat different kinetics at subsequent time points. The former group shows GO enrichment in RNA/mRNA metabolic processes and RNA-directed RNA polymerase activity, which may be related to the miRNA activities described above. Also included in these groups is MARCKS, a cellular substrate for protein kinase C, which becomes rapidly activated upon hormone treatment and controls a remodeling of the cytoplasm necessary for fertilization competency 55 and HURP, a protein involved in spindle formation that is associated with Aurora A 56 .</p><p>Conversely, groups pB and pE are characterized by dephosphorylation immediately following progesterone treatment. GO analysis shows functional enrichment in both groups for RNA binding, poly(A) binding, translation initiation factor activity, and cell-cell adhesion. These groups includes proteins involved in the expression of masked mRNAs, including Pin1 (discussed above), eIF4G-1, which appears to play a role in the translation of maternal mRNAs that code for proteins essential for the completion of meiosis I 57 , and symplekin, which is an essential component of the complex that directs cytoplasmic polyadenylation of maternal mRNAs 58 .</p><p>Fragile X mental retardation protein (FMRP) (group pD) is highly expressed in Xenopus oocytes 59 . The protein can repress translation by a direct interaction with ribosomes 60 . Perhaps of greater relevance in this case, FMRP in mouse neurons forms a complex with Ago2:miRNA to repress translation of specific mRNAs, which is reversed by dephosphorylation 61 . A subset of miRNAs is enriched in Xenopus oocytes, but is nearly absent in eggs 62,63 . Two (T505 and S506) out of four phosphorylated residues detected in oocyte FMRP1 show a marked decline immediately after progesterone treatment, which would be consistent with a role for miRNA-directed translational repression of maternal mRNAs that become activated upon oocyte maturation through a mechanism analogous to that in neurons.</p><p>Proteins in group pB are phosphorylated in stage VI oocytes (TP1) and do not immediately change upon progesterone treatment (TP2), but are dephosphorylated by GVBD (TP3). GO molecular function enrichment includes RNA, poly(A), and nucleic acid binding, while biological process enrichment includes cell cycle and protein dephosphorylation. The most noteworthy member of this group is Cdk1 whose transient dephosphorylation enables MPF to promote progression into meiosis II. The measured dephosphorylation of Mapk12 (p38γ/SAPK3) likely represents its inactivation as an independent kinase that targets Cdc25 and has been proposed to be a pathway to oocyte maturation that is complementary to the canonical MAP kinases 64 . Similarly, another member of this group, protein kinase Cδ has been reported to induce meiotic maturation when injected into oocytes 65 . The FRGY proteins are part of the storage RNP complexes that form on maternal mRNAs and are highly phosphorylated in oocytes on multiple CK2 sites 66 . We detect 10 sites of phosphorylation that match those identified earlier in oocytes and speculate that the observed dephosphorylation of this family is part of the unmasking process.</p><p>Fertilization. The marked increase in phosphorylation of proteins at GVBD (TP3) in groups pE and pH is essentially reversed at fertilization (TP4). These groups represent approximately 400 proteins and contain members of the pathways that lead to GVBD and sustain arrest of the egg at meiosis II (e.g., MEK1, Rsk1/2, Plk1, Plk3, Cdc25, Cdc20, APC subunit 1). Several of these cell cycle associated proteins become phosphorylated again during mitotic M phase (TP6). On the other hand, other members of these groups are associated with translational activation of maternal mRNA (e.g., CPEB1, eIF4G2, eIF3a, eIF4E binding protein, ELAV1) and these generally remain unphosphorylated after fertilization.</p><p>Several proteins in groups pC and pD, which exhibit dephosphorylation immediately after exposure to progesterone (TP2), become phosphorylated again at various times following fertilization. GO analysis detects some enrichment for nucleic acid binding activity, DNA metabolism, cell adhesion, mRNA metabolism, cell cycle, cytoskeletal organization and organelle organization. However, many of the proteins in these two groups are involved in diverse metabolic processes that presumably reflect the energetic demands of the approaching cell division.</p><p>The majority of proteins that comprise group pA show an increase in phosphorylation at various times following fertilization, although for a clear subset of this group, this increase has occurred at GVBD. Predominant GO process enrichment for this group includes various aspects of the (mitotic) cell cycle. An example of proteins that show cyclic phosphorylation are lamins B1 and B3, which exhibit maxima at GVBD (TP3) and again during formation of the cleavage furrow (TP6). The sites detected here are mostly targets of CDK1 and phosphorylation at these positions is required for disassembly of lamin filaments during mitosis 67 . Relatedly, CLASP1 is a multifunctional protein that is directed to the spindle midzone and kinetochores and plays an essential role in microtubule polymerization and bundling during mitosis 68 . CLASP1 is phosphorylated at many sites in a complex temporal pattern that seemingly reflects its dynamic interaction network. There are sites that exhibit a mutually exclusive periodicity that reflect its distinct roles. Maximum phosphorylation of S1227 and S1231 during M phase (TP3 and TP5/6) seemingly reflect its activity in spindle organization, whereas several residues in the amino terminal half of the protein, which show greatest phosphorylation during cortical rotation and formation of the cleavage furrow (TP4 and TP6, respectively), more likely reflect the role of CLASP1 in polymerization of noncentrosomal microtubules. CLASP1 is typical of several proteins in the database that show opposing sites of phosphorylation over the time interval studied here, which most certainly reflects regulated changes in protein activity at particular developmental stages. Indeed, this rapid interconversion, which must occur within the unusually short cell cycle of a cleavage stage embryo, could only be accomplished by reversible post-translational modifications such as phosphorylation.</p><p>GO functional enrichment in group pA includes RNA/nucleic acid binding, translation initiation, and several aspects of cell-cell adhesion that likely derive from sperm-egg fusion. While the range of biological function increases in this group compared to others, there remain examples of cell cycle control through phosphorylation such as MELK, a target of MPF and a possible regulator of the Cdc25 phosphatase 69 . The pre-replicative complex protein, MCM2, also falls into group pA. We detect phosphorylation at CDK and CK2 sites that have also been identified in human MCM2 38,70 . Phosphorylation of MCM2 by cyclin E/Cdk2 is required for assembly of the pre-replication complex and cell cycle reentry 71 . Thus, the dephosphorylated form of MCM2 in mature oocytes, along with the low levels of Cdc6, likely contributes to the inability to initiate replication.</p><p>A remarkable number of proteins in group pA are involved in spindle assembly. PCM1 is required for radial organization and anchoring of microtubules to the centrosome 72 . MAP4/p220 (microtubule associated protein 4) is extensively phosphorylated at TP3 and TP6 consistent with a role in determining meiotic and mitotic spindle assembly 73 . Mutation of sites phosphorylated by MPF alters MAP4 affinity for microtubules resulting in compromised chromosome movement 74 . While we detect phosphorylation of MAP4 at MPF and MAP kinase sites reported earlier 74 , there is a unexpected variety of additional sites that follow the same temporal pattern, including target sites for ATM, DNA-PK, CK1, and several sites for CK2. Similarly, another member of group pA, INCENP, is a component of the chromosomal passenger complex that interacts with CLASP1 and is required for spindle assembly. Plk3, unlike Plk1, has functions beyond cell cycle regulation. Although there have been no reports of phosphorylation of Xenopus Plk3, the modification has been detected in the mammalian cells and connected to a variety of processes: cell cycle progression, DNA damage, mitotic spindle disruption, and stress responses 75,76 .</p><p>GO analysis of group pA detected considerable enrichment of processes involved in intracellular/organelle organization that can be accounted for by the large number of proteins involved in assembly of the mitotic spindle and chromosome separation, as well as the phosphorylation of several proteins involved in nucleolar (e.g., nucleolin, Ki-67, nopp130, nop132) and nuclear membrane (e.g., lamins b1 and b3) structure.</p><!><p>Unsupervised clustering of phosphopeptides generated 22 groups with similar temporal patterns of phosphorylation (Supplemental Fig. S4, Supplemental Table S6). Consensus sequences (N 2 S/TN 4 ) were generated for each group in an effort to identify kinase activities that predominate during the different developmental stages. As might be expected, the vast majority of consensus sequences have a proline residue immediately flanking the phosphoamino acid and frequently at the −2 position as well (Fig. 6A). The cyclin dependent kinases, MEK1 and p42MAPK, all target these sequences and clearly account for the bulk of detectable protein phosphorylation during oocyte maturation as well as the first mitotic cell cycle.</p><p>A considerable number of sites predicted to be phosphorylated by CK2 were also identified. In many cases it appears that initial phosphorylation by a proline-directed kinase then created a site that was subsequently phosphorylated by CK2. Indeed, CK2 is one of just four known kinases whose recognition site can be created through a mechanism known as "hierarchical" or "primed" phosphorylation 77 . This phenomenon accounts for the ubiquitous phosphoserine stretches that occur throughout the human phosphoproteome 78 that we have also detected in this study.</p><p>The majority of clusters in which dephosphorylation occurs after progesterone treatment are not dominated by proline-directed kinase sequences; rather, as a group they are enriched in consensus sequences for CK1, PKA, and especially CK2 (Fig. 6B). In many instances, CK1 and CK2 appear to have functioned as priming kinases. In addition, consensus sequences for glycogen synthase kinase 3 (GSK3) occur in several proteins found in these clusters. GSK3, as a member of several disparate signaling pathways, phosphorylates many proteins, often in combination with CK1 and/or PKA (Fig. 7) 79 . Notably, GSK3 activity contributes to PI arrest and its inactivation in response to progesterone is necessary for oocyte maturation 80 . Upon progesterone treatment, decreased level of cAMP will reduce the activity of phosphorylase kinase that, combined with the inactivation of GSK3, should result in dephosphorylation of glycogen synthase. Indeed, we detect loss of glycogen synthase 1 phosphorylation at target sites for phosphorylase kinase, GSK3, PKA and CK1.</p><p>Another consensus sequence that occurs frequently is that for G protein-coupled receptor kinase (GRK). This observation is significant because prophase arrest requires constitutive G protein signaling and GRK3 has been implicated specifically in Xenopus oocyte maturation 81 . Phosphorylation of the receptor and subsequent binding of β-arrestin appears to be a common mechanism of desensitization. However, GRKs are known to have targets beyond receptor proteins, indicating their activity is not necessarily limited to just the immediate signaling pathway 82 . Additional targets of the GRK could serve as a means to amplify the signal of the agonist allowing for a more global and rapid response.</p><p>Our temporal analysis of the Xenopus phosphoproteome has detected distinct quantitative changes in phosphorylation upon oocyte maturation that correlate exceptionally well with the decline of kinase activities that maintain the cell at PI arrest (PKA, GSK3) and with the corresponding activation of MAPK and cyclin dependent kinases that control completion of meiosis I and arrest at meiosis II. While these are the principle regulatory kinases, it is clear that general kinases such as CK1 and CK2 play an essential supporting role that either amplifies or stabilizes the effect of the initial phosphorylation event (Fig. 7).</p><!><p>A fully-grown stage VI oocyte has accumulated most of the proteins that will be needed for fertilization and progression through the first mitotic cell cycle. The dormant or masked mRNAs of the immature oocyte that become activated upon progesterone-dependent maturation to an egg only modestly change the proteome. A notable exception is the synthesis of proteins necessary for fertilization. The vast majority of phosphorylation that occurs during egg formation can be attributed to proline-directed kinases from the MAPK pathway or cyclin-dependent kinases. However, we find that a considerable number of proteins possess multiple site of phosphorylation that are often clustered, which is consistent with recent examples in which GSK3 and CK1 work together in a hierarchical manner to create stretches of phosphorylated residues 77 . Examination of phosphopeptides identified here suggests that proline-directed kinases work with CK2, CK1, and possibly PKA in a similar fashion. This hyperphosphorylation may be a means to amplify signals or, relatedly, stabilize a particular protein conformation. While GO analysis yielded expected enrichments in molecular function and biological process related to translational control, cell-cycle regulation, and spindle organization, those related to regulation by miRNA were unexpected 83 . Thus, recent reports of translational control of maternal mRNAs by miRNAs may be more widespread and, like cytoplasmic polyadenylation, regulated by phosphorylation 53,54 . The excellent agreement between measurements made in this study with well-documented pathways that control oocyte maturation and the mitotic cell cycle engender confidence in the phosphoproteome data and its value for investigations into other processes controlled by this post-translational modification.</p><!><p>Materials. Xenopus laevis animals were purchased from Nasco (Fort Atkinson, WI USA). All animal procedures were performed according to protocols approved by the University of Notre Dame Institutional Animal Care and Use Committee. Complete, mini protease inhibitor cocktail and phospho-stop inhibitors, provided in EASYpacks, were purchased from Roche Diagnostics (Indianapolis, IN USA). Human Chorionic Gonadotropin (HCG), bovine pancreas TPCK-treated trypsin, progesterone, 0.5 M triethylammonium bicarbonate buffer, pH 8.5-8.6 (dissolution buffer for ITRAQ labeling), and cysteine were purchased from Sigma-Aldrich (St. Louis, MO USA). Pierce C-18 Spin Columns, and Pierce BCA Protein Assays were purchased from Thermo Scientific (Marietta, OH USA). Sep-Pak Vac (1 cc, 100 mg and 0.5 cc, 50 mg) C-18 Cartridges were purchased from Waters (Ireland). Centrifugal Filter Units with a 30,000 MWCO and Ziptips were purchased from Millipore (Carrigtwohill, CO USA). iTRAQ Reagent 8 Plex Kit and labeling reagent was purchased from Sciex (Framingham, MA USA). Oocyte and embryo collection and culture. A sample of ovary tissue was surgically removed, placed in OR2 buffer and oocytes were manually defolliculated. Stage VI oocytes were selected and maturation was induced by incubation at 18 °C with 10 µg/mL progesterone in OR2 buffer. For egg collection, female X. laevis were injected with 600 units of HCG 12-15 hours prior to spawning; testes were isolated from anesthetized male frogs. Eggs and minced testes were combined in a total volume of 2 to 5 ml 1/3 MMR (Marc's Modified Ringers) and incubated for 10 minutes. The sample was then flooded with 1/3MMR and incubated for another 20 minutes. Fertilized eggs were washed with 2% L-cysteine for 4 minutes to remove the jelly coat. Embryos were then allowed to develop at ambient temperature. All samples were snap frozen in liquid nitrogen immediately following collection to preserve the experimental time point.</p><!><p>Each sample was suspended in 600 μL NP40 buffer (containing phospho-stop and protease inhibitor) and processed as described previously 84 . Samples were prepared in biological duplicate.</p><!><p>Each protein sample (100 µg) was labeled according to the manufacturer's protocols.</p><p>Proteome fractionation. The sample complexity was reduced using high pH reversed phase fractionation prior to mass spectrometry analysis which has been described 84 .</p><!><p>A nanoACQUITY UltraPerformance LC (UPLC © ) system (Waters, Milford, MA USA) was used for peptide separation. Buffer A (0.1% FA in water) and buffer B (0.1% FA in ACN) were used as mobile phases for gradient separation. Peptides were automatically loaded onto a commercial C18 reverse phase column (Waters, 100 µm ID, 100 mm, 1.7 mm particle, BEH130C18, column temperature 40 °C) with 2% buffer B for 10 minutes at a flow rate of 1.00 µL/ min, followed by a 4-step gradient separation, 1 min from 2% to 8%, 87 minutes to 30% B, 1 minute to 80% B, and maintained at 80% B for the next 10 minutes. The column was then equilibrated for 10 minutes with 2% B before analysis of the next sample. The eluted peptides from the C18 column were pumped through a capillary tip for electrospray, and analyzed by a Q-Exactive ™ HF mass spectrometer (Thermo Fisher Scientific). For each sample, approximately 2 μg of peptide was analyzed Predicted kinase substrate sites were identified using PhosphoMotif Finder 86 . GRK (G protein-coupled receptor kinase), GSK3 (glycogen synthase kinase 3), CK2 (casein kinase 2), PKC (protein kinase C), PKA (protein kinase A), CDK2 (cyclindependent kinase 2), CK1 (casein kinase 1). CK2* indicates that phosphorylation at this site requires prior phosphorylation of a proximal priming site. per run. Electrospray voltage was 2.0 kV, and the ion transfer tube temperature was 280 °C. The S-Lens RF level was 60.00. Data acquisition was programmed in data-dependent acquisition (DDA) mode. For analysis using the Q-Exactive ™ HF, instrument settings included: a top 12 method, full MS scans were acquired in Orbitrap mass analyzer over 350-1500 m/z range with a resolution of 60,000, and the number of micro scans set to 1. Automatic gain control (AGC) target value was 3.00 E + 06, the maximum injection time was 30 ms. For MS/MS scans, the twelve most intense peaks with charge state ≥2 and <6 were sequentially isolated and further fragmented in the higher-energy-collisional-dissociation (HCD) cell following one full MS scan. The normalized collision energy was 33%, and MS/MS spectra were acquired in the Orbitrap mass analyzer with resolution 30,000. The first fixed mass was 100.0. The number of micro scans was 1 and the ion selection threshold was 1.0 E + 05 counts.</p><p>Phosphorylation enrichment. Sample preparation was scaled to prepare 500 µg for each channel. A 50 µL aliquot of Ti-IMAC was used for each mg of labeled protein. Beads were washed 3 times with 80% ACN + 6% TFA. Sample was then suspended in 80% ACN + 6% TFA and combined with the beads. The beaded sample was vortexed for 20 minutes at ambient temperature. Beads were washed 3 times with 80% ACN + 6% TFA three times before being washed 80% ACN and 80% ACN + 0.5 M glycolic acid respectfully. The sample was then eluted two times into a new tube of 50% ACN + 1% ammonium hydroxide. Samples were then lyophilized and reconstituted in 0.2% FA at ~1 µg/µL, before being placed on the UPLC-MS/MS system. Phosphoproteome fractionation. Samples were fractionated on a Thermo Ultimate 3000 at a flow rate of 0.5 mL/min. Sample was loaded onto a column (3 × 150, 1.7 µm, BEHC18, column temperature 65 °C). The mobile phase gradient was generated using buffer A (10 mM ammonium formate, pH 10) and buffer B (80% MeOH in 10 mM ammonium formate, pH 10). The sample was loaded onto the column followed by a 21-minute wash at 0% B, and then separated by a 3-step, 15-minute gradient, at a flow rate of 0.5 mL/min. 2 minutes at 0 to 25% B, followed by 8 minutes for 25-75% B, 1 minute of 75-100% B, and then maintained for 2.4 minutes at 100% B before re-equilibration at 0% B for remainder of the 15-minute gradient. Eluate from 3-14 minutes was collected at 41 second intervals and then every 9 th sample was combined (i.e., sample 1 added to 9, 2 to 10, etc.) to yield 8 total samples. Samples were then dried down and reconstituted in 0.2% FA before being placed on the MS.</p><!><p>A Thermo Fisher RPLC nano system was used for phospho-peptide separation. Buffer A (0.2% FA in water) and buffer B (80% ACN in 0.2% FA) were used as mobile phases for gradient separation. Peptides were automatically loaded onto a commercial C18 reverse phase column (Waters 75 µm × 25 cm, BEHC18, column temperature of 65 °C) with 0-4% buffer B for 11 minutes at a flow rate of 0.375 µL/min, followed by a 2-step gradient separation, 64 min from 4% to 55%, 1 minute from 55-100% B, and then maintained at 100% for the next 4 minutes. The column was then equilibrated for 10 minutes with 2% B before analysis of the next sample. The eluted peptides from the C18 column were pumped through an integrated capillary emitter for electrospray, and analyzed by an Orbitrap Fusion Lumos mass spectrometer (Thermo Fisher Scientific). For each sample, a 4 µL injection of peptide was placed on the column per run. Electrospray voltage was set at 2.1 kV and the ion transfer tube temperature was 280 °C. The S-Lens RF level was 20. Data acquisition was programmed in data-dependent acquisition (DDA) mode. For analysis using the Fusion, instrument settings included: a top 2 seconds, full MS scans were acquired in Orbitrap mass analyzer over 350-1500 m/z range with a resolution of 60,000, and the number of micro scans set to 1. The automatic gain control (AGC) target value was 1.0E6, the maximum injection time was 100 ms. For MS/MS scans, the most intense peaks with charge state ≥2 and <6 were sequentially isolated and further fragmented in the higher-energy-collisional-dissociation (HCD) cell following one full MS scan. The normalized collision energy was 35%, and MS/MS spectra were acquired in the Orbitrap mass analyzer with resolution 30,000 in centroid mode. The first fixed mass was 100.0. The number of micro scans was 1 and the ion selection threshold was 2.0 E + 05 counts. Peptide match and excludes isotopes were turned on. A maximum injection time of 118 ms was used for MS/MS collection. Data searching. Raw files were searched using MaxQuant 1.5.5.1 and the Genome 9 database (downloaded 31 October 2016 from Xenbase) using the default settings for deep proteome analysis. Carbamylation was set as a fixed modification. Variable modifications for the protein analysis included: Acetyl (K), Acetyl (Protein N-term), Deamidation (NQ), and Oxidation (M). For phosphoproteomic analysis the variable modifications used were: Acetyl (K), Acetyl (Protein N-term), Deamidation (NQ), Gln-> pyro-Glu;GlyGly (K), Oxidation (M) and Phospho (STY). A maximum of two missed cleavages was allowed. The minimum charge of the peptide was set to 2 and the max charge was set to 7. The maximum number of modifications per peptide was set to 8, and the maximum mass of a peptide was set to 4000 Da. A minimum of 1 unique peptide was required for protein identification. The false discovery rate (FDR) was set to 0.01 on peptide and protein levels. Data analysis. Data was corrected for each iTRAQ channel and normalized according to the mean of the individual protein across all 7 time points for both proteins and phosphorylation sites. To be included in the analysis, the protein or phosphorylation site was required to have an intensity value for analysis in our experiment.</p><p>Histograms, heatmaps, clustering, and individual graphs were generated in Matlab using default algorithms and the mean intensity of the protein or site. Protein lists for individual groups derived from the heatmaps were uploaded into the MetaCore TM software suite and subjected to analysis with the Gene Ontology (GeneGo) algorithms. Two GO functional ontologies, biological processes and molecular function, were generated using GO Term classification software CateGOrizer (v. 3.218).</p><p>For the clusters analyzed for consensus sequence homology, GProx was used to organize clusters with the following traits: no reference, no standard, 0.2 membership requirement, and threshold of −0.32-0.26 (equating to a 0.8-1.2-fold change minimum) 85 . Consensus sequences were aligned by the active phosphorylation site and generated using http://weblogo.berkeley.edu/logo.cgi. PhosphoMotif Finder was used to identify phosphorylation motifs within the sequence of phosphopeptides 86 .</p><p>Data availability. All raw files have been uploaded to the MassIVE database, including, Phospho enriched, proteome data, MaxQuant Processed files and MaxQuant search parameters (MQpar) (XML) at the following exchange ftp://MSV000081416@massive.ucsd.edu.</p>

Scientific Reports - Nature

Phospholipase iPLA2\xce\xb2 Averts Ferroptosis By Eliminating A Redox Lipid Death Signal

Ferroptosis, triggered by discoordination of iron, thiols and lipids, leads to accumulation of 15-hydroperoxy-arachidonoyl-PE (15-HpETE-PE) generated by complexes of 15-lipoxygenase (15-LOX) and a scaffold protein, PEBP1. As Ca2+-independent phospholipase PLA2 (iPLA2\xce\xb2, PLA2G6/PNPLA9 gene), can preferentially hydrolyze peroxidized phospholipids, it may eliminate ferroptotic 15-HpETE-PE death signal. Here we demonstrate that by hydrolyzing 15-HpETE-PE, iPLA2\xce\xb2 averts ferroptosis whereas its genetic or pharmacological inactivation sensitizes cells to ferroptosis. Given that PLA2G6/PNPLA9 mutations relate to neurodegeneration, we examined fibroblasts from a patient with a Parkinson\xe2\x80\x99s disease (PD)-associated mutation fPDR747W and found selectively decreased 15-HpETE-PE hydrolyzing activity, 15-HpETE-PE accumulation and elevated sensitivity to ferroptosis. CRISPR-CAS9-engineered PNPLA9R748W/R748W mice exhibited progressive parkinsonian motor deficits and 15-HpETE-PE accumulation. Elevated 15-HpETE-PE levels were also detected in midbrains of rotenone-infused parkinsonian rats and \xce\xb1-synuclein mutant SNCA-A53T mice with decreased iPLA2\xce\xb2 expression and PD-relevant phenotype. Thus, iPLA2\xce\xb2 is a new ferroptosis regulator and its mutations may be implicated in PD pathogenesis.

phospholipase_ipla2\xce\xb2_averts_ferroptosis_by_eliminating_a_redox_lipid_death_signal

INTRODUCTION<!>15-HpETE-PE is a preferred substrate of iPLA2\xce\xb2.<!>R747W mutation reduces iPLA2\xce\xb2 activity with 15-HpETE-PE.<!>Ferroptosis in fPDR747W and iPLA2\xce\xb2-deficient cells.<!>Parkinsonian phenotype of homozygous Pnpla9R747W mice.<!>PEox in midbrain of homozygous Pnpla9R748W mice.<!>PEox in midbrains of rotenone infused rats.<!>PEox in midbrains of A53T transgenic mice.<!>DISCUSSION<!>Materials.<!>Animal models.<!>Rat rotenone model:<!>A53T overexpression mouse model:<!>Pnpla9R748W mouse model:<!>Cell lines.<!>iPLA2\xce\xb2 knock-down.<!>PNPLA9 knockout in BeWo cells.<!>LPCAT3 knock-down (KD) in H109 and MEF cells.<!>Ferroptosis assay.<!>iPLA2\xce\xb2 cloning, expression, and purification.<!>Model system:<!>iPLA2\xce\xb2 activity in cells and tissues.<!>Western blot analysis.<!>LC-MS/MS analysis of ETE and 15-HpETE.<!>LC-MS/MS analysis of 15-HpETE-PE.<!>Analysis of oxygenated phospholipids by reverse phase LC/MS.<!>LC-MS analysis of catecholamine neurotransmitters and GSH.<!>Computational modeling of the full iPLA2\xce\xb2 dimer structure.<!>Construction and equilibration of neuronal lipid bilayer.<!>Molecular Dynamics (MD) Simulations of iPLA2\xce\xb2 dynamics, bound to the lipid bilayer.<!>in silico saturation mutagenesis analysis.<!>Pole test:<!>Rotarod test:<!>Catwalk test:<!>Statistical analysis.<!>Data availability:<!>Purification of iPLA2\xce\xb2 and analysis of 15-HpETE and its hydrolysis products.<!>Membrane composition and atomic structure used in MD simulations.<!>Comparison of conformational Flexibility of Acyl Chains.<!>Ability of the peroxidized group in 1-SA-2-HpETE-PE to come into close proximity of the protein surface.<!>Comparison of the intrinsic dynamics of 1-SA-2-15-HpETE-PE and1-SA-2-ETE-PE.<!>Homology modeling details of dimer iPLA2\xce\xb2 model.<!>Time evolution and histograms of interfacial contacts between iPLA2b residues and 15-HpETE-PE molecules observed in MD simulations.<!>iPLA2\xce\xb2-deficient cells are more sensitive to RSL3-induced ferroptosis compared to WT cells.<!>LPCAT3 KD protects mouse embryonic cells from RSL3 induced death.<!>Content of PE, oxygenated PE in midbrain of rotenone exposed rats and 8-months old WT and A53T mice.<!>

<p>The fidelity of biological systems depends on either re-programming or elimination of unnecessary or harmful cells and/or their organelles via several cell death programs1. Ferroptosis is a type of regulated cell death2 triggered by the discoordination of three major metabolic pillars – iron, lipids and thiols - culminating in lipid peroxidation3 The term "ferroptosis" reflects two specific roles of iron in: i) the production of membrane (phospho)lipid (PL) hydroperoxides and ii) "splitting" the weak hydroperoxy O-O bond4 to yield oxidatively-truncated electrophilic products, considered as the proximate executioners of ferroptotic death4. A seleno-peroxidase, glutathione peroxidase 4 (GPX4), reduces membrane phospholipid hydroperoxides to alcohols5, thus "neutralizing" the high-risk O-O-containing intermediates6.</p><p>Hydroperoxy-arachidonoyl-(C20:4, eicosatetraenoyl, ETE or AA)- and adrenoyl-(C22:4, docosa-tetraenoyl, DTE)-phosphatidylethanolamines (PE) (HpETE-PE and HpDTE-PE) have been identified as characteristic pro-ferroptotic signals2 Accordingly, esterification of these two fatty acid residues into PE by ACSL4 and re-acylation of lyso-PE (LPE) by LPCAT3 are important ferroptosis regulators7. High selectivity and specificity of ETE-PE and DTE-PE oxidation at their 15th and 17th carbons, respectively, is a feature of ferroptotic lipid peroxidation8 suggesting the involvement of enzymatic catalysis8. Indeed, 15-lipoxygenase (15-LOX) complexed with a scaffold-protein, PE-binding protein 1 (PEBP1), has been identified as a generator of pro-ferroptotic 15-HpETE-PE and 17-HpDTE-PE9. One can assume that a phospholipase A2 (PLA2) capable of hydrolyzing sn-2-oxygenated polyunsaturated fatty acid (PUFA)-PE residues would eliminate the ferroptotic signal whereas genetic or chemical ablation of this activity would be pro-ferroptotic. Ca2+-independent iPLA2β (PNPLA9 from PLA2G6 family) can hydrolyze oxidized phospholipids10, yet its catalytic competence towards ferroptotic signals has not been hitherto tested.</p><p>Ferroptosis has been implicated in a number of diseases, including Parkinson's disease (PD)11. Death of dopaminergic neurons in the substantia nigra pars compacta is one of the hallmarks of PD11. Enhanced oxidative stress and lipid peroxidation caused by the mishandling of iron and dopamine oxidation are important causative factors, along with the dysregulation of autophagy which otherwise regulates ferroptosis by degrading iron storage proteins12. Based on the striking similarity of the PD pathogenesis and pro-ferroptotic factors, it has been suggested that ferroptosis is involved in PD pathogenesis11. PNPLA9 mutations have been associated with several iPLA2β-related neurodegenerative diseases (PLAN). Based on the age of onset and progressive clinical features, several PLAN subtypes have been identified: infantile neuroaxonal dystrophy (INAD), atypical neuroaxonal dystrophy (ANAD) and a Parkinsonian syndrome with adult onset dystonia and autosomal recessive early-onset Parkinsonism13. Thus, we tested a hypothesis that deficiency in iPLA2β caused by genetic or pharmacological perturbations preserves the death signal, 15-HpETE-PE, hence propagates ferroptosis. Here we demonstrated that genetic abatement of iPLA2β in SH-SY5Y neuronal cells, H109 fibroblasts, and BeWo trophoblasts, or a naturally occurring mutation in the PNPLA9 gene encoding for iPLA2β (R747W) in fibroblasts from a dystonic PD patient (fPDR747W cells) result in: i) lowered hydrolytic activity towards 15-HpETE-PE, ii) elevated intracellular contents of 15-HpETE-PE, and iii) enhanced sensitivity to ferroptosis vs. WT controls. We documented that CRISPR-engineered Pnpla9R748W/R748W mice exhibit progressive Parkinsonian motor deficits along with 15-HpETE-PE accumulation. We detected decreased iPLA2β activity and elevated 15-HpETE-PE levels in the midbrains of rotenone-infused parkinsonian rats as well as 8-month-old SNCA-A53T mutant mice with decreased iPLA2β expression and PD-relevant phenotype14. These data along with computational modeling permitted us to decipher and predict the impact of impairments of iPLA2β and its R747W mutant on ferroptotic death relevant to PD pathogenesis.</p><!><p>To assess iPLA2β hydrolytic activity (Fig. 1a), we biosynthesized 1-SA-2-15-HpETE-PE and purified it to ~99% homogeneity (as evidenced by LC-MS analysis (Supplementary Fig. 1a). We also expressed, isolated and purified recombinant iPLA2β15 (Extended Data Fig. 1a) and tested its activity by LC-MS of the products: lyso-PE (1-SA-2-OH-PE) (Fig. 1a,b) and fatty acids (ETE or 15-HpETE) (Extended Data Fig. 1b,c). While both substrates (1-stearoyl-2-eicosatetraenoyl-phosphatidylethanolamine (1-SA-2-ETE-PE) and 1-stearoyl-2-15-hydroperoxy-eicosatetraenoyl-phosphatidylethanolamine (1-SA-2-15-HpETE-PE)) were readily hydrolyzed by the enzyme, the activity towards 1-SA-2-15-HpETE-PE was markedly higher than towards non-oxidized PE (Fig. 1a). Based on the kinetic characterizations of iPLA2β WT and iPLA2βR747W mutant and concentration dependencies with three substrates, 1-SA-2-ETE-PE, 1-stearoyl-2-15-hydroxy-eicosatetraenoyl-phosphatidylethanolamine (1-SA-2-15-HETE-PE) and 1-SA-2-15-HpETE-PE, we calculated the specificity constants using non-linear regression analysis (Extended Data Fig. 1d). In WT, the specificity constant for 1-SA-2-15-HpETE-PE and 1-SA-2-15-HETE-PE were 3.4- and 2.5-fold higher than the one for 1-SA-2-ETE-PE. The specificity constants for all substrates were decreased for the R747W mutant protein. Importantly, the specificity constant of the mutant protein towards 1-SA-2-15-HpETE-PE was decreased 2.6-fold (vs. WT) to the level close to that for WT with 1-SA-2-ETE-PE as the substrate (Supplementary Table 1). Thus 1-SA-2-15-HpETE-PE is the preferred substrate for the WT iPLA2β, and the specificity is obliterated in the R747W mutant.</p><p>We next performed computational modeling of the enzyme's interactions with membrane phospholipid substrates. Molecular dynamics (MD) simulations of a neuronal membrane containing 1-SA-2-15-HpETE-PE, 1-SA-2-ETE-PE and various PLs (Extended Data Fig. 2a,b and Methods), showed the ability of 1-SA-2-15-HpETE-PEs to migrate within the lipid bilayer and expose their peroxidized acyl chains, originally embedded in the membrane, to the surface (Fig. 1c and Movie 1–3), driven by the tendency of the peroxidized group to escape the hydrophobic environment.</p><p>Using MD simulations of the exposure and flexibility of sn-1 and sn-2 chains of 1-SA-2-15-HpETE-PE and 1-SA-2-ETE-PE (Extended Data Fig. 3a–b) we demonstrated that C15 carbon atom in 1-SA-2-15-HpETE-PE comes within 5Å of N-atom of the amino group but not in 1-SA-2-ETE-PE (Extended Data Fig. 3a, left, middle panels vs. right panel). This close proximity of the hydroperoxy-group to the membrane surface is similar to that observed in the Nuclear Overhauser experiments in support of the "whisker" model16. Indeed, the probability distribution of the sn-2 C15 (peroxidation site) positioning in the lipid bilayer shows that C15 of 1-SA-2-15-HpETE-PE is more exposed to the surface than its counterpart in 1-SA-2-ETE-PE (Extended Data Fig. 3b). Finally, we measured the probability distributions of C15 1-SA-2-15-HpETE-PE and 1-SA-2-ETE-PE distances from the iPLA2β surface and its catalytic site (Extended Data Fig. 4a,b). "Peroxidized" C15 of 1-SA-2-15-HpETE-PE has a higher probability to move closer to the protein surface and catalytic site, than C15 in 1-SA-2-ETE-PE. This corresponds with the experimentally observed higher iPLA2β activity towards 1-SA-2-15-HpETE-PE vs. 1-SA-2-ETE-PE. Additional computational experiments, compared with the published quantitative observables17 confirmed the statistical convergence of our modeling data (see Extended Data Fig. 5a,b). Notably, this behavior of 1-SA-2-15-HpETE-PEs is also similar to the migration of peroxidized triglycerides (TAG) to the surface of lipid droplets18.</p><p>We further examined the interactions of the catalytically active iPLA2β dimer surrounded by ankyrin (ANK) segments (Fig. 1d) with membrane-embedded 15-HpETE-PEs (Fig. 1e,f). We performed MD simulations of two models for the iPLA2β dimer: (i) dimer of CAT domain and (ii) dimer of CAT and ANK domains (see the Methods section and Extended Data Fig. 6 for modeling details). iPLA2β C651 was found to closely interact with the membrane, consistent with an earlier study15, thus bringing the catalytic dyad [S465 and D598] into proximity of the bilayer surface, and enabling frequent encounters with 1-SA-2-15-HpETE-PE (Fig. 2a,b). Quantitative analysis (Extended Data Fig. 7a,b) showed that select residues (e.g. K566PLP568 and H517 in monomer A and R656-P657, W661, T650, K665 and F668 in monomer B) had the strongest interactions with 1-SA-2-15-HpETE-PE.</p><!><p>Several mutations in the catalytic domain (CAT) of the PNPLA9 gene have been associated15 with neuropathies and PD19. We found that the activity of one of the mutant proteins (R747W, or R693W in the short variant) displayed a slightly lower activity towards non-oxidized ETE-PE and a markedly greater loss of activity towards 15-HpETE-PE (Fig. 1a). We next assessed the endogenous phospholipase activity in control and fPDR747W cells using two protocols to directly measure the hydrolytic potency of iPLA2β towards non-oxygenated substrates and 15-HpETE-PE. To distinguish between general PLA2 activity and specific iPLA2β activity, we employed an inhibitor, (S)-bromoenol lactone ((S)-BEL), a chiral-specific suicidal substrate that discriminates iPLA2β from all other phospholipases20 (Extended Data Fig. 8a). General PLA2 activity was slightly (1.3-fold) but significantly lower in fPDR747W cells than in control cells (Fig. 1g, left). The activity towards exogenously added 1-SA-2-15-HpETE-PE was sharply lower (>5-fold) in fPDR747W cells vs control cells (Fig. 1g, right). Western blotting (Fig. 1h) showed no differences in iPLA2β content between control and fPDR747W cells. Thus, R747W mutation specifically suppressed catalytic competence towards 15-HpETE-PE.</p><p>To explore the molecular basis of the lowered catalytic potency of PNPLA9 mutant R747W towards 1-SA-2-15-HpETE-PE, we repeated our simulations with the mutant dimer. The simulations showed that R747W mutation affected the interfacial packing between the two CAT domains of the dimer resulting in two effects: i) the catalytic site became less accessible, and ii) the association of the mutant with the membrane weakened (Fig. 2a,b). This supports the experimentally observed lower hydrolytic activity of the R747W mutant toward 1-SA-2-15-HpETE-PE.</p><!><p>Next we determined the content of phospholipids and phospholipid oxidation products in WT H109 and fPDR747W cells after triggering ferroptosis by a GPX4 inhibitor, RSL3. Using global phospholipidomics we found that phospholipid compositions of H109 cells and fPDR747W fibroblasts were very similar and included 86 species of PE, 78 species of PC, 28 species of PI and 32 species of PS (Fig. 3a and Extended Data Fig. 8b). PE included diacyl- (55.2±0.8 and 56.9±0.6% of total PE) and alkenyl- (44.8±0.8 and 43.1±0.6% of total PE) molecular species in H109 cells and fPDR747W cells, respectively. The contents of diacyl-PE species containing arachidonic, adrenic and docosahexaenoic acid in H109 and fPDR747W cells were 17.2±0.2, 7.4±0.1, 0.8±0.1% of total PE species and 15.9±0.1, 9.6±0.1 and 1.0±1.1% of total PE, respectively. No significant differences in the contents of alkenyl-PE species were detected: 24.4±0.5, 8.4±0.2, 1.9±0.1% of total PE species and 20.8±0.3, 10.3±0.2 and 2.0±0.1% for arachidonoyl, adrenoyl and docosahexanoyl-PE species, respectively. PC class contained mostly diacyl molecular species, 96.0±0.1% of total PC. The species containing arachidonic acid (36:4; 38:4 and 38:5) were predominant. The major arachidonoyl-PI species, 38:4, was accountable for 47.4±0.3 and 46.4±0.5 % of total PI in H109 and fPDR747W cells, respectively. PUFA-containing PS species were mainly represented by species with arachidonic (38:4) and adrenic acid (40:4).</p><p>Redox phospholipidomics revealed increased levels of a variety of mono-, di- and tri-oxygenated species of PE, PC, PS and PI species in fPDR747W vs H109 cells triggered to undergo ferroptosis by a GPX4 inhibitor, RSL3 (Fig. 3b,c and Extended Data Fig. 8c). An unbiased OPLS-DA analysis established that oxidized PE (PEox) species were the predominant oxidized phospholipids accumulating in WT H109 and fPDR747W mutant cells upon RSL3 treatment (Fig. 3d,e). Notably, RSL3 treated fPDR747W cells contained greater amounts of oxidized PE species (Extended Data Fig. 8c) and exhibited significantly higher levels of pro-ferroptotic 15-HpETE-PE both basally and after RSL3 treatment (Fig. 3c). 15-HpETE-PE level was 2.4-fold higher after RSL3 treatment in fPDR747W cells than in WT cells (Fig. 3c). Two of the major PC-ox species were lower in fPDR747W cells compared to H109 cells (Fig. 3d,e).</p><p>We asked if the human PD-associated R747W mutation in PNPLA9 increases the sensitivity of fPDR747W cells to pro-ferroptotic stimulation. The time-course and concentration-dependence of cell death triggered by RSL3 (Fig. 4a) showed a higher sensitivity to ferroptosis of fPDR747Wcells, both in terms of the lower RSL3 concentration required for the ferroptotic response and its earlier onset in the fPDR747Wcells. The specificity of RSL3-induced ferroptotic death was confirmed by negative responses to inhibitors of alternative death programs – apoptosis (z-VAD-fmk) and necroptosis (necrostatin-1s) (Fig. 4b), whereas four inhibitors of ferroptosis (Fer-1, DFO, Vitamin E, Baicalein) inhibited RSL3-induced cell death.</p><p>To explore whether the anti-ferroptotic potential of iPLA2β is realized in other cell types, we compared the sensitivity to ferroptosis in WT and several iPLA2β KD cells: H109 fibroblasts, SH-SY5Y neuronal cells and BeWo trophoblasts. A markedly enhanced sensitivity of iPLA2β-deficient cells to RSL3 induced ferroptosis vs WT controls was found (Fig. 4c and Extended Data Fig. 8d,e). Moreover, iPLA2β-deficient cells exhibited higher levels of pro-ferroptotic PE biomarkers detected by LC-MS (Extended Data Fig. 8f).</p><p>To demonstrate the role of remodeling of membrane phospholipids we knocked-down LPCAT3 in WT H109 cells (human fibroblasts) and mouse embryonic fibroblasts (MEF) and characterized the deacylation/reacylation processes by LC-MS. The levels of LPCAT3 expression assessed by western blotting decreased 2- and 1.5-fold in H109 and MEF cells, respectively (Fig. 4d and Extended Data Fig. 9a). This was accompanied by the increased resistance to ferroptosis by 50% and 60% (Fig. 4e and Extended Data Fig. 9b). Expectedly, LPCAT3 KD resulted in increased levels of the hydrolyzed arachidonoyl-PL metabolites, lyso-PE (1-SA-2-OH-PE) and lyso-PC (1-SA-2-OH-PC) (Fig. 4f and Extended Data Fig. 9c). In both cell lines the levels of the major pro-ferroptotic signal, peroxidized ETE-PE (1-SA-2-HpETE-PE) was increased upon exposure to RSL3; its content in LPCAT3 KD cells treated with RSL3 was significantly lower (Fig. 4g and Extended Data Fig. 9d). Human and mouse fibroblasts generated oxygenated arachidonoyl-PC in response to RSL3 treatment. Similarly, in RSL3-treated LPCAT3 KD cells, the level of 1-SA-2-HpETE-PC was lower vs RSL3 treated controls (Fig. 4g and Extended Data Fig. 9d). Analogous results were obtained for adrenoyl-PE and adrenoyl-PC species (Supplementary Fig. 2). RSL3 induced accumulation of PL-OOH, disorganized the membrane and facilitated the hydrolysis of both oxidatively modified PLs and non-oxidized PLs. Consequently, the amounts of lyso-PLs were higher than that of PL-OOH. These results support the LPCAT3 involvement in ferroptosis regulation via remodeling of membrane phospholipids (Lands cycle) and maintaining high AA-PE levels required for the generation of pro-ferroptotic death signals.</p><!><p>Assuming that human PNPLA9R747W mutation and the compromised control of pro-ferroptotic PE-derived signals may cause enhanced death of dopaminergic neurons, we employed by CRISPR/Cas genome editing, and created a mouse with an R748W point mutation at the mouse Pnpla9 locus, corresponding to R747W of human PNPLA9 (Supplementary Fig. 3. and 4a–c). In the homozygous Pnpla9R748W/R748W mice, we observed motor impairments (Fig. 5a–c and Supplementary Fig. 4d,e) documented using the pole test (Fig. 5a and Movie 4) and rotarod test (Fig. 5b and Movie 5) which started at 3–4 months. Quantitative catwalk tests indicated that Pnpla9R748W/R748W mice had reduced average walking speed, increased walking speed variation, and disrupted walking cycle, when compared with WT and Pnpla9WT/R748W mice (Fig. 5c). Motor impairments were not detected in heterozygous Pnpla9WT/R748W mice, confirming the human data that only homozygous mutants develop behavioral deficiencies related to PD pathogenesis21, which is consistent with its role in causing autosomal recessive PD.</p><p>In homozygous Pnpla9R748W/R748W mutant mice, the midbrain levels of tyrosine hydroxylase (TH), a marker for dopaminergic neurons, were significantly decreased (Fig. 5d), indicating a more than 40% reduction of dopaminergic neurons. Consistently, dopamine (DA) - the product of TH - and its primary metabolites 3,4-dihydroxyphenylacetaldehyde (DOPAL) and 3-methoxytyramine (3-MT) in the striatum of mutant mice were also significantly decreased (Fig. 5e). We also observed elevated levels of a chartacteristic peroxidation product, 4-hydroxynonenal, 4-HNE (Fig. 5f), and lower contents of intracellular antioxidant glutathione, GSH (Fig. 5g) in the mutant mouse midbrain. Monomeric α-synuclein and α-synuclein aggregates did not accumulate in the mutant mice (Supplementary Fig. 4f,g).</p><!><p>Redox phospholipidomics of midbrain samples revealed elevated levels of 15-HpETE-PE in Pnpla9R748W/R748W vs. WT mice (Fig. 5h). Assessment of the specific hydrolytic competency towards 15-HpETE-PE showed that the activity was sharply decreased (more than 75%) in the brain homogenates of PNPLA9 mutant vs WT mice (Fig. 5i and Supplementary Fig. 4h). Western blotting showed no differences in the iPLA2β protein contents between the two groups of mice (Fig. 5j).</p><p>PNPLA9 mutations have been linked to the pathogenesis of several neurodegenerative diseases19, including PD. To comparatively evaluate the effect of the R747W mutation on iPLA2β function and the pathogenicity potential vs all possible mutations we performed in silico saturation mutagenesis analysis using a machine learning tool Rhapsody22 (Supplementary Fig. 5). This analysis confirmed a highly deleterious nature and high pathogenicity score of the R747W.</p><!><p>By using phospho-lipidomics analysis in another PD model, rotenone infused rats, we identified 74 species of PE, 67 species of PC, 22 species of PI and 31 species of PS (Fig. 6a, Extended Data Fig. 10a and Supplementary Fig. 6a). The level of the PE plasmalogens was 1.9 times higher than that of diacyl-PEs. The contents of PE with arachidonic, adrenic and docosahexaenoic residues were 4.5±0.5, 7.1±0.3, 8.9±1.4% (of total PE) for diacyl species and 7.2±0.7, 1.0±0.1 and 11.4±1.2% for alkenyl PE species. In the PC class, diacyl-species were the most abundant and their content was 15.2 times higher vs. alkenyl-PC. The contents of arachidonoyl- and docosahexaenoyl species were 7.1±0.5 and 6.8±0.3 vs. 0.13±0.01 and 0.03±0.01 (% of total PC) for diacyl- and alkenyl PCs, respectively. The PS class included high amounts of arachidonic (38:4) and docosahexaenoic acid (40:6) species: 10.0±0.8 and 24.7±3.3 % (of total PS). Among PIs, the arachidonoyl species (38:4) was predominant - 62.7±1.6 %. No significant differences in the molecular speciation of major phospholipids and their relative contents were found between WT and Parkinsonian rats. Redox lipidomics revealed significantly elevated levels of pro-ferroptotic 15-HpETE-PE and 17-HpDTE-PE in midbrain tissue on days 10–14 after rotenone infusion (Fig. 6b,c and Extended Data Fig. 10b), when the characteristic manifestations of PD-related syndrome are clearly detectable23. Furthermore, the contents of PE species containing oxo-arachidonoyl-residues (1-SA-2-oxo-ETE-PE, 1-SA-2-oxo-hydroperoxy-ETE-PE as well as 1-SA-2-hydroxy-ETE-PE) were significantly increased in Parkinsonian rats vs control rats (Supplementary Fig. 6b). Direct assessments of the 15-HpETE-PE-hydrolyzing activity revealed significantly lower rates in rotenone-treated brain homogenates vs non-treated controls (Fig. 6d).</p><!><p>The SNCA-A53T mutation is associated with autosomal dominant PD and elicits a PD disease-relevant phenotype in rodents14. A53T αSyn can indirectly – via suppression of mitogen-activated protein kinase (MAPK) signaling – negatively regulate the expression of PNPLA924. Indeed, western blotting of iPLA2β revealed significantly decreased expression levels of the protein in the midbrain of 8-month-old SNCA-A53T mutant mice vs WT mice (Extended Data Fig. 10c). In transgenic mice overexpressing human α-synuclein (αSyn) with A53T mutation, redox lipidomics revealed higher levels of oxygenated PEs in midbrains (Extended Data Fig.10d) including the species that were identified as ferroptotic cell death signals (Extended Data Fig. 10e). PUFA-PE containing archidonic and docosahexaenoic acid were predominant species. No differences in the molecular speciation and contents of PE between midbrains of WT and A53T mice (Extended Data Fig. 10f) were detected.</p><!><p>Necro-inflammatory consequences of ferroptosis and its possible pathogenic role in major acute and chronic degenerative diseases stimulated studies of its regulation. Identification of enzymatic lipid regulating mechanisms of ferroptosis, - ACSL4, LPCAT325, along with 15-HpETE-PE generation by 15LOX/PEBP1 complexes9 - was followed by the discoveries of new regulatory cascades including FSP126, 27 and iNOS6. Here we demonstrate that iPLA2β-can act as an anti-ferroptotic guardian via elimination of the pro-ferroptotic signal, 15-HpETE-PE. The peroxidized acyl chain of 15-HpETE-PE is exposed to the membrane surface (the whisker model16), thus enabling interfacial interaction with membrane-bound catalytic domains of the dimeric iPLA2β. This alternative 15-HpETE-PE neutralizing mechanism may act in coordination with the reductive GPX4-dependent pathway or independently of it, serving as an additional check-point for preventing unnecessary or excessive ferroptotic death. The role of this pathway may be particularly important when the thiol-driven defenses become insufficient. Consequently, failure or deficiency in iPLA2β - caused by genetic factors or chemical/pharmacological poisoning - may be associated with increased sensitivity to ferroptotic death.</p><p>Replenishment of oxidatively modified phospholipids may occur via the Kennedy and the Lands pathways whereby the latter involves, as the first step, deacylation to produce lysophospholipids followed by their reacylation28. The deacylation step is the function of PLA2, and the reacylation is mediated by lysophospholipid acyltransferase. In ferroptosis, of particular importance is the elimination of the unstable primary molecular products of lipid peroxidiation, PL-OOH. This can be achieved through either direct reduction of hydroperoxy-group by GPX4/GSH system or hydrolysis of HOO-PUFA-phospholipids and subsequent reduction of the HOO-PUFA (eg, by non-heme peroxidases). As pro-ferroptotic peroxidation specifically targets arachidonoyl-containing PE species, this defines the specificity of ferroptotic phospholipid-metabolizing machinery. Both acyl-CoA synthetase long-chain family member 4 (ACSL4) and LPCAT3 display selectivity towards arachidonoyl-specific reactions25. Vectorial uptake of exogenous fatty acids by fatty acid transport proteins (FATPs) coupled with their activation to CoA-thioesters can participate in the metabolism of peroxidized phospholipids. FATP2, one of the six members of the FATP family, displays specificity for arachidonic (C20:4) and adrenic (C22:4) fatty acids29 pointing to its potential role in ferroptosis</p><p>Ca2+-independent PLA2s (iPLA2, group VI) are involved in the remodeling of membranes facilitated by the hydrolysis of oxidatively-modified PUFA-phospholipids followed by their re-acylation. The iPLA2β preference towards hydrolyzing arachidonoyl-substrates30 supports its anti-ferroptotic role in eliminating peroxidized arachidonoyl-PE species. Several studies in liposomes, isolated cell membranes and tissue homogenates support this role of iPLA2 in the "repair" of peroxidized phospholipids10, 31. As a member of the iPLA2 family, iPLA2β protects different cells from oxidative injury. Overexpression of iPLA2β in INS-1 insulinoma cells enhanced repair of oxidized mitochondrial cardiolipins (CL) and protected against mitochondria-driven prooxidant cell death32. In the retina, an inhibitor of iPLA2, bromophenol lactone (BEL), stimulated the accumulation of Fe-induced lipid peroxidation products33. Elevated Fe levels, oxidative stress and lipid peroxidation have been implicated in retinal damage in age-related macular degeneration (AMD)34. Given the selectivity of iPLA2β towards sn-2-arachidonoyl phospholipids, including PE30, one can assume that the enzyme deficiency may be a part of the pro-ferroptotic mechanism in the retina relevant to the AMD pathogenesis.</p><p>Metabolic features of dopaminergic neurons with their high abundance of iron required for dopamine biosynthesis and high levels of redox-active intermediates of dopamine oxidation create pro-oxidant microenvironments11. Among the most common manifestations of this are GSH deficiency, suppressed activity of GPX412, and elevated levels of lipid peroxidation products35 - all typical characteristics of cells primed to ferroptotic death2. Not surprisingly, the possible engagement of ferroptotic death in PD pathogenesis has been widely discussed, yet specific mechanisms for triggering ferroptosis in PD dopaminergic neurons remained enigmatic11. The present study demonstrates that a typical PD-associated mutation in iPLA2β results in a specific loss of catalytic activity towards one of the peroxidized PLs, the ferroptotic 15-HpETE-PE, in several cell culture and in vivo models. Interestingly, earlier work found no changes in the catalytic activity of iPLA2β mutants associated with idiopathic PD36. However these assessments did not examine oxygenated phospholipids, particularly 15-HpETE-PE, as substraters of the hydrolyitic activity.</p><p>Our work demonstrates that, in addition to direct effects on iPLA2β activity/expression, indirect regulatory factors altering iPLA2β levels in critical cells/tissues may affect the ferroptotic signaling via 15HpETE-PE. We found that iPLA2β expression was significantly reduced in a genetic model of rodent PD overexpressing the mutant A53T αSyn. Both WT and A53T αSyn suppressed MAPK signaling that regulates PNPLA9 expression24. This could be attributed to the direct interaction of αSyn with MAPKs such as ERK2 and its substrate ELK1, a transcription factor that binds the promoter of Pnpla937. We demonstrated that the decreased expression of iPLA2β leads to a dampened pro-ferroptotic arachidonoyl-PE metabolism. Interestingly, a recent study in Drosophila showed that the loss of iPLA2β leads to the shortening of phospholipid acyl chains, facilitating αSyn aggregation38. Thus, a vicious cycle involving αSyn overexpression and iPLA2β underexpression may be warranted.</p><p>Several Ca2+-independent phospholipases that belong to Group VI PLA(2) exhibit triacylglycerol (TAG) hydrolase activity39. In the context of A53T mutation in SNCA, lower levels of iPLA2β activity may be associated with the TAG accumulation in dopaminergic neurons. An increased fatty acid synthase expression and activation of Acyl-CoA synthetase40 in SNCA A53Tmutants may lead to higher levels of free fatty acids, production of their CoA-derivatives and activation of esterification leading to the accumulation of TAGs. While TAG accumulation has been reported in ferroptosis41, mechanistically their role in the execution of this death program has not been clearly defined. It can be speculated that the hydrolytic activity of iPLA2β towards TAGs may be viewed as a non-productive iPLA2β distraction from its participation in hydrolysis of pro-ferroptotic peroxidized PE species, leading to an indirect enhancement of ferroptosis.</p><p>Taken together, our experimental results and the in silico studies strongly suggest that ineffective control/destruction of pro-ferroptotic signals as a consequence of PD-related direct (PNPLA9), or indirect mutations leading to the reduced enzymatic hydrolytic activity of iPLA2β following pesticide or other environmental exposures, may contribute to PD pathogenesis.</p><!><p>1-Octadecanoyl-2-(5Z,8Z,11Z,14Z-eicosatetraenoyl)-sn-glycero-3-phosphatidylethanol-amine (1-SA-2-ETE-PE) was from Avanti Polar Lipids. 5Z,8Z,11Z,14Z-Eicosatetraenoic acid (ETE) and 15(S)-hydroxy-5Z,8Z,11Z,13E-eicosatetraenoic acid (15-HETE) was from Cayman Chemicals. Initially 15(S)-hydroperoxy-5Z,8Z,11Z,13E-eicosatetraenoic acid (15-HpETE) and 15(S)-hydroxy-5Z,8Z,11Z,13E-eicosatetraenoic acid (15-HETE) were synthetized and purified in our lab and later both 1-SA-2-15-HpETE-PE and 1-SA-2-15-HETE-PE became commercially available and were purchased from Cayman Chemicals. Unless otherwise stated, all other reagents were HPLC grade and purchased from ThermoFisher Scientific.</p><!><p>The experiments were approved by the Institutional Animal Care and Use Committee at the University of Pittsburgh (rat rotenone model) and the Laboratory Animal Ethics Committee at Jinan University (A53T overexpression and PNPLA9R748W mouse models). Animal care and handling were in accord with National Institutes of Health guidelines.</p><!><p>Rats were randomized in each study, and in all cases the surgeon and researchers evaluating the outcomes were blinded to the treatment group. Adult (7–9 months old, Charles River) male Lewis rats were subjected to a rotenone-induced model of PD as previously described42. Briefly, rats were injected intraperitoneally with vehicle or 3.0 mg/kg/day of rotenone (Sigma-Aldrich) either for one injection, five daily injections, or treated to Parkinsonian endpoint. Animals treated with rotenone to Parkinsonian endpoint (10–14 days) were sacrificed when animals displayed behavioral features including bradykinesia, postural instability/gait disturbances, and rigidity. Rat brains were removed from the skull, rinsed in cold 1×phosphate-buffered saline, placed on a cold Petri dish and cut in half into the right and left hemisphere. Using a blade and forceps precise micro-dissection of the ventral midbrain was performed and the tissue was flash frozen in liquid nitrogen and stored at 80°C.</p><!><p>Hualpha-Syn (A53T) transgenic C57BL/6J mice were from The Jackson Laboratory (Stock No: 006823) and bred. Animals were sacrificed at eight-months of age.</p><!><p>CRISPR/Cas-mediated genome engineering was used to create a C57BL/6N mouse model with point mutation (R748W) at Pnpla9 locus. Briefly, the gRNAs to mouse Pnpla9 gene, the donor oligo containing R748W (CGG to TGG) mutation, and Cas9 were co-injected into fertilized mouse eggs to generate targeted knockin offspring. Two silent mutations (GCC to GCG at A749) and (GGC to GGG at G755) were introduced to prevent the binding and re-cutting of the sequence by gRNA after homology-directed repair. F0 founder animals were identified by PCR followed by sequence analysis, which were bred to wild-type mice to test germline transmission and F1 animal generation. The homozygous mutant mice were further generated by inter-cross heterozygous mutant mice. The gRNA target sequences are as follows: gRNA1 (F1, matching reverse strand of gene): GATGCCGACCATCTCGCACCAGG; gRNA2 (R1, matching forward strand of gene): CTGTGGATCGGGCCCGGGCCTGG. Animals were sacrificed at seven-months-old.</p><!><p>H109 and fPDR747W human primary fibroblasts obtained from a healthy and a R747W PNPLA9 dystonia Parkinsonism patients43 and mouse embryonic fibroblasts (ATCC, CRL-2991) were cultured in Dulbecco's modified Eagle's medium (ATCC); SH-SY5Y cells (ATCC, CRL-2266) were cultured in 1:1 mixture of Eagle's Minimum Essential Medium and F-12K Medium and BeWo human trophoblast cells (ATCC, CCL-98) were cultured in F-12K medium. All media were supplemented with 10% fetal bovine serum (Sigma-Aldrich) and 1% penicillin-streptomycin (ThermoFisher Scientific) and cells were at 37°C, 5% CO2, and 95% humidity.</p><!><p>Lentivirus vectors encoding shPNPLA9 (GCTGACGCCCTAGTGAATTTC) were used to knockdown iPLA2β in H109 and SH-SY5Y cells according to the manufacturer's instructions (Integrated Biotech Solutions Co., Ltd.). The pGMLV-SC5 lentiviral vector system containing shPNPLA9 was transformed into DH5α cells and the plasmid DNA was extracted and sequenced. For transfection, the plasmid was diluted with serum-free OMEM (Gibco) and incubated with Lipofectamine 3000 (ThermoFisher Scientific) before adding to cells for 6 hrs. After replacing the media, cells were cultured for 40 hrs. Transfection efficiency was optimized using a range of plasmid and Lipofectamine 3000 concentrations.</p><!><p>BeWo cells that stably express the doxycycline-inducible Cas9 plasmid (Addgene plasmid #50661) were used for knocking out PNPLA9. Guide RNAs were designed using the CRISPR design tool (MIT), and expressed from pLKO5.sgRNA.EFS.tRFP657 (Addgene plasmid #57824) lentiviral system, initially in HEK293 cells, with virus used to infect BeWo cells (Guererro and Y Sadovsky, in submission).</p><!><p>Cells were transfected with a mix of two Dicer substrate siRNAs (DsiRNA) against LPCAT3 (hs.Ri.LPCAT3.13.1; 5' - CUUGGAACUGUAUUAGAUAAAAUCA −3'; 3'- UGGAACCUUGACAUAAUCUAUUUUAGU −5'; and hs.Ri.LPCAT3.13.2; 5'-CCAGUUCUCAAUGAAUCACUACATG-3'; 3'-GGGGUCAAGAGUUACUUAGUGAUGUAC-5') for H109 WT and (mm.Ri.LPCAT3.13.1; 5'- GUCUUGACACUGAAGCUAAUUGGGC-3', 3'-CACAGAACUGUGACUUCGAUUAACCCG-5 and mm.Ri.LPCAT3.13.2; 5'-GUUUCUCUUCUGCCAAUCUACUACG-3'; 3'-AACAAAGAGAAGACG GUUAGAUGAUGC −5') for MEF or with control DsiRNA (51-01-14-04) using Lipofectamine 3000 (Life technology) for 24 hrs, then counted and reseeded for experiments. Both si-NT or si-LPCAT3 cells (H109 and MEF) were treated with RSL3 after 48 hrs of transfection, and incubated for 20 hrs.</p><!><p>H109 and fPDR747W cells were treated with RSL3 (25nM) for 14 hrs in the absence or in the presence of ferrostatin-1 (Fer-1, 0.4μM) or z-VAD-fmk (50μM) or necrostatin-1s (20μM), or deferoxamine (10μM), or vitamin E (10μM), or baicalein (2μM). For SH-SY5Y WT and iPLA2β knockdown cells and BeWo WT and PNPLA9 KO cells, RSL3 (2μM) or (100nM) was used for 18 hrs or 12 hrs with or without Fer-1 (0.4μM). Cell death was detected by LDH release using the CytoTox-ONE™ Cyto-toxicity Detection Kit (Promega).</p><!><p>Short variant CHO iPLA2β was cloned and expressed as described previously15. The short variant lacks a 54 amino acid (396–450) insert between ankyrin repeats and catalytic domain. The numbering is adjusted accordingly, such that R747W in Human Long iPLA2β is R693W in the CHO construct. CHO and human proteins have 90.4% sequence identity. Briefly, the PNPLA9 gene cloned from CHO cells with a C-terminal 6XHisTag was cloned into pFastBac vector. CHO R693W mutant was obtained via Quikchange mutagenesis and was confirmed by complete sequencing (Source BioScience). The CHO iPLA2β protein was expressed in Sf9 cells (Invitrogen) and purified in purification buffer containing 25mM HEPES pH-7.5, 20% glycerol, 0.5M NaCl, 1mM TCEP on TALON cobalt resin (Clontech). iPLA2β and its mutant (R693W) were >95% pure as determined by Coomassie stained SDS-PAGE. Activity of purified proteins was confirmed using fluorescent phospholipase activity assay with Pyrene-PC (ThermoFisher Scientific #H361).</p><!><p>1-SA-2-ETE-PE (10μM) or 1-SA-2-15-HpETE (10μM) were incubated in 50mM PBS pH-7.4, containing 100μM DTPA in the presence of either WT iPLA2β or R747W mutant iPLA2β at 37°C. Both PEs were added as methanol solutions. Reaction was started by the addition of the enzyme (0.12μM). At different time points, lipids were extracted2 and enzymatic activity was assessed by the formation of the hydrolysis products –1-SA-2-OH-PE, ETE and 15-HpETE using LC/MS. To estimate the specificity constants, WT iPLA2β or R747W mutant iPLA2β were incubated in 50mM PBS (pH 7.4) containing DTPA (100μM) in the presence of different concentrations of 1-SA-2-ETE-PE (5–100μM) or 1-SA-2-15-HpETE-PE (5–150μM) or 1-SA-2-15-HETE-PE (50–100μM) for 5 min at 37°C. The reaction was stopped by the addition of chloroform:methanol (2:1, v/v). Hydrolysis product –1-SA-2-OH-PE - was extracted and resolved by LC/MS.</p><!><p>Cells (1×106) were harvested after treatment with trypsin-EDTA (0.25%; Gibco), washed with PBS, re-suspended in 20mM HEPES (pH 7.4) containing protease inhibitor cocktail (dilution 1:100) (ThermoFisher Scientific) and sonicated on ice. Cells and tissue homogenates were centrifuged at 12,000g for 15 min. Cell or tissue supernatants (200 μg protein) were added to 50mM HEPES (pH 7.4) containing 100mM NaCl, 5mM EGTA, 1-SA-2-ETE-PE (7μM) or 1-SA-2-15-HpETE-PE (7μM) and incubated for 30 min at 37°C. For (S)-BEL-treatetment, supernatants were pre-incubated with (S)-BEL (10μM), a chiral-specific non-reversible inhibitor of iPLA2β, for 10 min at 37°C. Reaction was stopped by chloroform:methanol (2:1, v/v), hydrolysis products were extracted and resolved by LC/MS. iPLA2β activity was expressed as pmols of 1-SA-2-OH-PE/min/mg of protein.</p><!><p>Cells and tissues were resuspended in lysis buffer (25mM Tris-HCl, pH-7.5, 150mM NaCl, and 1% SDS) containing protease-phosphatase inhibitor cocktail (ThermoFisher Scientific), sonicated to break down DNA and total protein amount was estimated by BCA protein assay kit (ThermoFisher Scientific). Samples diluted in Laemmli buffer were loaded in 8–16% Tris-glycine gradient gels (Life Technologies), proteins transferred to nitrocellulose or polyvinylidene difluoride (PVDF) membranes (Bio-Rad) and blocked with 5% milk or BSA in PBST (0.1% Tween). Protein expression was detected using anti-iPLA2β (polyclonal, PA5–27945, Thermo Fisher Scientific), anti-tyrosine hydroxylase (ab112, abcam), anti-α-synuclein (SC-7011-R, Santa Cruz), anti-4-HNE (ab46545, abcam), anti-LPCAT3 (ProSci, 16–999; 1:500 dilution), anti-GAPDH (FD0063, Fude Biotech), and anti-β-actin (mouse monoclonal, A3854, clone AC-15, Sigma-Aldrich) antibodies (1% BSA in PBST) overnight at RT, washed 3 times, and incubated with horseradish peroxidase (HRP)-conjugated secondary antibody-goat anti-rabbit IgG (A0545, Sigma-Aldrich, FDR07, Fude Biotech) and goat anti-mouse IgG (FDM07, Fude Biotech) (1 hr) in blocking solution before developing with SuperSignal West Pico Chemiluminescent Substrate (Thermo Fisher Scientific). Bands were visualized using X-ray film or imaged using Amersham Imager 600 (GE Health Care, Life Sciences) and quantified using ImageJ software (NIH) (https://imagej.nih.gov/ij/). The integrated density value was obtained by integrating the entire pixel values in the area of one band after correction for background and then normalized to loading control (actin/GAPDH).</p><!><p>ETE and 15-HpETE were analyzed by LC/MS using an Ultimate 3000 HPLC system coupled on-line to a Q-Exactive Hybrid Quadrupole-Orbitrap mass spectrometer (ThermoFisher Scientific) using a C18 column (Acclaim PepMap RSLC, 100Å, 5 μm, 150 mm×0.3 mm, 35°C, ThermoFisher Scientific). Gradient solvents A (20% methanol) and B (90% methanol) contained 5 mM ammonium acetate. Flow rate was maintained at 12 μl/min. The gradient was as follows: 30% solvent B to 95% solvent B, from 0–70 min, hold at 95%B from 70–80 min, 30%B from 80–83 min, 30%B from 83–90 min. Negative ion mode MS conditions were as follows: resolution, 140,000 for the full MS and data-dependent scan; scan range, m/z 150–600; isolation window 1.0 Da for MS and MS2 scans; capillary spray voltage and temperature 2.6 kV and 250°C, respectively; S-lens - 60. Analytical data were acquired and analyzed using Xcalibur 4.2 Quan Browser (ThermoFisher Scientific).</p><!><p>Lipids were extracted and phosphorus was determined by a micro-method as described previously2. Phospholipids were analyzed by LC/MS using an Ultimate 3000 HPLC system coupled on-line to a Q-Exactive Hybrid Quadrupole-Orbitrap mass spectrometer (Thermo Fisher Scientific) or Orbitrap Fusion Lumos mass spectrometer (ThermoFisher Scientific) using a normal phase column (Luna 3 μm Silica (2) 100Å, 150×1.0 mm, (Phenomenex)) as previously described2. Q-Exactive negative ion mode MS conditions were as previously described2. Fusion Lumos MS conditions: spray voltage, 4 kV; transfer tube temperature, 300 °C; RF-Lens level, 50%. Data were acquired in data-dependant-MS2 and targeted-MS3 mode. For MS: Ion detection, Orbitrap; mass resolution, 120,000; scan range, m/z 400–1800. For MS2: quadrupole isolation 1 Da; HCD, collision energy 24%; ion detection Orbitrap; mass resolution, 15,000. For MS3: isolation window, 2.0 Da; CID, collision energy, 35%; ion detection, ion trap. Analysis of raw LC/MS data was performed using Compound Discoverer™ 2.0 (ThermoFisher Scientific) with an in-house generated analysis workflow and database. Peaks with S/N ratio of more than 3 were identified and searched against oxidized phospholipids database. Lipids were further filtered by retention time and confirmed by MS3 analysis using the fragments utilized for their identification (www.Lipidmaps.org). Deuterated phospholipids (Avanti Polar Lipids) were used as internal standards. Values for m/z were matched within 5 ppm to identify the lipid species.</p><!><p>Total lipids were separated on an Thermo Ultimate 3000 LC system with a C30 reverse phase column (Accucore, 2.1mm×25 cm, 2.6μm, ThermoFisher Scientific). Solvent A: Acetonitrile/water (50/50); Solvent B: 2-propanol/acetonitrile/water (85/10/5) contained 5mM ammonium formate and 0.1% formic acid as modifiers. Gradient was as follows: 0–20 min, 30%−70% B; 20–55 min, 70–100% B; 55–70 min, hold at 100% B; 70–85 min, 100–30% B; 85–95 min, 30% B for equilibration. The flow rate was 100 μl/min and column temperature was 35°C. Analysis of phospholipids was performed on a Q-Exactive mass spectrometer (ThermoFisher Scientific). Negative ion mode MS conditions were as previously described9.</p><!><p>Catecholamines and GSH were determined by LC-MS using a Triple Quad 4500 Mass Spectrometer (SCIEX) (Extended Data Table 3). Analytes were separated on a C18 column (Acquity HSS T3, 1.8 μm, 2.1 mm×100 mm, 40°C, Waters) at a flow rate of 0.4 mL/min on an Exion LC AD system. Gradient: solvent A (water) and solvent B (acetonitrile), each containing 0.1% formic acid (v/v): 0–1.2 min isocratic of 2% B, 1.2–2.5 min linear gradient of 2 to 60% B, 2.5–4.5 min linear gradient of 60% to 95% B, 4.5–5.0 min isocratic of 95% B, and 5.0–6.5 min re-equilibration of 2% B. Positive and negative ion modes switching with a scheduled MRM scan method was used. Capillary spray voltage: 5.5 and −4.5 kV; ion source temperature, 600°C. Nebulizer gas and heater gas was set to 55 psi, and curtain gas was set to 30 abr. Data acquisition and processing were carried out using Analyst 1.6.2 software (SCIEX).</p><!><p>iPLA2β structure (752 residues) complexed with ankyrin fragments was constructed using the available structure (PDB code: 6aun15) and a combination of homology modeling and ab initio simulations. The structures of the short missing fragments in the X-ray structure of iPLA2β dimer (UniProt ID: A0A384E119), i.e. Y95-R115, L129-N145, R405-K408 and V652-A670, were reconstructed using SWISS-MODEL44 server and longer M1-A80 fragment of ANK repeats with I-tasser45 server. Detailed information on the quality of the model (released by SWISS MODEL and I-TASSER servers) has been provided in Extended Data Fig. 6. The full structure atomic coordinates of iPLA2β are shared as supplementary material.</p><!><p>Neuronal membrane containing 15-HpETE-PE (4%), ETE-PE (SAPE, 22%) and various phospholipids (SAPI-4%, SOPC-15%, PSPC-25%, SOPE-5%, OLPS-10%, PAPC-5%, SDPE-10%), the composition of which has been deduced from lipidomics data, has been prepared using CHARMM-GUI46 and in-house scripts. The membrane size was 50×50×50Å. The same composition and spatial distribution have been replicated to construct larger membranes (100×100Å2 and 250×250Å2 surface area and 50Å depth) to avoid end effects. Equilibration simulations were performed for 50×50×50Å membrane using NAMD47 software with CHARMM force field and explicit water model (TIP3P). The simulated system contained over 51,000 atoms, including the lipid and water molecules, and KCl ions. Six preparatory simulations were conducted before the main run following CHARMM-GUI protocols. Two main 100 ns MD simulations were performed where the additional polar oxygen atoms-containing acyl chains of ETE-PE reoriented and partially protruded from the membrane surface into the aqueous solution. Force field parameters for 15-HpETE-PE were created using SwissParam48. The structural model of 15-HpETE-PE was built and minimized using Maestro.</p><!><p>Equilibrated membrane with protruding acyl chain of 15-HpETE-PE was replicated to perform protein-membrane simulations. The dimeric protein orientation (CAT domain (M416-P752) and full structure of iPLA2β (M1-P752)) in the membrane was predicted by PPM server49. The systems contained over 320,000 atoms (100×100×115 Å box, CAT domains) and 750,000 atoms (200×200x×125 Å, iPLA2β dimer). The following protocol was adopted using NAMD package: 0.2 ns of water equilibration, 10,000 steps of minimization, 0.35 ns of heating from 0 to 300K, and 0.15 ns equilibration before initiating the production MD run. Eight trajectories (four for WT and R693W, each), with CAT domain dimer in close proximity to the membrane were generated, each 55 ns long with 2 fs timesteps. Four 20 ns trajectories with the full structure of iPLA2β dimer were also computed. A cutoff of 12Å was adopted for non-bonded interactions. Langevin dynamics and Langevin piston algorithm were used to maintain the temperature at 300K and the pressure at 1 atm. Analysis was performed using VMD and Pro Dy API together with in-house codes50.</p><!><p>The tool Rhapsody22 developed for predicting the functional consequences of single amino acid variants (SAVs) was used for automated scanning of all iPLA2β residue substitutions. The method uses a Random Forest-based classifier trained using an integrated dataset of 20,854 missense mutations functionally characterized to date. For each protein, eight structural, sequence-based, and dynamics-based features are calculated using multiple approaches, requiring iPLA2β structure (PDB id: 6aun) as input. The reported Rhapsody scores represent the probability of causing a deleterious effect on function22.</p><!><p>Pole test was used to evaluate the mouse movement disorder. The instrument consists of an iron stand (height, 60 cm; diameter, 0.8 cm) with a small ball wrapped with gauze at the top. Mice were placed on the top of a small ball and the time required for the mouse to climb down the pole was recorded. The test was measured 3 times/mouse and the maximum time was recorded.</p><!><p>Rotarod performance was used to assess mouse motor balance and coordination. Mice were trained for 3 days prior to treatment to adapt to the rotarod apparatus (Zhenghua Co.). After training, on the 7th day, mice were placed in the separate runway on the rod and at a constant speed of 25 rpm every day at the same time. Each mouse was tested three times. The latency to fall was recorded.</p><!><p>CatWalk Gait Analysis is a system for rodent gait analysis. The apparatus consists of a long glass walking plate, a fluorescent light beamed into the glass plate and a high-speed video camera under the glass plate. In a dark environment, the light was reflected downward and a camera mounted under the glass recorded the footprint of mouse on the walkway. Mice were trained to cross the glass walkway three days prior to the test. The mouse performed unforced and uninterrupted moving at least three times. The data of mouse gait was qualitatively and quantitatively analyzed by the automated gait analysis system CatWalk (Noldus Information Technology).</p><!><p>The results are presented as mean±standard deviation (s.d.) with a minimum of three replicates unless otherwise specified. Statistical analyses were performed by either Student's t-test or one-way/two-way ANOVA for normally distributed data using Prism 8.1 (GraphPad Software, Inc). The significance of differences was set at p<0.05. When the overall ANOVA revealed a significant effect, the data were further analyzed with the Dunnett/Sidak post hoc test to determine specific group differences.</p><!><p>Data generated during the study and included in this article are available from the corresponding authors upon request.</p><!><p>a, Western blot of purified recombinant WT and R693W mutant short variant CHO. The Short variant lacks a 54 amino acid (396–450) insert between ankyrin repeats and catalytic domain. The numbering is adjusted accordingly, such that R747W in Human Long iPLA2β is R693W in the CHO construct. Representative figure from 3 experiments b, Time-courses of ETE (upper panel) and 15-HpETE (lower panel) formation in reactions catalyzed by mutant R747W (red circles) or WT iPLA2β (black circles). 1-SA-2-ETE-PE and 1-SA-2-15-HpETE-PE were used as substrates. Data are means±s.d., N=3 independent experiments, *p=0.0381, ****p<0.0001 for 1-SA-2-ETE-PE data and *p = 0.0189, **p = 0.0044, ***P = 0.0003 for 1-SA-2-15-HpETE-PE data, WT iPLA2β vs iPLA2βR747W at respective time points, two-way ANOVA (Sidak's post-hoc test). We employed highly purified (>99%) substrates (1-SA-2-ETE-PE and 1-SA-2-15-HpETE-PE) thus free ETE and free 15-HpETE were undetectable in the samples with no enzyme added. c, Detection and identification of 15-HpETE, a hydrolysis product of 1-SA-2-15-HpETE-PE. Base peak chromatogram of molecular ions with m/z 335.2224 corresponding to 1-SA-2-15-HpETE (left insert). MS2 fragmentation pattern of molecular ions with m/z 335.2224 corresponding to 15-HpETE (right panel). Respective structure and fragments formed during MS2 analysis (right insert). d, Effect of substrate concentration on the velocity of iPLA2β WT (left panel) and iPLA2bR474W (right panel) catalyzed reaction. 1-SA-2-ETE-PE (blue circles), 1-SA-2-15-HETE-PE (light red circles) and 1-SA-2-15-HpETE-PE (dark red circles) were used as substrates. Data are presented as 1-SA-2-OH-PE mM/min, Data are means ± s.d, N=4 independent experiments for 1-SA-2-15HpETE-PE, 1-SA-2-15HETE-PE data and 3 for 1-SA-2-ETE-PE data.</p><!><p>a, Structural formulas of phospholipids used in simulations. b, Spatial distribution in the lipid bilayer. The lipid composition has been deduced from our lipidomics data (see Methods) Here, SAPE is ETE-PE, and SOOH is 15-HpETE-PE. ETE-PE and 15-HpETE-PE were included at the levels of 22% and 4%, respectively. See legend at the bottom for the complete composition. The same composition and spatial distribution have been replicated to construct a larger membrane (100 × 100 Å2 and 250 × 250 Å2 surface area and 50 Å depth/thickness) in a simulation box with a height of 125 Å enclosing the dimeric enzyme bound to the membrane and explicit water molecules.</p><!><p>a, Hydroperoxy-group in 15-HpETE-PE comes into close proximity of the membrane surface. Conformations reached in 100 ns MD simulations are shown for two 15-HpETE-PE molecules (left and middle) and one for ETE-PE (right). The distances between terminal amino group N-atom and the peroxidized C15-atom of the sn-2-acyl chain in 15-HpETE-PE and in ETE-PE (using the non-peroxidized equivalent C15-atom) are shown. A total of eight simulations were carried out, four with WT and four with mutant R747W iPLA2β. This distance was < 5 Å in at least two 15-HpETE-PE out of the total of 20 per MD snapshot. No such a short distance was observed for ETE-PE. Results for the two sets of four runs were very similar, showing that the behavior of the fatty acid chains is intrinsic to the oxidized fatty acids, irrespective of the iPLA2β. Similar behavior was also observed in the simulations of the membrane alone (in the absence of enzyme). The results in panel (a) are in qualitative agreement with the Nuclear Overhauser Effects (NOE) data obtained by Greenberg et al., 2007. The authors observed signals for irradiated –N(CH3)3 protons of the choline group and the terminal aldehydic group of the truncated sn2-oxidized fatty acid chain in oxidized lipids, consistent with 'whisker' model. However, no signals were observed for non-oxidized lipids, as these were embedded within the hydrophobic bilayer. Observation of a signal in NOE experiments, corresponds to a distance of < 5 Å. b, Probability distribution of the position of the peroxidized/non-peroxidized carbon (C15, sn-2) for 15-HpETE-PE and ETE-PE along the z-axis of the lipid bilayer. The interface between the two lipid monolayers (red dotted line in the inset) serves as a reference point (z = 0, shown in the inset) for the distances of C15 carbon atom from the center of the bilayer. Six random lipids for both, 15-HpETE-PE and ETE-PE, were chosen for the analysis. Carbon index (C15) refers to that shown on the chemical structure in Supplementary Fig. 5a. Shaded boxes denote the approximate range of distances at which maxima occur in the histograms. The C15 ETE-PE (orange box) remains embedded at ~ 5–6 Å, the C15 of 15-HpETE-PE moves closer to the membrane surface at 12–13 Å (blue box). The positions of the head groups at the membrane surface are indicated by grey box. The probability distributions show that the C15 atom ("peroxidized") of 1-SA-2-15-HpETE-PE adopts two distinct positions, one close to the membrane surface (blue box) the other buried near the central part of the lipid bilayer (z=0).</p><!><p>a, Probability distribution of the distance between the peroxidized carbon (C15, sn-2) of 1-SA-2-15-HpETE-PE or non-peroxidized carbon (C15, sn-2) of ETE-PE and: (i) protein surface (left panel) and (ii) catalytic site (right panel). For protein surface the closest atom of the lipid was taken into account whereas the reference point for catalytic site was computed based on the closest atom of the lipid to the center of mass of the highly conserved catalytic residue S465 and D598. Carbon index (C15) refers to that shown on the chemical structure in Supplementary Fig. 5a. Arrows denote the maxima in the histograms for C15 (which contains the OOH group, shown in panel b) in 1-SA-2-15-HpETE-PE (blue) and C15 in ETE-PE (orange). The analysis contains results from the second half of three MD trajectories for a WT iPLA2β dimer structure (50–100 ns period of time). b, A snapshot from MD simulations of iPLA2β dimer CAT domains with residues making close contacts with 1-SA-2-15-HpETE-PE and 1-SA-2-ETE-PE highlighted in space filling representation (green). The 1-SA-2-15-HpETE-PE (red-blue-cyan balls) and ETE-PE (grey surface) in the membrane are shown.</p><!><p>a, Chemical structures. Carbon atom indices are indicated in green for sn-2 and orange for sn-1. Pink boxes highlight the sn-2 chain which is different in the two lipids. Blue arrow points to the peroxidized carbon C15 in 1-SA-2-15-HpETE-PE. b, Order parameters computed from three sets of independent MD runs. Computationally predicted deuterium order parameters (SCD) for sn-1 (upper curves) and sn-2 (lower curves) chains, based on 3 MD runs performed for ETE-PE (thick lines) and three for 1-SA-2-15-HpETE-PE (thin lines). Blue arrow indicates position of C15. In general, the order parameter S = 3/2 <cos2α> - 1/2 varies in the range [−0.5, 1]; the two limits corresponding to complete order (parallel alignment of the probed bond with respect to the magnetic field, with the angular difference being α= 0) and antiparallel orientation (α = 90°). SCD = 0 for fully disordered states (<cos2α> = 1/3). The fully disordered state SCD = 0 is shown by the dotted line. Red arrow shows the position of the oxidized carbon in 1-SA-2-15-HpETE-PE, where a slight increase in order is induced upon oxidation.). Both chains display low order parameters, with the sn-2 chain being more disordered in general than sn-1, except for the terminal C-atom of sn-1 reflecting higher flexibility at the chain terminals. The results are shown for an ensemble of chains in each case, which exhibit highly reproducible patterns, also consistent with previous computational and experimental data. Note that the portion of 1-SA-2-15-HpETE-PE sn-2 near the peroxidation site (C15) exhibit relatively higher ordering, whereas the remaining portions show a mixed behavior. While these differences are small, they are reproducible in independent runs, lending support to the robustness of simulation data. The computed order parameters values are in a good agreement with previously, reported computational and experimental values. Furthermore, while the published results on the effects of peroxidation on lipids dynamics are somewhat contradictory, our computational observations of the peroxidized sn-2 acyl chains for 1-SA-2-15HpETE-PE are in general agreement with those computed for other peroxidized lipids using the same tools.</p><!><p>a, Swiss model results for homology modeling which includes global and local quality estimate values, sequence identity and coverage compared to the protein template (PDB code: 6aun) and sequence alignment. Red boxes on the sequence alignment denote regions which was not solved in X-ray structure (6aun) and was modeled using Swiss model server i.e. Y95-R115, L129-N146, R405-K408, V652-A670. Black box highlight region M1-A80 which was not present in the ANK repeats fragment. b, I-tasser results for homolog modeling of M1-A300 fragment of iPLA2β structure. Cyan ribbon diagram denote iPLA2β dimer solved in X-ray (6aun), red elements of the structures were modeled using Swiss Model server (shown in the panel a, red boxes). M1-A300 fragment of ANK repeat is shown in yellow ribbon diagram and alignment on the crystal structure. Estimated accuracy obtained by I-tasser server for M1-A300 model is also shown.</p><!><p>a) Results from 4 × 2 runs (labeled MD1–4) conducted for the WT and R747W mutants are displayed. Residues making contacts are listed along the ordinate, and the time evolution of contacts (atom-atom interactions closer than 3.5 A with any 15-HpETE-PE atom) is shown in in each case. Colored regions indicate the contacts made by chain A (cyan) and B (dark red). Note that most of the contacts are persistent once formed. b) Histograms of contacts. Residues making the largest number of contacts (counts based on snapshots collected every 50 picoseconds, summed over all runs) are listed, along with the corresponding counts for chains A (top) and B (bottom).</p><!><p>a, Total PLA2 activity in H109 and fPDR747W cells in the absence and in the presence of (S)-BEL. Cell supernatants were incubated with 1-SA-2-ETE-PE (upper panel) or 1-SA-2-15-HpETE-PE (lower panel) for 30 min at 37°C. Activity is presented as 1-SA-2-OH-PE, pmol/min/mg protein. The background levels of 1-SA-2-OH-PE in H109 and fPDR747W cell supernatants were low and estimated as 2.76±0.19 and 2.74±0.75 pmols per sample vs 70.5 ± 8.5 and 100.7 ± 10.5 pmols per sample accumulated in H109 and fPDR747W cell supernatants during incubation in the absence of S-BEL. Data are means ± s.d., *p=0.0014 for H109 cells, **p= 0.0073 for fPDR743w cells, ****p<0.0001, N=3 biologically independent experiments, one-way ANOVA (Tukey post hoc test). b, Heat map showing the content of phosphatidylethanolamine (PE), phosphatidylcholine (PC), phosphatidylserine (PS) and phosphatidylinositol (PI) molecular species in H109 and fPDR747W cells. dPE-diacyl species, pPE-plasmalogens, dPC-diacyl PC species, pPC-plasmalogen species of PC. Data presented as pmol/nmol of total phospholipids, N=3 biologically independent experiments. c, RSL3-induced accumulation of oxygenated PE (PEox, left), PC (PCox, upper right), PS (PSox, middle right) and PI (PIox, lower right) molecular species in H109 and fPDR747W fibroblasts. Cells were exposed to RSL3 (25nM) for 14 hrs. Data presented as pmol/mmol of total phospholipids, N=3 biologically independent experiments. d, RSL3 - induced ferroptosis in WT and iPLA2β KD SHSY5Y cells. Cell were treated with RSL3 (2 μM) for 18 hrs in the absence or in the presence of Fer-1 (0.4 μM). Inset: representative blot of iPLA2β. Data are means±s.d., ****p<0.0001 vs WT, N=3 biologically independent experiments, two-way ANOVA (Sidak post-hoc test). e, RSL3-induced ferroptosis in WT and PNPLA9 KO BeWo cells. Cells were incubated with RSL3 (100 nM) for 12 hrs in the absence or in the presence of Fer-1 (0.4 μM). Inset: Typical western blot of iPLA2β obtained from BeWo WT and PNPLA9KO cells. Ferroptosis quantified by LDH release. Data are means±s.d., ****p<0.0001 vs WT, N=3 biologically independent experiments, two-way ANOVA (Sidak post-hoc test). f, Content of 1-SA-2-15-HpETE-PE in WT and PNPLA9 KO BeWo cells. Data are means ± s.d., ***p=0.0002, ****p<0.0001 vs WT, N=3 biologically independent experiments, two-way ANOVA (Sidak post-hoc test).</p><!><p>a, Representative immunoblots and quantification of LPCAT3 in cell treated with non-targeted siRNA (si-NT) or LPCAT3 siRNA (si-LPCAT3). LPCAT3 levels were quantified from three biological replicates and normalized to actin. Data represent mean ± s.d., *p=0.0004 vs si-NT, unpaired two-tailed t-test. b, si-NT or LPCAT3 KD cells were exposed to RSL3 (100 nM) and cell death was monitored after 20 hrs by PI staining using flow cytometry. Data are mean ± s.d., N=3 biologically independent experiments; ****p<0.0001 vs si-NT control, ##p=0.0081 vs.si-LPCAT3 control, $$p=0.0078 vs si-NT/RSL3, one-way ANOVA. c, Quantitative LC/MS-based assessments of lyso-PE (1-SA-2-OH-PE, left) and lyso-PC (1-SA-2OH-PC, right) in MEF cells. Data are mean ± s.d., N=3 biologically independent experiments, ***p=0.0008, ****p<0.0001 vs si-NT, unpaired two-tailed student's t-test. d, The contents of oxygenated PE (1-SA-2-HpETE-PE, left) and PC (1-SA-2-15-HpETE-PC, right) in MEF cells. Cells were exposed to RSL3 (100nM) for 20 hrs. Data are mean ± s.d., N=3 biologically independent experiments, *p=0.0282, ****p<0.0001 vs si-NT control, one-way ANOVA, (Tukey's post-hoc test).</p><!><p>a, Content of PE in substantia nigra of control rats (treated with vehicle, DMSO) and Parkinsonian rats (treated with rotenone for 14 days at a dose of 3mg/kg/day). Data are presented as pmol/nmol of total phospholipids, N=6 biologically independent animals. d-diacyl species; p-alkenyl (plasmalogen) species. b, Quantification of oxygenated PE species in substantia nigra of control rats (treated with vehicle, DMSO) and Parkinsonian rats (treated with rotenone for 14 days at the dose of 3mg/kg/day). Data presented as pmol/mmol of total phospholipids, N=6 biologically independent animals. c, iPLA2β protein expression in midbrain of 8-months old WT and A53T mice. Inset: Typical western blot of iPLA2β. Data are means ± SD, ****p <0.0001, N = 5 biologically independent animals, unpaired two-tailed Student's t-test. d, Content of oxygenated PE species in midbrains of WT and A53T mice. Data are presented as pmol/mmol of total phospholipids. e, Content of 15-HpETE-PE in midbrains of WT and A53T mice. Data are presented as pmol of 1-SA-2-15-HpETE-PE per mmol of total phospholipids, **p=0.0010 WT vs A53T mice, unpaired two-tailed Student's t-test. N= 6 biologically independent animals. f, Content of PE in midbrains of WT and A53T mice. Data are presented as pmol/nmol of total phospholipids, N=5 biologically independent animals. d-diacyl species; p-alkenyl (plasmalogen) species</p><!><p>Pole test of representative 7-month-old wild-type and Pnpla9R748W mice.</p><p>HpETE-PE (shown in space-filling representation) observed in MD simulations to undergo rapid conformational changes enabling its lateral diffusion in the membrane and occasional movements toward the surface to expose the peroxidized head. The movie displays a trajectory of 50 ns, with frames recorded every 250 picoseconds). The lipid molecules are color coded as in Supplementary Fig. 2. The additional runs are displayed for HpETE-PE molecules, both resulting (within 50 ns) in the exposure of the peroxidation site of the AA-acyl chain to the membrane surface.</p><p>HpETE-PE (shown in space-filling representation) observed in MD simulations to undergo rapid conformational changes enabling its lateral diffusion in the membrane and occasional movements toward the surface to expose the peroxidized head. The movie displays a trajectory of 50 ns, with frames recorded every 250 picoseconds). The lipid molecules are color coded as in Supplementary Fig. 2. The additional runs are displayed for HpETE-PE molecules, both resulting (within 50 ns) in the exposure of the peroxidation site of the AA-acyl chain to the membrane surface.</p><p>HpETE-PE (shown in space-filling representation) observed in MD simulations to undergo rapid conformational changes enabling its lateral diffusion in the membrane and occasional movements toward the surface to expose the peroxidized head. The movie displays a trajectory of 50 ns, with frames recorded every 250 picoseconds). The lipid molecules are color coded as in Supplementary Fig. 2. The additional runs are displayed for HpETE-PE molecules, both resulting (within 50 ns) in the exposure of the peroxidation site of the AA-acyl chain to the membrane surface.</p><p>Rotarod test of representative 7-month-old wild-type and Pnpla9R748W mice.</p>

PubMed Author Manuscript

A high-throughput mass spectrometric enzyme activity assay enabling the discovery of cytochrome P450 biocatalysts

Assaying for enzymatic activity is a persistent bottleneck in biocatalyst and drug development. Existing high-throughput assays for enzyme activity tend to be applicable only to a narrow range of biochemical transformations, whereas universal enzyme characterization methods usually require chromatography to determine substrate turnover, greatly diminishing throughput. We present an enzyme activity assay which allows for the high-throughput mass-spectrometric detection of enzyme activity in complex matrices without the need for a chromatographic step. We demonstrate that this technology, which we call \xe2\x80\x9cProbing Enzymes with \xe2\x80\x98Click\xe2\x80\x99\xe2\x80\x93Assisted NIMS\xe2\x80\x9d (PECAN), can detect the activity of medically and biocatalytically significant cytochrome P450s in cell lysate, microsomes, and bacterial cells. Using our approach, we successfully screened a cytochrome P450BM3 mutant library for the ability to catalyze the oxidation of the sesquiterpene valencene.

a_high-throughput_mass_spectrometric_enzyme_activity_assay_enabling_the_discovery_of_cytochrome_p450

<p>Enzyme screening campaigns are frequently the bottleneck of drug, biomarker and biocatalyst discovery programs, making high-throughput enzyme activity assays essential to accelerating research in these areas[1–4]. Such screens often rely on fluorogenic or chromogenic enzyme substrate analogs, coupled assays, or biosensors, which – while high-throughput – are unsuitable for monitoring many important enzymatic transformations[5–7].</p><p>Mass spectrometry (MS) is a promising technology for the analysis of enzyme activity, given that the majority of enzymatic reactions change their substrates' masses. However, due to ionization suppression and isobaric ions, MS analysis is limited to low-complexity samples. This poses a significant disadvantage when screening large enzyme libraries for a desired catalytic activity, or evaluating enzymatic activity exhibited by clinical samples, both of which are typically conducted in complex biological matrices. Coupling MS analysis to a chromatographic step, such as in liquid chromatography-MS (LC-MS) or gas chromatography-MS (GC-MS), alleviates ionization and spectral complexity issues by physically separating analytes, but significantly decreases throughput.</p><p>One promising platform for high-throughput enzyme activity determination is Nanostructure-Initiator Mass Spectrometry (NIMS)[8–10]. NIMS involves laser desorption of analytes from a fluorophilic surface. Fortuitously, this surface has high affinity for perfluoroalkylated analytes through non-covalent fluorous interactions, enabling the substitution of lengthy chromatographic separations with an in-situ washing step. However, many enzymes may be unable to accommodate perfluoroalkylated substrates in their active sites, which may be why NIMS-based enzyme assays have thus far been applied only to carbohydrate-acting enzymes and acetyltransferases[8,9,11–13].</p><p>In this work, we carry out an enzymatic reaction on a substrate analog ("probe"), after which we couple the probe to a perfluoroalkylated affinity tag using copper-catalyzed "click" chemistry[14] (Fig. 1). This approach, which we call "Probing Enzymes with 'Click'–Assisted NIMS" (PECAN), requires only the incorporation of a relatively small "clickable" handle into the probe. When combined with acoustic sample deposition and MS Imaging[15], this permits high–throughput characterization of enzyme activity in complex biological matrices.</p><p>We apply PECAN to cytochrome P450s ("P450s") which are an important class of enzymes possessing the remarkable capacity to effect regio– and stereospecific C–H bond activation reactions. P450s play important roles in the biosynthesis of natural products[16] and steroids[17], and are central to human xenobiotic metabolism[18]. Their ability to catalyze reactions that are outside of the reach of traditional synthetic organic chemistry has made P450s valuable biocatalysts. Consequently, much research has been directed towards identifying P450s with the appropriate substrate–, regio– and stereospecificity for industrial biotransformations[19–21]. High-throughput P450 activity assays developed to facilitate these studies include chromogenic[22–24], fluorometric[25], luminogenic[26] and coupled assays[27], as well as MS-based approaches harnessing the high mass resolution of FT-ICR MS to distinguish analytes from matrix ions[28] or solid-phase extraction to enrich for analytes prior to MS analysis[29].</p><p>To benchmark the PECAN technology, we tested its ability to detect the activity of cytochrome P450BM3, a model P450 widely employed in biocatalysis[30]. P450BM3 is known to catalyze the hydroxylation of 1a to form 2a with a 34% e.e. of R-2a[31] (Fig. 2a), We designed an analog of 1a harboring an azide, 1b, to act as PECAN probe (Fig. 2b). We fed 1b (0.5 mM in 1% v/v DMSO) to lysates of E. coli expressing P450BM3 or Green Fluorescent Protein (GFP, serving as a negative control), along with a cofactor regeneration system (10 mM glucose 6-phosphate, 100 µM NADP+ and 0.1 unit/mL G6P dehydrogenase). After a 3-hour enzymatic reaction, the lysates were tagged with a perfluorinated alkyne through a copper(I)-catalyzed "click" reaction, and acoustically transferred onto a NIMS surface. The surface was washed with water and rastered on a MALDI-TOF mass spectrometer. The resulting MS image showed excellent signal-to-noise (Fig. 2c), and P450BM3 reactions could easily be distinguished from negative controls both qualitatively and quantitatively (Fig. 2d). The Z-factor – a commonly-used statistical measure of the quality a high-throughput assay[32] – was calculated to be 0.93, indicating an excellent assay. Similar results were obtained for biological replicates assayed in 96-deepwell microtiter plate format (Fig. S1), suggesting that PECAN can be applied to high-throughput enzyme screening experiments.</p><p>We also investigated whether PECAN could be used to monitor intracellular enzymatic reactions. Probing enzymatic activity in vivo rather than in lysate increases the relevance of screens to downstream whole-cell bioconversion applications, could improve experimental throughput by avoiding the lysis procedure, and decreases assay costs by avoiding the need for exogenous cofactors. Using PECAN, we could successfully monitor the oxidation of 1b fed to whole E. coli cells expressing P450BM3 (Fig. S2).</p><p>Eukaryotic P450s are typically membrane-bound and studied in microsomes (membrane fractions derived from the endoplasmic reticulum), prompting us to test the applicability of PECAN to the measurement of P450 activity in microsomes. We subjected the contraceptive medication 19-norethindrone to recombinant human microsomal P450 CYP3A4 and were able to detect its oxidation products (Fig. S3). Unlike 1b, which contains an azide, 19-norethindrone harbors an alkyne, requiring a reversal of the polarity of the tagging "click" reaction. This did not noticeably affect the efficiency of the tagging reaction or MS signal. CYP3A4 activity could be abolished in the presence of the known inhibitor clotrimazole (Fig. S3), suggesting that that PECAN may be suitable for drug-drug interaction screening.</p><p>As a model high-throughput enzyme screening campaign, we aimed to identify mutants of P450BM3 with the ability to oxidize valencene (3a) to the fragrance and insect repellent nootkatone (5a)[33], a commonly-studied biocatalytic process[34]. In such a process, a P450 hydroxylates 3a, producing either epimer of 4a. The P450 then again oxidizes the same position on 4a to yield 5a (Fig. 3a). While wild-type P450BM3 is unable to effect this transformation, mutants displaying this ability have been discovered in studies evaluating small, focused mutant libraries[35,36]. Although several P450BM3 variants were found to act on 3a, most are either unable to oxidize it beyond 4a, or oxidize 3a at more than one position, yielding non-specifically over-oxidized products (Figs. 3a, S4). Although PECAN cannot distinguish between isomers, it can estimate the extent of probe oxidation from its mass distribution, and thereby help reveal the library's most promising members. While peak heights of different ions in a mass spectrum cannot be directly quantified in the absence of isotopically-labeled internal standards, we expect that the analytes' desorption and ionization in the PECAN assay is largely determined by the tag, which is shared by all analytes. Regardless, the MS signal of an ion proportional to all other tagged ions is expected to rise monotonically with its concentration, allowing us to identify the most productive wells in a screen based on ion counts.</p><p>We performed combinatorial site-saturation mutagenesis (NNK codons) on two P450BM3 amino acid residues commonly mutagenized due to their proximity to the active site heme: F87 and A328[30,37]. While a subset of this library has been assayed for 3a oxidation before[36], no study has further explored its amino acid space, likely due to limited experimental throughput. A lack of library sequence bias was verified by Sanger sequencing 10 randomly-picked colonies. Like others, we observed color differences between the different wells of our P450BM3 library (Fig. S5), which has been attributed to the ability of some variants to oxidize indole to form the dyes indigo and indirubin[38]. Throughout our study, we found no clear correlation between colored wells and the enzyme's ability to oxidize 3a.</p><p>Employing 3b as our PECAN probe analog of 3a, we screened 1208 E. coli cell lysates generated from our P450BM3 library for the ability to oxidize 3b into products with the mass of 5b, in a 96-well format. Hits could be visually identified from the resulting MS image (Fig. 3c). For a quantitative analysis, we used the OpenMSI Arrayed Analysis Toolkit[39] to identify wells displaying m/z = 869 (i.e., 5c) relative ion intensities more than 10 standard deviations above eight GFP controls included on the same 96-well plate. These hits were de-replicated by Sanger sequencing to yield the variants listed in Table 1. The identified amino acid substitutions consist mostly of conservative small and hydrophobic residues. Several variants were recovered more than once. All variants were found to have 3b-oxidation activity in a confirmatory PECAN experiment (Table S1).</p><p>Because the PECAN screen could not distinguish between isomers, and because 4a is merely a surrogate substrate, we used GC-MS to assess the ability of P450BM3 variant hits to oxidize 3a in vitro (Table 1). Purified enzymes were employed to control for differences in protein expression levels. Nearly all screen hits were able to catalyze the conversion of 3a to 5a to some extent, whereas wild-type P450BM3 cannot. P450MB3 F87A, F87G, F87I, F87P and F87G/A238G were found to be hyper-active non-specific 3a oxidizers, which is unsurprising considering the small amino acid residues lining these variants' active sites. Three variants did not produce 5a, of which only one variant (F87V/A238P) was completely inactive on 3a.</p><p>Two of our hits, F87A/A283I and F87A/A283V have previously been identified by Seifert et al, who constructed an "enriched" library of P450BM3 mutants substituting positions F87 and A328 with A, I, L, V and F combinatorially (25 variants total) and tested for the ability to convert 3a to 5a using GC-MS. This verifies that the PECAN technology can identify enzymes with desired catalytic activities that could previously be discovered only using low-throughput approaches. However, despite our ability to screen a significantly more comprehensive mutant library using the PECAN technology, we still identified F87A/A328I as the optimal nootkatone-producing P450BM3 variant, suggesting that the "enriched library" strategy employed by Seifert et al. was indeed effective. Other amino acid substitutions yielding variants with substantial nootkatone production are S, T, Y, N and G. Glycine, in particular, appears in many highly-active variants, suggesting that future "enriched" site-saturation enzyme libraries may benefit from including this amino acid.</p><p>While the PECAN screen was able to identify P450BM3 variants capable of producing nootkatone, some hits were false positives, and the relative intensities of oxidation products observed using the PECAN screen (e.g., 4c vs 5c) did not always match those measured by GC-MS (e.g., 4a vs 5a, Table S1). This is likely because probe 3b is not a perfect surrogate for 3a. Compared to 3a, which has 15 heavy atoms, 3b has 18, presenting a 20% increase in size. Larger enzyme substrates, when similarly functionalized with a "clickable" functional group, would be expected to yield probes more representative of their archetype. To further improve hit identification, PECAN could be used in conjunction with P450 fingerprinting, in which enzyme variants' activities on multiple probes are correlated to activities on the substrate of interest measured for a subset of the library[40–42]. Alternatively, "label-free" technologies enabling the tagging of unfunctionalized biomolecules[43,44] promise to avoid this bias altogether.</p><p>While the potential for high-throughput NIMS-based analysis of complex enzyme reaction mixtures has frequently been alluded to, thus far all experiments describing the analysis of complex samples using NIMS have been limited to small studies[8,9,12], and all successful high-throughput NIMS experiments have been conducted with purified protein in low ionic strength buffer, avoiding the need for in situ fluorous affinity purification[43,45,46]. The screening campaign described here is the first demonstration of a high-throughput NIMS-based analysis of enzyme activity in a complex matrix.</p><p>In short, we have introduced the PECAN mass-spectrometric enzyme activity assay, and demonstrated its suitability for high-throughput screening of P450s in complex matrices such as cell lysates and microsomes. While here we have demonstrated the applicability of PECAN only to P450s, we believe that, after the appropriate method development, this technology could be applicable to any enzyme (or set of enzymatic reactions), that changes the mass of its substrate. Additionally, it may be possible to monitor isomerization reactions through the application of tandem MS, or enantioselective reactions through the use of isotopically labelled substrates[47]. As expected from any high-throughput assay, our PECAN screen yielded some false positives, and we re-screened a small subset of the library using low-throughput analytical technologies to verify the hits, as is routine in this field. Like spectrometric assays, the PECAN technology requires the synthesis of a specialized probe and enzymatic reactions are performed in microtiter plates. Therefore, we expect PECAN to have roughly equivalent throughput to – and act as a complementary method to – spectrometric assays. We expect PECAN to be a valuable technology for drug discovery or directed evolution campaigns that focus on enzymes for which no high-throughput screen currently exists.</p><p>Supporting information for this article is given via a link at the end of the document</p>

PubMed Author Manuscript

mTORC1 signalling and eIF4E/4E-BP1 translation initiation factor stoichiometry influence recombinant protein productivity from GS-CHOK1 cells

Many protein-based biotherapeutics are produced in cultured Chinese hamster ovary (CHO) cell lines. Recent reports have demonstrated that translation of recombinant mRNAs and global control of the translation machinery via mammalian target of rapamycin (mTOR) signalling are important determinants of the amount and quality of recombinant protein such cells can produce. mTOR complex 1 (mTORC1) is a master regulator of cell growth/division, ribosome biogenesis and protein synthesis, but the relationship between mTORC1 signalling, cell growth and proliferation and recombinant protein yields from mammalian cells, and whether this master regulating signalling pathway can be manipulated to enhance cell biomass and recombinant protein production (rPP) are not well explored. We have investigated mTORC1 signalling and activity throughout batch culture of a panel of sister recombinant glutamine synthetase-CHO cell lines expressing different amounts of a model monoclonal IgG4, to evaluate the links between mTORC1 signalling and cell proliferation, autophagy, recombinant protein expression, global protein synthesis and mRNA translation initiation. We find that the expression of the mTORC1 substrate 4E-binding protein 1 (4E-BP1) fluctuates throughout the course of cell culture and, as expected, that the 4E-BP1 phosphorylation profiles change across the culture. Importantly, we find that the eIF4E/4E-BP1 stoichiometry positively correlates with cell productivity. Furthermore, eIF4E amounts appear to be co-regulated with 4E-BP1 amounts. This may reflect a sensing of either change at the mRNA level as opposed to the protein level or the fact that the phosphorylation status, as well as the amount of 4E-BP1 present, is important in the co-regulation of eIF4E and 4E-BP1.

mtorc1_signalling_and_eif4e/4e-bp1_translation_initiation_factor_stoichiometry_influence_recombinant

Introduction<!>Materials<!>Cell culture and general sample preparation<!>35S-methionine incorporation assay<!>Pull-down assay using γ-aminophenyl-7-methyl-guanosine 5′-triphosphate agarose<!>Gene silencing by siRNA<!>SDS–PAGE and western blot analysis<!>Immunofluorescence microscopy<!>Comparison of cell productivity and translation rate in parental (Null 8) and mAbs producing cell lines (CHO42, CHO52 and CHO137).<!>Autophagy is activated towards the end of batch culture in GS-CHOK1 cells<!>Qualitative and quantitative analysis of factors regulating the translation initiation in parental and mAb-producing cell lines.<!>Translation initiation factors undergo an iterative series of phosphorylation changes throughout batch culture<!>Direct and indirect mTOR-mediated translational control mechanisms.<!><!>Amounts of the chaperone Hsp27 correlate with eIF4E levels<!><!>Expression of 4E-BP1 and eIF4E are co-regulated in GS-CHOK1SV cells, but changes in the eIF4E/4EBP1 ratio do not significantly influence recombinant mAb production<!>Role of PPM1G in modulating 4E-BP1 activity.<!>Does eIF4E and 4E-BP1 subcellular localisation relate to productivity of CHO cells?<!>Discussion

<p>Cultured mammalian cell expression systems, particularly Chinese hamster ovary (CHO) cells, are used to manufacture high-value biotherapeutic recombinant proteins (rPs) such as monoclonal antibodies (mAbs) [1]. Although the ability of such cell manufacturing 'factories' to produce rP products has been advanced over the last two to three decades, there is still an interest in further optimising the yields that can be delivered for novel and difficult to express proteins and the quality/homogeneity of the protein product. Two key parameters determining the amount of recombinant protein an expression system generates are (i) the maximum viable cell concentration achieved and the length of time this is maintained (the integral of viable cell concentration or IVC) and (ii) the cell-specific productivity of the cell (qP, or the amount of recombinant protein a cell manufactures per unit time, usually expressed as pg/cell/day). In this regard, improved growth medium and feeding strategies have resulted in dramatic increases in the maximum viable cell concentration and IVC achievable for mammalian cells in the bioreactor, and this has been associated with extended culture viability and enhanced recombinant protein yields [2]. Indeed, achieving high yields of recombinant protein of a clinically acceptable quality is dependent on a multiplicity of parameters, including the choice of the host expression system, the achieved viable cell mass [3,4], gene copy number [5], site-specific integration [6,7], the cellular processes responsible from gene to protein in the synthesis of the desired product [1,8], the bioreactor environment (e.g. nutrients and oxygen levels) [2,9], the authenticity and homogeneity of the product [10,11] and the yield and success of downstream processing [12,13].</p><p>There are now many reports suggesting that major cellular constraints upon recombinant protein production (rPP) can be post-transcriptional ([14–16]). One such control point is mRNA translation, with many reports having now reported that global and specific mRNA translation is a key parameter influencing rPP yields [14–16]. Other reports have suggested that ribosome biosynthesis also influences recombinant protein yields [17], and these reports collectively demonstrate that control of mRNA translation and ribosome biogenesis are important factors in cell growth and recombinant protein yield from mammalian cells. mRNA translation (protein synthesis) is catalysed by ribosomes, and mammalian target of rapamycin complex 1 (mTORC1) is a master regulator of ribosome biogenesis as well as mRNA translation [18]. With respect to the growth of recombinant cell lines, mTORC1 regulates these processes via the co-ordination of signalling pathways in response to growth factors, nutrient availability (amino acids), intracellular energy status (ATP levels) and diverse cell stresses [18], all factors that play key roles in regulating recombinant protein yields from mammalian cells. mTORC1 is thus likely to be a key global regulator of exactly those properties that are essential to achieving and maintaining high-level rPP from mammalian cells.</p><p>mTOR is the catalytic subunit of two functionally distinct complexes, mTORC1 and mTORC2. mTORC1 senses extra- and intra-cellular signals whereby growth factors, nutrients and energy promote mTORC1-dependent cell growth/proliferation and protein synthesis. At the same time, mTORC1 promotes ribosome biogenesis by enhanced transcription of ribosomal RNAs and translation of mRNAs for ribosomal (r-)proteins to increase protein synthetic capacity. Reduced mTORC1 activity leads to activated macroautophagy, which mediates the breakdown of cellular components into building blocks (e.g. amino acids and other small molecules) that may compensate for deficient nutrient supply [18]. The mTORC1 component raptor binds ribosomal S6 protein kinase 1 (S6K1) and eukaryotic initiation factor 4E-binding protein 1 (4E-BP1), thereby recruiting these substrates to be phosphorylated by mTOR. When hypophosphorylated, 4E-BP1 binds to eIF4E and prevents it from interacting with eIF4G to promote ribosome recruitment to mRNAs. It can thus repress the initiation of mRNA translation. By directly phosphorylating 4E-BP1 at multiple sites (in human 4E-BP1, Thr37/46, Thr70, Ser65), mTORC1 promotes its dissociation from eIF4E allowing the formation of the eIF4F complex and the initiation of cap-dependent translation [18]. Recent work has shown that increased 4E-BP1 phosphorylation is correlated with enhanced interferon-γ production in CHO cells. The authors suggest that this was due to the alleviation by mTORC1 of the repression of translation initiation [19].</p><p>With respect to rPP in mammalian cells, exogenous expression of mTOR has been reported to simultaneously improve key processes underpinning rPP from CHO cells, including cell growth, proliferation, viability and cell-specific productivity [20]. A further study has reported that in plasma cells (the cells that 'naturally' synthesise and secrete Igs), protein synthesis is regulated by cross-talk between endoplasmic reticulum stress and mTORC1 signalling [21]. Others [22] report that rapamycin differentially targets S6K and 4E-BP1, two downstream effectors of mTORC1. Here, we have examined the expression and phosphorylation state of many key proteins involved in mTORC1 signalling, as well as downstream effectors, using western blot analysis of CHO cell lysates collected throughout culture of cell lines with different productivity characteristics. Our data show that the flux of protein synthesis was altered across the culture time course reflecting mTORC1 signalling in the different cell lines. 4E-BP1 protein amounts were found to be elevated in a low-producer cell line alongside eIF4E amounts. We therefore sought to explore whether the relative amounts of eIF4E and 4E-BP1 influence the phenotype of cell lines (their productivity) in an initial set of four cell lines and a larger pool of cell lines [15]. These data show that the eIF4E/4E-BP1 translation initiation factor stoichiometry relates to recombinant protein productivity from glutamine synthetase (GS)-CHOK1 cells, and we discuss the implications of this for cell line engineering approaches.</p><!><p>Materials were obtained from Sigma–Aldrich unless otherwise indicated below.</p><!><p>GS-CHOK1 cells were from Lonza Biologics. Cells were grown in CD-CHO medium (Invitrogen) supplemented with 25 µM l-methionine sulfoximine. Cells were passaged three times prior to seeding 100 ml cultures for each cell line in 500 ml Erlenmeyer shaking flasks at 0.3 × 106 viable cells/ml. Cell counts were performed daily using a Vi-CELL 1.01 instrument (Beckman Coulter) to determine total and viable cell concentrations using the trypan blue dye exclusion method. Samples were taken each day (for 8–11 days), until cultures dropped below 60% viability. At each sampling point, 1 × 107 viable cells were removed, centrifuged at 1000 rpm for 3 min at 4°C and the supernatant removed (and immediately frozen at −20°C). The pellet was lysed in 200 µl of western lysis buffer [20 mM Tris–Cl (pH 7.5), 10 mM EDTA, 10 mM EGTA, 150 mM NaCl and 1% (w/v) Triton and 2 µl of protease/phosphatase inhibitor cocktail 100× (New England Biolabs)]. Samples were further centrifuged at 13 000 × g for 2 min at 4°C in order to sediment cell debris. The cytosolic fractions were then transferred to a fresh tube and sample buffer was added. The protein extracts were immediately stored at −20°C.</p><!><p>Viable cells (2 × 106) in 2 ml of medium were labelled with 762 kBq of [35S]methionine (PerkinElmer) in CD-CHO medium (Invitrogen) for 1 h, washed once with PBS and lysed in buffer containing 1% Triton X-100, 1 mM EDTA, 50 mM Tris–Cl, 1 mM EDTA, 0.1% β-mercaptoethanol, 1× protease/phosphatase inhibitor cocktail (#5872, Cell Signaling Technology).</p><!><p>Immobilised γ-aminophenyl-7-methyl-guanosine 5′-triphosphate (m7GTP)-agarose was purchased from Jena Bioscience. Beads (#AC-155S) were incubated with fresh CHO cell extracts in buffer containing 1% Triton X-100, 1 mM EDTA, 50 mM Tris–Cl, 1 mM EDTA, 0.1% (v/v) β-mercaptoethanol, 1× protease/phosphatase inhibitor cocktail (# 5872, Cell Signaling Technology) at 4°C for 2 h and then washed three times with cold PBS buffer. The proteins attached to the washed agarose were then subjected to 16% SDS–PAGE followed by western blotting.</p><!><p>Custom-made Stealth siRNAs were purchased from Invitrogen. Cells were seeded in six-well plates at a density of 750 000 cells/well and transfected with 4.5 (CHO-42) or 6.0 µl from a 20 nM siRNA pool against Chinese Hamster 4E-BP1 using Lipofectamine LTX (Invitrogen). Cell extracts were examined 48 h after transfection. For protein phosphatase magnesium-dependent 1 gamma (PPM1G), gene silencing was carried out using a 20 nM RNA Max stock from Eurofins and cells were transfected with Hi-Perfect (Qiagen).</p><!><p>Proteins were run on Tris–glycine gels [6, 10 and 16% (w/v) acrylamide, depending on the protein of interest]. After transfer to the polyvinylidene difluoride membrane, bound antibodies were detected using standard Enhanced Chemiluminescence analysis. Anti-β-actin antibodies (all diluted at 1/5000) were purchased from Sigma–Aldrich. Anti-4E-BP1 (clone 5H11) and eIF4G antibodies were purchased from Cell Signaling Technology. Secondary antibodies were either horseradish peroxidase-conjugated anti-rabbit or anti-mouse (both from Sigma–Aldrich). Anti-eIF4E antibodies were a kind gift from Prof. Simon Morley (Sussex). Phospho-S6 ribosomal protein (Ser240/244) (D68F8) XP rabbit mAb was purchased from Cell Signaling Technology.</p><!><p>Prior to the addition of CHO42 and CHO52, sterile circular coverslips were deposited into 24-well plates and coated with Corning Cell Tak Adhesive (at a concentration of 35 µg per ml, making sure the pH was in the range of 6.5–8). A 150 µl aliquot of a mid-exponential culture was added to the well. Following attachment, the cells were immediately fixed with 4% paraformaldehyde and permeabilised with 0.5% Triton in 1× PBS. All primary and secondary antibodies used in the present study were diluted 1/100 in 1% goat serum in 1× PBS. Goat anti-rabbit IgG (whole molecule)–TRITC (tetramethyl rhodamine isothiocyanate) antibody and goat anti-mouse were purchased from Sigma–Aldrich. Coverslips were mounted on slides with Vectashield with or without DAPI (at a final concentration of 0.1 µg/ml).</p><!><p>(A) For all cell lines, the growth profile was measured over the course of 9 days. (B) mAb secretion was also analysed in parallel by western blot analysis of the culture supernatant. (C) The overall rate of mRNA translation between cell lines was compared by adding 760 kBq of l-[35S] methionine to 2 × 106 live cells for 1 h, followed by a PBS wash. The neosynthetised proteins were separated using standard SDS–PAGE and were revealed on X-ray film autoradiography. The bands assigned as heavy and light chains are indicated by arrows. (D) The rate of appearance of mAbs was also examined following the procedure described in C, except that the cell fraction was incubated with protein A beads prior to gel separation. (E) Autophagy. The monitoring of the autophagosomal marker LC3-II was performed in the presence or absence of the lysosomal inhibitor chloroquine.</p><!><p>In the same cell line panel, we also investigated the intracellular accumulation of autophagy markers. Autophagy can reflect nutrient deprivation and activation of autophagy is a known strategy to preserve cellular fitness [24]. LC3-II amounts were therefore determined in cell lysates from the different cell lines throughout culture in the presence and absence of chloroquine by western blotting. These analyses showed that, in the CHO52 cell line, the least productive line and the one where cell viability decreased earlier than the other cell lines investigated, the onset of autophagy was observed earlier than in the Null8 and high-producing cell lines, as indicated by the increased amount of the modified LC3 (derived by lipidation of LC3-I to generate LC3-II; Figure 1E). This suggests that autophagy as a result of cellular stress, possibly nutrient deprivation, is activated earlier in the CHO52 cell line than the other cell lines investigated.</p><!><p>(A) Purification of mRNA cap-binding proteins using m7GTP-agarose beads using total cell extracts, done on different days (D) as stated. For each day, the panel shows SDS–PAGE/western blot analysis of the 4E-BP1, eIF4E, eIF4G and β-actin proteins in the input (I) and bound (B) fractions. (B) Total proteins were separated by SDS–PAGE, and the levels of total 4E-BP1, eIF4E, eIF4G and β-actin were examined by western blot analysis on stated days (D). (C) To detect faster running bands, D3 and D4 from CHO137 were obtained from an independent time course and separated at greater resolution.</p><!><p>We also investigated the amounts of total 4E-BP1, eIF4E and eIF4G proteins at each point of the time course (Figure 2B,C). Their total levels did not change appreciably across the time course (Figure 2B). However, the low-producer CHO52 cell line appeared to have higher amounts of both 4E-BP1 and eIF4E compared with the other three cell lines when compared with the β-actin loading control (Figure 2B). Samples from CHO137 lysates were also run from an independent time course and separated for 4E-BP1 western analysis at greater resolution in order to visualise the different bands more clearly.</p><!><p>(A) P-eEF2 Thr56 phospho-isoform and eEF2 examined by western blot analysis on stated days (D). (B) P-S6 Ser240/Ser244 phospho-isoform and S6 total examined by western blot analysis on stated days (D). For both set of gels, the ratio between eEF2-P/eEF2 and S6-P/S6 was calculated. The highest ratio value was set to 1 and used as the reference.</p><!><p>Relationship of 4E-BP1, eIF4E, 4E-BP1:eIF4E, and Hsp27 to product concentration and each other in recombinant antibody producing Chinese hamster ovary cell lines. (A–F) The relative levels of 4E-BP1, eIF4E and Hsp27 were measured in 15 different cell lines that exhibit differential recombinant protein productivity. The correlation coefficients between 4E-BP1 or Hsp27 levels and recombinant protein product concentration and P-values were calculated using SigmaPlot. Values in brackets correspond to recalculated R and P-values in the absence of the outlier(s) identified by Cook's distance. Outliers were defined as such if they were above three times the average Cook's distance.</p><!><p>Folding of newly synthesised polypeptides into the correct, active 3D shape is often assisted by proteins commonly termed chaperones, which include the heat shock proteins (Hsps). In addition, small Hsps can exert a protective role on their substrate. Of particular interest here, Andrieu et al. [30] showed that Hsp27 can interact with eIF4E and protect it from degradation. We therefore investigated whether there was a correlation between eIF4E amounts and Hsp27 amounts in our 15 model CHO cell lines. Our analysis in Figure 4E confirmed that not only did the amounts of eIF4E correlate with those of 4E-BP1 (Figure 4D), but also with Hsp27 levels (Figure 4E, R = 0.673, P < 0.01). The correlation coefficient (R = 0.625) and P-value were slightly lower (P = 0.0169) when we removed the outlier identified by Cook's distance (Figure 4E).</p><p>We have previously conducted a study on chaperone-assisted rPP in CHO cell lines [10]. This work revealed that transient overexpression of Hps27 could lead to enhanced yields of recombinant cytoplasmic Firefly luciferase but not a recombinant secreted Gaussia luciferase in CHOK1 cells. We therefore investigated the expression profile of Hsp27 (Supplementary Figure S1) across batch culture in the Null8, low- and high-producer cell lines from our original panel. In the Null8 cell line, Hsp27 remained more or less the same across the time course. However, Hsp27 levels clearly decreased as batch culture proceeded in the low (CHO52) and high (CHO137 and CHO42) producers (Supplementary Figure S1).</p><!><p>Modulation of eIF4E/4E-BP1 ratios in low (CHO42) and high (CHO52) producers. (A–D) CHO cells were transfected with human eEF4E variants together with mock siRNA (−) or siRNA (+) directed at 4E-BP1. Samples of cell lysate were analysed by western blot for the indicated proteins.</p><!><p>We also generated a series of human eIF4E (heIF4E) variants to assess the effect of heIF4E overexpression on 4E-BP1 levels. The constructs were V5-tagged to discriminate between endogenous and exogenous eIF4E. Because overexpression of any protein potentially has an impact on the translation machinery, we also created two mutants of heIF4E, [W73R] and [V69G], which have previously been reported [33]. heIF4E [W73R] and V69G cannot bind to eIF4G (or to 4E-BP1) and therefore cannot initiate translation at the mRNA cap. As the amounts of eIF4E are known to be regulated by many factors (e.g. AU-rich elements in the eIF4E 3′-UTR mediate binding of HuR to the eIF4E mRNA and its stabilisation [34]), we modified the heIF4E wild-type constructs by adding the hamster eIF4E 3′-UTR. We found that there was no detectable difference in eIF4E amounts when exogenous eIF4E-encoding plasmids were introduced into CHO cells when probing for total eIF4E (Supplementary Figure S2A,B). Detection of exogenously expressed eIF4E variants via the V5-tag showed that these variants were expressed (wild-type and 3′-UTR variants) or not detected at all (W73R variant) in CHO cells (Supplementary Figure S2A,B). Due to use of the V5-tag antibody to detect the exogenous variants, it is not possible to determine how the amounts of exogenous material directly compare with the endogenous material. In Figure 5C,D, overexpression of 4E-BP1 did not result in enhanced eIF4E protein amounts, suggesting again that the amount of eIF4E is tightly regulated or that the amount of exogenous 4E-BP1 expressed was insignificant/not sufficient to influence eIF4E amounts. One of the aims of this analysis was to assess whether increase in the eIF4E/4E-BP1 ratio could improve the productivity of low-producing cell lines such as CHO52; however, this proved difficult to achieve in this cell line at least and we could not detect any change in the overall rate of translation between the conditions tested as inferred from 35S labelling of new proteins (Supplementary Figure S3A,B). We also performed an ELISA assay to compare the production of IgGs from CHO42 and CHO52 in the presence of either 4E-BP1 siRNA or V5-tagged IF4E (data not shown). No significant difference in antibody was detected at 48 or 72 h post-transfection, demonstrating the complexity of regulation of 4E, 4E-BP1 and rPP, and that overexpressing or down-regulating individual members of a highly regulated network does not necessarily result in a modification of the desired phenotype.</p><!><p>(A) 4E-BP1 isoforms in CHO. (B) PPM1G knockdown (KD), mRNA levels were normalised to the β-actin mRNA. (C and D) Protein levels in control or silenced (KD) samples. (E) Effect of PPM1G knockdown on recombinant protein expression (IgG/mAb expression).</p><!><p>Lejbkowicz et al. [37] showed that while eIF4E is predominantly cytoplasmic, in mammalian cells a significant fraction (12–33%) localises to the nucleus, where it appears to co-localize with splicing factors [38]. The nuclear import of eIF4E is mediated by 4E-T (eIF4E transporter), which binds to eIF4E and simultaneously interacts with the nuclear import receptors importin α/β [39]. It is thought that nuclear eIF4E helps promote the export from the nucleus of a subset of mRNAs [40]. Furthermore, Rong et al. [41] showed that around a third of endogenous 4E-BP1 is localised to the nucleus in mouse embryo fibroblasts and that 4E-BP1 can regulate the subcellular localisation of eIF4E. Since our work showed (i) that the total levels of 4E-BP1 correlated with those of eIF4E, (ii) that the eIF4E/4E-BP1 ratio weakly correlates with cell productivity and (iii) that knockdown of 4E-BP1 resulted in a reduction in eIF4E amounts, we examined whether the distribution of eIF4E and 4E-BP1 was different in the low- (CHO52) and high-producer (CHO42) cell lines. Immunofluorescence, using antibodies against 4E-BP1 or eIF4E (Supplementary Figure S4), showed little difference in the localisation or intensity of the signal for the two proteins between the cell lines.</p><!><p>Here, we have investigated selected effectors of the mTORC1 signalling pathway that are involved in the control of protein synthesis in commercially relevant GS-CHOK1SV mAb-producing cell lines. In particular, we have investigated the relationship between 4E-BP1 amounts and phosphorylation and eIF4E amounts. Although the signalling pathways that converge on the mTORC1 kinase as well as mTORC1 downstream effectors have been widely reported, few studies have focused on mTORC1 regulation in rPP systems. Recent findings have implied that the mTORC1 signalling network could be exploited in bioprocessing. For example, exogenous expression of the mTOR kinase in CHO cell lines led to increased recombinant IgGs yields [20], although this study did not measure the amount of, or confirm, exogenous expression. Dadehbeigi and Dickson [22] also showed that inhibition of growth and rP titre in rapamycin-treated cells was transient, while 4E-BP1 phosphorylation remained stable. These two sets of results highlight that the interpretation of mTORC1 regulation in the context of rPP is complex [22]. More recently, Edros et al. [42] performed a transcriptomic analysis on a large panel of factors related to mTORC1 signalling in two different recombinant CHO cell lines with a 17.4-fold difference in mAb productivity. They showed that, across this pool of 84 genes, eight genes exhibited differences of >1.5-fold. These included upstream regulators of mTORC1 (AMPK, phospholipase D and Ras-related GTP-binding protein C) and one mTOR effector (the ribosomal protein S6). However, the present study only highlighted transcripts (mRNA levels), whereas other post-transcriptional and post-translational control mechanisms govern the mTOR network.</p><p>Besides cell proliferation, mTORC1 regulates diverse pathways including autophagy, a process which recycles broken-down intracellular components. The proteins ULK1, Atg13 and FIP200 complex link mTORC1 signalling to autophagy [43]. Our study indicates that a low-producing cell line (CHO52), where cell viability decreases during batch culture earlier than other cell lines, exhibits markers of autophagy (conversion of LC3-I into LC3-II) at an earlier stage compared with those cell lines where viability does not decrease at similar times (including both higher producing cell lines and a Null-producing cell line). Others have reported that autophagy can be beneficial for cell survival [24]. These data collectively suggest that a balance between maintaining cell viability and the benefits of 'recycling' of cellular components late in culture together can influence recombinant protein output. However, our data also suggest that activation of autophagy in a low-producing cell line is not helpful in terms of productivity as it is associated with a lower protein synthesis rate and a decline in viable cell numbers in the cell lines investigated here.</p><p>Previous reports have suggested that the yield of recombinant proteins from cultured mammalian cells is in part attributed to translation efficiency [8,44]. mTORC1 exerts its influence on translation via the mTORC1 effectors 4E-BP1, p70 S6 kinase and eEF2K. While we observed that 4E-BP1 amounts could differ across recombinant protein-producing cell lines, 4E-BP1 amounts alone did not correlate to cell productivity. Rather, the protein ratio of 4E-BP1 to eIF4E, a central parameter in cap-dependent translation, showed a higher degree of correlation with cell productivity. Furthermore, eIF4E amounts appear to be co-regulated with changing 4E-BP1 amounts when 4E-BP1 was reduced by knockdown experiments, but not when 4E-BP1 was reduced by manipulation of the PPM1G phosphatase. This may reflect a sensing of either change at the mRNA level as opposed to the protein level or the fact that the phosphorylation status and the amount of 4E-BP1 present are important in the co-regulation of eIF4E with 4E-BP1. With respect to 4E-BP1 phosphorylation, mammalian cells contain multiple 4E-BP1 isoforms, but the isoform pattern observed in rodent cells is simpler [43,44]. Consistent with the rat model, our work and that of others [45] show that in CHO cells, a total of three 4E-BP1 species are usually detected: a hyperphosphorylated form γ, a middle form β and a hypophosphorylated α form (see Figures 2B and 6A). However, lower resolution acrylamide/bis-acrylamide gels may only reveal two bands. The intensity of each band probably reflects the physiological state of the cell. Indeed, as inferred by the m7-agarose-binding experiment (Figure 2A), the phosphorylation status of 4E-BP1 is altered through the batch culture time course. mTOR-mediated phosphorylation of 4E-BP1 is a precisely orchestrated process, with phosphorylation in human 4E-BP1 on Thr37 and Thr46, seemingly acting as a priming event for subsequent phosphorylation on other serine/threonine sites [30, 31]. These events are generally thought to influence the affinity of 4E-BP1 for eIF4E and, in this way, regulate translation initiation, although this model has recently been disputed by Showkat et al. [46].</p><p>As 4E-BP1's state of phosphorylation has an impact on translation initiation, we wished to explore whether changing the amounts of 4E-BP1 as a whole or the isoforms with respect to each other would lead to discrete changes in recombinant protein synthesis. mTORC1 catalyses the phosphorylation of 4E-BP1, but also regulates many other cellular events; therefore, manipulating mTOR-mediated phosphorylation may lead to diverse consequences. Liu et al. [36] performed lentivirus-mediated silencing of a series of serine/threonine phosphatases and identified PPM1G as a phosphatase for 4E-BP1. We therefore induced PPM1G knockdown via siRNA-mediated silencing (Figure 6B) and observed a large drop in the overall amount of 4E-BP1, but only a subtle change in the phosphorylation pattern (the highest phosphorylated form being more abundant; Figure 6C,D). This may indicate that artificially altering the 'standard' mTOR-driven sequence of phosphorylation events in 4E-BP1 can lead to 4E-BP1 instability. Notably, the levels of eIF4E were unchanged (data not shown).</p><p>Analysis of IgG molecules by western blot (Figure 6E) showed that there was no increase in product yield following PPM1G knockdown, which is in contrast with the observations of Liu et al. [36]. Even if PPM1G knockdown led to an overall decrease in 4E-BP1, this should not significantly affect the amount of IgG being secreted from the cells, as we found that 4E-BP1 amounts did not correlate with cell productivity (Figure 4). Finally, as inferred by our [35S]methionine labelling experiment (Supplementary Figure S3), 4E-BP1 knockdown or overexpression at the levels achieved in the present study are not sufficient to reprogram the core capacity of a cell to undertake translation, this being observed in both high- (CHO42) or low- (CHO52) producing cell lines. This finding demonstrates that manipulation of single components within a complex regulatory process does not necessarily lead to a change in phenotype.</p><p>Another noteworthy finding in our study was that the levels of the chaperone Hsp27 declined over the time course. This was not observed in the non-producing cell line Null8. Tan et al. [47] observed that stably expressing Hsp27 could improve rPs yield in cells. Interestingly, when we looked at our panel of 15 recombinant cell lines, we saw no correlation between Hsp27 and cell productivity (Figure 4). However, Hsp27 levels correlated with eIF4E levels (Figure 4). A study by Cuesta et al. [48] suggested that Hsp27 targets eIF4E and eIF4G and attenuates protein translation in stressed cells by preventing the assembly of the cap initiation complex. Andrieu et al. [30] demonstrated that Hsp27 directly interacted with eIF4E. They suggested that Hsp27 has a protective role over eIF4E, thus protecting protein synthesis initiation. The exact mode of action of Hsp27 at the cap-binding stage remains unclear.</p><p>Our study highlights that the stoichiometry between the mTORC1 effector 4E-BP1 and the translation factor eIF4E differs across CHO recombinant cell lines and correlates with cell productivity. Interestingly, Yanagiya et al. [32] showed that a 90% reduction in eiF4E at the mRNA level only resulted in partial inhibition of mRNA translation. The same study shows that the 4E-BP1 hypophosphorylated form was less prone to degradation than hyperphosphorylated 4E-BP1, when eiF4E was knocked down. Our work shows that the stoichiometric balance between eiF4E and 4E-BP1 is constrained by the expression for each factor, stability, phosphorylation status, physical interaction and cellular localisation, which means that simply correcting the level of 4E-BP1 or eIF4E in the cells is not sufficient to convert a low-producing cell line into a high-producing cell line. Our analysis suggests that additional factors, such as eIF4G isoforms and Hsp27, are also likely to play a role in the tuning of this balance, and hence, upstream regulators of overall mTORC1 signalling are more likely to be targets whose manipulation can be utilised to drive cell phenotypes for enhanced rPP.</p>

PubMed Open Access

Liquid chromatographic nanofractionation with parallel mass spectrometric detection for the screening of plasmin inhibitors and (metallo)proteinases in snake venoms

To better understand envenoming and to facilitate the development of new therapies for snakebite victims, rapid, sensitive, and robust methods for assessing the toxicity of individual venom proteins are required. Metalloproteinases comprise a major protein family responsible for many aspects of venom-induced haemotoxicity including coagulopathy, one of the most devastating effects of snake envenomation, and is characterized by fibrinogen depletion. Snake venoms are also known to contain anti-fibrinolytic agents with therapeutic potential, which makes them a good source of new plasmin inhibitors. The protease plasmin degrades fibrin clots, and changes in its activity can lead to life-threatening levels of fibrinolysis. Here, we present a methodology for the screening of plasmin inhibitors in snake venoms and the simultaneous assessment of general venom protease activity. Venom is first chromatographically separated followed by column effluent collection onto a 384-well plate using nanofractionation. Via a post-column split, mass spectrometry (MS) analysis of the effluent is performed in parallel. The nanofractionated venoms are exposed to a plasmin bioassay, and the resulting bioassay activity chromatograms are correlated to the MS data. To study observed proteolytic activity of venoms in more detail, venom fractions were exposed to variants of the plasmin bioassay in which the assay mixture was enriched with zinc or calcium ions, or the chelating agents EDTA or 1,10-phenanthroline were added. The plasmin activity screening system was applied to snake venoms and successfully detected compounds exhibiting antiplasmin (anti-fibrinolytic) activities in the venom of Daboia russelii, and metal-dependent proteases in the venom of Crotalus basiliscus. Graphical abstractᅟ Electronic supplementary materialThe online version of this article (10.1007/s00216-018-1253-x) contains supplementary material, which is available to authorized users.

liquid_chromatographic_nanofractionation_with_parallel_mass_spectrometric_detection_for_the_screenin

Introduction<!>Chemicals and biological samples<!>LC, nanofractionation and MS detection<!>Plasmin bioassay<!>Reversed-phase (RP) LC and hydrophilic interaction liquid chromatography (HILIC) orthogonal separation strategy, fraction collection, proteomics<!>Tryptic digestion of bioactive proteins<!>NanoLC-MS/MS analysis of the tryptic digests<!>Evaluation of the plasmin bioassay<!><!>Profiling plasmin inhibition and proteolytic activity in Cb and Dr venoms<!><!>Identification of the bioactives<!>Conclusions<!>

<p>Plasmin is a trypsin-like enzyme, which plays a crucial role in the fibrinolytic pathway where it hydrolyses blood clots and maintains normal hemostasis. Plasmin originates from plasminogen, and is activated by tissue plasminogen activator (tPA) or urokinase plasminogen activator (uPA) [1]. Although the fibrinolytic activity of plasmin is regulated by the physiological inhibitors α2-antiplasmin and α2-macroglobulin, in many cases (e.g., major surgeries, hemophilia, von Willebrand syndrome, heavy menstrual bleeding), medicinal exogenous inhibitors are required to properly control its activity [1, 2]. Examples of plasmin inhibitors currently used successfully in the clinic are ε-aminocaproic acid (EACA) and tranexamic acid (TXA). However, their potency and selectivity are low, and consequently, these drugs have to be administered in high doses, which significantly increases the risk of adverse effects [1]. Therefore, there is a high demand for new, more potent and selective, anti-fibrinolytic agents.</p><p>Many snake venoms are a rich source of molecules that interfere with coagulation and fibrinolysis pathways in the hemostatic system [3]. Both anti- and pro-fibrinolytic activities have been observed, mostly in the venoms of the Viperidae snake family [4]. Antiplasmin proteins usually belong to the Kunitz/bovine pancreatic trypsin inhibitors (BPTI) family and so far a number of Kunitz-type protease inhibitors have been isolated and purified from snake venoms, including Kunitz inhibitor-I (DrKIn-I) and -II (DrKIn-II) from Daboia russelii [5], and the textilinins from Pseudonaja textilis [6].</p><p>The fibrin(ogen)olytic effect of snake venom toxins results from different mechanisms, e.g., stimulation of plasminogen activators from endothelial cells [7], direct plasminogen activation [8, 9], or direct cleavage and degradation of fibrinogen and fibrin [10–13]. Most of the molecules which exhibit fibrin(ogen)olytic activities belong to the snake venom serine protease (SVSP) and snake venom metalloproteinase (SVMP) toxin families [4, 14]. Both SVSPs and SVMPs are proteolytic enzymes of which the latter require zinc ions as metal co-factors for catalytic activity, and in some cases calcium for structural stability [14–16].</p><p>SVMPs have been recognized as one of the main components of venoms responsible for causing pathologies associated with snakebites [15]. Snakebites remain an important and neglected public health issue [17] especially in poorer, tropical areas of the world where access to healthcare is problematic. Such bites result in extensive mortality (~ 94,000 deaths/annum) and morbidity (~ 3–5 times the number of annual deaths), as the result of the hemorrhagic, coagulopathic, neurotoxic, and/or cytolytic effect of different snake venoms [18, 19]. Apart from their high hemorrhagic activities, which is most profound in the P-III class, SVMPs are reported to exhibit fibrin(ogen)olytic, apoptotic, inflammatory, and factor Xa and prothrombin-activating activities [14]. Considering the significant hazard that SVMPs represent to victims of envenomation, information on their presence, molecular structure and bioactive properties will help in the development of more specific, and thus more effective treatments for snake envenoming, such as aptamers [20], snake venom toxin inhibitors [21, 22], recombinant anti-venoms [23] or chelating agents [22, 24].</p><p>Due to the complex nature of snake venoms, multiple analytical techniques are required for the identification and biochemical characterization of individual toxins. A straightforward and effective methodology for screening venoms for selected bioactivities is nanofractionation analytics [25–27]. In this approach, a venom is first separated by liquid chromatography (LC) after which the column effluent is split, with one part being fractionated with high resolution into a high-density microtiter well plate (typically 96 to 1536 wells) and the other part being interrogated simultaneously by mass spectrometry (MS) analysis. The collected fractions on the microtiter plate are subsequently exposed to a bioassay. As chromatographic resolution is essentially maintained, the results of the bioassay readout can be constructed into a bioactivity chromatogram, which can be accurately correlated with the parallel obtained MS chromatogram. Besides venom research, the nanofractionation technique also has shown to be useful for drug-drug interaction profiling [23], environmental analysis [28], and profiling natural extracts [29].</p><p>In this study, we present nanofractionation analytics for the screening of plasmin inhibitors and proteases. We applied the developed methodology to the analysis of venoms from the medically relevant viperid snakes Daboia russelii and Crotalus basiliscus for the presence of plasmin inhibitors as well as (metallo)proteinases that exhibit fibrin(ogen)olytic activity similar to plasmin. Daboia russelii, also known as Russell's viper, is responsible for substantial proportion of snakebite-induced morbidity and mortality in Asia, and causes diverse symptoms such as coagulopathy, swelling, neurotoxicity and pain [30]. Crotalus basiliscus, known as the Mexican west-coast rattlesnake, as indicated by its name can be found mainly in Mexico. Envenomation by this snake leads to local tissue damage, systemic bleeding and hypotension [31]. The fibrin(ogen)olytic proteins were further evaluated for their activity dependency on metal ions by applying to the nanofractionated venoms the plasmin bioassay mixture enriched with zinc and calcium ions, and also in the presence of the metal chelating agents 1,10-phenanthroline or EDTA. Plasmin inhibitors and proteolytic enzymes detected by the applied screening methodology were subsequently characterized.</p><!><p>Human plasmin was purchased from Haematologic Technologies, Inc. (Essex Junction, VT, USA) and used at 670 μg/mL in 1:1 glycerol/water solutions (v/v). Fluorogenic substrate H-D-Val-Leu-Lys-AMC was purchased from BACHEM (Bubendorf, Switzerland) and dissolved in DMSO to a stock concentration of 40 mM. Leupeptin hydrochloride (dissolved in Milli-Q water to obtain a stock of 10 mM), 1,10-phenanthroline, bovine serum albumin (BSA), calcium chloride, DMSO, EDTA, iodoacetamide, trypsin from bovine pancreas, and Trizma were all obtained from Sigma (Zwijndrecht, The Netherlands). Hydrochloric acid (HCl) was received from Riedel-de-Haën (Zwijndrecht, The Netherlands), formic acid (FA) and β-mercaptoethanol were purchased from Merck (Darmstadt, Germany), and ULC-MS grade acetonitrile (ACN) was obtained from Biosolve (Valkenswaard, The Netherlands). Lyophilized snake venoms were from Daboia russelii (Dr) and Crotalus basiliscus (Cb), which both are from the Viperidae family. The Daboia russelii venom (Sri Lankan origin) was provided by Dr. Nicholas Casewell from the Alistair Reid Venom Research Unit, Liverpool School of Tropical Medicine, and the Crotalus basiliscus venom (Captive bred), was provided by Dr. Freek Vonk, Naturalis Biodiversity Center. Depending on the experiment performed, solutions of the venoms were prepared in Milli-Q water at concentrations of 4 mg/ml, 5 mg/mL or 0.5 mg/ml and stored at − 80 °C.</p><!><p>Bioassay optimization and analyses of crude snake venoms were performed using a Shimadzu high-performance liquid chromatography (HPLC) system ('s Hertogenbosch, The Netherlands) consisting of a Shimadzu SIL-30 AC autosampler and a Shimadzu LC-20AB binary pump, which were interfaced to an Ultima Q-TOF MS instrument (Waters, Bradfrod, UK). Dr and Cb venoms at a concentration 4 mg/ml (injected volume, 45 μL) and leupeptin standards (injected volume, 20 μL) at five different concentrations (ranging from 6 to 200 μM) were separated with a Waters XBridge C18 column (4.6 × 100 mm, particle size 5 μm) (Milford, MA). The analytical column was protected by a guard column comprising the same packing material. Both columns were kept in a CTD-30 column oven (Shimadzu) set to 37 °C. Gradient elution was performed employing mobile phase A (98% water, 2% ACN, 0.1% FA (v/v/v)) and mobile phase B (98% ACN, 2% water, 0.1% FA (v/v/v)). The gradient used for separation started at 0% B with a linear increase to 50% B in 20 min followed by an increase to 90% B in 2 min. The gradient was held at 90% B for 2 min, and then returned in 1 min to the starting conditions (0% B). Column re-equilibration was 5 min, resulting in a total analysis time of 30 min. The flow rate was 0.5 ml/min. Using a post-column flow split, 90% of the eluate was directed via a Shimadzu SPD-20A UV detector (set at 200 nm) to an in-house developed nanofractionation system, and the other 10% of the eluate was analyzed with an Ultima Q-TOF mass spectrometer equipped with electrospray ionization in positive ion mode (ESI+). The following settings of the ionization source were used: capillary voltage, 3.25 kV; source temperature, 125 °C; desolvation temperature, 200 °C; gas flow, 250 L/h. The mass range was set to 50–2000 m/z.</p><p>Detailed information on the nanofractionation system can be found elsewhere [28, 32]. Briefly, the system was built from a modified 235 Gilson autosampler, in which a fused silica capillary extension connected to LC tubing made of polyether ether ketone (PEEK) was mounted to the robotic arm, which allowed the capillary needle to move in xy directions. Fractions of 6 s were collected onto a black 384-well plate (Greiner Bio One, Alphen aan den Rijn, The Netherlands) in a serpentine way. After nanofractionation, the plates were evaporated overnight (O/N) using a RVC 2–33 CD plus maxi concentrator (Salm en Kipp, Breukelen, The Netherlands) with a rotation speed of 1500 rpm, pressure of 0.10 mbar, and temperature of 30 °C. After solvent evaporation, the dried plates were stored at − 20 °C prior to bioassaying. A simplified scheme showing of the analytical method can be found in Electronic Supplementary Material (ESM) Fig. S1.</p><!><p>Screening for snake venom components modifying the activity of plasmin was performed with an optimized bioassay mixture containing 100 ng/mL plasmin and 5 μM fluorogenic substrate H-D-Val-Leu-Lys-AMC dissolved in 100 mM TRIS-HCl buffer (pH 7.5) containing 0.1% BSA (w/v). The bioassay optimization results can be found in the ESM Section S1.1. The bioassay mixture was prepared by adding equal volumes of substrate and the enzyme solutions in 100 mM Tris-HCl buffer with 0.1% BSA. The buffer solution was at room temperature (RT), whereas the enzyme and substrate stock solutions were kept at − 20 °C. Prior to bioassaying, the enzyme and the substrate were thawed and diluted in buffer contained in two Greiner tubes. The solutions were carefully mixed (for 5 s) to prevent air bubble formation. Directly after preparing 50 μL of the bioassay mixture was pipetted over a black 384-well plate (Greiner Bio One, Alphen aan den Rijn, The Netherlands) using a Multidrop 384 reagent dispenser (Thermo Scientific, Ermelo, The Netherlands). A VarioSkan LUX microplate multimode reader (Thermo Scientific, Ermelo, The Netherlands) was used to measure fluorescence of each well kinetically at 380 and 460 nm excitation and emission wavelength, respectively. The temperature inside the plate reader was set at 37 °C. Each measurement comprised 10 cycles allowing generation of a kinetic curve, and subsequent determination of its slope, which was plotted against the time of nanofraction collection, producing a so-called bioactivity chromatogram. Negative peaks in the bioactivity trace represent inhibition of plasmin, and positive peaks indicate the presence of plasmin inducers or, more likely in the case of venoms, proteases exhibiting an enzymatic activity similar to plasmin.</p><p>The Z′ factor, introduced by Zhang et al. [33], was used as statistical parameter to assess the quality of the bioassay. For that, the inhibitor leupeptin (200 μM concentration representing full enzyme inhibition) and a sample containing mobile phase A were used as a positive and negative control, respectively. Both solutions were pipetted over a 384-well plate at volumes of 10 μL per well using the Multidrop 384 reagent dispenser. Next, the plates were vacuum-centrifuge evaporated and exposed to the plasmin bioassay. Measurements were performed in duplicate.</p><!><p>Identification of bioactive compounds from venoms of Dr and Cb involved nanofractionation onto 384-well plates with parallel high-resolution MS analysis using a Maxis HD quadrupole time-of-flight MS (Bruker Daltonics, Bremen, Germany). For nanofractionation and MS, 5 mg/ml Dr venom and 0.5 mg/ml Cb venom (50 μL injected using a SIL-20AC autoinjector; Shimadzu, Canby, OR, USA) were separated on an XBridge BEH300 C18 (4.6 × 150 mm; particle size, 3.5 μm) analytical column placed in a column oven (model CTO-20AC Shimadzu, Canby, OR, USA) kept at 30 °C. A binary mobile phase consisting of mixtures of solvent A (98% water, 2% ACN, 0.1% FA (v/v/v)) and solvent B (98% ACN, 2% water, 0.1% FA (v/v/v)) was used for gradient elution: 0–50% B in 20 min, 50–90% B from 20 to 24 min, isocratic at 90% B from 24 to 29 min, back to 0% B in 0.1 min. Column equilibration was then performed for another 10 min using 0% B, resulting in a total analysis time of 40 min. The flow rate was 0.5 ml/min The MS was equipped with an ESI source (Apollo) used in positive mode. The capillary voltage was 4.5 kV, and nebulization was done with nitrogen gas (0.4 bar). Ion desolvation was performed using nitrogen as a drying gas heated to 200 °C and flowing at a rate of 4 l/min. Mass accuracy was assured by internal calibration using an ESI-L low concentration tune mix (Agilent, Santa Clara, California, USA) in every run.</p><p>An orthogonal separation of the two crude venoms was performed using an Atlantis silica HILIC (4.6 × 150 mm; pore size, 100 Å; particle size, 3 μm) analytical column. The separation was performed using gradient elution with solvent A composed of 98% ACN, 2% water, 0.2% FA (v/v/v) and solvent B of 98% water, 2% ACN, 0.2% FA (v/v/v). The gradient started at 30% B and increased to 50% B in 15 min. The next 10 min, the gradient rose to 95% B and remained at 95% B for 10 min. Within 2 min, the gradient returned to initial conditions. Column equilibration was set for 23 min at 30% B resulting in a total analysis time of 60 min. The flow rate was 0.5 ml/min. MS conditions were the same as used for RPLC analysis.</p><p>The nanofractionated venoms were exposed to the plasmin bioassay by robotic pipetting of the bioassay mixture to every second well on the plate. As an eluting compound usually is distributed over multiple wells, the alternating wells (i.e., without bioassay mixture) could be used for other analyses (e.g., proteomics). From the bioactivity chromatograms, wells with bioactive compounds were identified and the content of adjacent wells was subsequently subjected to proteomic analysis. The content was dissolved in 100 μL water and after 30 min, 50 μL samples were transferred to low-protein binding Eppendorf tubes containing digestion buffer and reduction agent for tryptic digestion (see "Tryptic digestion of bioactive proteins" section).</p><p>Throughout the study, leupeptin was used as a time calibration point for accurate correlation of the bioactivity and MS chromatograms. Leupeptin was analyzed using RPLC and HILIC separation (ESM Figs. S2A and S2B, respectively). This experiment enabled calculation of the time delay, which is observed in the MS data acquisition compared to the nanofractionation process. This delay is due to differences in LC-tubing length between the two systems, the column used and the flow rates applied, and has to be measured after each change made in the system.</p><!><p>Bioactive proteins present in the venoms analyzed underwent tryptic digestions according the following protocol. Briefly, reduction of disulfide bonds was performed by adding 50 μL of each bioactive fraction (reconstituted in 100 μL water) from a nanofractionated snake venom to a solution containing 75 μL of 25 mM ammonium bicarbonate (pH 8.2) and 7.5 μL of 0.5% β-mercaptoethanol prepared in water (v/v). The solutions were shortly vortexed and then incubated at 95 °C for 10 min using a dry block heating thermostat (Bio TDB-100; Biosan Ltd., Riga, Lativa). After cooling down to room temperature, the solutions were centrifuged at a speed of 21,500 rpm for 10 s. Next, the reduced proteins were alkylated for 30 min with 15 μL of 100 mM iodoacetamide in the dark at room temperature. After the alkylation procedure, proteins were digested by adding 3 μL of a 0.1 μg/μL trypsin solution prepared in 25 mM ammonium bicarbonate (pH 8.2). The stock solution of the trypsin used was 1 mg/ml prepared in 1 mM HCl and stored at − 20 °C. The samples were incubated for 3 h at 37 °C followed by addition of an extra 3 μL of the same trypsin solution and then incubated O/N at 37 °C. The digestion was quenched with 5 μL of 5% formic acid in water (v/v) followed by centrifugation at 21,500 rpm for 10 s. The collected supernatant was transferred to LC-MS vials and analyzed with nanoLC-ESI-MS/MS (see "NanoLC-MS/MS analysis of the tryptic digests" section).</p><!><p>Following tryptic digestion, samples were separated and analyzed with an Ultimate 3000 Nano HPLC system (Thermo Scientific) interfaced to a Bruker Maxis quadrupole time-of-flight MS instrument. Detailed information on these analyses can be found in the ESM Section S1.2. Identification of proteins was performed with the search engine Mascot (version 2.301). Raw MS data were converted into centroided peak lists using DataAnalysis version 5.0 (Bruker). The parameters used for Mascot searches were set as follows: enzyme specificity, semiTrypsin; missed cleavages, 2; fixed modification, carbamidomethyl (C); variable modification, oxidation (M) peptide mass tolerance, 0.1%; fragment mass tolerance, 0.05 Da; no restriction for species was used. The instrument was set to ESI-QUAD-TOF.</p><!><p>Screening snake venoms for plasmin inhibitors and (metallo)proteinases exhibiting activity similar to plasmin was performed with a fluorescence based bioassay based on the protocol reported by Tervo et al. [34]. Bioassay optimization is described in the ESM section S1.1 and Fig. S3. The optimized bioassay mixture comprised 100 ng/mL end concentration of plasmin enzyme and fluorogenic aminocoumarin-based substrate H-D-Val-Leu-Lys-AMC at a final concentration of 5 μM, in 100 mM Tris-HCl buffer (pH 7.5) containing 0.1% BSA. BSA was used following the findings by Tervo et al. [34], who observed that it prolonged the enzymatic activity of plasmin. In our study, 0.1% BSA increased the signal-to-noise ratio (S/N) approximately three times and improved the baseline stability (data not shown).</p><!><p>Assessment of the bioassay performance. The Z′-factor for the bioassay was calculated to be 0.74. Positive control (C+) wells contain mobile phase A (n = 96); negative control (C−) wells contain 200 μM leupeptin (n = 96). The wells were evaporated to dryness prior to the bioassay</p><p>Evaluation of the overall performance of the nanofractionation system using the plasmin inhibitor leupeptin, which was nanofractionated at five concentrations injected. (a) 6.3 μM. (b) 25 μM. (c) 50 μM. (d) 100 μM. (e) 200 μM. (f) The extracted ion chromatogram (m/z 427.3, [M + H]+) of the analysis of 100 μM leupeptin with online LC-MS. RFU, relative fluorescence units</p><p>(a–f) Bioactivity chromatograms obtained for Cb and Dr venoms using the following assay variants. (a) Full plasmin bioassay mixture containing plasmin and substrate. (b) Bioassay mixture containing only the substrate. (c) Bioassay mixture containing only the substrate and buffer enriched with zinc ions. (d) Bioassay mixture containing only the substrate and buffer enriched with calcium ions. (e) Bioassay mixture containing only the substrate and buffer enriched with EDTA. (f) Bioassay mixture containing plasmin, substrate and 1,10-phenanthroline. (g) Total-ion chromatogram obtained for Cb and Dr venoms with online parallel LC-MS. The rectangular gray shades indicate the elution of the bioactive peaks. Photos of the snakes: Crotalus basiliscus (source: Shutterstock/Bernhard Richter); Daboia russelii (source: Shutterstock/Meet Poddar). RFU, relative fluorescence unit</p><!><p>The analysis of the venom of Dr resulted in bioactivity chromatograms showing clear inhibition of plasmin (Fig. 3a). Injection of a venom sample of 4 mg/ml showed multiple negative peaks eluting at 8.9 and 9.9 min, clearly indicating the presence of antiplasmin/anti-fibrinolytic activity. Dr venom has previously been reported to contain the Kunitz-type protease inhibitors DrKIn-I and -II which exert anti-fibrinolytic activity. Some reports, however, showed that Dr venoms also contain SVSPs and SVMPs, which could potentially exert fibrin(ogen)olytic properties [40, 41]. However, in our study, positive peaks were not observed.</p><p>For further investigation of the increased enzyme activities observed for the two venoms, six different variants of the plasmin bioassay were applied. In the first variant, the bioassay mixture was prepared without the addition of plasmin to test whether the active protein is a plasmin analogue that converts similar substrates or is a plasmin activator. In the second and third variants, the bioassay mixture was prepared without plasmin and was enriched with calcium (1 mM) or zinc (0.2 mM) ions, respectively, to determine whether bioactivity could be ascribed to a metalloproteinase and, if so, to differentiate between calcium or zinc dependency. To confirm that the resulting bioactive protein is a metalloproteinase, in the fourth and fifth variant of the bioassay, the metal ion chelators EDTA (50 mM) or 1,10-phenanthroline (5 mM) were added, while again excluding plasmin from the mixture. The chelator 1,10-phenanthroline specifically binds zinc ions, deactivating zinc-dependent metalloproteinases only, whereas EDTA chelates both zinc and calcium ions [42]. In the sixth and final variant, 1,10-phenanthroline was added to the full bioassay mixture, in order to deactivate SVMPs (if present) and allow the detection of inhibitors that are co-eluting with SVMPs.</p><p>Results from the analyses of both venoms using the six assay variants are shown in Fig. 3a–f for Cb and Dr, respectively. An extensive discussion of the results of these experiments can also be found in the ESM Section S2.1. Briefly, in the venom of Cb, the major positive peak was also observed when plasmin was not included in the bioassay mixture. This indicates that the bioactive protein detected is not a plasmin inducer, but an enzyme that is capable of converting the substrate. Subsequent experiments (bioassay variants 2–5) showed that the activity of the bioactive protein increased about threefold (compared to the results obtained with substrate only [variant 1]) when calcium ions were added to the bioassay mixture, indicating possible calcium dependency or increased stability of the protein. The addition of zinc ions did not affect the intensity of the bioactivity peak. To assure that the changes observed in the bioassay are not a result of the variation between different plates/bioassays tested, additional experiments confirming the dependence of the bioactive compounds on metal ions were performed. Two bioassay mixtures, with added calcium and zinc ions, respectively, were pipetted into alternating wells of one microtiter well plate containing nanofractionated venom. This was done to exclude potential inter-experiment variations between nanofractionated runs. The results (ESM Fig. S4a) showed an activity-inducing effect of the calcium ions on the eluted bioactive proteins, confirming the results of the earlier experiments. Figure 3 shows that, in the case of Dr venom, no increased enzyme activity was observed, indicating that the bioactive compounds detected (Fig. 3a–f) were predominantly plasmin inhibitors (anti-fibrinolytic agents) [43].</p><!><p>Correlation of bioactivity and MS chromatograms obtained for Cb (C. basiliscus) and Dr (D. russelii) venoms after RPLC and HILIC separations. The data were acquired online in parallel to the nanofractionation with LC-MS. XIC, extracted ion chromatogram; BPC, base peak chromatogram; RFU, relative fluorescence unit</p><!><p>When Dr venom (injected concentration, 5 mg/ml) was analyzed by RPLC-MS and HILIC-MS, the observed bioactivity peaks correlated with m/z 931.04 Da (charge state + 8; peptide mass 7440.25) and m/z 984.88 (charge state + 7; peptide mass 6887.10) (Fig. 4). Further analysis of the trypsin-digested bioactive well with nanoLC-MS and Mascot analysis obtained after RPLC revealed the presence of four proteins, and in the case of HILIC analysis, the number of found proteins was five. The matched proteins include the Kunitz-type serine protease inhibitor C6 from Daboia siamensis venom (protein coverage, 43%; protein score, 343) for which five peptides were found, and Kunitz-type serine protease inhibitors 3 and 4, both from Daboia russelii venom (protein coverage, 36%; protein score, 58, three matched peptides; and protein coverage, 33%; protein score, 125, one matched peptide, respectively). For the proteomic analysis of bioactive fractions collected after HILIC separation, the best match was for the co-eluting basic PLA2 VRV-PL-VIIIa (protein coverage, 87%; protein score, 2173), for which 17 peptides were found, and PLA2 VRV-PL-V (protein coverage, 71%; protein score, 1935), for which seven peptides were found. Both proteins were found to be from Daboia russelii venom. Kunitz-type serine protease inhibitor 3 was also found in the HILIC bioactive fraction, however, at a protein score of 65 and protein coverage of 20%, with a single peptide matched. Detailed results from Mascot analysis listing all protein and peptides found together with the protein and peptides scores, % coverage, observed and expected masses of the peptides can be found in ESM Table S3.</p><!><p>In this study, a nanofractionation platform was developed and applied for the screening of snake venoms for plasmin inhibitors and (metallo)proteinases with similar activity to plasmin. A 384-well format plate reader fluorescence assay for plasmin activity was optimized and implemented in the platform. The usefulness of the methodology was demonstrated by analyzing venoms of the snakes Daboia russelii and Crotalus basiliscus. In Dr venom, plasmin inhibitors were detected and identified by downstream proteomic analysis. In Cb venom, fibrin(ogen)olytic enzymes that act in a manner similar to plasmin were found. These enzymes were shown to be potentially calcium dependent proteases that exhibited similarity to known SVMP venom toxins. The method developed is a reliable and high-throughput screening tool that helps in discovering new drug leads by exposing individual proteins present in snake venoms to a bioassay mixture containing a drug target of interest. The parallel MS analysis and off-line proteomic studies that follow the bioactivity assessment allow for bioactive protein identification in a high-throughput manner. Understanding the composition of snake venoms, in particular toxins that are responsible for the life debilitating and life-threatening manifestation of snakebites, may also aid in the development of new toxin-specific anti-venoms that have already been demonstrated to have the potential to neutralize toxins in venoms of snakes species that are geographically and phylogenetically distinct [22].</p><!><p>(PDF 1.01 MB)</p>

PubMed Open Access

A dual role for Integrin \xce\xb16\xce\xb24 in modulating Hereditary Neuropathy with liability to Pressure Palsies

Peripheral myelin protein 22 (PMP22) is a component of compact myelin in the peripheral nervous system. The amount of PMP22 in myelin is tightly regulated, and PMP22 over or under-expression cause Charcot-Marie-Tooth 1A (CMT1A) and Hereditary Neuropathy with Pressure Palsies (HNPP). Despite the importance of PMP22, its function remains largely unknown. It was reported that PMP22 interacts with the \xce\xb24 subunit of the laminin receptor \xce\xb16\xce\xb24 integrin, suggesting that \xce\xb16\xce\xb24 integrin and laminins may contribute to the pathogenesis of CMT1A or HNPP. Here we asked if the lack of \xce\xb16\xce\xb24 integrin in Schwann cells influences myelin stability in the HNPP mouse model. Our data indicate that PMP22 and \xce\xb24 integrin may not interact directly in myelinating Schwann cells, however, ablating \xce\xb24 integrin delays the formation of tomacula, a characteristic feature of HNPP. In contrast, ablation of integrin \xce\xb24 worsens nerve conduction velocities and non-compact myelin organization in HNPP animals. This study demonstrates that indirect interactions between an extracellular matrix receptor and a myelin protein influence the stability and function of myelinated fibers.

a_dual_role_for_integrin_\xce\xb16\xce\xb24_in_modulating_hereditary_neuropathy_with_liability_to_pr

Introduction<!>Animal models and genotyping<!>Electrophysiology<!>Morphological assessments<!>Teasing and osmication of nerve fibers<!>Immunoelectron microscopy<!>Cell culture<!>Western blot<!>Immunoprecipitation<!>Immunohistochemistry<!>Proximity Ligation Assay<!>Quantitative real-time PCR<!>Mass spectrometry and data analysis<!>Gene ontology analysis<!>Statistical analyses<!>Ablation of integrin \xce\xb16\xce\xb24 has both beneficial and detrimental effects on the HNPP mouse model<!>Integrin \xce\xb16\xce\xb24-PMP22 do not colocalize in myelinating Schwann cells<!>Ablation of integrin \xce\xb16\xce\xb24 increases PMP22 protein levels<!>Identification of novel partners of \xce\xb16\xce\xb24 integrin to further elucidate the functional interaction with PMP22<!>Discussion

<p>Peripheral myelin protein 22 (PMP22) is a constituent of peripheral myelin (Welcher et al. 1991). Defects in PMP22 gene dosage cause the most common inherited diseases of peripheral nerves, which includes Hereditary Neuropathy with liability to Pressure Palsies (HNPP), due to deletion of a genetic region containing the PMP22 allele (Chance et al. 1993). Mice haploinsufficient for PMP22 recapitulate human HNPP pathology, that includes the formation of tomacula (smooth thickening of myelin typically at paranodes), progressive demyelination and failure of action potential propagation (Adlkofer et al. 1995, Adlkofer et al. 1997, Bai et al. 2010). Although PMP22 was identified 25 years ago, the function of PMP22 in myelin remains mysterious (Snipes et al. 1992). PMP22 may participate in actin-mediated cellular functions and in establishing lipid rafts (Lee et al. 2014). Other hypotheses regarding its function include compact myelin stabilization (D'Urso et al. 1999, Hasse et al. 2004), the promotion of Schwann cell differentiation, proliferation or apoptosis (Fabbretti et al. 1995, Amici et al. 2007, Sancho et al. 2001), calcium homeostasis (Nobbio et al. 2009) and maintenance of myelin permeability (Guo et al. 2014, Hu et al. 2016). β4 integrin interacts with the α6 integrin subunit to form a laminin receptor that provides stability and mechanical support to myelinated fibers (Feltri et al. 1994, Nodari et al. 2008, Quattrini et al. 1996),</p><p>During early stages of myelination, PMP22 and α6β4 integrin are co-expressed in Schwann cells and can be co-immunoprecipitated (Amici et al. 2006), but while PMP22 is present in compact myelin, α6β4 integrin is on the outer layer of the Schwann cell membrane. Therefore, it is uncertain how and where the interaction between PMP22 and α6β4 integrin may occur in nerves. Finally, deletion of Pmp22 or β4 integrin causes similar pathological features, but with drastically different age of onset, as Pmp22 haploinsufficient animals develop tomacula between 10 and 20 days of age (Adlkofer et al. 1995, Adlkofer et al. 1997) and mice with β4 integrin deletion in Schwann cells develop myelin folding and demyelination at 12 months (Nodari et al. 2008, Quattrini et al. 1996).</p><p>To elucidate the significance of the interaction between PMP22 and α6β4 integrin, we performed careful co-localization and biochemical analysis and identified novel interactors of the β4 integrin subunit. To ask if the lack of α6β4 integrin in Schwann cells influences myelin stability in the HNPP model, we generated mice lacking both PMP22 and β4 integrin in Schwann cells. We find that ablation of β4 integrin in HNPP does not modify myelin thickness or internodal length and does not worsen abnormal myelin folding, but worsens nerve conduction velocities in HNPP animals, and this correlates with worsening of the paranodal and internodal disorganization. These data indicate that integrin α6β4 and PMP22 cooperate to organize the internode and to allow proper propagation of the axonal signal along the axons.</p><!><p>All animal procedures were performed in accordance with San Raffaele Institute and University at Buffalo animal care committee regulations (n. 363 and UB1188M respectively). Itgb4 floxed mice (RRID:MGI:3804178), Pmp22 knock-out mice (RRID:MGI:3794448) and P0-Cre transgenic mice (RRID:IMSR_JAX:017927) have been described previously (Feltri et al. 1999, Adlkofer et al. 1995, Nodari et al. 2008). Animal's references can be retrieved with RRIDs at http://scicrunch.org/resources. Animals used as controls are Pmp22 +/+; Itgb4 fl/fl; P0-cre negative (Pmp22 +/+; Itgb4 SC +/+). Pmp22 +/− animals are Pmp22 +/−; Itgb4 fl/fl; P0-Cre negative (Pmp22 +/−; Itgb4 SC +/+). Itgb4 SC −/− animals are Itgb4 fl/fl; P0-Cre positive (Pmp22 +/+; Itgb4 SC −/−). Finally, Pmp22 +/−; Itgb4 SC −/− animals are Pmp22 +/−; Itgb4 fl/fl; P0-Cre positive (Pmp22 +/−; Itgb4 SC −/−). All mice were backcrossed to C57BL/6 to reach congenic background. Both males and females were included in the study. Mutant and control littermates were sacrificed at the indicated ages, and sciatic nerves were dissected. No animals were excluded from the study. Animals were housed in cages of no more than 5 animals in 12 h light/dark cycles. Most experiments were conducted with 3 animals per age and per genotype. A flow-chart of the experimental procedures is available in Fig. S5. Genotyping of mutant mice was performed by PCR on tail genomic DNA, as described previously (Feltri et al. 1999, Adlkofer et al. 1995, Nodari et al. 2008).</p><!><p>Animals were analysed at 30 days of age as described previously (Poitelon et al. 2015). Briefly, mice were anesthetized with tribromoethanol, 0.02 ml g−1 of body weight, and placed under a heating lamp to avoid hypothermia. Sciatic nerve motor conduction velocity and amplitude were obtained with subdermal steel monopolar needle electrodes: a pair of stimulating electrodes was inserted subcutaneously near the nerve at the ankle, then at the sciatic notch, and finally at the paraspinal region at the level of the iliac crest to obtain three distinct sites of stimulation, proximal and distal, along the nerve. Compound motor action potential were recorded with an active electrode inserted in muscles in the middle of the paw and a reference needle in the skin between the first and second digits.</p><!><p>Mutant and control littermates were euthanized at the indicated ages, and sciatic nerves were dissected. Tissues were fixed in 2% buffered glutaraldehyde and post fixed in 1% osmium tetroxide. After alcohol dehydration, the samples were embedded in Epon. Transverse sections (0.5 – 1 nm thick) were stained with toluidine blue and examined by light microscopy. Morphological measurements were quantified from anatomically comparable whole cross-section of sciatic nerve cut halfway between the sciatic notch and the knee. For G-ratio analysis of sciatic nerves (axon diameter/fiber diameter), 4 images per semithin section were acquired at the 100x objective. Axon and fiber diameters were quantified using the Leica QWin software (Leica Microsystem). Myelinated fibers and tomacula were quantified on the whole cross-section of sciatic nerve using ImageJ (imagej.nih.gov/ij). Data were analysed using GraphPad Prism 6.01. Images were quantified blindly.</p><!><p>Sciatic nerves were fixed in 2% glutaraldehyde overnight at 4 °C and washed 3 times in phosphate buffer (79 mM Na2HPO4, 21 mM NaH2PO4, pH 7.4). Sciatic nerves were stained in 1 % osmium, washed 4 times in phosphate buffer and incubated at 55 °C during 12 hours in 30% glycerol followed by 12 hours in 60% glycerol and 12 hours in 100% glycerol.</p><!><p>Immuno-electron microscopy analyses were performed as in (Quattrini et al. 1996) but with embedding modified as described in (Yin et al. 2000). Primary antibody to PMP22 (Abcam, ab61220, RRID:AB_944897), and 5 nM of gold conjugated secondary antibody were used. Immuno-electron microscopy analysis was performed once.</p><!><p>Rat 804G (RRID:CVCL_J122) and hamster CHO (RRID:CVCL_0213) cells were maintained in DMEM (Gibco), 10% FCS, penicillin and streptomycin. 804G permanently transfected with β4 integrin mutants were a gift from Dr. Giancotti (Spinardi et al. 1993). Transfection of HA-Pmp22 and β4 integrin were performed using Lipofectamine 2000 (Invitrogen), according to the manufacturer's instructions. Cells were analyzed 72h after transfection to insure efficient expression of the transfected plasmids. Each transfection was repeated at least three times.</p><!><p>Sciatic nerves were dissected, removed from their epineurium, frozen in liquid nitrogen then pulverized. Transfected hamster CHO cells (RRID:CVCL_0213) and sciatic nerves were resuspended in lysis buffer (95 mM NaCl, 25 mM Tris-HCl pH 7.4, 10 mM EDTA, 2% SDS, 1 mM Na3VO4, 1 mM NaF and 1:100 Protease Inhibitor Cocktail, Roche). Protein lysates were incubated at 4 °C for 30 min then centrifuged at 16,000 rpm for 30 min at 4°C. The concentration of the protein supernatants was determined by BCA protein assay (Thermo Scientific) according to manufacturer's instructions. Equal amounts of homogenates were diluted 3:1 in 4 X Laemmli (250 mM Tris-HCl pH 6.8, 8% SDS, 8% β-Mercaptoethanol, 40% Glycerol, 0.02% Bromophenol Blue), denatured 5 min at 100 °C, resolved on SDS-polyacrylamide gel, and electroblotted onto PVDF membrane. Blots were then blocked with 5% BSA in 1 X PBS, 0.05% Tween-20 and incubated over night with the appropriate antibody. Abcam anti-PMP22 (ab61220, RRID:AB_944897), Cell signaling anti-PAK1 (2602, RRID:AB_330222), GenWay anti-p-PAK1 (GWB-961E2C, RRID:AB_10275529), Millipore anti-integrin β1 (MAB1997, RRID:AB_2128202), anti-Itpr3 (AB9076, RRID:AB_571029), Novocastra anti-β-dystroglycan (NCL-b-DG, RRID:AB_442043), Roche anti-HA (11867423001, RRID:AB_390918), Santacruz anti-integrin β4 human (sc-9090, RRID:AB_2129021), anti-integrin α6 (sc-6597, RRID:AB_2128041) and Sigma anti-β-tubulin (T4026, RRID:AB_477577), anti-calnexin (C4731, RRID:AB_476845). Antibodies references can be retrieved with RRIDs at http://scicrunch.org/resources. Anti-integrin β4 mouse/rat (lot 2213) was gift of Dr. Brophy, Centre for Neuroregeneration, Univ. of Edinburgh). Membranes were rinsed in 1 X PBS and incubated for 1h with secondary antibodies. Blots were developed using ECL, ECL plus (GE Healthcare) or Odyssey CLx infrared imaging system (Li-Cor). Western blots were quantified using Image J software (http://imagej.nih.gov/ij). Each Western blot was repeated at least three times.</p><!><p>Rat 804G cells and sciatic nerves fibers were resuspended in lysis buffer (150 mM NaCl, 50 mM Tris-HCl pH 7.4, 0.5% Sodium deoxycholate, 1% NP-40, 1:100 Protease Inhibitor Cocktail, Roche). For each immunoprecipitation, 500 μg of proteins were incubated with 50 μl of Protein G Sepharose for 1 hour at 4 °C on a rotating wheel. After a short centrifugation, supernatants were transferred to a new vial, incubated overnight at 4 °C with 1 μg of antibody (Abcam anti-Integrin β4 (ab25254, RRID:AB_2129042) or Roche anti-HA (11867423001, RRID:AB_390918)) on a rotating wheel. 50 μl of Protein G Sepharose were added to the mix and incubated for an additional 3 hours. After centrifugation, supernatants were collected as unbound fraction (Ub), pellets were washed three time in lysis buffer, resuspended in Laemmli and processed as Western blot as the immunoprecipitated fraction (IP). Equal amounts of proteins were incubated with the target antibody, precipitated, separated by SDS/PAGE and probed. Each immunoprecipitation was repeated at least three times.</p><!><p>We tested various fixation and anti-PMP22 antibodies against nerves from Pmp22 −/− animals at various ages. Surprisingly, numerous antibodies anti-PMP22 (Abcam ab126769, RRID:AB_11129961; Assay Biotech C0306, RRID:AB_10686127; LifeSpan, LS-C118559-100, RRID:AB_10796545; LS-C122324-100, RRID:AB_10799898; Santacruz sc-58572, RRID:AB_785236; sc-18535, RRID:AB_2167000; sc-71911, RRID:AB_2167001; sc-65739, RRID:AB_2167002; Sigma-Aldrich SAB4502217, RRID:AB_10746275) gave positive stainings in all conditions tested even in null tissues, and were not further pursued. Only one anti-PMP22 (ab61220, RRID:AB_944897) showed no PMP22 signal in SCs of PMP22 −/− animals (Fig. S4). The antibodies references can be retrieved with RRIDs at http://scicrunch.org/resources. The following conditions gave no background staining for PMP22 in Pmp22 −/− null animals (data not shown). Unfixed sciatic nerve sections and sciatic nerve teased fibers were fixed with cold methanol for 1 min or cold acetone for 10 min, washed in 1 X PBS, blocked for 1 h in blocking solution (5% Fish skin gelatin, 0.5% Triton X-100, 1 X PBS), then incubated overnight with the following antibodies: Abcam anti-Integrin β4 (ab25254, RRID:AB_2129042), anti-PMP22 (ab61220, RRID:AB_944897), Alomone labs anti-Kv1.1 (APC-009, RRID:AB_2040144), Covance anti-Neurofilament M (PCK-593P, RRID:AB_663194), Millipore anti-Integrin β1 (MAB1997, RRID:AB_2128202), Invitrogen Rhodamine Phalloidin (R415, RRID:AB_2572408), Santacruz anti-Integrin α6 (sc-6597, RRID:AB_2128041), Sigma anti-Itpr3 (AB9076, RRID:AB_571029), anti-Claudin-19 was a gift by Dr. Shoichiro Tsukita and Dr. Mikio Furuse. Slides were rinsed in PBS, incubated 1 h with Jackson DyLight 488 or 549-conjugated secondary antibodies, stained with DAPI, and mounted with Vectashield (Vector Laboratories). Images were acquired with an upright microscope Leica DM5000, a Zeiss Apotome or a confocal microscope Leica SP5II. Each immunochemistry was repeated at least three times.</p><!><p>Unfixed sciatic nerve sections were fixed in cold methanol for 1 min, washed in in 1 X PBS then blocked for 1 h in blocking solution (5% Fish skin gelatin, 0.5% Triton X-100, 1 X PBS). Slides were then incubated overnight with the following antibodies: Abcam anti-Integrin β4 (ab25254, RRID:AB_2129042) and anti-PMP22 (ab61220, RRID:AB_944897) or Millipore anti-Neurofilament M (MAB-10651, RRID:AB_2150076) and anti-Neurofilament H (AB1989, RRID:AB_91202). MINUS probe was coupled to anti-Integrin β4 using In Situ Probemaker (Sigma). The following procedures were carried out using the Duolink starter kit (Sigma) products. PLA probes PLUS (rabbit) and MINUS (mouse) were diluted 1:5 in blocking solution and incubated 1 h at 37 °C. Slides were washed twice in Wash Buffer A for 2 min. Ligase and ligation buffer were diluted 1:40 and 1:5 were diluted in 1 X PBS, and incubated on slides at 37°C for 30 min. Slides were washed twice in Wash Buffer A for 2 min. Polymerase and amplification buffer were diluted 1:80 and 1:5 in 1 X PBS, and incubated on slides at 37°C for 120 min. Slides were washed three time in Wash Buffer B for 10 min. Cells were stained with DAPI, and mounted with media supplied by the kit. Images were acquired with an upright microscope Leica DM5000. The proximity ligation assay was performed twice.</p><!><p>For quantitative real-time PCR (qPCR), we sampled sciatic nerves from control and mutant mice at P5, P10, P30 and P90. Sciatic nerves were frozen in liquid nitrogen, crushed with a metallic pestle and total RNAs were isolated from sciatic nerve using TRIzol (Life Technologies). First strand cDNA was prepared from 1 μg or RNA using Superscript II, 50 μM oligo(dT)20 and 50 ng random hexamers, according to the manufacturer's instructions. qPCR was performed using Sybr Green (Roche) in a LightCycler 480 System (Roche) according to standard protocols. Normalization was performed using 18S rRNA as a reference gene. The relative standard curve method was applied using wild-type mice as reference. At least 3 animals were used for each genotype at each age. Primers were designed with the Roche Primer Design Center (http://qpcr.probefinder.com/organism.jsp) and their efficiency was assessed by standard curve. Only primers 90% efficient or above were selected. Primer dimerization was tested in vitro by Amplify X (http://engels.genetics.wisc.edu/amplify) and the homogeneity of the PCR products was assessed by a dissociation curve. We used the following primers to amplify 18S: 5′ ctcaacacgggaaacctcac 3′ and 3′ cgctccaccaactaagaacg 5′, Egr2: 5′ ctacccggtggaagacctc 3′ and 3′ aatgttgatcatgccatctcc 5′ Itga6: 5′ cctgaaagaaaataccagactctca 3′ and 3′ ggaacgaagaacgagagagg 5′, Itga7: 5′ agaaggtggagcctagcaca 3′ and 3′ gctgaacaccacacacttgg 5′, Itgb1: 5′ caaccacaacagctgcttctaa 3′ and 3′ tcagccctcttgaattttaatgt 5′, Itgb4: 5′ cttggtcgccgtctggta 3′ and 3′ tcgaaggacactaccccact 5′, Mpz: 5′ gctgccctgctcttctctt 3′ and 3′ tttccctgtccgtgtaaacc 5′, Pmp22: 5′ ccgtccaacactgctactcc 3′ and 3′ cgctgaagatgacagacagg 5′. Each quantitative real-time PCR was repeated twice.</p><!><p>Immunoprecipitated eluates were separated by 4–12% SDS–PAGE, stained with Coomassie Brilliant Blue (Bio-Rad) and excised in eight slices for LC-MS/MS analysis. Mass spectrometry analysis was performed by LC-MS/MS using an LTQ-Orbitrap mass spectrometer (ThermoScientific, Bremen, Germany). Tryptic digests for each band were first cleaned using Stage Tips as described previously (Rappsilber et al. 2007) and then injected in a capillary chromatographic system (EasyLC, Proxeon Biosystems, Odense, Denmark). Peptide separations occurred on a homemade column obtained with a 10-cm fused silica capillary (75 μm inner diameter and 360 μm outer diameter; Proxeon Biosystems) filled with Reprosil-Pur C18 3 μm resin (Dr Maisch GmbH, Ammerbuch-Entringen, Germany) using a pressurized 'packing bomb'. A gradient of eluents A [distilled water with 2 % (v v−1) acetonitrile, 0.1% (v v−1) formic acid] and B [acetonitrile, 2% (v v−1) distilled water with 0.1% (v v−1) formic acid] was used to achieve separation from 8% B (at 0 min, 0.2 ml/min flow rate) to 50% B (at 80 min, 0.2 ml min−1 flow rate). The LC system was connected to the orbitrap equipped with a nanoelectrospray ion source (Proxeon Biosystems). Full-scan mass spectra were acquired in the LTQ-Orbitrap mass spectrometer in the mass range m z−1 350–1500 Da and with the resolution set to 60000. The 'lock-mass' option was used for accurate mass measurements. The 10 most intense doubly and triply charged ions were automatically selected and fragmented in the ion trap. Target ions already selected for the MS/MS were dynamically excluded for 60 s (Olsen et al. 2005). Protein identification and quantification were achieved using the MaxQuant software version 1.3.0.5 (Cox & Mann 2008). Cysteine carbamidomethylation was searched as a fixed modification, whereas N-acetyl protein and oxidized methionine were searched as variable modifications. Mass spectra were analyzed by Andromeda plugin in MaxQuant using UniProt complete proteome Mus musculus 2013 database. Protein quantification was based on LFQ intensities. Peptides and proteins were accepted with a false-discovery rate of 0.01, two minimum peptides identified per protein of which one unique. The experiments were done in biological duplicate performing two technical replicates.</p><!><p>Gene ontology (GO) clustering analysis was performed using the Cytoscape (http://chianti.ucsd.edu/cyto_web/plugins/index.php) plugin Biological Network Gene Ontology (BiNGO) (http://www.psb.ugent.be/cbd/papers/BiNGO/index.html). The degree of functional enrichment for a given cluster and category was assessed quantitatively (P value) by hypergeometric distribution, a multiple test correction was applied using the false discovery rate (FDR) algorithm, fully implemented in BiNGO software. Overrepresented biological process categories were generated after FDR correction, with a significance level of 0.05. The full analysis is listed in Data S3.</p><!><p>Experiments were not randomized, but data collection and analysis were performed blindly to the conditions of the experiments. Researchers blinded to the genotype performed morphological analyses, nerve conduction velocities and morphometric analyses. The data obtained are presented as mean ± s.e.m. No statistical methods were used to predetermine sample sizes, but our sample sizes are similar those generally employed in the field. t-test and One-way ANOVA with Bonferroni's multiple comparisons test were used for statistical analysis of the differences among multiples groups according to the number of samples. Values of P ≤ 0.05 were considered to represent a significant difference. This study was not preregistered.</p><!><p>To test for functional interactions between α6β4 integrin and PMP22, we asked whether α6β4 integrin in Schwann cells modifies the timing or severity of myelin abnormalities associated with HNPP, exploiting an authentic model of Pmp22 loss-of-function (Adlkofer et al. 1995, Bai et al. 2010) and the conditional, Schwann cell specific, knock-out for integrin β4 (Nodari et al. 2008, Feltri et al. 1999). We generated double mutant (Pmp22 +/− and Itgb4 SC −/−) animals and analyzed myelination in the peripheral nervous system.</p><p>Analysis of Pmp22 +/− nerves by semithin sections from 20 to 90 days of age confirmed previous reports that haploinsufficiency of Pmp22 does not cause axonal loss or early demyelination (Fig. 1a–d), but promotes the formation of tomacula (Fig. 1a arrows, Fig. 1g open arrows) and a reduction in the length of myelin internodes (Fig. 1g–i) (Adlkofer et al. 1995, Amici et al. 2007, Adlkofer et al. 1997, Hu et al. 2016). Surprisingly, we observed that ablation of integrin α6β4 slightly delays the appearance of tomacula (Fig.1e, f). A significant reduction in the number of tomacula can be observed at 20 days of age (P20) while myelination is ongoing, but disappears at P200 once myelin is mature (Fig. 1e). We also noticed a moderate increase in the length of myelin segments (internodes) in double mutants as compared to Pmp22 +/− animals at P30 (Fig. 1g–i). The increase is significant for internodes of small caliber fibers (Fig. 1i). Together these data indicate that ablation of integrin α6β4 delays the onset of morphological defects caused by Pmp22 haploinsufficiency.</p><p>In contrast, electrophysiological analyses in 1-month old animals showed that ablation of α6β4 integrin worsens the nerve conduction velocities (NCV) of HNPP animals (Table 1). This was not due to a change in internodal length, which was increased in double mutants only in small myelinated fibers which do not contribute to NCVs (Li 2015). These data indicate that integrin α6β4 and PMP22 collaborate to allow proper propagation of the nervous signal along axons, in a way that is independent of myelin thickness (Fig. 1b) and of internodal length.</p><p>Overall, ablation of integrin α6β4 affects positively the HNPP phenotype by reducing the number of tomacula and restoring internodal length in small calibre fibers, but it also reduces NCV. While these observations are conflicting, they likely represent different pathogenic mechanisms of Pmp22 haploinsufficiency, which are modulated differently by integrin α6β4.</p><!><p>Because integrin α6β4 is present at the outer Schwann cell membrane (Feltri et al. 1994, Einheber et al. 1993), while PMP22 is localized in compact myelin (Haney et al. 1996) (see also Fig. S1), we hypothesized that the two proteins could interact in mature myelinated fibers in trans. We confirmed that integrin α6β4 and PMP22 can be co-immunoprecipitated from lysates of sciatic nerves at P30, as reported (Amici et al. 2006), and that this interaction is lost in PMP22-null or β4 integrin SC-null mice (Fig. 2a). However, using cell lines expressing various truncated forms of β4 integrin and transfected with an HA-PMP22 plasmid, we found that all truncated forms of β4 integrin appear to bind PMP22 (Fig. 2c). These data indicate that the two proteins do not interact in trans, but they could interact in cis via the only domain of β4 integrin that was not deleted, namely the transmembrane domain (i.e. amino acid 661–853) (Fig. 2d). From three distinct immunohistochemical approaches (i.e. proximity ligation assay, immunohistochemistry on sciatic nerve cross sections and on teased fibers), we showed that integrin α6β4 does not colocalize with PMP22 in the outer layer of the Schwann cell nor in compact myelin at P30 or as early as P3 (Fig. 2e–h, Fig S2, Data S2).</p><p>The discordance between the presence of co-immunoprecipitation from sciatic nerve lysates, the absence of colocalization and the persistence of co-immunoprecipitation in clones lacking most of the β4 integrin domains, may indicate that the interaction between integrin α6β4 and PMP22 is artefactual, and occurs in cell lysates, but not in vivo. This alternative explanation is supported by the observation that the interaction between β4 integrin and PMP22 can occur in vitro, after solubilization and mixing of the two proteins from two independent lysates, as shown in Fig. 2b. Overall, because we could not identify a specific β4 integrin domain of interaction with PMP22, nor could co-localize the two proteins in myelinating Schwann cells, we conclude that the two molecules probably do not physically interact in myelinating Schwann cells in vivo.</p><!><p>Even if PMP22 and integrin α6β4 do not interact directly, the modification of the HNPP phenotype observed in Pmp22 +/−; Itgb4 SC −/− mice supports the existence of a functional interaction between the two proteins. We thus asked if ablation of β4 integrin affects the expression of Pmp22. We did not observe an effect on the mRNA levels of Pmp22 or the myelin related genes Egr2 and Mpz (Fig. 3a). However, the protein levels of PMP22 were modestly increased by 25% in Pmp22 +/+; Itgb4 SC −/− and in double mutant animals (Fig. 3b). Conversely, it was shown that absence of PMP22 could affect the expression and organization specifically of the α6β4 integrin basal lamina receptor (Amici et al. 2006, Amici et al. 2007). Thus, we asked if Pmp22 haploinsufficiency could affect the expression of α6β4 integrin or other basal lamina receptors. However, we did not observe any difference in the mRNA, protein levels, or localization of any laminin receptors in HNPP sciatic nerves (Fig. S3).</p><p>NCV defects can be due to thinner myelin, shorter internodes, smaller axons, alterations in the electrical properties of myelin due to imbalanced lipid composition or abnormal paranodal axoglial junctions (Coetzee et al. 1996, Miyamoto et al. 2005, Dupree et al. 1998, Sherman et al. 2005, Court et al. 2004). We showed that there is no overall decrease in the number of myelinated fibers, internodal length or myelin thickness in Pmp22 +/−; Itgb4 SC −/− mice (Fig. 1). Therefore, it is possible that modifications of the NCVs are associated with other morphological features such as the electrical properties of the myelin sheath. Recently studies have reported that deficiency of Pmp22 in HNPP causes abnormal myelin permeability, also called "functional demyelination" (Guo et al. 2014, Hu et al. 2016). According to this model, the absence of Pmp22 affects the formation of tight and adherens autotypic junctions during development, and causes an increase of p21-activated kinase (PAK1) (Hu et al. 2016, Guo et al. 2014). The combined haploinsufficiency of PMP22 with increased PAK1 levels then leads to a further disruption of autotypic junctions and of F-actin organization in animals older than 3-months (Hu et al. 2016). It is possible that the absence of α6β4 integrin aggravates or accelerates the defects associated with myelin permeability in HNPP nerves. To test this hypothesis, we analyzed the localization of F-Actin at Schmidt-Lanterman incisures, Claudin-19 at paranodes and Kv1.1 at juxtaparanodes. In 1-month old sciatic nerves from Itgb4 SC −/−; Pmp22 +/− animals, we did not observe mislocalization of these proteins (Fig. 4a–b, d). In 3-month old animals, we confirmed that staining for F-Actin was increased in Schmidt-Lanterman incisures of PMP22 mutants (Fig. 4c, e), and ablation of β4 integrin alone did not affect the F-Actin localization. However, F-actin staining was increased in Schmidt-Lanterman incisures of PMP22 and double mutants at 3-months (Fig. 4e) and F-actin was more disorganized in numerous internodes of double mutants, both at 1-month and 3-months of age (Fig. 4c, white lines, Fig. 4f). We also analyzed the activation of PAK1 at 15 days and 3 months of age. We confirmed that p-PAK1 levels are increased in Pmp22 +/− mutants, by 30% at P15 and by 60% at P30 (Hu et al. 2016), and this appeared to be amplified in double mutants (Fig. 4g). Thus, ablation of Itgb4 may aggravate some of the defects in compact myelin organization observed in Pmp22 mutants.</p><!><p>To obtain more insights in the functional interaction between α6β4 integrin and PMP22, we used mass spectrometry to identify novel proteins that immunoprecipitate with β4 integrin from sciatic nerve lysates at 30 days of age (Fig. 5a). From two independent experiments, we identified 73 candidate partners for β4 integrin. All these partners were significantly enriched in immunoprecipitates of wild-type as compared to Itgb4 SC −/− nerves (Data S3). Among the 73 proteins, we identified several known partners of β4 integrin, including α6 integrin, vimentin, plectin and the laminin β1 subunit (Rezniczek et al. 1998, Homan et al. 2002, Lotz et al. 1990, Sonnenberg et al. 1988) (Data S3). Interestingly, we identified periaxin as a novel interactor of β4 integrin (Data S3). Both Prx −/− and Pmp22 +/− sciatic nerves develop similar tomacula and manifest decreased NCVs (Court et al. 2004, Gillespie et al. 2000), making it a candidate to functionally link β4 integrin and PMP22. However, the protein levels and localization of periaxin were normal in Pmp22 +/−; Itgb4 SC −/− animals (data not shown).</p><p>For an overview of the molecular functions of the β4 integrin network based on this list of proteins, we used BiNGO, a software which recognizes network linkage by gene ontology (GO) hierarchy. The most highly represented terms in the molecular function GO category were binding and protein binding (Fig. 5b and Data S3). Consistent with previous reports, β4 integrin interacts with several cytoskeletal and actin-binding proteins (Fig. 5b). This could potentially explain the slight disruption of F-actin organization that we observed in Itgb4 SC −/−; Pmp22 +/− internodes.</p><p>A careful examination of the map shows a statistically significant enrichment of interactors linked to calcium binding and calcium transport GO categories (Fig. 5b, red underlines). Specifically, we identified two novel Ca2+ transporters as β4 integrin partners, namely the plasma membrane calcium-transporting ATPase 4 (Atp2b4) and inositol 1,4,5-trisphosphate receptor type 3 (Itpr3) (Fig. 5b). ITPR3 is an inositol triphosphate receptor that controls intracellular Ca2+ concentrations. ITPR3 is localized in the endoplasmic reticulum, but has also been detected at the plasma membrane (Kuno & Gardner 1987) and in the nucleus (Nicotera et al. 1990). In Schwann cells, ITPR3 is localized at the nodal/paranodal region and in Cajal bands (Toews et al. 2007, Martinez-Gomez & Dent 2007). Of note, PMP22 may regulate intracellular Ca2+ concentration, and Nobbio et al. have proposed that PMP22 neuropathies could be caused by calcium dysregulation (Nobbio et al. 2009). To confirm that ITPR3 interacts with α6β4 integrin, we immunoprecipitated β4 integrin from wild-type nerves, and verified that ITPR3 and α6 integrin co-immunoprecipitation by western blot (Fig. 5c, left panel). In addition, the ITPR3 band was almost completely absent when the immunoprecipitation was performed from Itgb4 SC-null nerves, which retained very low levels of β4 integrin protein, likely deriving form perineurial cells of vessels (Fig. 5c, right panel). In contrast, some Atp2b4 could still be immunoprecipitated with β4 integrin from Itgb4 SC-null nerves (not shown) and we did not pursue Atp2b4 any further. We next examined the localization of ITPR3 in our mutants. In our hands, ITPR3 localized mainly to paranodal regions of wild-type nerves (Fig. 5d). α6β4 is present along all the outer Schwan cell membrane, including above the paranodal regions (Fig. 5d), suggesting that the two proteins co-localize and could interact in this location. Interestingly, we observed a reduction of the paranodal staining for ITPR3 in Itgb4 SC-null nerves that became more evident in Pmp22 +/−; Itgb4 SC −/− double mutants (Fig. 5e–f). Thus, ablation of Itgb4 alters the paranodal localization of ITPR3. Finally, we asked if this mislocalization of ITPR3 affects the concentration of calcium in Schwann cells. We performed a fluorometric determination of the intracellular Ca2+ concentration in Schwann cells co-cultured with DRG neurons from Pmp22 +/+; Itgb4 SC +/+ and Pmp22 +/−; Itgb4 SC −/− animals, but we did not observe any difference (data not shown).</p><!><p>There is evidence that PMP22 maintains myelin integrity, and part of PMP22 function may be mediated through its physical interaction with β4 integrin (Amici et al. 2006). To address the role of α6β4 integrin in PMP22 neuropathies, we genetically ablated β4 integrin in Schwann cells together with PMP22. We demonstrate that integrin α6β4 influences nerve conduction velocity in HNPP animals without affecting myelin thickness but possibly by enhancing the abnormalities in the architecture of myelinated fibers. Furthermore, a6β4 integrin delays the progression of myelin abnormalities in HNPP. Finally, we provide several lines of evidence that the effects of β4 integrin on HNPP pathophysiology are probably indirect, because integrin β4 and PMP22 do not colocalize in myelinating Schwann cells.</p><p>The role of laminin-integrin signaling in Schwann cell development has been defined in a large body of work–for review see (Feltri et al. 2016, Monk et al. 2015). Among laminin receptors, the function of α6β4 integrin is the most elusive and subtle. α6β4 integrin is localized at the abaxonal surface of myelinating SCs, where it regulates myelin maintenance, as its absence cause an anticipated appearance of myelin outfoldings at paranodes and juxtaparanodes, that is significantly worsened by additional ablation of another laminin receptor, dystroglycan (Nodari et al. 2008). Ablation of α6β4 integrin in Schwann cells also causes disorganization of Kv1.1 in the internodal mesaxon (Nodari et al. 2008), and of F-actin in HNPP non-compact myelin (this paper). This, together with the identification of a new candidate partner, ITPR3, enriched in the paranodal region of Schwann cells and with the worsened nerve conduction velocities of HNPP animals lacking β4 integrin, suggest a more general function for integrin α6β4 broadly related to the organization of the architecture of myelinated fibers. Similarly, laminins 211 and 511 are enriched in the nodal region (Occhi et al. 2005) and the laminin receptor dystroglycan is important for the organization of Schwann cell microvilli and the localization of axonal Na+ channels (Saito et al. 2003, Occhi et al. 2005, Colombelli et al. 2015). We postulate that the slight disorganization of non-compact myelin domains in internodes, Schmidt-Lanterman incisures and paranodes caused by ablation of Itgb4 in Schwann cells may exacerbate the functional defects of HNPP, where 35 – 45 % of the nodes have abnormal myelin foldings, leading to a worsening of NCVs.</p><p>Myelin instabilities such as outfoldings and tomacula may derive from a common mechanism (Goebbels et al. 2012). The results of our study on PMP22 tomacula and α6β4 integrin outfoldings could be viewed as conflicting with this hypothesis, as we do not observe a worsening of myelin instabilities in animals lacking both proteins. However, we detected a novel interaction of integrin α6β4 with an inositol triphosphate receptor, ITRP3. Phosphoinositol and AKT signaling have been associated with myelin instabilities (Pereira et al. 2012). Thus, further studies will need to focus on the potential role of phosphoinositol receptors at nodes of Ranvier, in the regulation of Ca2+ concentration at the paranodes, and in the maintenance of myelin stability.</p>

PubMed Author Manuscript

Binary and ternary binding affinities between exonuclease-deficient Klenow fragment (Kf-exo-) and various arylamine DNA lesions characterized by surface plasmon resonance

We used surface plasmon resonance (SPR) to characterize the binding interactions between exonulease-free Klenow fragment (Kf-exo-) and unmodified and modified dG adducts derived from arylamine carcinogens: fluorinated 2 aminofluorene (FAF), 2-acetylaminofluorene (FAAF), and 4-aminobiphenyl (FABP). Tight polymerase binding was detected with unmodified dG and the correct dCTP. The discrimination of correct versus incorrect nucleotides was pronounced with KD values in order of dCTP << dTTP < dATP < dGTP. In contrast, minimal selectivity was observed for the modified templates with Kf-exo- binding tighter to the FAAF (koff: 0.02s-1) and FABP (koff: 0.01s-1) lesions than to FAF (koff: 0.04s-1).

binary_and_ternary_binding_affinities_between_exonuclease-deficient_klenow_fragment_(kf-exo-)_and_va

<p>DNA is under constant assault by various endogenous and exogenous pathways, which result in different types of DNA damage. When a polymerase encounters a lesion, it can by-pass by replicative polymerase, either inserting the correct base (error-free) or incorrect base (error-prone)1. The environmental arylamine carcinogens are known to form C8-substituted dG adducts in vivo. We have shown that these lesions exist in a mixture of the base displaced stacked, major groove B-type, and wedge conformers, with each leading to potentially unique sequence-dependent mutation and nucleotide excision repair outcomes2.</p><p>It is important to understand the nature of interactions between polymerase and DNA lesions. Crystal structure and kinetic analyses have been used to elucidate details of polymerase action at an atomic resolution1. However, similar structural details of bulky DNA lesions have been challenging due to difficulties with obtaining crystals3,4. Consequently, various alternative techniques such as fluorescence, circular dichroism spectroscopy, gel mobility shift assays, and footprinting assays have been used5,6. However, these techniques are either qualitative or semi-quantitative, non-compatible with fast dissociation rates, and require labeling of at least one of the components of interest. Although gel-based assays are relatively simple and robust, samples of interest will not be in chemical equilibrium and the system's components are not amenable to testing across temperatures or salt concentrations6. Surface plasmon resonance (SPR) is a chip based, label free solution technique that allows real-time monitoring of binding interactions between DNA and proteins5-8.</p><p>In this report, a SPR study was conducted to examine-polymerase interactions of dG lesions derived from three fluorinated prototype arylamine carcinogens: 2-aminofluorene (FAF), 2-acetylaminofluorene (FAAF), and 4-aminobiphenyl (FABP) (Fig. 1c). We employed exonuclease-free E. coli DNA polymerase I Klenow fragment (Kf-exo- ) as it avoids complication of proofreading activity. The features of fluorinated arylamines as effective conformational probes are well documented2,9. The present study takes advantage of the sensitivity of Biacore T200 to conduct SPR analysis of the binary and ternary polymerase complexes of bulky carcinogen-DNA adducts.</p><p>Figure 1a and S1a show the construction scheme for a biotinylated hairpin-based template-primer strand on a gold sensor chip. The hairpin-DNA was used to improve stability of oligonucleotides during performance of kinetics experiments. Arylamine-modified 31-mer oligonucleotides were purified by HPLC and characterized by mass spectrometry (Fig. S2). The biotin-hairpin-template/primer strands were annealed, ligated, and purified by denaturing polyacrylamide gel (Fig. S1b). The incorporation of dideoxythymidine (ddT) was carried out using Kf-exo- and the 3′ terminal ddT allowed capture of the ternary polymerase/template-primer/dNTP complex without primer extension.</p><p>The kinetic assays were optimized with respect regeneration buffer, surface density, and surface testing, as described elsewhere10 (Fig. S3). The binding kinetics analysis was performed by injecting varying amounts of Kf-exo- to cover the hairpin template-primer DNA (Fig. 1b) coated on streptavidin surface in the absence (binary) and presence (ternary) of dNTPs (100 μM). The injections were repeated three times for each concentration in random, and the resulting data were fitted to the Langmuir model (1:1) (Fig. 2a). From the fitting, binding constants (kon, koff and KD) were calculated (Table 1 and S1) using Biacore's BIAsimulation software. The Chi-squared values for the 1:1 fitting were less than 1% of Rmax(0.002–0.003 for all experiments with Rmax in the range of 0.7–3.5RU) (Figs. S4 and S5). The KD values for ternary systems were determined using affinity analysis as the association rate (kon) reaches the near-diffusion limit. This procedure allowed the monitoring of interactions between unmodified or adducted DNA with different polymerases on a single chip. Furthermore, DNA over the chip surface was found to be stable for at least 7–10 days, without loss in binding activity under buffered reaction conditions.</p><p>The results from the binding assay (Fig. S6) are summarized in Table 1. The Kf-exo- bound tightly to unmodified DNA in the presence of a correct incoming dCTP opposite the templating dG. However, relative to dCTP binding, binding tightness was reduced by 30-, 60-, 34-, and 264-fold in binary, dATP, dTTP, and dGTP, respectively (Fig. 2b and Table 1). The discrimination ability of correct versus incorrect nucleotides was significant, as the Watson-Crick base pair dCTP bound tightly and dGTP does not bind significantly. In contrast, the discrimination effect on Kf-exo- binding was weaker for binding to FAF than for binding to unmodified DNA. The specificity of binding between the correct dCTP and incorrect nucleotides, as well as for the binary system, differed by only 2- to 16-fold. The tightness of Kf-exo- binding in the presence of dCTP was reduced by 4-fold, as compared to that of the unmodified control.</p><p>Moreover, the difference in binding affinity between dCTP and dATP was less for FAF (10-fold), as compared to that of unmodified DNA (60-fold)(Fig. 2b). The Kf-exo- bound more tightly to FAAF (koff=0.02s-1) and FABP (koff=0.01s-1) lesion sites than to the unmodified control (koff=0.13s-1) while kon values are similar (Table S1). However, discrimination between correct and incorrect nucleotides was not maintained with FAAF or FABP, for which binding affinities differed by only 1- to 3-fold (Fig. 2b).</p><p>Highly specific binding of Kf-exo- to unmodified DNA in the presence of dCTP opposite a dG templating base is in line with the polymerase undergoing conformational change from an open to a closed system to form Watson-Crick base pairs11. However, Kf-exo- does bind weakly with incorrect nucleotides, probably retaining the open polymerase conformation. In particular, the binding of dGTP is very poor compared to other nucleotides.</p><p>To further confirm that the binding of polymerase to DNA is 1:1, theoretical Rmax values were calculated and compared with experimental values. The data presented here are consistent with data from sedimentation studies in which polymerase was shown to bind template-primer junction in a 1:1 ratio12. Interestingly, the KD value for Kf-exo- binding to FAF adducts was higher in the presence of dCTP than with unmodified DNA (Table 1), indicating that the lesion prevents the nucleotide-induced, catalytically-favored closed conformation. Previous studies have shown that the carcinogenic aminofluorene orients into the energetically favorable solvent-exposed major groove, which causes less disruption at the replication fork, but may perturb the groove structures and the geometry in the active site of the polymerase3.</p><p>The aforementioned crystal structure of AF on T7 DNA polymerase showed fuzzy electron densities around the carcinogenic aminofluorene moiety in line with sequence-dependent conformational heterogeneity in solution4. The present kinetics data also fit with previously published findings from a single nucleotide insertion assay study in which dATP was the next preferred nucleotide after dCTP13.</p><p>The higher binding affinity of Kf-exo- to the bulky N-acetylated FAAF lesion, compared to unmodified DNA, could be due to the adduct perturbing the template-primer junction while maintaining some specific interactions with amino acids on the active site of the polymerase. It has been shown that the AAF lesion has two hydrogen bond interactions between the N2-amino group of the modified guanine and Asp-534, as well as between the N7-guanine and Arg-5664. In addition, the lesion adopts a syn-glycosidic conformation wherein the fluorene moiety is inserted between the hydrophobic pocket of the O-helix finger subdomain. These changes also keep the polymerase in the open and maintain a distorted conformation of the subdomain fingers, causing the Tyr-530 residue to occupy the binding region of the nucleotide and preventing interaction between the incoming nucleotide and polymerase4. The present data are also in agreement with previous results from tryptic digestion studies, in which the polymerase was shown to bind very tightly to unmodified DNA in the presence of the correct nucleotide and to be insensitive to digestion; FAAF did not exhibit any additional stability in relation to the incoming nucleotide14. FAF adducts are known to exist in a sequence-dependent equilibrium of B-type and base-displaced stacked conformers2,9. FABP is similarly N-deacetylated; however, its biphenyl moiety is not as coplanar as fluorene, thereby resulting in a lesser stacked conformer population15. Consequently, FABP may behave similar to FAAF at the replication fork in the active site of a polymerase.</p><p>In summary, tight binding of Kf-exo- was observed with unmodified dG in the presence of a correct dCTP in this study. Nucleotide selectivity was pronounced with KD values in the order of dCTP << dTTP < dATP < dGTP. In contrast, minimal selectivity was observed for the modified templates: Kf-exo- bound tightly to FAAF and FABP lesions as compared to FAF. The SPR results for FAF and FAAF agreed with those obtained from gel-based assays,16 demonstrating SPR as a powerful and superior tool for studying protein/DNA interactions with bulky DNA lesions as it provides kon and koff rates.</p>

PubMed Author Manuscript

Gold(I/III)-Phosphine Complexes as Potent Antiproliferative Agents

the reaction of gold reagents [HAucl 4 •3H 2 o], [Aucl(tht)], or cyclometalated gold(iii) precursor, [c^nAucl 2 ] with chiral ((R,R)-(-)-2,3-bis(t-butylmethylphosphino) quinoxaline) and non-chiral phosphine (1,2-Bis(diphenylphosphino)ethane, dppe) ligands lead to distorted Au(I), (1, 2, 4, 5) and novel cyclometalated Au(III) complexes (3, 6). These gold compounds were characterized by multinuclear nMR, microanalysis, mass spectrometry, and X-ray crystallography. the inherent electrochemical properties of the gold complexes were also studied by cyclic voltammetry and theoretical insight of the complexes was gained by density functional theory and TD-DFT calculations. The complexes effectively kill cancer cells with IC 50 in the range of ~0.10-2.53 μΜ across K562, H460, and OVCAR8 cell lines. In addition, the retinal pigment epithelial cell line, RPE-Neo was used as a healthy cell line for comparison. Differential cellular uptake in cancer cells was observed for the compounds by measuring the intracellular accumulation of gold using ICP-OES. Furthermore, the compounds trigger early -late stage apoptosis through potential disruption of redox homeostasis. Complexes 1 and 3 induce predominant G1 cell cycle arrest. Results presented in this report suggest that stable gold-phosphine complexes with variable oxidation states hold promise in anticancer drug discovery and need further development.Gold-based probe development and drug discovery remain a burgeoning area of biological research and treatment for disease indications such as cancer 1-5 , arthritis [6][7][8][9] , and microbial infection 10,11 following the FDA approval of tetra-O-acetylglucose-1-thiolgold(I) triethylphosphine complex (auranofin). Exploring the Au(I) and Au(III) chemical space has given rise to enormous diversity of gold compounds of biological relevance, influenced by creative ligand design [12][13][14][15][16] . Despite effective clinical and preclinical treatment of cancer and rheumatoid arthritis by gold complexes such as auranofin, the molecular basis of drug action remains unclear for gold(III) phosphine compounds present in this report. Years of research implicates a number of disease targets including: (i) proteasome-associated deubiquitinases [6][7][8][9] ; (ii) thiol-rich enzymes such as thioredoxin and glutathione reductase [17][18][19][20] ; (iii) thiol-dependent proteases 21 ; iv) autophagy induction 22 ; and superoxide/oxyradical ion generation 23 .Auranofin, which is under clinical and preclinical investigation for the treatment of a variety of cancers including leukemia 24,25 and ovarian malignancies [26][27][28] as well as microbial infections 29-31 is a phosphinogold complex. This has accelerated the development and discovery of several gold-phosphine complexes for therapeutic applications. Gold(I)-phosphine anti-cancer complexes have been identified to trigger apoptosis by targeting the mitochondria and inhibiting thioredoxin reductase [32][33][34] . Structural diversity of gold complexes bearing phosphine ligands have important implications for anticancer activity and probe development 20,35 . Work by Berners-Price et al. demonstrated the anticancer effect of gold-phosphine complexes and have also tried to improve the in vitro and in vivo efficacy of this class of compounds 2,5,[36][37][38][39][40][41][42] . Gold complexes bearing dithiocarbamate [43][44][45] and triorganophosphine ligands 33,46 of the type [(R 3 P) Au(S 2 CNR 2 )] display anticancer activity across a panel of cancer cells including ovarian cancer cells 47 . Recently, Darkwa and co-workers synthesized dinuclear phosphinogold(I) complexes bearing varied phosphine ligands including triphenylphosphine, and diphenylphosphino-alkanes and dithiocarbamates of the type [Au 2 Cl 2 (dppe)] and evaluated their anticancer activity 47 . The complexes displayed broad spectrum of activity in a number of cancer cell lines. Additionally, the anticancer activity of phosphinogold(I) complexes bearing thioglucose ligands as in the case of auranofin show higher potency than their thiolate counterparts even in cisplatin resistant cells. For example, the P -Au -S structural motif is prevalent in a number of gold-phosphine complexes such as the lupinylsulfide (OmS) or sulfanylpropenoate (sppa) 48 containing phosphinogold(I), [AuOmS) 2 (Ph 2 P(CH 2 ) 2 PPh 2 ] or[Au(PPh 3 )(sppa)], respectively and they exhibit good anticancer activity 49 . Improving the biological activity of gold-phosphine complexes require ligand tuning that expand diversity, lipophilicity, physiological stability, and high selective cytotoxicity in cancer cells over normal cells 50,51 .

gold(i/iii)-phosphine_complexes_as_potent_antiproliferative_agents

<!>Results and Discussion<!>Synthesis and characterization.<!>Density functional theory and TD-DFT calculations.<!>Mode of action studies.<!>conclusion<!>Methods<!>electrochemistry.<!>Cell viability assay.

<p>Whereas a lot of work has been conducted with linear phosphinogold(I), its high oxidation state counterpart gold(III) needs further exploration. Recent advancement of cyclometalated gold(III) in anticancer development show promising results 1,[52][53][54][55] . These ligands impart strong σ-donating character to the gold center for stability and offer the possibility of different ligands around the metal center, given its square-planar geometry 56 . Che and co-workers showed that dinuclear cyclometalated gold(III) phosphine, [(C^N^C) 2 Au 2 (μ-dppp)]CF 3 SO 3 ) 2 inhibit hepatocellular carcinoma in vivo by inducing ER stress 57 . There still remains the need to expand the structural diversity of gold-phosphine complexes by designing new gold(III)-phosphine complexes.</p><p>Another important feature of ligands in the context of biological efficacy is chirality, since they possess the property to tune substrates to respective biological targets for improved target engagement that may be elusive for non-chiral ones. The use of chiral ligands in gold drug discovery remain largely unexplored. Incorporating chiral ligands into gold(I) or gold(III) complexes will expand the chemical space to further opportunities in medicinal inorganic chemistry.</p><p>In this report, we synthesized gold(I) complexes bearing chiral or achiral phosphine ligands and in addition mononuclear (C^N)-cyclometalated gold(III) bearing chiral or achiral phosphine ligands. The complexes display potent cytotoxic activity in different cancer cell lines by triggering apoptosis through ROS induction. The study establishes the need for a broader scope of gold complexes for cancer therapy.</p><!><p>Rationale and approach. Stabilizing the gold metal center for biological utility remains an important aspect of metallodrug discovery. Gold compounds possess high redox potential (i.e. Au +3 + 2e − → Au +1 E° = 1.41 V and Au +1 + e − → Au (s) E o = 1.69 V) 58 . This is due to relativistic effects. Often, strong σ-donating ligands such as phosphines and N-heterocyclic carbenes are used to improved stability of gold complexes against rapid reduction 59 . The polarizable nature of gold, owing to its loosely held electrons facilitates its coordination to soft Lewis bases such as phosphorus, sulfur containing ligands. Linear gold(I) complexes are limited by their low coordination number. On the other hand, gold(III) compounds largely exist in a square-planar configuration and provides increased number of coordination sites. This can be harnessed for the development of anticancer agents bearing multidentate ligands with enhanced stability and activity. In addition the use of chiral phosphine ligands in gold scaffolds and their effects on geometry remains underdeveloped. A study that examines the effect of oxidation states of compounds bearing phosphine ligands will be beneficial. We envisioned the investigation of gold(I)/(III)-phosphino complexes as potential anticancer agents to offer preliminary structure-activity insights of gold-based metallodrugs. These gold(I)/(III) complexes could interact with biological targets, such as proteins, in a similar fashion to reported gold-phosphino complexes but with optimal kinetic lability to avoid premature deactivation. Within this context, work that studied the anticancer potential of gold-phosphino complexes offered impetus that the complexes synthesized in our laboratory will be active against cancer cells in vitro. Furthermore, the distorted gold(I) with phosphine ligands as well as gold(III) complexes bearing cyclometalated and phosphine ligands could exhibit the needed stability for effective therapeutic response, even in drug-resistant cancer.</p><p>The design of our compound library employed the use of the chiral (R,R)-(-)-2,3-bis(t-butylmethylphosphino) quinoxaline and achiral 1,2-bis(diphenylphosphino)ethane (DPPE) phosphine ligands due to their biocompatibility and moderate lipophilicity when present coordination complexes or as ancillary ligands in organometallic compounds. For further investigation into the geometry and oxidation states of these complexes and their effects on anticancer activity, we explored the synthesis of dinucleur gold(I) complexes and (C^N)-cyclometalated backbone for the formation of an organometallic gold(III), all bearing phosphine ligands used in this study. We hypothesized that the small set of compounds with rich diversity including, ligand type, stereochemistry, oxidation state, geometry and distortion, would possess interesting biological properties for gold-based drug discovery.</p><!><p>The gold-phosphine complexes under investigation in this study are depicted in Fig. 1. The chiral phosphine ligand, (R,R)-(-)-2,3-bis(t-butylmethylphosphino) quinoxaline 60 , which is well tolerated in mice was used for the synthesis of the respective Au(I) and Au(III) compounds, using tetrachloroauric acid trihydrate (HAuCl 4 •3H 2 O) or AuCl(tht), or (C,N)-cyclometalated Au(III)Cl 2 starting reagents, demonstrated in compounds 1, 2, and 3. To expand the diversity of gold-phosphine complexes, we synthesized compounds with the archetypical achiral phosphorus-donor ligand, DPPE, exemplified in 4 [61][62][63] , 5 [64][65][66] , according to previously reported protocols but with different anions, and the novel gold(III), 6. The synthetic approach for compound 1, was achieved by adding a chloroform solution of (R,R)-(-)-2,3-bis(t-butylmethylphosphino) quinoxaline to a cold solution of AuCl(tht) in chloroform. This complex was dried and recrystallized from chloroform/ether to give a pale yellow neutral compound. The synthesis of 2 was carried out by refluxing an equimolar concentration of HAuCl 4 •3H 2 O and (R,R)-(-)-2,3-bis(t-butylmethylphosphino) quinoxaline in chloroform for 1 h. After filtering through celite the reaction was purified by flash silica-gel chromatography to obtain 2 as a brick-red monocationic solid. We synthesized gold(III) analogs using the benzyolpyridine (C^N)-cyclometalated complex, 7, which was synthesized as previously reported 67 . The chiral ((R,R)-(-)-2,3-bis(t-butylmethylphosphino) quinoxaline) and achiral (DPPE) bisphosphine ligands were used in a ligand substitution reaction with 7 to give 3 and 6 respectively, after purification by silica-gel column chromatography. Phosphines are strong coordinating ligands which can readily displace the chlorine atoms. In the reaction that leads to complex 3, the formation of 2 in low yields is observed. This is as a result of reductive elimination promoted by the nucleophilic phosphine. It is a subject of intense investigation within our laboratory 68 . Complexes 4 and 5 bearing DPPE ligands were synthesized to investigate the effect of perchlorate counter anions on cellular activity. The complexes were purified by recrystallization from acetonitrile as white cationic solids.</p><p>The complexes were characterized by 1 H, 13 C, and 31 The high-energy bands observed in the UV-vis spectra, likely correspond to ππ*transitions from the ligands 69,70 or of ligand-to-metal charge transfer (LMCT) or metal-to-ligand charge-transfer (MLCT) character, which has been well-characterized by TDDFT calculations (vide infra). For example, in complex 7 there are two bands <300 nm, there is a high-energy band at 245 nm and another peak at 295 nm, which may be as a result of MLCT or LMCT transition [71][72][73] . There was no change in the absorbance profile of the gold(III) complex bearing cyclometalated and phosphine ligands (3) in PBS and minor changes in DMEM over the course of 48 h (Fig. S24-25). In contrast, the gold(I) bearing phosphine ligands experienced spectral changes in both PBS and DMEM over 48 h (Fig. S26-27). In PBS, the gold complexes show solubility and stability as shown from the absorption profiles in Fig. 2. A broad absorption band was noticeable for compound 1, which can be attributed to possible aggregation due to the relatively planar and lipophilic character of the dinuclear complex, 1, which can form π-stacking in aqueous solution. The representative examples offer insight into the role of cyclometalation and conjugated phosphine ligands to stabilize the gold center.</p><p>Complex 2, representing Au(I), shows a continuous decrease in absorption peaks at 248 and 337 nm over time in PBS. In contrast, complex 3, which is Au(III) showed no change in absorption bands. It can be deduced that the Au-P bonding in complex 2 is weakly stable in PBS compared to that of Au-P of complex 3. This is supported by computational results, in that, while the four Au-P bonds of complex 2 are ~2.5 Å, the two Au-P of complex 3 show shorter bond lengths of 2.38 and 2.5 Å, respectively, indicative of stronger bonds. The stability associated with 3, could be ascribed to both cyclometalated-Au bonding and adjacent Au-P bonds that enhance sigma-donation to the gold center. Complex 7, which is another Au(III) display significant changes to the absorption bands in DMEM. It appears that chloride, a good leaving group, is substituted by amino acids contained in DMEM 74 . On the contrary, it shows stability in PBS, this seems like PBS does not contain components to replace Cl of complex 7. Additionally, PBS contain 10 mM of NaCl that is likely to help stabilize the complex.</p><!><p>To gain theoretical insight into the gold complexes under investigation, DFT calculations were conducted. We used Gaussian 09 and the B3LYP functional was employed 75 . The basis set used was SDD on Au atom, taking into consideration relativistic effects and the Pople-type 6-31 G(d,p) was used for all other atoms. The geometry of the computed complexes was compared with the solved X-ray crystal structures to ensure the validity of the computations performed. The result is summarized in Fig. S31. Briefly, we considered the atomic groups around the gold atoms. For complex 1, two independent distances between Au atom and P atom were 2.22 and 2.23 Å, and the calculated values showed minimal discrepancy, with an error of 3%. Similarly, two independent bond distances between Au and Cl was 2.28~2.29 Å, whereas the calculated values were within ~2% error. Interestingly, the bond angle of P-Au-Cl, from the crystal structure of 1 was 171° and the corresponding calculated value was 171°. In the case of 2, the distorted structure of 2 had four P-Au bonds and its distances were all ~2.39 Å, and the calculated values were ~2.5 Å. Measured angles of three kinds of P-Au-P were 87.79°, 137.50°, and 106.48°, respectively, while the calculated values were 86.13°, 132.65°, and 111.58°. Unlike 1 and 2, compounds 4 and 5 possess dppe ligands, which also showed minimal discrepancy between the experimental and calculated bond distances or angles. For complex 4, the two gold metal atoms had two separate P-Au-P bonds, all four P-Au bond distances were 2.31 Å and the calculated values were ~2.38 Å. The X-ray structure revealed two P-Au-P angles as 166.8° and 177.8°, with the determined calculated angles were 169.1° and 171.0°, respectively. Complex 5 had a distorted structure like 2. The distance between the four P-Au distances were all 2.40 Å, while the calculated values was in the range of 2.35~2.36 Å, showing less than 3% error. The three P-Au-P angles were 86°, 117°, and 118°, and the calculated values were 91°, 128°, and 114°, respectively. Taken together, this shows that DFT calculations and the method employed can be considered suitable for providing accurate geometric information of gold complexes.</p><p>To corroborate peak assignments within the experimental absorption profiles and further provide insight into HOMO-LUMO transitions, we performed TD-DFT calculations.</p><p>In Fig. 3, the experimental UV-Vis spectrum for complex 6 was compared to the calculated spectrum. The oscillator strength and molecular orbital (MO) contributions obtained from the theoretical calculation is summarized in Table 1. The limitation of the method and basis sets used in the theoretical calculation as a result of the large relativistic effect of gold present some discrepancy, which is consistent with other gold systems in the literature. In the UV-Vis spectra, generally, 2-3 peaks were observed and the theoretical calculations (TD-DFT) were carried out to establish the relationship between the observed peaks and the associated electronic transition. The experimental absorption profile show a high-energy absorption peak located at approximately, 250 nm and a red-shifted shoulder band at ~275 nm. However, the calculated spectra show a high energy band at 350 nm and its related red-shifted band at 400 nm, confirming a discrepancy of about 100-150 nm from the experimental spectra. This is typical for gold systems 76,77 . Furthermore, for complex 1, the contribution of HOMO-2 to LUMO was the largest, corresponding to a MLCT transition. The HOMO-2 is largely localized on the gold and the neighboring Cl atoms and the LUMO on the phenyl/pyrazine of quinoxaline. A similar charge transfer is seen in HOMO-LUMO transition for compound 1 (Fig. S32). Additionally, MLCT is also seen in complexes 2 (Fig. S33) and 5 (Fig. S34). 2 has the same ligand as 1 and shows similar metastases: λ = 221 nm (HOMO to LUMO + 2) and λ = 249 nm (HOMO-1 to LUMO + 1), which are charge transfer from the gold center to both quinoxaline ligands. For λ = 326 nm, the HOMO-LUMO transition represents a similar transfer from the gold center to the www.nature.com/scientificreports www.nature.com/scientificreports/ periphery of the molecule. For complex 5, electrons distributed around the metal migrate to neighboring phenyl groups: λ = 243 nm (HOMO to LUMO + 3) and λ = 292 nm (HOMO to LUMO). Similarly, for complex 5, the high energy is caused by MLCT of HOMO to LUMO + 3, and the neighboring λ = 292 nm (HOMO to LUMO) by gold center to the surrounding phenyl groups of the the dppe ligand. In contrast, complexes 3, 4, 6 and 7 show mainly LMCT transitions. For 3, high energy peaks (λ = 256 nm) were observed in quinoxaline-to-gold charge transfer (HOMO-3 to LUMO and HOMO to LUMO + 1) and low energy peaks were considered as transition from benzopyridine to gold center (HOMO-2 to LUMO and HOMO-1 to LUMO, Fig. S35). Complex 4 is largely due to the transition of HOMO-2 to LUMO for λ = 241 nm, from the two phenyl rings of the dppe ligand to the gold, and HOMO to LUMO for λ = 294 nm from the phenyl to gold (Fig. S36). For 6, λ = 237 nm correspond to, HOMO-2 to LUMO, which is consistent with a LMCT transition based on the localization of the HOMO-2 and the LUMO and λ = 268 nm is as a result of a similar transition, HOMO to LUMO (Fig. 4). Complex 7 also possess a similar profile with λ = 234 nm corresponding to HOMO-2 to LUMO + 1, which is a LMCT transition, and λ = 285 nm, resulting from HOMO to LUMO with LLCT transition (Fig. S37). Although the complexes use www.nature.com/scientificreports www.nature.com/scientificreports/ similar ligands (quinoxaline, benzoylpyridine, and dppe), the type of charge transfer in the UV-Vis spectrum is different (MLCT, LMCT, or LLCT) [69][70][71][72][73] because the different types of ligands, combination of ligands, and the different coordination number of the gold make a difference. The Au(III) complexes all display LMCT transitions and the Au(I) complexes display MLCT transition with the exception of 4. Overall, the study provide insight into the MO contributions of the experimentally obtained absorption profiles of these cytotoxic compounds.</p><p>Variable temperature NMR. Prior to biological investigation of these complexes, we assessed the thermal stability of representative compounds in DMSO-d 6 and D 2 O. Given that stock solutions of the complexes were prepared in DMSO prior to biological, photophysical, or electrochemical evaluation DMSO-d 6 was used. Consequently, all the biological evaluation is in aqueous base medium, thus studying their stability in D 2 O was appropriate. We measured the 1 H-NMR of complexes 1-6 within a temperature range of 24-80 °C (Fig. S38-50). There were no obvious changes in the 1 H-NMR spectra for the respective complexes studied over the temperature range, indicative of stability of these complexes under harsh conditions. In summary, these gold compounds show thermal stability in DMSO and D 2 O, which is an important characteristic for biologically relevant transition metal complexes.</p><p>X-ray crystallography. Single crystals of four complexes out of the six compounds studied were obtained and the crystal structures were determined by X-ray crystallography. Crystal structures with optimized structures for 1, 2, 4, and 5 are shown in Fig. 5. We note that 4 [61][62][63] , and 5 [64][65][66] share cationic similarity to structures previously reported, for different salts of these cationic complexes. A comparison of the previously reported structures and ours reveal the perchlorate anions and no significant differences in the overall geometry of the gold complex. Moreover, the dinuclear gold compound, 4 crystallizes in the triclinic P1 space group, while 5 crystallizes in the orthorhombic space group, Pca2 1 . Crystallographic information and selected interatomic distances for compounds 1, 2, 4, and 5 can be found in Table S1-4. For the dinuclear complex, 1, it crystallizes in the orthorhombic space group, P2 1 2 1 2 1 . There is a slight distortion in the linear geometry of the P-Au-Cl bonds, for example, the P1-Au1-Cl1 angle is 171.77 (9). The rigid chiral bisphosphine ligand may be the culprit for the observed distortion. Additionally, the bond angles C-P-Au were different for the methyl and tert-butyl group bonded to the phosphorus. For example, C9-P1-Au1 is 108.1(3), whereas C10-P1-Au1 is 112.6(3), indicative of a larger angle influenced by the bulkier tert-butyl fragment. Importantly, the dinuclear gold compound does not have a bond between the two gold atoms. Compound 2 crystallizes in the monoclinic, C2 space group which is a deviation from 1, despite the similar chiral ligand environment. The complex exhibits an unusually distorted square planar geometry for Au(I) species. The Au-P bond distances for both 1 and 2 are slightly invariant as we observe a longer bond distance for Au-P in 2, when compared to Au-P distances in 1. For example, complex 1, Au1-P1: 2.225(2) Å and Au2-P2: 2.236(2) Å, and for the complex 2, Au1-P1: 2.3889( 14) Å, Au1-P4: 2.390(2) Å, Au1-P2: 2.3928( 19) electrochemistry. To elucidate the reduction and oxidation properties of these gold complexes, we measured the cyclic voltammetry in DMSO with sodium perchlorate (NaClO 4 ) as supporting electrolyte at a Pt electrode, using CH-600D potentiostat equipment. One reversible reduction event was observed for the gold(I) complex, 1 at a potential of −1.20 V (Fig. S52). In an oxidative sweep, an oxidation peak appears at −0.98 V. This phenomenon is observed in the free ligand, ((R,R)-(-)−2,3-bis(t-butylmethylphosphino) quinoxaline (QuinoxP), which possess a more negative reduction potential at −1.68 V. Complex 2, displays two successive reversible processes with reduction potentials at −1.26 V and −1.38 V respectively (Fig. 6). Note that 2 is supported by two QuinoxP ligands with distortions from a classical linear Au(I) complex that may lead to two closely distinct redox events. The result and the absence of metallic gold at the platinum working electrode led to the conclusion that the overall reduction for this class of Au(I) complexes is ligand-centered rather than metal-centered 78 .</p><p>The electrochemical behavior of 4 and 5, which are Au(I) complexes bearing DPPE ligands were investigated. Complex 4 shows irreversible reduction peak at −0.84 and −1.54 V (Fig. S53). Similarly, 5 show irreversible reduction events at −0.81 V and −1.55 V with no oxidation events (Fig. S54). The events for 4 and 5 are considered to occur on the ligand, given that the reduction potentials of compounds 4 and 5 were similar and consistent with dppe as well as the lack of gold precipitation. We attribute the observation to stability of the complexes and ligand-centered redox events. Furthermore, we studied the electrochemical behavior of the gold(III) complexes, 3 and 6 in DMSO using cyclic voltammetry. As shown in Fig. S55, irreversible reduction peak at −0.80 V was observed and a reduction peak at −1.46 V for 3. In addition, a corresponding oxidation event was observed at −1.27. 6 showed two reversible reductions at −1.01 and −1.37 V, and the paired oxidation events at -0.86 and −1.31 V were observed, respectively (Fig. S56). Typically, gold(III) undergo two separated Au(III)/Au(I) and Au(I)/Au(0) steps in non-aqueous environment or three electron 79 , Au(III)/Au(0) reduction in aqueous solutions due to disproportionation of Au(I) in aqueous solution [80][81][82] . Our observation lean more towards the former with an initial Au(III)/Au(I) reduction for 3 and 6.</p><p>Reactivity with BSA. Spectrophotometric investigations of the gold complexes (1-6) described in this report in a reaction with bovine serum albumin (BSA) was performed under physiological conditions. Taking advantage of the intense absorption bands of the gold complexes and BSA, we monitored the progress of the reaction using 1:1 ratio of BSA and buffered solutions of each gold complex over 24 h. Serum albumin is a major soluble protein component present in the circulatory system and has many physiological functions 83 . Importantly, BSA acts as a carrier for various pharmacological agents. It must be noted that BSA has been extensively studied, and shares homology with human serum albumin (HSA) 84 . Often, gold compounds bind methionine and cysteine residues in BSA via the sulfur atoms. The inherent absorption peaks for complexes 2-6 were minimally affected by the addition of the BSA solution over a 24 h period (Figs 7 and S57-S61). It is also worth pointing out that the peak corresponding to the absorption of BSA at 280 nm was unaffected under the experimental conditions. The ability for compound 1 to aggregate in aqueous solution limited the ability to evaluate it under the experimental conditions. However, a solution of compound 1 with BSA did not affect the peak attributed to BSA (Fig. S62). For compound 2, while a decrease in the absorption band corresponding to MLCT at ~250 nm was observed, no changes in the band at 325 nm was observed in the course of the experiment. The observed decrease is consistent with the time-dependent study of 2 in PBS. Also, the peak corresponding to BSA remained unchanged throughout the 24 h period. Complexes 3 and 6 are gold(III) compounds with cyclometalated ligands but different bisphosphine ligands. Interestingly, none of the peaks associated with the complexes or BSA changed, indicative of stability in the presence of BSA over 24 h. While the shoulder peak at 300 nm disappeared in the case of complex 4 (LMCT), the BSA peak at 250 nm was unmodified. Complex 5 on the other hand did not display any alteration in its peak. In general, there was no indication of the formation of metallic gold as no brown precipitate formed in any of the reaction over the duration of the experiments. Using HPLC (Fig. S63-S64) we characterized the extent of interaction of the test compounds and BSA. This approach can be used to quantify potential binding of gold compounds with BSA by evaluating the retention times and area of peaks associated with individual agents as well as reaction solutions of test compounds and BSA. Following the UV-vis studies, we used compounds 2 and 3 for the HPLC study based on the common chiral ligands but different oxidation states. There were no changes in the peaks, indicative of no covalent modification of BSA or changes to the gold compounds. These compounds by virtue of their coordinated ligands and cyclometalation demonstrate high stability even towards proteins like BSA. Detailed studies by Minghetti and co-workers 85 on the reactivity of selected gold(III) complexes with serum albumin under similar experimental conditions showed varied stability of the gold complexes in the presence of BSA. This result has important implications for the pharmacological activity of these gold complexes, in that they can avoid premature deactivation until they reach their target and also reduce off-target effects. cellular toxicity studies. The antiproliferative properties of these gold complexes were evaluated in a panel of cancerous cell lines using crystal violet assay for H460 and OVCAR8. We used an ATP-dependent luminescence cell assay, cell titre glo, for K562 cells. To extend the therapeutic utility of these novel drug candidates, we performed cytotoxicity studies with normal retinal pigment epithelium, RPE-NEO. Auranofin and cisplatin were used as controls. We obtained dose-response curves from the cell viability experiments and subsequently derived IC 50 values (concentration required to kill 50% of cells) summarized in Table 2. Complexes 1-6 displayed high nanomolar to low micromolar cell killing which are 2-10 folds better than cisplatin. None of the gold-phosphine complexes display cross-resistance evidenced by indifferent toxicities in cisplatin resistant cells, including the well-characterized high grade-serous ovarian cancer (HGSOC) cell line, OVCAR8, which demonstrate the example of high potency of these novel Au complexes in clinically relevant tumor cells. Generally, the gold compounds studied are slightly less potent toward RPE-Neo cells, indicative of selective toxicity for cancerous cells compared Table 3 summarizes the electrochemical potentials, LUMO eigenvalues and representative cytotoxicity of the gold complexes studied in this report. In general, the redox active behavior of the complexes may be suggestive of redox induced cell-death. However, we could not establish a positive correlation between reduction potentials, LUMO eigenvalues and cytotoxicity. Several factors affect the induction of cell death by metal-based drugs and such correlations must be cautiously applied. cellular uptake studies. To gain insight into the intracellular behavior of the gold reagents, we studied the whole cell uptake of 1-6 and auranofin as well as subcellular distribution of Au for 1 and 3. OVCAR8 cells were incubated with the gold compounds (5 μM) for 15 h. The Au content was measured by using ICP-OES, calibrated with HAuCl 4 standard. Differential whole cell uptake was observed for all compounds with neutral complexes, auranofin and 1 displaying the highest uptake respectively and 6 taken up the least (Fig. 8). Generally, increased cellular uptake correlates well with high cytotoxicity and biological effects thereof. Previous work using other gold phosphine complexes show that increased uptake is accompanied by cell growth inhibition, and that a parabolic dependence between cell growth inhibition, uptake, and lipophilicity does exist 86,87 . Thus, it is possible to predict whether changes in the oxidation number of gold complexes and changes in lipophilicity of the complexes would have an effect on uptake and growth inhibition. However, the set of test compounds investigated in this report did not establish significant correlations of uptake and cytotoxicity. A clear observation of the data revealed higher uptake for dinuclear complexes (1 or 4) over mononuclear gold complexes bearing chiral or achiral phosphine ligands respectively. It is likely that the dinuclear gold increases lipophilicity of the agents. Additionally, the use of chiral QuinoxP ligands have a higher lipophilic character over the dppe ligands. We evaluated 1 and 3, which bear the same chiral ligand but different oxidation states in a series of biological experiments and although IC 50 values do not show significant differences, apoptosis assay (vide infra) reveal a much higher early to late-stage apoptosis for 1, which has a higher cellular uptake than 3. Following the incubation of OVCAR8 cells with 1 or 3, we isolated the nuclear component from the cytoplasm and the residual cell pellet. As shown in Fig. S65, complex 1 or 3 was localized predominantly to the cytoplasmic fraction at 200 pmol/10 6 cells concentration. In contrast, a lower localization within the nuclear fraction at 15 pmol/10 6 cells was observed for both 1 and 3. The similar Au concentration in the subcellular fractions despite the different structural scaffolds may be the clue to the similar growth inhibition observed. The concentration found within the cell residue (pellet) was similar to that of the nuclear fraction, within margin of error. These results demonstrate that the new Au constructs largely localize and may target cytoplasmic proteins owing to their lipophilic chiral ligands with minimal potential to target genomic DNA.</p><!><p>To further understand the cellular responses evoked by 1 or 3, the compounds were examined for apoptotic effects. Mitochondria dysfunction and ER stress can result in apoptosis. Characteristically, cells undergoing apoptosis have associated cell membrane disorientation. This leads to efficient phosphatidylserine residue translocation from the interior of the cell to the exterior membrane, which can be readily detected by Annexin V molecules. We performed a dual-stain experiment involving FITC-labeled Annexin V and propidium iodide (PI) for flow cytometry analysis. The HGSOC cell line, OVCAR8, were treated with 1, 3, auranofin or cisplatin for 72 h. We observed a large population occurring as late-stage apoptotic cells for the Au(I) compounds, i.e. 1 and auranofin. In general, all the Au complexes induce significant apoptosis over cisplatin as shown in Fig. 9. Taken together, these Au reagents display enhanced in vitro potency when compared with the conventional platinum(II)-based agent, cisplatin. Suspicious of mitochondria dysfunction and ROS production as primary inducers of cell death, we conducted intracellular reactive oxygen species (ROS) experiments using human cancer cell line, OVCAR8 (Fig. 10b). Mitochondria are the power hub that control cell death signaling within cells and generate large amounts of ROS within the cell. That said, the mitochondria is also susceptible to ROS attack. Thus, measuring ROS production induced by the gold compounds presented in this report is rational. Under the experimental condition, compound treated cells revealed a low amount of ROS. It is fair to say that all compounds inhibit ROS, especially the Having established the effects of 1-6 on intracellular ROS, mitochondrial membrane potential studies were conducted to monitor mitochondrial insults. Notably, 1 and 2, largely depolarized the mitochondria membrane by measuring the fluorescence of rhodamine 123 using flow cytometry (Figs 10c, S66). Well studied organic small-molecule that modulate ROS are known to induce mitochondrial membrane potential depolarization resulting in cell death 88 . Additionally, a few organometallic gold complexes induce mitochondria depolarization as a result of mitochondrial dysfunction or ER stress 1,89 . Mitochondria membrane depolarization often stimulates caspase signaling, hence, using immunoblotting analyses, we monitored changes in the expression of biomarkers related to apoptosis and DNA damage response (Fig. 10d). OVCAR8 cells incubated with 1-3 for 48 h revealed marked increase in caspase 3 and 9 levels 2,41,42,90 . While 3 induced differential expression of cleaved PARP and γ-H2AX, compounds 1 and 2 did not, suggesting that 3 may impose a broader mechanism of action. The generation of ROS is capable of compromising mitochondria function, leading to the activation of caspase-3 and cleavage of PARP. Our experimental results parallel the existing mode of action with a unique differentiation. For example, complexes 1 and 2 are not involved in ROS production and do not induce PARP cleavage. This suggests that within the apoptotic program, 1 and 2 may cause late apoptosis as rightly observed for 1 (Fig. 9) since PARP cleavage occurs early during apoptosis in order to avoid energy (NAD and ATP) depletion needed for later stages in the apoptotic program 91,92 . Overall, these studies confirm the differentiated mechanism of novel chiral or non-chiral gold complexes that are capable of inhibiting ROS in cells, setting off mitochondrial membrane potential depolarization as a result of perturbed redox homeostasis that lead to apoptosis in cancer cells. The preliminary mechanistic activity of the gold candidates was characterized in OVCAR8 cells by assessing cell cycle effects. Using flow cytometric analyses on OVCAR8 cells after treatment with 1 or 3, incubating the cells with each agent at ~two-fold higher concentrations than their respective in vitro IC50 values for either 24, 48 or 72 h. The study revealed that 1 and 3 stalled the cell cycle at the G0/G1 phase in a time-dependent manner. Notably, within 24 h, there was an increase in the G2/M phases, which diminished to levels relative to untreated controls by 72 h (Figs 10a and S67). This observation differs from cell cycle patterns of cisplatin-treated cells, which primarily induce G2/M phase arrest by 72 h for most ovarian cancer cells. Undoubtedly, a more rigorous biological characterization is required to fully elucidate the mechanism.</p><!><p>The synthesis and characterization, electrochemical, and biological evaluation of six gold complexes bearing chiral and achiral bisphosphine ligands were investigated. All compounds display potent cytotoxic effects in cancer cells over normal cells. Particularly, the gold compounds described show high antiproliferative activity in a panel of cell lines including the clinically relevant HGSOC OVCAR8 cell line, ranging from 2-10-fold lower www.nature.com/scientificreports www.nature.com/scientificreports/ IC 50 values when compared to cisplatin. The complexes investigated offer a broader mechanism of action than cisplatin or auranofin including apoptosis, mitochondria depolarization and G0/G1 cell cycle arrest. Despite differential intracellular accumulation of complexes 1-6, the cytotoxic effects were not significantly invariant. The redox potential of these complexes is affected by the choice of ligands. Additionally, theoretical insight using DFT and TD-DFT calculations demonstrated HOMO-LUMO orbital distributions which can inform future design and donor effect of ligands for stability and prediction of excited states. Although this study uses a relatively modest set of compounds, they represent an important proof-of-concept study for the development of new gold-phosphine complexes that possess ligand variability and different oxidation states. Ongoing studies are being performed to expand the scope of stable gold(III)-phosphine complexes as effective anticancer agents.</p><!><p>General information. Auranofin and Cisplatin were purchased from Santa Cruz Biotechnology Inc. and Strem, respectively and used without further purification. All other reagents were purchased from Strem, Sigma Aldrich, or Alfa Aesar and used without further purification. All reactions were carried out under normal atmospheric conditions. Deuterated solvents were purchased from Cambridge Isotope Laboratories (Andover, MA). 1 H, 13 C, 31 P, COSY, and HSQC NMR spectra were recorded on a Varian Unity 400/500 NMR spectrometer with a Spectro Spin superconducting magnet in the University of Kentucky NMR facility. Chemical shifts in 1 H, 13 C, COSY, and HSQC NMR spectra were internally referenced to solvent signals ( 1 H NMR: DMSO at δ = 2.50 ppm; 13 C NMR: DMSO at δ = 39.52 ppm), and those in 31 P NMR spectra were externally referenced to 85% H 3 PO 4 in D 2 O (δ = 0 ppm). Electrospray ionization mass spectrometry (ESI-MS) was performed on an Agilent Technologies 1100 series liquid chromatography/MS instrument. High-resolution mass spectra (HRMS) were obtained by direct flow injection (injection volume = 5 or 2 μL) ElectroSpray Ionization (ESI) on a Waters Qtof API US instrument in the positive mode (CIC, Boston University). Typical conditions are as follows: capillary = 3000 kV, cone = 35 or 15, source temperature = 120 °C, and desolvation temperature = 350 °C. In addition to spectroscopic characterization, bulk purity of all new compounds was assessed by combustion elemental analysis for C, H, N. Elemental analysis was carried out at the microanalysis lab at University of Illinois Urbana Champaign using Perkin Elmer 2440, Series II with a combustion temperature of ~2000 °C and accuracy of 0.3% abs. Reactions were monitored using aluminum backed silica-gel thin-layer chromatography (TLC) plates (Silicycle, TLA-R10011B-323, Canada) and visualized under low-wavelength light (254 nm) or stained with iodine on silica for visualization with the naked eye. Purification of reactions was performed using silica-gel (Silicycle, P/N: R10030B (SiliaFlash ® F60, Size: 40-63 μm, Canada) chromatography. The CombiFlash ® www.nature.com/scientificreports www.nature.com/scientificreports/ 126. 34, 126.04, 124.58, 124.47, 54.92, 37.91, 37.68, 36.97, 36.73, 27.11, 27.09, 26.53, 26.51, 26.48, 26.45, 25.34, 25.32, 6.43, 6.10, 5.62, 5.01, 4.66, 4.37. 31 UV-Vis spectrophotometric measurements. Stock solutions (1 mM) of complexes 1-7 were prepared in DMSO. For studies with PBS and DMEM, the gold solutions were diluted with the respective diluent (PBS or DMEM) to achieve a final gold concentration of 25 μM or 50 μM. Background scans were performed with the corresponding DMSO or solvent mixture. For the reactivity experiment with BSA, the sample solution was prepared by the above-mentioned method, and BSA was separately dissolved in PBS to prepare 1 mM stock solution and diluted with PBS. The sample solution was mixed with the BSA solution immediately before the measurement. Single scan or time-based scanning was performed as needed. The UV-Vis spectrometer used was a Shimadzu UV-1280 model.</p><!><p>Cyclic voltammetry was performed in dimethyl sulfoxide (DMSO) with NaClO 4 , which were purchased (Aldrich, USA) and used without further purification. All complexes (1−6) were dissolved in DMSO in the concentration of 1.0 mM, followed by addition of 0.01 M NaClO 4 as a supporting electrolyte. Electrochemical measurements were performed at ambient temperature using CH-600D potentiostat (CH Instruments, USA) for cyclic voltammetry. The voltammetric measurements were performed in a three-electrode cell containing a platinum working electrode, a non-aqueous Ag/AgCl reference electrode and a platinum wire as counter electrode.</p><!><p>Various established human ovarian, leukemia and lung cancer cell lines were seeded in 96-well plates (2 × 10 3 cells/well) and were incubated with RPMI 1640 supplemented with 10% FBS (150 μL) for 24 h and at 37 °C. They were then treated with cisplatin, auranofin, or 1-6 at increasing concentrations for 72 h from stock solutions following serial dilutions. Cisplatin stock solution was prepared using PBS and the gold compounds were prepared in DMSO. Thereafter, cellular viability was assessed via the established crystal violet colorimetric assay. In brief, crystal violet reagent (50 μL of a 0.5% solution in glutaraldehyde) was added to each well and allowed to incubate with cells for 30 min. The plates were washed with water by running them under a gentle flow of tap water. The plates were air dried and dissolved with methanol (100 μL/well) and the plates were rocked for an additional 10 min. Measurements of absorbance were subsequently performed using a Genios plate reader at 570 nm (peak absorbance). All experiments were conducted in triplicate. For K562 cells, a cell density of 4,000 cells/well was used and cellular viability was assessed after treatment, using the established luminescent cell titre glo assay. at 37 o C. Cells were then incubated with the test compounds (5 μM) in fresh RPMI 1640 medium (10 mL) and subsequently incubated for a given period of time (~24 h) at 37 °C. The medium was then removed and cells were collected via trypsinization. Cells were then washed with PBS (1 mL x 3). The cells were digested by adding 0.5 mL of concentrated HCl and briefly placing them on an agitator. Cells were then transferred to a new tube containing 4.5 mL of DI water. The gold content was analyzed by ICP-OES to obtain the whole cell uptake after quantification.</p><p>Intracellular distribution. To measure the intracellular cellular uptake, ~1 million OVCAR8 cells were treated with 5 μM compounds 1, or 3 at 37 °C for ~17 h. The media were removed, and the cells were washed with PBS solution (1 mL × 3), harvested, and centrifuged. The cell pellet was suspended in an appropriate volume of PBS to obtain a homogeneous suspension (100 μL). The nuclear and cytoplasmic extraction kit (NE-PER, Thermo Scientific Inc.) was used to extract the separate cytoplasmic, nuclear, and pelleted fractions. The fractions were mineralized with 70% HNO 3 and then heated at 95 °C for 10 min. The gold content was analyzed by ICP-OES. Cellular gold levels were expressed as pmol of Au per million cells. Results are presented as the mean of three determinations for each data points. cell cycle analysis. OVCAR8 cells were seeded in 6-well plates (2 × 10 5 cells/well) and were treated with PBS, 1, or 3 for 24 or 48 h at a concentration of 2 µM. The cells were re-suspended in ice-cold PBS, fixed with 70% ethanol in PBS at 4 °C overnight, subsequently washed with ice-cold PBS (x2), and then incubated with RNase A (1 μg/mL) for 20 min at 37 °C. They were then stained with PI (10 μg/mL for 30 min in the dark) and their DNA content and cell cycle distribution were measured, using a FACSCalibur flow cytometry (BD Biosciences, USA) and as determined with ModFit software.</p><p>Apoptosis study. OVCAR8 cells were seeded in 6-well plates (5 × 10 5 cells/well), treated with PBS, cisplatin, auranofin, 1 or 3 at a fixed Au concentration (2 µM) for 48 h, and harvested by trypsinization. The Annexin V-FITC Apoptosis Detection Kit (BD Biosciences) was used to determine the fraction of cells (from a total of 1 × 10 4 cells) that underwent apoptosis, using fluorescence-activated cell-sorting sorting (BD Biosciences, USA) and by following the manufacturer's protocol. Data were analyzed using FlowJo software.</p><p>Mitochondria membrane potential. OVCAR8 cells (175,000 cells/well) were plated in 6-well plates taking into account respective controls (untreated and unstained, untreated and stained) and test compounds. After overnight incubation of cells at, 37 °C and 5% CO 2 , 5 μM of complexes was added to the desired wells and incubated overnight. We collected cells via trypsinization and centrifugation, and re-suspended cells in a solution of 70% ethanol in PBS and placed in freezer for at least 5 minutes. Rhodamine-123 solution (1 mg/ml in ethanol) was dissolved and diluted with PBS to a final concentration of 0.005 mg/mL. The solution was added (0.5 mL) to the cells re-suspended, and incubated for 30 min. Cells were pelleted by centrifugation, washed with 1 mL of PBS, and re-suspended in 0.5 mL of PBS for flow cytometry analysis.</p><p>Western blot. OVCAR8 cells (500,000 cells/well) were plated on a 6-well plates. The cells were then treated with 5 μM gold compounds and incubated for 24 h and 48 h at 37 °C, after which media was removed and cells were washed with PBS. The cells were scraped into SDS-PAGE loading buffer (64 mM Tris-HCl (pH 6.8)/9.6% glycerol/2% SDS/5% β-mercaptoethanol/0.01% bromophenol blue) and incubated at 95 °C on a heat block for 10 min. The samples were cooled and stored at −20 °C until ready for use. Whole cell lysates were resolved by 4-20% sodium dodecylsulfate polyacrylamide gel electrophoresis (SDS-PAGE; 200 V for 25 min) followed by electro transfer to a polyvinylidene difluoride membrane, PVDF (350 mA for 1 h). Membranes were blocked using 5% (w/v) bovine serum albumin (BSA) in PBST (PBS/0.1% Tween 20) and incubated with specific primary antibodies (Cell Signaling Technology and Santa Cruz Biotechnology) overnight at 4 °C. On the following day, after washing with PBST (3 × 5 mL), the membrane was incubated with horseradish peroxidase-conjugated secondary antibodies (Cell Signaling Technology) in fresh BSA blocking solution. Immune complexes were detected with the ECL detection reagent (BioRad) and analyzed using a BioRad imager fitted with a chemiluminescence filter. Intracellular ROS using murine polymorphonuclear neutrophils.</p><p>X-ray crystallography. Crystals of 1, 2, 4, and 5 were grown at room temperature by vapor diffusion of diethyl ether into a DMF or DCM solutions of each complex. Suitable crystals were selected by microscopic examination through crossed polarizers, mounted on a fine glass fibre in polyisobutane oil, and cooled to 90 K under a stream of nitrogen. A Bruker D8 Venture diffractometer with graded-multilayer focused MoKα X-rays (λ = 0.71073 Å) was used to collect the diffraction data from the crystal. The raw data were integrated, scaled, merged and corrected for Lorentz-polarization effects using the APEX3 package [96][97][98] . Space group determination and structure solution and refinement were carried out with SHELXT and SHELXL 99,100 , respectively. All non-hydrogen atoms were refined with anisotropic displacement parameters. Hydrogen atoms were placed at calculated positions and refined using a riding model with their isotropic displacement parameters (U iso ) set to either 1.2U iso or 1.5U iso of the atom to which they were attached. The structures, deposited in the Cambridge Structural Database, were checked for missed higher symmetry, twinning, and overall quality with PLATON 101 , an R-tensor 102 , and finally validated using CheckCIF 101 . See Tables S1-S8 for structure details. Of note is the different counter anions (ClO4 − ) for 4 and 5, which make them crystallographically distinct from similar published structures.</p><p>cell lines and cell culture conditions. All ovarian cancer cells (OVCAR8), lung cancer cells (H460), and the leukemia cell line (K562) were maintained in the Roswell Park Memorial Institute (RPMI) 1640 medium.</p>

Scientific Reports - Nature

Core/Shell Magic-Sized Semiconductor Nanocrystals

Magic-sized semiconductor nanocrystals (MSNCs) grow via discrete jumps between specific sizes. Despite their potential to offer atomically precise structures, their use has been limited by poor stability and trap-dominated photoluminescence. Recently, syntheses have been reported that produce CdSe MSNCs over a larger size series. We exploit such particles and demonstrate a method to grow shells on CdSe MSNC cores. Thin CdS shells lead to dramatic improvements in the emissive properties of the MSNCs, narrowing their fluorescence linewidths, enhancing photoluminescence quantum yields, and eliminating trap emission. While thicker CdS shells lead to decreased performance, CdxZn1-xS alloyed shells maintain efficient and narrow fluorescence lines. These alloyed core/shell crystallites also exhibit a tetrahedral shape, in agreement with a recent model for MSNC growth. Our results indicate that MSNCs can compete with other state-of-the-art semiconductor nanocrystals. Furthermore, these core/shell structures will allow further study of MSNCs and their potential for atomically precise growth.

core/shell_magic-sized_semiconductor_nanocrystals

<p>Semiconductor nanocrystals (NCs) exhibit optical properties that depend on their size and shape. 1 Consequently, NC researchers have strived to develop syntheses that produce monodisperse (i.e. identical) NCs. 2 These efforts have yielded high-quality quasi-spherical nanocrystals known as quantum dots (QDs), as well as a variety of other shapes, such as quasi-onedimensional nanorods, 3 quasi-two-dimensional nanoplatelets, 4 and tetrapods. 5 However, despite these advances, the synthesis of monodisperse NCs with perfect uniformity is (perhaps unsurprisingly) still not possible. Thus, samples always exhibit inhomogeneous optical properties.</p><p>For example, even for the best samples, single-NC spectroscopy reveals variations in fluorescence spectra between crystallites. 6 These differences are at least in part due to distributions in size and shape. 7 This leads to a fundamental question: how atomically precise can nanocrystals be? 8,9 One potential route to truly monodisperse NCs (at least in principle) is through magic-sized clusters (MSCs). [10][11][12] While conventional QDs grow by continual addition of atoms, leading to a quasicontinuum of potential sizes, MSCs grow through discrete jumps such that only certain sizes are allowed. 13 Previous studies have indicated that MSCs are tetrahedral in shape, 14,15 and we have recently shown that this can explain their discrete sizes. 16 MSCs grow layer by layer, and the addition of a full layer on any one of the four exposed facets of the tetrahedron can yield the nextlarger tetrahedron in a series. This process involves a kinetic barrier, as a partially completed layer costs energy due to its extra dangling bonds. Upon completion of the layer, these extra dangling bonds disappear, causing a sudden decrease in energy. Thus, "magic" sizes appear due to a series of complete tetrahedra that represent local energy minima with kinetic barriers in between. This rationalizes why only complete tetrahedra are experimentally observed. However, while the existence of MSCs has long been known, 17,18 they have suffered from some disadvantages compared to other NCs, preventing their wider study and application. Notably, the photoluminescence (PL) of MSCs has been dominated by a very broad emission feature with a large Stokes shift from the lowest-energy absorption peak. 19 Because this feature is reminiscent of "deep trap" emission in QDs, it has been presumed to originate from surface traps. 20 Thus, careful tailoring of the MSC surface could lead to improved optical properties. 21 Poor PL due to surface defects is not limited to MSCs, but is a common issue among semiconductor NCs. 22,23 One strategy to eliminate trap states is to coat the particle with a shell. 24 The shell material serves to passivate dangling bonds of the optically active core and isolate it from the surrounding medium. 25 The shell typically consists of a wide-band-gap semiconductor, which helps confine photogenerated charge carriers (electrons and holes) to the NC core. This approach has been applied to many NC shapes and structures, yielding particles with high photoluminescence quantum yields (PLQYs). [26][27][28][29] In the case of CdSe NCs, common shells include other II-VI semiconductors with wider band gaps, such as CdS, ZnSe, ZnS, or their alloys. 27,[30][31][32] Clearly, the same approach could potentially be used to improve the poor PL of MSCs. One issue is that MSCs have been limited to very small sizes, which are typically hard to isolate and study. Fortunately, discrete growth is now possible to larger sizes. 11,14,16 In particular, our group has recently introduced a synthesis that can produce up to 11 discrete species of CdSe crystallites.</p><p>As their sizes extend well beyond the traditional cluster regime, we refer to the entire series as magic-sized nanocrystals (MSNCs). 16 The larger particles can be easily isolated, providing discretely growing CdSe NCs that are stable under ambient conditions for months. 16 They present an opportunity to grow shells on magic-sized particles.</p><p>Herein, we exploit this opportunity to synthesize core/shell CdSe MSNCs. We first isolate a specific MSNC to carefully control the size of our cores. We then deposit a shell to enhance the optical properties of these MSNCs. By growing thin layers of CdS on CdSe, we are able to eliminate the trap emission of the parent cores. This is accompanied by significant increases in the PLQY and narrowing of the linewidth of the band-edge fluorescence. However, growth of thicker CdS shells leads to sharp drops in the PLQY of our MSNCs. We find that the addition of a zinc precursor to our shelling reactions allows CdSe/CdxZn1-xS MSNCs with thicker shells to be obtained. Furthermore, we show that the unique physical properties of MSNCs, as compared to conventional QDs, have a dramatic impact on the properties of the resulting core/shell NCs. The resulting bright, stable core/shell particles also enable study and application of MSNCs as an important nanomaterial class.</p><p>We introduce our process with zincblende CdSe MSNCs that have their lowest-energy absorption feature at 494 nm (MSNC494, Figure 1a). Slight modifications were made to our previously reported synthesis to obtain these cores. 16 (See sections S1 and S2 in the Supporting Information for detailed descriptions of our materials and synthetic methods.) In particular, MSNC494 was synthesized in hexadecane instead of 1-octadecene (ODE) to prevent the formation of poly(ODE), 33 and additional washing steps were carried out. Our MSNC494 cores already exhibit relatively strong, band-edge fluorescence with a small Stokes shift, but trap emission is still visible (Figure 1b). The PLQY of these MSNC494 cores was measured up to 28%, however this number fell to 18% when only the band-edge fluorescence was considered. (See section S3 in the Supporting Information for optical-characterization methods.)</p><p>We have previously observed that CdSe MSNCs can grow from one size to the next in the absence of Se precursors via a ripening process. 16 Thus, to prevent core growth during shelling, which would potentially lead to a mixture of MSNC sizes, the shell should be added at a lower temperature than that used to obtain the core. We can also exploit the fact that progressively higher temperatures are required to grow a MSNC to the next-larger size. Thus, the enhanced thermal stability of our larger MSNCs allowed shelling reactions at relatively high temperatures, which is beneficial for obtaining good shells and high PLQYs. 34 Specifically, we combined cadmium oleate and the MSNC494 cores in ODE and heated the mixture to 180 ºC. At this reduced temperature, formation of poly(ODE) is not expected. 33 A sulfur source in ODE was then continuously added. At 180 ºC, the shelling reaction is sufficiently below the 240 ºC at which MSNC494 was grown that core growth due to ripening is prevented. For the sulfur source, either sulfur or acetyl sulfide can be used, with negligible differences in the optical properties of the final core/shell MSNCs (Figure S1 in the Supporting Information). The only major difference between precursors is that the shell growth is faster with acetyl sulfide than with elemental sulfur (see section S2 in the Supporting Information). Upon completion, the reaction was cooled to room temperature and the product was isolated and cleaned via precipitation three times with methyl acetate followed by redispersion in hexanes, yielding MSNC494/CdS core/shell MSNCs.</p><p>The addition of the CdS shell shifts the absorption spectrum of the parent MSNC494 to longer wavelengths (Figure 1a). Because CdSe and CdS have conduction bands with similar band offsets, a photoexcited electron in a CdSe/CdS core/shell NC can delocalize to the shell, explaining this red shift. 35 Indeed, it has been widely demonstrated that thicker CdS shells on CdSe NCs induce larger red shifts of the lowest-energy absorption feature relative to the CdSe core itself. 36,37 We observed a continuous red shift in the absorption of MSNC494/CdS NCs with shell growth (Figure S2 in the Supporting Information), in contrast to the discrete jumps seen for the MSNC core. This is indicative of continuous shell growth.</p><p>Upon shelling, the fluorescence of MSNC494 dramatically improved. The low-energy trap emission observed in MSNC494 was completely eliminated by the addition of the shell, with MSNC494/CdS only exhibiting band-edge fluorescence (Figure 1b). The PLQY of MSNC494/CdS was also increased, up to 78%. Finally, the fluorescence linewidth of MSNC494/CdS narrowed to 116 meV, compared to 140 meV for the original MSNC494 cores.</p><p>A similar procedure was applied to smaller MSNCs, which exhibit a lowest-energy absorption feature at 434 nm (MSNC434, Figure 1c). Prior to adding a shell, the PL of MSNC434 also showed a band-edge feature, but emission from trap states was more pronounced (see Figure 1d) than in spectra for MSNC494. The total PLQY of MSNC434 was measured up to 38%, but the band-edge contribution was only up to 12%. Because the synthesis of the smaller MSNC434 was at 180 ºC (see section S2 in the Supporting Information), shells were added at 150 ºC to prevent core growth due to ripening. As with MSNC494/CdS, a continuous red shift of the absorption features was seen during shell growth for MSNC434/CdS (Figure S3 in the Supporting Information). The trap emission that dominated the PL spectrum of the MSNC434 cores was almost completely eliminated in MSNC434/CdS (Figure 1d). The band-edge-only PLQY increased to up to 55%, with a narrowing of the fluorescence linewidth from 161 meV in the original MSNC434 cores to 141 meV.</p><p>We analyzed these core/shell MSNCs via transmission electron microscopy (TEM) and scanning TEM (STEM) and measured the projected area of our particles (see section S4 in the Supporting Information). While our previous work suggested that such MSNCs are truncated tetrahedra, 16 we assume below that the MSNC cores and core/shell particles are quasi-spherical for ease of comparison. This allows us to convert the measured projected areas into effective diameters (see section S5 of the Supporting Information). Prior to shelling, MSNC494 had an average effective diameter of 2.67 ± 0.22 nm, which grew to 3.13 ± 0.19 nm in the completed MSNC494/CdS (Figure 2a,b). Prior work has shown that one monolayer (ML) of CdS grown completely around a zincblende CdSe nanoplatelet core (i.e., on both sides) leads to a total thickness increase of 0.61 nm. 38 Thus, our MSNC494/CdS corresponds to the growth of ~0.75 MLs (0.46 nm) of CdS around the MSNC494 CdSe core. In the case of MSNC434, we measured an initial average effective diameter of 2.13 ± 0.29 nm, which increased to 2.52 ± 0.21 nm for MSNC434/CdS (Figure 2c,d).</p><p>This suggests a growth of 0.64 MLs (0.39 nm) of CdS. We note that we did not observe triangular projections in the STEM images, which would be a clear indication of tetrahedral-shaped particles. However, this is not surprising given the small particle sizes. Indeed, our earlier work on MSNC494 suggested faceted particles, but they were too small for a precise shape determination. 16 Here, the addition of the sub-ML-thick shells would potentially further smoothen any surface facets.</p><p>The shell thicknesses extracted from electron microscopy can be further corroborated by experiments using colloidal atomic layer deposition (c-ALD). 39 c-ALD proceeds via self-limiting half reactions of ionic precursors adding to a surface. 40 For example, in the case of cadmium-rich surfaces of CdSe, a layer of sulfur can be deposited, followed by a layer of cadmium to yield 1 ML of added CdS. This process, which is typically carried out at room temperature, has been shown to be efficient, [38][39][40] adding 1 ML of CdS with each full cycle (one layer of S and one layer of Cd).</p><p>Multiple cycles can then yield thicker shells.</p><p>Consequently, c-ALD has been suggested as an alternative approach to core/shell MSNCs. 9 However, we observed much poorer optical properties in our core/shell MSNCs grown via c-ALD when compared to those grown with our high-temperature synthesis. For example, the PLQY of MSNC494 with 1 ML of CdS grown via c-ALD was only up to 7%. Nevertheless, we can use c-ALD to further characterize our high-temperature shell growth. The lowest-energy absorption feature of our MSNC494/CdS is at 514 nm, close to the 518 nm peak position for MSNC494 with 1 ML CdS added via c-ALD (Figure S4 in the Supporting Information). For MSNC434, our hightemperature shell growth yielded an absorption feature at 458 nm, compared to ~478 nm for 1 ML of CdS added via c-ALD (Figure S5 in the Supporting Information). These results are consistent with our conclusion above that the MSNC/CdS samples in Figure 1 have a CdS shell just under 1 ML thick. It has been observed that thicker CdS shells on a CdSe core often lead to higher PLQYs compared to thin shells. 41,42 Therefore, we attempted to grow thicker shells on MSNC494 by modifying the amount of cadmium and sulfur precursors used, as well as by varying the reaction times (see section S2 in the Supporting Information). Absorption and PL spectra were taken at various time points as the shell was added (Figure 3). The absorption peaks of the resulting core/shell crystallites (which we refer to as MSNC494/CdS+) are shifted to even longer wavelengths than those seen for our thin-shelled MSNC494/CdS, consistent with the addition of more CdS. The PL of MSNC494/CdS+ also remains sharp, with trap emission completely absent (Figure 3b, inset).</p><p>However, the PL intensity drops rapidly between 60 and 120 min. The final washed species has a PLQY only up to 10%, much lower than the thin-shelled MSNC494/CdS above. When one monolayer of CdS was grown on MSNC494 via c-ALD, the PL peak was observed at 553 nm (grey dashed line, Figure 3b). The drop in PL intensity that we observe when growing CdS on CdSe MSNC494 occurs as we approach one monolayer of CdS. A similar and even more dramatic effect is seen when a thicker shell is grown on MSNC434 (Figure S6 in the Supporting Information). The PLQY of this species (MSNC434/CdS+) is only up to 3%.</p><p>To investigate why thicker CdS shells lead to poor optical properties, we performed a TEM size analysis. It revealed that the MSNC494/CdS+ have an average effective diameter of 3.73 ± 0.54 nm. This is an increase of 1.05 nm from the MSNC494 core and corresponds to ~1.7 ML of CdS. The MSNC494/CdS+ particles do not appear tetrahedral (Figure S7 in the Supporting Information). If they were tetrahedral, we should observe triangular projections at this size, as we did previously for MSNCs with an effective diameter of 3.27 nm. 16 In addition, the STEM images show particles with enhanced aggregation, as compared to MSNC494/CdS in Figure 2b. In the case of MSNC434/CdS+, we again see particles with no clear triangular projections, with an effective diameter of 3.26 ± 0.84 nm, corresponding to the growth of 1.8 MLs of CdS. Here many long worm-like particle aggregates are visible in STEM images (Figure S8 in the Supporting Information). These data suggest that growth of CdS shells above 1 ML on our CdSe MSNCs leads to a steep drop in PL intensity. We postulate that the observed shape and aggregation of these particles results from uneven shell growth to relieve strain. Due to the lattice mismatch between CdSe and CdS (~4%), the edges and corners of the tetrahedral MSNC cores exhibit increasing strain as the shell growth proceeds. We believe this enhanced strain leads to the poor optical properties in CdSe MSNCs with thicker CdS shells.</p><p>Alloying is a common strategy to reduce interfacial strain in core/shell NCs. [43][44][45] In particular CdxZn1-xS grown on CdSe has led to core/shell particles with excellent optoelectronic properties. 27,46 ZnS has a wider band gap than CdS, leading to better passivation of the CdSe core.</p><p>However, ZnS has a larger lattice mismatch with CdSe (~12%). [30][31][32] The alloyed shell leads to reduced interfacial strain as compared to a pure ZnS shell, while still maintaining the benefits of the wider band gap of ZnS.</p><p>To grow a CdxZn1-xS shell on our MSNC494, we modified our earlier CdS procedure. Specifically, we added a zinc precursor (zinc oleate) and oleylamine along with the cadmium precursor (cadmium oleate); elemental sulfur was used as the sulfur source (see section S2 in the Supporting Information). These precursors were previously exploited to obtain CdxZn1-xS shells on QDs and nanoplatelets. 29,47,48 As with the growth of CdS on MSNCs above, we observed a gradual red shift of the absorption features as the reaction proceeds, indicative of continuous shell growth (Figure S9 in the Supporting Information). After the addition of the shell, the isolated crystallites exhibited a lowest-energy absorption peak at 534 nm, with strong PL (up to 80% PLQY) without trap emission (Figure 4a). This ~40 nm red shift in the lowest-energy absorption feature is larger (smaller) than that expected for a shell of pure ZnS (CdS). 39,49,50 Thus, this observation is consistent with the addition of an alloyed shell. TEM images of these MSNC494/CdxZn1-xS confirmed that we had grown thick shells on our initial MSNC494 cores (Figure 4b). The average effective diameter of MSNC494/CdxZn1-xS was 5.16 ± 0.47 nm, an increase of 2.49 nm from the MSNC494 core. Most strikingly, the MSNC494/CdxZn1-xS crystallites clearly exhibit triangular projections in STEM, as we would expect for tetrahedral-shaped particles. Assuming that they are indeed tetrahedral, we can determine effective edge length. Converting from the projected area (section S5 in the Supporting Information) gives us an effective edge length of 6.95 ± 0.63 nm. To determine whether this tetrahedral shape is a result of growing on a tetrahedral core or due to the shelling conditions, we synthesized zincblende QDs, 51 which are known to be quasi-spherical. In particular, we grew QDs with a lowest-energy absorption feature at 494 nm (QD494, Figure S10 in the Supporting Information), matching the peak position of MSNC494. We then added CdxZn1-xS shells on these QD494 with the same method as with MSNC494/CdxZn1-xS. The resulting QD494/CdxZn1-xS shows similar optical properties to MSNC494/CdxZn1-xS, with longer-wavelength absorbance and PL without trap emission (Figure S11 in the Supporting Information). However, TEM images of QD494/CdxZn1-xS reveal that these particles have irregular shapes (Figure S12 in the Supporting Information) without discernable triangular projections. Thus, we conclude that the triangular projections observed for MSNC494/CdxZn1-xS are due to the shape of the underlying MSNC494 core.</p><p>We note that while previous data has suggested that small CdSe MSNCs are tetrahedral, 14,15 it had been unclear if this holds for all MSNCs. 9,13 Thus, in addition to enhanced optical properties, our core/shell MSNCs can give us details about structure of the underlying MSNC core.</p><p>In conclusion, we have demonstrated the synthesis of core/shell crystallites grown from CdSe MSNCs. Thin CdS shells dramatically improve the optical properties of MSNCs, leading to an elimination of low-energy emission from trap states, a decrease in the fluorescence linewidth, and large increases in PLQY. Analysis of these particles via TEM shows these thin CdS shells are below 1 ML thick. Thicker CdS shells can also be grown, but this leads to diminished PLQY. TEM of particles with thicker shells reveals quasi-spherical shapes that tend to aggregate. To MSNCs with thick shells with high PLQY, we grew CdxZn1-xS shells. Unlike our MSNCs with thick CdS shells, our MSNC494/CdxZn1-xS species is clearly tetrahedral. These results demonstrate a route to overcome the poor stability and poor optical properties that have plagued MSCs.</p><p>Furthermore, these core/shell structures will enable the fundamental structural and optoelectronic properties of MSNCs to be studied, expanding our understanding of this special class of nanomaterial. 3.13 ± 0.19 nm, (c) 2.13 ± 0.29 nm, and (d) 2.52 ± 0.21 nm. The increase in size corresponds to a ~0.75 ML thick CdS shell for MSNC494/CdS and ~0.64 ML for MSNC434/CdS. Scale bars are 10 nm. We note that while strategies to minimize electron-beam damage in STEM images were followed (see section S4 in the Supporting Information), the images are still potentially affected. shows the distribution of effective edge lengths of 603 particles assuming equilateral triangles. The scale bar is 20 nm. We note that while strategies to minimize electron-beam damage in STEM images were followed (see section S4 in the Supporting Information), the images are still potentially affected.</p>

ChemRxiv

Enantioselective Construction of Trialkyl Tertiary Centers via Ni-Catalyzed Markovnikov Hydrocarbofunctionalizations of Unactivated Olefins and Electrophiles

Routes to efficient construction of fully aliphatic substituted tertiary chiral centers are highly challenging and desirable. Herein, a Ni-catalyzed enantioselective hydroalkylation of unactivated alkenes at room temperature is described, providing a general and practical access to tertiary stereogenic carbon centers with three alkyl substituents. This reaction involves the regio-and stereoselective hydrometalation of unactivated alkenes with Markovnikov selectivity, followed by coupling with unactivated alkyl electrophiles to access tertiary chiral centers with full alkyl substituents. The mild and robust conditions enable the use of terminal and internal unactivated alkenes as well as primary and secondary unactivated alkyl, benzyl, propargyl halides for the construction of diverse trialkyl tertiary stereogenic carbon centers with broad functional group tolerance.

enantioselective_construction_of_trialkyl_tertiary_centers_via_ni-catalyzed_markovnikov_hydrocarbofu

<!>ASSOCIATED CONTENT

<p>Carbon-Carbon bonds between sp 3 -hybridized carbon centers compromise the major portion framework of organic molecules. 1 Thus, great efforts have been attracted to build the saturated stereogenic carbon centers and to avoid flat molecules, which play a crucial rule in chiral auxiliaries, pharmaceutical agents, natural products and bioactive molecules. 2 Transitionmetal-catalyzed enantioselective cross-coupling of alkyl halides with alkyl metal nucleophiles is a powerful and established strategy for sp 3 -hybridized C-C bonds forming (Scheme 1a, left). 3,4 This strategy requires the use of stoichiometric amount ofgennerally reactive and sensitive organometallic reagents, which usually require time-consuming preformation. 5 Alternatively, reductive cross-coupling would be an appealing alternative due to the mild conditions and avoiding use of metallic reagents. Specifically, cross-coupling between two alkyl electrophiles using chiral metal catalyst under reductive conditions to forge alkyl-alkyl bonds with good levels of enantiomeric excess is a promising alternative to construct saturated carbon centers yet challenging (Scheme 1a, right). 6 On the other hand, olefins are among the most important feedstock to synthesize value-added targets due to their easy accessibility and diverse reactivity profiles. 7 Consequently, catalytic asymmetric intermolecular carbofunctionalizations of olefins have been emerging as an attractive strategy to access stereogenic carbon centers by constructing C-C bonds to increase saturation of molecules. 8 Among these transformations, enantioselective reductive C-C bond forming processes involving olefins are particularly attractive. 8a,9 Recently, significant progresses have been achieved on the reductive enantioselective carbofunctionalizations of activated olefins, 10 including alkylarylation, 10a-c hydroarylation, 10d-f and hydroalkylation 10g-k of activated olefins. The reductive enantioselective hydrofunctionalizations of unactivated olefins remain a formidable challenge because of their low reactivity and poor selectivity control issues. 11 Seminal Scheme 1. Construction of Saturated Tertiary Stereogenic Centers.</p><p>work from Buchwald disclosed the anti-Markovnikov hydroamination of unactivated olefins to deliver linear alkyl amines by copper catalysis. 11b,12 Catalytic hydrofunctionalization of unactivated of olefins in Markovnikov selectivity remains elusive due to the repulsion of ligated metal center with alkyl chain (Scheme 1b). 13 To date, only a few examples for racemic transformations of Markovnikov hydrofunctionalization of unactivated olefins are developed. [14][15][16] In 2020, Hong reported a racemic Ni-catalyzed Markovnikov hydroamination of unactivated olefins assisted by 8-aminoquinoline (AQ) amide group, which could suppress β-H elimination of alkyl-metal species by its rigid and saturated coordination. 14 In particular, catalytic Markovnikov hydrocarbofunctionalizations of unactivated olefins are attractive by introducing one stereogenic carbon center not adjacent to heteroatoms or carbonyls. In 2020, Koh and Wang reported Ni/Mn-catalyzed hydroarylation/alkylation of unactivated olefins in a racemic fashion facilitated by a bidentate 8-aminoquinoline. 15 No example of intermolecular, enantioselective, Markovnikov hydrofunctionalization of unactivated olefins has been reported (Scheme 1c). As part of our continuous interests in asymmetric alkyl-alkyl cross-coupling, 10g,k we envisioned developing an enantio-and Markovnikov-selective hydrocarbofunctionalization of unactivated olefins. To address the challenging enantioselective Csp3-Csp3 coupling of both electrophiles nonadjacent to an activating group (aryl, vinyl, carbonyl, heteroatom), we reported a regio-and enantioselective hydroalkylation, hydrobenzylation, hydropropargylation of unactivated olefins with unactivated electrophiles in the presence of a weak coordinating and removable directing group.</p><p>To test the feasibility of the proposal, we commenced to identify the reaction parameters using terminal olefin 1a and 1-iodo-2-phenylpropane 2a as prototype substrates in the presence of silane and base. After extensive preliminary optimization, we chose NiBr2ꞏglyme (10 mol%) as the nickel catalyst precursor, (MeO)3SiH as hydride source and potassium phosphate monohydrate as base in THF (0.1 M) for further evaluation (Tables S1-10). Then, a wide range of chiral ligands were tested for this reaction (L1-L8). To our delight, using the chiral BOX ligand L1 with a methyl substituent at 5-position of oxazolidine ring could catalyze the regio-and enantioselective hydroalkylation reaction of unactivated olefin, delivering the Markovnikov hydroalkylated product 3a in 86% yield with 80% ee. Only trace amount of anti-Markovnikov hydroalkylated regioisomer 3a' was detected. Increasing the steric hindrance at 5-position of the chiral ligand from methyl to propyl improved the enantioselectivity of the desired product (L1-L3), furnishing 3a in 79% yield with 86% ee with L3 as anchoring ligand. Further increasing the size of the substituent at 5-position of the BOX ligand (L4) delivered 3a in diminished yield with identical enantioselectivity. Alternating the substituent R 2 on BOX ligand from methyl to ethyl (L5), n-propyl (L6), benzyl (L7), or 4-tert-butylbenzyl (L8) led to lower yields and enantioselectivities. Further evaluation the solvent effect of this reaction disclosed that 3a was obtained in 72% yield with 91% ee in dioxane (0.1 M). Table 1. Condition Evaluation of the Reaction a a The reaction was conducted using 1a (0.1 mmol) and 2a (0.2 mmol) in the presence of (MeO)3SiH (0.6 mmol) and potassium phosphate monohydrate (0.6 mmol) in 1 mL of solvent under indicated conditions for 10 h unless otherwise stated. Yield was determined by GC using n-dodecane as internal standard. Isolated yield is shown in the parentheses. The enantiomeric excess was determined by HPLC using a chiral stationary phase. b NiBr2·glyme (5 mol%), (R,R)-L3 (6 mol%) were used. c The reaction was run in dioxane (0.05 M).</p><p>Testing the effect of additives revealed that the use of N-methyl-4-trifluoromethylphenylsulfonamide (A1, 30 mol%) increased the yield of 3a to 81% with 92% ee. Decreasing the loading of A1 could further increase the enantioselectivity of 3a without erasing the efficiency of this reaction. After routine optimization, we defined the use of NiBr2·glyme (5 mol%) and (R,R)-L3 (6 mol%) as catalyst, A1 (6 mol%) as additive in the presence of (MeO)3SiH (0.6 mmol) and potassium phosphate monohydrate (0.6 mmol) in dioxane (0.05 M) as standard conditions, affording the desired product 3a in 86% yield with 96% ee. 16 With the optimized conditions in hand, we turned to evaluate the scope of this reaction. The protocol tolerated a wide variety of functional groups and substitution patterns for this enantioselective Markovnikov hydrocarbofunctionalization of unactivated olefins (Figures 1 and 2). In general, this reaction afforded tertiary stereogenic carbon center with full alkyl substitution in high effeciency with excellent regioselectivity. First, we tested the scope of unactivated olefins (Figure 1). N-Aryl 3-enamides with electron-withdrawing or electrondonating groups on the aromatic ring were well-tolerated under the standard conditions, undergoing Markovnikov hydroalkyation in good yields (58-83% yield) with excellent regio-(rr = 8:1 to 26:1) and enantioselectivities (87-96% ee) (3b-3m). Notably, aryl halides, such as fluorides and iodides were compatible in the reaction, delivering the desired products (3g and 3h) in 83% and 64% yields with 93% and 92% ee, respectively. Unsaturated functional groups, including ester (3j), ketone with acidic protons (3k), nitrile (3l), were compatible in the reaction, furnishing corresponding products (3j-3k) in 67-79% yields with 87-95% ee. Heteroaromatic aniline derived olefin underwent the desired transformation smoothly, Table 2. Scope for the Reaction in Terms of Olefins a 6k, 87%, 96% ee, rr > 20:1 6p, 51%, 97% ee, rr > 20:1 6q, 74%, 96% ee, rr = 20:1 6f, 43%, rr = 6:1 dr = 1 (91% ee):1 (91% ee) d 6m, 68%, 80% ee, rr = 9:1 6n, 73%, 80% ee, rr = 8:1 6j, 76%, 94% ee, rr > 20:1 6l, 80%, 93% ee, rr = 11:1 6i, 72%, 92% ee, rr = 7:1 a Standard conditions, see Table 1 for detail. Alkyl iodide was used unless otherwise noted. rr = ratio of branched product and linear product. rr was determined by GC. b The reaction was conducted on 2.0 mmol scale. c 3.0 equiv of alkyl halide was used. d NiBr2‧glyme (10 mol%), (R,R)-L3 (12 mol%), A1 (50 mol%) were used.</p><p>delivering 3n in 55% yield with 87% ee. Alkyl amine based amide tethered olefins could be tranformed into corresponding amides with a trialkyl substituted tertiary stereogenic center in synthetic useful yields with 73% ee (3o and 3p). Notably, internal olefins were also good substrates for this reaction, giving diverse trialkyl steregenic carbon centers in good yields (70%-80%) with 92-95% ee (4a-4d), representing one of the few examples for transition metal-catalyzed enantioselective conversion of unactivated internal olefins with high levels of enantiocontrol. 11b Next, the scope of organic halides for this reaction was evaluated (Table 3). A wide variety of alkylhalides (including iodides and bromides), benzylchlorides, and propargyl bromides are tolerated in this reaction, delivering a myriad of enantioenriched trialkyl tertiary stereogenic carbon center in high efficiency with excellent levels of enantioselectivity. First, primary alkyl halides were tested. 2-Phenyl-1-iodopropane was converted to trialkyl chiral amide 5a in 82% yield with 95% ee. The reaction worked well on 2.0 mmol scale, affording 5a in 73% yield with 97% ee. Linear and α-branched alkyl iodides could be transformed into corresponding amides (5b-5g) in 65%-84% yields with 94%-96% ee. Functional groups, such as esters, amides, ethers, silyl ethers, acetals, and nitriles were compatible in the reaction, delivering the desired hydroalkylation products (5h-5o) in 67%-82% yields with 83%-95% ee. Ketone with acidic protons could be tolerated to give corresponding hydroalkylation product 5p in 62% yield with 92% ee. Amides with free N-H bond were also tolerated in the reaction, delivering desired products (5q and 5r) in synthetic useful yields with 93% and 97% ee. Free alcohol could be tolerated in the reaction, giving desired product 5s in 54% yield with 90% ee. Heterocycles, such thiophene, indole, derived alkyl iodides were transformed to desired products (5t and 5u) in 72% and 83% yields with 94% and 95% ee, respectively. Moreover, α-bromoesters could be coupled in the reaction to give desired products (5v and 5w) in 76% and 84% yields with 88% and 90% ee, respectively.1,1,1-Trifluoro-3-iodopropane and (iodomethyl)trimethylsilane were successfully converted to corresponding products (5x and 5y) in 69% and 65% yields with 96% and 87% ee. Next, secondary alkyl halides were tested. Interestingly, cyclic and acyclic secondary alkyl halides were all good substrates for this reaction (6a-6g). Carbon, nitrogen or oxygen tethered cyclic secondary alkyl iodides could be transformed to corresponding products (6a-6d) in moderate to good yields with 90-95% ee. 2-Iodopropane was coupled to give the hydroalkylated product 6e in 54% yield with 93% ee. Unsymmetrical acyclic iodides were successfully involved in the reaction to give the desired products (6f and 6g) in moderate yields with the same level of enantioselectivity for both diastereomers (91% ee), indicating the facile control of vicinal saturated carbon centers from two unactivated racemic electrophiles. 4h,17 Next, benzyl halides were tested for this reaction. It is found that benzyl chlorides with electron-donating or electron-withdrawing group could undergo Markovnikov hydrobenzylation to give corresponding products with a trialkyl substituted tertiary stereogenic center (6h-6k) in 72-87% yields with 91-96% ee. Heretoaryl containing benzyl chlorides were well-tolerated under the reaction conditions, giving desired products (6l and 6m) in 80% and 68% yields with 93% and 80% ee, respectively. Furthermore, propargyl bromides were examined for this reaction. Methyl, phenyl, and trimethylsilyl substituted propargyl could be coupled to give the desired hydropropargylation products (6o-6q) in moderate to good yields with 91%-97% ee, representing the first example of enantioselective hydropropargylation of unactivated olefins.</p><p>Figure 1. Application of the reaction for late-stage functionalization using alkyl iodide. For reaction conditions, see Table 1 for detail. rr = ratio of branched product and linear product. rr was determined by GC. a NiBr2‧glyme (10 mol%), (R,R)-L3 (12 mol%), A1 (50 mol%) were used.</p><p>To demonstrate the robustness and usefulness of this protocol, we applied this condition to late-stage functionalization of complex molecules, including natural product and drug molecules (Figure 1). derivative could be tolerated to give construct the trialkyl tertiary carbon center 7a in 65% yield with 98:2 dr. Theobromine derived alkyl iodide was converted to corresponding product 7b in 69% yield with 96% ee. Menthol was also compatible under the reaction conditions, providing menthol containing product 7c in 76% yield with 97:3 dr. Drug molecules, such as isoxepac and oxaprozin, were transformed into corresponding products 7d and 7e in 70% and 65% yields with 94% and 95% ee. Moreover, Indomethacin could be incorporated in the reaction to deliver 7f in 70% yield with 92% ee.</p><p>To further prove the synthetic potential of this protocol, the enantioenriched β-chiral amide product was employed to convert to other functionalized compounds (Figure 2). First, enantioselective γ-chiral amine 8a was obtained from 5a (97% ee) in 81% yield with 97% ee. Second, 5a could be transformed into γ-chiral alcohol 8b in 74% yield with 98% ee. Moreover, β-chiral carboxylic acid 8c could be furnished from 5a in 79% yield with 96% ee. Furthermore, 3a was converted to unfunctionalized enantiopure alkane 8d with a trialkyl tertiary stereogenic carbon center in 65% yield with 95% ee, providing a route to pure carbon chiral molecules challenging to access otherwise. The aforementioned derivatizations render this protocol amenable to diverse chiral compounds bearing different functional groups with excellent levels of enantioselectivity. To gain insight into the reaction process, we carried out a series of experiments to probe the reaction mechanism. First, reactions using deuterated silane were conducted (Fig. 3a). The reaction of terminal olefin with deuterated silane (PhSiD2) under otherwise identical to standard conditions afforded deuterated hydroalkylation product 9 in 70% yield with 95% ee. Deuterium incorporation (>94% D) was exclusively delivered to γposition of 9. No deuterium incorporation was found at β-or αposition of 9. The reaction of internal olefin with deuterated silane (PhSiD2) delivered the desired product 10 in 63% yield with 89% ee in single diastereosisomer. The results indicated that Ni-H insertion onto olefin to form alkyl-Ni species In summary, a unified protocol for Ni-catalyzed regio-and enantioselective hydrocarbofunctionalizations of unactivated olefins with organohalides under mild conditions has been developed. The use of a modified chiral bisoxazolidine ligand enables the Markovnikov hydroalkylation, hydrobenzylation, and hydropropargylation of unactivated olefins in the presence of a removable and transformable directing group in good yields with excellent levels of enantioselectivity, providing a straightforward access to fully alkyl substituted tertiary saturated carbon centers which are otherwise challenging to access.</p><!><p>General procedures for the enantioselective hydrocarbofunctionalizations of olefins, condition optimization, characterization of new compounds, X-ray data of 5e (CCDC 2054450), copies of NMR spectra, and HPLC traces (PDF). The Supporting Information is available free of charge on the ACS Publications website.</p>

ChemRxiv

The waste ban in China: what happened next? Assessing the impact of new policies on the waste management sector in China

The 2017 ban on the waste import and new policies for the waste management sector in mainland China had wide-spread impact. After decades of poor environmental and public health impacts from the sector, a study is needed which focuses on policies updates and waste management. This provides a direction for the survival of local waste management industries and consider similarities with the ban promulgated in China on the restriction of waste import from other countries. We review the waste management situation in China before national legislation prevented the import of waste, highlight the status of landfill mining in China, and review the dynamics of domestic policies before and after the promulgation of the ban in China. The impact of the COVID19 pandemic on the waste management system is starting to emerge, providing both challenges and opportunities for the sector in China. We see the impact of the ban on the range of imported waste and domestically generated materials. The ban results in price increases for domestic recycling that forces companies to introduce more formal recycling processes and to drive the consumption behaviours to more reasonable and environmentally friendly options. The driver in China is to reduce pollution in the environment and improve health, but a negative impact has been from increased landfill mining which has impeded the original aim of the waste ban and requires further technological development. The dynamic of domestic policies in China shows higher level of activity of updates and revisions or introduction of new policies from 2015 onwards and the concept of ‘zero waste cities’ brings new hope for improvement of the Chinese waste management system. The pandemic also suggests an important step to establish sustainable management systems despite evidence of increased “fly-tipping”. The rebound of the waste ban may have stimulated in the short term negative impacts on local environments both in China and internationally.

the_waste_ban_in_china:_what_happened_next?_assessing_the_impact_of_new_policies_on_the_waste_manage

Introduction<!><!>Introduction<!><!>Changes in waste import in the run-up to the ban<!>Reasons why the ban started<!>The status of landfill mining in China<!>The dynamics of domestic policies in China<!>Challenges and opportunities for waste management companies<!>Impact of the COVID19 pandemic on waste management system<!>Conclusion and recommendations<!><!>Author contribution<!>Funding<!>Data availability<!>Conflict of interest<!>Human or animal rights<!>Consent for participate<!>Consent for publication

<p>Waste management is still a challenging global issue because of a rapid increase in the amount of waste produced as a result of rapid economic growth, development of technology, population growth, and overconsumption. Around 2.01 billion tonnes of waste was generated worldwide in 2016, and the waste generation is expected to increase to 3.40 billion tonnes annually by 2050 (Kaza et al., 2018). Solid waste is typically recycled in developed countries, exported to developing countries, and has provided a source of income as a resource. However, a ban on the import of foreign waste and reform of the administrative system for solid waste was announced by the Chinese government in 2017. This raises a number of questions: What has the impact of the solid waste ban in China been on the internal waste management industry? Has the ban achieved its target of environmental protection and improvement of human health? Will the approval and acceptance of applications for the import of solid waste finally be stopped from 2021 (Creech, 2020)?</p><p>China's solid waste ban not only changed the global recycled solid waste supply chain, which diverted solid waste to other markets, like Malaysia and other parts of Southeast Asia (Tran et al, 2021); but also led to the construction of new infrastructure for solid waste management in these new destination countries to provide the capacity to handle sudden growth in waste streams (Hook & Reed, 2018). In addition, the diverted plastic scrap appears to be handled by small-scale waste processors, operating under little to no environmental regulations (Hook & Reed, 2018). It also increases the flow of potentially hazardous waste, for example, Waste Electrical and Electronic Equipment (WEEE) into Thailand that recycled the waste into new plastic materials (Hook & Reed, 2018). As a result, Malaysia, Vietnam, and Thailand are due to freeze the import of solid waste, which will lead to major waste producers such as the USA, UK, Japan, and Australia facing the challenge to establish alternative mechanisms to deal with the solid waste internally (WMR, 2018).</p><p>The introduction of the ban has locally initiated a positive impact on the environment in China and, with an increased demand for long-term sustainable development (United Nations, 2015), has resulted in waste management companies switching their target material or stopping their operations in the short term (Qu et al., 2019). However, there are challenges for the survival of companies that use solid waste plastic as resource material in China: a lot of small groups and some waste management companies relying on imported waste as a feedstock have had to stop trading due to a shortage of raw material (Schulz, 2020) while some waste management companies moved to other countries to continue their business after the ban (Parajuly & Fitzpatrick, 2020). Immediate and direct impacts of the Chinese import ban have highlighted that an assessment for evaluating shipment policies linked to waste management is still required for a more comprehensive long-term impacts rather than the short-term economic benefits (Parajuly & Fitzpatrick, 2020).</p><!><p>Comparison of resale prices (2015 and 2018) for waste paper for stakeholders in the paper recovery supply chain in Beijing, China (Yang et al, 2020)</p><!><p>There has been a significant gap in the strength of environmental policies and regulations between a developed country and developing countries like China (as per United Nations definitions) which makes waste trade possible based on the difference in regulations and financial costs in each country. The evolving Chinese waste management system is separated as municipal solid waste (it is also called domestic garbage) management, industrial solid waste, and hazardous waste. There has been a rapid increase in Chinese domestic waste production over the last two decades with the increase of municipal solid waste (MSW) from 214 million tons (135 million tons in Chinese cities and 79 million tons in counties) in 2001 to 311 million tons (242 million tons in Chinese cities and 69 million tons in countries) in 2019 (MHURD, 2020). The impact of the ban on domestic Chinese and global waste flows has been analyzed (Huang et al., 2020) using (i) An structural path analysis combined with a multi-regional input–output model to know how the consumption patterns drive plastic waste import to China, (ii) A ecological network analysis to identify dominant controller in the global plastic waste trade network, (iii) A hypothetical extraction method to understand the added value for China and increased waste treatment capacity requirement for other economies. The results indicated that there was a lack of recycled plastic material due to the import ban, which accounted for 28.1% of domestic plastic waste recycling. The key controllers of plastic waste flow are China, the US, Germany, and the wider EU, and it is difficult for other large economies to replace China's role within the trade network in the short term. This has a higher relative impact on the developed countries, which need to rapidly increase their waste treatment capacity than for the situation in China (Huang et al., 2020).</p><p>Environmental impacts from mechanical recycling of waste plastics, incineration, and landfill with municipal solid waste were evaluated with a life-cycle assessment (Chen et al., 2019). This demonstrated the environmental benefit of the current treatment in end-of-life (EOL) waste plastics through the analysis of current recycling technologies and the impact of the ban. It was found that the ban decreased the transport distances of waste plastics, which thereby reduces related environmental impacts such as a reduction of 84.8% marine eco-toxicity potential (Chen et al., 2019). A challenge is that more than 95% of the labelled plastic is associated with WEEE, which creates additional hazards when mechanical recycling is used (Hook & Reed, 2018). The wider environmental impact is influenced by a number of important factors: (i) the percentage of the recyclable plastic within the imported waste, (ii) whether it is recycled or goes straight to landfill, and (iii) illegal import (smuggling) of WEEE in containers falsely documented as containing plastic waste (Hook & Reed, 2018).</p><p>It is quite reasonable to assume that China announced the ban to protect the domestic environment and improve human health, and a number of studies have been carried out to assess the impact. However, the positive effect is negated because of a number of internal concerns. Growing Chinese urbanisation produces a lot of constructions and demolition waste which is sent to landfills, therefore there is an increasing need for landfill space. This waste is also secretly, and illegally, dumped in the countryside ("fly-tipping"), becoming more common in many jurisdictions during the covid-19 pandemic, with a direct impact on the environment and citizens' health. Data on this is relatively poorly developed in the region (Lee et al, 2021). However, to reduce this environmental impact, new regulations are necessary to force further processing with only residual waste to be landfilled. We provide a perspective on issues impairing the positive impact of the waste ban in a domestic context, considering the landfill mining situation in China, urbanisation, and policy change.</p><!><p>Import data for the main recycled resources to the PRC for the period 2014–2020 (CMC, 2016–2019, Intracen, 2021)</p><p>The number of main categories of recyclable waste recovered in PRC between 2014 and 2018 (unit: × 106tons) (CMC, 2016, 2017, 2018, 2019)</p><p>aThe amount of the iron and steel scrap recovered by small and medium iron and steel enterprises and the amount of scrap steel used in the foundry and forging industries have been included in data since 2014</p><p>bThe amount of waste zinc recovered from hot galvanised slag, zinc ash, flue ash, gas mud ash has been included in the statistical scope since 2014</p><p>Main categories of recyclable waste generated in PRC and value of recovered material between 2014 and 2018 (unit: billion Yuan) (CMC, 2016, 2017, 2018, 2019)</p><p>Internal prices of main categories of recyclable waste generated in PRC before and after the ban (unit: Yuan/t)</p><p>Evidence of increased illegal dumping of domestic and Construction and Demolition Wastes during COVID-19 (photo credit Na Song 2020)</p><p>Frequency of revision of national laws and regulations in the PRC (1991–2020)</p><p>Time series analysis of the amount of solid waste produced for incineration in a southern city of the PRC (2016–2020)</p><!><p>As shown in Fig. 2, the import of iron and steel scrap and non-ferrous metal scrap both show a consistent decrease from 2014 to 2016, with a slight increase from 2016 to 2017, but then a significant decrease of 42.2% and 30.5%, respectively in 2018. The amount of iron and steel scrap imported was 18.4 × 10 kilo-tons, 2.7 × 10 kilo-tons in 2019 and 2020, respectively (Intracen, 2021), with a decrease of 86.3% and 85.3% compared to the previous year. Whilst the amount of non-ferrous metal scrap imported was 294.9 × 10 kilo-tons and 198.6 × 10 kilo-tons in 2019 and 2020, respectively, with the decrease of 26.2% and 32.7% compared to the previous year. The waste plastic shows a relatively gentle decrease of over 30% from 2014 to 2017, with a reduction in import of more than 99% by 2018 (583 × 10 kilo-tons to 5.0 × 10 kilo-tons), then almost zero in 2019 and 2020. The waste paper imports increase slightly from 2014 to 2015, then decline slowly from 2015 to 2017 with more than 33.8% reduction in 2018, then decline again from 2018 to 2020 with reduction of 35.7% and 37.05% compared with those in the previous year. For the import of iron and steel scrap, non-ferrous metal scrap and waste paper, there are still permitted licences application for companies to import higher levels than standard during the following two years until 1st January 2021 when the Ministry of Ecology and Environment of the People's Republic of China canceled licenced permission acceptance and permission of permitted licence application (MEE, 2018, 2020).</p><p>It should be noted that the ban on imports was not introduced suddenly, rather increasingly strict pollution prevention measures have been implemented since 1996 (Sun, 2019). The import of waste plastic reveals the biggest reduction from 2014 to 2020 with a complete import ban on 1st January 2021 according to (MEE, 2020). Besides, in order to control plastic pollution, an ambitious five-year plan was released on 18th January 2020 to ban or restrict the production, sales, and use of environmentally unfriendly plastic at the Chinese national level (State Council, 2020) which means that regular plastic demand will be cut by 2025 across the whole country. The amount of iron and steel scrap and waste paper imported were categorised in 'Catalogue of Restricted Import Solid Wastes that Can Be Used as Raw Materials' (MEE, 2019) and as no licences are being issued in 2021, these products are currently banned. However, it should be noted that import licences may be permitted in the future. The import of non-ferrous metal scrap, higher standard copper-containing waste, and aluminium-containing waste continues at the time of writing (July 2021).</p><!><p>One of the main reasons for the ban on waste import is the serious environmental contamination and the associated human health issues derived from handling Waste Electrical and Electronic Equipment (WEEE) imports. For example, prior to the ban, 70% of the WEEE worldwide was imported by China through various paths. Guiyu Guangdong on the southeast coast of China was the biggest site to deal with electrical waste in China, with over a 20-year history of electrical waste recycling. There are fewer than 200 companies approved for e-waste recycling, which means it is impossible to deal with a huge volume of waste in the market. Greenpeace has conducted environmental surveys in Guiyu and surrounding villages since 2005. The test results of these samples consistently show that soil barium (Ba) concentrations in the villages are 20 times that of the background; chromium (Cr) and lead (Pb) content exceeded that of the Chinese standard (Environmental quality standard for soils (GB15618-1995) by > 1,800 times and > 2,000 times, respectively (Xu, 2018).</p><p>The mean values of blood lead levels of children in Guiyu were 144.3 ± 69.3 µgL−1, whilst 69.9% of children were considered of high lead burden with values of their blood lead levels exceeding 100 µgL−1 (Liu, 2009). In addition, organic pollutants such as Polybrominated Diphenyl Ethers) are also widely distributed in Guiyu (Leung et al., 2007) (Li et al., 2019). Local water is so contaminated that residents could only rely on tap water from other places or bottled water. Another reason is that a large amount of unreported WEEE may be illegally shipped to developing countries like China where valuable component materials are recycled inappropriately, impacting directly on human health and the environment (Geeraerts et al, 2015).</p><p>One more reason for increasingly stringent regulation is the increasing amount of municipal waste from daily domestic life generated for recycling. The waste arising (mass) of materials commonly recycled for the period 2014–2018 is shown in Table 1 with data presented by value in the domestic market in Table 2 below.</p><p>From Table 1, it is apparent that whilst the internal recyclable waste is still substantial, the reduction in imported materials has impacted on waste paper associated with industrial activity, indicating a shortage from 2018. Iron and steel scrap showed an increase from 2015 to 2018, with the similar trend for iron and steel scrap of large iron and steel enterprises whilst the iron and steel from other industries indicated a decrease in the supply chain since 2017.</p><p>Finally, the economic development is another reason to promote the implementation of the ban. For example, the import of plastic wastes is a method to mitigate the shortage of resources in China with the plastic wastes being utilised as secondary material from the mid-1990s to the early 2010s when economic development was increased and raw materials demand was rising thanks to Chinese opening-up policy (Shi et al., 2021). Enterprises that engaged in solid waste recycling had been offered tax refunds since 1995 by the Chinese government for economic reasons (Shi et al., 2021). However, with the development of the Chinese economy, the governmental budget to deal with land, water, and air pollution increased; the pressure and costs of environmental remediation have continually increased for the Chinese Government and have exceeded the benefit of waste importation, therefore resulting in the promulgation of the import ban in 2017.</p><p>There are only a few companies that get involved in the domestic waste recycling market due to the low revenue streams from domestic recycling, the high cost of wages, the chaotic situation of the domestic waste management system, and the challenge of cheaper 'secondary resource' material from waste import. As shown in Fig. 3, prices of main categories of recyclable waste generated in PRC all increased from 2016 to 2018 that might stimulate improvement of the domestic recycling. The import ban is a strategy partially to encourage more people to get themselves involved with domestic recycling and partially to invest more research and development money within waste recycling industries.</p><!><p>Many formal landfill sites were already full in advance of the introduction of new waste import laws. This created significant pressure on the waste management sector for the Chinese Government. A catastrophic landslide event occurred in Shenzhen city, China in 2015 because of overfilling of the landfill, injured 17 people and killed at least 73 people with an estimated total economic loss of more than 0.13 billion USD (Yang et al., 2017). To solve the problems of high cost and limited land to develop a new landfill, landfill mining provides an opportunity to relieve the pressure (Zhou et al., 2015). The import ban aims to reduce the amount of solid waste entering the country, therefore, decreasing the quantity of waste going to landfills, enhancing the environmental footprint and protecting human health. However, landfill mining brings new challenges as the nature of the waste within old landfill sites may be problematic to handle. For instance, excavated plastic bags have impurities even after normal cleaning techniques, this will hinder effective recycling of aged plastic wastes or their use within energy from waste processes such as hydrogenation, gasification, and pyrolysis (Zhou et al., 2014). A further threat may be the most practical processing method of landfill mining plastic wastes is incineration or being treated as residue-derived fuels for energy recovery (Zhou et al., 2014). Moreover, Construction and Demolition Waste (CDW) accounts for 5% of material besides soils (87%) and plastics (2.1%) in the excavated landfills (Hölzle, 2019), which also shows potential problems because of energy consumption as well as emissions resulting from operations, such as excavation, processing and transportation in landfill mining. A report in 2013 shows that about 1 billion tonnes of CDW are generated in China annually, which was five times the amount of municipal solid waste produced in China the same year; the reused and recycled CDW was only 5% (Duan & Li, 2016). Landfill space is reducing rapidly and emissions from previously disposed of wastes increase (Lee et al, 2020), which might lead to higher cost of landfill and more illegal dumping of domestic CDW shown in Fig. 4.</p><p>Though the negative impact from landfill mining might have impeded the original aim of the waste ban, the practise of landfill mining would be a choice to reduce landfill space demands, especially for those in big cities where urbanisation expands fast. Moreover, landfill mining could bring more opportunities if technological advances in waste re-processing were accelerated.</p><p>Table 1: Main categories of recyclable waste generated in PRC and the amount recovered between 2014 and 2018 (unit: × 106tons).</p><p>Table 2: Main categories of recyclable waste generated in PRC and value of recovered material between 2014 and 2018 (unit: billion Yuan).</p><!><p>There are frequent updates of major policies related to waste and the environment since 1991, well before the waste import ban was announced in 2017. As shown in Fig. 3, the frequency of updates or revisions of major policies shows higher level of activity from 2015 onwards. It includes major policy changes related to environmental issues and reflects broader changes in formal environmental regulation intensity (Zhang et al., 2021), which shows significant cross-regional variation in regulation of emission to air, water, and re-use of wastes. The disposal of waste is still dominated by landfills although combustion has increased, and more regulations and policies are being adopted to encourage a more circular economy. The 'Green Fence' policy has been implemented since 2013 to restrict copper scrap imports to mainland China and the impact of the policy on the Circular Economy has been assessed and shows that China still has to pay more attention to the domestic recycling industry and keep the import of high-quality copper scrap, which could provide China a transition to a more circular economy on copper in future (Dong et al., 2020). To reduce the source of waste, an announcement on matters related to the ban on the import of solid waste totally (MEE, 2020) was issued on 24th November 2020 by the Ministry of Ecology and Environment (Gov.CN, 2020). It has been implemented since 1st January 2021.</p><p>There is increased attention on the transition from fossil fuels to renewable energy sources that not only reduce China's carbon footprint reduction but also leads to reduction of waste generation and whilst maximising recycling of various waste as secondary carbon raw material, which is the pilot study to develop zero waste cities (GOSC, 2018). There are "11 + 5" cities and regions labelled as zero waste construction pilot city or region alltogether in 2018 on basis of Chinese administrative division, including eleven cities: Shenzhen city, Xining city and so on; one new district: state-level Xiong'an new area in Hebei province; one development zone: Beijing Economic and Technological Development Area; one international cooperative: Sino-Singapore Tianjin Eco-City, one county: Guangze county of Fujian province; one county-level city: Ruijin county-level city of Jiangxi province (Sohu, 2019). Xining city is the only city located in northwest China and the biggest area among all pilot cities. The accumulative budgeted investment for 26 "non-waste cities" construction projects in Xining city is nearly 0.646 billion US dollar. There are 10 solid waste utilization and disposal chains formed within Xining City. The structure of traditional industries such as the chemical industry and refining is optimised to promote industrial recycling, resource utilization, and ecological development, with a reduction of 7.5% on energy consumption per unit of GDP compared to last year.</p><p>In terms of agriculture, five production bases are built in the implementation of the agricultural and livestock product quality and safety assurance project and it occupied 57,730 ha, including the national green food raw material (broad bean) standardised production base for broad bean and potato in Huangzhong District and Datong County, respectively. Through using alternative technologies such as using organic fertilizers, crop rotation, and improved irrigation, there has been a reduction of 43,000 ha of land that has been treated with chemical fertilizers and pesticides. In this model, 630,000 acres of chemical fertilizers and pesticides have been reduced, and 77,000 tons of organic fertilizers have been subsidized to increase efficiency in 2020. The use of chemical fertilizers and pesticides in 2020 was reduced by 41.9% and 32.9%, respectively compared with those of 2018. In addition, "Enterprise recycling, farmer participation, government supervision, and market promotion" agricultural film recycling system is also built-in Xining city with the recycling rate of the agricultural film more than 90%. For citizens daily life, the domestic waste classification in Xining city covers a total of 304,000 households with a recycling rate of 37% on domestic waste, including more than 1,300 tons of waste in plastic, textile, metal, paper, electronic products, and other recyclables and perishable waste that enter the recyclable and non-hazardous disposal network separately.</p><p>Land use policy is further impacted by recent plans to develop zero waste cities (Lee et al., 2020): a large amount of land will be used more efficiently due to less industrial solid waste exposure. Through the establishment of green waste-free cells in society, promotion and guidance of the concept of waste-free are vigorously promoted to form a consensus of waste-free for citizens. According to the investigation conducted by the Qinghai Provincial Social Situation and Civil Investigation Center, the popularisation rate of publicity, education, and training for the construction of a "waste-free city" in Xining City is 85.34%. The degree of satisfaction for the construction of the "waste-free city" was 83.02% for the public. The construction of green cells consists of green posts, green restaurants, green mines, and green buildings, etc.</p><p>In addition, more regulations and policies have been introduced since 1995 to restrict the waste import properly, which shows that the Chinese government has set a basic regulation frame all the time and is trying to build a better recycling or waste management system over time.</p><!><p>The life cycle cost of recycled paper manufacturing in China has been assessed (Li et al., 2020). A ban on unsorted recovered paper was announced in 2018 and the import quota for recovered paper has been tightened, leading to a dramatic drop in the amount of imported recovered paper (Li et al., 2020). As the recovered paper is a key raw material for the recycled papermaking industry, alternatives (e.g., straw, imported wood pulp, and imported deinked pulp) are now being assessed to deal with the decrease in the feedstock. The results indicated that imported deinked pulp might be the trend for the recycled papermaking industry in China with a growth in the price of the domestic recovered paper (Li et al., 2020). This change might bring additional research and development of potential alternatives in the future to satisfy the increasing need for imported recycled material due to the waste import ban.</p><p>The environmental impact on the life cycle of used polyethylene terephthalate (PET) was also analysed under many post-ban scenarios (Ren et al., 2020). The result shows that the absence of imported used PET in China leads to an increase in virgin PET fibre production using manufacturing processes that are based on carbon-intensive coal. This brings an additional environmental impact because of the higher air pollution emissions from production (Ren et al., 2020). The treatment of air pollution may become new challenge; however, improving the local PET recycling rates and searching for production alternatives are emerging new opportunities.</p><p>There are recycled commodities, which may be sold at a cheaper price to downstream companies because of higher quality compared to domestic products. Without importing waste material, the costs of downstream companies would increase because they cannot use cheaper recycled items directly, or they need to find an alternative source. However, with more initiatives available to boost collection and recycle businesses (Zuo et al., 2020), there are now opportunities for people to start new, or adapt current, businesses in different ways and bring new jobs.</p><p>Businesses who deal with waste still need a significant and consistent supply of waste materials, which pushes the domestic waste recycling companies to move their attention to domestic waste as an alternative, for example, domestic copper scrap recycling and application (Dong et al., 2020; Liu et al., 2020; Wang et al., 2019). Some of the larger companies may survive after small businesses fail: the trash ban is a threat to the existence of small companies; however, it can be an opportunity for a big company to enlarge and update its technology (Zhang, 2020a). Some companies will change their processing to higher efficiency and value to survive, for example, developing a new idea to apply the recycled resource. It may even force companies to consider product recycling at the design stage, such as extended producer responsibility (EPR) (Zhang, 2020b). Therefore, instead of just disposing of the waste directly and paying fees to get the raw recycled waste material directly from abroad, companies are pushed to design products, which consider new waste sorting and recycling methods.</p><p>For waste management companies in Southeast Asian countries, an increase in solid waste import after the China import ban is now a threat to business sustainability from the potential imposition of similar restrictions by their governments. There is no detailed data covering company failure as a result of the ban and the number of jobs lost, which is quite important for the national-society. However, it seems that some Southeast Asian countries have already tightened their restrictions by announcing statements or revoking their import licences for local companies that process plastic waste or e-waste (WMR, 2018).</p><p>For companies in mainland China, technologies such as the Beidou navigation system, the Internet of Things, Internet and Artificial Intelligence, etc. are applied by Yixun Intelligent Environmental Sanitation System to manage the whole process of people, trucks, materials, machines involved in environmental sanitation management in real-time. It can help design environmental sanitation management models, rationally, improve the quality of sanitation operations, reducing environmental operation costs, and assessing the effectiveness of the management through data (Gov.CN, 2019). It is a new start and possibility for the structure of the domestic waste management system thanks to the involvement of the Internet and big data. Whilst some information has been derived from structural changes of the global waste trade network (Wang et al., 2020) (Pu et al., 2019; Tran et al., 2021), and mechanisms of response (Tan et al., 2018), 'Big data' application for the waste management systems could be a potential bright path for the transition of domestic companies.</p><!><p>The global pandemic has influenced the production and management of the various industrial sectors, in terms of waste to recycling generation of end-of-life vehicles were stopped because of the shutdown of companies upstream, and there was an insufficient supply of raw materials for many manufacturing companies. Whilst for domestic WEEE, there is a consequential increase in the generation of e-waste due to increasing utilization of electrical and electronic devices for online teaching or work from home during the pandemic or post COVID19 in a developing country (Adejumo & Oluduro, 2021). In terms of waste to energy, for example, waste incineration for power generation (business), the closure of restaurants and service companies leads to the amount of food collection and volume of transport in some cities or regions was only 25% ~ 33% of pre-pandemic levels of pre-pandemic levels due to lockdown and social restrictions. On the other hand, the pandemic caused an increase in medical waste due to increased production of protective materials (masks, gowns, etc.). The more complicated situation poses a potential challenge for waste management system because of shortage of the raw materials for manufacture, restricted traffic, infection risk, higher cost. In addition, large-scale sanitation and disinfection work during the pandemic increased company operating costs such as labour, machinery, materials, and transportation (Finance, 2020). However, studies have shown that lower levels of industrial wastes are generated because of lower production and exports due to reduced demand during the shutdown (Maliszewska et al., 2020). More waste has been sent directly to incineration in order to reduce the risk of infection, resulting in the over-working of incineration equipment in some cities and regions. Besides, there is also the impact of the extended construction period, increased investment, and secondary urban environmental protection issues on projects under construction of waste incineration for power generation. According to the current local control levels, the construction period has been impacted with an average of 55 days. The overlap of the processing period is longer because of the lag in the impact of materials, equipment, logistics, and other parts of the supply chains. In addition, the investment will also increase due to the following: (i) the increase in the cost of personnel, machinery, and materials of the project, (ii) the increase in the cost of capital, and (iii) the loss of operating income caused by the delay in commissioning (Finance, 2020). Higher costs and lower-income bring a dilemma for investors and the construction of waste incineration facilities in the future. It is difficult to make decisions on the treatment of the municipal solid waste and medical waste, to consider separate or not during the difficult periods during the pandemic.</p><p>The total mass of the portion of municipal solid waste from a city in Southern China, transported to a waste to energy incineration plant is shown in Fig. 6. The annual average and detailed monthly amounts received between 2016 and 2020 are presented. As a yearly average, a slight increase in 2019 from roughly comparable levels in 2016–2018 is enhanced in 2020. However, the detailed monthly variation is also informative. In general, there is a big drop in the amount of solid waste produced in February because of the lockdown (coming into effect early in 2020 in China) and people returning to their hometown for the Spring Festival. In pre-pandemic, variation across the year tends to show decrease in amounts over the popular summer holiday months, picking up particularly in the period of the early autumn festivals. For 2020 this variation is suppressed and a steady increase in waste generation is seen as restrictions on movement, maintain a constant population in the city and consumption rates of produce increase along with increased disposal of PPE. Changes in annual average waste production prior to the pandemic are less easy to rationalise, and the dip from 2016 to 2018 and increase from 2018 to 2020 may be partly attributed to the waste import ban and further restrictions.</p><p>The pandemic increases pressure on the waste management system on top of the waste ban and implementation of a series of related regulations and policies. However, it also provides some dynamic evidence of the response of the waste management system. The post-pandemic recovery should provide an opportunity to establish more resilient, sustainable management systems in the future.</p><!><p>The impact of the waste ban and new policies on the waste management sector in mainland China has been reviewed in the context of recent challenges from the COVID19 pandemic as national legislation prevented the import of waste. We highlight the landfill mining situation in China, assessing the dynamics of domestic policies before and after the implementation of the ban in China and challenges and opportunities for waste management companies. The integrated assessment of the impact of the ban in China has relevance for future influence on the waste management system in China.</p><p>In conclusion, the change in waste import in the run-up to the ban leads to different situations for the variety of waste imports in the future in China. The ban drove price increases in the domestic recycling sector that forces companies to introduce more formal recycling processes and to drive the consumption behaviours to more reasonable and environmentally friendly options. The impact in China is ultimately to reduce the pollution of the environment and improve health, but the negative impact from landfill mining has impeded the original aim of the waste ban that requires further technological development. The dynamic of domestic policies in China shows a higher level of activity of updates and revisions or promulgations of new policies from 2015 onwards and 'zero waste cities' brings new hope for improvement of the Chinese waste management system. Lessons could be learned for other countries, such as countries in Southeast Asia, after the import ban in China through discussion on challenges and opportunities for waste management companies in China. 'Big data' application could be a transition for the domestic companies. Whilst the restrictions on imports have provided a stimulus for improved waste management practices, changes in those restrictions in other territories may have further impacts on the economic viability of businesses once these countries lift the restrictions similar to the ban promulgated in China. Moreover, the pandemic brings more tricky challenges for the waste management system than before; however, it also gives a clue to the world that how quickly the waste management system can respond, particularly post-pandemic will be an important step to establishing sustainable management systems in the future.</p><p>The detailed analyses of the companies that ceased trading or of the rise in unemployment in the sector due to the waste ban, the consideration of advantages and disadvantages on 'zero waste city' and 'big data' application would be urgent priorities for further research.</p><!><p>Publisher's Note</p><p>Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.</p><!><p>ASH, NS: Conceptualization, Methodology, Validation, Investigation, Formal analysis, Writing—Original Draft & Review & Editing, Supervision. WL: Investigation, Data curation, Formal analysis. IMcL & ZW: Investigation, Methodology, Data curation, Formal analysis Reviewing & Editing.</p><!><p>Na Song acknowledges the support from the Scottish Alliance for Geoscience, Environment & Society (www.sages.ac.uk) for a PECRE award in February 2019.</p><!><p>Data used in this paper was derived from publicly available information and the result of ongoing research. No personal or sensitive information was used.</p><!><p>The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.</p><!><p>No animals were involved in this study.</p><!><p>No human subjects were involved in this study.</p><!><p>No human subjects were involved in this study.</p>

PubMed Open Access

Bis benzothiophene Schiff bases: synthesis and in silico-guided biological activity studies

Since benzo [ b ] thiophene scaffold is one of the privileged structures in drug discovery as this core exhibitsactivities for different biological problems, in this study bis (benzo[ b ]thiophene-2-yl) alkyl methanimine derivatives (1-9) were synthesized by reacting benzo[ b ]thiophene-2-carbaldehyde with diamines. All newly compounds were characterized by IR, 1H NMR and 13C NMR spectroscopic methods. Synthesized compounds were investigated using binary QSARbased models on therapeutic activity prediction of synthesized compounds and they showed high predicted activities in following diseases: bacterial, angina, allergy, depression and obesity. Thus, they were then tested for their antimicrobial and antileishmanial activities as a result of this theoretical study. Compound 1(N, N’- (propane-1,3-diyl) bis (1-(benzo [ b ] thiophene-2-yl)) methanimine) was found the most active compound in both diseases. Thus, its molecular docking studies were also carried out.

bis_benzothiophene_schiff_bases:_synthesis_and_in_silico-guided_biological_activity_studies

1. Introduction<!>2.1. General method for the synthesis of compounds (1-9)<!>2.1.1. Synthesis of N, N’-(propane-1,3-diyl) bis(1-(benzo[ b ] thiophene-2-yl) methanimine) (1):<!>2.1.2. Synthesis of 1-(benzo[ b ]thiophene -2-yl)-N-(4-((benzo[ b ]thiophene-2-ylmethylene) amino) butyl) methanimine(2):<!>2.1.3. Synthesis ofN,N’-(pentane-1,5-diyl)bis(1-(benzo[ b ] thiophene-2-yl)methanimine) (3):<!>2.1.4. Synthesis of 1-(benzo[ b ]thiophene-2-yl)-N-(6-((benzo[ b ]thiophene-2- ylmethylene) amino) hexyl) methanimine (4):<!>2.1.5. Synthesis of N,N’-(heptane-1,7-diyl)bis(1-(benzo[ b ] thiophene-2-yl)methanimine) (5):<!>2.1.6. Synthesis of 1-(benzo[ b ]thiophene-2-yl)-N-(8-((benzo[ b ]thiophene -2- ylmethylene) amino) octyl) methanimine (6):<!>2.1.7. Synthesis of N,N’-(nonane-1,9-diyl)bis(1-(benzo[ b ] thiophene-2-yl)methanimine) (7):<!>2.1.8. Synthesis of 1-(benzo[ b ]thiophene-2-yl)-N-(10-((benzo[ b ]thiophene-2- ylmethylene) amino) decyl) methanimine (8):<!>2.1.9. Synthesis of 1-(benzo[ b ]thiophene-2-yl)-N-(12-((benzo[ b ]thiophene-2- ylmethylene) amino) dodecyl) methanimine (9):<!>2.2. MetaCore/MetaDrug applications<!>3. Ligand preparation<!>4. Protein preparation<!>5. Molecular docking<!>6.1. Synthesis<!><!>6.1. Synthesis<!>6.2. In silico predictions of therapeutic activities of synthesized compounds<!>6.3. In vitro antileishmanial activity studies<!><!>6.3. In vitro antileishmanial activity studies<!>6.4. In vitro antibacterial and antifungal activity studies<!><!>6.4. In vitro antibacterial and antifungal activity studies<!><!>7. Conclusion<!>

<p>Benzo [b] thiophene derivatives display remarkable biological activities such as antiinflammatory, analgesic, antifungal, antidepressant, antiangiogenic, estrogen receptor modulating, antimitotic, anticancer, kinase inhibitors, antituberculosis, anticonvulsant, antimalarial, anthelmintic, antihyperglycemic and pesticide. Benzothiophene derivatives have been used as potential diagnostic agents and amyloid binding in neurodegenerative diseases, treatment of fatty acid amide hydrolase inhibitors (FAAH), BMP-2 upregulators, Alzheimer's disease (AD), human nicotinamide phosphoribosyltransferase inhibitors, BRAF kinase inhibitors, Rho kinase inhibitors, selective linear tachykinin NK2 receptor antagonists, protein tyrosine phosphatase 1B inhibitors, histamine H3 antagonists, antiallergic agents and many other activities [1–7]. By replacing the substituents attached to the benzothiophene ring, some of the FDA-approved anticancer drugs such as arzoxifene and raloxifene were also developed [8,9]. Schiff bases, including azomethine functional group (N=CH), possess biological properties such as analgesic, antifungal, antibacterial, antidepressant, anticancer, anticonvulsant and antiinflammatory [10–15]. Schiff bases are also known to be highly effective in synergistic effects on insecticides and plant growth regulators [16–18].</p><p>Antibiotics are widely used in the treatment of bacterial infections for many years in the early 20th century. Penicillin was the first antibiotic used to treat bacterial infections. Antibiotics, which are effective by killing or stopping the growth of bacteria, cause the emergence of drug-resistant pathogens and the development of resistance if used unconsciously and excessively. In addition, the long-term use of antibiotics damages our microbiota and losing the beneficial microorganisms has negative consequence. Approximately 16 million people die annually due to bacterial infections, while developing new approaches to combating drug-resistant pathogens and infectious diseases. New substances have been synthesized due to technological developments and have been used in various bacterial infections for therapeutic purposes [19,20].</p><p>Leishmaniasis is a disease caused by a protozoon called Leishmania which is transmitted by the bite of infected female sand flies. Poor nutrition lack of sanitation weakened immune system and poverty facilitate spread of the disease in underdeveloped and developing countries. The disease occurs in 3 different forms: cutaneous (most common), visceral (also known as kala-azar and the most serious form of the disease), and mucocutaneous. An estimated 700,000 to 1 million new cases are reported annually, and 26,000 to 65,000 disease-related deaths are reported by the World Health Organization.While cutaneous form is endemic in Southeastern Anatolia, visceral form is mostly seen in Mediterranean, Central Anatolian and Aegean Regions andimportant constitutes an important health problem for Turkey [21,22].</p><p>Recent advances made in molecular biology and computational chemistry open new avenues in designing novel therapeutic compounds. Molecular biology has provided the crystal structures and cryo-electron microscopy structures of crucial targets for different diseases, which can be used as accurate templates in modeling studies. Computational chemistry offers a range of simulation, multiscale modeling and virtual screening tools for definition and analysis of protein-ligand, protein-protein interactions. Development of new techniques on statistical methods and free energy simulations help to predict novel optimal ligands. Thus, in this study, we performed in silico-guided biological activity studies for a set of newly synthesized compounds. We aimed to synthesize a new series of Schiff bases including benzo [ b ] thiophene for more efficacious biologic activities. In order to predict the biological activities of synthesized compounds before the experimental studies, therapeutic activities of these novel compounds were predicted. The therapeutic activity value (TAV) of each compound was calculated by binary QSAR models of 25 common disease QSAR models. In silico results represented that synthesized compounds may show potential activity against in following diseases: bacterial, angina, allergy, depression and obesity. Therefore, these compounds were then tested for their antimicrobial and antileishmanial activities. Moreover, molecular docking study for one of the most active compounds as antibacterial agent was performed and its molecular interactions were studied.</p><!><p>Benzo [ b ] thiophene-2-carbaldehyde (2 mmol) and diamines (1 mmol) were mixed and heated in oil bath without solvent for 2–3 h at 150–160 °C. The reaction content was controlled with TLC examination and cooled to room temperature. The precipitate formed was purified by recrystallization from ethanol-diethylether.</p><!><p>Yield: 87.00%, m.p. 183–184 °C. IR (KBr, cm−1): 3056 (=CH), 2936 (CH), 1633 (C=N), 1600 (C=N), 1527 (C=C); 1H NMR (400 MHz, DMSO-d6) δ: 8.64 (s, 2H, N=CH), 7.92–7.97 (m, 4H, Arom.H), 7.82 ( bs, 2H,thiop.H), 7.42 (bs, 4H, Arom.H), 3.68 (s, 4H, N-CH2) , 1.98 (s, 2H, N-CH2-C H 2); 13C NMR (100 Hz, DMSO-d6) δ: 156.23, 143.05, 139.99, 139.55, 128.78, 126.64, 125.28, 123.22, 58.28, 32.05.</p><!><p>Yield: 88.20%, m.p. 166–167 °C. IR (KBr, cm−1): 3056 (=CH), 2939 (CH), 1628 (C=N), 1526 (C=N), 1464 (C=C); 1H NMR (400 MHz, DMSO-d6) δ: 8.61 (s, 2H, N=CH), 7.87–7.97 (m, 4H, Arom.H), 7.79 (2H, bs, thiop.H), 7.37–7.44 (m, 4H, Arom.H), 3.64 (s, 4H, N-CH2), 1.68 (s, 4H, N-CH2-C H 2); 13C NMR (100 Hz, DMSO-d6) δ: 155.87, 143.11, 139.96, 139.65, 128.66, 126.62, 125.22, 123.23, 60.33, 28.64.</p><!><p>Yield: 82.65%, m.p. 110–111 °C. IR (KBr, cm−1): 3053 (=CH), 2935 (CH), 1632 (C=N), 1560 (C=N), 1525 (C=C); 1H NMR (400 MHz, DMSO-d6) δ: 8.58 (s, 2H, N=CH), 7.84–7.93 (m, 4H, Arom.H), 7.75 (2H, bs, thiop.H), 7.37–7.42 (m, 4H, Arom.H), 3.59 (s, 4H, N-CH2), 1.65–1.70 (m, 4H, N-CH2-C H 2), 1.35–1.39 (m, 2H, N-CH2-CH2-C H 2); 13C NMR (100 Hz, DMSO-d6) δ :155.73, 143.15, 139.64, 128.56, 126.58, 125.19, 123.20, 60.44, 31.44, 24.87.</p><!><p>Yield: 80.28%, m.p. 143–144 °C. IR (KBr, cm−1): 3057 (=CH), 2934 (CH), 1628 (C=N), 1560 (C=N), 1526 (C=C); 1H NMR (400 MHz, DMSO-d6) δ: 8.58 (s, 2H, N=CH), 7.88–7.95 (m, 4H, Arom.H), 7.76 (2H, bs, thiop.H), 7.41 (bs, 4H, Arom.H), 3.58 (s, 4H, N-CH2), 1.63 (bs, 4H, N-CH2-C H 2), 1.37 (bs, 4H, N-CH2-CH2-C H 2); 13C NMR (100 Hz, DMSO-d6) δ: 155.73, 143.15, 139.64, 128.56, 126.58, 125.19, 123.20, 60.44, 31.44, 24.87.</p><!><p>Yield: 80.85%, m.p. 117–118 °C. IR (KBr, cm−1): 3056 (=CH), 2934 (CH), 1673 (C=N), 1629 (C=N), 1517 (C=C); 1H NMR (400 MHz, DMSO-d6) δ: 8.57 (s, 2H, N=CH), 7.87–7.94 (m, 4H, Arom.H), 7.76 (2H, bs, thiop.H), 7.37–7.43 (m, 4H, Arom.H), 3.56 (s, 4H, N-CH2), 1.62 (bs, 4H, N-CH2-C H 2), 1.37 (bs, 6H, N-CH2-CH2-C H 2-C H 2); 13C NMR (100 Hz, DMSO-d6) δ: 155.63, 143.16, 139.95, 139.64, 128.54, 126.59, 125.19, 123.22, 60.53, 30.70, 28.93, 27.14.</p><!><p>Yield: 84.85%, m.p. 122–123 °C. IR (KBr, cm−1): 3056 (=CH), 2924 (CH), 1628 (C=N), 1525 (C=N), 1465 (C=C); 1H NMR (400 MHz, DMSO-d6) δ: 8.59 (s, 2H, N=CH), 7.91 (bs, 4H, Arom.H), 7.78 (2H, bs, thiop.H), 7.41 (bs, 4H, Arom.H), 3.39 (s, 4H, N-CH2) , 1.62 (bs, 4H, N-CH2-C H 2), 1.32 (bs, 8H, N-CH2-CH2-C H 2-C H 2); 13C NMR (100 Hz, DMSO-d6) δ: 155.65, 143.17, 140.91, 128.56, 126.01, 125.20, 123.23, 60.58, 30.78, 29.36, 27.18.</p><!><p>Yield: 81.20%, m.p. 115–116 °C. IR (KBr, cm−1): 3057 (=CH), 2923 (CH), 1672 (C=N), 1629 (C=N), 1525(C=C); 1H NMR (400 MHz, DMSO-d6) δ: 8.57 (s, 2H, N=CH), 7.89–7.96 (m, 4H, Arom.H), 7.78 (2H, bs, thiop.H), 7.41 (bs, 4H, Arom.H), 3.56 (s, 4H, N-CH2), 1.60 (bs, 4H, N-CH2-C H 2), 1.30 (bs, 10H, N-CH2-CH2-C H 2−C H 2-C H 2); 13C NMR (100 Hz, DMSO-d6) δ: 155.64, 143.17, 139.95, 139.66, 128.56, 126.60, 125.21, 123.23, 60.57, 30.77, 29.41, 29.18, 27.18.</p><!><p>Yield: 83.25%, m.p. 141–142 °C. IR (KBr, cm−1): 3050 (=CH), 2924 (CH), 1625 (C=N), 1591 (C=N), 1525(C=C); 1H NMR (400 MHz, DMSO-d6) δ: 8.58 (s, 2H, N=CH), 7.88–7.94 (m, 4H, Arom.H), 7.77 (2H, bs, thiop.H), 7.40 (bs, 4H, Arom.H), 3.56 (s, 4H, N-CH2), 1.58 (bs, 4H, N-CH2-C H 2), 1.28 (bs, 12H, N-CH2-CH2-C H 2−C H 2-C H 2); 13C NMR (100 Hz, DMSO-d6) δ: 155.64, 143.18, 139.70, 139.66, 128.55, 126.59, 125.21, 123.23, 60.57, 30.77, 29.83, 29.41, 27.33.</p><!><p>Yield: 80.25%, m.p. 122–123 °C. IR (KBr, cm−1): 3054 (=CH), 2918 (CH), 1632 (C=N), 1560 (C=N), 1526 (C=C); 1H NMR (400 MHz, DMSO-d6) δ: 8.58 (s, 2H, N=CH), 7.90–7.95 (m, 4H, Arom.H), 7.79 (2H, bs, thiop.H), 7.41 (bs, 4H, Arom.H), 3.38 (s, 4H, N-CH2), 1.59 (bs, 4H, N-CH2-C H 2), 1.24 (bs, 10H, N-CH2-CH2-C H 2−C H 2-C H 2), 1.06 (bs, 4H, N-CH2-CH2-CH2−CH2-CH2-C H 2); 13C NMR (100 Hz, DMSO-d6) δ: 155.65, 139.70, 128.57, 126.60, 125.21, 123.23, 60.57, 30.74, 29.40, 29.18.</p><!><p>Studied compounds were examined for their therapeutic activity against 25 common diseases using Meta-Core/MetaDrug platform of Clarivate Analytics (Philadelphia, PA, USA). Training set of compounds were collected from FDA approved compounds, candidate drugs in clinical phases and hit compounds from in vivo activities. Results showed that synthesized compounds can be used as antibacterial agents. Following training and test set compound numbers were used in bacterial QSAR model constructions: Training set N = 530, test set N = 97, sensitivity = 0.87, specificity = 0.90, accuracy = 0.89, Matthews correlation coefficient (MCC) = 0.77.</p><!><p>The synthesized molecules are prepared for further calculations via LigPrep module [23] of Schrodinger's molecular modeling suite with OPLS2005 force field [24]. Epik [25] was used to determine the protonation states of these compounds at neutral pH.</p><!><p>The X-ray diffraction (XRD) structure of the target was used from the Protein Data Bank (PDB) with PDB IDs of 5TW8. Protein preparation module was utilized [26], [27]. Missing side chains are filled via prime module and PROPKA [28]. was subjected to determine the protonation of side chains at physiological pH. Finally, a restrained minimization in OPLS2005 force field was applied to these optimized structures.</p><!><p>The molecules which possessed desirable pharmacokinetic and toxicity properties were subjected to a grid-based Glide/XP docking protocol from Schrodinger's Maestro molecular modeling package [29]. Binding pocket was determined from cocrystallized ligand and default protocol was employed. Shortly, we used flexible docking approach. Hydroxyl and thiol groups of binding pocket residues were allowed for rotation throughout the docking. Extra precision (XP) protocol of Glide was used. Ten poses were requested at the docking and postdocking minimization was performed.</p><!><p>Synthesis of bis (benzo[b]thiophene-2-yl) alkyl methanimine derivatives (1-9) was performed according to the reaction outlined in Scheme.</p><!><p>Synthesis of Schiff base derivatives with benzo [b] thiophene (1-9).</p><!><p>The peaks of NH2 and C=O belonging to the starting amines and aldehyde respectively disappeared in the IR spectra of the compounds (1-9). Proton signal of imine group (N=CH) obtained by Schiff base reaction was observed as singlet at 8.57–8.64 ppm in the 1H-NMR spectra of the compounds (1-9). Aromatic protons belonging to benzo[ b ]thiophen ring resonated at 7.37–7.97 ppm and CH2 alkyl protons were seen at 1.28–3.68 ppm in the 1H-NMR spectra of the compounds as expected from alkyl groups. When carbon spectrawere examined, N=CH imine carbon was observed at 155.64–156.23 ppm in the 13C-NMR spectra of compounds 1-9. Aromatic carbon peaks of benzo [ b ] thiophene were seen at 123.20–143.19 ppm and alkyl carbon peaks resonated at 24.87–60.58 ppm in the 13C-NMR spectra of the compounds. As a result, spectral data supports structures of the compounds (1-9).</p><!><p>Binary quantitative structure-activity relationships (QSAR) common disease models from Clarivate Analytics MetaCore/MetaDrug platform were used for the therapeutic activity predictions of synthesized compounds. For this aim, 25 different common diseases binary QSAR models were used. All the nine compounds were screened on MetaCore/MetaDrug platform and therapeutic activity values were predicted. Therapeutic activity values in MetaCore/MetaDrug are normalized between 0 and 1 (while 0 represents inactive compound, 1 represents active compounds). Although predicted therapeutic activity value higher than 0.5 indicates compound that may show activity, in the current study we used a higher cutoff value (0.75) for being in the safe zone. Table S1 at the supplementary information show corresponding predicted therapeutic activity values of each studied compound that has activity value equal or more than 0.75. When we check diseases in Table S1 it can be seen that synthesized compounds can be considered mainly for bacterial, angina, allergy, depression and obesity models. Since the investigated compounds are analogs of each other they showed similar TAV values in certain disease models such as antibacterial profile. Thus, from these diseases we considered to perform antibacterial in vitro assays using these compounds. Furthermore, since known antifungal and antileishmanial effect of benz [ b ] thiophene derivatives, we also performed antifungal and antileishmanial in vitro assays for these compounds.</p><!><p>Antileishmanial activity results of nine compounds as a result of evaluation Figures 1–2 and minimum inhibitory concentration (MIC) values are given in Table 1. The tests in the positive and negative control wells were found to work as expected.</p><!><p>Antileishmanial activity results of the compounds against standard Leishmania infantum promastigotes. Control drug: Amphotericin B. Dilution concentrations 10,000 μg/mL to 312 μg/mL. N.C.: Negative control; P.C.: Positive control.</p><p>Antileishmanial activity results of compounds against standard Leishmania tropica promastigotes. Control drug: Amphotericin B. Dilution concentrations 10,000 μg/mL to 312 μg/mL. N.C.: Negative control; P.C.: Positive control.</p><p>Minimum inhibitory concentration (MIC) values of the compounds against Leishmania infantum and Leishmania tropica promastigotes.</p><!><p>Compound 1 was found to be the most effective compound (MIC: 1250 μg / mL) among the compounds whose antileishmanial activities were evaluated against Leishmania infantum promastigotes. In addition, compounds 3, 7 and 9 were found to have antileishmanial activity at different concentrations (MIC: 5000–10,000 μg/mL).</p><p>It was found that other compounds did not have antileishmanial activities against Leishmania infantum promastigotes at the studied concentrations. Compound 1 was also found to be the most effective compound against the Leishmania tropica promastigotes from compounds whose antileishmanial activities were evaluated (MIC: 1250 μg/mL). It was determined that other compounds did not have antileishmanial activities against Leishmania tropica promastigotes at studied concentrations.</p><p>The study of Maina et al. revealed the antileishmanial activity of Clerodendrum myricoides and Salvadora persica . They reported, Clerodendrum myricoides water extract demonstrated the best potential antileishmanial activity against Leishmania major promastigotes (MIC = 625 μg/mL). Also, the dichloromethane and petroleum ether extract were reported moderate to weak activity against Leishmania major promastigotes (MIC = 1250 μg/mL; 2500 μg/mL) and amastigotes respectively [30].</p><p>In the study of Ogeto et al., Aloe secundiflora water extract was found active against L. major at the lowest concentration of 2000 μg/mL. Methanollic plant extract reported less antileishmanial activity as compared to Pentostam and amphotericn B with MIC of 1000 μg/mL, 250 μg/mL and 125 μg/mL, respectively. They marked higher antileishmanial activities though not of comparative concentrations than most of the reference drugs. The results also showed that plant extracts had lower toxicity against Vero cells as compared to the standard drug amphotericin B [31].</p><p>Amphotericin B, which is used as a standard drug, was found to be effective at all concentrations (MIC: <312 μg / mL) for both leishmania species.</p><!><p>Different concentrations of some compounds against standard bacterial isolates and yeast isolates were found to have antimicrobial activity, while some compounds were found to be ineffective at the concentrations studied. The test images of compound 1 which have the most effective antibacterial and antifungal effect are given in the Figure 3. The MIC values of the compounds are given in Table 2. It was determined that the compounds showed antibacterial and antifungal activity against 6 of the studied standard bacteria and yeast isolates.</p><!><p>Antibacterial and antifungal activity results of compound 1. Dilution concentrations: 10,000 μg/mL–312 μg/mL. N.C.: Negative control; P.C.: Positive control.</p><p>Minimum inhibitory concentration (MIC) values of all compounds against bacteria and fungi.</p><!><p>The compound 1 showed the strongest antibacterial effect against Shigella flexneri and Yersinia enterocolitica (MIC:1250 μg/mL) (Figure 3). Compound 1 was found to be effective against all bacteria and yeast (Table 2). Antibacterial and antifungal activity were determined in 6 compounds (1, 2, 4, 6, 8 and 9).</p><p>When the results of synthesized compounds were evaluated, it was found that compound 1 was the most effective compound for all bacteria and yeast isolates studied. Compound 1 was found to have different levels of antimicrobial activity against Shigella flexneri , Yersinia enterocolitica , Staphylococcus aureus (MRSA), Escherichia coli , Salmonella enteritidis , Candida tropicalis , Pseudomonas aeruginosa , and Klebsiella pneumoni (respectively MIC values: 1250 μg/mL, 1250 μg/mL, 2500 μg/mL, 2500 μg/mL, 2500 μg/mL, 2500 μg/mL,5000 μg/mL, and 5000 μg/mL).</p><p>It is found that Klebsiella pneumoniae is the least affected one by the compounds. All concentrations of the compounds 3, 5 and 7 were found to have no effect on any bacterial and yeast isolates (MIC: >10,000 μg/mL).</p><p>Thus, because of its promising results at both disease models, compound 1 can be considered as lead compound. However, further animal in vivo studies must be carried out.</p><p>Since the most active compound was found as compound 1 as antibacterial agent, we used crystal structure of wild-type S. aureus penicillin binding protein 4 (PBP4), PDB ID: 5TW8, and performed dockin simulations. Figure 4 shows 2D and 3D ligand interactions diagram of compound 1 at the binding pocket of the target structure. It can be seen that main nonbonded interactions are constructed with Tyr291 and Tyr268 (π-π stacking interactions) and with Glu297 (ionic interactions). These strong interactions (docking score, -8.33 kcal/mol) show that compound 1 can be considered as promising anti-bacterial agent. Both in silico predictions (ligand-based and target-driven based) as well as performed in vitro assays validate each other for the antibacterial effect of the novel compound (compound 1). Our hybrid molecular modeling approache (combined ligand-based and structure-based) predicted the biological activity of a set of new compounds and these predictions were investigated by in vitro studies.</p><!><p>2D and 3D ligand interactions diagrams of compound 1 at the binding pocket of 5TW8 PDB coded structure.</p><!><p>Nine bis (benzo[b]thiophen-2-yl) alkyl methanimine derivatives were synthesized. All newly compounds were characterized by IR, 1H NMR and 13C NMR spectroscopic methods.Synthesized compounds were investigated using 25 different binary QSAR disease models. Results showed that synthesized compounds may be considered mainly for bacterial, angina, allergy, depression and obesity models. Our in silico-guided design study lead to a hit compound (Compound 1) for its antibacterial effects. In vitro antileishmanial, antifungal and antibacterial activity studies were performed on all synthesized compounds. According to the test results; Compound 1 showed antileishmanial activity (MIC = 1250 μg/mL) and it was concluded that further studies could be continued for this compound 1. The purpose of the control drug study was made to test whether the experimental study was working properly. The substances studied showed antileishmanial activity even if they had a higher value than the MIC of the standard drug Amphotericin B. In this antileishmanial activity study, while aiming to determine the concentration of the compound that is effective against leishmania, the reliability, toxic effect and side effects of the substances should also be evaluated for new drug candidates. Due to the toxic effects of currently used drugs and the increasing resistance to these drugs, the discovery and development of new therapeutic agents is important.Compound 1 also showed better antibacterial and antifungal activity.</p><!><p>Supplementary Materials</p><p>Click here for additional data file.</p>

PubMed Open Access

Identification of a biliverdin geometric isomer by means of HPLC/ESI–MS and NMR spectroscopy. Differentiation of the isomers by using fragmentation “in-source”

AbstractA commercially available biliverdin sample was analyzed by means of HPLC/ESI–MS and NMR spectroscopy. It was been found that beside the main IXα 5Z,10Z,15Z isomer, the sample contains also the geometric isomer IXα 5Z,10Z,15E. It was also found the isomers behave differentially upon “in-source” fragmentation in negative ion mode (in contrast to the their behavior upon “in-source” fragmentation in positive ion mode and to their behavior upon MS/MS fragmentation in both modes): the relative abundances of deprotonated molecules and fragment ions are significantly different for both isomers, which can be used as an analytical tool to differentiate between the isomers.Graphical abstract Electronic supplementary materialThe online version of this article (10.1007/s00706-018-2161-7) contains supplementary material, which is available to authorized users.

identification_of_a_biliverdin_geometric_isomer_by_means_of_hplc/esi–ms_and_nmr_spectroscopy._differ

Introduction<!><!>Results and discussion<!><!>Results and discussion<!><!>Results and discussion<!>Conclusions<!>Experimental<!>

<p>Biliverdin (BV) is a tetrapyrrolic pigment, a product of heme catabolism. Usually the term "biliverdin" refers to the main isomer of biliverdin, namely to IXα isomer. However, one should note that other positional isomers of biliverdin have been identified (e.g. XIIIα or IXδ) [1–11], and also besides the "natural" (5Z,10Z,15Z) isomer the geometric isomers (e.g. (5Z,10Z,15E), at the exocyclic double bonds are possible [12, 13]. Furthermore, the commercially available biliverdin may contain other positional isomers as well [14]. Therefore, we decided to check using HPLC/ESI–MS if biliverdin obtained from a commercial source as hydrochloride contains other isomers.</p><!><p>Single ion chromatograms of ions [BV + H]+ (top) and [BV−H]− (bottom)</p><!><p>As clearly shown in Fig. 1, for both ions two chromatographic peaks are obtained, thus the analyzed biliverdin sample contains two isomers. Taking into account the height of the peaks in the positive ion mode, the ratio of the main isomer to the minor isomer is about 5/1. A similar ratio was obtained by HPLC–UV/Vis analysis as shown in the Supplementary Material. As described further a similar ratio was also obtained by NMR spectroscopy. In the negative ion mode the peak ratio is different (Fig. 1), it indicates that the isomers may behave differentially upon ESI(−) conditions and this problem is discussed further in the text.</p><p>It can be taken for granted that the main isomer is IXα. However, the question is what is the minor isomer. To identify it by HPLC-ESI/MS we should have a respective isomer standards (to compare the retention times and ESI mass spectra). Fortunately, using NMR spectroscopy we were able to identify the minor isomer also as IXα, however, as Z,Z,E-isomer (obviously, the main isomer is Z,Z,Z), as discussed in detail below. The structures of both isomers (geometric isomers) are shown in Scheme 1.</p><!><p>1H NMR spectrum of biliverdin sample in CD3CN/D2O (298 K)</p><p>1H NMR spectrum of biliverdin sample in CD3OD/D2O/NaOD (298 K): full spectrum (top), selected regions (middle and bottom)</p><p>ESI mass spectra of bilverdin isomers obtained upon HPLC-ESI/MS analysis (CV = 50 V)</p><!><p>As clearly seen in Fig. 4, the spectra of both isomers obtained in the positive ion mode are quite similar. There is no difference in relative abundances of fragment ions. There are only minor differences in relative abundances of protonated dimers and sodium adducts. However, the spectra of both isomers obtained in the negative ion mode are different. Beside the differences in relative abundances of deprotonated dimers and sodium adducts, there is a significant difference in relative abundances of fragment ions (Fig. 4). It should be noted that in a few papers ESI/MS has been successfully applied for biliverdin analysis [16–21]. However, to the best of our knowledge, our finding is the first one which demonstrates the different ESI/MS behavior of two biliverdin isomers.</p><p>We also performed HPLC/ESI–MS/MS analysis of commercial biliverdin sample. However, the MS/MS spectra were very similar in both positive and negative ion mode. The results of HPLC/ESI–MS/MS analysis are presented in the Supplementary Material. In other words, there are differences in MS behavior of the isomers upon "in-source" fragmentation in negative ion mode, however there are no differences in MS behavior of the isomers upon MS/MS fragmentation (in collision chamber). The fragmentation of ions upon MS/MS experiments occurs later than that "in-source". Therefore, it is reasonable to conclude that before the isomer ions reach the collision chamber, they isomerize to the identical structure. The fragmentation "in source" occurs almost immediately after the transfer of the ions from solution to the gas phase, thus the fragmentation reflects the structural differences of the biliverdin isomers present in solution.</p><p>The key question is why the differences upon fragmentation "in-source" of the isomers are in negative ion mode and not in positive ion mode. It is reasonable that in positive ion mode, protonation occurs at nitrogen atom of C ring. Such protonated biliverdin molecules may isomerize due to the resonance structures shown in Scheme 3.</p><p>In negative ion mode, deprotonation of biliverdin molecule occurs at a carboxyl group. It is clear that the isomerization of deprotonated biliverdin molecules is not easy as the isomerization of protonated biliverdin molecules.</p><!><p>The breakdown plots of ions [2BV−H]−, [BV−H]− and main fragment ion at m/z = 285, against cone voltage (V). The abundances of ions correspond to the respective peak areas obtained upon HPLC-ESI/MS analysis</p><!><p>As shown in Fig. 5 the gas phase stability of the [2BV−H]− ion (deprotonated dimer) of main isomer is definitely higher than the gas phase stability of the [2BV−H]− ion of the minor isomer (it is difficult to rationalize why at a cone voltage of 70 V we deal with an increase in [2BV−H]− ion abundances for both isomers). The gas phase stability of the [BV−H]− ion of the main isomer is also definitely higher than the gas phase stability of [BV−H]− ion of the minor isomer. For both isomers decomposition of ions [BV−H]− begins from the cone voltage 40 V (fragment ion at m/z = 285 is formed, Fig. 5). However, for the main isomer decomposition of [BV−H]− ion is amply compensated by the sensitivity increase (at higher cone voltage more ions reach the high vacuum region). It is also worth adding that the decomposition of ions [BV−H]− is compensated by the decomposition of ions [2BV−H]−. Taking into account the behavior of ions [2BV−H]− (Fig. 5) it is clear that compensation of ion [BV−H]− for the main isomer is more effective.</p><p>We have performed the breakdown plots of respective positive ions against cone voltage. As shown in the Supplementary Material, the breakdown plots for both isomers are similar.</p><!><p>Using HPLC-ESI/MS and NMR spectroscopy, the minor isomer IXα 5Z,10Z,15E was found in a commercially available biliverdin sample (beside the main IXα 5Z,10Z,15Z isomer). The isomers behave differentially upon "in-source" fragmentation in the negative ion mode (in contrast to their behavior upon "in-source" fragmentation in the positive ion mode and to their behavior upon MS/MS fragmentation in both modes). It is difficult to rationalize why this very geometric isomer is present in the analyzed biliverdin sample. The geometric isomers are often formed as a result of exposure to light. However, our sample had been stored in the dark and frozen. It should be emphasized that our finding does not exclude such a commercial biliverdin sample from its use for scientific (e.g. analytical) purposes. Quite the opposite, this sample may be useful for analysis of both isomers, at least for semi-quantitative analysis.</p><!><p>Biliverdin (as hydrochloride) was obtained from Sigma-Aldrich (Poznań, Poland) and used without purification.</p><p>1H NMR spectra were recorded on Agilent DD2 800 spectrometer, operating at frequency 799.83 MHz. All spectra were measured at 298 K. The signal assignment has been made on the basis of 2D spectra (gCOSY, gHSQCAD, gHMBCAD) and 1D selective NOE measurements (mixing time 500 ms). Samples were prepared by dissolution of 5 mg of biliverdin hydrochloride in 0.7 cm3 of [2H]4-methanol, containing 10% of [2H]2-water and 0.05% of NaOD or in 0.7 cm3 of [2H]3-acetonitrile, containing 10% of [2H]2-water.</p><p>1H NMR (CD3CN/D2O): δ = 7.97 (bs, 1H, H-10), 6.69 (bm, 1H, H-21), 6.51 (bs, 1H, H-15?), 6.49 (bs, 1H, H-5?), 6.17 (bm, 1H, H-32), 5.78 (bd, 1H, J = 11.4 Hz, H-22), 5.64 (bd, 1H, J = 18.0 Hz, H-22), 5.41 (bs, 1H, H-33), 4.98 (bs, 1H, H-33), 3.23 (bs, 4H, H-24, H-27), 2.69 (bs, 4H, H-25, H-28), 2.28 (s, 3H, H-31?), 2.27 (s, 3H, H-23?), 2.10 (s, 3H, H-30), 1.48 (s, 3H, H-20) ppm.</p><p>1H NMR data assigned to the 5Z,10Z,15Z-isomer (main) (CD3OD/D2O/NaOD): δ = 7.02 (s, 1H, H-10), 6.73 (dd, 1H, J = 11.7, 17.9 Hz, H-21), 6.52 (dd, 1H, J = 11.6, 17.6 Hz, H-32), 6.13 (s, 1H, H-15), 6.08 (s, 1H, H-5), 5.97 (dd, 1H, J = 2.2, 17.6 Hz, H-33), 5.62 (dd, 1H, J = 1.6, 17.9 Hz, H-22), 5.57 (dd, 1H, J = 1.6, 11.7, H-22), 5.36 (dd, 1H, J = 2.2, 11.6, H-33), 2.92 (m, 4H, H-24, 27), 2.35 (m, 4H, H25, 28), 2.18 (s, 3H, H-31), 2.13 (s, 3H, H-23), 2.11 (s, 3H, H-30), 1.84 (s, 3H, H-20) ppm.</p><p>1H NMR data assigned to the 5Z,10Z,15E -isomer (minor) (CD3OD/D2O/NaOD): δ = 6.93 (s, 1H, H-10), 6.70 (dd, 1H, J = 11.7, 17.9 Hz, H-21), 6.50 (dd, 1H, J = 11.6, 17.8 Hz, H-32), 6.13 (s, 1H, H-15), 6.03 (s, 1H, H-5), 5.87 (dd, 1H, J = 2.0, 17.8 Hz, H-33), 5.62 (dd, 1H, J = 1.6, 17.9 Hz, H-22), 5.56 (dd, 1H, J = 1.6, 11.7, H-22), 5.37 (dd, 1H, J = 2.0, 11.6, H-33), 2.92 (m, 2H, H-24), 2.89 (m, 2H, H-27), 2.35 (m, 4H, H25, 28), 2.19 (s, 3H, H-31), 2.14 (s, 3H, H-23), 2.08 (s, 3H, H-30), 1.86 (s, 3H, H-20) ppm.</p><p>The HPLC-ESI/MS analyses were performed using a Waters model 2690 HPLC pump (Milford, MA, USA), a Waters/Micromass ZQ2000 mass spectrometer (single quadrupole type instrument equipped with electrospray ion source, Z-spray, Manchester, UK). The software used was MassLynx V3.5 (Manchester, UK). Using an autosampler, the sample solutions were injected onto the XBridge C18 column (3.5 µm, 100 × 2.1 mm i.d., Waters). The injection volume was 10 mm3 of biliverdin-containing solution at concentration 0.05 mg/cm3. The solutions were analyzed using linear gradient of CH3CN–H2O with a flow rate of 0.3 cm3/min. The gradient started from 0% CH3CN—95% H2O with 5% of a 10% solution of formic acid in water, reaching 95% CH3CN after 10 min, and the latter concentration was maintained for 10 min.</p><p>The mass spectra were recorded in the m/z range 200–1200, in positive and negative modes simultaneously (during the HPLC/ESI–MS analyses the mass spectrometer was switched in the fast mode between the positive and negative ion modes). The electrospray source potentials were: capillary 3 kV, lens 0.5 kV, extractor 4 V, and cone voltage 20–70 V (indicated in each mass spectrum shown). The source temperature was 120 °C and the desolvation temperature 300 °C. Nitrogen was used as the nebulizing and desolvation gas at the flow rates of 100 and 300 dm3/h, respectively.</p><!><p>Supplementary material 1 (DOC 412 kb)</p><p>Electronic supplementary material</p><p>The online version of this article (10.1007/s00706-018-2161-7) contains supplementary material, which is available to authorized users.</p>

PubMed Open Access

An integrated photocatalytic/enzymatic system for the reduction of CO₂ to methanol in bioglycerol–water

A hybrid enzymatic/photocatalytic approach for the conversion of CO 2 into methanol is described. For the approach discussed here, the production of one mol of CH 3 OH from CO 2 requires three enzymes and the consumption of three mol of NADH. Regeneration of the cofactor NADH from NAD + was achieved by using visible-light-active, heterogeneous, TiO 2 -based photocatalysts. The efficiency of the regeneration process is enhanced by using a Rh(III)-complex for facilitating the electron and hydride transfer from the H-donor (water or a water-glycerol solution) to NAD + . This resulted in the production of 100 to 1000 mol of CH 3 OH from one mol of NADH, providing the possibility for practical application.

an_integrated_photocatalytic/enzymatic_system_for_the_reduction_of_co₂_to_methanol_in_bio

Introduction<!>Results and Discussion<!>Conclusion<!>Synthesis of photosensitized TiO 2<!>Synthesis of (NH 4 ) 2 [CrF 5 (H 2 O)]<!>Synthesis of [Cp*Rh(bpy)H 2 O]Cl 2<!>UV-vis characterization<!>Regeneration of NADH from NAD +

<p>The reduction of CO 2 to fuel is a technology that could contribute to the recycling of large quantities of carbon. Among the various routes, the enzymatic reduction of CO 2 in inexpensive H-donor solvents to produce methanol (or other C 1 molecules) is an attractive option. Redox enzymes are of industrial interest as they may catalyze reactions in which the use of conventional chemical catalysts is restricted [1]. Unfortunately, their application is quite limited due to the high cost of their cofactors. A huge effort is being made for the in situ regeneration of these cofactors using various approaches such as: the use of secondary enzymes, electrochemical regeneration, or even the use of living cells. Noteworthy is that the regeneration of the cofactor often involves the potential production of a variety of isomers of the active species or even the formation of dimers that may not be active in promoting the enzymatic reaction or may even act as inhibitors. In nature, cofactors are usually regenerated via enzymatic reactions. One of the most interesting cofactors is nicotinamide adenine dinucleotide (NAD + ), a cofactor of the oxydoreductase class of enzymes. NAD + , together with its reduced form, 1,4-NADH, plays an essential role in many metabolic processes of living cells. NADH is also important in industrial biocatalysis, namely in the process of reductive synthesis of chiral organic compounds [2,3].</p><p>Currently, the enzymatic reduction of carbon dioxide is under investigation as a possible route for fuel production [4]. A specific application, which leads to the production of methanol, occurs in water through three 2esteps based on the use of three enzymes, namely: formate dehydrogenase (F ate DH), formaldehyde dehydrogenase (F ald DH), and alcohol dehydrogenase (ADH). These enzymes promote the cascade reduction of CO 2 to methanol through formic acid (F ate DH), formaldehyde (F ald DH) and aldehyde (ADH). The reduction process is enabled by NADH, which is oxidized to NAD + . However, the production of one mole of CH 3 OH from CO 2 requires the consumption of three moles of NADH (Figure 1). Therefore, the regeneration of NADH is necessary for practical application of the described process. In nature, the endergonic process of NAD + reduction to NADH is performed with the help of solar energy during photosynthesis. Presently, a substantial effort is being made for the regeneration of 1,4-NADH using a variety of strategies, including the use of enzymatic catalysis, as well as chemical, electrochemical and photochemical methods [5,6]. For industrial application, such reduction would require implementation of the most energetically and economically convenient technologies, such as visible-lightdriven photocatalysis, as chemical methods are either too expensive or not compatible with the enzymes [7]. To date, photocatalysis has shown great potential for photodegradation of environmental pollutants [8,9]. Most interesting is that photo-catalysis may play a relevant role in the conversion of large quantities of CO 2 into fuel by using water or waste organics as the hydrogen source [7,10,11]. Therefore, integrated photochemical CO 2 reduction/organic oxidation and H 2 O splitting have received significant interest due to their potential environmental and resource preservation benefits [12,13]. Interestingly, the oxidized forms of some common waste organics may find practical application.</p><p>The use of heterogeneous photocatalysts in the process of NADH regeneration from NAD + would be of great interest due to their low cost, moderate (ambient) operational conditions and acceptable environmental impact. The most extensively applied photochemical processes are based on the use of TiO 2 as a photocatalyst in oxidation reactions [14][15][16]. While pure TiO 2 has a band gap energy of 3.2 eV (which is not compatible for use with visible light), modified TiO 2 is known to be more suitable for carrying out photocatalytic processes utilizing the visible part of the solar spectrum [17].</p><p>In previous work, systems based on stable, encapsulated enzymes [7] for the enzymatic reduction of CO 2 to CH 3 OH in water combined with the near-UV-vis light driven photoregeneration of NADH for increasing the CH 3 OH/NADH molar ratio was described [7]. However, this approach is a hybrid process involving: the enzymatic reduction of CO 2 to CH 3 OH promoted by the reduced form of cofactor NADH, and the in situ photocatalytic reduction of NAD + to NADH under visible-light irradiation, using semiconductors in water/bioglycerol mixtures. Bioglycerol is being produced in increasing volumes for biodiesel production from oleaginous seeds. New applications for this product are being investigated [7,11,18,19] and its use as a H-source providing its oxidized derivatives (or even C 2 molecules [11]) may be an interesting path for the economically viable use of large volumes of CO 2 . Our ultimate goal is to attain a highly efficient and selective reduction of NAD + to 1,4-NADH or other equally active isomers upon visible-light irradiation in order to make the enzymatic reduction of CO 2 to CH 3 OH viable.</p><!><p>In previous work [7] we demonstrated that encapsulated F ate DH, F ald DH, and ADH enzymes are able to rapidly (<1 min) reduce CO 2 in proton-donor solvents under pH-controlled conditions, resulting in CH 3 OH at room temperature. The regeneration of NADH was attempted using both chemical and photochemical techniques. The former affects the enzymes that are quickly deactivated, while the latter techniques are more interesting and produce up to a few mol of methanol per mol of NADH (with respect to 3 NADH per methanol) as shown in Figure 1. As previously discussed [7], the success of the regeneration depends on the separation of the enzymatic reduction from the photoregeneration of NADH, thus a two-compartment reactor was used (see the Experimental section). In fact, the light most likely affects the enzyme activity by inducing structural modifications. In [7] a photocatalyst was used that operates on the border of the UV-vis spectrum. As described above, the goal of this research is to work in the visible-light range, possibly using direct irradiation with solar light. Therefore, we have synthesized a number of semiconducting materials showing photocatalytic activity under visiblelight irradiation. The most active material was selected and fully characterized regarding its photoelectrical-and chemical-properties. Optimal conditions for use with visible light for the reduction of NAD + using bioglycerol as a H-donor and a Rh(III)-complex as an e − -H + transfer agent were found. It was shown that the photocatalyst, the electron mediator and the H-donor have suitable energy levels that can be combined together for an effective recycling of NAD + . The cofactor can be used several times in combination with the encapsulated enzymes, which promote the reduction of CO 2 to methanol. In this work we discuss in detail the utilization of a new TiO 2 photocatalyst (Degussa P25 or 10 nm particles produced in-house [20]) modified with the inorganic complex [CrF 5 (H 2 O)] 2− [20]. Additionally, the properties of other photocatalysts (Cu 2 O, InVO 4 , and TiO 2 , which are less active than the Cr-modified TiO 2 photocatalysts), which were modified with the organic compound "rutin", are briefly presented. Figure 2 shows the transformed diffuse reflectance spectra of the photocatalysts converted by the Kubelka-Munk function.</p><p>The properties of the photocatalysts used can be summarized as follows. Cu 2 O is a visible-light-absorbing, red-colored, p-type semiconductor with the band gap energy equal of 2.1 eV. InVO 4 is an n-type semiconductor (E bg = 2.8 eV). TiO 2 modified with the organic compound rutin (rutin@TiO 2 ) shows an electron injection into the conduction band of TiO 2 as the result of a direct molecule-to-band charge transfer (MBCT) within the surface-formed, colored, charge-transfer complex of titanium(IV) [21]. TiO 2 with an adsorbed chromium(III) anionic complex [CrF 5 (H 2 O)] 2acts as a photosensitizer by injecting electrons into the conduction band of titania [20].</p><p>Photocatalytic tests of NADH regeneration using these materials have been performed using visible light irradiation (λ > 400 nm). The results are summarized in Figure 3.</p><p>Figure 3a shows that the conversion yield of NAD + into NADH is similar for the various photocatalysts. Additionally, the selec-tivity towards 1,4-NADH appears to be not very high. The concentration of the photogenerated, reduced form of the cofactor after 6 hours of visible-light irradiation in the presence of water is within the range of 0.1-0.15 μM. Noteworthy is that no or negligible reaction was observed when any of the components of the system was missing: photocatalyst, light, NAD + , or water. TiO 2 alone was tested as a reference material yielding only minor traces of a complex mixture of products under the same experimental conditions. HPLC in combination with NMR analysis was employed for the detection of the various isomers of the NAD + -reduction products, namely the 1,4-NADH, 1,2-NADH, 1,6-NADH, or dimeric species. We have observed that when the photocatalysts alone were used, the selectivity towards 1,4-NADH was significantly lower than 100%. In fact the low amount of 1,4-NADH reported in Figure 3a is due to NAD + conversion resulting in a mixture of compounds (including dimers) with a low selectivity (approximately 5%) towards 1,4-NADH. This is most likely due to the fact that the reactions take place on the surface of the photocatalyst without any selectivity. Conversely, the regeneration of 1,4-NADH was much more efficient via an indirect route of H + -e − transfer, using hydride-transfer agents coupled to the photocatalytic materials. To implement such a strategy, the above described photoactive materials were coupled to the well-known [22] [Cp*Rh(bpy)(H 2 O)]Cl 2 [aquo(2,2'-bipyridine)(pentamethylcyclopentadienyl)]rhodium(III), where Cp* = pentamethylcyclopentadienyl. For comparison we have also used its iridium analog, with phenantroline as a bidentate N-ligand replacing bpy. Iridium showed interesting activity, comparable to that of Rh. A key point in our approach was to demonstrate that the H + -e − transfer system (photocatalyst, transition metal complex) we designed had an ideal potential for e − transfer and could operate in combination with the H-donor and the enzyme to eventually convert CO 2 into CH 3 OH. The success of this system was evidenced by measuring the amount of photogenerated 1,4-NADH using water or water/glycerol as electron donor. Figure 3b illustrates the results of the photocatalytic regeneration of NADH in presence of various materials after 6 hours of irradiation. Interestingly, for [CrF 5 (H 2 O)] 2− @TiO 2 the yield is 70 times higher in the presence of the electron mediator (under the same experimental conditions). The Cr-modified TiO 2 showed the highest activity among the tested photocatalysts [20] and thus, the focus of the discussion is now shifted to this photocatalyst. Although CrF 3doped TiO 2 is reported in the literature [8] to be active in oxidation processes of waste organics, neither TiO 2 loaded with anionic [CrF 5 (H 2 O)] 2− nor CrF 3 on TiO 2 (used for reduction purposes) has been described so far.</p><p>The 19 F NMR spectrum confirms the presence of the anionic form on the TiO 2 surface with a signal at 121 ppm, while free [CrF 5 (H 2 O)] 2− shows a signal at 122 ppm (see Experimental section). In contrast, CrF 3 has signals in a completely different region (−127 and −128 ppm). Therefore, we conclude that the supported form of Cr on TiO 2 is the anionic complex [CrF 5 (H 2 O)] 2− . Another issue in this process was to demonstrate whether the reduction is selective towards 1,4-NADH and no other isomers or if dimers were formed. The answer to this question was provided by using 1 H and 13 C NMR in connection with HPLC. The anionic pentafluorochromate-modified TiO 2 coupled to the Rh(Ir)-transition metal complex appeared very selective towards 1,4-NADH, while neither isomers nor dimers were formed. Figure 4 shows typical 1 H NMR spectra recorded during the course of the reduction: a pseudo-quartet centered at 2.65 ppm increases with time showing that the reaction occurs. This signal is due to 1,4-NADH, as shown in Figure 5. Here, the 1 H NMR spectrum of standard NADH (a commercially available product) is compared with the spectra of the reduction products formed in presence and absence of the hydride-transfer agent used together with the [CrF 5 (H 2 O)] 2− @TiO 2 photocatalyst. The green and blue spectra were taken after 6 h of irradiation with solar light or white light under the same operative conditions with and without the Rh complex. They show that the presence of the Rh mediator improves the conversion rate.</p><p>It is known that the reduction of the [Cp*Rh(bpy)(H 2 O)] 2+ 1 complex to [Cp*Rh(bpy)] 2 adds a proton and results in the conversion into a hydrido form. This product is an efficient and selective reduction catalysts of NAD + to 1,4-NADH [22]. The resulting active hydrido form, [Cp*Rh(bpy)H] + 3, transfers a hydride ion to the 4-position of NAD + (coordination to the @TiO 2 is −0.58 V vs NHE, as measured in the present study using a previously published methodology [24]. The electrode covered by [CrF 5 (H 2 O)] 2− @TiO 2 generates a photocurrent upon visible light irradiation, proving a photoinduced electron transfer from the excited chromium(III) com-plex to the conduction band of TiO 2 (Figure 6). The following step, that is, the transfer of electrons from the conduction band of the photocatalyst to the oxidized form of the rhodium complex (according to Equation 1), is thus thermodynamically feasible. The photogenerated holes can regain electrons via the oxidation of glycerol. The reduced complex (Rh(I)) reacts with a proton yielding a Rh(III)-hydrido species (Equation 2). The resulting Rh-hydrido-species transfers the hydride to NAD + affording NADH (Equation 3 and Figure 7). Such steps, already hypothesized in the literature [23,25], are clearly demonstrated in the present work through the following experiments. First, [Cp*Rh(bpy)H 2 O] 2+ was converted into [Cp*Rh(bpy)H] + upon reaction with elemental hydrogen. The UV-vis absorption spectrum recorded after the reaction shows the appearance of a band at 521 nm that is characteristic of the formation of the rhodium hydride. This was confirmed by taking the spectrum of the isolated complex. The addition of NAD + resulted in NADH formation (a band at around 344 nm) in concurrence with the disappearance of the 521 nm band (Figure 8). The formation-disappearance of the hydride was further confirmed by 1 H NMR where a signal at −7.5 ppm (in the same region as the analog [Cp*Rh(6,6'-dimethyl-2,2'bipy)H] + [22]) was evident. This 1 H NMR signal was correlated with the disappearance of the 521 nm band in the UV-vis spectrum, along with the appearance of the characteristic band at approximately 344 nm. The process was cyclic and the appearance-disappearance of the hydride signal followed the change in position of the UV-vis band from 521 to 344 nm and back.</p><p>The spectral changes in the spectrum of [CrF 5 (H 2 O)] 2− are reported in Figure 9 together with the cyclic voltammograms. The new combined system described in this paper has very high activity and specificity upon visible light irradiation. The extraordinary activity of [CrF 5 (H 2 O)] 2− @TiO 2 can be explained by an efficient photoinduced electron transfer from Cr(III) to the conduction band of TiO 2 and further to the adsorbed rhodium complex. A hindered back electron transfer from the Rh species to the photocatalyst can also be responsible for the overall effi- ciency of the process. Substitution of Rh with Ir and of bipyridine with phenanthroline did not improve the yield to appreciable extent, thus the Rh-complex was used. Other photomaterials which are active upon visible light irradiation were also tested, such as Fe/ZnS, Co/ZnS, Ag/ZnS, ZnBiO 4 , AgVO 4 , NiO, CrF 3 @TiO 2 , tiron@TiO 2 . However, they did not show significant activity in the NADH photoregeneration process when compared to the [CrF 5 (H 2 O)] 2− @TiO 2 photocatalyst. As mentioned above, glycerol was considered as an electron donor in aqueous solution. When only water was used in the photoreduction, the reaction rate was very low due to the weak ability of H 2 O to transfer electrons. Figure 10 shows the influence of the nature of the electron donor and the concentration of the electron mediator, on the reduction rate. The NADH regeneration rate exhibits a strong dependence on the concentration of glycerol, which plays the crucial role of a sacrificial electron donor. Figure 10 also shows that the increase of the glycerol concentration from 0 M (the lowest point on the black line in Figure 10) to 0.001 M (the second point in the same figure) results in an extremely rapid increase of the reaction rate. It must be emphasized that the reaction in the pres-ence of other electron donors, such as triethanolamine or isopropanol, was also tested but were not comparable with glycerol. For practical applications the mixture of glycerol 0.1-0.5 M in water was used. These findings demonstrate that the photocatalysts described in this paper are able to utilize visible light for the generation of the electron-hole couples (exciton). The excited electrons are transferred to reaction centers at the surface of the photo-materials, and the conversion of NAD + to NADH occurs, involving the electron mediator and the oxidation of glycerol. The overall mechanism is summarized in Figure 11. The first and only product of glycerol oxidation is 1,3-dihydroxyacetone (1,3-DHA), as demonstrated by NMR studies carried out on a fresh reaction mixture ( 1 H resonances found at 4.34 ppm, as compared with literature data: 1 H at 4.40 ppm) [26]. The dihydroxyacetone species is not very stable and can be easily converted into other monomeric or polymeric species under the reaction conditions. Oxidation of glycerol is still under investigation as the isolation of 1,3-DHA would add value to the process. Other parameters, such as the optimization of the photosystem for a more efficient NAD + reduction, the improvement of the reaction selectivity, the increase of the absorption coefficient, and the increase of the chemical stability and photostability of photocatalysts, are also under continuous investigation for their further improvement and bring the whole reaction closer to a potential application.</p><!><p>Titanium dioxide modified with chromium(III) complex, [CrF 5 (H 2 O)] 2− @TiO 2 , exhibits great ability to drive the in situ selective reduction of NAD + cofactor to 1,4-NADH. This process works particularly well in the presence of a [Cp*Rh(bpy)H 2 O]Cl 2 complex playing the role of the electron transfer mediator. Our studies demonstrate that the photocatalyst is able to utilize visible light for generation of the electron-hole couples. The excited electrons are transferred to an e − -transfer mediator at the surface of the materials, where the conversion of NAD + to NADH occurs, involving the eventual oxidation of glycerol. The overall mechanism is summarized in Figure 11.</p><p>Anodic photocurrents generated by the photocatalyst upon visible light irradiation suggest the electron transfer from the excited sensitizer (chromium(III) complex) to the conduction band of TiO 2 . The reduction of the rhodium complex was confirmed by spectroscopic studies made in the presence of irradiated TiO 2 and upon electrochemical reduction of the complex. Selective reduction of NAD + to 1,4-NADH has also been experimentally evidenced by 1 H NMR and HPLC measurements. Finally, the enzymatic route of the CO 2 reduction was confirmed to occur as was previously described [7,11]. The novelty of the photochemical NADH regeneration described in this paper consists of the use of visible light and the coupling of an inexpensive photocatalyst with robust e − -and H − -transfer mediators, which can be used for days. Furthermore, we have demonstrated the entire mechanism of the electron flow from the photo-excited photosensitizer to NAD + , resulting in NADH generation, which can be further used in enzymatic processes including carbon dioxide reduction.</p><p>The product of glycerol oxidation is 1,3-dihydroxyacetone. This species is not very stable and can be converted into other monomeric or polymeric species. Oxidation of glycerol is still under investigation together with other parameters, such as the optimization of the photosystem for a more efficient NAD + reduction, the improvement of the reaction selectivity, the increase of the absorption coefficient, and the increase of the chemical stability and photostability of photocatalysts.</p><p>The photocatalytic system discussed in this paper is much more effective as compared to ZnS-based photocatalysts as previously presented in [7,11] and represents a significant step towards potential application of this hybrid technology for CO 2 reduction to methanol.</p><!><p>The modification of TiO 2 with rutin was carried out as reported in the literature [27]. A titanium dioxide powder, P25 Evonik (500 mg), was added to 10 cm 3 of aqueous rutin solution (10 −2 mol dm −3 ). The suspension was sonicated (10 minutes) and the colored precipitate was collected, washed 3 times with water and dried in air at 60 °C.</p><p>[CrF 5 (H 2 O)] 2− @TiO 2 was prepared by impregnation of TiO 2 particles (P25, Evonik or 10 nm particles) [20] by (NH 4 ) 2 [CrF 5 (H 2 O)] under ultrasonic stirring. The suspension was sonicated for 15 minutes, left for 24 hours, and the nanoparticles were isolated washed with water and dried under vacuum.</p><p>Copper(I) oxide was prepared in the reaction of an aqueous solution of glucose (10 mL, 0.8 M) dropped into an alkaline solution of CuSO 4 (50 mL, 0.2 M) in presence of polyvinylpyrrolidone K-30 (0.3 g) at 80 °C. After 1 hour a red precipitate was separated by filtration, washed with water and dried.</p><p>Vanadates were prepared as reported by Hu et al. [28] with minor modifications: NaOH and V 2 O 5 powders in a molar ratio of 6:1 were dissolved together in water and stirred. Subsequently, the solution of In(VO 3 ) 3 or AgNO 3 was added. Precipitates, which appeared immediately, were aged at room temperature for 1 hour, washed and dried.</p><!><p>5 mL of an ammonia solution was added to 50 mL of an aqueous solution of NH 4 •HF (15 g). A CrF 3 solution was added dropwise to a hot (363 K) solution of NH 4 •HF in the presence of zinc powder. A green precipitate was obtained, isolated and analyzed, resulting in the title compound.</p><!><p>[Cp*Rh(bpy)H 2 O]Cl 2 was obtained from [Cp*RhCl 2 ] 2 as reported in [29]. Phenanthroline was used instead of bpy to</p><!><p>The UV-vis diffuse reflectance spectra of the photocatalysts were recorded using a UV-3600 spectrophotometer (Shimadzu) equipped with an integrating sphere. Powder samples were ground with BaSO 4 (1:50 wt ratio). Barium sulfate was used as a reference.</p><!><p>Photocatalytic tests of NADH regeneration were performed in a borosilicate glass reactor (V = 10 mL). The photocatalyst (1 g L −1 ) was suspended in deoxygenated phosphate buffer (pH 7). NAD + (0.8 mM) was added. Glycerol was used as an electron donor. [Cp*Rh(bpy)H 2 O]Cl 2 was used as electron mediator (concentration range: 0-0.5 μM). The suspension was irradiated in the sealed reactor, under nitrogen atmosphere, using a 50 W LED illuminator (λ > 400 nm) as light source or concentrated solar light. 2 mL samples were collected during irradiation, filtered and analyzed by HPLC (column Zorbax SB-Aq) and NMR.</p><p>Reduction of CO 2 to CH 3 OH using the assembled photocatalytic/enzymatic system The reduction of CO 2 to methanol using encapsulated enzymes is described in [7]. Here we recall that the enzymes F ate DH, F ald DH, and ADH were encapsulated into silicate cages made from Ca alginate and tetraethoxysilanes (TEOS) and used at a controlled pH of 7. CO 2 was admitted in a continuous flow reactor at such a rate to generate a 20% excess with respect to the stoichiometric amount required by the enzymes. Excess CO 2 was avoided to prevent its emission into the atmosphere since a goal of the research is to reduce CO 2 emissions during the synthesis of methanol. For this reason, the recovery of methanol by stripping was carried out using N 2 and not CO 2 itself, which would also be possible. A more complex experiment with CO 2 recovery is currently under investigation, so that CO 2 can be used as reagent and carrier of methanol. In this study, air influenced the enzymes and was not used. The beads (Figure 12) were suspended in water in compartment A of the reaction cell below (Figure 13) in presence of the required amount of NADH, and CO 2 was slowly admitted. When the formation of methanol (monitored by GC on withdrawn samples of the reaction solution, within 1 min maximum) was at the maximum, the solution was pumped into compartment B, which contained the photocatalyst and the heterogenized Rh complex, while the encapsulated enzymes remained in compartment A. Irradiation (0.5-1 h) with visible light (or solar light) caused the conversion of NAD + into NADH to occur at the maximum yield (very close to 100%). The solution was again pumped into compartment A where the reduction of CO 2 occurred with formation of CH 3 OH. The cycle was repeated until no more CH 3 OH was formed. At given intervals, CH 3 OH was extracted in compartment B (stripping was realized by bubbling N 2 and condensing of the vapors) to avoid an increasing concentration which might block the enzymes. The difference in the rate of the enzymatic reaction and the photocatalytic regeneration of NADH is a barrier to practical utilization of this hybrid technology. Further efforts for improving the convergence of the time of reaction in the two steps is currently underway.</p>

Beilstein

Allosteric modulation of protein oligomerization: an emerging approach to drug design

Many disease-related proteins are in equilibrium between different oligomeric forms. The regulation of this equilibrium plays a central role in maintaining the activity of these proteins in vitro and in vivo. Modulation of the oligomerization equilibrium of proteins by molecules that bind preferentially to a specific oligomeric state is emerging as a potential therapeutic strategy that can be applied to many biological systems such as cancer and viral infections. The target proteins for such compounds are diverse in structure and sequence, and may require different approaches for shifting their oligomerization equilibrium. The discovery of such oligomerization-modulating compounds is thus achieved based on existing structural knowledge about the specific target proteins, as well as on their interactions with partner proteins or with ligands. In silico design and combinatorial tools such as peptide arrays and phage display are also used for discovering compounds that modulate protein oligomerization. The current review highlights some of the recent developments in the design of compounds aimed at modulating the oligomerization equilibrium of proteins, including the “shiftides” approach developed in our lab.

allosteric_modulation_of_protein_oligomerization:_an_emerging_approach_to_drug_design

Introduction<!>Mechanisms of modulating the oligomerization equilibrium of proteins<!><!>Mechanisms of modulating the oligomerization equilibrium of proteins<!>The tumor suppressor p53<!><!>The tumor suppressor p53<!><!>The tumor suppressor p53<!>Viral proteins<!><!>Viral proteins<!><!>Viral proteins<!><!>Viral proteins<!>Fibril-forming proteins<!>Conclusion<!>The shiftides concept<!><!>The shiftides concept<!>Conflict of interest statement

<p>Oligomerization is a common property of proteins and takes place in all biological systems. It is estimated that at least 35% of all proteins in cells are oligomeric (Jones and Thornton, 1996; Goodsell and Olson, 2000). The properties of protein oligomers are highly diverse: Protein oligomers may be homoligomers (Wang et al., 1994; Eisenstein and Beckett, 1999; Yun et al., 2007; Thulin et al., 2011) or heterooligomers (Fermi et al., 1984; Cramer et al., 2001; Unwin et al., 2002; Gomez et al., 2011) and can range from dimers (Sapienza et al., 2007) to high order structures such as capsids (Wu and Rossmann, 1993; Nam et al., 2011) and fibrils (Craig and Woodhead, 2006; Bedrood et al., 2012). Some proteins form one specific active oligomeric state. This is the case with hemoglobin, which exists in red blood cells as an α2β2 heterotetramer (Fermi et al., 1984), and with the acetylcholine receptor that is a pentamer (Unwin et al., 2002). Other proteins are involved in dynamic oligomerization equilibria between several states with different activities, and switch between these states as part of regulating their normal function. For example, the enzyme carbamoyl phosphate synthase (CPS) exists in equilibrium between inactive dimers and active tetramers. The equilibrium is regulated by allosteric inhibitors that stabilize the dimer and allosteric activators that stabilize the tetramer (Kim and Raushel, 2001; Mora et al., 2002). In many cases, the dynamic assembly and disassembly of oligomers plays a central role in regulating the activity of a protein, as in the case of actin (Ono, 2007), which induces cell motility by this mechanism.</p><p>Correct protein oligomerization is critical for its function and is therefore tightly regulated by various factors. For example, the GroEL-GroES chaperonin complex assists in the folding of polypeptides under thermal stress conditions. The complex undergoes small changes in the inter-subunit interactions when the temperature increases from 37 to 42–45°C. This may enable it to distinguish normal temperatures from stress temperatures (Cabo-Bilbao et al., 2006). Protein oligomerization can also be regulated by the binding of partner proteins (Grossman, 2001; Fernandez-Fernandez et al., 2005; Marinho-Carvalho et al., 2006; De Meyts, 2008; Rajagopalan et al., 2008), metal cofactors or small molecule allosteric effectors (Kim and Raushel, 2001; Lawrence et al., 2008; Selwood et al., 2008; Semenova and Chernoff, 2012). In a notable example, Krojer et al. discovered that the bacterial HtrA protease exists in an inactive hexameric state, which undergoes an extensive conformational change upon binding to effector peptides. This change involves the rearrangement of the active site to a catalytically active structure and induces the formation of 12-mer or 24-mer oligomeric states (Krojer et al., 2010). Changes in the strength and geometry of interactions between subunits can also be modulated by ATP hydrolysis (Zhang et al., 2010).</p><p>Since oligomerization is very common and is crucial for protein activity, modulating this process is a highly promising therapeutic strategy that can be applied to many different diseases involving oligomeric proteins (Hayouka et al., 2007; Lawrence et al., 2008; Christ et al., 2012; Gabizon et al., 2012). Several compounds that are already in clinical use were later discovered to act via modulation of protein oligomerization. A well-known example is that of the anti-cancer drug Taxol, which was discovered in the stem bark of the pacific yew during the screening of plant-derived compounds for cytotoxic activity (Wani et al., 1971). It was later discovered that Taxol binds to β-tubulin (Löwe et al., 2001) and allosterically inhibits the assembly and disassembly dynamics of microtubules (Jordan et al., 1993; Derry et al., 1995), thus interfering with mitosis and inhibiting the division of cancer cells.</p><p>Many compounds that modulate protein oligomerization are still being discovered indirectly by screening methods that do not target protein oligomerization. However, a growing number of studies aim to discover molecules that directly target the oligomerization of a well characterized protein. The understanding of oligomeric protein structures and how they are regulated significantly progressed in recent years. Structures determined by NMR and X-ray crystallography enable detailed characterization of the oligomerization interfaces and the interactions that stabilize the oligomers (see for example Fermi et al., 1984; Jeffrey et al., 1995; Lange-Savage et al., 1997; Luger et al., 1997; Walters et al., 1997; Whitson et al., 2005; Sharma et al., 2010). Methods for precise determination of the oligomeric states of a protein and their relative populations, such as size exclusion chromatography (SEC) (Mateu and Fersht, 1998; Gotte et al., 2012; Yu et al., 2013), analytical ultracentrifugation (AUC) (Weinberg et al., 2004a; Murugan and Hung, 2012; Szymanski et al., 2013) and single molecule methods (Groulx et al., 2011; Paredes et al., 2012; Calebiro et al., 2013), improve our understanding of the thermodynamics and kinetics of oligomerization processes. Based on this knowledge, molecules that shift the oligomerization equilibrium of target proteins are being developed using techniques ranging from de novo design to combinatorial screening. This review will cover the latest developments in this field and their application to various disease-related proteins. The compounds discussed in this study are summarized in table S1.</p><!><p>Modulation of protein oligomerization can take place by various mechanisms. In the simplest mechanism, inhibition of oligomerization can be achieved by molecules that bind directly to the oligomerization interface and competitively block it (He et al., 2005), thus preventing oligomerization (Figure 1A). These competitive oligomerization inhibitors do not act by stabilizing a specific oligomeric state and are thus outside the scope of this review. Alternatively, molecules may stabilize a specific oligomer by binding near the oligomerization interface (Kessl et al., 2009) (Figure 1B) or by binding to several monomers simultaneously (Teufel et al., 2007) (Figure 1C).</p><!><p>Schematic illustrations of mechanisms of modulation of protein oligomerization. (A) Direct blocking of oligomerization interfaces; (B) Binding to or near the oligomerization interface and stabilizing it; (C) Binding to several monomers simultaneously; (D) Morpheein mechanism (Jaffe, 2005). The conformations determine the stoichiometry of the full oligomer. In this example, a ligand binds the monomer and stabilizes the conformation that promotes formation of a different oligomer.</p><!><p>A different mechanism for allosteric modulation of oligomerization equilibria was described by Jaffe et al. (Jaffe, 2005). In this mechanism, a lower oligomeric form of the protein can exist in different conformations (called morpheeins), and each conformation dictates a defined stoichiometry for the higher oligomer. Transition between the oligomeric states requires dissociation of the oligomer and a change in conformation before the other oligomeric state is formed. Therefore, molecules that allosterically stabilize a certain conformation of the lower oligomer will shift the oligomerization equilibrium toward the corresponding higher oligomer (Figure 1D). This mechanism has been shown for the enzyme phorphobilinogen synthase (PBGS), which is involved in tetrapyrrole metabolism and plays a crucial role in cellular respiration (Jaffe and Lawrence, 2012). PBGS exists in equilibrium between active octamers and inactive hexamers. The transition between the two states requires dissociation into dimers followed by a conformational change in the dimer and reassociation. This process can be modulated by allosteric effectors, such as magnesium, which specifically stabilizes the octamer (Jaffe, 2005). Lawrence et al. used in silico screening to develop a compound that inhibits pea PGBS by binding specifically to the inactive hexamer (Lawrence et al., 2008), thus proving the feasibility of modulating the activity of proteins by shifting their oligomerization equilibrium. Many proteins exhibit characteristics indicating that their oligomerization dynamics follow the morpheein mechanism (reviewed in Selwood and Jaffe, 2012).</p><!><p>The tumor suppressor p53 is a transcription factor that is activated and accumulated in the nucleus in response to oncogenic stress. Following its induction, p53 binds specific promoters in the genome and activates the transcription of a wide array of target genes, aimed at eliminating the threat of malignant transformation (Levine, 1997; Vogelstein et al., 2000; Ryan et al., 2001; Michael and Oren, 2002). p53 is mutated in over 50% of all cancer cases, highlighting the vital role it plays in tumor suppression. The majority of cancer-associated mutations in p53 occur in its DNA binding core domain (Levine, 1997).</p><p>p53 is active as a homotetramer (Chene, 2001) and its tetramerization is mediated by a structurally independent tetramerization domain (p53Tet, residues 326–355) (Clore et al., 1994; Lee et al., 1994; Jeffrey et al., 1995). Tetramerization of p53 is vital to its function and plays a central role in the regulation of p53 activity. p53 tetramers bind p53 DNA response elements more tightly than dimers and monomers, and only tetramers can induce transcription of p53 target genes (Weinberg et al., 2004b; Menendez et al., 2009). Tetramerization also affects the cellular localization of p53: the Nuclear Export Signal (NES) of p53 is located within the tetramerization domain and is shielded in p53 tetramers, preventing nuclear export of p53 tetramers (Stommel et al., 1999). However, in monomers and dimers of p53 the NES is exposed and p53 is thus exported from the nucleus to the cytoplasm, where it is degraded via the ubiquitin-proteasome pathway.</p><p>The oligomerization equilibrium of p53 is regulated by interactions with other proteins, such as proteins from the 14-3-3 and S100 families (Fernandez-Fernandez et al., 2005, 2008; Rajagopalan et al., 2008; Słomnicki et al., 2009; Van Dieck et al., 2009a) and numerous kinases (Delphin et al., 1997; Gotz et al., 1999). Post translational modifications also have an effect on p53 oligomerization, either by directly affecting tetramer stability (Nomura et al., 2009; Yakovlev et al., 2010) or by modulating the interactions of p53 with other proteins (Rajagopalan et al., 2008; Van Dieck et al., 2009b). Recently, using fluorescence correlation spectroscopy in single cells, Gaglia et al. showed that DNA damage causes the shifting of the oligomerization equilibrium of p53 toward tetramers, and that this change is sufficient to activate the transcription of p53 target genes even without the net accumulation of p53 (Gaglia et al., 2013). The importance of tetramerization for p53 function makes p53 an attractive therapeutic target for compounds that modulate protein oligomerization. Several recent projects utilized different strategies to shift the oligomerization equilibrium of p53 toward the active tetramer.</p><p>Ligands containing several spaced cationic groups bound the p53 tetramerization domain and stabilized p53 tetramers. These ligands were developed using a combination of intuitive design with computational and combinatorial methods. Salvatella et al. designed a tetraguanidinium ligand that binds to a patch of negatively charged residues on the surface of the p53 tetramerization domain (Figure 2A), facing outwards from the dimer-dimer interface (Salvatella et al., 2004). This ligand was used by Martinell et al. as a basis for the computational design of a peptide with four arginine residues with similar spacing as the guanidinium groups in the original ligand. The new peptide (CAN4) bound p53Tet with a Kd of 8 μM and increased the thermal stability of p53Tet by 2°C. The same group later synthesized a library of modified peptides and tested their binding to p53Tet. Several peptides in the library bound p53Tet with affinities as low as 0.8 μM (Martinell et al., 2006).</p><!><p>Modulation of p53 oligomerization. (A) Left: Structure of the tetraguanidinium ligand described in Salvatella et al. (2004) and sequence of the peptide CAN4 described in Martinell et al. (2006). Right: Stick model and space filling model (indicated residues in red) of the acidic patch on the surface of the p53 tetramerization domain targeted by the two molecules. Structures from pdb 1PET (Lee et al., 1994); (B) Left: Structure of guanidinium-calix[4]arene described in Gordo et al. (2008) (1) and imidazole calix[6]arene described in Kamada et al. (2010) (2). Right: structure of the acidic patch targeted by these molecules; (C) Discovery of peptides that bind specifically to tetrameric p53. A peptide array derived from p53-binding proteins was screened for binding to p53CTD (left). The binding of the identified peptides was quantified by fluorescence anisotropy (center) and their effect on p53 oligomerization was characterized using AUC (right).</p><!><p>Gordo et al. focused on the cancer associated R337H mutant of p53, in which a critical hydrogen bond is lost due to the mutation, resulting in destabilization of the p53 tetramer. They designed a calix[4]arene with 4 guanidiniomethyl groups, in which a hydrophobic calixarene group binds to the hydrophobic pocket formed in the dimer-dimer interface of the p53Tet tetramer and the guanidinium groups form hydrogen bonds with acidic residues located on different monomers (Figure 2B). The designed ligand bound the R337H mutant and increased the thermal stability of the mutant to the same level as the wild type p53Tet (Gordo et al., 2008). In a later work, Gordo et al. showed that increasing the flexibility of the ligand increases its affinity to the R337H by enabling it to accommodate the optimal geometry for binding more easily (Gordo et al., 2011). However, the activity of this ligand was tested in water, and experiments at physiological ionic strength did not show any activity (Kamada et al., 2010). Based on this work, Kamada et al. Designed larger ligands with a calix[6]arene group with different positively charged end groups. One of the ligands, containing imidazole groups, increased the thermal stability of the R337H and enhanced the transcriptional activity of p53 R337H in cells (Kamada et al., 2010).</p><p>A different approach to discovering molecules that stabilize p53 tetramers was used in our laboratory (Gabizon et al., 2008, 2012). We used the natural protein-protein interactions of the p53 tetramerization domain or nearby regions for developing peptides that bind p53 in or near its tetramerization domain and may thus stabilize the tetramer. The interaction between p53Tet and the HIV-1 Tat protein (Longo et al., 1995) was characterized using peptide mapping. Two peptides from HIV-1 Tat bound p53Tet, and the interaction of a peptide derived from the arginine rich motif of Tat with p53Tet was characterized. The Tat-derived peptide bound all oligomeric forms of p53Tet without preference and thus was not a potential candidate for modulating p53Tet oligomerization (Gabizon et al., 2008). In a following work, we used a combinatorial approach and designed a peptide array derived from proteins known to bind the C terminal domain (CTD) of p53 (p53CTD, residues 293–393). Screening the array with recombinant p53CTD resulted in the identification of 10 peptides that bound p53CTD. Several of these peptides increased the thermal stability of the p53CTD and bound specifically to p53 tetramers in AUC experiments (Gabizon et al., 2012) (Figure 2C).</p><p>Other cancer related proteins may also be potential targets for molecules that modulate their oligomerization. Gray et al. developed a fluorescent monoclonal antibody assay for determining the extent of oligomerization of the oncoprotein AGR2 (Gray et al., 2013). They found that a peptide derived from the disordered N-terminal domain (NTD) of AGR2 stabilizes AGR2 oligomers, and that oligomerization of AGR2 enhances its binding to its chief partner protein reptin. This highlights the therapeutic potential of compounds that modulate AGR2 oligomerization.</p><p>Mdm2 and mdmX are negative regulators of p53 that mediate polyubiquitination and degradation of p53. Graves et al. developed compounds that bind mdm2 and mdmX and induce dimerization of the proteins. The p53-binding interface is buried in the dimerization interface, thus inhibiting the binding of p53 by mdm2 and mdmX (Figure 3A). The compounds activate the p53 transcriptional pathway in cells and induce apoptosis of cancer cells (Graves et al., 2012).</p><!><p>(A) Crystal structure of mdmX with the inhibitor RO-2443. A dimer of RO-2443 (sticks, yellow, and cyan) induces dimerization of mdmX (each monomer in different shade of gray). The binding sites for p53 (red and magenta) are blocked in the dimer. Structure from pdb 3U15 (Graves et al., 2012); (B) Crystal structure of pyruvate kinase M2 (PKM2) with the activator DASA-58. The four monomers are in red, green, blue and yellow and are semitransparent. DASA-58 is in cyan. The second activator TEPP-46 binds at the same site. Structure from pdb 3ME3 (Anastasiou et al., 2012).</p><!><p>Another example of anti-cancer compounds that act by promoting protein oligomerization is given by the work of Anastasiou et al. (2012). In cancer cells, inhibition of the enzyme pyruvate kinase M2 (PKM2) by phosphotyrosine-containing proteins increases the availability of glycolytic metabolites for the support of cell proliferation. The authors characterized two small molecule activators of PKM2. The compounds activate PKM2 by stabilizing its tetrameric state, prevent inhibition of PKM2 by phosphotyrosine-containing protein, alter the metabolism of cancer cells and inhibit cancer cell proliferation. Structural studies of the compounds with PKM2 revealed that the compounds bind at the interface between two monomers and stabilize the tetrameric structure (Figure 3B).</p><p>Another potential target may be BAK, a mitochondrial membrane protein involved in apoptosis. Following cell stress, p53 binds to BAK and induces BAK oligomerization (Leu et al., 2004; Pietsch et al., 2008), a critical stage in mitochondrial apoptosis. Thus molecules that bind and induce oligomerization of BAK may have potential as anti-cancer drug leads.</p><!><p>Oligomeric proteins play vital roles in the replication cycles of viruses. Many viruses encode proteases that catalyze the processing of viral polyproteins or the maturational processing of precapsids. These proteases are frequently oligomeric, as in HIV (Lange-Savage et al., 1997) and hepatitis C virus (Li et al., 2010). The function of viral capsid proteins is derived from their oligomerization properties, and the high order structures of numerous capsids have been characterized, as in the case of canine parvovirus (Wu and Rossmann, 1993), Hepatitis B virus (HBV) (Katen et al., 2013), and HIV (Zhao et al., 2013). Other prominent viral proteins that are active as oligomers include integrases (Cherepanov et al., 2003) and reverse transcriptases (Smerdon et al., 1994). Targeting the oligomerization of these proteins is emerging as a promising therapeutic strategy.</p><p>Considerable work has been performed on directly inhibiting the oligomerization of viral proteins using molecules that bind to their oligomerization interfaces. Peptides derived from the oligomerization interfaces can inhibit the oligomerization and activity of such proteins, as was shown for the HIV-1 reverse transcriptase (Divita et al., 1994; Morris et al., 1999; Depollier et al., 2005) and integrase (Sourgen et al., 1996; Maroun et al., 2001; Zhao et al., 2003). Non peptidic inhibitors of protein oligomerization have also been characterized (Rodríguez-Barrios et al., 2001; Bonache et al., 2005; Koh et al., 2007; Vidu et al., 2010; Tintori et al., 2012). This subject has been reviewed elsewhere and is outside the scope of the current review (Camarasa et al., 2006).</p><p>Inhibition of viral proteins can also be achieved by ligands that do not bind the oligomerization interface, but rather modulate the oligomerization of the protein allosterically. One of the main targets for this approach is the HIV-1 integrase protein (IN), which catalyzes the integration of the viral cDNA into the host genome (Delelis et al., 2008), a crucial step in the HIV-1 replication cycle (Sherman and Greene, 2002). Integration proceeds by two steps, each performed by a specific IN oligomer: (i) 3′ end processing, in which IN removes a GT dinucleotide from the 3′ termini of the long terminal repeats (LTRs) in the viral DNA. This step is performed by IN dimers (Deprez et al., 2001; Guiot et al., 2006) bound to each LTR in the cytoplasm; (ii) strand transfer, in which the viral DNA is integrated into the host DNA following nuclear import of the stable synaptic complex (Engelman et al., 1991). This step is performed by tetrameric IN (Li and Craigie, 2005; Li et al., 2006) in the nucleus with the assistance of cellular proteins, especially LEDGF/p75, which promotes IN tetramerization and tethers it to the chromosomes (Cherepanov et al., 2003; Maertens et al., 2003; Emiliani et al., 2005). The protein-protein interactions of IN are emerging as promising therapeutic targets (Maes et al., 2012), in particular the IN-LEDGF/p75 interaction (Christ and Debyser, 2013).</p><p>The importance of the integration step within the HIV replication cycle and the lack of mammalian homologs for IN both make IN an attractive therapeutic target. However, only two IN inhibitors (Raltegravir and Elvitegravir) are currently used in the clinic as anti-HIV drugs (Serrao et al., 2009; Messiaen et al., 2013). The rapid development of resistant strains (Mouscadet et al., 2010) emphasize the need to develop new drugs that function by different mechanisms, and efforts are made by numerous laboratories to discover novel IN inhibitors. In most cases, the modulation of IN oligomerization was not the intended result and candidate compounds were screened for inhibition of IN catalytic activity or inhibition of IN-LEDGF/p75 binding. However, many of the most promising IN inhibitors discovered recently were later shown to act primarily by shifting the oligomerization equilibrium of IN.</p><p>Kessl et al. studied IN inhibitors derived from chicoric acid, which inhibit the strand transfer activity of IN at micromolar concentrations (Kessl et al., 2009). They developed a method to measure the rate of subunit exchange between IN oligomers. The method uses His6-labeled IN and non-labeled IN, which are allowed to equilibrate separately and then mixed. The exchange of subunits is monitored by pulling down the His6-IN with Ni+2 beads and measuring the amount of non-labeled IN pulled down by SDS PAGE. One of the compounds significantly reduced the extent of subunit exchange, indicating that the oligomers were stabilized and thus dissociated more slowly. Molecular docking and mutational analysis indicated that the compound binds at a cleft located at the interface between two monomers of the IN catalytic core domain (CCD), while forming contacts with both monomers, thus stimulating oligomerization (Figure 4A). The authors postulate that the compound induces the formation of conformationally rigid IN tetramers that are unable to bind DNA in the necessary orientation for catalysis.</p><!><p>Inhibiting of HIV-1 IN by modulating its oligomerization equilibrium. (A) Structure of the IN inhibitor described by Kessl et al. (2009) and Possible binding sites. The two monomers of the IN CCD are in two shades of gray. The possible binding sites for the inhibitor are in orange and magenta. Structure from pdb 1EXQ (Chen et al., 2000). (B) Structure of the complex between the IN CCD (dimer, two shades of gray) and the IN binding domain of LEDGF/p75 (green). The LEDGF sequences that mediate IN binding are residues 361–370 (red) and 402–411 (magenta). The sequences of peptide inhibitors derived from LEDGF (Hayouka et al., 2007), Rev (Hayouka et al., 2008) and from combinatorial screening (Maes et al., 2009) are given. Structure from pdb 2B4J (Cherepanov et al., 2005); (C) Structures of IN inhibitors described in the review (from left to right): c(MZ 4-1), a cyclic derivative of LEDGF 361–370 (Hayouka et al., 2010b). Amino acid residues are shown in red; Inhibitors CX05168, CX05045, and CX014442 Described by Christ et al. (2010, 2012); inhibitor BI 224436 (Fader et al., in press).</p><!><p>In our laboratory we discovered peptides that bind IN and shift its oligomerization equilibrium toward the tetramer, thus inhibiting the 3′-end processing catalytic activity and preventing viral replication. We termed these peptides "Shiftides." The first anti-IN shiftides were derived from the IN-binding site in the cellular protein LEDGF/p75 (Hayouka et al., 2007) (Figure 4B). These peptides bound IN, shifted its oligomerization equilibrium toward the tetramer and inhibited IN activity in vitro and in HIV-1 infected cells (Hayouka et al., 2010a). We then synthesized a series of cyclic peptides derived from the IN binding sites in LEDGF/75. One of the peptides inhibited IN as much as the linear peptide while being more stable in cells (Figure 4C) (Hayouka et al., 2010b). The mechanism of binding of the cyclic peptides to IN depended on their ring size—while peptides with smaller rings bound preferably to IN dimers and stabilized them, larger rings promoted binding and stabilization of IN tetramers (Hayouka et al., 2010b). A similar mechanism of inhibition was discovered for peptides derived from the viral protein Rev (Figure 4B) (Hayouka et al., 2008), which binds and inhibits IN (Rosenbluh et al., 2007). IN-inhibitory shiftides were also developed by a combinatorial approach: a 20 residue peptide termed IN1, which was identified by a yeast two hybrid assay, bound to tetrameric IN, shifted the oligomerization equilibrium of IN toward the tetramer and inhibited IN activity (Figure 4B) (Armon-Omer et al., 2008; Maes et al., 2009). In comparison, peptides comprising residues 1–10 or residues 11–20 of IN1 bound to IN dimers, indicating that IN1 may stabilize tetrameric IN by bridging two IN dimers (Maes et al., 2009).</p><p>A major class of novel IN inhibitors are termed Allosteric IN Inhibitors (ALLINI's) (Tsantrizos et al., 2007; Christ et al., 2012; Tsiang et al., 2012; Engelman et al., 2013; Fader et al., in press). Christ et al. used a combination of computational and experimental methods to discover IN-inhibiting compounds. The authors initially aimed to find inhibitors for the IN-LEDGF/p75 interaction and used virtual screening to design a library of compounds that may bind to the LEDGF/p75 binding sites in IN and inhibit the IN-LEDGF/p75 interaction. Several compounds with micromolar activities were discovered (Christ et al., 2010). Further optimization yielded the compound CX14442, which inhibited the IN-LEDGF/p75 interaction at submicromolar concentrations (Figure 4C) (Christ et al., 2012). The authors discovered that beyond inhibiting the IN-LEDGF/p75 interaction, the compounds enhance oligomerization of IN and directly inhibit IN activity. Kessl et al. further studied the mechanism of action of these compounds as well as a similar IN inhibitor discovered by high throughput screening for 3′ processing activity (Figure 4C) (Tsantrizos et al., 2007). The ALLINI's bound at the dimerization interface between two CCD monomers, promoted IN multimerization and increased the thermal stability of IN. Studies of an IN mutant which is resistant to inhibition by ALLINI's showed that inhibition of IN activity is achieved mainly by the promotion of IN multimerization and not by inhibition of the IN-LEDGF/p75 interaction (Feng et al., 2013). Furthermore, Jurado et al. found that ALLINI's promote IN multimerization in virions, and that virions produced in the presence of ALLINI's are not infectious and have no reverse transcriptase or integrase activity (Jurado et al., 2013). One of the compounds, BI 224436, has an EC50 of 11–27 nM and has recently entered clinical trials (Fader et al., in press). These studies highlight the versatility and potency of compounds that inhibit IN by modulating its oligomerization equilibrium.</p><p>Viral capsid proteins are also a promising target for compounds that modulate oligomerization. Viral capsids are large, highly symmetric oligomers comprised of one or several types of monomers, and may contain hundreds of subunits (Mateu, 2013). The function of capsid proteins requires that they form highly stable capsids that can withstand high internal pressures (Molineux and Panja, 2013) and yet be able to dissociate upon cell entry and release the viral DNA and proteins into the host cell. Incorrect formation of the viral capsid can be detrimental to the replication of viruses and thus mutations that destabilize or alter the structure of the capsid highly reduce the infectivity of viruses, as in the case of HIV (Noviello et al., 2011) and HBV (Tan et al., 2013). Therefore, many inhibitors targeted against capsid proteins are currently being developed.</p><p>The HIV-1 capsid protein (CA) is one of the main targets for inhibition. CA is composed of two helical domains, the NTD and the CTD, which are connected via a flexible linker. The basic structural unit of the capsid is a hexameric ring formed by NTD-NTD interactions, and reinforced by intermolecular NTD-CTD interactions (Figure 5A). Dimeric CTD-CTD interactions link between the hexameric rings, and the full capsid has a fullerene-like structure (Pornillos et al., 2009). Correct assembly of the capsid may be very sensitive to small changes in the strength and geometry of the inter-subunit interactions. Unlike in the case of HIV-1 integrase, most HIV inhibitors that target CA were intended for inhibition of capsid formation and were discovered by rational structure-based approaches or screenings that used in vitro capsid assembly assays.</p><!><p>Molecules that target capsid assembly. (A) Structure of HIV-1 capsid protein hexamer. Different monomers are in Green, red, blue, magenta, cyan, and orange. The NTD and CTD of every monomer are in different shades of the same color. Structure from pdb 1VUU (Zhao et al., 2013). (B) Structure of HIV-1 capsid protein NTD in the absence (left) and presence (right) of CAP-1. Phe32 is shown in yellow, CAP-1 is shown in magenta. Structures from pdb files 1VUU (Zhao et al., 2013) and 2JPR (Kelly et al., 2007). The formulas of CAP-1 and a representative Benzimidazole (Goudreau et al., 2013) and Benzodiazepine (Tremblay et al., 2012) are given at the bottom. (C) Formula of CA-binding compound described by Lemke et al. (2013). (D) Formula of the CA-binding compounds PF74 (described by Shi et al., 2011) and BI-2 (described by Lamorte et al., 2013) (E) Structure of HIV-1 capsid protein CTD in the absence (left) and presence (right) of the capsid assembly inhibitor (CAI). CAI is in orange. The helix from which the peptide CAC1 was derived (Bocanegra et al., 2011) is in cyan. The sequence of CAI and the structure of the optimized peptide NYAD-1 (Zhang et al., 2008) are given below. Structures from pdb files 1AUM (Gamble et al., 1997) and 2BUO (Ternois et al., 2005).</p><!><p>Tang et al. used virtual screening to search for compounds that bind and inhibit CA. One of the compounds discovered, CAP-1, inhibits HIV-1 infectivity by specifically interfering with the formation of viral capsids (Tang et al., 2003). CAP-1 inhibited capsid formation in vitro and caused the formation of viral capsids with abnormal morphologies. 1H-15N HSQC measurements showed that CAP-1 binds the NTD at the apex of the helical bundle, and inhibits the NTD-CTD interaction necessary for hexamer stabilization. Kelly et al. further characterized the binding of CAP1 to CA using X-ray crystallography and NMR, and observed that CAP1 binds into a hydrophobic pocket formed by the displacement of Phe32 (Kelly et al., 2007) (Figure 5B).</p><p>Other compounds that inhibit CA by a mechanism similar to CAP1 have been studied. Lemke et al. developed a high throughput assay for testing the effect of compounds on capsid formation and used it to screen a large library for CA-inhibitors. This led to the discovery of new inhibitors derived from Benzimidazole (BM) and benzodiazepine (BD) (Lemke et al., 2012) (Figure 5B). While both families bound CA at the same site (which is identical to the CAP-binding site), their modes of binding were different and had distinct effects on HIV-1 replication: while BD compounds inhibited capsid formation and release, BM compounds inhibited the formation of the mature, conical capsid after release from the cell. In later studies, the compounds were modified and improved - Tremblay et al. systematically modified the functional groups in a BM scaffold, yielding CA-inhibitors with IC50 values below 0.1 μM (Tremblay et al., 2012). Goudreau et al. developed BD-based compounds with IC50 values below 1 μM (Goudreau et al., 2013).</p><p>Recently, Goudreau et al. characterized a novel family of BM-based CA inhibitors that bind the NTD in a distinct site from CAP-1(Goudreau et al., 2013). One of the compounds induces the formation of an NTD-dimer with a non-native geometry (Lemke et al., 2013) (Figure 5C).</p><p>Several compounds that target the CA NTD were discovered by screening of compound libraries for inhibition of HIV replication in cells, with further studies revealing that CA was the target. Blair et al. and Shi et al. discovered the compound PF-74, which inhibits HIV replication in cells by binding to the NTD of CA and destabilizing the capsid structure (Blair et al., 2010; Shi et al., 2011). This causes premature uncoating of the virion during the replication cycle. On the other hand, Lamorte et al. discovered pyrrolopyrazolone based HIV inhibitors that bind CA at a similar binding site to PF74 and increase the stability of the viral capsid, thus interfering with the nuclear import of the stable synaptic complex (Lamorte et al., 2013) (Figure 5D). These results indicate that the viral replication cycle can be very sensitive to small changes in capsid stability in either direction, which further highlights the therapeutic potential of CA-binding molecules.</p><p>The CTD of CA is also a target for inhibition. Sticht et al. used phage display screening to identify peptides that bind CA and inhibit capsid formation. The authors discovered a peptide, termed capsid assembly inhibitor (CAI), that binds to the CTD and inhibits the formation of the mature capsid (Sticht et al., 2005). X-ray crystallography of the peptide-bound CTD (Ternois et al., 2005) showed that the peptide adopts a helical conformation and binds into a hydrophobic groove in the CTD, forming a five helix bundle. The binding of the peptide significantly alters the geometry of the dimerization interface (Figure 5E). Therefore, although the peptide does not directly bind the dimerization interface and does not destabilize the CTD dimer, it alters the geometry of the dimer interface, thus preventing the formation of the mature capsid. In a later study, the sequence of CAI was optimized using hydrocarbon stapling in order to stabilize the helical secondary structure of the peptide. This resulted in the peptide NYAD-1, which had improved cell penetration and activity in vivo (Zhang et al., 2008).</p><p>Bocanegra et al. used a rational design approach to inhibit capsid assembly. The authors synthesized a peptide, termed CAC1, derived from a helix in the dimerization interface of the CTD (Figure 5E), and also made several modifications in the peptide to increase its solubility and its affinity to the CTD. The peptides bound directly to the dimerization interface of CTD, inhibited capsid assembly and also had antiviral activity in cells (Bocanegra et al., 2011).</p><p>Capsid-binding molecules have been studied for other viruses. Plevka et al. used X-ray crystallography to study the effect of the capsid binding inhibitor WIN 51711 on the replication of enterovirus 71, which is associated with foot and mouth disease (Plevka et al., 2013). They discovered that WIN 51711 binds a pocket in one of the subunits involved in capsid formation and increases the stability of the capsid, thus restricting the dynamics of the capsid necessary for genome release and lowering the infectivity of the virus (Figure 6A).</p><!><p>(A) Structure of the enterovirus 71 capsid proteins with WIN 51711. Three subunits of the capsid are in blue, green and red. WIN 51711 is in yellow. Based on Plevka et al. (2013); (B) Formulas of HBV capsid binding inhibitors (Stray et al., 2005; Stray and Zlotnick, 2006).</p><!><p>The HBV is also being targeted by capsid-binding inhibitors. Heteroaryldihydropyrimidines (HAPs) are a class of compounds that inhibit HBV replication in tissue culture by interfering with capsid formation (Deres et al., 2003). The HBV capsid is made of 120 dimers of capsid protein (Cp) arranged in icosahedral symmetry (Bottcher et al., 1997). Stray et al. studied the effects of the two HAP compounds, HAP-1 and BAY 41-4109, on capsid formation in vitro (Stray et al., 2005; Stray and Zlotnick, 2006) (Figure 6B). While at substoichiometric concentrations the compounds increase the rate of capsid formation, at high concentrations they misdirect capsid formation and induce the formation of aberrant structures. The mechanism by which the compounds inhibit viral replication is still unclear—it is possible that the compounds do not antagonize HBV by directly inhibiting capsid formation, but rather disrupt the coordination of capsid assembly with other stages of the replication, or inhibit structural transitions necessary for the formation of mature, infectious capsids.</p><!><p>The equilibrium between monomers and fibrils plays a major role in many diseases, particularly neurodegenerative diseases such as Alzheimer's disease (Gilbert, 2013), Parkinson's disease and prion diseases (Salvatella, 2013). The inhibition of amyloid fibrillation is a major goal in drug development. Amyloid β fibrillation can be inhibited by short peptides derived directly from the sequences that mediate fibrillation (Tjernberg et al., 1996). Others methods involve rational design of inhibitors based on the structure of the interface between monomers (Sato et al., 2006; Sievers et al., 2011). This subject has been further reviewed elsewhere (Soto and Estrada, 2005; Belluti et al., 2013).</p><p>The monomer-fibril equilibrium can also be shifted toward the fibril, as has been shown in our laboratory for non-muscle myosin type II (NMII). NMII undergoes dynamic filament assembly and participates in cellular processes such as cell migration and cytokinesis. Ronen and Rosenberg et al. investigated the role of the non-helical C-terminal tailpiece in filament assembly (Ronen et al., 2010). They found that the tailpiece is intrinsically disordered and is divided into two oppositely charged regions. The positively charged part of the tailpiece interacts with an assembly incompetent fragment of NMII and induces its filamentation, while the negatively charged part affects the morphology of the filaments.</p><!><p>Oligomerization plays a crucial role in the activity of many disease-related proteins and is therefore a promising target for therapeutic intervention. The molecules presented here were discovered by methods ranging from combinatorial screening methods such as phage display to rational, structure-based design. In most cases a combination of several methods was used. Many of the compounds were discovered in experiments that did not aim for modulating protein oligomerization, and their mechanism of action was elucidated later. However, as the knowledge of the oligomerization states and structure of the target proteins accumulates and methods that measure changes in the stability or structure of the oligomers become widespread, a growing number of active compounds are designed and screened for allosteric modulation of protein oligomerization. It is therefore not surprising that most of the compounds being developed target proteins with well characterized structures and oligomerization equilibria.</p><!><p>In our laboratory we focus on peptides as tools for modulating protein oligomerization and we term them "shiftides." These peptides bind specifically to a particular oligomeric state of the target protein and stabilize it. By doing so, these peptides shift the oligomerizaton equilibrium toward this specific oligomeric state. This way it is possible either to activate a protein by stabilizing an active oligomer or inhibit a protein by stabilizing an inactive oligomer. In this review, we have demonstrated the development of shiftides that target several proteins such as HIV-1 integrase (Hayouka et al., 2007; Maes et al., 2009), p53 (Gabizon et al., 2012) and non-muscle myosin IIC (Ronen et al., 2010). Shiftides can be discovered using rational design based on the sequences of proteins known to bind the target protein, using combinatorial approaches, or combining the two methods. The versatile chemistry of peptides enables the facile optimization of shiftide activity as well as improving pharmacological properties such as cell permeability (Wang et al., 2014) and proteolytic stability (Moretto et al., 2006; Dong et al., 2012) (Figure 7). Therefore, the shiftide approach can be applied effectively to a wide range of disease related proteins with dynamic oligomerizaion equilibria.</p><!><p>Shiftides—peptides that modulate the oligomerization equilibrium of proteins. Shiftides can be developed using combinatorial screening or rational design, and can be modified easily to improve their activity and pharmacological properties. An example of a protein in a dimer—tetramer equilibrium is given, but the principle can be applied to any oligomerization equilibrium.</p><!><p>The number of disease-related oligomeric proteins that are being studied and characterized continues to grow (Lawrence et al., 2008; Ferré et al., 2010). A prominent example is G-protein coupled receptors (GPCRs). Many GPCRs are now known to form homo-and hetero-oligomers (Ferré et al., 2010), and the oligomerization of GPCRs can be critical to their activity (Jones et al., 1998) or significantly alter it (Azdad et al., 2008). The binding of agonists and antagonists can also induce conformational changes in the dimerization interface of GPCRs (Guo et al., 2005). Therefore, allosteric modulation of oligomerization may be a powerful approach for the development of drugs that target GPCRs in the future. As the number of potential targets grows, we expect that modulation of protein oligomerization will become a central therapeutic strategy for a variety of diseases.</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

Two novel nonlinear optical carbonates in the deep-ultraviolet region: KBeCO3F and RbAlCO3F2

With the rapid developments of the all-solid-state deep-ultraviolet (deep-UV) lasers, the good nonlinear optical (NLO) crystal applied in this spectral region is currently lacking. Here, we design two novel NLO carbonates KBeCO 3 F and RbAlCO 3 F 2 from the first-principles theory implemented in the molecular engineering expert system especially for NLO crystals. Both structurally stable crystals possess very large energy band gaps and optical anisotropy, so they would become the very promising deep-UV NLO crystals alternative to KBBF. Recent experimental results on MNCO 3 F (M 5 K, Rb, Cs; N 5 Ca, Sr, Ba) not only confirm our calculations, but also suggest that the synthesis of the KBeCO 3 F and RbAlCO 3 F 2 crystals is feasible. Deep-ultraviolet (deep-UV, l , 200 nm) nonlinear optical (NLO) crystals, as a key component of all-solidstate deep-UV lasers, have played important roles in many advanced scientific and technical areas 1-3 . Numerous attempts have been performed on the explorations of new deep-UV NLO materials with good performances [4][5][6][7][8] , and the dominant research field has been focused on the borates 9-11 . In particular, Potassium beryllium fluoroborate (KBe 2 BO 3 F 2 , KBBF) exhibits excellent NLO performance in the deep-UV region; till now KBBF is the sole NLO crystal that can practically generate the deep-UV lasers by direct SHG process 1 . However, the applications of KBBF are heavily hindered by the layering tendency in the single crystal growth processes 12 . Therefore, it is urgently demanded the emergence of new types deep-UV NLO materials.To be a good deep-UV NLO material, the following four criterions are commonly considered 13,14 : (i) a wide UV transparency range down to the deep-UV region, corresponding to the large energy band gap and high damage threshold; (ii) a relatively large efficient second harmonic generation (SHG) coefficient (d ij $ 0.39 pm/V, d 36 of KH 2 PO 4 ); (iii) a relative large birefringence (Dn $ 0.08) to achieve the phase matching condition in the deep-UV region; and (iv) good chemical stability and mechanical properties. According to the anionic group theory 15,16 , the planar [BO 3 ] 32 microscopic anionic groups have the dominant contribution to the macroscopic optical anisotropies in crystal, as in the KBBF case. Analogously, [CO 3 ] 22 and [NO 3 ] 2 anionic groups are expected to be the good NLO micro-structural units as well since they have the similar planar triangle structure with the p-conjugated molecular orbitals which can produce the large birefringence and second-order susceptibility. The nitrates, however, are not considered as the NLO candidates for their hydrolysis. Thus, it is greatly desirable to explore the deep-UV NLO materials in carbonates.The rapid developments of scientific computational resources make it possible to predict the new advanced materials directly from the first-principles theory [17][18][19] , although search for the best candidate for a special property is still a major task 20 . In this work, aiming at the additional NLO candidates in the deep-UV region, we design two structurally stable carbonates KBeCO 3 F and RbAlCO 3 F 2 . The energy band gaps and the linear and nonlinear optical properties of both crystals are determined by ab initio calculations with high precision, which clearly demonstrate their promising applications as good optoelectronic functional materials in the deep-UV region. ResultsIn order to search for the suitable deep-UV carbonates, we employed the anionic group theory combined with first-principles calculations which are implemented in the molecular engineering expert system especially for NLO crystals originated from our group 9,15 . After numerous efforts, we have theoretically discovered two carbonates, KBeCO 3 F and RbAlCO 3 F 2 , according to the following materials design considerations: (i) the candidates

two_novel_nonlinear_optical_carbonates_in_the_deep-ultraviolet_region:_kbeco3f_and_rbalco3f2

<!>Discussion<!>Methods

<p>are the fluoride carbonates since they can possess the larger energy band gaps compared with other carbonates 21 ; (ii) the A-site cations are the metal cations without unclosed d or f electrons because the dd or f-f electronic transitions have negative influences to the energy band gap. Especially the alkaline cations and the lightweight metal cations in the Group IIA and IIIA in the Periodic Table are taking into account; (iii) all [CO 3 ] 22 anionic groups are parallel (flat-lying) with respect to the overall structural layering. This structural feature has strong optical anisotropic responses to the incident light and exhibit large optical birefringence; and (iv) the orientation of all [CO 3 ] 22 groups is parallel to each other. This arrangement is favorable to the additive superposition of the microscopic second susceptibilities in the anionic groups, so produce large macroscopic SHG effects in crystal.</p><p>The structures of KBeCO 3 F and RbAlCO 3 F 2 are plotted in Fig. 1 (their crystallographic data see Table S1 in the Supplementary Information). In KBeCO 3 F, the [BeO 2 F 2 ] are connected with the [CO 3 ] triangles by sharing their edges, forming the zerodimensional [BeCO 3 F 2 ] microscopic structures isolated by the K 1 cations. In RbAlCO 3 F 2 , the [AlO 3 F 2 ] trigonal bipyramids and [CO 3 ] triangles are alternately arranged in a trigonal pattern and connected via common O corners, generating a two-dimensional infinite [AlCO 3 F 2 ] layer parallel to the a-b plane. The Rb 1 cations are located between these layers to balance charge and also hold the layers together through the coordination with O and F anions. Both structures clearly exhibit the parallel flat-lying arrangement of the [CO 3 ] 22 groups in crystals, so they are expected to have large SHG coefficients and birefringences. More importantly, the structural stabilities of both crystals are carefully verified by the first-principles methods 18,19 . Figure 2 displays the phonon spectra and the total energy as a function of volume per atom in KBeCO 3 F and RbAlCO 3 F 2 . In the phonon spectra (Figure 2(a)) none of the imaginary phonon modes is observed, and in the curves of the total energy as a function of volume per atom (Figure 2(b)) there has a single minimum in the wide range of volume modification. Both evidences clearly demonstrate that the titled crystals are kinetically stable.</p><p>The calculated linear and nonlinear optical properties for KBeCO 3 F and RbAlCO 3 F 2 are listed in Table 1, and the experimental results for KBBF are also shown as a comparison. It is clear that both KBeCO 3 F and RbAlCO 3 F 2 possess very large energy band gaps, 7.61 eV (,164 nm) and 8.21eV (,152 nm), respectively. Meanwhile, the birefringences and SHG coefficients in both KBeCO 3 F and RbAlCO 3 F 2 crystals are larger than those in KBBF. The large birefringences (Dn 5 0.1297 in KBeCO 3 F and Dn 5 0.0998 in RbAlCO 3 F 2 ) guarantee the achievement of the SHG phase-matching condition in both crystals down to their UV absorption edge, thus their shortest phase-matching wavelengths very approach (or even exceed) the corresponding wavelength in KBBF (,161 nm) 12 . It should be emphasized that RbAlCO 3 F 2 is optically uniaxial, which would be beneficial to its practical applications. The excellent optical properties in KBeCO 3 F and RbAlCO 3 F 2 demonstrate that both crystals are comparable to KBBF for the deep-UV SHG capabilities. Upon being obtained, these crystals would have wide applications as the optoelectronic functional materials in the deep-UV spectral region.</p><!><p>To verify the reliability of our ab initio calculations on the optical properties in carbonates, we also applied our calculated methods on a family of recently synthesized NLO fluoride carbonates, MNCO 3 F (M 5 K, Rb, Cs; N 5 Ca, Sr, Ba) 22 . The preliminary optical measures revealed that these compounds possess wide energy band gap (UV absorption edge , 200 nm) and qualitatively exhibit good capabilities for the UV harmonic generation 22 . However, due to the very small size of the samples the obtained experimental data are not enough to determine their application prospects in the deep-UV region. Our calculated energy band gaps and SHG coefficients match the available experimental results very well (see Table S2 and S3 in the Supplementary Information), which strongly prove the validity and high precision of the first-principles studies on the UV NLO carbonate crystals. Since the MNCO 3 F series possess strong NLO effect and large birefringence, we predict that they are suitable to be good NLO crystals and excellent birefringent materials in the UV region. Nevertheless, these carbonates cannot be applied in the deep-UV region due to their relatively small energy band gaps (E g , 6.3 eV).</p><p>The detailed electronic structure analysis show that the small energy gaps in MNCO 3 F are mainly due to the O 2p non-bonding states exclusively occupied at the valence band maximum, which have negligible small overlap with other electronic states and directly determine the energy band gap 23 (see Figure S1 in the Supplementary Information). The energy spanning of these non-bonding states is as large as about 2.5 eV, and their complete elimination can significantly increase the energy band gap of crystals to more than 8.3 eV (UV absorption edge , 150 nm). Compared MNCO 3 F with our proposed crystals, the substitution of Ca, Sr or Ba atoms with the lightweight metal cations in the Group IIA or IIIA, Be or Al, can effectively remove the non-bonding orbitals in carbonates, analogous to the borate cases such as beryllium in KBBF 24,25 and aluminum in BaAlBO 3 F (BABF) [26][27][28] . Indeed, Figure 3 clearly shows that the energy spanning of the non-bonding regions in KBeCO 3 F and RbAlCO 3 F 2 reduces about 1.1 eV and 2.3 eV, respectively, compared with that in MNCO 3 F, exemplified by KSrCO 3 F. Therefore, the UV absorption edge of both KBeCO 3 F and RbAlCO 3 F 2 are significantly blue-shifted. In addition, the similarity of the micro-structural features in the MNCO 3 F series and our proposed crystals suggests that it is feasible to synthesize KBeCO 3 F and RbAlCO 3 F 2 in experiments.</p><p>We surveyed throughout Inorganic Crystal Structure Database (ICSD) 29 , and found that only 14 fluoride carbonates whose energy band gaps may be large enough to transmit the UV radiation have been discovered and synthesized in experiments (see Table S4 in the Supplementary Information), in which seven are noncentrosymmetric and satisfy the SHG requirement. The very few report on UV NLO carbonates actually implies that the researches on this type of optoelectronic functional materials have been long-termly neglected and the broad developing space is available. We believe that our work would have great implications on the search and design of new NLO crystals in the deep-UV spectral region with the great help of advanced synthesis and single-crystal growth approaches.</p><!><p>First-principles computational methods and geometry optimizations. The first-principles calculations are performed by the plane-wave pseudopotential method 30 implemented in the CASTEP package 31 based on the density functional theory (DFT) 32 . The ion-electron interactions are modeled by the optimized normal-conserving pseudopotentials for all elements [33][34][35] . The kinetic energy cutoffs of 900 eV and Monkhorst-Pack k-point meshes 36 with the spanning of less than 0.03/A ˚3 in the Brillouin zone are chosen. The supercell volume and the atomic positions for KBeCO 3 F and RbAlCO 3 F 2 are fully optimized using the quasi-Newton method 37 . The convergence thresholds between optimization cycles for energy change, maximum force, maximum stress, and maximum displacement are set as 10 25 eV/atom, 0.03 eV/ A ˚, 0.05 GPa, and 0.001 A ˚, respectively. The above computational set ups are sufficiently accurate for present purposes 38 .</p><p>Calculated methods for structural stability. To verify the structural stability, phonon spectra calculations on KBeCO 3 F and RbAlCO 3 F 2 are performed. The positive eigenvalues for all phonon modes is the most important evidence for the structural stability in crystal. The structural stability can also be demonstrated by the total energy as a function of volume per atom. If there has only a single minimum in the wide range of volume modification, the studied geometrical structure would be stable. The studied methods on structural stability were also adopted by Yao 18 and Sheng 19 .</p><p>Calculated methods for optical properties. When determining the linear optical properties the imaginary part of the dielectric function e 2 is calculated 39 , and then its real part is determined by Kramers-Kronig transform, from which the refractive indices (and the birefringence) are obtained. Moreover, the second-order susceptibility x (2) , i.e., the SHG coefficient d ij , is calculated by the following formula developed by our group 40 :</p><p>where x abc (VE), x abc (VH) and x abc (two bands) denote the contributions from virtual-electron processes, virtual-hole processes and two-band processes, respectively. The formulae for calculating x abc (VE), x abc (VH) and x abc (two bands) are given as follows 40 :</p><p>x abc (twobands)~e</p><p>Here, a, b and c are Cartesian components, v and v9 denote valence bands, and c and c9 denote conduction bands. P(abc) denotes full permutation and explicitly shows the Kleinman symmetry of the SHG coefficients. The band energy difference and momentum matrix elements are denotes as hv ij and p a ij , respectively, and they are all implicitly k dependent.</p><p>It is well acknowledged that the DFT calculations with the exchange-correlation (XC) functional of generalized gradient approximation (GGA) 41 always underestimate the energy band gap of crystals. For calculating the optical coefficients, a scissors operator 42,43 is usually introduced to shift up all the conduction bands to agree with the measured band gap, which is not ''purely'' ab initio. Recently, we have examined several XC functionals beyond GGA on the determination of energy band gaps and optical properties in UV NLO crystals 44 . It was found that the hybrid functionals such PBE0 45 , P3LYP 46 and sX-LDA 47 can predict the UV absorption edge very well, but the calculated electronic band structures and optical properties are not well reproduced compared to the scissors-corrected GGA method 44 . Indeed our calculations show that in the MNCO 3 F crystals the GGA energy band gaps are in large discrepancy to the experimental results (absolute error excesses 2.4 eV), while the PBE0 values are in good consistency with the measurements (see Table S2 in the Supplementary Information). Therefore, in this work the difference between PBE0 and GGA energy band gaps is set as the scissors operator, which is then used to determine the optical properties by GGA. The further tests have also revealed that the x (2) values is insensitive to the modification (e.g. , 60.5 eV) of the scissors operator because of the very large band gap (.6 eV) in the studied carbonates. Therefore, this computational procedure ensures that our first-principles studies are self-consistent without adjusting any parameter from experiments.</p>

Scientific Reports - Nature

A Soft Mechanical Phenotype of SH-SY5Y Neuroblastoma and Primary Human Neurons Is Resilient to Oligomeric A\xce\xb2(1\xe2\x80\x9342) Injury

Aggregated amyloid beta (A\xce\xb2) is widely reported to cause neuronal dystrophy and toxicity through multiple pathways: oxidative stress, disrupting calcium homeostasis, and cytoskeletal dysregulation. The neuro-cytoskeleton is a dynamic structure that reorganizes to maintain cell homeostasis in response to varying soluble and physical cues presented from the extracellular matrix (ECM). Due this relationship between cell health and the ECM, we hypothesize that amyloid toxicity may be directly influenced by physical changes to the ECM (stiffness and dimensionality) through mechanosensitive pathways, and while previous studies demonstrated that A\xce\xb2 can distort focal adhesion signaling with pathological consequences, these studies do not address the physical contribution from a physiologically relevant matrix. To test our hypothesis that physical cues can adjust A\xce\xb2 toxicity, SH-SY5Y human neuroblastoma and primary human cortical neurons were plated on soft and stiff, 2D polyacrylamide matrices or suspended in 3D collagen gels. Each cell culture was exposed to escalating concentrations of oligomeric or fibrillated A\xce\xb2(1\xe2\x80\x9342) with MTS viability and lactate dehydrogenase toxicity assessed. Actin restructuring was further monitored in live cells by atomic force microscopy nanoindentation, and our results demonstrate that increasing either matrix stiffness or exposure to oligomeric A\xce\xb2 promotes F-actin polymerization and cell stiffening, while mature A\xce\xb2 fibrils yielded no apparent cell stiffening and minor toxicity. Moreover, the rounded, softer mechanical phenotype displayed by cells plated onto a compliant matrix also demonstrated a resilience to oligomeric A\xce\xb2 as noted by a significant recovery of viability when compared to same-dosed cells plated on traditional tissue culture plastic. This recovery was reproduced pharmacologically through inhibiting actin polymerization with cytochalasin D prior to A\xce\xb2 exposure. These studies indicate that the cell\xe2\x80\x93ECM interface can modify amyloid toxicity in neurons and the matrix-mediated pathways that promote this protection may offer unique targets in amyloid pathologies like Alzheimer\xe2\x80\x99s disease.

a_soft_mechanical_phenotype_of_sh-sy5y_neuroblastoma_and_primary_human_neurons_is_resilient_to_oligo

INTRODUCTION<!>Characterization of Oligomeric and Fibrillated A\xce\xb2.<!>Baseline Studies: Oligomeric A\xce\xb2 Reduces Cell Viability More than Fibrillated A\xce\xb2 Does.<!>Toxicity to Oligomeric A\xce\xb2(1\xe2\x80\x9342) is Matrix-Dependent.<!>Actin Polymerization and Cell Stiffness Adjust in Response to Both Matrix Stiffness and Aggregated A\xce\xb2.<!>Potential Physical Mechanisms That Regulate Neuronal Survival.<!>Preparation of Oligomeric and Fibrillated A\xce\xb2.<!>Structural Characterization of A\xce\xb2(1\xe2\x80\x9342) Preparations via AFM.<!>2D Polyacrylamide Matrix Preparation.<!>Cell Culture.<!>Confocal Microscopy.<!>3D Collagen Matrix Preparation and Rheological Characterization.<!>MTS Proliferation Assay.<!>Lactate Dehydrogenase Release Assay.<!>Nanoindentation to Measure Cell Stiffness.<!>Western Blot Analysis to Measure Total Actin Expression.

<p>In healthy brains, the low nanomolar concentration of amyloid beta (Aβ) has been reported to modulate neurogenesis and help maintain synaptic plasticity.1 However, excess Aβ production and the resulting aggregates are associated with numerous pathways of neuronal dysfunction and toxicity.2 In an early study to demonstrate Aβ toxicity, Yankner et al. reported that toxicity was observed in hippocampal neurons at concentrations of 20 μM after 24 h.3 However, the same dose of Aβ did not have any observed toxicity to glial cells. Recently, a shift in the amyloid cascade hypothesis has moved the primary focus from total Aβ concentration to the small oligomeric, intermediate species.4 Associated with this toxicity, Aβ has been shown to dysregulate the actin cytoskeleton, with subsequent synaptic and dendritic malfunction, which generally precedes neuronal toxicity, loss and amyloid-induced dementia.5</p><p>Abnormalities in cytoskeletal organization are a common feature of many neurodegenerative disorders, including Alzheimer's disease,6,7 Parkinson's disease,8 and Huntington's disease.9 Neurons are polarized cells, with two distinct functional compartments, and the neuro-cytoskeleton is crucial in maintaining shape and intercompartment polarity differences for efficient signal transmission.10 Thus, an appropriately organized cytoskeleton is necessary during nervous system development, maintenance, and recovery after injury. Cytoskeletal defects including alterations in microtubule stability, stalled actin dynamics, and prolonged/fragmented F-actin bundles have all been observed in several neurodegenerative conditions.11 For example, Alzheimer's disease is associated with both neurofibrillary tangles and microtubule defects caused by the aberrant hyperphosphorylation of the tau protein.12</p><p>A key regulator of cytoskeletal remodeling and neuronal function is the extracellular matrix (ECM). Physical characteristics of the ECM—stiffness, composition, and topology—are tissue specific, and resident cells have evolved to be optimally functional within a specific ECM. An ECM that is excessively stiff may prompt actin polymerization and stress fiber formation, an increase in RhoA-mediated intracellular tension, and/or an increase in the release of enzymes that degrade the ECM. Alternatively, a matrix that is too soft may trigger actin depolymerization, a decrease in cell spreading, and increased collagen secretion to potentially rebuild and fortify the surrounding ECM.13 Neuropotentiation has been further shown to increase overall brain tissue stiffness, while age, injury, and disease state have been associated with a decrease overall brain tissue stiffness. Brain tissue is soft with reported Young's moduli ranging from 0.1 to 16 kPa and is significantly more compliant than other tissues in the body such as endothelial, muscle, or bone. Moreover, the brain ECM is not constant, and elastography imaging has shown that brain tissue softness decreases with age, and for patients suffering from Alzheimer's disease (AD) this tissue softening is enhanced.14 It is unclear, however, how a changing brain ECM may contribute to the pathology of AD.</p><p>Herein we explore the ECM's potential role in aggregated Aβ toxicity in both SH-SY5Y neuroblastoma and primary neurons. ECM stiffness helps regulate cell shape, adhesion, differentiation, and survival, among numerous other cellular behaviors. The ECM also affects the endocytosis of both nanoparticles and peptides, including our previous work that displays an increased uptake for both the 40- and 42-residue isoforms of Aβ on soft polyacrylamide (PA) matrices relative to Aβ uptake into cells plated on tissue culture plastic.15 Notably, this increase in uptake on soft matrices is enhanced for the 42-residue isoform, suggesting that cell–matrix interactions could influence uptake in a mechanism-specific manner.</p><p>In addition to mechanosensitive endocytosis, it has previously been reported that cells can be more susceptible to certain toxins, such as acrylamide or acetaminophen, when they are plated in soft matrices compared to plastic substrates.16,17 This susceptibility in soft alginate substrates was argued to be a result of increased Rho GTPase activity, a key regulator of the actin cytoskeleton. For AD, it has been shown that the involvement of Rho GTPases led to an increased level of RhoA and decreased neurite length in dystrophic neurons located at the periphery Aβ plaque deposits. Song et al. determined that the overproduction of Aβ in a septal neuronal cell line induced the formation of actin stress fibers.18 In addition, Aβ(1—42) has been shown to stimulate actin polymerization in these neurons through the activation of the Rho GTPases Rac1 and Cdc42.19</p><p>The cytoskeleton provides mechanical stiffness to the cell, and thus changing cell stiffness (whether due to soluble or physical factors) may serve as a marker of cytoskeletal status. Ungureanu et al. observed a decrease in cell stiffness when the cells were dosed with 10 μM Aβ for 2 h in artificially aged primary neurons.20 This decrease in cell stiffness was not observed for cells that were cultured for 1 week, and they concluded that membrane elasticity changes are induced with a "sublethal" concentration of preaggregated Aβ. Additionally, aged neurons were found to be more susceptible to amyloid toxicity than nonaged or "young" neurons.</p><p>To better understand the influence of cytoskeletal dysregulation and AD, Furukawa and Mattson observed that pretreatment with cytochalasin D rescued primary hippocampal neurons from fibrillated Aβ toxicity in a dose dependent manner.21 Cytochalasin D, a pharmacological agent that depolymerizes actin, was further reported to decrease intracellular calcium levels, even after exposure to Aβ. Cells dosed with Aβ displayed an increase in intracellular calcium, while this increase was not observed in cells that were treated with cytochalasin D prior to exposure. In addition, actin remodeling was considered the key contributor to this neuroprotection, because destabilization of microtubules did not prevent Aβ induced toxicity. The destabilization of actin limited the amount of intracellular calcium in the cells and decreased excitotoxicity.21–23</p><p>Although these studies show varying consequences of actin and microtubule remodeling, they collectively confirm that there is a definitive, cytoskeletal-mediated, cellular response to Aβ(1—42). As cytoskeletal dysregulation has been implicated in multiple neurodegenerative disorders, this work aims to determine whether cytoskeletal remodeling, prompted by physical modifications to the ECM, can adjust Aβ toxicity. If successful, we anticipate that detailing the specific, mechano-biochemical pathways that orchestrate this actin response would offer novel therapeutic targets in the treatment of Alzheimer's disease.</p><!><p>To confirm the preparation protocol generated two different aggregated states of Aβ each preparation was analyzed using atomic force microscopy (AFM). Previous studies have analyzed Aβ aggregation using SDS-PAGE electrophoresis; however, these techniques may lead to apparent aggregates that arise from technical artifacts due to matrix effects.24 Additionally, AFM imaging permits identification of the aggregate conformation, either globular or fibrillated. As shown in Figure 1, when our aggregate preparations were analyzed using AFM, distinct fibrils were observed for samples prepared in 10 mM HCl at 37 °C. Alternatively, when the sample was prepared in cell culture media at 4 °C, globular aggregates, which were considerably smaller in size, appeared. This is consistent with the results described in the protocol reported by Stine et al.24 Note that in Figure 1 some differences were observed for Aβ preparations using material sourced to different vendors.</p><!><p>Prior to exploring the potential role of the ECM in amyloid-mediated toxicity, it is important to first establish baseline viability data using traditional, tissue culture plastic. Monomeric Aβ is considered benign, particularly at low micromolar concentrations, and indeed recent studies have reported that low nanomolar concentrations of monomeric Aβ exert a positive influence on synaptic plasticity, neurogenesis, and memory formation.25 An early study reported by Lorenzo and Yankner tested the toxicity of monomeric and fibrillated Aβ, and they observed no toxicity when primary rat hippocampal neurons were treated with 20 μM of monomeric Aβ(1–40) or Aβ(1–42) over 72 h.26 However, significant toxicity was observed for neurons dosed with fibrillated Aβ. Pretreatment with Congo red, a commonly used inhibitor of amyloidogenesis, restored viability to control levels, and thus, it was concluded that aggregation was requisite for Aβ toxicity.</p><p>For our study, SH-SY5Y cells and primary neurons were treated with 1, 5, and 10 μM concentrations of oligomeric or fibrillated Aβ. Note that, in Figure 2, each of these aggregated species reduced cell viability in a concentration-dependent fashion. However, exposure to the oligomeric species is consistently more toxic than that to fibrillated Aβ for both cell types, and at all concentrations. A modest reduction in cell viability was observed for fibrillated Aβ only at the highest in-well concentration, and for both cell lines cell viability remained at near 60–70% (relative to untreated control cells) after a 48 h exposure to fibrillated Aβ. Our results are similar to those reported by Cecchi et al. whereby undifferentiated SH-SY5Y cells exposed to 1 and 10 μM doses of small (ca. 50 nm diameter), globular aggregates of Aβ(1—42) reduced MTT viability to near 55% after 24 h.27 In the same study, retinoic acid differentiated SH-SY5Y cells displayed a significant resistance to Aβ injury, suggesting a strong, phenotype-specific, toxicity response to Aβ. The relationship between chemical differentiation of SH-SY5Y and sensitivity to amyloid toxicity was further explored Krishtal et al.28 At a 10 μM concentration of HFIP-disaggregated Aβ (1–42), undifferentiated SH-SY5Y cells displayed no reduction in WST-1 viability relative to untreated control cells. Upon differentiation, however, WST-1 viability reduced significantly and to varying extents, depending upon the specific differentiation protocol. One clear exception is SH-SY5Y cells differentiated through retinoic acid followed by tetradecanoylphorbol acetate to a dopaminergic phenotype that was completely resistant to Aβ at either 10 or 20 μM concentrations.</p><p>The mechanism of amyloid-mediated toxicity is also sensitive to the specific aggregation state of Aβ. Gharibyan et al. studied amyloidogenic protein aggregation and its associated toxicity using lysozyme as a model protein.29 Early stage aggregates of lysozyme formed oligomers that induced apoptosis in SH-SY5Y cells, while late-stage aggregates formed mature fibrils that induced a necrotic cell death primarily through disruption of the plasma membrane. Consistent with our studies, Gharibyan et al. found the oligomeric species to be significantly more toxic to the neuroblastoma than either the monomeric or fibrillated species; however, the specific mechanism(s) that distinguish the pronounced toxicity of oligomeric lysozyme remain unclear. Recent treatment strategies for Alzheimer's disease have been developed to exclusively target oligomeric Aβ, rather than the monomers or mature fibrils. For example, Limbocker et al. treated SH-SY5Y cells with trodusquemine, a compound that facilitates nucleation and Aβ fibrillation, which led to a significant reduction in toxicity, despite a significant increase in overall plaque load.30</p><!><p>Neuronal differentiation is not restricted to chemical cues only, but the physical influence of the ECM—topography, dimensionality, stiffness, microstructure, and pore size—all play a critical role in directing the terminal fate of progenitor cells.31–35 Analogous to the chemical differentiation studies described above, cells exposed to different extracellular matrices may express a distinct, "mechanical" phenotype. Cancer cells are well-known to display a mechanical phenotype often markedly softer than that for corresponding normal tissue cells. As reported by Lin et al., this mechanical transformation to a soft phenotype occurs regardless of cancer type—breast, bladder, cervix, or pancreas—and is driven largely by suppressing caveolin-1 expression to reduce stiffness sensing and traction force at cell-ECM interface.36 The authors suggest that this loss of stiffness sensing may confer this soft phenotype with an antiapoptotic character that permits the anchorage-independent growth which is characteristic of many cancer types.</p><p>The potential for matrix-dependent apoptosis has significant bearing for Alzheimer's disease. The brain ECM changes in composition and stiffness with age, and the alternative adhesion signaling in resident neurons results in cytoskeletal restructuring. To test the potential matrix-dependency of oligomeric Aβ(1–42) toxicity to SH-SY5Y cells and primary neurons, each cell type was plated on (1) soft and stiff 2D PA matrices, (2) 3D collagen gels, or (3) tissue culture plastic and the cell viability was assessed with an MTS proliferation assay. Note that, for our experiments, the "soft" PA matrix had a Young's modulus of near 3 kPa and the "stiff" PA matrix had a Young's modulus of near 24 kPa. The shear modulus of the 3D collagen gel (2 mg/mL) was determined to be 54.7 Pa, and we estimate the Young's modulus to be 0.156 kPa assuming a Poisson's ratio of 0.5. This approximation is reasonable because a hydrated gel is mostly composed of water, an incompressible fluid. Specific details about the rheological measurements are provided in the Supporting Information; however, the composition and stiffness data for all matrices used in this study are collected for reference in Table 1. Our studies using the 2D matrices were performed using collagen type IV as an ECM protein because it is similar to the nonfibrous proteins found in vivo in the brain ECM. This linker protein is the primary point of integrin-mediated adhesion and helps communicate matrix stiffness to the cell; thus, changing this linker protein would have definite implications in modifying cell behavior. In our work, additional studies were completed using fibronectin and laminin. However, for each of these ECM proteins, sufficient cell attachment was not achieved on the softest matrices (data not shown), which led to poor cell counts and challenged follow-on analyses by flow cytometry.</p><p>Each sample was exposed to 10 μM oligomeric Aβ(1–42) for 48 h. Note that SH-SY5Y cell and primary neuron proliferation are matrix dependent, and thus, MTS results could not be directly compared across the different matrices because the raw cell counts were different for different matrices. To address this challenge, each sample is compared to an untreated control of cells plated on the same corresponding matrix. As displayed in Figure 3, for the 2D PA matrices, both cell types displayed a significant recovery in MTS viability when compared to same-dosed cells on traditional tissue culture plastic. When comparing the viability of SH-SY5Y cells on soft and stiff PA matrices, no significant difference was observed at α = 0.05; however, at the same significance level, the human primary neurons on the soft PA matrix displayed a higher viability than that observed for cells on the stiff PA matrix. Further shown in Figure 3, both SH-SY5Y cells and primary neurons suspended in the soft, 3D collagen gel yielded the highest overall increase in viability relative to both PA matrices and tissue culture plastic. Overall, both ECM stiffness and dimensionality may play a significant role in regulating amyloid toxicity.</p><!><p>Characterizing the mechanical properties of individual cells is sensitive to cell type, sample preparation (e.g., live/fixed), and method of measurement. Note that cell stiffness is generally reported as a Young's modulus (E) and an increasing E indicates lower compressibility and higher cell stiffening. Consistent with our previous studies, live human primary neurons were plated on the soft and stiff PA matrices, as well as a glass coverslip, prior to indentation measurements. Because AFM indentation requires physical contact between the AFM probe and the sample, cell stiffness could not be obtained for cells suspended in the 3D collagen gel. Previous indentation measurements of cell stiffness were found to be 0.23–0.52 kPa for cortical neurons,37,38 2–11 kPa for human fibroblasts, and 1.2–2.8 kPa for human lung cells.39 As shown in Figure 3, our AFM nanoindentation results are consistent with these previous studies and confirm that primary human neurons are sensitive to changes in matrix stiffness. Neurons adapting to the softest PA matrix displayed a Young's modulus of 0.18 kPa. On the stiffest PA matrix, a modest, yet significant, increase to 0.22 kPa was observed. On glass, a material like traditional tissue culture plastic with GPa stiffness, neurons display a stiff phenotype with a spread morphology and a Young's modulus of 0.42 kPa, which is nearly double that observed on the softer PA matrices. Our stiffness results for live neurons plated on a glass substrate are comparable to previous results reported in the literature.39</p><p>Further shown in Figure 3, the primary neurons display a clear, matrix-dependent morphology similar to that observed previously for SH-SY5Y cells. On the softest matrix, neurons display a rounded morphology, yet when plated onto the stiffest matrix a spread morphology is adopted. An ImageJ analysis of neuronal spread area is provided in the Supporting Information. Cells adherent to a stiff matrix generally adopt a spread morphology with an increased apical cell area. A stiff matrix can better resist cell-generated traction forces leading to an increase in actomyosin activation, stress fiber formation, and intracellular tension. Qualitative confocal fluorescence images of primary neurons plated on the stiff PA matrices displayed a clear increase in phallodin-stained F-actin and stress fiber formation, while the rounded neurons on the softest matrix stained weakly and thus qualitatively display the lowest F-actin fraction. In no instance for the 2D matrices did we see a traditional neuronal morphology with long neurites extending from a well-defined soma. Only in the 3D collagen matrix was this traditional, neuronal morphology observed.</p><p>To better quantify the extent of matrix-dependent actin reorganization, total actin expression was measured for cells that were plated on the each of the 2D PA matrices. As displayed in Figure 3, both neuroblastoma and primary neurons significantly increased actin production when plated on traditional tissue culture plastic. However, when comparing total actin levels from cells plated on either soft and stiff PA matrices, no significant differences were observed. Attempts to isolate G- and F-actin pools separately were unsuccessful, and therefore, total actin is reported (normalized to GADPH levels). However, we suspect that if F-actin were isolated, its fraction would decrease as matrix stiffness decreased. The large increase in actin production from cells plated on plastic is directly in line with the large increase in cell stiffness observed. Overall, for neurons and SH-SY5Y neuroblastoma, the morphological changes in cell spreading clearly correspond to reorganization of the actin cytoskeleton that subsequently manifests as a change in cell stiffness. This inter-relationship between cell area, F-actin, and cell stiffness had previously been reported by Solon et al. for NIH 3T3 fibroblasts.40 Using a similar series of PA matrices, fibroblasts preferentially adopted a spread morphology (higher cell area) on a stiff matrix and a rounded morphology (lower cell area) on softer matrices. Plotting fibroblast cell stiffness against cell area displayed a positive correlation and furthermore displayed an increase in F-actin polymerization.</p><p>Apart from physical stimuli, chemical cues presented to the cell membrane from the extracellular space can also modify cell stiffness.41,42 To investigate the potential of aggregated Aβ inducing changes in cell stiffness, each cell type was exposed to 1 μM of oligomeric or fibrillated Aβ for 24 h. All stiffness experiments were performed at 37 °C with live cells cultured onto a glass coverslip. We selected a subtoxic concentration of aggregated Aβ because our interest was whether cell-Aβ interactions may equally potentiate changes in cell stiffness. If toxic doses were used, then cytoskeletal restructuring that results from either apoptotic blebbing or necrotic membrane damage would confound any stiffness changes introduced by cell-Aβ interactions alone. As seen in Figure 4, for both cell types, an increase in cell stiffness was observed when exposed to the oligomeric Aβ; however, exposure to the same concentration of fibrillated Aβ produced no change in stiffness for either cell type. Referring to Figure 2, this concentration and exposure would result in a minor (if any) reduction in viability, and therefore, the aggregation dependent increase in cell stiffness is not likely a result of actin restructuring due toxicity, but rather it indicates that oligomeric and fibrillated Aβ interact with neuronal cells in distinctly separate fashions.</p><p>Multiple studies have explored cell—Aβ interactions using cell stiffness to better detail the mechanism of amyloid-mediated toxicity.43–45 Lulevich et al. reported that untreated N2a neuroblastoma have a Young's modulus of 0.9 MPa and after exposure to oligomeric Aβ(1—42) the Young's modulus increased to 1.85 MPa.44 Exposure to monomeric or fibrillated Aβ yielded either no change (monomeric) or a modest increase (fibrillated) in cell stiffness. An even more pronounced change in stiffness was observed for HT22 hippocampal neurons, whereby exposure to oligomeric Aβ prompted a greater than 3-fold increase in cell stiffness from 1.73 MPa (untreated) to 5.5 MPa (treated). In a separate study, the stiffness of primary hippocampal neurons, exposed to oligomeric Aβ, was further studied Ungureanu et al.; however, depending upon the "age" of the neuron, exposure to Aβ(1–42) resulted in a significant decrease in cell stiffness.20 For cortical neurons harvested from 17 day old Sprague—Dawley rats, the viscoelastic response was monitored as a function of oligomeric Aβ(1—42) exposure using microrheology. Using a 1 μM Aβ(1—42) dose, the scaled shear modulus (Go) of cortical neurons was found to increase from control values (~42 Pa) to ~55 kPa after a 6 h exposure. With continued exposure up to 24 h, this increased rigidity displayed no further change. Similarly, the viscosity increased from 3.0 to ~4.7 Pa·s after a 3 h exposure and remained constant thereafter for 24 h.</p><!><p>The relationship between cytoskeletal structure and survival suggests that amyloid toxicity is inherently sensitive to changes in the ECM. ECM stiffness has previously been shown to affect the survival of neurons and cancer cells exposed to various toxins, nanoparticles, and chemotherapeutics. As reported by Senthebane et al., the ECM proteins of the tumor microenvironment are an essential component to understanding chemoresistance in esophageal squamous cancer cells (ESCCs).46 A PCR analysis of the decellularized ESCC biopsies revealed increased mRNA expression for multiple ECM proteins: collagen, laminin, and fibronectin. Moreover, cells seeded into these fortified matrices displayed an enhanced resistance to common chemotherapies relative to cells grown on plastic or in collagen- and fibronectin-lacking matrices. Ramamoorthi and co-workers furthermore reported that glioblastoma exposed to acrylamide, acetaminophen, quinidine, or cadmium chloride displayed a clear matrix-dependent cell viability.17 Glioblastoma cells were plated on traditional tissue culture plastic or suspended in two different 3D alginate matrices designated as soft and stiff, and cells grown in the soft alginate matrix were found to be more susceptible to each of the four toxins. In a separate study, Wang et al. found when two breast cancer cell lines were plated on 2D substrates with stiffness values ranging from 1—25 kPa, the cells grown on the softest matrix were resistant to paclitaxel treatment.47 Combined these studies confirm cell survival is sensitive to the ECM; however, the specific mechanism that promotes this sensitivity is unknown.</p><p>An initial consideration in mechanosensitive toxicity is endocytosis. In general, when the physical properties of a cell change, it would be anticipated that endocytic pathways could be affected. This uptake sensitivity to the ECM would, however, be pathway-specific, because different endocytic pathways require a specific membrane deformation, actin organization, and/or protein recruitment. From our previous studies on monomeric Aβ endocytosis, undifferentiated SH-SY5Y cells plated on progressively softer PA matrices displayed a clear reduction in spread area and cell stiffness.15 This soft, rounded phenotype of SH-SY5Y cells also displayed an increased uptake of monomeric Aβ. This finding was consistent with a recently published theoretical model that describes the energetics and kinetics of membrane deformation and wrapping.48 Further support is found from the experimental study that detailed the substrate dependent uptake of carboxylated polystyrene particles into bovine endothelial cells.49</p><p>While the expectation may be that softening promotes endocytosis, monomeric and oligomeric Aβ display completely different physicochemical properties and thus are not expected to be internalized through the same endocytic path. As shown in the Supporting Information, our initial uptake studies using oligomeric Aβ (prepared as described earlier but doped with fluorescently labeled monomeric Aβ) indicate that SH-SY5Y cells on soft and stiff matrices internalize less oligomeric Aβ than SH-SY5Y cells grown on traditional cell culture plastic. Therefore, the improved viability of SH-SY5Y cells exposed to oligomeric Aβ may be partly attributed to the reduced uptake. The internalization of oligomeric Aβ into primary neurons, however, displayed no significant sensitivity to the underlying substrate. Therefore, the improved viability observed on the softest matrix does not stem from differences in endocytic uptake. Further attempts to assess uptake of oligomeric Aβ in cells suspended in the 3D matrix were unsuccessful due to a limited cell number. Note that, despite a micrometer-sized pore structure for 2 mg/mL collagen gels, the lower diffusivity of oligomeric Aβ into the 3D matrix may also contribute to the improved viability for both SH-SY5Y cells and primary neurons.50</p><p>Oligomeric Aβ(1–42) exerts toxicity through multiple mechanisms, some of which—oxidative stress and calcium dysregulation—are known to be sensitive to the actin cytoskeleton and intracellular tension.51–56 It has previously been shown that primary hippocampal neurons, when pretreated with cytochalasin D for 1 h, display a dose-dependent increase in viability when challenged with 50 μM Aβ(25–35) after 48 h.21 Cytochalasins are a class of fungal metabolites that act to destabilize the cytoskeleton by binding to the plus-end of actin microfilaments and preventing further actin fiber elongation and polymerization. However, when the microtubules were destabilized using a 100 nM colchicine pretreatment, there was no recovery in viability, suggesting that specifically actin depolymerization contributed to the improved viability. To investigate the role of actin turnover in our studies, both SH-SY5Y cells and primary neurons were pretreated with 100 nM cytochalasin D, prior to a 10 μM challenge of oligomeric Aβ(1–42) for 48 h. As shown in Figure 5, cells pretreated with cytochalasin D demonstrated improved MTS viability and reduced LDH toxicity when compared to cells treated with oligomeric Aβ alone. Thus, both the pharmacological and matrix-mediated depolymerization of the actin cytoskeleton can provide protection against oligomeric Aβ.</p><p>Actin dynamics and cell—ECM adhesion signaling are interrelated.57–59 Adhesion to the ECM is predominantly integrin-mediated, and depending upon the specific integrin activation different scaffolding and adaptor proteins are recruited to the intracellular focal adhesion. This focal adhesion furthermore connects to the actin cytoskeleton and through its composition and structure may signal secondary messengers that modulate actin turnover. An early, upstream regulator of cytoskeletal status is paxillin; its recruitment and phosphorylation state helps regulate cell adhesion and cytoskeletal structure. Both SH-SY5Y cells and primary neurons display a significant decrease in paxillin expression (normalized to the GAPDH band) when plated on the soft PA matrices, indicating that paxillin expression is highly matrix-dependent. as shown in Figure 6. The decrease of total paxillin is consistent with decreased cell adhesion and actin polymerization, and this correlates with improved viability; however, further studies are underway to identify whether paxillin (or its phosphorylation status) directly participates in matrix-mediated, neuronal survival. As shown in the Supporting Information, our initial attempts to agonize integrin activation using MnCl2 resulted in reduced viability for SH-SY5Y cells exposed to 5 μM oligomeric Aβ, but no significant effect was observed for the primary neurons. Given that paxillin and its phosphorylation state are associated with amyloid-mediated cytoskeletal dysfunction in Alzheimer's disease,60–62 understanding how adhesion dynamics may protect against neuronal loss in an aging and/or pathological ECM would offer multiple untapped targets in the search for effective, new therapies in neurodegenerative disorders.</p><p>In conclusion, Aβ aggregates are a key component of Alzheimer's disease, and their mechanism of toxicity is a primary focus when developing next-generation therapeutic options. Consistent with previous studies in alternative cell lines, oligomeric Aβ is significantly more toxic to both SH-SY5Y neuroblastoma and primary cortical neurons than fibrillated Aβ. We find that this toxicity, however, is matrix-dependent and cytoskeletal remodeling may directly influence amyloid toxicity. Cells cultured on soft 2D or 3D matrices and exposed to aggregated Aβ display a significant recovery in MTS viability compared to same-dosed cells plated on conventional tissue culture plastic. For cells adherent to stiff matrices, there is an increase in actin polymerization and stress fiber formation, which translates to an increase in cell stiffness. For both cell types, treatment with a subtoxic concentration of oligomeric Aβ further prompted significant cell stiffening. Both SY-SY5Y cells and primary neurons grown on the softer matrices displayed a rounded, soft mechanical phenotype that weakly stained for F-actin. We suggest this is due to reduced cell adhesion to the soft ECM, and this is consistent with the reduced paxillin levels observed. We anticipate that cells grown in a matrix that more closely mimics the mechanical environment of the brain are buffered against cell stiffening introduced by oligomeric Aβ. The precise mechanism that confers this matrix-dependent protection is unclear, and future studies are underway, but our results suggest that actin depolymerization can protect against amyloid toxicity. Our focus on the actin cytoskeleton is furthermore supported by pharmacological depolymerization of actin using cytochalasin D, which improves the viability for both cell types when challenged with oligomeric Aβ. Overall, recognizing that physical cues presented from the ECM may regulate cell survival introduces a variety of new sites for potential therapeutic intervention in Alzheimer's disease.</p><!><p>Amyloid beta used for AFM characterization was purchased from California Peptide and American Peptide Company. Further toxicity and AFM cell stiffness analyses were done using only Aβ(1–42) purchased from California Peptide. Aβ(1–42) oligomers and fibrils were prepared by following the protocol reported by Stine et al.24 Any pre-existing aggregates were removed by dissolving the Aβ to 5 mM in hexafluoro-2-propanol (HFIP), aliquoted into microcentrifuge tubes, and the allowed to evaporate in a fume hood overnight. These samples were stored at −20 °C in a desiccant jar until further use. For sample preparation, Aβ(1—42) was dissolved to 5 mM in dimethyl sulfoxide (DMSO). For oligomeric samples, Aβ(1–42) was further diluted to 100 μM in serum-free,phenol-free cell culture medium and incubated at 4 °C for 24–48 h. For fibrillated samples, Aβ(1–42) was diluted to 100 μM in 10 mM HCl and incubated at 37 °C in a heat block for 24 h prior to AFM characterization or dosing.</p><!><p>Height images were recorded using a Molecular Force Probe 3D (Asylum Research) atomic force microscope. Fibrillated and oligomeric Aβ samples were prepared by drop casting Aβ solutions (100 μM) onto a freshly cleaved atomically flat mica (V–I grade, SPI Supplies) substrates. AFM height images of deposited Aβ samples were collected in air using the AC mode with silicon nitride AFM probes, a nominal spring constant of 0.3 N/m, and a typical tip radius of curvature of 8 nm (Mikromasch, CSC37). In this study, images were collected with a spatial resolution of 5 nm and 1 Hz scan rate. Raw AFM data and a detailed experimental analysis are provided in the Supporting Information.</p><!><p>Polyacrylamide (PA) matrices were prepared exactly as described in our previous publication.15 A detailed protocol can be found in the Supporting Information. Note that, consistent with our previous studies, we selected a "soft" and "stiff" PA matrix to contrast with results provided from cells plated on traditional, tissue culture plastic. Our "soft" and "stiff" PA matrices have Young's moduli of approximately 3 and 24 kPa, respectively. Rat tail collagen type IV was selected as an ECM linker protein to promote cell adhesion. This protein was covalently bound to the surface of each PA matrix using standard sulfo-SANPAH chemistry. Note that all traditional cell culture plates (plastic) also had a thin layer of collagen type IV deposited onto the surface to assist adhesion.</p><!><p>SH-SY5Y human neuroblastoma cells were obtained from American Type Culture Collection. Cells were maintained at 37 °C and 5% CO2 and were grown in Opti-MEM supplemented with 10% fetal bovine serum, 1 mM sodium pyruvate, MEM nonessential amino acids (Gibco), and 1% penicillin/streptomycin. All uptake experiments used Opti-MEM without phenol in the media. SH-SY5Y cells would not be used when the passage number exceeded 30. Primary human cortical neurons were obtained from Neuromics (Edina, MN) and grown in neuronal culture medium, also purchased from Neuromics. These neurons were cultured strictly according to the culture conditions advised by Neuromics. Primary neurons were passaged 2–3 times before beginning anew with frozen neurons.</p><!><p>After primary neurons were plated onto the 2D polyacrylamide matrices, the cells were fixed in ice cold, 4% paraformaldehyde. Cells were then permeabilized with 0.1% Triton-X, and stained with FITC-phalloidin (green). The cells were mounted on microscope slides with a drop of VectaShield mounting media containing DAPI for staining cell nuclei blue. The cells were imaged using a Zeiss 710 confocal microscope located in the Central Microscopy Research Facility at the University of Iowa.</p><!><p>Collagen matrices were prepared according to the method reported by Baker et al.63 Briefly, the gels were prepared by first combining 100 mM HEPES in DPBS with high concentration collagen I (Corning) in equal proportions. This solution was mixed with cell culture media and cells to get a final concentration of 2 mg/mL and 1 × 105 cells/mL. The mixture was plated into 96-well plates at 100 μL per well. The plate was incubated at 37 °C, 5% CO2 for 2 h while the collagen matrix spontaneously self-assembled. After the matrix had fully formed, 100 μL of complete medium was added above each gel and the cells were grown for 24 h before staining or treatment.</p><p>The viscoelastic properties of these collagen gels were determined using a RheoStress One rheometer with a 35 mm stainless steel, 4 ° cone. Collagen mixtures of different concentrations were prepared and allowed to polymerize on the rheometer for 2 h at 37 °C. Frequency and stress sweeps were first performed to determine the linear viscoelastic range of each matrix using an oscillatory stress of 250 Pa while varying the frequency from 0.1 to 10 Hz and measuring 8 points per decade. Then, using a frequency of 1 Hz, the oscillatory stress was varied between 0.04 and 10 Pa. Each frequency and stress sweep were performed at 37 °C and is the average of three independently prepared samples.</p><!><p>SH-SY5Y cells or primary neurons were plated in 12-well plates on 2D PA matrices or tissue culture plastic at 1.0 × 106 cells per well. Otherwise SH-SY5Y cells or primary neurons were plated in 2 mg/mL 3D collagen matrices or on tissue culture plastic in 96-well plates at 1.0 × 105 cells per well as described earlier. After 24 h of growth, cells were treated with either oligomeric or fibrillated Aβ at various concentrations for 48 h in serum-free and phenol-free media. After this exposure period, the media was aspirated and the cells were gently washed with DPBS. MTS reagent was added at 1 mL reagent plus 10 mL of Hank's balanced salt solution with 1 mg/mL glucose. The plate was incubated at 37 °C for an additional 4 h, and then the absorbance was measured at 490 nm using a plate reader. Viability for all samples was compared to that of untreated control cells that were plated on the same corresponding matrix.</p><!><p>SH-SY5Y cells or primary neurons were plated in 12-well plates on 2D PA matrices or tissue culture plastic at 1.0 × 106 cells per well. After 24 h of growth, cells were treated with either oligomeric or fibrillated Aβ at various concentrations for 48 h in serum-free, phenol-free media. Thirty minutes before analysis, 10 μL of 10× lysis buffer was added to positive control wells for each cell line and on each substrate. Lactate dehydrogenase (LDH) release was measured using a LDH release assay kit (Thermo Scientific), according to the manufacturer's instructions. After 50 μL of sample media was transferred to a 96-well plate, 50 μL of the working reagent was added to each well. The plate was stored in the dark at room temperature for 1 h. Then 50 μL of a stop solution was added to each well. The absorbance of the plate was read at 490 nm on a plate reader. Toxicity is reported as the percentage of LDH release relative to the completely lysed sample.</p><!><p>The AFM nanoindentation for cells grown on either the 2D PA matrices or glass was done exactly as described in our previous paper. A detailed experimental protocol is further provided in the Supporting Information.</p><!><p>To measure actin expression for each of the 2D PA substrates, PA gels were prepared as previously mentioned. SH-SY5Y cells or primary neurons were plated on each substrate at 1.5 × 106 cells per well on the softest substrate and at 1.0 × 106 cells per well on the stiff and plastic substrates and were grown for 3 days in complete culture medium. Cells were lysed and prepared for a Western blot. Next, 20 μg/well of each sample was loaded onto a 10% SDS-PAGE gel and run at 90 V. The samples were transferred electrophoretically onto a nitrocellulose membrane and then blocked in 5% BSA in TBST for 2 h. The membrane was then probed with an actin primary antibody (Cytoskeleton, Inc.) and a GAPDH housekeeping control (University of Iowa Hybridoma Bank) according to the vendors' instructions. The membrane was washed twice with TBST and then probed with a rabbit secondary antibody (Santa Cruz, 1:20,000) for 1 h. The membrane was washed twice with TBST and then analyzed using a chemiluminescent substrate according to the manufacturer's instructions (ThermoScientific). Each sample is presented relative to the amount of GAPDH expression.</p>

PubMed Author Manuscript

Conformational Populations of \xce\xb2-(1\xe2\x86\x924) O-Glycosidic Linkages Using Redundant NMR J-Couplings and Circular Statistics

Twelve disaccharides containing \xce\xb2-(1\xe2\x86\x924) linkages and displaying systematic structural variations in the vicinity of these linkages were selectively labeled with 13C to facilitate measurements of multiple NMR spin\xe2\x80\x93spin (scalar; J) coupling constants (JCH and JCC values) across their O-glycosidic linkages. Ensembles of spin-couplings (2JCOC,3JCOCH, 3JCOCC) sensitive to the two linkage torsion angles, phi (\xcf\x95) and psi (\xcf\x88), were analyzed by using parametrized equations obtained from density functional theory (DFT) calculations, Fredholm theory, and circular statistics to calculate experiment-based rotamer populations for \xcf\x95 and \xcf\x88 in each disaccharide. With the statistical program MA\xe2\x80\x99AT, torsion angles \xcf\x95 and \xcf\x88 were modeled as a single von Mises distribution, which yielded two parameters, the mean position and the circular standard deviation (CSD) for each angle. The NMR-derived rotamer populations were compared to those obtained from 1 \xce\xbcs aqueous molecular dynamics (MD) simulations and crystallographic database statistical analyses. Conformer populations obtained exclusively from the MA\xe2\x80\x99AT treatment of redundant J-couplings were in very good agreement with those obtained from the MD simulations, providing evidence that conformational populations can be determined by NMR for mobile molecular elements such as O-glycosidic linkages with minimal input from theory. The approach also provides an experimental means to validate the conformational preferences predicted from MD simulations. The conformational behaviors of \xcf\x95 in the 12 disaccharides were very similar, but those of \xcf\x88 varied significantly, allowing a classification of the 12 disaccharides based on preferred linkage conformation in solution.

conformational_populations_of_\xce\xb2-(1\xe2\x86\x924)_o-glycosidic_linkages_using_redundant_nmr_j-

I. INTRODUCTION<!>IIA. Synthesis of 13C-Labeled Disaccharides 2 and 4\xe2\x80\x9314<!>IIB. NMR Spectroscopy<!>IIIA. Geometric Optimization of Disaccharides 2 and 4\xe2\x80\x9314<!>IIIB. DFT Calculations of NMR Spin-Coupling Constants in Disaccharides 2 and 4\xe2\x80\x9314<!>IIIC. Parametrization of J-Coupling Equations for O-Glycosidic Linkage Conformational Analysis<!>IIID. Description of the Computational Algorithm Used in the MA\xe2\x80\x99AT Software, and Determinations of O-Glycosidic Linkage Conformational Models from NMR J-Coupling Ensembles<!>IIIE. Molecular Dynamics Simulations of Disaccharides 2 and 4\xe2\x80\x9314<!>IVA. Classification of \xce\xb2-(1\xe2\x86\x924) Linkage Conformations Based on Qualitative Comparisons of \xcf\x95- and \xcf\x88- Dependent J-Coupling Ensembles<!>IVB. Preferred Linkage Conformations in Disaccharides 2 and 4\xe2\x80\x9314 Determined from DFT-Derived Potential Energy Surfaces<!>IVC. Structural Dependencies of \xcf\x95- and \xcf\x88-Dependent J-Couplings in \xce\xb2-(1\xe2\x86\x924)-Linked Disaccharides 2 and 4\xe2\x80\x9314<!>IVC1. Three-bond 13C\xe2\x80\x9313C Spin-Couplings 3JC2\xe2\x80\xb2,C4, 3JC1\xe2\x80\xb2,C5, and 3JC1\xe2\x80\xb2,C3<!>IVC2. Two-Bond 13C\xe2\x80\x9313C Spin-Coupling 2JC1\xe2\x80\xb2,C4<!>IVC3. Three-Bond 13C\xe2\x80\x931H Spin-Couplings 3JC4,H1\xe2\x80\xb2 and 3JC1\xe2\x80\xb2,H4<!>IVC4. Effects of C\xe2\x80\x93O and C\xe2\x80\x93C Bond Conformation on Calculated Trans-O-Glycoside J-Couplings<!>IVD. Statistical Modeling of \xcf\x95 and \xcf\x88 in Disaccharides 2 and 4\xe2\x80\x9314 Using J-Coupling Ensembles<!>IVE. Crystallographic Analysis of \xcf\x95 and \xcf\x88 in \xce\xb2-(1\xe2\x86\x924)-Linked Disaccharides<!>IVF. Behavior of \xcf\x95 and \xcf\x88 in Disaccharides 2 and 4\xe2\x80\x9314 Determined from Aqueous Molecular Dynamics Simulations<!>V. CONCLUSIONS

<p>Molecular flexibility is a hallmark feature of saccharides in their monomeric (monosaccharide), oligomeric (oligosaccharide), and polymeric (polysaccharide) forms.1-6 In contrast to many biomolecules, saccharides are rich in electron lone pairs contributed by multiple hydroxyl groups appended to their carbon scaffolds. These lone pairs are important determinants of structure and thus reactivity. For example, the rotation of the C–O bonds of exocyclic hydroxyl groups results in substantial structural change in saccharides due to stereoelectronic interactions between lone-pair orbitals and the antibonding orbitals of proximal C–H and C–C bonds.7-10 These interactions are superimposed on those involving the endocyclic ring and anomeric oxygens (e.g., endo- and exo-anomeric eflects11-19). These behaviors render the solution structures, and thus intrinsic reactivities, of a free saccharide different from those of the same saccharide bound to a receptor (e.g., protein or enzyme), because hydroxyl conformations are likely to differ in both states. Establishing quantitative relationships between saccharide covalent structure and solution conformation and dynamics is central to achieving a deeper understanding of how these molecules perform their functions in vivo and to deciphering the glyco-code.20-22</p><p>The conformational properties of saccharides in solution are difficult to determine, especially those of the O-glycosidic linkages found in more complex structures.23-25 For saccharides of modest size (MW < 20 000 Da), NMR spin–spin coupling constants (scalar or J-couplings) are valuable experimental parameters to investigate conformational properties in the presence of motional averaging, given their high abundance and the fact that they average linearly.26-34 For example, 50 J-couplings involving 1H and 13C atoms are available in methyl β-d-glucopyranoside 1 that are sensitive to either ring or exocyclic hydroxymethyl conformation (Scheme S1, Supporting Information). Groups of J-couplings (ensembles) often report on the same conformational element (redundancy). For example, at least four J-couplings are sensitive to ϕ and six to ψ for O-glycosidic linkages composed of two C–O bonds (Scheme 1). A total of 17 J-couplings are available for O-glycosidic linkages composed of three bonds (Scheme 2). This redundancy is especially useful in conformational studies when the individual J-couplings in the ensemble exhibit unique dependencies on the same conformational element (e.g., a molecular torsion angle). In some cases, J-couplings depend on two conformational elements, which allows correlated conformation in solution to be determined.35</p><p> </p><p>Modern experimental studies of NMR J-couplings in saccharides are often accompanied by density functional theory (DFT) calculations to expand their utility.10,29,30,36 These calculations provide quantitative relationships between a specific J-coupling and one or more molecular parameters. For example, although accurate J-couplings across O-glycosidic linkages can be measured in 13C-labeled compounds, these parameters have limited value unless they can be related quantitatively to linkage geometry. DFT calculations provide these relationships. Furthermore, because the bonds comprising an O-glycosidic linkage are potentially mobile, analyses of J-coupling ensembles must accommodate the possible presence of multiple conformations in solution. Because it remains difficult to determine conformational populations based solely on experimental data, molecular dynamics (MD) simulations and other theoretical calculations are relied upon heavily to obtain this information directly or to assist in the interpretation of experimental data.37 This situation prevents independent experimental validations of the results of MD simulations.</p><p>This study was undertaken to improve the conformational assignments of O-glycosidic linkages in oligosaccharides based on experimental observables, in this case, NMR J-couplings, and minimal input from theory. The present work extends prior studies of trans-O-glycosidic J-couplings26-34,38,39 in four significant ways: (1) 12 structurally related β-(1→4)-linked disaccharides have been studied (Scheme 3) containing systematic covalent modifications in the vicinity of their glycosidic linkages to potentially broaden the range of preferred ϕ and ψ values, (2) the dependencies of specific trans-O-glycosidic J-couplings on coupling pathway structure have been investigated and parametrized by DFT, (3) an improved mathematical algorithm has been employed to extract conformational models of these linkages from ensembles of redundant NMR J-couplings, and (4) aqueous MD simulations of the same disaccharides studied experimentally were conducted to permit direct comparisons between NMR- and MD-based models of linkage conformation. We show that NMR-based conformational models are in close agreement with those obtained from MD simulations, that relatively small differences in linkage conformation are detectable by the method, and that conformational classifications of O-glycosidic linkages can be achieved using this approach.</p><!><p>Twelve disaccharides were prepared by chemical synthesis: methyl β-d-galactopyranosyl-(1→4)-β-d-glucopyranoside (methyl β-lactoside) (2),40 methyl β-d-galactopyranosyl-(1→4)-α-d-glucopyranoside (methyl α-lactoside) (4),41 methyl β-d-glucopyranosyl-(1→4)-β-d-glucopyranoside (methyl β-cellobioside) (5),42,43 methyl β-d-glucopyranosyl-(1→4)-α-d-glucopyranoside (methyl α-cellobioside) (6),44 methyl β-d-galactopyranosyl-(1→4)-α-d-mannopyranoside (7),45 methyl β-d-xylopyranosyl-(1→4)-β-d-mannopyranoside (8),46 methyl β-d-mannopyranosyl-(1→4)-2-acetamido-2-deoxy-β-d-glucopyranoside (9),47,48 methyl 2-acetamido-2-deoxy-β-d-glucopyranosyl-(1→4)-2-acetamido-2-deoxy-β-d-glucopyranoside (methyl β-chitobioside) (10),43,49 methyl 2-deoxy-β-d-glucopyranosyl-(1→4)-β-d-glucopyranoside (11),50 methyl β-d-galactopyranosyl-(1→4)-β-d-allopyranoside (12),51 methyl β-d-galactopyranosyl-(1→4)-β-d-xylopyranoside (13),52 and methyl β-d-mannopyranosyl-(1→4)-β-d-xylopyranoside (14) (Scheme 3).53 Compounds 2 and 4–14 were prepared with 13C-labeling at two carbons to facilitate the measurement of J-couplings across their internal β-(1→4) O-glycosidic linkages. Labeling Mode 1 was employed for 6, 7, 8, 9, 12, and 14, and Mode 2 was employed for all others (Scheme 4).</p><!><p>High-resolution 1H and 13C{1H} NMR spectra were obtained on ~20 and ~100 mM aqueous (2H2O) solutions, respectively, at 22 °C, using 5 mm NMR tubes on a 600-MHz FT-NMR spectrometer equipped with a 5 mm 1H–19F/15N–31P dual broadband probe. 1H NMR spectra were collected with a ~2800 Hz spectral window and ~4 s recycle time, and FIDs were zero-filled to give final digital resolutions of <0.01 Hz/pt. 13C{1H} NMR spectra were collected with a ~12 800 Hz spectral window and ~5 s recycle time, and FIDs were zero-filled to give final digital resolutions of <0.05 Hz/pt. FIDs were processed with resolution enhancement (Gaussian or sine-bell functions) to improve resolution and facilitate the measurement of small J-couplings (≥0.5 Hz), and reported J-couplings are accurate to ±0.1 Hz, unless otherwise stated. Coupling signs were determined using the projection resultant rule54 and/or from DFT calculations.</p><!><p>Density functional theory (DFT) calculations were conducted within Gaussian0955 using the B3LYP functional56,57 and 6-31G* basis set58 for geometric optimization. The calculations included the effects of solvent water, which were treated using the self-consistent reaction field (SCRF)59 and the integral equation formalism (polarizable continuum) model (IEFPCM).60 Geometries were calculated in fully substituted models of the 12 disaccharides (2 and 4–14; Scheme 3). Exocyclic torsion angles in each model structure were constrained as described in Scheme S2 (Supporting Information); only one set of torsion angles was investigated at each incremented value of ϕ and ψ unless otherwise noted. Torsion angles ϕ and ψ (Schemes 1 and 5) were varied in 15° increments through 360° and held constant during geometry optimization, yielding 576 final structures of each disaccharide.</p><!><p>JHH, JCH, and JCC values were calculated in each geometry-optimized structure of 2 and 4–14 using DFT and the B3LYP functional56,57 in Gaussian09.55 The Fermi contact,61-63 diamagnetic and paramagnetic spin–orbit, and spin–dipole terms61 were recovered using a specially designed basis set, [5s2p1dl3s1p],10,64 and raw (unscaled) calculated couplings are reported and are accurate to within ±0.2–0.3 Hz on the basis of prior work.64 The J-coupling calculations also included the effects of solvent water, which were again treated using the self-consistent reaction field (SCRF)59 and the integral equation formalism (polarizable continuum) model (IEFPCM)60 as implemented in Gaussian09. Plots of DFT-calculated J-couplings vs ϕ or ψ were generated using the graphics software, Prism.65</p><!><p>Equations relating DFT-calculated J-couplings to either ϕ or ψ were parametrized using the scipy and numpy packages in Python.66 During parametrization, phase shifts of +120° or −120° were applied to ϕ and ψ, respectively, as needed to give equations that use conventional definitions of ϕ and ψ (ϕ = H1′–C1′–O4–C4; ψ = H4–C4–O4–C1′) (Scheme 5). Equations were parametrized using J-couplings calculated in all conformers and in conformers whose energies were less than or equal to a 10 kcal/mol energy cutoff (see discussion below). A second constraint was also applied when needed to remove DFT-derived structures containing distorted aldopyranosyl rings. To screen consistently for the latter distortions, Cremer–Pople puckering parameters were calculated from DFT-generated Cartesian coordinates and a θ value of 35° was used for the barrier.67,68 The goodness-of-fit of each equation is reported as a root mean squared (RMS) deviation.</p><!><p>An experimental J-coupling for a molecule in solution is related to calculated (predicted) J-couplings associated with explicit conformers of the same molecule by calculating J-couplings in all possible conformers and weighting them on the basis of the probability distribution (relative population) of each conformer present in solution. For an experimental J-coupling, Jexp, that is sensitive to a specific molecular torsion angle θ, this relationship can be modeled using eq 1 where p(θ) is the probability at angle θ, and J(θ) is the J-value associated with torsion angle θ.69-71 Integration of eq 1 across all values of θ (0−360°) yields a predicted J value that is used to determine the relative populations of conformers in solution. Prior computational approaches to calculate continuous probabilities have assumed that p(θ) is zero outside a finite number of θ values, so that eq 1 simplifies to eq 2.72 By treating different Jexp values that report on the same θ, eq 2 can be solved as a system of n variables (typically three or less; e.g., three staggered conformations about the C–C bond of a X1– C1–C2–X2 molecular fragment). This approach is limited by assumptions made about the identities of allowable conformers in solution. When information needed to constrain θ is unreliable or unavailable, or when the assumption that discrete conformations about θ exist in solution is suspect, the approach becomes problematic. This method also assumes essentially zero libration about the preferred value(s) of θ. (1)Jexp=∫02πJ(θ)p(θ)dθ (2)Jexp=∑i=0nJ(θi)p(θi)</p><p>The computational approach applied in this work models p(θ) as a continuous distribution across all values of θ. The population can be modeled as a single or sum of several (typically two or three) Gaussian-like distributions, whose relative heights, widths, and positions are allowed to vary freely. Using the additional structure afforded by this approach, an expression for the expected J values of any given population can be derived. This expression is used to back-calculate expected J values, and the distributions are then recursively optimized to fit the expected values to the experimental data. Because of the intrinsic speed of the calculations, a complete computation can be performed over all possible θ values. The resulting fit is less biased than that obtained using previous methods69-71,73-75 for a diverse set of conformational scenarios. In this modified treatment, no assumptions are made about the mean positions or relative populations of conformers. The mathematical expressions used to describe the rotamer populations yield quantitative information about the mean positions and abundances of the stable conformers, and in some cases the degree of libration about each maximum.</p><p>In the present work, NMR J-coupling ensembles and DFT parametrization were combined with Fredholm theory and circular statistics to determine the conformational populations of two-bond, β-(1→4) O-glycosidic linkages (Scheme 1) using the statistical software, MA'AT. O-Glycosidic torsion angles, ϕ and ψ, were modeled as a single von Mises distribution, which yields two fitting parameters, the mean position and the circular standard deviation (CSD) of ϕ or ψ. Von Mises distributions were chosen for modeling because they are the most commonly encountered distribution for independent random variables on a circular axis, according to the central limit theorem. Monte Carlo methods were used to generate model parameters and least-squares methods were used to minimize the RMS deviation between the experimental and predicted J-couplings. The mean position and CSD of each model were calculated using the circular package in R.76,77 Approximate standard errors for the model parameters were computed by taking the square root of the diagonal elements from the estimated covariance matrix.78</p><p>Continuous and discrete models of molecular torsions angle using J-couplings have been used previously in structural studies of various macromolecules, including proteins, nucleic acids, and saccharides.69-75 When these models are assessed, accuracy, generalizability, computational requirements, and the number of available experimental restraints need to be considered. The computational approach used here differs from prior approaches in three key respects. First, no assumption is made about the mean position(s) of the rotamer population(s). Second, the two fitting parameters, mean position and CSD, are calculated simultaneously, eliminating bias. The use of only two parameters reduces computational requirements. Third, the reduced number of model parameters allows unique solutions to be determined in most cases using two or three experimental structure restraints (in this case, J-couplings). It should be appreciated that the computational algorithm encoded in MA'AT is not limited to modeling only a single von Mises distribution. However, methods to determine the uniqueness of solutions of multistate models are still under development. In this work, a single-state model was chosen out of necessity because, at present, only three J-couplings are available in disaccharides 2 and 4–14 to evaluate ϕ and ψ. Applying more complex two- or three-state models to treat this limited number of experimental restraints is unjustified mathematically. However, future modeling of pyranosyl and furanosyl ring, and exocyclic hydroxymethyl, conformations using the MA'AT software and different distributions (e.g., von Mises,77 wrapped Cauchy,77 uniform,79,80 raised cosine,79,80 and Cartwright's power of cosine distributions;81 see recent study of O-acetyl side-chain conformation82) will permit two- and three-state models to be considered because significantly greater numbers of redundant J-couplings are available in these systems. Additional J-coupling restraints for the modeling of ϕ and ψ are also possible (Schemes 2 and 3), which may allow more complex models to be tested in the future. Additional structure restraints, including NOEs,37,83 residual dipolar couplings (RDCs)84-86 and/or residual chemical shift anisotropies (RCSAs)84,87-89 can also be used to expand the number of experimental observables and allow more complex models and distributions of ϕ and ψ to be tested. The MA'AT software in its current form is capable of treating multistate models and using different distributions as circumstances allow.</p><!><p>Initial structures of disaccharides 2 and 4–14 (Scheme 3) were built using the Carbohydrate Builder module available at the GLYCAM Web site (http://www.glycam. org).90 The GLYCAM0691 (version j) force field was employed in all simulations. The disaccharides were solvated with TIP3P92 water using a 12 Å buffer in a cubic box, using the LEaP module in the AMBER14 software package.93 Energy minimizations for the solvated disaccharides were performed separately under constant volume (500 steps steepest descent, followed by 24 500 steps of conjugate-gradient minimization). Each system was subsequently heated to 300 K over a period of 50 ps, followed by equilibration at 300 K for a further 0.5 ns using the nPT condition, with the Berendsen thermostat94 for temperature control. All covalent bonds involving hydrogen atoms were constrained using the SHAKE algorithm,95 allowing a simulation time step of 2 fs throughout the simulation. After equilibration, production simulations were carried out with the GPU implementation96 of the PMEMD.MPI module, and trajectory frames were collected every 1 ps for a total of 1 μs. One to four nonbonded interactions were not scaled,97 and a nonbonded cutoff of 8 Å was applied to van der Waals interactions, with long-range electrostatics treated with the particle mesh Ewald approximation. The output from each MD simulation was imported into Prism65 for visualization.</p><!><p>NMR J-couplings across the β-(1→4)-linkages of 2 and 4–14 were measured in aqueous solution (Table 1; see representative NMR spectra in Figures S1–S3, Supporting Information). The disaccharides in Table 1 were chosen for study with the expectation that different covalent structure proximal to the linkage will affect linkage conformation and thus provide a means to (a) assess the sensitivities of specific NMR J-couplings to linkage conformation, (b) classify linkage conformations qualitatively on the basis of comparisons of J-coupling ensembles, (c) develop and validate statistical models of linkage torsion angles using J-coupling ensembles, and (d) compare the conformational models obtained from (c) to those predicted from MD simulations.</p><p>Disaccharides 2 and 4–14 give essentially identical ensembles of J-couplings that depend on ϕ (Table 1). Average ϕ-dependent J-couplings with their standard deviations were calculated as follows: 2JC1′,C4 = −1.9 ± 0.1 Hz 3JC4,H1′ = 4.0 ± 0.1 Hz, and 3JC2′,C4 = 3.1 ± 0.1 Hz. The standard deviations are within the experimental errors of the measurements. Changes in configuration at sites proximal to and remote from the internal β-(1→4) linkage do not affect the ϕ-dependent J-couplings, including differences in configuration at C2′. For example, a 3JC2′,C4 value of 2.9 Hz is observed in 13, which bears an equatorial O2′, and in 14, which bears an axial O2′. The removal of O2′ from 5 to give 11 has little effect on 2JC1′,C4, 3JC4,H1′, and 3JC2′,C4 values. These findings show qualitatively that the differences in covalent structure between disaccharides 2 and 4–14 do not affect the conformational preferences of ϕ in solution; the behavior of ϕ in the 12 disaccharides is virtually identical. This finding is likely explained by mutually reinforcing stereoelectronic,11-19,98 steric and other noncovalent (e.g., H-bonding) forces, or to some competitive interplay between them, that heavily control ϕ and are largely unaffected by covalent structure in the vicinity of the linkages.</p><p>In contrast, the ensembles of ψ-dependent J-couplings in disaccharides 2 and 4–14 differ significantly (Table 1), allowing a classification into three groups based mainly on the values of 3JC1′,C3 and/or 3JC1′,C5. Group I disaccharides (containing 2 and 4–11) have the same configurations at C3 and C5, and neither the configuration at the anomeric carbon bearing the OCH3 aglycone nor the configuration at C2, C2′, C4′ or C5′, affect the conformational properties of the β-(1→4) linkage. Removal of the sterically demanding exocyclic hydroxymethyl (CH2OH) side chain at C5, however, triggers a major change in conformation about ψ (e.g., compare 2 and 13). This change is unaffected by configuration at C2′; disaccharides 13 and 14 comprising Group III give similar ψ sensitive J-couplings despite their different C2′–O2′ bond orientations. The ψ-sensitive J-coupling ensemble for 12 (Group II) contains a 3JC1′,C5 value that is ~1 Hz larger than those observed in Group I disaccharides, and a 3JC4,H1′ value that lies between those observed for the disaccharides in Groups I and III, indicating that configuration at C3 influences the conformational preference of ψ in solution.</p><p>As shown above, qualitative inspections of ϕ- and ψ-dependent J-coupling ensembles allow O-glycosidic linkages to be classified on the basis of linkage conformation, which is difficult to achieve by existing experimental methods. In the 12 β-(1→4) linkages studied, this classification was based solely on ψ, because conformational behaviors about ϕ were essentially identical in this set of linkages. Disaccharides such as 2 and 4–14 provide conformational information on "isolated" β-(1→4) linkages. When embedded into larger oligosaccharides, these linkages will be influenced by additional structural factors that may favor different conformations from those found in the simple disaccharide. Importantly, simple qualitative inspections of ϕ- and ψ-sensitive J-coupling ensembles quickly reveal conformational changes caused by differences in structural context.</p><!><p>DFT-Derived potential energy surfaces (PESs) for disaccharides 2 and 4–14 are very similar (see Figure S4, Supporting Information), and global and local energy minima extracted from these surfaces are summarized in Table 2. The energy data do not affirm the conformational classification that evolved from inspections of ϕ- and ψ- dependent J-coupling ensembles (Table 1). Values of ϕ between 0° and +60° are highly represented in the lowest energy structures, as expected from stereoelectronic considerations,11-19,98 but ϕ values near 180° commonly appear within ~2 kcal/mol of these structures.99 In one case (13), a ϕ of 181° is found in the lowest energy structure. However, because the energy differences between the three lowest energy structures are <3 kcal/mol in nearly all cases, assignments of preferred ϕ torsion angles cannot be made with confidence. Similar arguments pertain to ψ, with values ranging from −30° to +30° commonly observed in the lowest energy structures, but values of ~180° also occurring in structures within a few kcal/mol of these structures.83,100-102 The greater uniformity of ϕ values, relative to that of ψ values, in the most stable structures suggests that the former are more constrained in the 12 disaccharides, in agreement with the J-coupling data. However, the PES data were collected over a very small fraction of the complete energy hypersurface, because only a very small subset of exocyclic torsion angles was sampled in the DFT calculations (Scheme S2, Supporting Information). Thus, whereas the PES results provide clues about the preferred regions of ϕ /ψ space, reliable determinations of the preferred values of ϕ and ψ in 2 and 4–14 cannot be made from these data.</p><!><p>Equations were parametrized by DFT for six of the ten J-couplings that depend on ϕ and ψ in disaccharides 2 and 4–14 (Scheme 1): for ϕ, 3JC4,H1′, 2JC1′,C4, and 3JC2′,C4; for ψ, 3JC1′,H4, 3JC1′,C3, and 3JC1′,C5. Equations for a given J-coupling were then compared, and those that were very similar across all structures were combined to give a single generalized equation (individual equations for each J-coupling in disaccharides 2 and 4–14 are available in the Serianni Laboratory Data Archive at the following URL: https://www3.nd.edu/~aserilab/Disaccharide_Database/Home/DATA.html; (contact A. Serianni for access privileges).</p><p>Prior work has shown that trans-O-glycoside J-couplings depend primarily on either ϕ or ψ.26-34 In this study, systematic effects caused by the second torsion angle were observed for some 3JCOCC values. Modeling these dual dependencies requires equations with at least one two-dimensional component to describe the significant features of the data adequately. Some of these features, however, may not be practically important because they occur in regions of ϕ /ψ space occupied by highly strained structures that are poorly sampled in solution. For example, a 2D plot of calculated 3JC1′,C5 values in βGal14βXylOCH3 13 against ϕ and ψ shows that 3JC1′,C5 depends primarily on ψ as expected, but a secondary dependence on ϕ is also observed, especially at ψ values near 120° (Figure 1). The differential secondary dependence of 3JC1′,C5 on ϕ is more apparent in a conventional plot of calculated 3JC1′,C5 vs ψ, where the dispersion at a given ψ value reveals the secondary dependence on ϕ (Figure 2A). The magnitude of the latter dispersion varies with ψ, with the greatest dispersion observed at ψ = ~ 120° (i.e., in geometries in which C1′ and C5 are eclipsed). Values of ψ near 120° are rarely, if ever, adopted by β-(1→4)linkages; for example, at ψ = 120°, the exocyclic CH2OH side-chain in the βGlcp residue of 2 is in close proximity to O5 of the βGalp ring, destabilizing this conformation. Linkage geometries with ψ near 120° produce strain in geometry-optimized structures, leading to perturbations in the calculated J-couplings and to the increased ϕ-induced dispersion at ψ values near 120° in Figure 2A.</p><p>The effects of the above-noted structural perturbations on parametrized equations were reduced by using calculated J-couplings in geometry-optimized structures of disaccharides 2 and 4–14 within 10 kcal/mol of the global energy minimum structures found in DFT-derived potential energy surfaces. A secondary constraint was applied to remove structures containing an aldopyranosyl ring that was distorted significantly from the 4C1 ring form (see Computations). This data reduction did not much affect the overall shape of the fit lines, as shown in Figure 2, but removed much of the dual dependency, allowing simpler and more realistic parametrized equations. The regions of the fit curve most likely to affect its usefulness for evaluating ψ in solution (ψ values of 0° ± 60°) remained largely unaffected. An alternative solution to this problem involves weighting J-couplings from each rotamer according to the probability of finding that rotamer based on the Boltzmann distribution. This approach was not taken because the Boltzmann distribution weights rotamers located near the DFT-calculated global energy minimum too heavily. The accuracy of this minimum cannot be assumed because (1) the complete conformational space of each disaccharide was not sampled due to current practical limitations of DFT calculations, and (2) solvent (water) effects on the energy landscape are not adequately described because the solvent treatment does not capture the effects of hydrogen bonding.103 Applying two data reduction parameters gave larger standard errors in each equation than those previously reported.30 These previous studies generated equations by sampling a much smaller percentage of ϕ/ψ space (typically three staggered rotamers) than was sampled in this work. The larger errors indicate that the variance of possible J-coupling values at any value of ϕ or ψ may be greater than previously thought.</p><!><p>The behaviors of the three 3JCOCC values for β-(1→4) glycosidic linkages (3JC2′,C4, 3JC1′,C3, 3JC1′,C5) cannot be captured by a single generalized equation because they involve coupling pathways that differ with respect to internal and terminal electronegative substituents.30-34,39,104 These differences affect curve amplitudes and/or the locations of curve maxima and minima.</p><p>The dependencies of 3JC2′,C4 on ϕ in disaccharides 2 and 4–14 are shown in Figure 3. The scattering of lines observed between 210° and 330° arises from the secondary dependence of 3JC2′,C4 on ψ (C2′ and C4 are eclipsed at ϕ = 240°); this scattering occurs at ϕ values that are not preferred on the basis of stereoelectronic considerations.11-19,98 Because similar ϕ dependencies are observed in 2, 4–7, and 9–12 in which the terminal hydroxyl group on C2′ is equatorial, a generalized equation was derived from the data on these nine structures (eq 3). Disaccharides 8 and 14 bear axial hydroxyl groups on the terminal C2′ carbon of the coupling pathway, which causes a ~10° shift in the curves relative to those observed for disaccharides bearing an equatorial O2′. Equations for 3JC2′,C4 obtained for 8 and 14 were thus combined into a generalized equation (eq 4). Disaccharide 11 bears no hydroxyl group at C2′, and its removal affects the dependence of 3JC2′,C4 on ϕ; for example, the minimum in the curve at ϕ = ~135° is located between that for 3JC2′,C4 (O2′ eq) and 3JC2′,C4 (O2′ ax) (Figure 3A). A separate equation was parametrized to treat this case (eq 5). (3)JC2′,C43(O2′eq)=1.56+0.24cos(ϕ)+0.45sin(ϕ)−0.13cos(2ϕ)+1.76sin(2ϕ)rms0.41Hz (4)JC2′,C43(O2′ax)=1.48+0.66cos(ϕ)+0.39sin(ϕ)−0.50cos(2ϕ)+1.55sin(2ϕ)rms0.41Hz (5)JC2′,C43(O2′deoxy)=1.70+0.40cos(ϕ)+0.11sin(ϕ)−0.64cos(2ϕ)+1.83sin(2ϕ)rms0.43Hz</p><p>Prior studies of 3JC1,C6 and 3JC3,C6 values in aldohexopyranosyl rings such as 1 have shown that coupling between the carbons is enhanced when the terminal oxygen substituents lie in the Cα–O–C–Cβ plane (Cα and Cβ are antiperiplanar in these pathways).30,105,106 The same pathway bearing out-of-plane terminal oxygen substituents appeared to exert the same effect on 3JCC magnitude as the pathway devoid of terminal oxygen substituents.30 The C2′–C1′–O1'–C4 coupling pathway pertinent to 3JC2′,C4 orients the coupled carbons antiperiplanar at ϕ = 60°. In this geometry, the terminal O2′ cannot assume an in-plane orientation regardless of whether it is axial or equatorial. The three curves for eqs 3–5 shown in Figure 3B overlap at ϕ = 60°, reaffirming the effects of in- and out-of-plane terminal electronegative substituents on 3JCOCC observed for analogous coupling pathways in aldohexopyranosyl rings.</p><p>The behavior of 3JC1′,C5 in βGal14βXylOCH3 (13) and βMan14βXylOCH3 (14) differs from that observed for the remaining disaccharides, giving curves with somewhat larger amplitudes at ψ values of 120° and 300° (Figure 4A). The differences are attributed to the lack of a −CH2OH side chain in 13 and 14 to which the coupled C5 is attached. Two equations were parametrized to treat the pathways bearing a −CH2OH substituent (eq 6) and those in which this substituent is replaced by hydrogen (eq 7) (Figure 4B). (6)JC1′,C53(CH2OH)=2.05+0.49cos(ψ)−0.79sin(ψ)−1.01cos(2ψ)−2.29sin(2ψ)rms0.68Hz (7)JC1′,C53(H)=2.42+0.65cos(ψ)−0.54sin(ψ)−1.44cos(2ψ)−2.45sin(2ψ)rms1.00Hz</p><p>The C1′–O1′–C4–C5 coupling pathway in β-(1→4)-linked disaccharides mimics the C1–O5–C5–C6 coupling pathway in aldohexopyranosyl rings and is subject to the same terminal electronegative substituent effects (see above).30,34,105 The C1′–O1′–C4–C5 torsion angle is ~180° at ψ = ~300°, and the associated calculated 3JC1′,C5 values lie between 5.5 and 6.0 Hz (Figure 4A). In the latter geometry, the in-plane O5 is expected enhance 3JC1′,C5 values relative to the analogous 3JC1′,C3 values for C1′–O1′–C4–C3 coupling pathways that lack this feature (see below).</p><p>The dependence of 3JC1′,C3 on ψ in βGal14βAllOCH3 12, which bears an axial O3, differs slightly from that in the remaining 11 disaccharides, which bear equatorial O3 atoms (Figure 5A). An ~15° phase shift in the curve is observed for 12 relative to those found for the remaining disaccharides. These terminal electronegative atom effects were captured in parametrized eqs 8 and 9 to treat 3JC1′,C3 values in β-(1→4)-linked disaccharides bearing equatorial and axial O3 atoms, respectively (Figure 5B). (8)JC1′,C33(O3eq)=1.85+0.20cos(ψ)+0.71sin(ψ)−1.81cos(2ψ)+1.87sin(2ψ)rms0.49Hz (9)JC1′,C33(O3ax)=1.98+0.49cos(ψ)+0.72sin(ψ)−1.27cos(2ψ)+1.49sin(2ψ)rms0.66Hz</p><!><p>Prior studies have shown that geminal 2JCOC values across O-glycosidic linkages depend primarily on ϕ and the C–O–C bond angle.38 The superimposition of both factors adds complexity to plots of 2JC1′,C4 vs ϕ (Figure 6); because rotation about ψ also affects the C–O–C bond angle, this torsion angle affects 2JC1′,C4 values indirectly. These interdependencies are revealed in 2D contour plots, one of which is shown in Figure 7 for βGal14βGlcOCH3 2. 2JC1′,C4 is largely independent of ψ at ϕ torsion angles of 0° ± 60°, but a significant effect of ψ on 2JC1′,C4 is observed at the remaining ϕ values. In the current parametrization, the curves in Figure 6 were combined to give a single generalized equation (eq 10), with the realization that the effects of ψ are not encoded completely but with the expectation that this limitation will not detract from its usefulness in evaluating ϕ. (10)JC1′,C42−2.53+0.94cos(ϕ)−0.84sin(ϕ)−0.12cos(2ϕ)−0.44sin(2ϕ)rms0.50Hz</p><!><p>Two trans-O-glycosidic 3JCOCH values show primary dependencies on either ϕ (3JC4,H1′) or ψ (3JC1′,H4) and exhibit dynamic ranges of ~10 Hz (Figure 8).28,29,31,34 Visualizations of DFT data in 2D contour plots, shown for βGal14βGlcOCH3 2 in Figure 9, showed that 3JC4,H1′ is largely unaffected by ψ and 3JC1′,H4 is largely unaffected by ϕ. Because equation parametrization in disaccharides 2 and 4–14 showed very similar dependencies of 3JC4,H1′ and 3JC1′,H4 on the C–O–C–H torsion angle, the resulting 24 equations were combined to give a generalized eq (eq 11) to treat both 3JCOCH values (θ in eq 11 denotes either ϕ or ψ). (11)JCOCH3=3.81−1.91cos(θ)+0.08sin(θ)+3.92cos(2θ)+0.34sin(2θ)rms0.72Hz</p><p>Plots of DFT-calculated J-couplings as a function of ϕ or ψ are shown in Figures S5–S7 (see Supporting Information) for disaccharides 2, 12, and 13, respectively. These plots illustrate the extent of the secondary dependencies of each J-coupling and the best fit of the data, from which parametrized eqs 3–11 were derived. The data shown for 2, 12, and 13 are representative of similar data obtained on the remaining nine disaccharides.</p><!><p>The above findings confirm those of earlier studies30-32,34,38,39,107 that different orientations (i.e., axial vs equatorial) of electronegative (hydroxyl) groups affect the parametrization of J-coupling equations when these groups are attached to internal and terminal carbons in the coupling pathway (Figures 3 and 5). J-couplings may also be affected by the conformations of C–O bonds, and possibly of C–C bonds, especially when these bonds involve coupling pathway carbons.32,34 DFT calculations were conducted on disaccharides 2 and 4–14 in which all hydroxyl group conformations (i.e., exocyclic C–O bond torsions) were fixed at discrete values (Scheme S2, Supporting Information). To determine the magnitudes of the conformational effects of C–O and C–C bonds, a single linkage conformer of βGal14βGlcOCH3 2 containing fixed ϕ (28°) and ψ (−8°) torsion angles was used to determine the effects of C2′–O2′, C3–O3, and C5–C6 bond conformations on calculated trans-O-glycoside JCH and JCC values. One bond (C2′–O2′, C3–O3, or C5–C6) was rotated in 15° increments through 360° during geometry optimization while ϕ, ψ and all remaining exocyclic torsion angles were fixed at values shown in Scheme S3 (Supporting Information). J-coupling calculations were then conducted on each conformer of 2. The results show that rotations of the three bonds do not affect the calculated J-couplings significantly, with variations of ~0.5 Hz observed if only staggered C–O and C–C bond conformations are considered (Figure S8, Supporting Information). These findings support the conclusion that eqs 3–11 are sufficiently quantitative to model ϕ and ψ in disaccharides 2 and 4–14.</p><!><p>Parametrized J-coupling eqs 3–11 and experimental J-couplings (Table 1) were used to generate single-state von Mises models of the rotamer distributions about ϕ (Figure 10A) and ψ (Figure 10B) in each disaccharide. The mean positions, circular standard deviations (CSDs), and RMS errors of these models are shown in Table 3. The RMS errors of 0.3–0.4 Hz indicate a good fit of the experimental J-couplings to the von Mises model. The similar mean positions of ϕ (23–33°, with errors of ±12°) show that the conformational properties of ϕ are essentially identical in the 12 disaccharides insofar as J-coupling ensembles are able to discriminate between different single-state models. This finding is consistent with the qualitative analyses of ϕ discussed in section IVA, and with expectations for ϕ based on stereoelectronic considerations (the exo-anomeric effect11-19,98 favors ϕ values near +60° for β-glycosides) (Scheme 5).</p><p>Although the differences in the experimental J-couplings sensitive to ψ are small (maximum differences of ~2 Hz; Table 1), the mean positions of ψ determined from a von Mises model vary significantly (–22° to +17°, with errors of ±8°) (Figure 10B, Table 3), thus providing a measure of the sensitivity of J-coupling ensembles to relatively minor changes in the behaviors of molecular torsion angles. Unlike ϕ, the models show that the mean positions of ψ fall into three groups. The mean positions of ψ in Groups II (disaccharide 12) and III (disaccharides 13 and 14) differ by ~39°. Mean positions of ψ in Group I (disaccharides 2 and 4–11) locate between those in Groups II and III (Table 3). The three sets of overlapping curves in Figure 10B average to give three distinct ψ distributions associated with Groups I–III (Figure 11). Newman projections for ψ in each group, superimposed on the average von Mises distributions of ψ shown in Figure 11, are shown in Scheme 6. These results demonstrate quantitatively that ψ is significantly more affected than ϕ by the structural differences between disaccharides 2 and 4–14 and provide a structural justification for classifying these disaccharides into three groups on the basis of qualitative inspection of their ψ dependent J-coupling ensembles (Table 1).</p><p>The single-state von Mises models used to treat ϕ and ψ yield two parameters (mean and CSD), which allow a determination of the uniqueness of the model by visual inspection of the parameter space (Figure 12 and Figures S9 and S10, Supporting Information). Unique solutions were observed in many, but not all, cases (for example, see data for ψ in disaccharide 13 in Figure S10, Supporting Information). The parameter space of each model contained two or three minima, with the global minimum giving the smallest RMS in most cases. However, in some cases, two minima gave very similar RMS errors; for example, ψ in 13 gave two minima with RMS errors of 0.33 Hz. Nuclear Overhauser effect (NOE) and/or residual dipolar coupling (RDC) constraints will be needed to reduce and/or eliminate local minima and to more completely test the uniqueness of fit. Additional J-coupling constraints may also be available, including the germinal couplings 2JH1′,C2′, which depends on ϕ, and 2JC3,H4, 2JC5,H4, and 2JC3,C5, which depend on ψ (Scheme 1).34,108</p><!><p>The ϕ and ψ torsion angles in the crystal structures of disaccharides 2, 4, 5, 7, 8, and 12–14 differ significantly from the idealized values of ±60° and 180°, with mean values of +35° ± 9° and –30° ± 13°, respectively (Table 4). Although the behaviors of ϕ and ψ in crystal structures may not faithfully reflect those in solution, the data show that analyses of experimental observables (e.g., J-couplings) measured in solution cannot assume the exclusive presence of staggered rotamers about these torsion angles. The larger standard deviation for ψ also suggests that ψ may be prone to more disorder than ϕ; that is, ψ is influenced more strongly by linkage structure than ϕ. This difference in behavior may reflect the additional stereoelectronic constraint (exo-anomeric effect11-19,98) on ϕ that favors the g+ rotamer (ϕ = +60°) (Scheme 5).</p><!><p>Distributions of ϕ and ψ obtained from aqueous 1 μs MD simulations on disaccharides 2 and 4–14 are summarized in Table 5 (MD histograms for 2, 12, and 13 are shown in Figure 13; corresponding data for the remaining nine disaccharides are shown in Figures S11–S19 in Supporting Information). Mean values of ϕ range from 40° to 46°, which are larger than the NMR-determined mean values of 23°–33° (Table 3). Mean values of ψ range from –20° to +16°, which compares favorably to the corresponding NMR-determined range of –22° to +17° (Table 3). The MD results indicate that ϕ is less disordered than ψ and less influenced by linkage structure, results consistent with those obtained from J-coupling analyses. For example, the MD data for βGal14βGlcOCH3 2 (Group I), βGal14βAllOCH3 12 (Group II), and βGal14βXylOCH3 13 (Group III) show essentially no differences in the mean values of ϕ but significant differences in the mean values of ψ (Figure 13). The absence of an exocyclic hydroxymethyl group at C5 in 13 leads to a more positive ψ value and promotes greater libration about ψ, as reflected by the larger CSD, compared to the case for 2. In contrast, the presence of an axial O3 in 12 compared to an equatorial O3 in 2 shifts ψ to a more negative value, presumably due to altered steric interactions between O5′ and C3. Newman projections for ϕ and ψ in disaccharides 2, 12, and 13 superimposed on the statistical distributions of ϕ and ψ determined by NMR and MD (Figure 14) reveal an ~15° difference in the mean values of ϕ, and very good agreement in the behavior of ψ, between the two treatments.</p><p>MD simulations of disaccharides 8, 13, and 14 produced erroneous results with respect to the conformational behaviors of their constituent β-d-xylopyranosyl rings (Figures S20–S22, Supporting Information). In these simulations, the β-xylo ring partitioned almost equally between the 4C1 and 1C4 ring conformers, in contrast to NMR J-coupling data that show a very high (>95%) preference for the 4C1 form in aqueous solution. This behavior points to a weakness in the GLYCAM06 force field that will require corrective action. In this work, the MD trajectories obtained for 8, 13 and 14 were purged of all structures containing β-xylo rings in the 1C4 conformation, and the resulting histograms were used to calculate the means and CSD values shown in Table 5.</p><!><p>This investigation aimed to develop NMR-based models of the conformationally mobile O-glycosidic torsion angles, ϕ and ψ, in a group of structurally related β-(1→4)-linked disaccharides. The intent was to derive these models from an analysis of redundant trans-O-glycosidic NMR J-couplings, aided by density functional theory to parametrize equations relating specific JCH and JCC values to either ϕ or ψ. The experiment-based models were compared to those determined from aqueous molecular dynamics simulations and from analyses of crystal structures. The end products of this work (a) demonstrate the feasibility and reliability of deriving conformational models of O-glycosidic linkages based solely on J-coupling data and (2) provide a means of validating models obtained from MD simulation.</p><p>Unbiased experimental modeling of β-(1→4) O-glycosidic linkages confirms that ϕ and ψ do not adopt perfectly (or near perfectly) staggered rotamers in solution. The latter are often assumed in simpler interpretations of J-couplings in conformationally flexible molecular fragments in solution (e.g., hydroxyl109,110 and hydroxymethyl35,64,111 conformations in saccharides; side-chain conformations in peptides/proteins75,112). The glycosidic torsion angles in 12 disaccharides were well modeled by a single-state von Mises distribution, based on the low RMS errors obtained in the data fitting. The NMR-determined means and CSDs for ϕ and ψ were in very good agreement with those obtained from MD simulation. The number of experimental observables used to treat ϕ and ψ in this work (three for each torsion angle) precluded the testing of two- and three-state models on statistical grounds. For example, to test a two-state model, a minimum of three adjustable parameters is required if assumptions about which conformations are allowed or disallowed are not made a priori. If additional information on the conformational disorder of each state is desired, the number of adjustable parameters increases to five. Thus, expanding the data analysis to include two- and three-state models requires more experimental observables than were available in this work. Future improvements in the experimental modeling of O-glycosidic linkages will require additional experimental constraints in the form of additional J-couplings, nuclear Overhauser eflects (NOEs), residual dipolar couplings (RDCs), and/or residual chemical shift anisotropy (RCSAs).</p><p>An inspection of the ϕ- and ψ-dependent J-coupling ensembles for disaccharides 2 and 4–14 showed that ϕ is virtually unaffected by structural changes in the vicinity of the linkage, whereas ψ is affected appreciably. Qualitative inspection of the ψ-dependent ensembles led to a classification of these disaccharides into three groups. Subsequent statistical analyses of the ψ-dependent ensembles confirmed this classification and, importantly, provided explicit conformational models of ψ characterized quantitatively by preferred mean values and CSDs. An inspection of the ψ-dependent J-couplings in 2 and 4–14 showed that even modest changes in the ensembles lead to discernible differences in the preferred mean values of ψ. This finding suggests that larger conformational perturbations of these linkages, potentially caused by their insertion into larger structures, will be detectable through analyses of J-coupling ensembles. It should also be appreciated that the classification of 2 and 4–14 into three distinct groups based on different preferred values of ψ may be a consequence of a relatively small data set. As more β-(1→4) linkages are examined, a more continuous distribution of ψ values may result.</p><p>An underlying assumption of this approach is that DFT-parametrized equations for a specific linkage, derived using an appropriate disaccharide model structure, are applicable to the same linkage in different structural contexts. The fact that generalized equations could be derived for J-couplings sensitive to both ϕ and ψ in 12 structurally distinct β-(1→4)-linked disaccharides provides evidence that this assumption is valid. With these equations in hand, it should be possible to determine the conformational behaviors of β-(1→4) linkages in any context and to identify the structural factors that influence their conformational properties. These linkages occur in human milk oligosaccharides,113,114 complex N- and O-linked N-glycans,115-117 and in other biologically relevant oligo- and polysaccharides.118-120 The methods described herein to treat β-(1→4) linkages should also be applicable to other types of O-glycosidic linkages found in oligo- and polysaccharides, provided that DFT-parametrized equations for these linkages are available. These studies will provide new opportunities to improve present understandings of the relationships between oligosaccharide conformation in solution and biological function. The statistical method employed here to model O-glycosidic linkages also reduces the current heavy reliance on MD simulations to assign the conformational properties of oligosaccharides.</p>

PubMed Author Manuscript

Tuning Fluorescence Direction with Plasmonic Metal\xe2\x80\x93Dielectric\xe2\x80\x93 Metal Substrates

Controlling the emission properties of fluorophores is essential for improving the performance of fluorescence-based techniques in modern biochemical research, medical diagnosis, and sensing. Fluorescence emission is isotropic in nature, which makes it difficult to capture more than a small fraction of the total emission. Metal\xe2\x80\x93 dielectric\xe2\x80\x93metal (MDM) substrates, discussed in this Letter, convert isotropic fluorescence into beaming emission normal to the substrate. This improves fluorescence collection efficiency and also opens up new avenues for a wide range of fluorescence-based applications. We suggest that MDM substrates can be readily adapted for multiple uses, such as in microarray formats, for directional fluorescence studies of multiple probes or for molecule-specific sensing with a high degree of spatial control over the fluorescence emission. SECTION: Physical Processes in Nanomaterials and Nanostructures

tuning_fluorescence_direction_with_plasmonic_metal\xe2\x80\x93dielectric\xe2\x80\x93_metal_substrate

<p>Fluorescence detection has become a dominant technology in the biosciences. Until recently, almost all efforts to modify molecular fluorescence have been based on modification of the chemical structure of the fluorophore. In recent years, this laboratory and others have focused on the modification of fluorescence spectral properties using nearby metallic structures.1–8 These efforts represent a fundamental shift from using the intrinsic far-field emission from fluorophores to using near-field interaction, for modifying the quantum yields, anisotropy, and directionality of the emission. These effects depend on collective oscillation of electrons in metals, which are called surface plasmons. The optical fields near fluorophores can couple to the surface plasmon oscillations of metallic substrates as localized excitations for metal nanoparticles or as propagating surface waves for thin metal films. This offers exciting new opportunities for directing and controlling the flow of optical energy using near-field effects rather than classical far-field optical components.5–7 Over the past several years, we observed that emission from fluorophores located near thin metal films on a glass substrate enters the substrate at an angle defined by the surface plasmon resonance (SPR) angle for the emission wavelength.1,2 This phenomenon of surface-plasmon-coupled emission (SPCE) has several remarkable features that can complement and improve fluorescence technology.9,10 One of the advantages of SPCE is that it directs the fluorescence emission in a sharply defined angle, leading to a narrow cone of emission. This is very useful because fluorescence emission is isotropic in nature, which makes it difficult to detect more than a small fraction of the total emission. With SPCE, most of the emission can be captured by a detector, leading to significant improvements in the fluorescence collection efficiency.1,2 However, SPCE typically occurs at a large angle from the surface normal axis which is a disadvantage in multiplexing or array-type applications. It would certainly be more advantageous if we could obtain the fluorescence emission as a single, highly collimated beam with greater control over the emission directionality.</p><p>In this Letter, we demonstrate that the emission from randomly oriented fluorophores can be converted into narrow emission beaming normal to the surface, with planar metal–dielectric–metal (MDM) substrates. Importantly, our MDM structures can be made with simple vapor deposition methods without the need for top-down nanofabrication of features in the substrate plane. The MDM structures can support both photonic as well as plasmonic modes that can act together under suitable conditions. We show that these structures can be easily adapted for controlling the direction of emission as well as other fluorescence properties, like the emission spectra and lifetimes of fluorophores embedded within the dielectric medium of the substrate. The beaming emission not only improves the fluorescence collection efficiency but also provides the opportunity for a wide range of fluorescence based applications. We suggest that MDM substrates can be integrated in lab-on-chip formats to monitor different kinds of interactions in a single frame with microarray technology. They can be incorporated in a regular microscope for fluorescence imaging or single-molecule studies. They can also be used for directional fluorescence studies of multiple probes or for molecule-specific sensing, with a high degree of spatial control.</p><p>The MDM substrates in the present study were prepared by thermal vapor deposition of 50 nm Ag films on cleaned glass slides, followed by spin coating an aqueous solution of poly(vinyl alcohol) (PVA) containing the dyes. The dielectric thicknesses on the substrates were varied from ∼120 to 400 nm by using different weight percentages of PVA (see Supporting Information, Figure S1). A layer of 50 nm Ag was then vapor-deposited above the PVA layer to obtain the final substrates. The MDM samples were illuminated in the reverse Kretschmann (RK) configuration, which in this case is the incident light impinging on the top Ag surface. The fluorescence was collected using an optical fiber (Scheme 1).</p><p>Both the detector and the sample were placed on a rotation stage to enable excitation and detection at different angles. Further details of the experimental setup and the illumination geometry are provided in the Supporting Information.</p><p>Figure 1 shows the variation in the angular distribution of emission from sulforhodamine101 (S101) embedded in dielectric PVA layers of different thicknesses between 50 nm thick Ag films (Ag-PVA-Ag, MDM substrates). Interestingly, no emission is observed for the samples with <4% PVA concentration. At 4% PVA concentration, which corresponds to a film thickness of ∼150 nm, the emission is highly directional and is mostly concentrated in a narrow beam normal to the substrate, with a half angle spread of ∼18° from the normal. With increasing PVA concentrations, the emission pattern gradually changes, with the s- and p-polarized components appearing at definite angles. At 8% PVA concentration, beaming emission is observed again but is also accompanied by an s-polarized component at a different angle.</p><p>Metallo-dielectric layered structures are currently being investigated for applications like waveguides for photon transport, tunable color filters, and electro-optic devices.11–13 To the best of our knowledge, this is the first observation of beaming emission normal to the surface using planar MDM substrates, without nanoscale lateral features. Recently, plasmonic beaming of fluorescence has been studied with optical antennas consisting of circular gratings and nano-holes.3,4 This approach was confined to a small area and involved complex and expensive nanofabrication. In this context, the observation of beaming emission with simple, large-area Ag-PVA(4%)-Ag substrate, is quite exciting. The significance of this observation is further strengthened by the fact that the emission pattern is unaffected by the direction of excitation. Similar beaming emission is observed even when the excitation angle is changed to 40° instead of the normal excitation at 0° (see Supporting Information, Figure S2).</p><p>On the basis of our previous studies on SPCE, it is now understood that SPCE is related to SPR, a technique widely used for bioaffinity studies. In brief, the SPCE phenomenon arises due to the near-field interaction of excited fluorophores with the metal film. Near-fields around fluorophores have large wave vectors and are thus able to create surface plasmons in the metal film. The surface plasmons in turn radiate into the substrate at definite angles and with interesting polarization properties. An intuitive understanding of SPCE can be obtained from the principles of SPR. In SPR, a thin metal film is illuminated through a prism (Kretschmann, KR configuration). The film is highly reflective, except at a particular angle called the SPR angle. At this angle, the film absorbs incident light due to the creation of surface plasmons in the metal. The emission patterns for SPCE satisfy the same conditions required for observing the minimum reflectivity from the substrates in SPR.1,2,9,10,14,15 We anticipated that this should also be the case for the present MDM substrates. Figure 2 shows the reflectivity curves calculated using the software package, TF Calc, for S101 emission (600 nm) in the Ag-PVA(4%)-Ag and Ag-PVA(8%)-Ag substrates. Indeed, the calculated angle-dependent reflectivity curves are found to be in close agreement with the angular emission patterns of S101 observed with these substrates. More specifically, at 150 nm PVA thickness, the reflectivity is at a minimum for light normal to the surface, which corresponds to the normal emission for 4% PVA shown in Figure 1. Similarly, for 350 nm PVA thickness, there is a reflectivity minimum for light normal to the surface along with s-polarized reflectivity minima at the corresponding emission angles (∼120 and 240°) observed for 8% PVA, as shown in Figure 1.</p><p>It is also useful to compare the emission pattern of the Ag-PVA-Ag, MDM substrate with that observed for a substrate consisting of S101 in 4% PVA above a single 50 nm Ag film (PVA(4%)-Ag). In the latter case, the emission is strictly s-polarized and is observed at an angle of ∼135° (and ∼225° due to symmetry), leading to a cone of emission, with no emission normal to the surface (see Supporting Information, Figure S3). This is in accordance with our previous studies on SPCE and waveguide effects in thick dielectric films.1,2Figure 3 shows the emission spectra of S101 in the Ag-PVA-Ag, MDM substrate and in PVA-Ag. It is interesting to note that the SPCE spectra of S101 in the PVA(4%)-Ag sample shows considerable wavelength dispersion, and the emission appears to be different at the various observation angles. The emission spectra of S101 in the Ag-PVA(4%)-Ag substrate remain unchanged at different observation angles but are narrower toward the red edge. The dispersion effect in the PVA(4%)-Ag sample can be explained by considering the broad emission bandwidth of S101 and the condition that each wavelength in the spectrum couples to the propagating surface plasmons in the metal–dielectric interface only at a defined angle. So, although the emission appears different at different observation angles, the overall SPCE emission spectra obtained by summation of all individual spectra matches with the free space emission of S101 (see Supporting Information, Figure S4). To understand the lack of dispersion and narrowing in the spectra of S101 in Ag-PVA(4%)-Ag, we have to consider the interference effect of light in multilayer structures.</p><p>The Ag-PVA-Ag, MDM structures can be considered to be a planar microcavity consisting of two metallic mirrors separated by a dielectric film, thus forming a Fabry–Perot resonator. Accordingly, the emission properties of a fluorophore embedded within this microcavity will be affected by the geometry of the substrate. The resonant modes of this microcavity will be at wavelengths (λresonance) where the following conditions are satisfied:</p><p>Here m is an integer, n is the refractive index of PVA (n = 1.52), and deffective is the effective cavity thickness. The effective thickness comprises dPVA, the distance between the Ag films determined by the physical thickness of the PVA layer, and dphase-change that arises due to the phase change on reflection at the Ag films.16,17 The first resonance condition (m = 1) for the emission wavelength of 600 nm (corresponding to the emission maximum of S101) should be satisfied at an effective thickness (deffective) of 300 nm. For ideal reflecting mirrors, the phase change on reflection is π, and in this case deffective = ndPVA, so the resonance condition would be observed at a PVA thickness of ∼197 nm. Because the beaming emission is observed for the Ag-PVA(4%)-Ag substrate with PVA thickness (dPVA) of 150 nm that is less than the ideal case, it implies that the phase change due to reflection from the two Ag surfaces in the MDM substrate contributes toward the effective thickness (dphase-change). Below a dielectric thickness of 150 nm, no resonant mode exists for the present Ag-PVA-Ag substrates and hence no emission is observed from S101 for PVA concentrations less than 4% (Figure 1). Furthermore, a dielectric thickness of 150 nm cannot support modes with higher wavelengths. This explains the narrowing in the red-edge of the emission spectra of S101 in the Ag-PVA(4%)-Ag substrate. The second Fabry–Perot resonance mode for the Ag-PVA-Ag substrate appears at a dielectric thickness of 350 nm and matches well with the beaming emission observed in the Ag-PVA(8%)-Ag sample. From the above discussion, it can be easily understood that the emission angle of a fluorophore can be tailored in a relatively straightforward manner by changing the thickness or type of the dielectric layer or by altering the phase changes on reflection at the metallic surfaces. This can be possible either by changing the thickness of the Ag film or even using other metals such as Au or Al depending on the emission wavelength.</p><p>Considering that the MDM substrate corresponds to a planar microcavity, it is expected that the spontaneous emission rate for a fluorophore placed within the microcavity will be increased due to the alteration in the photonic mode density (Purcell effect).18,19 A significant reduction in the fluorescence lifetime of S101 is in fact observed in the Ag-PVA(4%)-Ag substrate (average lifetime ∼3 ns) in comparison with the lifetime of S101 (∼4 ns) in PVA(4%) spin-coated on glass (see Supporting Information, Figure S5). This is a favorable result because a fluorophore that spends less time in the excited state can undergo more excitation emission cycles prior to photodecomposition. Hence, a reduction in the fluorescence lifetime should lead to increased photostability of the fluorophore. The intensity decay is multi- or non-exponential in the MDM substrate. This suggests that the fluorophore– plasmon coupling depends on the location of the fluorophore on the dielectric, the orientation of the emission dipole relative to the planar metal surface, or both.</p><p>To examine further the emission directionality of the MDM substrate, we have studied the emission patterns for another fluorophore, Cy5, that is widely used for biochemical studies. For Cy5 (having emission maximum ∼670 nm), the beaming emission is observed with the Ag-PVA(6%)-Ag substrate, which corresponds to a dielectric thickness of ∼175 nm (see Supporting Information, Figure S6). In this case, the half angle spread of the beaming emission is ∼15° from the normal. It may be recollected that at ∼175 nm dielectric thickness, the emission from S101 is observed at different angles depending on polarization. This result has enormous significance because the emission directionality also carries information about the nature of the fluorophore. In other words, the Ag-PVA-Ag, MDM substrates can be used for multicolor directional fluorescence sensing of multiple probes. This is demonstrated more clearly by recording the angular distribution of emission for a mixture of S101 and Cy5 in the Ag-PVA(6%)-Ag substrate (Figure 4). The emission from Cy5 is observed as a narrow beam normal to the substrate, whereas the emission from S101 is observed at ∼145°. Accordingly, the emission spectrum recorded at 180° shows the usual Cy5 spectral features, whereas the emission spectral features of S101 are clearly evident at 145°. It is important to mention that the changes in the angular dependence of the emission intensities with change in the emission wavelengths is more pronounced for the present MDM substrates in comparison with the usual SPCE substrates with a single metal layer.1</p><p>The observation of beaming emission from the MDM substrates opens up a broad subject area, and numerous fluorescence-based applications can be envisaged. Figure 5 shows a schematic representation and real color photographs of the images formed by projecting the emission from S101 embedded in the Ag-PVA-Ag and PVA-Ag substrates on a screen. The PVA-Ag substrate generates a ring of fluorescence for each excitation spot, whereas the PVA-Ag substrate generates a ring of fluorescence for each excitation spot, the Ag-PVA-Ag substrate shows distinct fluorescence spots corresponding to each excitation. This Figure thus serves to demonstrate the ease with which the present MDM substrates can be adapted for fluorescence studies in microarray formats (also see Supporting Information, Figure S7). Fluorescence microarray technology is a powerful tool for high throughput bioanalytical studies due to its miniaturization, low material consumption, ability for multiplexing, and automation.20 We believe that the high spatial control and beaming emission from the MDM substrates will permit focusing onto an imaging detector with simple and inexpensive optics and will reduce crosstalk between adjacent spots, thus improving data quality.</p><p>One limitation of the MDM substrates is that the fluorophores are embedded within the dielectric between the Ag layers and are not amenable for dynamic studies or binding interactions. For bioanalytical applications it is important to be able to access the analyte surface so that chemistry can be performed. Recently, there have been studies of energy transfer (ET) across a metal film mediated by surface plasmon polaritons (SPPs).21,22 The mechanism of SPP-ET is different from conventional nonradiative Forster ET and involves the coupling of donor emission to the surface plasmon modes of the first metal–dielectric interface, followed by cross coupling of the SPPs on opposite sides of the metal film and ET to the acceptor molecules on the second metal–dielectric interface. These concepts can be utilized in the present MDM substrates for bioassays or bioanalytical applications. A schematic representation of SPP-ET with the MDM substrates is presented in Figure 6. The topmost Ag layer can be coated with biomarkers capable of attaching to specific fluorescence donor labeled biomolecules. The dielectric layer between the metals can contain the fluorescence acceptor. Under suitable binding conditions, SPP-ET will lead to the observation of directional emission from the embedded acceptor molecules. This configuration will thus allow molecule-specific biosensing, with the advantage of a high degree of spatial control over the fluorescence emission. To demonstrate the feasibility of this concept, we have studied the SPP-ET from the donor, tris-(8-hydrxyquinoline)aluminum (AlQ3) spin coated on the top Ag layer, to the acceptor, rhodamine6G (Rh6G) embedded inside the substrate. In these substrates, the donor, AlQ3, is spin coated above the top Ag layer using a poly(methyl methacrylate) (PMMA) solution in chloroform. An increase in the fluorescence intensity of Rh6G accompanied by a decrease in the fluorescence intensity of AlQ3 is observed for the donor–acceptor samples. This suggests the occurrence of SPP-ET from AlQ3 to Rh6G. A quantitative estimate of the SPP-ET efficiency (FET) can be determined from the area under the donor–acceptor (IDA) and acceptor only (ID) curves, as FET ≈ (IDA – IA)/IDA. 21,22 For the present MDM substrates, the fraction of the total donor–acceptor sample emission attributable to SPP-ET is estimated to be ∼0.35. A more rigorous test for the ET process is to examine the intensity decays of the donor and acceptor samples. A long lifetime component, corresponding to that of AlQ3, is observed for Rh6G in the presence of the donor, which confirms SPP-ET from AlQ3 to Rh6G (see Supporting Information, Table S1).</p><p>In summary, the present study demonstrates that MDM substrates can channel the fluorescence into a narrow beam and allow greater control over the emission directionality, which is highly desirable for improved fluorescence detection. These structures also allow us to modify the fluorescence lifetimes and polarization of the emission. Importantly, the MDM substrates can be easily and reproducibly fabricated using standard thermal evaporation techniques. So, the fluorescence from different dyes can be conveniently tuned to suit our needs by changing the nature and thickness of the metal or the dielectric medium. The MDM substrates are promising for both fundamental light matter interactions and device applications. We believe that many novel fluorescence based studies can be designed with the present MDM plasmonic substrates.</p>

PubMed Author Manuscript

Fracture Behavior of Mullite Reticulated Porous Ceramics for Porous Media Combustion

Mullite reticulated porous ceramics (RPC) are one of the key components for porous media burner, the mechanical properties of mullite RPC decided the service life of the burner. However, the irregularities of cellular structure made it difficult to reveal the fracture behavior of mullite RPCs. In this study, the three-dimensional (3-D) structures of mullite RPCs were analyzed by X-ray computed tomography. The strength and damage behavior of mullite RPCs were respectively investigated via the compression tests and finite element modeling based on the actual 3-D model, also the corresponding strengthening mechanism was proposed. The results indicated that the reconstructed 3-D model exhibited the real microstructure of mullite RPCs, containing the hollow struts and strut defects. The Young's modulus calculated from actual 3-D structures was lower than that from Gibson-Ashby theory. In addition, the surface defects preceded triangular tips to generate the area of stress concentration, leading to the fracture behavior first occurred at the strut defects. With the formation of dense strut in mullite RPCs, the stress uniformly distributed in the whole solid skeleton, thus significantly improving the damage resistance of mullite RPCs.

fracture_behavior_of_mullite_reticulated_porous_ceramics_for_porous_media_combustion

Introduction<!>Preparation of Mullite RPCs<!>RPCs Characterization<!>Mechanical Characterization<!>Microstructure and Physical Properties of Mullite RPCs<!><!>Microstructure and Physical Properties of Mullite RPCs<!><!>Microstructure and Physical Properties of Mullite RPCs<!><!>X-ray Computed Tomography<!><!>Model Reconstruction<!><!>Model Reconstruction<!>Effective Elastic Modulus<!><!>Effective Elastic Modulus<!><!>Stress Distribution<!><!>Stress Distribution<!><!>Stress Distribution<!>Conclusions<!><!>Data Availability Statement<!>Author Contributions<!>Conflict of Interest

<p>With the increasing of the energy crisis and consciousness of environmental issue, porous media combustion has attracted people's attention due to its low pollutant emission, high powder density and combustion efficiency (Trimis and Durst, 1996; Wood and Harris, 2008; Yu et al., 2013). Because of the heat recirculation from the burned hot downstream gas to the unburned ones within the burners, the premixed gases can be preheated, thereby the superadiabatic flame or excess enthalpy flame will be yielded (Wu et al., 2015). This combustion technique is being widely used in the field of clean recovery of low calorific gas (Mujeebu et al., 2009). Mullite reticulated porous ceramics (RPCs) with their own characteristics of open cells, three-dimensional networks structure, lower thermal expansion coefficient, higher thermal radiation, and economical to produce are considered as one of the most promising porous media for high-temperature burners (Prochazka and Klug, 2010; Maurath and Willenbacher, 2017; Sandoval et al., 2019). As the key component of porous burner, mullite RPCs must subject to severe thermal stress during the switch on/off of the burner (Pickenäcker et al., 1999). Therefore, the improved mechanical performance of the porous media is critical for their applications.</p><p>Generally, RPCs are prepared by template replication with ceramic slurry. This method commonly used the polymer sponge as porous template, and coated with a thixotropic ceramic slurry. After sintering, the RPCs had the same morphology to the original sponge, which exhibited the open cell and three-dimensional structure (Brockmeyer et al., 1990). However, with the removal of polymeric sponge at elevated temperature, the strut defects, such as surface cracks, hollow struts, and their inside tricuspid tips will generate in RPCs. These flaws affect the mechanical properties of RPCs, thereby weakening the life of porous burner (Vogt et al., 2010). Several approaches have been proposed to strengthen the porous ceramics. For instance, the coherence ability between ceramic slurries and sponge was improved via pretreatment of mechanical stretching or sodium hydroxide, thereby increasing the weight of the coated slurry (Yao et al., 2005). Furthermore, the recoating process was developed to eliminate the surface defects of RPCs, and correspondingly thicken the ceramic skeleton (Pu et al., 2004). The novel template with carbon black slurry coated sponge was used to reduce the stress concentration within the hollow struts of RPCs (Jun et al., 2006). Besides, the liquid silicon was immersed into the SiC skeleton to fill up the hollow strut (Fuessel et al., 2011; Ortona et al., 2012). Although the above methods on increasing the strength of RPCs, their fracture behavior and strengthening mechanism are far from being well-understood. The irregular pores and strut defects within the hollow struts lead to the low elastic limit of RPCs because the cracks will appear before the detectable elastic deformation, which are failed to assess the fracture mode and the elastic modulus of the RPCs using the experimental methods (Oliveira et al., 2006; D'Angelo et al., 2012).</p><p>Up till now, plenty of simulation approaches are performed to reveal the elastic properties of RPCs. In general, the space-filling polyhedron models including cubic model, Kelvin model and Weaire-Phelan model are widely used for the simulation of porous foams (Gibson and Ashby, 1997; Zhu et al., 1997; Buffel et al., 2014). These models are composed of regular polyhedrons, which are constructed with classic beam shell theory. Meanwhile, the cell struts with circular, triangular, and plateau border cross sectional shapes of Kelvin models were modified to simulate the real structure of foams (Gong et al., 2005; Jang et al., 2008). However, the microstructures of porous ceramics are much more complex than those of the above models, such as the varied pore morphology, pore size, and strut thickness. In order to increase the akin to the real foams, the tessellation-based models including voronoi tessellation and Laguerre tessellation models are built. Tessellation-based models are capable of incorporating foam microstructural variability and irregularity, such as different cell size, strut thickness variation, strut curvature (Song et al., 2010; Colloca et al., 2012; Chen et al., 2015). However, the microstructural characteristics, like the internal strut micromorphology of hollow struts with triangular voids, surface crack, and defects in RPCs are still too complicated for tessellations to construct.</p><p>It is well-recognized that the global properties of RPCs are dependent on the reticulated structure and strut microstructure of RPCs (Rezaei et al., 2014). The model construction based on the real ceramic struts is the critical to obtain the elastic properties of RPCs. X-ray computed tomography (CT) is usually used to in-situ observe the microstructure inside the materials, which is an effective technique to reconstruct the complicated morphology of porous ceramics (Fischer et al., 2009; Ortona et al., 2010; Chen et al., 2017). In this work, the mullite RPCs with two kinds of strut structures of hollow struts and dense struts are fabricated via polymer sponge replica technique as well as vacuum infiltration of ceramic slurry. The strut microstructure of mullite RPCs were analyzed by means of scanning electron microscope and their 3-D structures were reconstructed via X-ray computed tomography. The effects of strut structures and morphological features on the mechanical properties and fracture behavior of mullite RPCs were investigated based on the compression tests and finite element method. Furthermore, the strengthening mechanism for mullite RPCs are revealed, also the corresponding approaches are proposed.</p><!><p>The commercial available mullite powder (325 mesh, Jiangsu Jinxing Co., Ltd., China) was used as the main raw material. Andalusite (<7.9 μm, Y60, Imerys) and α-Al2O3 (<2.0 μm, Kaifeng Special Refractories Co., Ltd., China) were the sintering aids. The additives used for preparing mullite slurry were polycarboxylate, sodium carboxymethyl-cellulose, contraspum K1012, and ammonium lignosulfonate, whose detailed manufactures were same to the previous paper (Liang et al., 2019). The above additives were firstly dissolved in deionized water to prepare solution. The mixture powders containing 80 wt% mullite, 15.2 wt% andalusite, and 4.8 wt% α-Al2O3 were subsequently added into the solution and stirred for 30 min to produce mullite slurry with solid content of 78 wt%.</p><p>In this experiment, mullite RPCs were fabricated via the sponge replica method. Polyurethane sponge (10 pores per inch, 50 × 50 × 20 mm3, F. M Co. Ltd., Germany) was used as the porous template, then it was coated with the as-prepared mullite slurry via impregnation process. The coated polyurethane sponges were dried at room temperature for 24 h, subsequently treated at 1,500°C for 3 h to prepare mullite RPCs with hollow struts, which named as sample MH.</p><p>Furthermore, in order to obtain mullite RPCs with dense struts, the coated template was first pre-fired at 1,300°C to burn out polymer sponges, thereby producing mullite preforms. Secondly, the infiltration slurry used for the vacuum infiltration process was prepared. The additives of dispersant agent (polycarboxylate, BASF Group, Germany) and antifoam agent (contraspum K1012, Zschimmer & Schwarz, Germany) were dissolved in deionized water by stirring for 5 min. The starting materials of mullite powder, α-Al2O3, and andalusite powder were mixed with the solution. After ball-milling for 3 h, the infiltration slurry was obtained. Finally, the mullite preforms were treated by vacuum infiltration using the same method given in Liang et al. (2016). The as-infiltrated mullite preforms were, respectively, sintered at 1,500°C for 3 h before cooling to room temperature. The sintered mullite RPC were marked as sample MD.</p><!><p>The bulk density of mullite RPC (ρb) was calculated by the ratio of mass to the volume of the whole sample. In order to analyze the changes of strut diameter in mullite RPC before and after vacuum infiltration, the macro image of RPC was taken by a digital camera, and the strut diameter was measured by the means of Image-Pro Plus software (Media Cybernetics, Inc., Netherlands). The apparent density (ρs) of mullite struts were tested using mercury porosimeter (AutoporeIV9500, Micromeritics Instrument Corp., USA). Meanwhile, the relative density of mullite RPC was the ratio of bulk density (ρb) to the strut density (ρs). The micromorphology of the sintered struts and strut surface were observed using scanning electron microscope (SEM, Quanta 400, FEI Company, USA). Furthermore, the internal structure of mullite RPC was analyzed via the scanning of X-ray computer tomography (μ-CT, d2, diondo GmbH, Germany). The device was equipped with a 160 KV X-ray source and a Dexela detector 1512 (PerkinElmer, Germany) with a resolution of 1,944 × 1,526 active pixels. In the tomography scanning process, the device was operated at 120 kV and 50 mA, and the reconstructed voxels had a resolution of 25 μm. Based on the obtained tomography image, VG Studio Max 3.0 (Volume Graphics GmbH, Germany) was used as the visualization software to reconstruct the 3D model.</p><!><p>The strength of mullite RPC was characterized by the cold crushing strength (CCS), which was tested on the universal testing machine (ETM, Wance, China). The method of CCS test is consistent with the (Goretta et al., 1990), six samples in group were tested to get the average strength to represent the CCS of mullite RPC, also a load speed of 0.5 mm/min was applied during the strength measurement. The Young's modulus of mullite RPCs (rectangular sample of 20 × 20 × 100 mm3) were determined by the impulse excitation technique at room temperature (RFDA, Genk, Belgium). The fracture strength of mullite strut was measured as the clod modulus of rupture (CMOR) of the sintered mullite slurries. In addition, the FEA modeling was performed to simulate the effective Young's modulus, Von Mises distribution of the struts during uniaxial compression via ANSYS software.</p><!><p>In order to investigate the strut structure of the prepared mullite RPCs, the micrographs of ruptured struts as well as the strut surface in mullite RPCs are analyzed, shown in Figure 1, respectively. In sample MH, the hollow strut with obvious triangle tips was observed because of the burnt out of polyurethane template (Figure 1A). However, the hollow void within mullite strut was filled up and the triangle tips disappeared in sample MD with the vacuum infiltration of mullite slurry. Meanwhile, mullite skeleton was coated by infiltration slurry, thus forming the dense strut in sample MD (Figure 1B). The coated infiltration slurry on the surface of mullite skeleton could repair the strut cracks and defects in mullite RPCs. It was clearly seen that the large longitudinal cracks and surface defects appeared in sample MH with hollow strut (Figure 1A), while the smooth strut with crack-free showed in sample MD (Figure 1B). Furthermore, the large strut defects were eliminated in the formed dense struts.</p><!><p>SEM micrographs of the cross section of mullite struts (A) Sample MH (B) Sample MD.</p><!><p>The physical properties of mullite RPCs are given in Table 1. It was clearly seen that the strut structure decided the physical properties of mullite RPCs. As for sample MD with dense struts, its bulk density was 0.49 g/cm3, which was much larger than that of sample MH. In mullite RPCs with dense struts, the triangular void within hollow strut was completely filled up, and the strut surface was coated by infiltration slurry. Thus, the increased relative density and the strut diameter obtained in sample MD. Furthermore, the dense strut resulted in a higher strength of mullite RPCs. With the formation of dense struts in mullite RPCs, the CCS of mullite RPCs increased from 0.26 to 0.63 MPa.</p><!><p>Physical properties of mullite RPCs.</p><!><p>The stress-strain curves of mullite RPCs under uniaxial compression are shown in Figure 2. It is apparent that the stress of mullite RPCs increased with the strain, whose curves exhibited the zigzag shape. The stress and strain of mullite RPCs with dense struts were larger than that of sample MH. It was noteworthy that the strains of mullite RPCs were both smaller than 0.08.</p><!><p>The stress-strain curves of mullite RPCs in the uniaxial compression test.</p><!><p>The 3-D reconstructed structures of mullite RPCs consisted of open-cells skeleton and unregular struts are shown in Figure 3. The reconstructed structures could truly characterize the microstructure of mullite RPCs. In the reconstructed X-ray CT slice of sample MH, the hollow struts with surface defects were observed in Figure 4A. In the reconstructed structure of sample MD, the surface defects and triangular voids within the struts disappeared and dense struts formed (Figure 4B), which was same to the microstructure of mullite RPCs (Figure 1). Furthermore, compared with the slice images of sample MH and MD, it could be seen that the strut diameter of sample MD was larger than that of sample MH. The results exhibited the similar trend to the measured ones (Table 1).</p><!><p>The reconstruction image of the 3-D structure of mullite RPCs.</p><p>The slice image of the 3-D structure of mullite RPCs (A) MH, (B) MD.</p><!><p>The mechanical properties of mullite RPCs were modeled using the finite element analysis based on the reconstructed models. In order to reduce the computing time, the typical model with the dimension of 8.8 × 8.8 × 8.8 mm3 was chosen, showed in Figure 5. The effective physical parameters of the solid struts in mullite RPCs were characterized using the slip casted mullite slurry. After firing at 1,500°C, the Young's modulus (E) and Poisson ratio (γ) of the casted mullite slurry were tested by impulse excitation technique. The strength of mullite strut was measured as the clod modulus of rupture (CMOR) of sintered mullite slurries. From the experimental results, the value of E and γ for the simulation of reconstructed models were tested as 38.12 GPa and 0.27, respectively. The fracture strength of solid strut was 38.6 MPa (Table 2). Furthermore, the Young's modulus of mullite RPCs was much lower than that of mullite slurry because of the high porous structure. It was worth noting that the elimination of strut defects was beneficial for the improvement of Young's modulus in mullite RPCs. The Young's modulus of sample MH was 0.10 GPa, while the value increased to 2.26 GPa in sample MD.</p><!><p>Reduced 3-D structure of mullite RPCs for finite element modeling (A) MH, (B) MD.</p><p>Physical parameters of mullite slurry and mullite RPC after firing at 1,500°C.</p><!><p>Three deformation steps of mullite RPCs in the process of uniaxial compression were simulated based on the reconstructed models. The deformation was achieved by applying a uniaxial displacement to all the nodes belonging to two parallel faces of the model, and the other faces were fixed. Furthermore, the applied displacements were 5 × 10−4, 1.0 × 10−3, and 5.0 × 10−3 mm, respectively.</p><!><p>The uniaxial apparent strain along the loading direction was simulated as 0.006, 0.011, and 0.055%, respectively (Table 3). Because the effective Young's modulus of mullite RPCs was determined by the linear elasticity behavior, the smallest strain was chosen to calculate the reactive force in the uniaxial direction when mullite RPCs were under compression. Therefore, the Young's modulus of mullite RPCs were simulated using the actual 3-D models when the strain of 0.006% was applied. The simulated Young's modulus of sample MH and MD were 0.77 and 1.48 GPa, respectively. For comparison, the Young's modulus of the mullite RPCs was also calculated with the Gibson-Ashby model, and the equation was as follows (Fuessel et al., 2011):</p><p>Where ρrel was the relative density of RPCs, E and Es was the effective Young's modulus of mullite RPCs and that for their solid struts (Table 2). Because the RPCs prepared in this work had a complete open cell skeleton, here n was chosen as two and C1 was a proportionality factor close to one (Knackstedt et al., 2005).</p><!><p>Steps in FEA calculation and calculated reaction force and apparent stress.</p><!><p>Table 4 showed the results of effective Young's modulus of mullite RPCs based on Gibson-Ashby model and the actual 3-D structure. It was clearly seen that the Young's modulus calculated by the finite element analysis was lower than that from Gibson-Ashby model. Because the reconstructed models from X-ray tomography exhibited the real microstructure of mullite RPCs, the strut defects and irregular pores were included (Figure 4). However, Gibson and Ashby model consumed that porous materials were ideal matrix, which consisted of the equiaxed open-cell cubic shaped pores.</p><!><p>The calculated Young's modulus from finite element analysis and Gibson-Ashby theory.</p><!><p>The varied strut structures of mullite RPCs led to the difference in stress distribution. In order to reveal the strengthening mechanism of RPCs, the von Mises stress of mullite RPCs with hollow strut and dense strut were calculated during the uniaxial compression. The damage behavior of mullite RPCs was evaluated via the comparison of the von Mises stress to the flexural strength (CMOR) of the solid mullite strut (Table 2). When the von Mises stress was larger than flexural strength, the mullite strut was simulated to be fracture.</p><p>Figure 6 shows the von Mises stress of model MH under the applied strains. With the application of a small strain of 0.006%, the stress concentration only appeared at the surface defects of mullite struts, which indicated that the linear elastic behavior occurred (Figure 6A). It verified the validity of the effective Young's modulus calculated by the above finite element method. As the increase of strain to 0.011%, the area of stress concentration enlarged and the value of maximum stress increased correspondingly. Meanwhile, the stress concentration with small areas began to be detected at the triangular tips within the hollow struts (Figure 6B). Under the strain of 0.055%, the large area of damage zone appeared near the strut defects (red regin). Furthermore, the remarkable stress concentration existed at the triangular tips inside the hollow struts (Figure 6C).It was obviously seen that the stress concentration was more likely to form at the surface defects of struts than that at the inner triangular tips in mullite RPCs with hollow struts. In addition, the fracture first occurred at the strut defects during the uniaxial compression. As further increasing the applied strain, the defect-free struts and the struts with larger diameter began to fracture, which resulted in the stress-strain curve with the zigzag shape (Figure 2).</p><!><p>The Von Mises stress and fracture in reduced 3-D structure reconstruction model MH (A) at the strain of 0.006%, (B) at the strain of 0.011%, and (C) at the strain of 0.055%.</p><!><p>The von Mises stress of mullite RPCs with dense struts under the uniaxial compression are presented in Figure 7. It was noteworthy seen that the triangular voids within the struts were completely filled up, also the surface defects disappeared from the reconstructed model MD. When the strain was 0.006%, the stress uniformly distributed in the whole mullite skeleton, and the maximum stress was <10 MPa (Figure 7A). Under the strain of 0.011%, the whole stress was nearly unchanged, only local stress appeared at the strut with small area (Figure 7B). With the strain increasing to 0.055%, the stress increased among the whole struts, especially in the area of finer struts. However, the fracture behavior of mullite RPCs did not occur even under the largest strain (Figure 7C). With the formation of dense struts in mullite RPCs, the stress concentration of mullite RPCs disappeared, thus resulting in the enhanced compression resistance.</p><!><p>The Von Mises stress and fracture in reduced 3-D structure reconstruction model MD (A) at the strain of 0.006%, (B) at the strain of 0.011%, and (C) at the strain of 0.055%.</p><!><p>It can be seen from Figures 6, 7, the whole structures were not completely failed in model MH and MD when the applied strain was 0.055%. Therefore, the stresses of sample MH and MD could be calculated as 0.42 and 0.81 MPa based on the effective Young's modulus of mullite RPCs from FEA. However, the measured strength of sample MH and MD were 0.29 and 0.63 MPa respectively, which were much lower than the simulated ones. The stress difference between FEA results and experimental tests might be related to the uneven surface of mullite RPCs. During the compression test, the non-uniform loading induced by the rough surface of mullite RPCs caused the stress concentration, thus seriously reducing the strength of RPCs. From the FEA results, the elastic deformation occurred when the strain was much <0.055%. Therefore, the loading displacement should be <1.1 μm during the testing process, which was difficult to be achieved in the actual compression test. The stress concentration induced by the rough surface led to the lower strength of mullite RPCs than the simulation ones, which was unavoidable.</p><!><p>The strut microstructure and mechanical properties of mullite RPCs were, respectively, investigated via compression tests and finite element analysis based on the reconstructed 3-D models, also the fracture behavior of mullite PRCs was discussed. The following conclusions could be summarized:</p><!><p>The reconstructed 3-D models from XCT exhibited the real strut microstructure of mullite RPCs, containing the hollow struts and strut defects. The Young's modulus calculated from the finite element analysis with actual 3-D structures was lower than that from Gibson-Ashby model.</p><p>The surface defects and triangular tips within hollow struts led to the stress concentration of mullite RPCs during uniaxial compression. The surface defects preceded triangular tips to generate the area of stress concentration, resulting in the fracture behavior first occurred at the strut defects. With the formation of dense strut in mullite RPCs, the hollow strut was completely filled up and the strut defects were eliminated. The dense struts made the stress uniformly distributed in the whole solid skeleton, thus significantly improved the damage resistance of mullite RPCs.</p><!><p>All datasets generated for this study are included in the article/supplementary material.</p><!><p>XL, YL, TZ, and QW contributed conception and design of the study. LP and BL organized and analyzed the database. CA revised the manuscript. All authors contributed to manuscript revision, read, and approved the submitted version.</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

Templated‐Assembly of CsPbBr3 Perovskite Nanocrystals into 2D Photonic Supercrystals with Amplified Spontaneous Emission

AbstractPerovskite nanocrystals (NCs) have revolutionized optoelectronic devices because of their versatile optical properties. However, controlling and extending these functionalities often requires a light‐management strategy involving additional processing steps. Herein, we introduce a simple approach to shape perovskite nanocrystals (NC) into photonic architectures that provide light management by directly shaping the active material. Pre‐patterned polydimethylsiloxane (PDMS) templates are used for the template‐induced self‐assembly of 10 nm CsPbBr3 perovskite NC colloids into large area (1 cm2) 2D photonic crystals with tunable lattice spacing, ranging from 400 nm up to several microns. The photonic crystal arrangement facilitates efficient light coupling to the nanocrystal layer, thereby increasing the electric field intensity within the perovskite film. As a result, CsPbBr3 2D photonic crystals show amplified spontaneous emission (ASE) under lower optical excitation fluences in the near‐IR, in contrast to equivalent flat NC films prepared using the same colloidal ink. This improvement is attributed to the enhanced multi‐photon absorption caused by light trapping in the photonic crystal.

templated‐assembly_of_cspbbr3_perovskite_nanocrystals_into_2d_photonic_supercrystals_with_amplified_

<!>Introduction<!>Results and Discussion<!><!>Results and Discussion<!><!>Results and Discussion<!><!>Results and Discussion<!><!>Conclusion<!>Conflict of interest<!>

<p>D. Vila-Liarte, M. W. Feil, A. Manzi, J. L. Garcia-Pomar, H. Huang, M. Döblinger, L. M Liz-Marzán, J. Feldmann, L. Polavarapu, A. Mihi, Angew. Chem. Int. Ed. 2020, 59, 17750.</p><!><p>Halide perovskite nanocrystals (NCs)1 have emerged as a new class of efficient and tunable light sources for various optical and photonic applications and have already shown great promise in solar cells, photodetectors, LEDs and lasers.1b, 2 The optical properties of halide perovskite NCs are readily tunable across the entire visible light spectrum, by means of their halide composition as well as their dimensionality.1c, 3 We have witnessed tremendous research progress regarding the shape and composition‐controlled synthesis of perovskite NCs for tailoring their emission color, as well as improving quantum efficiency.3f, 4 Among all available compositions, Pb‐based perovskite NCs stand out for their stability and optical efficiency, even as a two‐photon‐pumped lasing medium.5 Because of the interest in perovskite nanocrystals as light sources, a light management strategy is often required. An appealing strategy to engineer the optical response of the material without recurring to changes in its composition comprises molding it into higher order architectures. This approach has been previously applied to colloidal gold nanocrystals, self‐assembled into 2D plasmonic supercrystals (SCs) exhibiting tunable lattice plasmon resonances, but also to PbTe nanocrystals, for which the arrangement was modulated into a variety of motifs.6 Self‐assembly concepts have also been recently extended to colloidal perovskite NCs,7 including the formation of colloidal SCs with distinct optical properties.4b, 8 Despite these early efforts, the self‐assembly of perovskite NCs into 2D SC arrays with subwavelength patterns remains a daunting challenge, and therefore their optical properties remain unexplored.</p><p>Photonic crystals (PhCs) are materials exhibiting a periodicity in the refractive index on the order of the wavelength of incident light.9 Light propagation in these materials can be described in terms of photonic bands and band gaps (forbidden intervals). PhCs, however, also exhibit many other interesting phenomena. 2D photonic crystals, very much as other gratings, can diffract light very efficiently, which has been used e.g., to enhance the optical path in solar cells.10 Many materials can be organized into photonic crystals by periodically arranging them in the sub‐micrometer range and thereby exhibit optical properties differing from those of the bulk material.6b, 11 In this work, we propose a simple and scalable method to fabricate large area 2D photonic crystals using colloidal solutions of perovskite nanocubes. We used template‐induced self‐assembly to produce large‐area arrays from colloidal perovskite NC dispersions. The obtained perovskite 2D PhCs have lattice parameters on the order of visible light wavelengths, with a strong iridescence indicating the high quality of the structure. Furthermore, the photonic architecture is engineered to couple the 800 nm pumping light into photonic modes propagating in the perovskite PhC layer. This enhanced electric field intensity in the material, as revealed by FDTD simulations, is used in turn to produce amplified spontaneous emission (ASE) in PhC films under fluence that is too low to obtain ASE from planar films obtained under similar conditions. The combination of the NCs optical properties with the photonic environment provides an efficient use of optical pumping to boost the light intensity‐dependent mechanism of two‐photon absorption.</p><!><p>The colloidal CsPbBr3 nanocube ink used for the fabrication of 2D photonic crystals was prepared by ligand‐assisted ultrasonication (see experimental section in the Supporting Information SI). The synthesized NCs are highly monodisperse, with an average size of 10.7±0.9 nm (Figure S1). The colloidal dispersion was purified by centrifugation to obtain a perovskite NC ink in hexane at a 100 mg mL−1 concentration. The perovskite NCs exhibited green photoluminescence (PL) with a maximum at 513 nm (Figure S1). Figure 1 a illustrates the self‐assembly process using poly(dimethylsiloxane) (PDMS) templates12 with pre‐patterned square arrays (see AFM image of the PDMS surface in Figure S2).</p><!><p>a) Schematic illustration of PDMS template‐assisted fabrication of 2D photonic supercrystals made of perovskite NCs. b),c) Representative SEM images of 2D photonic supercrystals made of CsPbBr3 arrays with lattice parameters of 600 nm and 1700 nm, respectively. The inset in (b) is a photograph of different CsPbBr3 SC arrays on glass substrates, displaying diffraction under white light illumination.</p><!><p>The modified template‐assisted self‐assembly process is illustrated in Figure 1. In the first step, a drop (10 μL) of CsPbBr3 colloidal dispersion is deposited on a glass substrate and immediately covered with a PDMS template with holes facing the droplet (Figure 1 a‐I). The resulting confinement forces the colloidal dispersion into the holes of the PDMS template, from where the solvent is left to evaporate for 1 hour (Figure 1 a‐II). Homogeneous distribution of the dispersion and conformal contact between the substrate and the PDMS mold was ensured by placing a 700 g weight on top of the stamp during the entire drying step (see experimental section in SI). Finally, the PDMS mold is gently lifted‐off from the substrate, yielding an array of NCs with the negative image of the template pattern on the substrate (Figure 1 a‐III). The obtained PhC films exhibit strong green PL under UV light illumination (see left inset in Figure 4 b below) and appear iridescent under white light illumination as shown in the inset of Figure 1 b. The diffraction of white light into different colors originates from the subwavelength periodicity of the NC assemblies on the substrate. The 2D perovskite photonic crystals were further characterized by scanning electron microscopy (SEM). SEM images of the films fabricated using a PDMS template with a lattice parameter (L) of 600 nm clearly show large‐area 2D arrays of NC assemblies (Figure 1 b). Interestingly, the CsPbBr3 NCs in the assemblies exhibit cubic close‐packing into an ordered supercrystal, as revealed by high magnification SEM images (Figure 1 b). The lateral dimension of the supercrystal is ≈200 nm. These supercrystals are formed during evaporation of the solvent inside the holes of the PDMS template. However, parameters such as ligand concentration and solvent composition should be investigated further to better understand the mechanism of supercrystal formation under the template assembly.6b</p><p>Perovskite photonic crystals with different lattice parameters were fabricated by different choice of PDMS molds. Optimization of the system allowed us to successfully assemble the NCs into structures with L ranging from 1700 nm down to 400 nm, thereby overcoming the increasing difficulty involved in pattern miniaturization. SEM images of PhCs with different lattice parameters are shown in Figure S3. Bright field optical microscopy overview images of the samples confirm the formation of uniform and large‐area 2D PhCs (Figure S4). Shown in Figure 1 c is a representative SEM image of a perovskite 2D PhC with a lattice parameter of 1700 nm, prepared on a glass substrate using the corresponding PDMS template. In this particular array, high magnification SEM reveals that the NCs are closely packed within individual SCs (Figure 1 c). In order to better visualize the individual CsPbBr3 NCs, we assembled the same NCs over carbon‐coated grids and analyzed them by transmission electron microscopy (TEM, Figure S5). TEM images of these assemblies and their 2D fast Fourier transform (FFT) analysis indeed confirm the formation of self‐assembled perovskite supercrystal arrays (Figure S5). In addition, the assemblies can be seen on the substrate as well after a thorough washing of the surface with methylacetate or acetone (Figure S6). However, this process partially destroys the NCs and the assemblies.</p><p>In addition to the dimensions of the template, we found that both concentration and viscosity of the CsPbBr3 colloidal dispersion play a critical role on the final structure of the PhC. We fabricated a series of photonic crystals with fixed geometrical parameters and varying concentration of NCs (Figure 2). We observed that isolated SC pillars were effectively achieved with concentrations below 10 mg mL−1, whereas increasing the concentration of nanocrystals resulted in residual nanocubes connecting the pillars. For the highest NC concentrations (80–100 mg mL−1) the final photonic architecture was composed of an array of CsPbBr3 nanopillars on top of a residual layer of nanocubes. The height of the pillars is in the range of 80–100 nm (Figure S6). This latter configuration is particularly useful in the case of light emission studies (amplified spontaneous emission and lasing), where thick layers of active material are required. We focus now on the optical properties of perovskite 2D photonic crystals with lattice parameters of 400, 500 and 600 nm, produced from 100 mg mL−1 dispersions. These PhCs consist of arrays of 100 nm high pillars, on top of a 250 nm flat film of NCs, as schematically illustrated in Figure 3 c. This film thickness is required to observe a meaningful photoluminescence signal from the sample. We first inspect the extinction (1—transmittance) from the PhCs and compare it to the extinction from a flat NC film (Figure 3 a).</p><!><p>SEM characterization of photonic crystals prepared by self‐assembly of NCs from dispersions at concentrations of a)–c) 100 d)–f) 50 and g)–i) 10 mg mL−1. The concentration of NCs and ligands has a strong influence on the viscosity of the sample, altering the distribution of NCs along the surface. Isolated pillars are formed at low concentrations, while a residual layer of NCs is formed all over the sample at 100 mg mL−1.</p><p>Lattice resonances and field enhancement: Experimental (a) and theoretical (b) extinction (1−T, where T is transmittance) spectra for a NC thin film (green curve) and for different PhCs with lattice parameters of 400 (blue), 500 (black) and 600 nm (red). The vertical dashed line in (a) corresponds to the 800 nm laser excitation wavelength used in the nonlinear experiments. c)–f) FDTD simulated spatial distribution of the electric field enhancement illustrating the effect of light diffraction, for the case of 600 nm lattice parameter versus a perovskite flat film (e,f).</p><!><p>The flat reference sample shows an excitonic peak at 512 nm, while PhCs with lattice parameters of 400, 500 and 600 nm exhibit excitonic peaks at 515, 519 and 527 nm, respectively. However, a series of new peaks appear for the patterned samples, which redshift with increasing lattice spacing. These features are Rayleigh anomalies associated to the offset of diffraction. These peaks indicate that light impinging normally to the sample is diffracted by the periodicity of the array.6b We carried out finite difference time domain (FDTD) simulations from Lumerical Solutions Inc., to theoretically obtain the corresponding extinction curves (Figure 3 b) and to extract the spatial distribution of 800 nm incident light on a 600 nm lattice parameter perovskite PhC, versus a flat film (see numerical details in SI). The field profiles reveal that the electric field within the photonic crystal is 7.5 times higher than that in a flat film of the same thickness (Figure 3 c–f). Effectively, a perovskite PhC diffracts the incoming light and couples it to the perovskite residual layer under the pillars. This light trapping strategy is typically used in photovoltaics to couple light more efficiently to the active layer of a solar cell, by increasing the optical path of light in the material.10 Similarly, this scheme has been exploited in the literature to enhance the two‐photon absorption (TPA) of light emitters, a third order process whose absorption cross section scales non‐linearly with the intensity of the incoming light (see numerical details in SI).13 However, unlike previous reports in which quantum dot‐decorated PhCs14 were fabricated by costly lithographic techniques such as electron beam lithography (EBL), our approach readily produces a high quality PhC configuration within minutes. We thus applied this strategy to optimize the use of the 800 nm pumping light and look for an improved PL enhancement, following recent studies on bulk‐perovskite metasurfaces that showed enhanced linear and nonlinear PL.2a, 11</p><p>We selected the perovskite PhCs fabricated from 100 mg mL−1 dispersions (Figure 4 a) for PL characterization, since these resulted in the thicker architectures reproducing the light trapping scheme shown in Figure 3 c. We collected PL under 400 nm excitation from the PhC, as well as from a flat region outside of the patterned area, and compared it with the PL from the NC dispersion (Figure 4 b). The NCs in dispersion have a PL centered at 513 nm, with a Gaussian FWHM of 21 nm. Instead, both the NCs in the PhC pattern and in the flat reference region exhibited a spectrally redshifted emission with respect to the PL of the NC dispersion, respectively at 528 nm (FWHM: 23 nm) and 524 nm (FWHM: 23 nm). Importantly, this redshift is not only observed in the emission but also in the extinction spectra (Figure 3 a), as a redshift of the 1s excitonic transition. Such a redshift is known to stem from self‐assembly of NCs into colloidal superlattices.8, 15 The physical origin of the redshifted PL stems from the close vicinity of neighboring NCs, which leads to overlap of their electronic wavefunctions. This coupling phenomenon between NCs leads to the formation of minibands, both in the conduction and in the valence bands. The split energy levels result in a lower energy for excitonic absorption and recombination (i.e., the observed redshifts).8a In our case, the redshifted absorption and emission from the PhCs underlines a higher degree of electronic coupling in such structures, with respect to the NCs in the reference flat area. Therefore, the SCs obtained inside the holes of the PDMS templates are of higher quality than those spontaneously formed in the unpatterned area, which evidences the potential versatility of this fabrication technique.</p><!><p>a) SEM close view of SC arrangement in a perovskite PhC obtained from a 100 mg mL−1 NC dispersion, used for ASE studies. b) Photoluminescence spectra obtained from inside the patterned area (red), outside of the patterned area (blue) and from the initial colloidal dispersion (green).</p><!><p>Finally, to probe the effect of the field enhancement obtained by the presence of PhCs on the light emitting properties of perovskite NCs, we photoexcited the system with an 800 nm pulsed laser system, with a repetition rate of 1 kHz and a pulse width of 100 fs. This excitation wavelength corresponds to the electric field spatial distribution observed for the PhCs and is attributed to the increase in light diffraction, which is absent for unpatterned NCs (see Figure 3 a). CsPbBr3 NC films have been shown to exhibit below band gap absorption and nonlinear amplified spontaneous emission (ASE).16 We therefore explored the possibility to obtain ASE through nonlinear photon absorption in our PhC pattern, varying the intensity of the photoexcitation and comparing the observation to the ASE obtained from unpatterned NCs. ASE was reproducibly observed from our PhC pattern (see Figure 5 a), as evidenced by the emergence of an additional narrow emission peak on the low‐energy side of the PL spectra. By integrating the PL intensity versus the pump fluence we obtained an ASE threshold intensity of 10.9 mJ cm−2 (Figure 5 b). Employing an unpatterned film from the same NC ink as a reference, we observed no ASE, even for the maximum excitation fluence of 13 mJ cm−2 (Figure S7). These results suggest that 2D perovskite PhCs make more efficient use of the excitation light by coupling to photonic modes with more intense field profiles at 800 nm, as predicted from FDTD simulations (Figure 3 e). This results in a decrease of the photoexcitation intensity necessary to achieve population inversion, hence ASE. This finding indicates how PhCs can be employed to tailor the field enhancement effects in optoelectronic devices based on perovskite NCs. We envision that the optimized use of the optical pumping provided by the ordered nanostructuring of the NC ink not only broadens the range of applicability of this material beyond its visible band gap, but also will be particularly beneficial to light emitting applications under NIR excitation.</p><!><p>a) Fluence‐dependent PL (evolution of ASE) of CsPbBr3 2D photonic crystals film and b) the corresponding emission intensity obtained by integration of the spectra. The inset in (a) is a photograph of the photonic crystal film showing waveguiding above excitation threshold for ASE. The dotted lines in (b) are the best linear fits for intensities below (green) and above (red) threshold. The spectra were acquired with 800 nm femtosecond laser excitation.</p><!><p>In conclusion, we have demonstrated the fabrication of large‐area CsPbBr3 2D photonic crystals by template‐assisted self‐assembly, using 10 nm nanocrystal inks. We produced homogeneous photonic architectures with lattice parameters ranging from 400 nm to 1700 nm. The PhCs obtained exhibit optical resonances in the extinction spectra under normal incidence produced by the onset of diffraction from the photonic crystal geometry. These resonances can be tuned through the choice of lattice parameter (L) in the architecture. We reproduced the experimental spectra using FDTD simulations and investigated the field distribution in the structure, revealing a strong field enhancement at below band‐gap energies. Furthermore, we have taken advantage of this light trapping scheme to enhance the two‐photon absorption of NIR light and obtained ASE. We anticipate that these results will have broader impact in the field of perovskites beyond ASE and open new opportunities to tailor the optical properties of this novel material at the nanoscale.</p><!><p>The authors declare no conflict of interest.</p><!><p>As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.</p><p>Supplementary</p><p>Click here for additional data file.</p>

PubMed Open Access

Modulating NHC catalysis with fluorine

Fluorination often confers a range of advantages in modulating the conformation and reactivity of small molecule organocatalysts. By strategically introducing fluorine substituents, as part of a β-fluoroamine motif, in a triazolium pre-catalyst, it was possible to modulate the behaviour of the corresponding N-heterocyclic carbene (NHC) with minimal steric alterations to the catalyst core. In this study, the effect of hydrogen to fluorine substitution was evaluated as part of a molecular editing study. X-ray crystallographic analyses of a number of derivatives are presented and the conformations are discussed. Upon deprotonation, the fluorinated triazolium salts generate catalytically active N-heterocyclic carbenes, which can then participate in the enantioselective Steglich rearrangement of oxazolyl carbonates to C-carboxyazlactones (e.r. up to 87.0:13.0).

modulating_nhc_catalysis_with_fluorine

Introduction<!>Results and Discussion<!>Conclusion<!>Experimental

<p>Molecular editing using fluorine is a powerful strategy to modulate the conformation and reactivity of small molecule organocatalysts [1][2][3]. The negligible steric penalty associated with H→F substitution, together with the polarised nature and stability of aliphatic C-F bonds, render this unit attractive from the perspective of molecular design [4]. The low-lying antibonding orbital (σ C-F *) can interact with an array of vicinal substituents ranging from non-bonding electron pairs, such as in the case of the fluorine anomeric effect [5], to electron rich sigma bonds (σ→σ*). The stereoelectronic gauche effect in 1,2difluoroethane is the most prominent example (1; Figure 1) [6][7][8][9]. The counterintuitive preference of vicinal fluorine substituents to adopt a gauche preference (Φ F-C-C-F ≈ 60°) can be rationalised by invoking two stabilising hyperconjugative interactions (σ C-H →σ C-F *). This conformational preference is conserved in numerous systems in which one of the fluorine atoms has been substituted by another electron withdrawing group (X (δ+) ; X (δ+) -C α -C β -F δ− ). Often this modification leads to the introduction of a stabilising electrostatic component, thus enhancing the interaction: this is exemplified by the pioneering work of O'Hagan and co-workers [10][11][12].</p><p>In recent years, this laboratory has strategically employed the aforementioned effects in the design of functional small mole- [18]. Only the synclinal-exo arrangement of 6 and 7 is shown [22]. cules [13][14][15][16][17][18][19][20][21][22], often for application in organocatalysis [1]. Common to these studies has been the strategic incorporation of a fluoro substituent vicinal to a catalytically active amino group. Subsequent generation of a (partial) positive charge at nitrogen generates the requisite X-C α -C β -F δ− system (X = N + ), thus providing a facile approach to controlling rotation around this bond (Φ XCCF ≈ 60°). In this study, the influence of fluorination on catalyst behaviour is extended to the study of triazolium salts such as 2, which can be converted to the respective N-heterocyclic carbenes (NHCs) by simple deprotonation.</p><p>Given the importance of NHCs in modern organic synthesis [23][24][25][26][27][28] it was envisaged that these systems would be intriguing candidates for investigation. Moreover, structural information gleaned from the triazolium salt pre-catalysts regarding con-formation [18,22], assist in rationalising the behaviour of the NHCs generated in situ.</p><p>Herein, the synthesis and catalytic efficiency of a series of fluorinated, bicyclic triazolium salts 2 is disclosed. The effect of molecular editing by hydrogen to fluorine substitution is evaluated in the NHC-catalysed, enantioselective Steglich rearrangement of oxazolyl carbonates 3 to C-carboxyazlactones 4 [29], recently reported by Smith and co-workers [30][31][32][33][34][35][36].</p><p>Fluorination sites were selected based on their proximity to the ring junction nitrogen of the triazolium system (Figure 2). Consequently, two distinct β-fluoroamine sub-classes may be generated. The first site positions the β-fluorine atom on a freely rotatable (sp 3 -sp 3 ) exo cyclic group (5, 6 and 7), conceivably allowing for both synclinal-exo and synclinal-endo conformations to be populated: this is consistent with the recently reported fluorine-NHC gauche effect [22]. The second fluorination site embeds the β-fluoroamine within the bicycle framework of the triazolium salt, thus restricting conformational freedom (e.g. 8). This later scenario was inspired by the elegant work of Rovis and co-workers, which demonstrated that backbone fluorination of bicyclic NHCs improves enantioselectivity in Stetter reactions of heterocyclic aldehydes with nitroalkenes [37][38][39][40]. Finally, one hybrid system was prepared containing both β-fluoroamine classes (7). The trifluoromethylated triazolium salt 9 and the non-fluorinated equivalent 10 served as electronic and steric control catalysts for this study.</p><!><p>Pre-catalyst synthesis Scheme 1: Synthesis of the difluorinated triazolium salt 7 starting from commercially available N-Boc-trans-4-hydroxy-L-proline methyl ester (11).</p><p>the preparation of triazolium salts 5 and 6 [18,22]. The route to target 7 began by treating N-Boc-trans-4-hydroxy-L-proline methyl ester (11) with diethylaminosulfur trifluoride (DAST) in CH 2 Cl 2 to install the first fluoro substituent (12) with clean configurational inversion (88%, Scheme 1).</p><p>Oxidation of the pyrrolidine to the corresponding lactam 13 using a Ru(III)/NaIO 4 system proceeded smoothly, followed by TFA-mediated Boc deprotection to yield 14 (75%, 2 steps). Reduction of the methyl ester to the primary alcohol (15, 18%), and subsequent protection as the TBDMS ether delivered the cyclisation substrate 16 in good yield (92%). A three step, one pot sequence consisting of methylation, treatment with phenylhydrazine and subsequent cyclisation furnished the triazolium salt 17 in 76% yield (3 steps). Finally, DAST-mediated TBDMS deprotection/deoxyfluorination completed the synthetic sequence to give 7 in 45% yield.</p><p>Synthesis of the monofluorinated pre-catalyst 8 (Scheme 2) commenced with an Appel reaction of alcohol 15 to prepare the primary bromide 18. Owing to the potentially labile nature of the primary bromide, this material was used without further purification in the next step. Reduction (H 2 , Pd/C) furnished the lactam 19 (21% over 2 steps) in preparation for the cyclisation sequence. As previously described, successive treatment with the Meerwein salt, phenylhydrazine and methyl orthoformate yielded the target triazolium salt 8 in 61% over 3 steps. The pre-catalysts 9 and 10 required for control experiments were prepared by an analogous reaction sequence (Scheme 3). Commercially available (S)-(+)-(trifluoromethyl)pyrrolidine 20 was protected (21, quantitative), oxidised to the corresponding lactam (22, 38% over 2 steps) and processed to the target triazolium salt 9 (46%, 3 steps). The non-fluorinated catalyst 10 (Scheme 3; lower) was prepared in a short synthesis starting from the primary bromide 23 [22]. Hydrogenolysis (24, 67%) [41] and subsequent conversion to the triazolium salt completed the short synthesis (52% over 3 steps). X-Ray structural analysis of 5, 6 and 7</p><p>The X-ray crystal structures of triazolium salts 5, 6 and 7 were then to examine the conformation of the β-fluoroamine motifs that were the major motivation for this study (Figure 3) [42]. In previous analyses of (S)-2-(fluorodiphenylmethyl)pyrrolidine derivatives, the synclinal-endo conformation was almost exclusively observed in the solid state [13,15,16,18,21,22]. This was also found to be the case in triazolium salt 5 (Φ NCCF −54.0°), with the diphenylfluoromethyl group adopting a quasi-equatorial orientation, presumably to minimise non-bonding interactions as a consequence of the sterically demanding phenyl groups.</p><p>Deletion of these Ph units from the exocyclic group ( 6) resulted in a switch to the synclinal-exo conformation (Φ NCCF +67.9°), with the monofluoromethyl group occupying a quasi-axial orientation. Interestingly, this synclinal-endo → synclinal-exo switch is also observed in the corresponding pyrrolidino systems [13,21]. The hybrid structure 7 containing both β-fluoroamine types again showed the synclinal-exo arrangement (Φ NCCF +63.65°) as expected, although the fluorine group on the ring system did little to alter the conformation when compared with 5 and 6.</p><p>Having completed the synthesis of the fluorinated triazolium salts (5-10) for this study, their effectiveness in catalysing the Steglich rearrangement of an oxazolyl carbonate derivative (25) a Representative reaction protocol: To a suspension of 6 in the appropriate solvent was added the base indicated. The mixture was then stirred for 15 min before a solution of 25 (20.0 mg, 76.0 µmol) in toluene was added. The mixture was stirred for a further 18 h, after which time the solution was concentrated in vacuo and filtered over a plug of silica gel (CH 2 Cl 2 as eluent). The resulting solution was then concentrated in vacuo. b The conversion and enantiomeric ratio of the product were determined by HPLC on an Agilent 1260 series system using a reprocil chiral-OM 4.6 mm column. Percent conversion was determined by integration of the starting material and product peaks, correcting for response factors.</p><p>to the corresponding C-carboxyazlactone 26 was investigated (Table 1). For this initial study, the monofluorinated triazolium salt 6 was arbitrarily chosen (10 mol %), with toluene being used as the reaction medium and KHMDS as the base [30]. Gratifyingly, complete conversion was observed after 18 h and with good levels of enantioselectivity (e.r. 80.5:19.5). Variation in the choice of solvent proved detrimental to both the conversion and enantioselectivity ( 1, entry 10) led to higher levels of enantioselectivity. A control reaction using solid KHMDS, rather than the commercial 0.5 M solution in toluene, revealed a lower conversion but did not alter the enantiomeric ratio (Table 1, entry 11). However, a commensurate performance was noted with Cs 2 CO 3 (Table 1, entry 12, >99%, e.r. 80.5:19.5). Alterations in reaction concentration had little influence on the selectivity (Table 1, entries 13 and 14, 0.02 or 0.5 mol•L −1 , e.r. 80.5:19.5 and 79.0:21.0, respectively). However, catalyst loading did dramatically alter the selectivity outcome (Table 1, entries [15][16][17]. Given that similar enantioselectivities were recorded in reactions using Cs 2 CO 3 (cf. KHMDS), an analogous set of reactions were run for complete- a Representative reaction protocol: A suspension of the catalyst (7.6 µmol) in toluene (200 µL) was treated with Cs 2 CO 3 (2.5 mg, 7.6 µmol) and stirred for 15 min. A solution of 25 (20.0 mg, 76.0 µmol) in toluene (200 µL) was then added. The mixture was stirred for a further 18 h, after which time the solution was concentrated in vacuo and filtered over a plug of silica gel (CH 2 Cl 2 as eluent). The resulting solution was then concentrated in vacuo. b The conversion and enantiomeric ratio of the product were determined by HPLC on an Agilent 1260 series system using a reprocil chiral-OM 4.6 mm column. Percent conversion was determined by integration of the starting material and product peaks, correcting for response factors. c Reversal in the sense of enantioselectivity. Having evaluated a series of parameters for the catalytic Steglich rearrangement using catalyst 6, efforts were then focussed on a logical process of molecular editing to clarify the effect of H→F substitution (Table 2). Once again, toluene was employed as solvent, and reactions were run at rt for 18 h at a concentration of 0.19 M. Due to the similar enantioselectivities observed when using KHMDS and Cs 2 CO 3 (Table 1), it was deemed prudent to perform the study using both bases independently. Initially, the bulky diphenylfluoromethyl-containing triazolium salt 5 was subjected to the optimised conditions. It was envisaged that one of the phenyl rings might assist in the facial discrimination of the activated electrophile, as a consequence of the fluorine gauche effect (Φ NCCF −54.0°, Figure 3). However, the product C-carboxyazlactone 26 was isolated essentially in racemic form ( 2, entries 9 and 10). Deletion of both fluorine atoms from the catalyst core (10) was accompanied by a drop in enantioselectivity (e.r. 62.5:37.5), although the reactions did not display the same sensitivity to changes in base (Table 2, entries 11 and 12).</p><!><p>In conclusion, the ability of fluorine to modulate the catalytic performance of N-heterocyclic carbenes in the Steglich rearrangement of oxazolyl carbonates has been demonstrated. A focussed molecular editing study (Figure 4) has revealed that the introduction of a single fluorine atom on the exocyclic unit leads to enhanced enantioselectivities (6 versus 10, e.r. 80.5:19.5 versus 62.5:37.5). Further augmentation can be achieved by introduction of a second fluorine substituent on the catalyst core (7; e.r. 87.0:13.0, 99% conversion). However, the reinforcing role of these two fluorine substituents in orchestrating enantioinduction requires clarification and will be the subject of future investigations. What is apparent is that fluorine incorporation can confer significant advantages in (organo)catalyst optimisation and design.</p><!><p>Full experimental data is provided in Supporting Information File 1.</p>

Beilstein

TMEM106B p.T185S regulates TMEM106B protein levels: implications for frontotemporal dementia

Frontotemporal lobar degeneration (FTLD) is the second leading cause of dementia in individuals under age 65. In many patients, the predominant pathology includes neuronal cytoplasmic or intranuclear inclusions of ubiquitinated TAR DNA binding protein 43 (FTLDTDP). Recently, a genome-wide association study identified the first FTLD-TDP genetic risk factor, in which variants in and around the TMEM106B gene (top SNP rs1990622) were significantly associated with FTLD-TDP risk. Intriguingly, the most significant association was in FTLD-TDP patients carrying progranulin (GRN) mutations. Here we investigated to what extent the coding variant, rs3173615 (p.T185S) in linkage disequilibrium with rs1990622, affects progranulin protein (PGRN) biology and TMEM106B protein regulation. First, we confirmed the association of TMEM106B variants with FTLD-TDP in a new cohort of GRN mutation carriers. We next generated and characterized a TMEM106B-specific antibody for investigation of this protein. Enzyme-linked immunoassay analysis of PGRN levels showed similar effects upon T185 and S185 TMEM106B overexpression. However, overexpression of T185 consistently led to higher TMEM106B protein levels than S185. Cycloheximide treatment experiments revealed that S185 degrades faster than T185 TMEM106B, potentially due to differences in N-glycosylation at residue N183. Together, our results provide a potential mechanism by which TMEM106B variants lead to differences in FTLD-TDP risk.

tmem106b_p.t185s_regulates_tmem106b_protein_levels:_implications_for_frontotemporal_dementia

Introduction<!>Study population<!>Genotyping and association analyses<!>Construction and mutagenesis of cDNA clones<!>Antibodies<!>Cell culture transfection and Western blotting<!>Immunofluorescence<!>Immunohistochemistry<!>PGRN ELISA<!>Drug Treatments<!>TMEM106B deglycosylation<!>RNA isolation and real-time PCR<!>Statistical Analysis<!>Genetic analysis of TMEM106B in a new cohort of GRN mutation carriers<!>TMEM106B antibody characterization and protein localization<!>T185 TMEM106B is more highly expressed than S185 TMEM106B<!>TMEM106B isoforms differ in their rate of degradation<!>Glycosylation at position N183 might be involved in TMEM106B isoform stability<!>Discussion

<p>Frontotemporal lobar degeneration (FTLD) is the second most frequent neurodegenerative disorder in individuals under the age of 65, accounting for 5–10% of dementia patients (Neary et al. 1998, Ratnavalli et al. 2002). FTLD patients often present with behavioral changes and personality dysfunction, corresponding to the pathology in frontal and temporal lobes (Boxer & Miller 2005, McKhann et al. 2001). Post-mortem analyses of FTLD brains determined that the most common pathological subtype involves intracellular inclusions of ubiquitinated TAR DNA binding protein 43 (TDP-43). In 2006, mutations causing a 50% loss in progranulin protein (PGRN) were found in the progranulin gene (GRN), accounting for nearly 20% of patients affected by this FTLD subtype (FTLD-TDP) (Baker et al. 2006, Cruts et al. 2006, Gass et al. 2006, Gijselinck et al. 2008). More recently, an expansion in a hexanucleotide repeat in the chromosome 9 open reading frame 72 gene was identified as an additional major genetic cause of FTLD-TDP (DeJesus-Hernandez et al. 2011, Renton et al. 2011).</p><p>Genetic studies have also served as a successful tool in the discovery of risk factors for FTLD-TDP. In 2010, Van Deerlin and colleagues performed a genome-wide association study in which single nucleotide polymorphisms (SNPs; top SNP rs1990622 T>C) located at the transmembrane protein 106 B gene locus (TMEM106B) on chromosome 7p21 were identified as FTLD-TDP risk factors (Vass et al. 2011). Moreover, the risk association of these SNPs was greatest in GRN mutation carriers (Vass et al. 2011, Finch et al. 2011), in which there was a highly significant decrease in the frequency of homozygous carriers of the rs1990622 minor allele, suggesting a protective effect of this allele in these patients. The exact relationship between TMEM106B polymorphisms and TMEM106B regulation and/or function, however, remains poorly understood. Initial studies showed a dose-dependent decrease in TMEM106B mRNA expression associated with the rs1990622 minor allele (Vass et al. 2011); however, this could not be confirmed in subsequent studies (Cruchaga et al. 2011, van der Zee et al. 2011). Also, even though individuals homozygous for the protective minor allele of rs1990622 showed higher plasma PGRN levels (Finch et al. 2011), only a minor absolute increase in PGRN was observed, suggesting other disease mechanisms may be at play. Finally, we and others identified that rs1990622 is in perfect linkage disequilibrium with p.T185S (rs3173615), a coding variant in TMEM106B; however, thus far no studies have been reported to elucidate the functional consequence of p.T185S on TMEM106B.</p><p>At the time of these genetic discoveries, TMEM106B was a relatively uncharacterized transmembrane protein; however, recent publications have indicated that TMEM106B is a glycoprotein predominantly localized at the lysosomal membrane where it might interact with intracellular PGRN (Lang et al. 2012, Chen-Plotkin et al. 2012, Brady et al. 2012). Though these reports investigated the effect of TMEM106B expression on PGRN levels in vitro, the data is conflicting, and the effect of the coding variant p.T185S was not extensively studied. To further characterize TMEM106B and its role in FTLD-TDP, we investigated the risk (T185) and protective (S185) isoforms of TMEM106B and their effects on TMEM106B. First, we replicated the importance of TMEM106B polymorphisms on FTLD-TDP risk in GRN mutation carriers in a new cohort of patients. We further confirmed that TMEM106B is located in the lysosomes where it might interact with PGRN. Importantly, we showed that the protective (S185) TMEM106B isoform is consistently expressed at lower levels than the T185 TMEM106B isoform due to an increased rate of protein degradation, possibly resulting from changes in TMEM106B glycosylation. Thus, we provide the first insight into a functional difference between the risk (T185) and protective (S185) isoforms of TMEM106B.</p><!><p>A total of 29 newly identified white patients with GRN mutations (14 females, 15 males) were included in the genetic association study. Patients were either identified by sequencing analysis or by low PGRN levels on a plasma PGRN ELISA test (Adipogen), and mutations included c.29T>C, c.90_91insCTGC, c.102delC, c.234_235delAG, c.328C>T, c.415T>C, c.592_593delAG, c.813_816delCACT, c.918C>A, c.933+1G>A, c.1072C>T, c.1252C>T, c.1428_1431delGGAT and c.1477C>T. The mean age at diagnosis was 63.3 ± 10.5 years (range 44–76). Patients were ascertained from a total of 5 centers: Mayo Clinic Jacksonville (N=5), Mayo Clinic Rochester (N=1), University of British Columbia, Canada (N=5), University of Western Ontario, Canada (N=2), and IRCCS Istituto Centro San Giovanni di Dio Fatebenefratelli, Brescia, Italy (N=13). An additional 3 patient samples were obtained from the Mayo Clinic Jacksonville Brain Bank. All patients agreed to be in the study and biological samples were obtained after informed consent with ethical committee approval from the respective institutions.</p><!><p>Genotyping of rs1990622 and rs3173615 TMEM106B variants was performed using Taqman SNP genotyping assays (assay numbers C_11171598_10 and C_27465458_10, respectively; Invitrogen) on the 7900HT Fast Real Time PCR system. Genotype calls were made using the SDS v2.2 software (Applied Biosystems).</p><!><p>cDNA constructs encoding the V5-tagged p.T185S TMEM106B genes were constructed from heterozygous human cDNA. Briefly, total RNA was extracted from human frontal cortex brain tissue heterozygous for rs3173615 using the RNeasy Plus kit (Qiagen). Reverse transcription was performed using the Superscript III system (Invitrogen), and cDNA was PCR-amplified with primers specific to TMEM106B to create BamHI and XhoI restriction sites, as well as the V5 tag. Amplified cDNA was digested with BamHI and XhoI restriction enzymes and cloned into the pAG3 and pAAV1 plasmids. The N183S (N4) mutation was induced independently in wild-type and mutant TMEM106B clones using the QuickChange site-directed mutagenesis kit (Stratagene, Agilent Technologies) and mutagenesis primers per the manufacturer's instructions. All clones were confirmed by sequencing analysis and transformed into Stbl3 competent E. Coli cells (Invitrogen), after which DNA for cell culture studies was obtained using the Nucleobond Xtra Maxi Plus EF kit (Clontech). Primer sequences are listed in Supplemental Table 1.</p><!><p>A polyclonal rabbit TMEM106B antibody was generated through Thermo Fisher Pierce Protein Research against an N-terminal TMEM106B peptide comprised of amino acids 14:27 (KEDAYDGVTSENMR). The following antibodies were used for Western blotting experiments: rabbit polyclonal anti-TMEM106B (1:50,000 to 1:200,000), pre-immune rabbit serum (1:100,000), mouse monoclonal anti-V5 (1:200,000; Invitrogen), mouse monoclonal anti-LAMP-1 (1:500; Santa Cruz) and mouse monoclonal anti-GAPDH (1:500,000; Meridian Life Science). Anti-mouse or anti-rabbit secondary HRP-conjugated antibodies (1:5000; Promega) were used for chemiluminescence. The following antibodies were used for immunofluorescence: rabbit polyclonal anti-TMEM106B (1:5,000), mouse monoclonal anti-V5 (1:50,000; Invitrogen), mouse monoclonal anti-PGRN (1:100; R&D Systems), and lysosomal-associated membrane protein 1 (LAMP-1) (1:100; Santa Cruz). The anti-TMEM106B antibody and pre-immune rabbit serum were used as a concentration of 1:1000 for immunohistochemical analyses.</p><!><p>HeLa cells were maintained in Eagle's Minimum Essential Medium supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin at 37° C, 5% CO2. One day prior to transfection, the cells were plated at 200,000 cells per well in a 6-well culture dish. The next day, the cells were transfected with 2 μg of control plasmid DNA (tGFP- (Origene) or empty pAAV1) or V5-tagged TMEM106B plasmid DNA in either the wild-type or mutant form using the Lipofectamine 2000 transfection reagent (Invitrogen). Cells were harvested in radioactive immunoprecipitation assay (RIPA) buffer (Boston BioProducts) for Western blot analysis. Cell lysates were diluted in an equivalent volume of 2X Novex Tris-glycine sodium-dodecyl sulfate (SDS) sample buffer (Invitrogen) supplemented with β-mercaptoethanol (to 5%) and denatured at room temperature for 30 min. Equal volumes were run on 10% SDS-polyacrylamide gels (Invitrogen), transferred to Immobilon membranes (Millipore), and immunoblotted with the primary antibody. The next day, blots incubated with an HRP-conjugated secondary antibody and bands were detected by enhanced chemiluminescence using Western Lightning Plus-ECL reagents (Perkin Elmer). Peptide preabsorption experiments were performed for 2 hrs using 1 μg of the immunizing peptide before adding to the membrane. Western blot quantifications were normalized to GAPDH levels.</p><!><p>HeLa cells were plated and transfected on coverslips. Three days post transfection, the cells were fixed in 4% paraformaldehyde (PFA) in phosphate buffered saline (PBS), and permeabilized in 0.3% Triton X-100 in PBS. Coverslips were then incubated in 10% bovine serum albumin in PBS at room temperature for 1 h before labeling with the primary antibody. The next day, the coverslips were labeled with the secondary antibody, incubated with Hoechst stain in PBS (1:5000; Molecular Probes), and mounted on glass slides using Fluoromount-G (Southern Biotech). Images were acquired with a Zeiss LSM 510 META confocal laser-scanning microscope running either AIM or Zen2009 software (both from Carl Zeiss, Germany) using a Plan-Apochromat 63X/1.4NA oil immersion objective at a scan zoom of 1.7 and optical slice depth of 0.7 μm. For colocalization analyses, TMEM106B and LAMP-1 weighted colocalization coefficients were obtained using the Zen2009 software.</p><!><p>Tissue sections of the amygdala, hypothalamus, lentiform nucleus, neocortex, hippocampus, thalamus, midbrain, pons, medulla and cerebellum were studied in 12 neurologically normal elderly individuals (8 females, 4 males) and 13 FTLD-TDP (4 female, 9 male) cases. Sections were cut at 5 μm thickness from formalin-fixed paraffin-embedded blocks, deparaffinized in xylene, rehydrated, and stained as previously described (Bieniek et al. 2012). Sections were processed with either pre-immune serum or the TMEM106B antibody and counterstained with hematoxylin.</p><!><p>To determine PGRN expression levels in cell culture media and lysates, we used the Human Progranulin Quantikine ELISA kit (R&D Systems) following manufacturer's instructions. Samples were undiluted, performed in duplicate, and the provided human recombinant PGRN was used as a standard.</p><!><p>Cycloheximide was used at a concentration of 20 μg/ml to block protein synthesis. Lysosomal degradation was inhibited using 10 μM leupeptin and proteasomal degradation was obtained with 1 μM epoxomicin or 10 μM MG-132. For all drug treatments on T185- and S185-transfected cells, cells were harvested 0, 1, 3, 6, and 8 hrs after treatment for analysis by Western blotting.</p><!><p>Transfected HeLa cells were incubated on ice for 15 minutes in lysis buffer (150 mM NaCl, 50 mM Tris-HCl pH 7.5, 2 mM EDTA, 1% NP-40, protease inhibitor cocktail) and centrifuged at 6000 rpm, at 4° C for 5 minutes. The supernatant was transferred to clean tubes and 10 μg of protein per sample was denatured in 10X Glycoprotein Denaturing Buffer (New England BioLabs Inc.) at room temperature for 30 minutes. To perform deglycosylation, 500 units of recombinant Edoglycosidase H (EndoH) (New England BioLabs Inc.) in 10X Reaction Buffer were used. Samples incubated in 10X Reaction Buffer without the presence of the enzyme served as a control. The samples with and without EndoH were incubated at 37° C for 1 hour and the reaction was stopped with the addition of denaturing buffer 2X Sample Buffer (10% β-mercaptoethanol) at room temperature for 30 minutes. Deglycosylated TMEM106B was separated and analyzed using SDS-PAGE.</p><!><p>Total RNA was extracted from transfected HeLa cells using the RNeasy Plus Mini Kit (Qiagen), and its quality was assessed on an Agilent 2100 Bioanalyzer. RNA samples were normalized to 50 ng/μl and using 300 ng as template, a reverse transcription reaction was performed using the Superscript III system (Invitrogen). Real-time quantitative PCR using an ABI7900 was performed in triplicate for each sample using gene expression probes for TMEM106B (Hs00998849_m1), GAPDH (Hs00266705_g1) and RPLP0 (Hs00420895_gh). Results were analyzed using SDS software version 2.2 and relative quantities of TMEM106B mRNA were determined.</p><!><p>For experiments in which only two groups were compared, significance was measured using a two-sample t-test. For analyses involving more than two groups, GraphPad Prism 5.04 (GraphPad Software) was utilized to perform a one-way ANOVA followed by the Tukey's Multiple Comparison test. Using the same software, the drug time course experiments were analyzed by linear regression.</p><!><p>To provide further support for TMEM106B SNPs in modifying the disease risk in GRN mutation carriers, we genotyped rs1990622 and rs3173615 (p.T185S) in a cohort of 25 newly identified symptomatic GRN mutation carriers. In this series, both SNPs were in complete linkage disequilibrium and only one patient was homozygous for the minor allele (rs1990622: TT n=15, TC n=9, CC n=1). The frequency of homozygous carriers of the minor allele was significantly less (1/29, 3.4%) compared to what would be expected based on a previously genotyped control population (157/822, 19%; p=0.03, OR=0.15) (Finch et al. 2011). Interestingly, the patient carrying two copies of the minor allele is currently 71 years old with a diagnosis of mild cognitive impairment (MCI) and had been identified as a GRN mutation carrier based on low PGRN plasma levels by ELISA (Ghidoni et al. 2012).</p><!><p>To better investigate the role of the TMEM106B protein in culture and in vivo, a rabbit polyclonal antibody raised against the N-terminus (aa 14–27) of the human TMEM106B protein was generated and tested both in HeLa cells transfected with V5-tagged TMEM106B and in human frontal cortex brain tissue (Fig. 1A). In both sample types, a predominant band was detected at approximately 40 kDa using the TMEM106B antibody, with a less prominent higher molecular weight species at approximately 70–90 kDa. Both the low and high molecular weight species in TMEM106B-overexpressing HeLa cell lysates were also detected when blotted with a V5 antibody (Fig. 1B), and were no longer visible after preabsorption with the TMEM106B immunogen peptide (Fig. 1D). In brain samples, the higher molecular weight band appeared to be non-specific as it was also detected with the pre-immune rabbit serum, as well as after peptide incubation of our TMEM106B antibody. In contrast, the ~40 kDa species appeared specific in these tissue samples (Fig. 1C & 1D). Immunocytochemical experiments further supported the specificity of our antibody showing substantial overlap of V5 and TMEM106B antibody labeling in V5-tagged TMEM106B-overexpressing HeLa cells upon confocal imaging (Fig. 1E).</p><p>Immunohistochemistry in paraffin sections of normal and disease brains with the TMEM106B antibody showed cytoplasmic punctate or granular immunoreactive structures in both neurons and non-neuronal cells, most notably microglia, in a pattern similar to that seen with immunohistochemistry for PGRN (Supplemental Fig. 1) (Ahmed et al. 2007). We did not observe differences in the distribution of neuronal TMEM106B in controls as compared to FTLD-TDP cases (Supplemental Fig. 2).</p><p>Previous reports have shown that TMEM106B is a type II transmembrane glycoprotein located within the late endosome and lysosomal compartments of the endomembrane system (Chen-Plotkin et al. 2012, Lang et al. 2012). Consistent with these studies, immunocytochemical analyses of cells transfected with either T185 or S185 TMEM106B showed punctate cytoplasmic labeling with our TMEM106B antibody, suggesting its localization to a subcellular compartment (Fig. 1F & 1G). Co-labeling of TMEM106B with a late endosomal and lysosomal marker, LAMP-1, showed overlap of these two antibodies, further suggesting that TMEM106B is, at least in part, located within the late endosomes or lysosomes in our cell culture system, with more colocalization of the T185 than S185 TMEM106B (T185 weighted colocalization coefficient 0.878 ± 0.010 as compared to S185 weighted colocalization coefficient 0.750 ± 0.019; p<0.0005) (Fig. 1F & 1G).</p><p>To determine to what extent the TMEM106B isoforms colocalize with PGRN, cells were transfected with either T185 or S185 TMEM106B and labeled with TMEM106B and PGRN antibodies. We observed a striking overlap between PGRN and TMEM106B proteins, regardless of TMEM106B isoform (Fig. 2A & 2B). We also observed that cells transfected with T185 or S185 TMEM106B appeared to have more intense fluorescence with the PGRN antibody compared to non-transfected cells, indicating increased endogenous levels of PGRN in cells overexpressing TMEM106B (Fig. 2A & 2B). To more accurately determine whether T185 and/or S185 TMEM106B levels differently alter endogenous PGRN levels, we harvested T185-and S185-transfected HeLa cells 2 days after transfection and quantified both intra- and extracellular levels of PGRN using a PGRN ELISA. TMEM106B overexpression of either isoform caused a significant increase in PGRN levels in both intracellular and in the media as compared to control-transfected cells (Fig. 2C). These results confirm earlier observations published by Brady et al. and indicate that a direct effect on PGRN levels is unlikely to explain the risk associated with p.T185S in TMEM106B (Brady et al. 2012).</p><!><p>Upon overexpression of T185 and S185 TMEM106B as part of our PGRN-related studies, we consistently observed a more prominent TMEM106B-immunoreactive band in T185-transfected cells (Fig. 2D). Upon quantification of TMEM106B protein levels, S185 expression was approximately 37% that of T185 expression both 2 and 3 days post transfection (p<0.0001) (Fig. 3A & 3B). To determine whether these differences in TMEM106B isoform expression were due to changes at the RNA level, TMEM106B RNA levels were measured in T185- and S185-transfected cells at the same time points. Quantitative PCR analyses confirmed that TMEM106B RNA levels are not different in T185 versus S185 overexpressing cells at 2 or 3 days (Fig. 3C & 3D). To eliminate the possibility that differences between T185 and S185 protein expression are plasmid-specific, we further cloned the TMEM106B isoforms into a pAG3 mammalian expression plasmid and transfected these constructs into HeLa cells for 3 days. Consistent with our pAAV TMEM106B constructs, S185 TMEM106B protein was significantly less expressed than T185 TMEM106B (56% of T185 levels; p<0.05), independent of RNA levels (Supplemental Fig. 3A,C,E). To ensure that these observations were not specific only to HeLa cells, we also transformed T185 and S185 TMEM106B into human embryonic kidney (HEK-293T) cells for 3 days, after which protein and RNA levels were quantified. Similar to HeLa cells, HEK-293T cells also expressed S185 TMEM106B significantly less than T185 (55% of T185 levels; p<0.05), independent of RNA levels (Supplemental Fig. 3B,D,F). Taken together, these results suggest that a post-translational mechanism is responsible for the differences in protein levels between the risk (T185) and protective (S185) isoforms of TMEM106B.</p><!><p>To determine whether T185 and S185 TMEM106B proteins are degraded at different rates, we treated T185- and S185-transfected cells with 20 μg/ml of cycloheximide to block protein synthesis. Western blotting of these lysates showed a gradual and marked decrease in TMEM106B levels of both isoforms with increasing time of cycloheximide treatment (Fig. 4A). We next quantified the changes in TMEM106B T185 or S185 levels over time and showed that the rate of degradation of the S185 isoform was significantly faster than the degradation rate of T185 TMEM106B (slopes are different; p = 0.016) (Fig. 4C). Based on our raw data, it would take approximately 8 hrs for 50% of T185 TMEM106B protein to be degraded as compared to only 2 hrs for S185 TMEM106B.</p><p>To rule out the possibility that different rates of protein synthesis of T185 and S185 TMEM106B might also contribute to their varying protein expression levels, we blocked TMEM106B degradation to observe the rate of TMEM106B intracellular accumulation. Since it remains unclear as to what subcellular compartment is involved in TMEM106B's degradation, we first pharmacologically abrogated degradation in the proteasome using 10 μM MG132 and in the lysosome using 10 μM leupeptin for 8 hrs. Western blotting analysis of TMEM106B-transfected HeLa cells treated with MG132 did not show significant changes in TMEM106B expression after 8 hrs (Supplemental Fig. 4A). Similar results were obtained using a second proteasomal inhibitor, epoxomicin (data not shown). However, an increase in TMEM106B expression of both isoforms was observed upon treatment with leupeptin at 8 hrs (Supplemental Fig. 4B), indicating the lysosome as the predominant subcellular compartment involved in TMEM106B degradation. Thus, we performed a time course experiment in which TMEM106B-transfected cells were harvested 0–8 hrs after leupeptin treatment in order to observe the rate of TMEM106B protein synthesis after T185 or S185 overexpression. As indicated by Western blotting, leupeptin treatment of TMEM106B-transfected cells caused a gradual increase in TMEM106B levels in both T185 and S185 overexpressing cells (Fig. 4B); however, the rate of protein synthesis of the T185 and S185 isoforms was not different (slopes are not different; p=0.44) (Fig. 4D). Together these studies suggest that the differences observed between T185 and S185 protein expression is unlikely to be due to differences in the rate of protein synthesis, but more likely due to differences in the rate of their degradation.</p><!><p>In the recent report by Lang et al., TMEM106B was found to be a glycosylated protein with 5 N-glycosylation sites throughout the protein at amino acids N145, N151, N164, N183, and N256, each with an N-X-T/S consensus sequence (Lang et al. 2012). Simple glycan modifications are added to mature T185 TMEM106B at the first three N-glycosylation sites, whereas complex glycans were detected on the last two N-glycosylation sites of the protein. Proper protein glycosylation is critical for adequate protein folding and subsequent stability, localization, and function. Interestingly, the p.T185S variant in the TMEM106B protein is located within the glycosylation consensus sequence for the fourth N-glycosylation site. Therefore, p.T185S might contribute to differences in glycosylation at N183. To address this question, we first extended the time of SDS gel separation of lysates from HeLa cells overexpressing T185 and S185 TMEM106B isoforms; however, no obvious difference in their molecular weights was identified (data not shown). To more specifically determine whether N183 glycosylation differs between T185 and S185 TMEM106B, we subjected lysates from T185- or S185-overexpressing cells to digestion with endoglycosidase H (EndoH). This enzyme is only capable of removing simple glycans, leaving complex modifications intact. Thus, if simple glycans are added to N183 of S185 TMEM106B, we expected to see a lower molecular weight product post EndoH digestion than with T185. EndoH treatment of both the T185 and S185 TMEM106B proteins however resulted in similar TMEM106B-immunoreactive bands (Fig. 5A), indicating that complex glycosylation is likely preserved at N183 in the S185 TMEM106B protein. EndoH digestion did not rule out the possibility that more subtle changes in complex gycosylation at N183 might account for differences in protein stability between T185 and S185 TMEM106B.</p><p>To further study the involvement of N183 in the stability difference observed between the TMEM106B isoforms, we introduced a N183S mutation in both T185 and S185 TMEM106B and compared expression levels of the mutants. As compared to non-mutated T185 and S185 TMEM106B, the N183S mutations caused a downward shift in the molecular weight of the TMEM106B protein (Fig. 5B), likely due to complete loss of glycosylation at amino acid 183 (Lang et al. 2012). Moreover, protein levels of the overexpressed N183S mutants resulted in a robust decrease in TMEM106B expression (Fig. 5B), underlining the significance of the N183 site for TMEM106B maturation. Importantly, quantification of mutant expression levels indicated that when the N183S mutation was introduced, TMEM106B RNA levels were still the same between T185 and S185 TMEM106B, but now the T185 TMEM106 protein levels were no longer significantly higher than that of S185 (Fig. 5C & 5D).</p><!><p>It has become increasingly clear that variants in or near the TMEM106B gene play a critical role in the risk of developing FTLD-TDP. Strong association was especially apparent in FTLD-TDP patients carrying GRN mutations, suggesting that TMEM106B might modify the disease through regulation of PGRN levels or function (Vass et al. 2011). In this study, we further confirmed the association of TMEM106B with FTLD-TDP in an additional 29 patients with loss-of-function GRN mutations. Only one patient (3.4% of patients) was homozygous for the minor protective allele of rs1990622. Consistent with our previous findings, the phenotype in this patient is relatively mild, with a diagnosis of mild cognitive impairment at the age of 71 years. In our laboratory we previously identified 2 out of 127 patients (1.6%) to be homozygous for the minor allele of rs1990622. Cumulatively, among unrelated probands of GRN mutation families, we identified only 3 out of 156 (1.9%) patients to be homozygous for the minor allele of rs1990622 as compared to 157 out of 822 (19.1%) control individuals studied to date (Vass et al. 2011, Finch et al. 2011, and this study). These genetic findings add to the growing body of evidence confirming TMEM106B as a disease risk factor and potential modifier in GRN-related FTLDTDP.</p><p>To provide insight in the molecular mechanisms associated with TMEM106B variants, we focused our studies on the coding variant p.T185S, which is in complete linkage disequilibrium with rs1990622, as a potential functional variant implicated in FTLD-TDP risk. Since the TMEM106B variants were associated most strongly with FTLD-TDP risk in GRN mutations carriers, TMEM106B might confer risk by directly affecting PGRN levels or function. Here we show that both isoforms of TMEM106B (T185 and S185 TMEM106B) colocalize, in part, with lysosomal compartments. Immunofluorescence analyses of T185 and S185 TMEM106B also showed that both TMEM106B isoforms colocalize with PGRN. This is in agreement with two recent reports that also observed significant overlap of the TMEM106B and PGRN (Brady et al. 2012, Chen-Plotkin et al. 2012). Moreover, we found that PGRN levels were significantly increased as compared to control-transfected cells in the media and lysates of T185- and S185-transfected cells at 2 days. Even though the first characterization of TMEM106B did not reveal TMEM106B-induced changes in PGRN (Lang et al. 2012), we are now the third group to observe an increase in PGRN with TMEM106B overexpression (Chen-Plotkin et al. 2012, Brady et al. 2012). However, similar to findings by Brady et al. (Brady et al. 2012) both TMEM106B T185 and S185 isoforms similarly affected PGRN levels. These results, therefore, fail to explain the decrease in FTLD-TDP risk in people who are homozygous for the rs3173615 TMEM106B minor allele.</p><p>In previous reports discerning the association between TMEM106B and FTLD-TDP, attention has been drawn to the potential effect of the TMEM106B SNPs on TMEM106B mRNA expression levels (Vass et al. 2011, Chen-Plotkin et al. 2012, Brady et al. 2012). This began with the observation that individuals homozygous for the protective C-allele of rs199022 had lower TMEM106B RNA levels in brain tissue (Vass et al. 2011). However, this finding was performed using a small subset of samples and has not been replicated by other groups (Cruchaga et al. 2011, van der Zee et al. 2011). Although we cannot exclude that rs1990622 or variants in linkage disequilibrium with rs1990622 located in non-coding regulatory regions of TMEM106B might contribute to difference in TMEM106B expression levels in vivo, our study now provides strong evidence implicating p.T185S as a functional TMEM106B variant modulating TMEM106B protein levels. Using multiple cell lines and expression vectors we consistently showed that the risk T185 TMEM106B isoform was expressed nearly two-fold greater than S185 TMEM106B. Subsequent analyses showed that the difference in expression resulted from a more rapid degradation of the S185 TMEM106B isoform in our cell culture system. Brady el at. did not report differences between these two TMEM106B isoforms; however, discrepancies might result from their use of a mouse cell line (Brady et al. 2012). Additionally, the transfection levels were not reported for each isoform (Brady et al. 2012). Our findings support the hypothesis that higher TMEM106B protein levels are, at least in part, contributing to the risk differences between the T185 and S185 TMEM106B isoforms and provide the first variant-related difference in the post-translational regulation of TMEM106B.</p><p>The p.T185S coding variant of the TMEM106B protein is a part of the N-X-T/S glycosylation consensus sequence for N-glycosylation at position 183 (Lang et al. 2012). Protein glycosylation is a critical post-translational modification that enhances functional diversity as well as biological activity and expression level of a wide-range of glycoproteins. Even though a T or S residue at position 185 is expected to be sufficient for N-glycosylation at TMEM106B N183, we speculate that glycosylation may be different between the two isoforms with a slight change in the glycan composition and/or complexity at N183 in TMEM106B T185 compared to S185. In support of this hypothesis, it has been previously shown that T and S amino acids can affect N-glycosylation transfer rates in vitro, with less efficient interaction between glycotransferase enzymes in S-containing consensus sequences (Bause & Legler 1981). Previous work showed that complete loss of glycosylation at N183 results in retention of TMEM106B to the endoplasmic reticulum (ER) (Lang et al. 2012). Based on our results, cells overexpressing S185 TMEM106B still show TMEM106B localized to the lysosomes suggesting that more subtle changes in glycosylation may be at play. Subtle differences in S185 versus T185 TEM106B N183 glycosylation could explain why no gross differences in the molecular weights or EndoH digestion products of T185 versus S185 TMEM106B were observed. Confirming differences in the composition of complex N-glycans at TMEM106B amino acid 183 would require extensive mass spectrometry analyses and/or specific high performance liquid chromatography beyond the scope of this study. Thus, while our current data do not specifically confirm N183 glycosylation as the functional process involved in regulating T185 versus S185 TMEM106B expression levels, abnormal glycosylation of TMEM106B S185 could explain the enhanced degradation of this TMEM106B isoform. In line with this hypothesis, introduction of an artificial N183S glycosylation-defective mutant within the TMEM106B T185 and S185 isoforms ablated the observed differences in protein expression between these two isoforms.</p><p>In conclusion, our study is the first to demonstrate that the p.T185S coding variant, genetically associated with FTLD-TDP risk, acts as a functional variant to regulate TMEM106B protein levels, which we speculate are due to changes in glycosylation. As TMEM106B levels have been shown to be important for determining proper endolysosomal homeostasis, these results support a critical role for lysosomal dysfunction in the development of FTLD-TDP. We also confirmed an effect of TMEM106B expression on PGRN levels. However, in contrast to what would be expected from an FTLD risk factor, all published studies observed an increase (not decrease) in PGRN after overexpression of TMEM106B. Since this increase in PGRN may be the consequence of lysosomal dysfunction it remains unclear how this finding has to be interpreted and whether it has any relevance with regards to FTLD-TDP disease risk. One possibility for the strong association of TMEM106B SNPs in patients with GRN mutations may be that these patients merely reflect a patient population vulnerable to additional genetic modifying factors such as TMEM106B. The recent observation of an effect of TMEM106B genotypes on cognition in amyotrophic lateral sclerosis patients and the presence of TDP-43 pathology in Alzheimer's disease patients, two neurodegenerative diseases in which patients are thought to have normal PGRN levels, is of interest in this regard (Vass et al. 2011, Rutherford et al. 2012).</p><p>Together, our findings support a critical role for p.T185S in FTLD-TDP risk by regulating TMEM106B protein levels providing a promising novel avenue for disease intervention in FTLD-TDP and related TDP-43 proteinopathies.</p>

PubMed Author Manuscript

Expanding the Versatility of Microbial Transglutaminase Using \xce\xb1-Effect Nucleophiles as Noncanonical Substrates

The substrate promiscuity of microbial transglutaminase (mTG) has been exploited in various applications in biotechnology, in particular for the attachment of alkyl amines to glutamine-containing peptides and proteins. Here we expand the substrate repertoire to include hydrazines, hydrazides, and alkoxyamines, resulting in the formation of isopeptide bonds with varied susceptibilities to hydrolysis or exchange by mTG. Furthermore, we demonstrate that simple unsubstituted hydrazine and dihydrazides can be used to install reactive hydrazide handles onto the side chain of internal glutamine residues. The distinct hydrazide handles can be further coupled with carbonyls, including ortho-carbonylphenylboronic acids, to form site-specific and functional bioconjugates with tunable hydrolytic stability. The extension of the substrate scope of mTG beyond canonical amines thus substantially broadens the versatility of the enzyme, providing a new approach to facilitate novel applications.

expanding_the_versatility_of_microbial_transglutaminase_using_\xce\xb1-effect_nucleophiles_as_noncan

Introduction<!>\xce\xb1-Effect nucleophiles as substrates for mTG<!>Stability of the isopeptide bond analogues toward mTG-mediated hydrolysis and exchange<!>Functionalization of a peptide/protein with a side chain hydrazide group<!>Application of the installed hydrazide handles in bioconjugation<!>Discussion of other potential applications<!>Conclusion

<p>A number of enzyme-based methods have been developed to incorporate novel functionalities into biomolecules.[1] These strategies make use of the ability of the enzymes to accept diverse substrate analogues. One prominent member of this enzymatic toolbox is transglutaminase (TG), of which the simpler single-domain and calcium-independent variants of microbial origin are commonly used.[2] Microbial transglutaminase (mTG) catalyzes a transamidation reaction between certain surface-exposed glutamine (Gln) and lysine (Lys) residues on protein substrates, crosslinking the two via an isopeptide bond. In addition to Lys, mTG is also known to recognize a wide range of primary amine substrates, thus providing the basis for mTG's utility to equip proteins with assorted functional moieties.[3] The enzyme, for instance, has been employed to generate protein-polymer and antibody-drug conjugates (ADCs) for pharmaceutical applications.[4]</p><p>The catalytic mechanism of mTG involves a nucleophilic attack of the target acyl donor, Gln, by the enzyme's active site cysteine to form a thioester intermediate. Subsequent nucleophilic attack of the thioester by an amine substrate serving as the acyl acceptor results in the formation of the transamidated product. Given the mechanistic role of the amine as a nucleophile and the broad substrate tolerance of the enzyme, we envisaged that the scope of the acyl acceptor substrate could be extended to include other nucleophiles, such as α-effect amines.</p><p>Indeed, evidence of hydrazine and hydrazide acting as TG substrates can be found in early studies in the 1970s that aimed to elucidate the mechanism of hydrazine-/hydrazide-functionalized agents as inhibitors of mammalian tissue TG.[5] Detection of isotopic labeling by C14-isonicotinic acid hydrazide (isoniazid) was observed for N-benzyloxycarbonyl-L-glutaminylglycine (ZQG, a common model acyl donor substrate for TG) and a variety of proteins using guinea pig liver TG. Hydrazinonaphthalazine (hydralazine) reportedly led to the same result. Meanwhile, hydroxylamine is used as a substrate in the classical hydroxamate assay used to measure TG activity.[6] Despite these early indications that TG can recognize α-effect nucleophiles, contemporary use of mTG in bioconjugation chemistry has been exclusively focused on alkyl amine substrates. Herein, we explore the expansion of the applied substrate palette to include α-effect nucleophiles (Scheme 1) and we show that the versatility of mTG can be bolstered by using these "rediscovered" substrates.</p><!><p>We first set out to verify mTG's substrate tolerance to α-effect nucleophiles, as evidence from previous studies were primarily based on mammalian TG. The ZQG acyl donor substrate was incubated with mTG and benzoic hydrazide, the latter being isostructural to isonicotinic acid hydrazide. Analysis by ESI-MS found a product with an m/z corresponding to ZQG incorporated with benzoic hydrazide (Figure 1a). The structure of the isolated product was determined by NMR (Figure S1), which indicated that the terminal amino group of the hydrazide (not the amide-like nitrogen) was the site of conjugation to the acyl group of Gln in ZQG.</p><p>Having demonstrated that benzoic hydrazide is an effective mTG substrate in a model system, we next examined whether it could serve as a substrate in a more typical site-specific protein conjugation. Given the growing prevalence of mTG in the assembly of site-specific ADC therapeutics, a trastuzumab antibody targeting the oncogenic human epidermal growth factor receptor 2 (HER2) was used as our model protein. In particular, we used a mutant, referred as Tras LC-Q, that has an LLQG tag engineered at the C-terminus of the light chain. Since native immunoglobulin G antibodies do not have an mTG-reactive Gln, enzymatic conjugation would occur site-specifically at the engineered Gln on the light chain.[4b, 4c] Indeed, treatment of Tras LC-Q with benzoic hydrazide under fairly typical mTG coupling conditions resulted in the successful conjugation to the light chain, but not the heavy chain (Figure S2), demonstrating the conserved specificity of the enzymatic reaction using the noncanonical hydrazide substrate.</p><p>Kinetics of mTG-catalyzed conjugation of benzoic hydrazide to Tras LC-Q was then compared with its amine counterpart, phenethylamine. Both demonstrated very similar rates of conjugation (Figure 1b; Figure S4). This suggests that the enhanced nucleophilicity of the hydrazide did not result in an increase in the rate of displacement of the thioester acyl-enzyme intermediate.</p><p>Next, we surveyed the structure-activity relationship (SAR) for a series of hydrazide derivatives, including alkyl hydrazide (1a-b), semicarbazide (1d-e), thiosemicarbazide (1f) and sulfonyl hydrazide (1g), as well as a few hydrazines (1h-j) and alkoxyamines (1k-l). The efficiency of conjugation of the tested substrates was calculated based on the loading on the light chain of Tras LC-Q as determined by ESI-MS after antibody reduction (Figure S3). Encouragingly, many of the assorted α-effect nucleophile substrates (1b-d, 1h, 1l) exhibited high efficiency of conjugation (>80%). Some were less efficient substrates under our initial conditions (Table 1; Condition 1). For example, a semicarbazide containing an adjacent carboxylic acid (1e) had a comparatively lower extent of conjugation (67%) than the parent semicarbazide (1d; 86%); and 4-hydrazinobenzoic acid (1i) had no apparent conjugation, in contrast to 99% for phenylhydrazine (1h). This result is consistent with previous observations that having a carboxylic acid group proximal to the amine hampers conjugation,[7] suggesting that the SAR for amine substrates may be transferrable to α-effect nucleophiles.</p><p>In order to increase the yield for the less efficient substrates, optimization was performed on a number of parameters, including pH, temperature, and the equivalents of mTG and substrate (Figure 2). Unexpectedly, reduced temperature (4°C) significantly improved the loading as compared to 22°C and 37°C. We speculate that this may be due to improved enzyme stability. Negligible effect was observed when increasing the enzyme concentration, similar to a previous observation from another study.[4d] Meanwhile, increased substrate concentration contributed positively to loading. Using thiosemicarbazide as the representative substrate for this optimization, the loading was increased from ~16% to ~95%. The optimized conditions enabled complete or near-complete loading of many of the substrates that were less efficiently conjugated under the parent conditions (Table 1; Condition 2). The exception in our list is p-toluenesulfonyl hydrazide (1g), which exhibited only 10% loading even under the optimized conditions, suggesting that (aryl) sulfonyl hydrazides are poor substrates for mTG.</p><!><p>For the typical mTG-catalyzed transamidation of Gln with an alkyl amine, the resultant linkage is an amide bond (known as an isopeptide bond). On the other hand, transamidation with α-effect amines result in the formation of isopeptide analogues, which are akin to classical peptide bond surrogates, such as hydrazinopeptide, aminoxypeptide, and azapeptide, that have been employed for peptide backbone modifications.[8] The α-heteroatom in these backbone amide surrogates can influence the electronics and the torsional preference of adjacent sites. The unique chemical and structural properties of these pseudopeptide linkages can contribute to enhanced proteolytic stability, hence their application as replacements of scissile bonds in peptide-based peptidomimetics.[9] Moreover, in a study demonstrating the use of hydrazine and hydrazide as substrates for the sortase enzyme, it was stated that the product, which bears similar pseudopeptide linkages, is no longer recognized by the enzyme and thus cannot be reversed.[10] The stability improvement associated with these "α-effect" pseudopeptide linkages therefore prompted us to explore whether the isopeptide bond analogues formed from the α-effect amine substrates (Table 1) confer similar protective ability against proteolysis, specifically by mTG itself. Although TG is known for its bond-forming reaction, it is also able to catalyze the reverse reaction.[11] In the absence of excess amine substrate, the enzyme can hydrolyze the isopeptide linkage, releasing the formerly attached amine substrate and irreversibly converting the original Gln residue to a glutamic acid. Alternatively, in the presence of an excess of another amine substrate, the enzyme can promote transamidation, or the exchange of one amine for another. We therefore examined whether the isopeptide analogues generated by mTG and α-effect amine substrates can prevent this reversal or exchange reaction, thus offering an advantage over traditional amine substrates.</p><p>To compare the stability of the different isopeptide derivatives, Tras LC-Q conjugates with phenethylamine (2a; as reference for the typical amide linkage), benzoic hydrazide (2b), benzylhydrazine (2c), phenylhydrazine (2d), and o-benzylhydroxylamine (2e) were prepared (Figure 3a). The conjugates were then subjected to incubation with mTG in phosphate-buffered saline (PBS), pH 7.4 at 37 ˚C and their integrity was monitored over time (up to five days) by ESI-MS. Reversibility was detected based on the appearance of a lower molecular weight peak corresponding to the deamidated species. The extent of hydrolysis of the conjugates was observed in the order, from most to least: benzoic hydrazide > o-benzylhydroxylamine > phenethylamine > benzylhydrazine > phenylhydrazine (Figure 3b; Figure S5). The same trend was observed when excess of a different amine substrate, dansylcadaverine, was incubated along with the conjugate and the enzyme, though instead of the deamidated species, the transamidated product with dansylcadaverine was observed (Figure 3c; Figure S5). Notably, in both scenarios of enzymatically induced hydrolysis or exchange, the two conjugates of Tras LC-Q with hydrazine substrates, benzylhydrazine (2c) and phenylhydrazine (2d), exhibited the least turnover, with the latter showing ≤ ~10% deamidation after five days under these forcing conditions. Curiously, under identical conditions but in the absence of mTG, 2d and, to a small extent, 2c, underwent hydrolysis over the five day incubation (Figure S6). The reason for this deviation is unclear given the general regard of the hydrazide linkage to be relatively stable.[12]</p><p>The stability evaluation based on Tras LC-Q conjugates 2a-2e suggests that the isopeptide subtype generated by conjugation of hydrazine substrates are less susceptible to reversibility or exchange by mTG than the classical isopeptide linkage. As enzymatic reactivity of the acyl donor substrate may differ depending on nearby residues and its local conformation,[13] we note that the trend may differ depending on the site and the protein. Nevertheless, the overall differential stabilities of the isopeptide derivatives toward hydrolysis or exchange by mTG present a new design consideration when using the enzyme as a tool to assemble bioconjugates, particularly for bioconjugates assigned to a system where mTG or its homologue is present.</p><!><p>In an early study looking at diamines as bifunctional substrates with the capacity of both ends to crosslink two Gln residues, hydrazine in its unsubstituted form was examined in the context of a diamine.[14] Using guinea pig liver TG and ZQG as the partner substrate, the authors observed both the expected hydrazide product (in which one end of hydrazine is conjugated) as well as the crosslinked dimer product. Formation of the monomer species could be favored over the dimer by increasing the concentration of hydrazine, which presumably competes with the monomer product as substrate. Therefore, in the presence of TG and sufficient excess of hydrazine, the Gln side chain could be transformed into γ-glutamic acid-hydrazide. While this previous work used mammalian TG and aimed to address a physiological question about diamines in native protein crosslinking, its results inspired us to explore the application of hydrazine as a TG substrate to introduce a side chain hydrazide moiety in peptides and proteins.</p><p>We first verified that hydrazine also acts as a substrate for mTG, resulting in the conversion of ZQG to Z(Q-hydrazide)G as was found with mammalian TG (Figure S7). Furthermore, continuing with the rationale of using difunctional hydrazines to convert the carboxamide group of Gln into a hydrazide, the simplest dihydrazides, namely carbohydrazide and thiocarbohydrazide, were also explored as substrates (Figure 4a). In both cases, ZQG was converted to the respective Z(Q-carbohydrazide)G and Z(Q-thiocarbohydrazide)G as confirmed by ESI-MS (Figure S8–9).</p><p>Next, the ability to install the various hydrazides on a protein was examined, again using Tras LC-Q. Full modification of the engineered Gln on the light chain was observed using unsubstituted hydrazine as substrate (Figure S12). For carbohydrazide, in addition to the modified light chain, a species with a mass equivalent to a dimer of two light chains crosslinked through one molecule of carbohydrazide was observed on ESI-MS (Figure S10). As expected based on the previous study,[14] formation of the functionalized monomer can be maximized by increasing the carbohydrazide concentration (Figure S10–11). Meanwhile, derivatization with thiocarbohydrazide can be achieved using the optimized conditions in Table 1 (Condition 2) (Figure S12).</p><!><p>Installation of the reactive hydrazide handle can be used in a number of downstream applications. For one, it can be applied as a bioorthogonal handle for coupling with aldehydes or ketones.[15] The resultant acid-labile hydrazone is well known for serving as a pH-sensitive linker in triggered release systems. To demonstrate their application in bioconjugation, Tras LC-Q derivatized with an alkyl hydrazide (3a), carbohydrazide (3b), or thiocarbohydrazide (3c) was allowed to react with 10 eq. of an aldehyde fluorophore (coumarin aldehyde) at pH ~7.4 in the presence of 4-amino-phenylalanine as a catalyst for the reaction (Figure 4a).[16] Reducing SDS-PAGE analysis showed that only the modified light chain was fluorescently labeled, consistent with the chemoselective nature of the reaction (Figure 3b, lanes 1–4). Tras LC-Q-hydrazide (3a) and -carbohydrazide (3b) had similar coupling efficiency (~70% based on ESI-MS; Figure S14). Tras LC-Q-thiocarbohydrazide (3c) had a relatively low conjugation yield (~30%), at least under the unoptimized condition; for this reason, we focus our discussion on 3a and 3b. However, we note that the thiocarbohydrazide handle may be useful for other applications (vide infra).</p><p>Periodic reimaging of the gel stored in acidic destaining solution (10% acetic acid, 50% methanol) showed that the fluorescence of the bands corresponding to the conjugates of 3a and 3b decreased over time (Figure 4b, lanes 2 & 3), indicating hydrolysis of the hydrazone conjugates in the acidic milieu. Importantly, fluorescence of the acyl hydrazone conjugate (lane 2) diminished more quickly than the carbohydrazone (lane 3), suggesting that the carbohydrazone linkage is more hydrolytically stable. Our observation is in line with the reportedly shorter half-life of acyl hydrazone compared to semicarbazone,[17] the latter sharing resemblance to a carbohydrazone. This indicates that mTG can be used to incorporate cleavable linkers with varied rates of hydrolysis. This can allow for the fine-tuning of the cleavage rate, which is an important design factor for ADCs that contain such hydrolytic linkages.[18]</p><p>Although the reversible hydrazone construct has been leveraged to enable controlled payload delivery, it is not ideal for other applications where the stability of the linkage is essential. Moreover, hydrazone formation is known to be slow unless catalyzed or performed under acidic condition.[15] Recent developments in a variation of the classical aldehyde/ketone condensation involving placement of a boronic acid ortho to an aromatic aldehyde or ketone offer opportunities to improve both product stability and reaction kinetics.[19] Intramolecular Lewis acid catalysis from the ortho-boronic acid increases the condensation rate by several orders of magnitude compared to simple carbonyls. In particular, the reaction between 2-formylphenylboronic acid (2fPBA) and a hydrazide rapidly forms a hydrazone as an intermediate that further cyclizes into a boron-nitrogen heterocycle.[20] Stability of this 2,3,1-benzodiazaborine (DAB) product can vary by changing the nature of the hydrazide, with certain substituted hydrazides and semicarbazides capable of forming stable DABs that are suited to serve as robust coupling scaffolds.[20–21] To extend the utility of the mTG-modified protein-hydrazides in areas that are limiting for conventional hydrazone conjugates, bioconjugation using the 2fPBA-based subtype of carbonyl chemistry was also examined.</p><p>Tras LC-Q-hydrazide (3a), -carbohydrazide (3b), and -thiocarbohydrazide (3c) were allowed to react with 5 eq. of a 2fPBA-functionalized fluorophore, Texas Red-2fPBA (Tx Red-2fPBA), at pH 7.4 (Figure 4a). Reducing SDS-PAGE analysis showed that, once again, only the modified light chain was fluorescently labeled (Figure 4b, lanes 5–8; Figure S15). In-gel fluorescence of the DAB conjugate of Tras LC-Q-hydrazide (lane 6) decreased over time when the gel was kept in acidic destaining solution, consistent with our previous finding that the DAB product of 2fPBA and an alkyl hydrazide is acid-labile.[20] On the other hand, the fluorophore-labeled conjugate of Tras LC-Q-carbohydrazide (3b) showed a negligible difference over time (up to two weeks), demonstrating remarkable stability (Figure 4b, lane 7). Notably, the DAB conjugate of 3b persisted even when the hydrazone conjugate with coumarin aldehyde had disassembled. (compare lane 2–3 with lane 7). This demonstrates that the carbohydrazide-2fPBA ligation results in a DAB linkage with stability that exceeds the hydrolytic limit of typical hydrazones.</p><p>To ensure that the bioconjugate generated via the chemoenzymatic process retains the function of the biomolecule, Tras LC-Q-carbohydrazide coupled to a 2fPBA-functionalized fluorophore (Janelia Fluor 669–2fPBA) was prepared (Figure S17) and tested for its functional activity in antigen binding. SK-BR-3 (HER2+) and MDA-MB-231 (HER2−) breast cancer cells were treated with the antibody-fluorophore conjugate (AFC). Clear fluorescence localization at the cell membrane of SK-BR-3 cells, but not MDA-MB-231 cells, was observed by fluorescence microscopy (Figure 4c), demonstrating the unperturbed function of the AFC in binding its HER2 target on the cell surface.</p><!><p>It is worth noting that isopeptide hydrazides may find applications in areas beyond bioorthogonal conjugations. For example, the mTG-derived hydrazide may serve as a stable precursor of a thioester for native chemical ligation (NCL), commonly employed in protein semisynthesis. Oxidation of the hydrazide to an acyl azide or conversion to an acyl pyrazole results in an intermediate that can undergo thioesterification and subsequent NCL for coupling with an N-terminal cysteine peptide/protein.[22] A number of strategies have thus been developed to functionalize peptides/proteins with a hydrazide, specifically at the C-terminus of the protein backbone.[10, 23] Indeed, the protein synthesis application demonstrated for these methods makes it appropriate to have the hydrazide placed at the C-terminus, but they also limit the scope of protein architectures to only the canonical linear scaffolds. Methods to introduce the hydrazide group on the side chain at an internal amino acid residue is also desirable, as it can allow the thioester to be generated on the side chain. This can enable synthetic access to unique branched or lariat protein structures, which may endow novel functions to the engineered peptide/protein. Our method of introducing a thioester synthon in the form of a hydrazide at the side chain thus provides a facile enzyme-mediated approach that supplements the current toolbox for creating unique peptide/protein scaffolds and biomaterials.</p><p>The introduction of a hydrazide on the protein also complements existing strategies for functionalization with aldehydes and ketones.[24] In addition to simply switching the two reactive handles, placement of the hydrazide on the protein can have its own value as well. The hydrazide can pair with a carbonyl tailored to have certain motif, such as a suitably positioned atom with the capability of hydrogen bonding or coordination interaction. Acyl hydrazones with such functions can serve as molecular switches[25] or ligands for metals[26] that can be potentially integrated into a protein. Importantly, the benefit of hydrazide functionalization at the side chain of an internal amino acid residue enables the functionality to be built into internal sites. Furthermore, the ability to introduce other hydrazide derivatives, including carbohydrazide and thiocarbohydrazide, enables the generation of hydrazone variants that may have different isomerization (switching) propensity or metal affinity.[27] Together, the assorted hydrazide functionalities that can be installed onto the protein by mTG may provide a novel synthetic route supporting the de novo design of protein switches and metalloproteins.</p><!><p>In summary, we have demonstrated that α-effect nucleophiles can be used as acyl acceptor substrates for mTG, thus significantly expanding the substrate palette for this widely used bioconjugation technology. The distinct isopeptide linkages that result from conjugation of hydrazines, hydrazides, and alkoxyamines are reminiscent of the pseudopeptides used to modulate the proteolytic stability of peptidomimetics. Their differential stability profiles toward hydrolysis or exchange by mTG may be a pertinent factor to consider in the design of new bioconjugates, in particular for situations where the bioconjugate may encounter mTG or its homologues.</p><p>In addition to providing diversity in the linkages generated via direct conjugation, the expanded substrate repertoire also offers a way to directly install a reactive functional group for chemoenzymatic conjugation. Unsubstituted hydrazine and dihydrazides, which are widely accessible and inexpensive, can be used as substrates for mTG to install assorted hydrazide derivatives in peptides or proteins. The introduced hydrazide handle can subsequently participate in bioorthogonal couplings with conventional aldehydes/ketones or ortho-carbonylphenylboronic acids to form dynamic or stable bioconjugates.</p><p>Importantly, the distinguished side chain position at which the hydrazide is presented enables functionalization of an internal amino acid residue, which complements existing methods that typically introduce the hydrazide at the C-terminus or the complementary carbonyl group. As such, the direct introduction of hydrazide can potentially open doors to unique applications in protein semisynthesis and de novo protein design. The diversification of the substrate scope of mTG with α-effect amines therefore further unravels the versatility of this important enzyme and invites the possibility of continued expansion of the applied substrate repertoire with other nucleophilic moieties.</p>

PubMed Author Manuscript

Bioorthogonal Tetrazine Carbamate Cleavage by Highly Reactive Trans-Cyclooctene

The high reaction rate of the 'click-to-release' reaction between allylic substituted trans-cyclooctene and tetrazine has enabled exceptional control over chemical and biological processes. Here we report the development of a new bioorthogonal cleavage reaction based on trans-cyclooctene and tetrazine with up to 3 orders of magnitude higher reactivity compared to the parent reaction, and 4 to 6 orders higher than other cleavage reactions. In this new pyridazine elimination mechanism, wherein the roles a reversed, a trans-cyclooctene activator reacts with a tetrazine that is substituted with a methylene-linked carbamate, leading to an 1,4-elimination of the carbamate and liberation of an amine. Through a series of mechanistic studies, we identified the 2,5-dihydropyridazine tautomer as the releasing species and found factors that govern its formation and subsequent fragmentation. The bioorthogonal utility was demonstrated by the selective cleavage of a tetrazine-linked antibody-drug conjugate by trans-cyclooctenes, affording efficient drug liberation in plasma and cell culture. Finally, the parent and the new reaction were compared at low concentration, showing that the use of a highly reactive trans-cyclooctene as activator leads to a complete reaction with antibody-drug conjugate in seconds vs. hours for the parent system. We believe that this new reaction may allow markedly reduced click-to-release reagent doses in vitro and in vivo and could expand the application scope to conditions wherein the trans-cyclooctene has limited stability.

bioorthogonal_tetrazine_carbamate_cleavage_by_highly_reactive_trans-cyclooctene

Introduction<!>Results and Discussion<!>Conclusion

<p>Bioorthogonal cleavage reactions have emerged as powerful strategies to control the release or activation of small molecules and biomolecules in chemical and biological settings. [1][2][3] Most organic cleavage reactions were derived from their click conjugation counterparts and include the reactions between tetrazines and vinyl ethers, [4][5][6] vinylboronic acids, 7 3-isocyanopropyls, 8 cyclooctynes 9 and benzonorbornadienes, 10 the iminosydnone cyclooctyne reaction, 11,12 the azide-to-amine reduction by trans-cyclooctene (TCO), 13,14 in addition to the use of the Staudinger reaction 15,16 and ligation. 17 We reported that the fastest bioorthogonal conjugation reaction, the inverse-electron-demand Diels-Alder (IEDDA) between TCO and tetrazine derivatives, 18 widely used for selective and efficient bioconjugations in vitro and in vivo, 19,20 could be transformed into a bioorthogonal cleavage reaction. 21 In this IEDDA pyridazine elimination reaction, termed 'click-to-release' (Scheme 1A), a carbamate-linked payload is installed on the allylic position of TCO. Following reaction of the TCO-carbamate with a tetrazine, the resulting 1,4-dihydropyridazine intermediate rapidly eliminates the amine-containing payload and CO2.</p><p>The high reactivity and selectivity of the IEDDA pyridazine elimination reaction has led to its widespread application, such as in in vivo cleavage or unmasking of TCO-containing antibody-drug conjugates (ADCs), 22,23 prodrugs, 24,25 proteins, 26,27 and peptide antigens, 28 by the administration of a tetrazine activator. In addition, this click-to-release approach has been used in a range of diverse in vitro applications, such as uncaging of fluorogenic compounds 29,30 and enzyme substrates, 31 cell-specific proteome labelling, 32 oligonucleotide delivery into cells, 33 and purification of solid phase synthesized oligonucleotides. 34 Nevertheless, a further increase of the click-to-release reaction rate would be beneficial for a number of applications. For example, complete in vivo activation of a target-localized protein or ADC requires the intravenous administration of a large excess of tetrazine activator. A higher click reaction rate may allow a lower dose of the activator, which would facilitate clinical translation and may open up other prodrug approaches. Likewise, a higher reactivity would enable in vitro assays that use low concentrations. Here, the TCO is a limiting factor, as it has a reduced reactivity due to the allylic-positioned payload and as it needs to remain stable for hours or days when used as a linker or mask. The latter precludes increasing the reactivity by designing more strained TCO derivatives as these will likely become too unstable. 35 Furthermore, TCOs in general and highly strained TCOs in particular do not combine well with high thiol concentrations, low pH or UV light, which may affect their application scope in, for example, in vitro assays or chemistry.</p><p>We therefore set out to develop a new click-to-release strategy, still based on the robust IEDDA reaction between TCO and tetrazine and the unique and versatile properties of the dihydropyridazine intermediates, but now with the TCO being the activator and the tetrazine the linker. In such a system, wherein the roles are reversed, the TCO does not require the allylic substitution, and its reactivity can be boosted further as the stability requirements for the activator are less stringent than for the linker, which typically has to withstand harsher conditions and/or remain intact for a longer time. Another advantage of such a system would be that relatively simple, even commercially available, TCOs could be used. IEDDA reactions give 4,5-dihydropyridazines (DHPs), which usually rapidly tautomerize to 1,4-and 2,5-dihydropyridazines. 36,37 We envisioned that the dihydropyridazine IEDDA intermediates, of which the 1,4-dihydropyridazine leads to an electron cascade elimination of a carbamate from the part originating from the allylic position on the TCO, 21,31 may also be enlisted for an electron cascade elimination of a carbamate from the part originating from the tetrazine. Specifically, we hypothesized that a reaction of a TCO with a tetrazine substituted with a methylene-linked carbamate and subsequent tautomerization of the 4,5-dihydropyrididazine (4,5-DHP) to the 1,4-and 2,5-dihydropyridazines (1,4-DHP and 2,5-DHP) would lead to an 1,4-elimination of the carbamate from either 1,4-DHP or 2,5-DHP, or both, liberating the amine, CO2 and pyridazine P2 (Scheme 1B). 38 Here we show that the dihydropyridazine product from the reaction between trans-cyclooctene and tetrazine can indeed be enlisted to induce an 1,4-elimination of a methylene-linked carbamate on the tetrazine. Through a series of mechanistic studies we found that 4,5-DHP preferentially tautomerizes to 2,5-DHP and that, fortuitously, 2,5-DHP is the releasing species. Depending on the TCO used the reactivity increased 6-and 800-fold compared to the parent pyridazine elimination reaction, and affording release yields of 67 to 93 %. The bioorthogonality of the system was demonstrated in the context of an antibodydrug conjugate comprising the new tetrazine linker, which could be efficiently reacted and cleaved in biological conditions.</p><!><p>We commenced with the preparation of a range of model 1,2,4,5-tetrazines with methylene carbamate derivatives on the 3-position, or both the 3-and 6-positions, comprising benzylic or aromatic amines with and without alkyl substituents on the amine and the methylene bridge (Figure 1A). These compounds were then screened for stability and TCO 12-triggered carbamate cleavage in 20 % acetonitrile (ACN) / phosphate buffered saline (PBS) at 37 °C (Figure 1B,C). To our delight, already among the first three prepared compounds, symmetrical tetrazines 1-3, we observed good TCO-triggered cleavage and reasonable stability for 3. While compounds 1 and 2 showed release as well, they were highly unstable, indicating a destabilizing effect from the primary carbamate NH on the tetrazine. Based on these results, we designed non-symmetrical model compounds 4-10 that could form the basis of a click-cleavable linker. Perusal of Figure 1C shows triggered cleavage half-lives between 1.5 h and 10 h and a correlation between release rate and stability half-lives. Installing a methyl substituent on the methylene appears to facilitate the cleavage (5 vs. 4, and 2 vs. 1). This is in line with structure activity relationships established for the widely used self-immolative para-aminobenzyloxycarbonyl linker wherein a methyl substituent on the methylene bridge leads to more efficient liberation due to stabilization of the increasing positive charge upon release of the carbamate. 39,40 Furthermore, benzylamine-derived tetrazine carbamates were more stable than aniline-derived carbamates (5 vs. 7, 6 vs. 8) and both carbamate types could be stabilized further by replacing the N-methyl by the bulkier N-isopropyl substituent (6 vs. 5, 8 vs. 7). However, comparing 5 with 9 suggests that the increased steric bulk of a branched vs. linear alkyl substituent on the 6-position of the tetrazine does not further improve the stability.</p><p>On the contrary, the analogous tetrazine 10 with a phenyl substituent showed a 6-fold higher stability. While this was accompanied by a slower release than the isopropyl analog 5, we believed the enhanced stability was more important for most applications, and we continued our investigations with this tetrazine motif.</p><p>To further study the release, we prepared 11 (Figure 2A), the dimethylamine analog of 10, and started by monitoring its TCO-triggered cleavage in a range of ACN/PBS ratios with liquid chromatography-mass spectrometry (LC-MS). To facilitate analysis, reaction mixture aliquots were rapidly oxidized to give stable mixtures of (aromatic) pyridazines P1 and P2 (Figure 2A), the ratio of which affording the release yield. The structure of P2 was confirmed by reference compound synthesis (see Supporting Information), indicating that 1,4-elimination of the carbamate is followed by hydration of the exocyclic double bond of the elimination product (EP, Figure 2A, 2D). While the IEDDA cycloaddition between 11 and 12 was instantaneous in all mixtures, the ensuing dimethylamine elimination from the dihydropyridazine intermediate showed a strong correlation with the water content (Figure 2B). Reaction in 25 % ACN/PBS afforded dimethylamine liberation with a half-life of 20h, but in 50 % ACN/PBS, the release slowed down to a 54h half-life and, interestingly, no cleavage was observed in 100 % ACN. This trend, together with the 10h release half-life observed in 20 % ACN for the product of 10 and 12 (Figure 1C), indicated that the release half-life in fully aqueous conditions was likely to be less than 3h.</p><p>Subsequently, the reaction between 11 and 12 was studied in CDCl3 with 1 H NMR, UV-Vis and IR, showing instantaneous formation of the initial 4,5-dihydropyridazine product (4,5-DHP) with -max of 279 nm as a mixture of two diastereomers (83 vs. 17 %), arising from the stereocenter on the methylene (Figure 2E, F, Supporting Information Section S3.2). In these conditions, 4,5-DHP did not tautomerize further and did not release dimethylamine. Addition of 0.1 v/v% formic acid resulted in complete tautomerization 31,41 within 3h to give a mixture in a 73/27 ratio of the 2,5-and 1,4-dihydropyridazines (2,5-DHP and 1,4-DHP), each present as two diastereomers, as demonstrated by the loss of peaks at 7.9 and 3.2 ppm and the appearance of 4 new sets of peaks between 6.0 and 5.4 ppm and 3.8 and 3.1 ppm (Figure 2E; see Figure S6 for 2D NMR characterization). This tautomerization was accompanied by the appearance of an IR resonance at 3340 cm -1 for the N-H moiety (Figure S9), and the appearance of a UV absorbance at 340 nm in addition to the band at 280 nm (Figure 2F). To understand this, the UV spectra of the three tautomers were simulated using time-dependent density functional theory (TDDFT) in conjunction with implicit solvent effects employing the COSMO model with parameters for water, leading to a predicted -max of 276, 290/307 and 346 nm for respectively the 4,5-, 2,5 and the 1,4-tautomer, matching the trend in the previous observations (Figure 3). Extending the incubation in acidic conditions to 184h resulted in conversion of 2,5-DHP into 1,4-DHP, affording a 26/74 mixture and a concomitant increase of the 340/280 nm absorbance ratio, as predicted (Figure 2E, F). Also some oxidation to the aromatic pyridazine (P1) occurred, similar to the previously reported pyridazine elimination reaction. 21,31 Even though all three tautomers could be formed in CDCl3, no release was observed, which is possibly due to the need to stabilize the charges on the carbamate and methylene that develop upon release. 39,40 Alternatively, protonassisted mechanisms wherein water acts as the proton carrier can be a critical pathway in the cleavage mechanism. The ability to control the tautomerization in chloroform was subsequently applied to identify the releasing tautomer by comparing the release profiles when starting from 4,5-DHP, or a composition rich in respectively 2,5-DHP or 1,4-DHP. This was achieved by starting the reaction between 11 and 12 in CDCl3 and by controlling the tautomer formation as shown above. Samples with the desired tautomer were concentrated and then dissolved in 25 % ACN/PBS, incubated at 37C, followed by oxidation of aliquots and evaluation of the P1/P2 ratio with LC-MS (Figure 2C). Starting from 4,5-DHP led to a steady release of eventually 85 % with a half-life of the maximum cleavage yield of 14h, similar to when the release was started directly in 25 % ACN/PBS (Figure 2B). Interestingly, starting from a 70/30 mixture of 2,5-/1,4-DHP afforded a much faster liberation, with a half-life of the maximum cleavage yield of 3h, leading to a maximum cleavage of 71 %. On the contrary, starting from a 24/76 mixture of 2,5-/1,4-DHP led to a reduced maximum release yield of only 26 %. The strong correlation between the 2,5-DHP percentage and release yield as well as the marked release rate increase when starting from 2,5-DHP clearly indicates this is the releasing species. To confirm this, we set out to identify the non-releasing tautomer by incubating tetrazine 11 and TCO 12 in 25 % ACN/PBS at 37C for 3 days. Given the release rate observed in these conditions (see Figure 2B) it follows that the releasing tautomer(s) would have to be consumed at that time. Indeed, subsequent lyophilization and NMR in CDCl3 clearly showed that 1,4-DHP was still present, together with some oxidized P1, but the signals belonging to the 2,5-DHP had disappeared, confirming this is the releasing species (Figure 2G). In line with this and the established UV signatures of the tautomers, LC-MS of non-oxidized aliquots from of the 70/30 mixture of 2,5-/1,4-DHP in 25 % ACN/PBS, showed the two tautomers as adjacent peaks with respectively a -max of 278 and 327 nm, with the 278 nm peak disappearing in time, in conjunction with the formation of hydrated elimination product (EP-H2O) (Figure 2H and D). To confirm that the formation of EP-H2O and the corresponding P2 is the result of 1,4elimination followed by addition of water to the exocyclic double bond of EP, instead of hydrolysis of the carbamate bond of 2,5-DHP, 11 and 12 were incubated in 25 % ACN/PBS containing the N-acetyl cysteine (NAC) to trap EP. As expected, LC-MS of the reaction mixture clearly showed the formation of the cysteine adduct EP-NAC in addition to EP-H2O (Figure 2D). The foregoing demonstrates that in aqueous conditions the 4,5-DHP predominantly tautomerizes to the 2,5-DHP and to a minor extent to the 1,4-DHP, of which 2,5-DHP then liberates the carbamate. Furthermore, the formation of EP-H2O, EP-NAC and P2 indicates that this release occurs via the envisioned 1,4-elimination. To further support the mechanistic findings we calculated the minimal energies of 4,5-DHP, 2,5-DHP and 1,4-DHP (Figure 3) and several other possible tautomers of the reaction product of 11 and 12, including the 1,2-dihydropyridazine and its exocyclic analog (Table S1). Using a B3LYP exchange-correlation functional and water as an implicit solvent, the 2,5-DHP and 1,4-DHP were indeed shown to have the lowest energy, with the 1,4-DHP being slightly lower. Calculations in vacuum showed essentially the same result, but with a slightly smaller energetic difference between the two tautomers (Table S1). These values, combined with the NMR study (Figure 2E), suggest that upon 4,5-DHP tautomerization 2,5-DHP is kinetically favoured and can slowly convert to the thermodynamically favoured 1,4-DHP. The high release yield in 25 % ACN/PBS indicates that this 2,5-to 1,4-tautomerization does not readily occur in neutral aqueous conditions.</p><p>Having identified the releasing tautomer, we evaluated the potential reactivity advantage of this system. Click reaction of 11 with 12 in ACN afforded a k2=3.14 +/-0.10 M -1 s -1 at 20C, which is already 6-fold faster than the parent reaction of allylic substituted TCO with 3,6-bisalkyl-tetrazine. 21 Conformationally strained TCO (sTCO) based on a cis-fusion of a cyclopropyl to the TCO ring represents the most reactive TCO that still has good utility in bioorthogonal conjugations. 42 Reaction of 11 with this much more reactive sTCO-acid 14 was found to have a very high k2= 420 ± 49 M -1 s -1 , 800-fold higher than the parent click-torelease reaction (Figure 2I). 21 As expected, in 25 % ACN/PBS these rates increased further to 287 ± 10 M -1 s -1 for 12 and to the very high rate of 23800 ± 400 M -1 s -1 for 14.</p><p>To demonstrate the proof of principle of this novel click-to-release reaction in a biological environment, we developed an antibody-drug conjugate (ADC) comprising the drug monomethyl auristatin E (MMAE) linked via the tetrazine to a pegylated CC49 diabody that targets tumor associated glycoprotein 72 (TAG72). 23 Based on model compound 11 we prepared tetrazine linker 18 via a modified Pinner synthesis (Figure 4A) followed by PNP carbonate formation and introduction of the MMAE. After Boc-deprotection, tetrazine-MMAE 21 was conjugated to PEG derivative 22, affording maleimide-functionalized Tz-MMAE linker 23. Linker-drug 23 was then site-specifically conjugated to four engineered cysteine residues in the CC49 diabody providing tz-ADC with a drug-to-antibody ratio (DAR) of 4 (61 kDa, Figure 4B). Tz-ADC exhibited excellent stock stability (PBS, 4°C) as no tetrazine degradation or spontaneous drug liberation was observed in 1 year (Figure S22). In addition to TCO 12 and sTCO-acid 14 we also prepared the DOTA chelate-conjugated analogs 13 and 15, for increased hydrophilicity and to be able to monitor their reaction with tz-ADC by labeling the chelate with a radiometal (Figure 4D). The four TCOs were subsequently examined for their ability to liberate MMAE from the tz-ADC (Figure 4C,E-G). For all TCOs, mass spectrometry of the diabody in PBS showed efficient conversion of the ADC (30691 Da for the scFv monomer, each monomer linked to 2 MMAE) to the IEDDA product, followed by formation of species that had eliminated 1 or 2 MMAE moieties. Whereas TCO 12 and its DOTA analog 13 afforded a MMAE release yield of 93 %, the sTCO motif had a lower maximal cleavage yield of ca. 67 % (Figure 4F). Similar results were obtained in mouse plasma, while only low levels of free MMAE were found when the ADC was incubated without a TCO (Figure 4G). We hypothesized that the lower release from sTCO may be caused by a slower tautomerization of the 4,5-DHP derivative, which may allow for more oxidative deactivation to P1. Indeed, when we used UV to evaluate the tautomerization of the 4,5-DHP product from the reaction between 3-methyl-6-phenyl-tetrazine and, respectively TCO 12 and sTCO ethyl ester 16 in 25 % ACN/PBS, we found a markedly slower tautomerization for the sTCO (Figure 2J). Despite the lower overall release from sTCO-derived 2,5-DHP, the remarkable reactivity increase offered by these TCOs provides a compelling argument for their use when the application is relatively demanding (e.g. at low concentrations). Furthermore, the cleavage yield is on par with the release observed for the widely used 3-pyrimidyl-6-methyl-tetrazine based activators 29,30 and TCO-linked ADC (data not shown). ADCs allow the safe and effective use of highly potent drugs like MMAE, which are too toxic to be used as free drugs. After ADC targeting of the cancer cell, these drugs typically need to be cleaved from their antibody linker to exert their therapeutic effect. As a proof of principle of the TCO-cleavable tz-ADC we evaluated the cytotoxicity of tz-ADC and TCO 15 alone and in combination in a human colorectal cancer cell culture. TCO 15 alone was not toxic while tz-ADC alone only exhibited a relatively moderate activity (Figure 4H). However, when a fixed dose of 3.3 μM of TCO 15 was combined with the tz-ADC, the cytotoxicity increased 100-fold, affording an EC50 value of 0.67 nM, matching the toxicity of the parent drug MMAE, clearly underlining the efficacy of this new cleavage system. Finally, to demonstrate the reactivity advance offered by click-to-release from tetrazine vs. click-to-release from TCO, we compared the click reaction between a previously reported TCO-linked MMAE diabody ADC (tco-ADC) 23 and 177 Lulabelled DOTA-tetrazine activator (S16) with the reaction of tz-ADC with 111 In-labelled DOTA-sTCO 15 in PBS at 37C at a very low concentration of 0.6 µM in nearly equimolar conditions. Figure 4I shows the striking difference between the two systems, with the conjugation of 15 complete within 1 minute, while the conjugation of the tetrazine activator had a 3h half-life.</p><!><p>We have developed a new and highly reactive bioorthogonal elimination reaction that enables traceless release of an amine-containing payload from a tetrazine following reaction with a trans-cyclooctene. This was achieved by switching the roles of the tetrazine and TCO in the parent IEDDA pyridazine elimination, now permitting the use of highly reactive sTCO derivatives as activators, boosting the reactivity 3 orders of magnitude compared to the parent click-to-release reaction, and 4 to 6 orders of magnitude compared to the other known cleavage reactions. [4][5][6][7][8][9][10][11][12][13][14][15][16][17] Even the non-conformationally strained TCOs already gives 6fold higher reactivity than the parent reaction combined with near quantitative release yields. Through mechanistic studies we demonstrated that the releasing tautomer is the 2,5-dihydropyridazine, which fortuitously is the favoured IEDDA product, and we found that the formation rate of this species is dependent on the type of TCO being used. By trapping the elimination product, we could show that the release occurs via the hypothesized 1,4-elimination. Furthermore, the strong dependence on water indicates a direct role for water in the electron cascade elimination mechanism. The new click-cleavable linker was subsequently applied in an ADC, which exhibited TCO-triggered drug cleavage half-lives down to 2h in fully aqueous conditions and showed good controlled release in biological conditions in plasma and in cell culture. Activation of the ADC led to effective cancer cell killing with the same potency as the free drug. Furthermore, a head-to-head comparison of the click reaction of TCO-linked ADC with radiolabelled tetrazine and tetrazine-linked ADC with radiolabelled TCO at very low concentration, underlined the pronounced advance offered by the highly reactive TCO-triggered dihydropyridazine elimination, for example for cases when the activator cannot be used in large excess.</p><p>These results hold promise for in vivo drug release and unmasking applications, potentially allowing substantially reduced activator doses. We also envision the use of this reaction in chemical biology, lifescience assays, and material chemistry, enabling the controlled (dis)assembly of molecules, proteins, cells, or biomaterials at low concentration or in conditions that are incompatible with the TCO. For example, contrary to TCOs, tetrazines are typically stable at low pH. While the click-to-release from TCO offers faster release kinetics, we expect that the versatile click-to-release from tetrazine harbours ample opportunity for further improvements in cleavage rate and yield. Finally, the presented release chemistry can be performed with commercially available TCOs and uses tetrazine linkers and masks that are easier to prepare than TCO linkers, making highly reactive click-to-release chemistry available to a larger research community.</p>

ChemRxiv

Design, synthesis and biological evaluation of edaravone derivatives bearing the N-benzyl pyridinium moiety as multifunctional anti-Alzheimer’s agents

AbstractA series of multi-target directed edaravone derivatives bearing N-benzyl pyridinium moieties were designed and synthesised. Edaravone is a potent antioxidant with significant neuroprotective effects and N-benzyl pyridinium has previously exhibited positive results as part of a dual-site binding, peripheral anionic site (PAS) and catalytic anionic site (CAS), acetylcholinesterase (AChE) inhibitor. The designed edaravone-N-benzyl pyridinium hybrid compounds were docked within the AChE active site. The results indicated interactions with conserved amino acids (Trp279 in PAS and Trp84 in CAS), suggesting good dual-site inhibitory activity. Significant in vitro AChE inhibitory activities were observed for selected compounds (IC50: 1.2–4.6 µM) with limited butyrylcholinesterase inhibitory activity (IC50’s >160 µM), indicating excellent selectivity towards AChE (SI: 46 – >278). The compounds also showed considerable antioxidant ability, similar to edaravone. In silico studies indicated that these compounds should cross the blood–brain barrier, making them promising lead molecules in the development of anti-Alzheimer’s agents.

design,_synthesis_and_biological_evaluation_of_edaravone_derivatives_bearing_the_n-benzyl_pyridinium

Introduction<!><!>Introduction<!><!>Chemistry<!>Synthesis of 4–(3-Methyl-5-oxo-4H-pyrazol-1-yl)-N-(pyridine-4-ylmethyl)benzamide (3)<!>General procedure for the synthesis of compounds 5a–l<!>1-Benzyl-4-({[4–(3-methyl-5-oxo-4H-pyrazol-1-yl)phenyl]foramido}methyl)pyridin-1-ium bromide (5a)<!>1-[(2-Fluorophenyl)methyl]-4-({[4–(3-methyl-5-oxo-4H-pyrazol-1-yl)phenyl]foramido} methyl)pyridin-1-ium bromide (5b)<!>1-[(3-Fluorophenyl)methyl]-4-({[4–(3-methyl-5-oxo-4H-pyrazol-1-yl)phenyl]foramido} methyl)pyridin-1-ium bromide (5c)<!>1-[(4-Fluorophenyl)methyl]-4-({[4–(3-methyl-5-oxo-4H-pyrazol-1-yl)phenyl]foramido} methyl)pyridin-1-ium bromide (5d)<!>1-[(2-Chlorophenyl)methyl]-4-({[4–(3-methyl-5-oxo-4H-pyrazol-1-yl)phenyl]foramido} methyl)pyridin-1-ium bromide (5e)<!>1-[(3-Chlorophenyl)methyl]-4-({[4–(3-methyl-5-oxo-4H-pyrazol-1-yl)phenyl]foramido} methyl)pyridin-1-ium bromide (5f)<!>1-[(2-Bromophenyl)methyl]-4-({[4–(3-methyl-5-oxo-4H-pyrazol-1-yl)phenyl]foramido} methyl)pyridin-1-ium bromide (5g)<!>1-[(3-Bromophenyl)methyl]-4-({[4–(3-methyl-5-oxo-4H-pyrazol-1-yl)phenyl]foramido} methyl)pyridin-1-ium bromide (5h)<!>1-[(4-Bromophenyl)methyl]-4-({[4–(3-methyl-5-oxo-4H-pyrazol-1-yl)phenyl]foramido} methyl)pyridin-1-ium bromide (5i)<!>4-({[4–(3-Methyl-5-oxo-4H-pyrazol-1-yl)phenyl]foramido}methyl)-1–(2-methylphenyl)-pyridin-1-ium bromide (5j)<!>4-({[4–(3-Methyl-5-oxo-4H-pyrazol-1-yl)phenyl]foramido}methyl)-1–(3-methylphenyl)-pyridin-1-ium bromide (5k)<!>4-({[4–(3-Methyl-5-oxo-4H-pyrazol-1-yl)phenyl]foramido}methyl)-1–(4-methylphenyl)-pyridin-1-ium bromide (5l)<!>Ache molecular docking studies<!>Cholinesterase inhibition assay<!>Antioxidant assay<!>In silico blood–brain permeability predictions<!>Results and discussion<!><!>Results and discussion<!><!>Results and discussion<!><!>Results and discussion<!><!>Results and discussion<!>Conclusions<!>

<p>Alzheimer's disease (AD) is a progressive neurodegenerative disorder that is mainly prevalent in the older population (>65 years of age)1–3. Approximately fifty million people are diagnosed with dementia, with AD accounting for 60–70% of these cases4. The disease can be characterised by an array of symptoms which include; memory loss, cognitive impairment, behavioural and psychiatric abnormalities3. Due to the complex and multifactorial nature of AD, the exact aetiology of the disease is unknown. Multiple pathways and hypotheses have been indicated in the pathology of the disease such as the Aβ cascade-, cholinergic- and the oxidative stress hypotheses1,3,5.</p><p>Senile plaques are considered a pathological hallmark of AD. The primary constituent of these senile plaques is Aβ and are believed to play a central role in the pathogenesis of the disease6,7. The amyloid cascade hypothesis suggests that one of the main driving forces behind AD development is the buildup and deposition of Aβ peptide aggregation within the brain8,9. Recently, it was discovered that the amyloid precursor protein (APP) gene undergoes mutations that induce an increase in Aβ formation. The APP mutations are situated near the sites where proteases, β and γ-secretase, cleave the APP. These mutations result in the favouring of the Aβ1–40 and Aβ1–42 peptide fragment formation7,8,10. Aβ1–40 and Aβ1–42 are more inclined to self-aggregate to form amyloid beta fibrils. With the persistent imbalance of the production and clearance of the Aβ fragments, the consequential result is the genesis of insoluble senile plaques. These senile plaques result in the blockage of parenchymal spaces between neurons in the brain leading to eventual neuronal cell death8,9.</p><p>The cholinergic hypothesis describes that the hydrolysis of the neurotransmitter acetylcholine (ACh) by cholinesterases, acetylcholinesterase (AChE) and butyrylcholinestrase (BuChE), leads to a drastic decrease in ACh levels. The loss of cholinergic transmission due to the decreased levels of ACh has been correlated with loss of memory and cognitive ability11. AChE is the predominant enzyme that hydrolyse ACh in the healthy brain12. The AChE enzyme contains a pocket with two binding sites, the catalytic anionic site (CAS) and peripheral anionic site (PAS). Interactions with both these sites are crucial for the inhibition of AChE activity and potential neuroprotective effects13. The PAS possesses a non-cholinergic role that through protein–protein interactions, bind to and promotes the formation and deposition of insoluble Aβ fibrils leading to neurotoxicity. Recent studies have shown that the inhibition in the PAS did not only improve the memory in a transgenic APP/PS1 murine model, but also significantly stemmed the amount of Aβ plaques in the brain14–16.</p><p>Reactive oxygen species (ROS) are known to play a significant role in the progression of neurodegenerative disorders such as AD17–19. The most significant ROS include hydroxyl radicals, superoxide anions and peroxyl radicals19. As ROS begins to accumulate and antioxidant levels begin to reduce, detrimental effects in the brain begin to occur. This process is further increased with aging17,19. The brain is especially susceptive to the neurotoxic effects of ROS due to its high demand for oxygen as well as the large amounts of peroxide susceptible lipid cells18. ROS have also been observed to cause disruptions in neuronal cell integrity and to modify and inactivate several proteins that are important for glucose metabolism and ATP synthesis resulting in mitochondrial dysfunction10,18. The occurrence of the neurotoxic effects of ROS in the development of AD, coupled with the presence of Aβ, supports the role of oxidative stress in the pathogenesis of AD18,20,21. Several lines of evidence have revealed a connection between oxidative stress and Aβ formation10,19,22. Aβ exhibits the ability to enhance the formation of ROS and vice versa22,23. Aβ produces ROS through the promotion of oxidative modification and inhibition of important transmembrane transports systems within the neuronal and glial cells, Aβ-induced lipid peroxidation and protein oxidation22,24,25. In addition, ROS also stimulate the enhanced activity of proteases, β and γ-secretase, which increases the formation of Aβ1–40 and Aβ1–4223.</p><p>At present, there are no therapeutic agents that are able to reverse, halt or slow the progression of the disease and the current options are only able to treat AD symptomatically2,11,26–28. All of these treatments follow the much researched "one-molecule–one-target" drug discovery approach with minimal success. Therefore, more researchers are exploring the development of multi-target directed ligands (MTDL)26. MTDLs are conceived from the molecular hybridisation of various pharmacophoric moieties from recognised bioactive compounds. The MTDLs are designed to interact with multiple targets involved in the multifactorial pathogenesis of AD. The rational decision to combine these pharmacophores results in greater selectivity for the targets of AD, leads to fewer side effects and potentially improves the compounds' neuroprotective abilities29,30.</p><p>Edaravone (Figure 1(a)) is a potent free radical scavenger used to treat acute cerebral infarction in Japan31. In addition, edaravone has also exhibited beneficial neuroprotective effects in amyotrophic lateral sclerosis (ALS) and Parkinson's disease animal models32. Edaravone's neuroprotective effects are believed to be caused by its ability to scavenge ROS. The decrease in ROS levels in turn reduces oxidative stress and oxidative damage to neuronal cells33,34. In previous studies, edaravone has demonstrated the ability to attenuate Aβ-induced oxidative stress and neurotoxicity, inhibit Aβ aggregation, disaggregate preformed Aβ fibrils and attenuate downstream pathologies including tau-hyperphosphorylation, neuroinflammation and neuronal cell loss31,35.</p><!><p>The two moieties combined to synthesise the novel MTDLs in this study. (a) Edaravone. (b) R-substituted N-benzyl pyridinium.</p><!><p>In search for potentially potent and selective AChE inhibitors, benzyl pyridinium salts have been extensively researched (Figure 1(b))36–39 and N-benzyl pyridinium moieties have demonstrated excellent activity against AChE. Previous research has found that the best AChE inhibitory activity is reached when the N-benzyl pyridinium moiety is bound to another privileged molecule, using the MTDL strategy, to form a dual-site (PAS and CAS) binding compound37,38. Substitutions, e.g. halogens and methyl groups, at various positions on the benzyl group of the moiety has demonstrated improved AChE inhibitory activity compared to an unsubstituted benzyl ring36,40.</p><p>Thus, we describe here the docking, synthesis and biological evaluation of new edaravone-N-benzyl pyridinium hybrid compounds (Figure 2). These compounds are expected to exhibit strong dual-site AChE inhibitory activities and significant antioxidant capacity, which could lead to promising MTDL neuroprotective effects.</p><!><p>Edaravone-N-benzyl pyridinium hybrid compounds designed and evaluated in this study. R = H, Br, F, Cl, or CH3.</p><!><p>All the reagents used to synthesise the desired compounds were acquired from Sigma-Aldrich® or Industrial Analytical (Pty) Ltd. All the reagents were used without further purification. Solvents used in the synthesis and purification of the compounds were obtained from a variety of commercial sources. Thin-layer chromatography (TLC) was used to monitor all reactions and was carried out on 0.20 mm thick aluminium silica gel sheets (TLC silica gel 60 F245 Merck KGaA). Visualisation of the samples was achieved using UV light (254 nm and 366 nm) and iodine vapours. Mobile phases were prepared on a volume-to-volume basis. Infra-red spectra were acquired using a Perkin Elmer Spectrum 400 spectrometer. The IR spectrometer was equipped with a diamond attenuated total reflectance (ATR) attachment. The spectra were then acquired from PerkinElmer, Inc. Spectrum version 10.5.4 IR software. The MS spectra of the compounds were acquired from a Waters SYNAPT G2 high resolution mass spectrometer. The melting points of the samples were acquired using a Lasec Melting Point SMP 10 apparatus and capillary tubes. Proton (1H) and carbon (13 C) spectra were acquired using a Bruker Avance IIIHD Nanobay 400 MHz instrument that is equipped with a 5 mm BBO probe. Tetramethylsilane (TMS) was used as the internal standard and deuterated dimethyl sulfoxide (DMSO-d6) as the deuterated solvent. Chemical shifts (δ) and coupling constants (J) were reported in parts per million (ppm) and hertz (Hz) respectively. The internal standard (δ = 0 ppm) and DMSO-d6 (δ = 2.5 ppm) were used as the reference peaks. The multiplicities of the respective signals were indicated using the following abbreviations: s – singlet, d – doublet, t – triplet, m – multiplet. The atom numbering of the target compounds used for 1H NMR data are depicted on each respective compound found in the supplementary data.</p><!><p>The 4-(aminomethyl)pyridine moiety was conjugated to the carboxylic group of 1 via HATU activational chemistry. One equiv. edaravone-COOH (1) and four equiv. of N,N-diisopopylethylamine (DIPEA) was stirred at room temperature for 20 min. Thereafter, the carboxylic acid of 1 was activated using the HATU activational agent in a 1 equiv.:1.2 equiv. ratio in an appropriate quantity of dimethylformamide (DMF). The mixture was stirred at room temperature for 1 h and monitored using TLC (3 ethanol: 2 ethyl acetate: 4 diethyl ether). Once the reaction was complete, 4-(aminomethyl) pyridine (2) was added to the mixture and stirred under reflux at 40–50 °C for 1 h and monitored using TLC. Once the reaction was complete, toluene was added to the mixture in a ratio of 3 equiv. toluene: 1 equiv. DMF. The reaction was then rotary evaporated until just off dry. The mixture was left to precipitate out overnight in a refrigerator. Finally, the precipitate was filtered and washed with distilled water. The precipitate was placed in a vacuum oven and allowed to dry rendering the desired compound 3.</p><p>Physical data: Yield: 72.52%; light pink solid; mp: 234 °C; Rf: 0.45; 1H NMR: (400 MHz, DMSO-d6), δH: 9.10–9.13 (t, 1H, J = 5.60, 5.83, 11.43 Hz, H – 12), 8.50–8.51 (d, 2H, J = 5.88 Hz, H—16, 17), 7.95–7.98 (d, 2H, J = 8.68 Hz, H—6, 10), 7.85–7.87 (d, 2H, J = 8.64, H—7, 9). 7.30–7.32 (d, 2H, J = 5.76 Hz, H—15, 18), 4–49-4.51 (s, 2H, H—13), 2,13 (s, 3H, H—1); 13 C NMR: (400 MHz, DMSO-d6): 166.36, 149.98, 149.15, 128.66, 122.62, 120.10, 42.21, 14.70; IR: (FT-IR, cm−1): 3217, 3035, 1713, 1637; MS: (HR-ESI+), [M + H+], m/z: calcd.: 309.1273, found: 309.1347.</p><!><p>Compound 3 (1 equiv.) and 1.3 equiv. of the respective substituted benzyl bromide derivative (4) were dissolved and stirred under reflux, at 40–50 °C, in 5–6 ml of DMF. The compounds were monitored using TLC (3 ethanol: 2 ethyl acetate: 4 diethyl ether) for 4–6 h. Once the reaction was complete, 3 equiv. toluene: 1 equiv. DMF was added. The mixture was then rotary evaporated to dryness. Thereafter, 20 ml of diethyl ether was added to the dried mixture. The mixture was then left to precipitate out overnight in a refrigerator. Thereafter, if solid, the precipitate was filtered off and washed with diethyl ether. If the precipitate exhibited a waxy/oily appearance the mixture was diluted in a minimal amount of ethanol and transferred into a polytop. Finally, the precipitate or waxy/oily substance was dried, rendering the desired compounds 5a–l.</p><!><p>Physical data: Yield: 95.29%; light grey solid; mp: 225 °C; 1H NMR: (400 MHz, DMSO-d6), δH 9.33–9.36 (t, 1H, J = 5.60, 5.72, 11.32 Hz H—12), 9.11–9.12 (d, 2H, J = 6.32 Hz, H—16, 17), 8.06–8.07 (d, 2H, J = 6.24 Hz, H—15, 18), 7.97–7.99 (d, 2H, J = 8.60 Hz, H—6, 10), 7.87–7.89 (d, 2H, J = 8.64 Hz, H—7, 9), 7.43–7.53 (m, 5H, H—21, 22, 23, 24, 25), 5.83 (s, 2H, H—19), 4.72–4.74 (d, 2H, J = 5.28 Hz, H—13), 2.13 (s, 3H, H—1); 13 C NMR: (400 MHz, DMSO-d6): 166.68, 162.79,160.74, 159.84, 144.80, 134.87, 129.83, 129.72, 129.59, 129.23, 128.96, 128.84, 126.55, 119.36, 117.36, 63.14, 42.80, 17.19; IR: (FT-IR, cm−1): 3184, 3034, 1717,1639; MS: (HR-ESI+), [M-Br]+, m/z: calcd.: 399.1815, found: 399.1820.</p><!><p>Physical data: Yield: 97.17%; black solid; mp: 191 °C; 1H NMR: (400 MHz, DMSO-d6), δH: 9.33–9.36 (t, 1H, J = 5.72, 5.83, 11.55 Hz, H—12), 9.03–9.04 (d, 2H, J = 6.44 Hz, H—16, 17), 8.06–8.08 (d, 2H, J = 6.52 Hz, H—15, 18), 7.87–8.00 (m, 4H, H—6, 7, 9, 10), 7.58–7.62 (t, 1H, J = 7.68, 7.56, 15.24, Hz, H—22), 7.51–7.55 (m, 1H, H—23), 7.34–7.35 (d, 1H, J = 1.88 Hz, H—24), 7.30–7.32 (d, 1H, J = 7.96 Hz, H—25), 5.93 (s, 2H, H—19), 4.74–4.75 (d, 2H, J = 5.40 Hz, H—13), 2.14 (s, 3H, H—1); 13 C NMR: (400 MHz, DMSO-d6): 166.69, 162.78, 162.20, 161.08, 145.01, 132.65, 132.57, 131.96, 128.83, 126.53, 125.81, 125.78, 121.87, 121.72, 117.37, 116.62, 116.41, 57.83, 42.84; IR: (FT-IR, cm−1): 3201, 3034, 1717, 1638; MS: (HR-ESI+), [M-Br]+, m/z: calcd.: 417.1721found: 417.1719.</p><!><p>Physical data: Yield: 98.48%; brown solid; mp: 220 °C; 1H NMR: (400 MHz, DMSO-d6), δH: 9.33–9.36 (t, 1H, J = 5.60, 5.83, 11.43 Hz, H—12), 9.11–9.13 (d, 2H, J = 6.64 Hz, H—16, 17), 8.07–8.08 (d, 2H, J = 6.56 Hz, H—15, 18), 7.87–8.00 (m, 4H, H—6, 7, 9, 10), 7.48–7.54 (m, 1H, H—23), 7.45–7.48 (d, 1H. J = 10.2, H—21), 7.37–7.39 (d, 1H, J = 7.72 Hz, H—25), 7.26–7.31 (m, 1H, H—24), 5.85 (s, 2H, H—19), 4.73–4.75 (d, 2H, J = 5.32, H—13), 2.14 (s,3H, H—1); 13 C NMR: (400 MHz, DMSO-d6): 166.68, 163.88, 161.44, 160.89, 144.89, 137.26, 137.18, 131.91, 131.82, 128.84, 126.58, 125.48, 125.45, 119.39, 117.36, 116.87, 116.47, 62.34, 42.82, 17.18; IR: (FT-IR, cm−1): 3201, 3036, 1716, 1638; MS: (HR-ESI+), [M-Br]+, m/z: calcd.: 417.1721, found: 417.1732.</p><!><p>Physical data: Yield: 87.13%; light grey solid; mp: 215 °C; 1H NMR: (400 MHz, DMSO-d6), δH: 9.32–9.35 (t, 1H, J = 5.60, 5.80, 11.43 HZ, H—12), 9.09–9.10 (d, 2H, J = 6.64 Hz, H—16, 17), 8.05–8.07 (d, 2H, J = 6.48 Hz, H—15, 18), 7.87–7.99 (m, 4H, H—6, 7, 9, 10), 7.61–7.64 (m, 2H, H—22, 25), 7.28–7.32 (t, 2H, J = 8.80, 17.60 Hz, H—21, 25), 5.81 (s, 2H, H—19), 4.72–4.73 (d, 2H, J = 5.44 Hz, H—13), 2.13 (s, 3H, H—1); 13 C NMR: (400 MHz, DMSO-d6): 166.67, 164.27, 161.82, 160.76, 144.72, 131.93, 131.85, 131.09, 131.06, 128.94, 128.83, 126.55, 117.36, 116.74, 116.53, 62.29, 42.81, 17.19; IR: (FT-IR, cm−1): 3201, 3036, 1716, 1638; MS: (HR-ESI+), [M-Br]+, m/z: calcd.: 417.1721, found: 417.1729.</p><!><p>Physical data: Yield: 96.53%; black solid; mp: 211 °C; 1H NMR: (400 MHz, DMSO-d6), δH: 9.36–9.38 (t, 1H, J = 5.60, 5.72, 11.32 Hz, H—12), 9.00–9.02 (d, 2H, J = 6.32 Hz, H—16, 17), 8.07–8.09 (d, 2H, J = 6.28 Hz, H—15, 18), 7.98–8.01 (d, 2H, J = 8.64 Hz, H—6, 10), 7.88–7.90 (d, 2H, J = 8.68 Hz, H—7, 9), 7.59–7.61 (d, 1H, J = 7.44 Hz, H—22), 7.47–7.54 (m, 3H, H—23, 24, 25), 5.97 (s, 2H, H—19), 4.76–4.77 (d, 2H, J = 5.32 Hz, H—13), 2.13 (s, 3H, H—1); 13 C NMR: (400 MHz, DMSO-d6): 166.71, 162.79, 161.17, 145.19, 133.77, 132.09, 131.98, 130.61, 129.58, 128.96, 128.84, 128.67, 126.42, 119.38, 117.37, 61.11, 42.83, 17.81; IR: (FT-IR, cm−1): 3212, 3034, 1711, 1638; MS: (HR-ESI+), [M-Br]+, m/z: calcd.: 433.1425, found: 433.1441.</p><!><p>Physical data: Yield: 21.31%; black solid mp: 228 °C; 1H NMR: (400 MHz, DMSO-d6), δH: 9.30–9.33 (t, 1H, J = 5.60, 5.83, 11.43 Hz, H—12), 9.09–9.11 (d, 2H, J = 6.6 Hz, H—16, 17), 8.05–8.07 (d, 2H, J = 6.44 Hz, H—14, 18), 7.96–7.98 (d, 2H, J = 8.72 Hz, H—6, 10), 7.87–7.89 (d, 2H, J = 8.76 Hz, H—7,9), 7.69 (s, 1H, H—21), 7.47–7.52 (t, 3H, J = 13.62, 6.22, 19,84 Hz, H—23, 24, 25), 5.81 (s, 2H, H—19), 4.73–4.74 (d, 2H, J = 5.40 Hz, H—13), 2.13 (s, 3H, H—1); 13 C NMR: (400 MHz, DMSO-d6): 166.69, 160.90, 144. 89, 137.00, 134.16, 131.62, 129.83, 129.31, 128.83, 128.05, 126.60, 62.31, 42.84, 14.62; IR: (FT-IR, cm−1): 3213, 3035, 1710, 1639; MS: (HR-ESI+), [M-Br]+, m/z: calcd.:433.1425, found: 433.1430.</p><!><p>Physical data: Yield: 94.2%; black solid; mp: 204 °C; 1H NMR: (400 MHz, DMSO-d6), δH: 9.53–9.37 (t, 1H, J = 5.60, 5.83, 11.43 Hz, H—12), 8.98–9.00 (d, 2H, J = 6.52 Hz, H—16, 17), 8.07–8.09 (d, 2H, J = 6.40 Hz, H—15, 18), 7.87–8.00 (m, 4H, H—6, 7, 9, 10), 7.76–7.78 (d, 1H, J = 7.60 Hz, H—25), 7.49–7.53 (t, 1H, J = 7.06, 7.40, 14.46 Hz, H—22), 7.36–7.44 (m, 2H, H—23, 24), 5.94 (s, 2H, H—19), 4.76–4.78 (d, 2H, J = 5.44, H—13), 2.13 (s, 3H, H—1); 13 C NMR: (400 MHz, DMSO-d6): 166.71, 162.77, 161.21, 145.24, 133.91, 133.56, 132.04, 131.98, 129.18, 128. 96, 128.84, 126.41, 123.98, 63.17, 42.84, 14.62; IR: (FT-IR, cm−1): 3217, 3034, 1711, 1638; MS: (HR-ESI+) [M-Br]+, m/z: calcd.: 477.0920, found: 477.0928.</p><!><p>Physical data: Yield: 17.73%; dark grey solid; mp: 230 °C; 1H NMR: (400 MHz, DMSO-d6), δH: 9.30–9.33 (t, 1H, J = 5.60, 5.72, 11.32 Hz, H—12), 9.09–9.10 (d, 2H, J = 6.4 Hz, H—16, 17), 8.05–8.07 (d, 2H, J = 6.32 Hz, H—15, 18), 7.96–7.98 (d, 2H, J = 8.60 Hz, H—6, 10), 7.87–7.89 (d, 2H, J = 8.64 Hz, H—7, 9), 7.83 (s, 1H, H—21), 7.63–7.65 (d, 1H, J = 8.00 Hz, H—23), 7.52–7.54 (d, 1H, J = 7.76 Hz, H—25), 7.39–7.54 (t, 1H, J = 7.74, 7.90, 15.64 Hz, H—24), 5.80 (s, 2H, H—19), 4.73–4.74 (s, 2H, J = 5.36 Hz, H—13), 2.13 (s, 3H, H—1); 13 C NMR: (400 MHz, DMSO-d6): 166.69, 160.89, 144.88, 137.24, 132.74, 132.15, 131.86, 128.83, 128.43, 126.60, 122.72, 62.26, 42.84; IR: (FT-IR, cm−1): 3216, 3035, 1711, 1638; MS: (HR-ESI+), [M-Br]+, m/z: calcd.:477.0920, found: 477.0940.</p><!><p>Physical data: Yield: 97.42%; black solid; mp: 190 °C; 1H NMR: (400 MHz, DMSO-d6), δH: 9.33–9.36 (t, 1H, J = 5.95, 11.90 Hz, H—12), 9.08–9.10 (d, 2H, J = 6.80 Hz, H—16, 17), 8.06–8.07 (d, 2H, J = 6.64 Hz, H—15, 18), 7.95–7.99 (m, 2H, H—6, 10), 7.89–7.87 (d, 2H, J = 8.88 Hz, H—7, 9), 7.65–7.68 (d, 2H, J = 8.40 Hz, H—22, 24), 7.48–7.50 (d, 2H, J = 8.44 Hz, H—21, 25), 5.81 (s, 2H, H—8), 4.72–4.74 (d, 2H, J = 5.52 Hz, H—13), 2.13 (s, 3H, H—1); 13 C NMR: (400 MHz, DMSO-d6): 166.67, 162.77, 160.85, 144.84, 134.11, 132.64, 131.58, 128.96, 128.84, 126.56, 123.34, 62.32, 42.80; IR: (FT-IR, cm−1): 3217, 3034, 1715, 1638; MS: (HR-ESI+), [M-Br]+, m/z: calcd.: 477.0920, found: 477.0943.</p><!><p>Physical data: Yield: 21.75%; light grey solid; mp: 194 °C; 1H NMR: (400 MHz, DMSO-d6), δH: 9.34–9.37 (t, 1H, J = 5.60, 5.83, 11.43 Hz, H—12), 8.93–8.95 (d, 2H, J = 6.64 Hz, H—16, 17), 8.06–8.08 (d, 2H, J = 6.56 Hz, H—15, 18), 7.98–8.00 (d, 2H, J = 8.80 Hz, H—6, 10), 7.87–7.89 (d, 2H, J = 8.80 Hz, H—7, 9), 7.20–7.37 (m, 3H, H—23, 24, 25), 7.12–7.14 (d, 1H, J = 7.52 Hz, H—26), 5.88 (s, 2H, H—19), 4.75–4.76 (d, 2H, J = 5.40 Hz, H—13), 2.29 (s, 3H, H—22), 2.13 (s, 3H, H—1); 13 C NMR: (400 MHz, DMSO-d6): 166.70, 162.77, 160.81, 144.98, 137.38, 132.82, 131.43, 129.87, 129.57, 128.95, 128.83, 127.20, 126.49, 117.37, 61.36, 42.81, 19.23, 17.19; IR: (FT-IR, cm−1): 3215, 3035, 1715, 1639; MS: (HR-ESI+), [M-Br]+, m/z: calcd.: 413.1972, found: 413.1968.</p><!><p>Physical data: Yield: 98.3%; black powder; mp: 212 °C; 1H NMR: (400 MHz, DMSO-d6), δH: 9.31–9.34 (t, 1H, J = 5.60, 5.83, 11.43 Hz, H—12), 9.08–9.09 (d, 2H, J = 6.64 Hz, H—16, 17), 8.04–8.06 (d, 2H, J = 6.52 Hz, H—15, 18), 7.96–7.98 (d, 2H, J = 8.88 Hz, H—6, 10), 7.87–7.89 (d, 2H, J = 8.80 Hz, H—7, 9), 7.33–7.35 (d, 1H, J = 5.24 Hz, H—21), 7.29–7.31 (t, 2H, J = 2.35, 7.40, 9.75 Hz, H—24, 25), 7.23–7.24 (d, 1H, J = 6.84 Hz, H—26), 5.77 (s, 2H, H—19), 4.72–4.73 (d, 2H, J = 5.44 Hz, H—13), 2.30 (s, 3H, H—23), 2.13 (s, 3H, H—1); 13 C NMR: (400 MHz, DMSO-d6): 166.68, 160.68, 144,76, 139.10, 134. 76, 130.46, 129.75, 129.63, 128.83, 126.52, 126.32, 63.19, 42.81, 21.38, 17.1; IR: (FT-IR, cm−1): 3208, 3034,1716, 1638; MS: (HR-ESI+), [M-Br]+, m/z: calcd.:413.1972, found: 413.1973.</p><!><p>Physical data: Yield: 60.13%; light grey solid; mp: 230 °C; 1H NMR: (400 MHz, DMSO-d6), δH: 9.30–9.33 (t, 1H, J = 5.60, 5.83, 11.43 Hz, H—12), 9.06–9.08 (d, 2H, J = 6.6 Hz, H—167, 17), 8.03–8.05 (d, 2H, J = 6.52 Hz, H—15, 18), 7.96–7.98 (d, 2H, J = 8.68 Hz, H—6, 10), 7.86–7.89 (d, 2H, J = 8.76 Hz, H—7, 9), 7.41–7.43 (d, 2H, J = 7.96 Hz, H—21, 26) 7.24–7.26 (d, 2H, J = 7.88 Hz, H—22, 25), 5.76 (s, 2H, H—19), 4.71–4.72 (d, 2H, J = 5.44 Hz, H—13), 2.29 (s, 3H, H—24), 2.13 (s, 3H, H—1); 13 C NMR: (400 MHz, DMSO-d6):166.67, 160.64, 144.67, 139.89, 131.89, 130.24, 129.29, 128.82, 126.50, 63.01, 42.81, 21.21, 17.1; IR: (FT-IR, cm−1): 3224, 3032, 1714, 1638; MS: (HR-ESI+), [M-Br]+, m/z: calcd.: 413.1972, found: 413.1974.</p><!><p>Molecular Operating Environment (MOE) 2018.10 software package41 was employed to predict the interactions and binding modes of the novel compounds within the active site of TcAChE. The co-crystallised structure of TcAChE with donepezil (PDB accession code: 1EVE) was acquired from the Protein Data Bank (PDB)38. The following protocol was employed, as previously described42, to simulate the orientation and binding interactions of the test compounds. Firstly, the test compounds were drawn using ChemSketch v2019.2.1 and saved as a mol files. Secondly, the enzyme structures were inspected for missing atoms, bonds and contacts. Thirdly, partial charges and hydrogens were added using MOEs' protonate 3 D application. Fourthly, the ligands were assembled employing the builder module in MOE and energy minimisation (MMFF94x) was performed. Thereafter, the ligands were docked, using the MOEdock application, within the AChE active site. Finally, the retained best poses, as per their binding affinity scores, were inspected visually and analysis of the interactions within the active aromatic gorge of AChE was conducted. To determine the accuracy of this docking protocol, the co-crystallised ligand, was re-docked into the AChE active site. This procedure was repeated three times and the best ranked solution exhibited an RMSD value of less than 2.0 Å from the position of the co-crystallised ligand. The RMSD value in this case is smaller than 2.0 Å indicating that the docking protocol is capable of accurately predicting the binding orientation of the co-crystallised ligand43. This protocol was thus deemed to be suitable for the docking of inhibitors into the active site model of AChE.</p><!><p>A modified Ellman's method was employed to determine the ChE inhibitory activities of the synthesised compounds44. eeAChE, eqBuChE, 5,5′-dithiobis-(2-nitrobenzoic acid) (DTNB, commonly known as Ellman's reagent), S-butrylthiocholine iodide (BTCI), acetylthiocholine iodide (ATCI) and donepezil were purchased from Sigma-Aldrich®. eeAChE, eqBuChE, DTNB, BTCI and ATCI were diluted with a buffer solution (tris hydrochloride (50 mM), pH 8). Each well of a 96 well plate, contained the following; 148 µL of 1.5 mM DTNB, 50 µL of either 0.22 U/ml eeAChE or 0.12 U/ml eqBuChE and 2 µL of either test compound, control (donepezil) or blank [dimethyl sulfoxide (DMSO)]. The test compounds and control were dissolved in DMSO and added to the well to yield various concentrations (1000 µM, 100 µM, 10 µM, 1 µM, 0.1 µM and 0.01 µM). The quantity of DMSO per well accumulated to below 0.01%45. Each concentration of test compound and control was conducted in triplicate to ensure consistent results. The well-plates were then incubated at 25 °C for 10 min. Thereafter, 30 µL of either the ATCI or BTCI substrate was added to each well. The plate was then placed inside a Rayto 6500 spectrometer and the absorbance was read at 405 nm every 45 s for 5 min. The percentage activity was calculated using the following equation: [(absorbance of blank − absorbance of test compounds)/absorbance of blank × 100)]. All data analysis was conducted on GraphPad Prism version 8.2.1 for Mac OS, GraphPad Software (San Diego, CA, USA)</p><!><p>Antioxidant activity was studied with the DPPH+ free radical scavenging assay46,47. DPPH+ and trolox were purchased from Sigma-Aldrich®. The following method was employed, as previously described48, to determine the antioxidant activity of the test compounds. Each well of a 96 well plate, contained the following: 180 µL of 0.12 mM DPPH+ dissolved in methanol and 20 µL of either test compound, control (trolox) or blank (DMSO). The test compounds and control were dissolved in DMSO and added to the well to yield various concentrations (1000 µM, 100 µM, 10 µM, 1 µM and 0.1 µM). Each concentration of test compound and control was conducted in triplicate to ensure consistent results. The well plate was incubated at 25 °C for 30 min within a dark space. A change of colour from dark purple to light yellow was observed. The quantity of DMSO per well accumulated to below 0.01%45. The plate was then placed inside a Rayto 6500 spectrometer and the absorbance was read at 517 nm three times to ensure statistical viability. The percentage activity was calculated using the following equation: [(absorbance of blank − absorbance of test compounds)/absorbance of blank × 100)]. All data analysis was conducted on GraphPad Prism version 8.2.1 for Mac OS, GraphPad Software (San Diego, California, USA).</p><!><p>An in silico model was used to determine the blood–barrier permeability of the synthesised compounds. The BBB predictor used can be found on an intergrade cloud computing server called AlzPlatform49. The BBB predictor was designed to determine whether a ligand is permeable across the blood brain barrier (BBB+) or not (BBB−). The BBB predictor was developed by applying the LiCABEDS and support vector machine (SVM) algorithms on four types of fingerprints of 1593 reported compounds50–52. The BBB predictor software employed is available at http://www.cbligand.org/BBB/.</p><!><p>Previous studies have shown that molecules that contain the N-benzyl pyridinium moiety interact with the CAS of the AChE enzyme36,37,40. Therefore, to determine if the edaravone portion of the hybrid molecules would exhibit the proposed interactions with the PAS as well as potential interactions exhibited by the N-benzyl pyridinium moiety with the CAS, molecular docking studies were performed. The molecular docking studies were conducted using the Molecular Operating Environment (MOE) 2018.10 software package53. The co-crystallised structure of TcAChE with donepezil (PDB accession code: 1EVE)38 was utilised to establish the starting model for AChE active site docking54. The findings from these studies were then used to rationalise the synthesis and evaluation of these compounds as potential MTDLs.</p><p>Results show that the majority of the hybrid compounds exhibit potential interactions with important conserved residues within the PAS site (Figures 3 and 4, and Supplementary Material). In general, the compounds interacted with a combination of important residues within the PAS, which included Trp 279 and Tyr 33440,55. The compounds were also observed in close proximity to Arg 289. The Arg 289 residue is found in site I that is part of one of the four putative binding sites within the PAS that was shown to play an important role in Aβ formation38. It has also previously been reported that Trp 279 plays a role in Aβ formation54,56,57. These findings support the hypothesis that the edaravone-N-benzyl pyridinium hybrid compounds could inhibit AChE and significantly reduce the formation of AChE induced Aβ plaques.</p><!><p>The active site cavity of AChE exhibiting the binding and interactions of representative compound 5f (Scheme 1). (a) Two-dimensional (2D) representation of the docked compound 5f. The close proximity of the Arg 289 residue to edaravone's pyrazoline ring can be observed. (b) Three-dimensional (3D) representation of the docked compound 5f, showing the orientation and positing of 5f within the AChE active site cavity.</p><p>The active site cavity of AChE exhibiting the binding and interactions of representative compound 5c (Scheme 1). (a) 2 D representation of the docked compound 5c. Interactions with Trp 279 (PAS) and Trp 84 (CAS) can be observed. (b) 3D representation of the docked compound 5c, showing the orientation and positing of 5c within the AChE active site cavity.</p><!><p>In general, the hybrid compounds exhibited interactions with the AChE mid-gorge recognition site residues Phe 330 and Phe 331 (Figures 3 and 4, and Supplementary Material)55. These interactions are π–π interactions with the pyridinium ring and the carbon linker between the pyridinium ring and amide. It has also previously been observed that the interaction between the pyridinium ring and Phe 330 leads to stabilisation in the orientation of a compound. This in turn, increases the probability for an interaction with Trp 279 within the PAS37. The benzyl moiety of the hybrid molecules was mainly found to interact with the residue Trp 84 within the CAS55. Trp 84 is a key residue found within the CAS and is responsible for the molecular recognition of cationic substrates in cholinesterases. Previous studies have shown that the replacement of residue Trp 84 with alanine caused a drastic decrease in human AChE's ability to hydrolyse acetylthiocholine57. Interactions were also observed with His 440, a residue of the catalytic triad found in the CAS and with Glu119, found in the oxyanion site of the CAS. All these residues play a variable role in the inhibitory activity of AChE55. From these results, it is clear that the majority of the hybrid compounds exhibit similar orientational and positional conformations within the AChE active site. The only exception was found in compound 5i (Scheme 1 and Supplementary Material) in that its most stable conformations were flipped within TcAChE active site. The pyrazoline of the edaravone moiety in 5i interacted with Trp 84 through H–π stacking and the pyridinium ring interacted with Trp 279 through π–π stacking as well as H–π stacking. The intermediate 3, lacking the benzyl moiety (Scheme 1 and Supplementary Material), interacts with residues Phe 288 and Trp 84 present in the mid-gorge and CAS respectively, and is not able to travers into the PAS of the AChE active site. This indicates the potential importance of the benzyl moiety for optimal dual site binding AChE inhibitory activity in these compounds. These results therefore indicate that the docked compounds could exhibited the potential to act as dual-site AChE inhibitors.</p><!><p>Synthesis pathway of the edaravone-N-benzyl pyridinium derivatives 5a–5l. Reagents and conditions: (i) HATU, DMF, DIPEA, 2 h, stirring under reflux. (ii) DMF, stirring under reflux, 4–6 h at 40–50 °C. (1) Edaravone-COOH, (2) 4-(aminomethyl) pyridine, (4) R-benzyl bromide derivatives.</p><!><p>The synthesis of the novel MTDLs was carried out in two steps (Scheme 1). Firstly, an amide intermediate (3) was formed by reacting edaravone-COOH (1) with 4-(aminomethyl) pyridine (2) via 1-[bis(dimethylamino)methylene]-1H-1,2,3-triazolo[4,5-b]pyridinium-3-oxid hexafluoro-phosphate (HATU) activational chemistry. Thereafter, the final edaravone-N-benzyl pyridinium derivatives (5a–5l), containing the pyridinium moiety, were synthesised via N-benzylation of compound 3 with benzyl bromide containing various substitutions (H, Br, F, Cl or CH3) at different positions of the benzene ring.</p><p>Analytical instruments and techniques, described in the experimental section, were employed for structural elucidation of the synthesised compounds. All of the protons and carbons of the aromatic groups, CH2 groups and CH3 moiety of the pyrazoline ring, were observed for all the final compounds and exhibited similar peaks on the 1H- and 13 C-NMR spectra (Supplementary Material). The only variance in NMR peaks were as a result of the different functional group substitutions on the benzyl moiety. The formation of the positively charged nitrogen of the pyridinium group in 5a–5l, can be observed by a shift of the pyridine 1H NMR doublet peaks from ∼ 8.5 ppm and ∼7.3 ppm to ∼9.1 ppm and ∼8.0 ppm respectively when compared to the NMR of intermediate 3. HRMS also confirmed the molecular masses and identity of the synthesised compounds. Refer to the supplementary material for all NMR spectra and a complete discussion on NMR and HRMS results.</p><p>Edaravone is known to exist in different neutral tautomeric forms58. In all 1H NMR spectra where DMSO-d6 was used additional peaks, visible at δ = 5.4 ppm and 11.7 ppm, were observed suggesting that the novel synthesised compounds may also exist in different tautomeric states (Figures 5 and 6, and Supplementary Material). The peaks, at δ = 5.4 ppm and δ = 11.7 ppm, were found to belong to the respective CH-group and NH-group of the amine tautomer portion of the edaravone of the synthesised compounds59. In previous research, it has been found that the amine tautomer of edaravone may be more stable in aprotic polar solvents such as DMSO compared to the keto tautomer59,60. The NMR experiment was repeated using the protic polar solvent, methanol-d4. Using this solvent system, the keto tautomer was found to be more stable as no peaks were present at δ = 5.4 ppm and δ = 11.7 ppm (Figure 6). This finding is in accordance to previous research as it has been found that the enol and keto tautomeric forms are both found within this solvent system. In addition, the keto tautomer was found to be the predominant of the two tautomeric forms in methanol61. The enol tautomer 1H NMR peak for the CH-group at δ = ∼6.2 ppm62 was not present and further confirms the presence of the keto tautomer within the methanol-d4 solvent system. The formation of the edaravone tautomers are dependent on certain solvation effects and electrostatic interactions between the solvent and molecule and further studies are to be conducted to explore the tautomeric nature of these hybrid molecules. Based on the data in this study, the keto tautomer of these hybrid molecules seems to be more stable in polar protic solvents whereas its amine tautomeric form is more stable in polar aprotic solvents. As the biological evaluations on these compounds were carried out in protic environments it is expected that the keto tautomer will be the predominant form present. The effects of the different tautomers on the biological profile of these compounds should therefore be taken into consideration for future pharmaceutical development.</p><!><p>Two major tautomeric forms of the edaravone-N-benzyl pyridinium hybrid compounds and their respective 1H NMR chemical tautomeric shifts in deuterated solvents.</p><p>Overlaid 1H NMR spectra of compound 5g in methanol-d4 (blue) and DMSO-d6 (red). (a) Singlet that represents the CH-group of the amine tautomer.</p><!><p>The inhibitory activities of the synthesised compounds were evaluated using eeAChE (electric eel) and eqBuChE (equine serum) according to a modified method of Ellman44. Donepezil, a known potent ChE inhibitor, was used as the reference compound for both assays. This reference compound was chosen as the structure of donepezil contains similarities to the N-benzyl pyridinium moiety of the synthesised compounds. Compounds 5b–5g exhibited some of the best inhibitory activities (Table 1). These compounds have either a fluorine or chlorine substituted at various positions of the benzyl moiety and exhibited IC50 values between 1.2 and 4.6 µM. It can be deduced that smaller and more electronegative substitutions, such as fluorine or chlorine, in comparison to larger substitutions, such as bromine (5h–5j) or methyl (5g–5i), is preferred as it results in superior inhibitory activity (Table 1). The majority of the final compounds also exhibited improved activity compared to the unsubstituted benzyl moiety of 5a.</p><!><p>In silico BBB predictions and IC50 values (µM) of the test compounds and controls for eeAChE, eqBuChE, and DPPH+.</p><p>aAChE selectivity index = IC50(eqBuChE)/IC50(eeAChE).</p><p>bA value higher than 0.02 is predicted to cross the BBB using AlzPlatform's intergrade cloud computing server49,63.</p><p>cIC50 values of donepezil reported by reference [64].</p><p>dIC50 values of edaravone reported by reference [65].</p><p>n.d.: not determined.</p><!><p>In addition, all the compounds with substituents in the ortho position of the benzyl ring (5b, 5e, 5g, 5j) exhibited the highest inhibitory activities when compared to their meta and para counterparts. This finding is similar to that reported in previous studies37,66. The superior activity observed for 5a–5l when compared to intermediate 3, could be due to the increase in the length of the molecule. Molecular modelling corresponds with this observation in that 3 is too short to interact with both the PAS and CAS of the AChE active site. This also confirms the importance of the N-benzyl pyridinium group for optimal AChE inhibitory activity within these compounds.</p><p>Compounds 5a–5l also exhibited highly selective AChE inhibitory activity over BuChE (Table 1, SI:11 – >278). It can be speculated that the poor BuChE inhibitory results exhibited by all the compounds are due to the phenyl ring of edaravone and the pyridinium moiety exhibiting π–π interactions with the aromatic residues of AChE but are not able to exhibit similar interactions with BuChE's aliphatic and/or aromatic residues67,68. The N-benzyl pyridinium moiety presents a similar structure to donepezil's benzyl-pyridine moiety69. Therefore, the high selectivity of the hybrid compounds for AChE compared to BuChE was expected. Selectivity to AChE is advantageous as it has been shown to result in lower incidences of cholinergic side effects and wider therapeutic indices compared to non-selective cholinesterase inhibitors70. Most of the hybrid compounds exhibited proficient AChE inhibitory activity when compared to previously designed, AChE acting MTDLs30.</p><p>The 2,2-diphenyl-1-picrylhydrazyl (DPPH+) assay was employed to determine the antioxidant ability of the synthesised compounds. Trolox, a known potent antioxidant, was used as the reference compounds for this assay71. Compounds 3, 5d–j and 5l (IC50: 9.5–19 µM) exhibited similar or greater antioxidant activity when compared to the control, Trolox (IC50 = 13.1 µM). In addition, these compounds retained the antioxidant activity of edaravone (IC50 = 4.7 µM)65. Compound 5j had the best IC50 of 9.5 µM. The rest of the compounds within the series exhibited greater antioxidant activity compared to 1. These results correspond with previous literature in that a large lipophilic substitution on the 4-position of the phenyl ring of the edarvone moiety improves its antioxidant activity when compared to a smaller carboxylic group on the 4-position of the phenyl ring of the edaravone structure58,72,73.</p><p>Potential AD therapeutic agents are required to cross the blood–brain barrier (BBB) to act in the CNS. Therefore, an in silico model49,63 was used to determine the blood–barrier permeability of the synthesised compounds. The BBB permeability prediction score represents the compounds ability to cross the BBB. A threshold score of over 0.02 is considered that the compounds are BBB permeable (BBB+) and a score below 0.02 is considered that the compounds are BBB impermeable (BBB−). The results are shown in Table 1. All the compounds exhibit scores of above 0.02 and are therefore predicted to effectively cross the BBB.</p><!><p>The main goal of this study was to design and synthesise a novel series of multi-target directed edaravone-N-benzyl pyridinium hybrid compounds that exhibit cholinesterase inhibitory activity and significant antioxidant ability. The molecular modelling results showed that these compounds should be able to form significant interactions within the PAS and CAS of the AChE active site, which in turn should lead to notable inhibitory activities. The in vitro cholinesterase results indicated excellent selective AChE inhibitory activity. Compounds 5b–g demonstrated the best AChE inhibitory activity showing that smaller substitutions, e.g. fluorine and chlorine, especially in the ortho benzyl position is important for AChE inhibitory activity. Compounds 5d–j and 5l exhibited potent antioxidant activity that was comparable to trolox and edaravone. In silico blood–brain barrier evaluations predicted that all these hybrid compounds should cross the BBB. Compound 5d–g presented as the most promising MTDL candidates against AD. These compounds exhibited excellent selective AChE inhibitory activities (IC50: 1.9–3.6 µM), promising antioxidant abilities (IC50: 11.5–19 µM) and are predicted to cross the BBB. Further exploration of these compounds' abilities to exhibit Aβ inhibitory activity, neuroprotection and their pharmacokinetic- and toxicity profiles are recommended.</p><!><p>Click here for additional data file.</p>

PubMed Open Access

Composite of bentonite and cyclodextrin as an efficient catalyst for promoting chemical transformations in aqueous media

Combining the encapsulating capability of cyclodextrin and instinctive features of bentonite clay, a versatile metal free catalyst has been developed that could promote various chemical reactions such as Knoevenagel condensation, synthesis of xanthan and octahydroquinazolinones in aqueous media under ultrasonic irradiation. To prepare the catalyst, bentonite was Cl-functionalized and then reacted with isatin and guanidine successively to furnish amino functionalized bentonite. The latter then reacted with tosylated cyclodextrin. The resultant catalytic composite was characterized via XRD, SEM, EDS, BET, elemental mapping analysis, TGA and FTIR. The catalytic activity tests approved excellent activity of the catalyst as well as broad substrate scope. Notably, the catalyst could be simply recovered and reused for several reaction runs. Moreover, the activity of the composite was superior to that of its components.Clays are one of the mostly attractive natural materials that have been extensively applied for the catalysis. Apart from their low cost and availability, the diversity in their structure and chemical composition as well as their high stability make them valuable choices both as catalysts and supporting materials. Among layered clays, bentonite, Bent, is one of the mostly utilized one for developing heterogeneous catalysts. This clay possesses empirical formula of Rx(H 2 O) 4 {(Al 2 -x,Mgx) 2 [(Si,Al) 4 O 10 ](OH) 2 }, where R refer to the exchangeable cations of alkali and alkali-earth metals between the layers. Moreover, there are numerous -OH functionalities on Bent that allow introduction of other functional moieties on Bent 1-4 .Conjugation of carbohydrates and clays is a well-established strategy to produce composites that benefits from the properties of both components 5,6 . Among various carbohydrate, cyclodextrins, CDs, have attracted immense attention due to their cone shapes that allow CDs to act as nanoreactors for hosting guests with appropriate chemistry and sizes [7][8][9][10][11][12][13][14][15][16][17][18][19] . Using this feature of CD, many of chemical processes with hydrophobic reagents were accomplished in aqueous media. In fact, CDs were applied as shuttles for accommodating the hydrophobic reagents in their hydrophobic cavities and carrying them into the hydrophilic media.Using ultrasonic irradiation (US) for accomplishing chemical transformations is a well-known protocol that can accelerate the reaction rate and provide environmentally benign protocols 20 . Ultrasonic irradiations cause cavitation phenomenon. In more detail, US resulted in the formation of bubbles with high pressure and temperature. The collapse of these bubbles generates active spots for promoting the reactions 21 .In pursuit of our research on employing natural compounds for the fabrication of effective and low-cost catalytic systems [22][23][24] , herein we wish to present a versatile metal-free catalyst for performing chemical transformations in aqueous media under ultrasonic irradiation. For the fabrication of the catalyst, Bent was Cl-functionalized and then reacted successively with isatin and guanidine. The resultant compound was then tolerated reaction with the as-prepared tosylated CD to furnish a composite that possesses both clay and carbohydrate in its framework (Fig. 1). One reason for using these two components is that they are natural, biocompatible and non-toxic materials that are relatively low-cost. The second reason is related to the properties of these two components. According to the literature, cyclodextrin that is a cyclic carbohydrate with a hydrophobic cavity and hydrophilic surface can encapsulate the hydrophobic guests and then transfer them in aqueous media. In other word, it can act as a molecular shuttle [7][8][9][10][11][12][13][14][15][16][17][18][19][20] . The reason for using bentonite is that this clay has intrinsic catalytic activity. Moreover, this clay is locally abundant and very low-cost. On the other hand, similar to other clays it

composite_of_bentonite_and_cyclodextrin_as_an_efficient_catalyst_for_promoting_chemical_transformati

Result and discussion<!>Catalyst activity.<!>Elucidating the merit of conjugation of Bent and CD.<!>Proposed mechanism.<!>Evaluation of recyclability of Bent-Gu-CD.<!>Experimental section<!>Catalysts preparation. Synthesis of Bent-Cl.<!>Synthesis of Bent-IS.<!>Synthesis of Xanthan.<!>Synthesis of octahydroquinazolinones.<!>Conclusion

<p>Validation of the structure of Bent-Gu-CD. Bent, Bent-IS, Bent-IS-Gu and Bent-Gu-CD (Fig. 1) were all characterized by FTIR spectroscopy (Fig. 3). According to the literature, the absorbance bands of Bent are the bands at 1032 cm −1 (Si-O-Si), 795 cm −1 (Si-O), 1636 cm −1 (H 2 O), 3631-3425 cm −1 (-OH), and 526 cm −1 (Al-O-Si) 1,25,26 . These characteristic bands are observable in the spectra of Bent-IS, Bent-IS-Gu and Bent-Gu-CD, approving the fact that Bent structure is maintained in each preparation step. In the FTIR spectrum of Bent-IS, an additional band appeared at 1742 cm −1 that is representative of -C=O in the IS structure. In the FTIR spectrum of Bent-IS-Gu, the band at 1685 cm −1 can be ascribed to the -C = N functionality in GU. FTIR spectrum of the catalyst is similar to that of Bent-IS-Gu. This issue is due to the fact that the characteristic bands of CD overlapped with those of Bent-IS-Gu. XRD is a powerful analysis that can establish whether Bent instinct structure will be maintained in the course of functionalization. Figure 4 illustrated the comparison of XRD patterns of Bent and Bent-Gu-CD. It can be corroborated that the two patterns are similar and exhibit the characteristic peaks of Bent at 2θ = 7°, 20.8°, 21.9°, 26.6°, 27.7°, 31.7°, 36°, 50°, 62°, 73.5° and 76°2 7,28 . This result clearly confirmed that Bent was stable and did not structurally alter during surface modification.</p><p>SEM images of Bent-Gu-CD is provided in Fig. 5. As shown, the catalyst possesses an aggregated morphology that is distinguished from the morphology of the used Bent (Supplementary Fig. S1).</p><p>The results of EDS analysis of Bent-Gu-CD is presented in Fig. 6A. According to the literature, Si, Fe, Mg, O, Al and Ca are indicative of Bent structure 29 . C and O atoms can be attributed to the CD structure. Moreover, N, C and O atoms can be ascribed to IS-Gu. High dispersion of C and N atoms in the elemental mapping analysis, Fig. 6B, approved homogeneous functionalization of Bent with the organic functionality.</p><p>TG analysis of the catalyst, Supplementary Fig. S2, indicated that apart from the known weight losses of Bent, i.e. weight losses as a result of dehydration (at ~ 200 °C) and Bent degradation (~ 460-560 °C), an additional www.nature.com/scientificreports/ weight loss at 280 °C can be discerned that is due to the degradation of the incorporated organic modifying agents. Using TGA, the content of CD in Bent-Gu-CD was estimated to be 7 wt%.</p><p>Using BET, the specific surface area of Bent-Gu-CD was estimated as ~ 6 m 2 g −1 . This value was lower than that of Bent (40.62 m 2 ).</p><!><p>Using Bent that is a natural clay with instinct catalytic activity and CD that can act as a molecular shuttle, a novel metal free catalyst, Bent-Gu-CD, is fabricated. It was assumed that this catalyst can serve as a versatile catalyst with utility for promoting various chemical transformations in aqueous media. To examine this postulation, a model Knoevenagel condensation was first carried out through reaction of benzaldehyde and malononitrile in the presence of scant amount (10 mg) of Bent-Gu-CD in H 2 O, Fig. 2. Furthermore, to accelerate the reaction and provide environmentally benign procedure, the reaction was performed under ultrasonic irradiation. Gratifyingly, it was found that under Bent-Gu-CD catalysis, the reaction proceeded rapidly and resulted in 100% conversion and yield after 5 min.</p><p>Encouraged by high activity of the catalyst, more complicated reaction, synthesis of a model xanthan, was examined via ultrasonic-assisted reaction of benzaldehyde, dimedone and catalytic dosage of Bent-Gu-CD (20 mg) in aqueous media at ambient temperature. Interestingly, it was shown that 100% conversion and yield were achieved after 10 min.</p><p>To further establish the diversity of the catalyst, ultrasonic-assisted synthesis of octahydroquinazolinones in aqueous media was appraised. To this purpose, the catalyst dosage was optimized (Supplementary Table S1). The results confirmed that using 40 mg of Bent-Gu-CD, the model reaction, reaction of benzaldehyde, dimedone and urea, proceeded to give 100% conversion and yield after 15 min.</p><p>Next, the substrate scope was examined for the synthesis of octahydroquinazolinones (Table 1). As tabulated, the presented protocol could be generalized to various aldehydes, including aldehydes with different groups and heterocyclic aldehydes. In all cases, high reaction yield was achieved in a very short reaction time. However, it can be recognized that use of heterocyclic aldehydes resulted in slightly lower yield. On the other hand, comparison of the aldehydes with same functional groups indicated that the position of the functionality can affect the obtained yield, Table 1, entries 1 and 2. As shown, presence of the functional group on para position was favorable and led to superior yield. This issue can be rationalized by considering the fact that encapsulation of aldehydes substituted on para position is more facile that other positions.</p><!><p>Next, the merit of conjugation of Bent and CD was appraised by comparing the efficiency of Bent, CD, and Bent-IS-GU with that of Bent-Gu-CD for the synthesis of the model octahydroquinazolinone. Performing the reaction under optimum condition in the presence of Bent indicated that only 40% yield of the product was furnished after 15 min. Regarding the efficiency of CD, the obtained yield was 35%. Apart from low yield, the homogeneous nature of CD was troublesome. In the case of Bent-IS-Gu, the catalytic efficiency improved to 70%. This can be due to the role of amino functionalities on activation of the reagents. As mentioned before, the activity of Bent-Gu-CD was superior to all of the control samples and reached to 100%, implying that conjugation of amino-functionalized Bent and CD is favorable for the reaction.</p><!><p>Regarding the reaction mechanism of formation of octahydroquinazolinones, it can be proposed that aldehyde was first encapsulated in CD cavity and transferred into the aqueous media. Then, it was activated by the catalyst and underwent reaction with enol form of dimedone. Subsequently, the formed intermediate tolerated dehydration and then reacted with urea. The final product was furnished through cyclization and dehydration (Fig. 7) 31 .</p><p>Study of nature of catalysis: Sheldon test. To validate the nature of catalysis and establish the real heterogeneity of Bent-Gu-CD, Sheldon test has been carried out 35 to perform this test, the reaction of synthesis www.nature.com/scientificreports/ of the model octahydroquinazolinone has been halted after 5 min and the catalyst has been separated. Then, the reaction was continued in the absence of the catalyst. Measurement of the yield of the reaction before and after catalyst removal showed that after separation of the catalyst the reaction did not proceed. This observation approved heterogeneous nature of the catalysis.</p><!><p>To evaluate the recyclability of the developed Bent-Gu-CD, the performance of the recycled catalyst for the synthesis of octahydroquinazolinones was examined. In this line, the recovered Bent-Gu-CD from the first reaction run was thoroughly washed and dried and reused for the second run of the same process under exact similar condition. As shown in Fig. 8A, Bent-Gu-CD maintained its www.nature.com/scientificreports/ activity for the second run. Recovering-reuse cycle was repeated up to eight runs. As illustrated in Fig. 8A, after the second run, the yield of the reaction slightly decreased and reached to 75% at eighth run.</p><p>To find out the origin of the observed loss of the activity of Bent-Gu-CD upon recycling, FTIR spectroscopy was employed. As shown in Fig. 8B, the comparison of the FTIR spectra of recycled and fresh Bent-Gu-CD www.nature.com/scientificreports/ indicated that although the structure of Bent-Gu-CD was preserved, some additional absorbance bands as well as band broadening were discerned. This observation can be ascribed to the deposition of organic substrates on the surface of Bent-Gu-CD. In fact, the coverage of the active spots on Bent-Gu-CD can justify the observed reduction of the catalytic activity.</p><!><p>Materials and instrument. For the fabrication of the catalyst the following chemicals were applied:</p><p>(3-chloropropyl) trimethoxysilane (CPTES), isatin (IS), triethylamine (TEA), guanidine (Gu), β-CD, p-toluenesulfonyl chloride (p-TsCl), K 2 CO 3 , toluene, EtOH and Bent. Bent was obtained from Madan Kavan Co. Iran. The other reagents were purchased from Sigma-Aldrich. For performing the ultrasonic-assisted chemical transformations, benzaldehyde, malononitrile, dimedone and urea (provided from Sigma-Aldrich) have been used. Bent-Gu-CD synthesis was verified by Fourier transform infrared (FTIR), Thermo gravimetric analysis (TGA), X-ray diffraction (XRD), scanning electron microscope (SEM), energy dispersive spectroscopy (EDS) and elemental mapping analysis. The FTIR spectrum were collected using PERKIN-ELMER Spectrum 65. For TGA, METTLER TOLEDO apparatus was employed. The tests were carried out under N 2 atmosphere at heating rate of 10 °C min −1 . XRD pattern of Bent-Gu-CD was gathered via Siemens, D5000 apparatus with Cu Kα as a radiation source. SEM /EDX analyses were carried out on MIRA 3 TESCAN-XMU. Brunauer Emmett Teller (BET) analysis was carried out via a Belsorp Mini II apparatus. Degassing was performed at 150 °C for 3 h.</p><!><p>Initially, the surface of Bent was functionalized with CPTES.</p><p>To this purpose, Bent (4 g) was dispersed in dry toluene (70 mL) under constant stirring, and then 3 mL of CPT-ES was injected drop by drop into the Bent suspension. Subsequently, the obtained mixture was continuously stirred under reflux condition at 110 °C overnight. In the next step, the product, Bent-Cl, was filtered, washed with toluene (30 mL) repeatedly and dried overnight at 70 °C.</p><!><p>A solution of IS (15 mmol in 20 mL of EtOH) was added to the homogeneous suspension of Bent-Cl (3 g, in EtOH). Then, TEA (20 mL) was added slowly into the above mention suspension, and the resulting mixture was refluxed for 24 h. Afterward, the reaction mixture was subjected to centrifugation and the resulting solid, Bent-IS, was washed with MeOH (2 × 10 mL) to remove the unreacted materials and dried in an oven at 70 °C overnight.</p><p>Synthesis of Bent-IS-Gu. Bent-IS (3 g) was dispersed in EtOH (80 mL) and magnetically stirred for 30 min to achieve a homogenous suspension. Next, to the Bent-IS suspension, Gu (2 g) was added, and the subsequent mixture was refluxed for 24 h at 100 °C. Finally, the obtained product, Bent-Gu, was filtered out by centrifugation and then washed with EtOH. The obtained Bent-Gu was dried at 60 °C overnight.</p><p>Synthesis of Bent-Gu-CD. Bent-Gu-CD was prepared in two steps. In the first step, CD was tosylated with p-TsCl. To this purpose, the mixture of CD (15.86 mmol) and p-TsCl (7.9 mmol) in pyridine (200 mL) was kept in the refrigerator at 0 °C for 48 h. Afterward, an oily product was furnished by adding distilled water (75 mL). Subsequently, cold water was poured to the oily product and the solid CD-OTs, was purified through recrystallization from water.</p><p>In second step, Bent-Gu was modified with CD. In this regard, Bent-Gu (2 g) suspension in H 2 O/ EtOH (2:1, 60 mL) was dispersed via ultrasonic irradiation (120 W, 30 min). Then, the solution of CD-OTs (1.5 g) and K 2 CO 3 (0.1 g) in H 2 O (15 mL) was injected to the so-called suspension and the resultant mixture was refluxed at 100 °C for 24 h. At the end, the reaction mixture was subjected to centrifugation and, the solid Bent-Gu-CD was washed with H 2 O (10 × 20 mL) and EtOH (3 × 10 mL) and dried in an oven overnight (Fig. 1).</p><p>Examining the catalytic activity: ultrasonic-assisted chemical transformations. Knoevenagel condensation. In a typical procedure, benzaldehyde (1 mmol), malononitrile (1.2 mmol) and Bent-Gu-CD (10 mg) were mixed in H 2 O (10 mL) in a reaction vessel. The reaction proceeded under ultrasonic irradiation (20 kHz frequency for 5 min) at room temperature. After the completion of the reaction (traced by thin layer chromatography), Bent-Gu-CD was separated via centrifuge and then washed several times with MeOH (Supplementary Figs. S3 and S4).</p><!><p>Typically, a mixture of benzaldehyde (1 mmol), dimedone (2 mmol), H 2 O (15 mL) and Bent-Gu-CD (20 mg) were sonicated at 25 °C for 10 min. The ultrasonic apparatus was equipped with thermal sensor and in the case of temperature evolution, the temperature was controlled with cold water. The reaction was followed by TLC. Then, the reaction mixture diluted with MeOH and the catalyst was filtered off. The recovered Bent-Gu-CD was washed and dried at 100 °C in an oven (Supplementary Figs. S5 and S6).</p><!><p>In a typical process, a mixture of benzaldehyde (1 mmol), dimedone (1 mmol), urea (1 mmol) and Bent-Gu-CD (40 mg) in 15 mL H 2 O were subjected to ultrasonic irradiation for 15 min. After completion of the reaction, indicated by TLC, Bent-Gu-CD was filtered, washed with MeOH and dried in oven at 70 °C. To examine recyclability of the catalyst, the dried Bent-Gu-CD was applied as a catalyst for eight runs. In Fig. 2, the three ultrasonic-assisted reactions are depicted (Supplementary Fig. S7).</p><!><p>Covalent composite of Bent and CD is fabricated through reaction of Ben-Cl with IS and GU, followed by reaction with tosylated CD. The catalyst exhibited excellent activity for promoting various chemical reactions, including Knoevenagel condensation, synthesis of xanthan and octahydroquinazolinones in aqueous media under ultrasonic irradiation. It was believed that CD in the framework of the catalytic composite acted as a molecular shuttle for hosting the substrates in its cavity and carrying them to the aqueous media. Bent on the other hand, showed catalytic activity instinctively and rendered the composite heterogeneous. Generality of the protocol, rapidness, facile recovery and recyclability of the catalyst as well as using naturally occurring compounds are other advantageous of the present protocol.</p><p>Received: 20 October 2020; Accepted: 15 February 2021</p>

Scientific Reports - Nature

Emerging roles of DYRK2 in cancer

Over the last decade, the CMGC kinase DYRK2 has been reported as a tumor suppressor across various cancers triggering major antitumor and proapoptotic signals in breast, colon, liver, ovary, brain, and lung cancers, with lower DYRK2 expression correlated with poorer prognosis in patients. Contrary to this, various medicinal chemistry studies reported robust antiproliferative properties of DYRK2 inhibitors, whereas unbiased ‘omics’ and genome-wide association study-based studies identified DYRK2 as a highly overexpressed kinase in various patient tumor samples. A major paradigm shift occurred in the last 4 years when DYRK2 was found to regulate proteostasis in cancer via a two-pronged mechanism. DYRK2 phosphorylated and activated the 26S proteasome to enhance degradation of misfolded/tumor-suppressor proteins while also promoting the nuclear stability and transcriptional activity of its substrate, heat-shock factor 1 triggering protein folding. Together, DYRK2 regulates proteostasis and promotes protumorigenic survival for specific cancers. Indeed, potent and selective small-molecule inhibitors of DYRK2 exhibit in vitro and in vivo anti-tumor activity in triple-negative breast cancer and myeloma models. However, with conflicting and contradictory reports across different cancers, the overarching role of DYRK2 remains enigmatic. Specific cancer (sub)types coupled to spatiotemporal interactions with substrates could decide the procancer or anticancer role of DYRK2. The current review aims to provide a balanced and critical appreciation of the literature to date, highlighting top substrates such as p53, c-Myc, c-Jun, heat-shock factor 1, proteasome, or NOTCH1, to discuss DYRK2 inhibitors available to the scientific community and to shed light on this duality of protumorigenic and antitumorigenic roles of DYRK2.

emerging_roles_of_dyrk2_in_cancer

<!>DYRK2 regulates 26S proteasome function<!>DYRK2 phosphorylates HSF1 and modulates proteotoxic response<!>DYRK2-p53 tumor suppressor link<!><!>c-Myc<!>c-Jun<!>SNAIL<!>SIAH2<!>EDVP E3 ubiquitin ligase<!>STAT3<!>TBK1<!>NOTCH1<!>Transcriptional/epigenetic mechanisms<!>TNBC<!>Other breast cancer subtypes<!>DYRK2 in lung cancer: unresolved issues<!><!>Small-molecule inhibitors of DYRK2<!>Conclusion and future perspectives<!>Conflict of interest

<p>Edited by Alex Toker</p><p>DYRK2 belongs to the DYRK family within the CMGC superfamily and is mutated in cancer.A, DYRK2 is a class II DYRK on the CMGC superfamily branch of the kinome. B, structure of DYRK2 indicating the major structural domains and cancer-associated mutations (derived from PDB ID: 3K2L) with a hypothetical effect on DYRK2 structure/function. C, the domain diagram providing a 2D comparative image of the domains of class II DYRK2 and class I DYRK1B. Class I DYRKs exhibit two NLS sequences, a C-terminal PEST domain and a lack of NAPA domain characteristic of class II. The autophosphorylation of Tyr and hydroxylated Pro Y382/P441 (DYRK2) and Y273/P332 (DYRK1B) are shown. CMGC, Cyclin-dependent kinases, Mitogen-activated protein kinases, Glycogen synthase kinases, and CDC-like kinases; DYRK, Dual-specificity tYrosine phosphorylation–Regulated Kinase; NAPA, N-terminal autophosphorylation accessory; NLS, nuclear localization sequence; PEST, Pro-Glu-Ser-Thr.</p><p>Like most CMGC kinases, DYRKs have an amino acid motif of preference on their substrates. DYRKs prefer an arginine (R) at the −3 position of the phosphoserine/threonine residue along with a strong preference for a proline (P) at +1: Rxx(pS/T)P motif (7). Being a preferred motif across all members, redundancies have been observed wherein multiple CMGC kinases phosphorylate the same site on the substrate (reviewed in Boni et al. [8]). Although both the −3 R and +1 P are strongly preferred, some DYRK substrates lack the +1 P such as histone H3 for DYRK1A (9), 26S proteasome regulatory subunit 6B RPT3 (10), and heat-shock factor 1 (HSF1) (11) for DYRK2, whereas the −3 R is lacking on multiple DYRK2 substrates such as p53 (12, 13), c-Jun, c-Myc (14), and SIAH2 (15). For those lacking the +1 P, the substrates have exhibited no redundant kinases within the related CMGC superfamily thus far (10). Among the DYRKs, DYRK2 often functions in tandem with related CMGC kinase, GSK3, in sequentially phosphorylating various substrates (7, 9, 16, 17). DYRK2 provides a priming phosphorylation for further GSK3 activity (7, 9, 16, 17). DYRK2 has been identified in all eukaryotes (10, 18), and interestingly across all orthologues, the conserved biological function of the DYRK2 isoform is regulation of cell division and/or tissue development (10, 18). A recent work has shown that DYRK2 is an essential kinase during embryogenesis, and mouse embryos with homozygous deletion of DYRK2 exhibit stunted development and pups die just before birth (19).</p><p>Of all the DYRK isoforms, DYRK2 is the only member that functions as a kinase activity–independent scaffold for an E3 ubiquitin ligase complex (20, 21, 22, 23, 24). DYRK2 is an integral part of the EDVP (EDD [ubiquitin protein ligase] + DDB1 [damage-specific DNA-binding protein] + VPRBP [HIV-1 Vpr-binding protein]) E3 ubiquitin ligase complex that carries out phosphorylation-mediated degradation of various cell cycle components to ensure smooth transition of G2/M stages of cell cycle (20, 21, 22, 23, 24). Some of the cancer mutations in Figure 1B are thought to affect efficient EDVP complex formation (3). Thus, over the past few decades, many groups have identified various molecular mechanism and substrates for DYRK2 playing diverse roles in cellular growth, proliferation, and developmental processes with a focal point being its role in cancer (10, 25, 26, 27, 28).</p><p>Besides DYRK2, the other DYRK isoforms, especially the class I's, have a long history in the field of cancer. Although DYRK1B has an overall protumorigenic role specifically in pancreatic and ovarian cancers, DYRK1A exhibits a more controversial role with reports of both protumorigenic and antitumorigenic mechanism in different cancers (reviewed in Boni et al. [8]). Within the class II DYRKs, very little is known about DYRK3 and DYRK4 with limited literature pointing to a more protumorigenic role for both (29, 30). DYRK2, on the other hand, is the most extensively studied class II isoform, and the high-profile substrates reported, such as p53 (12, 13), c-Jun (14), c-Myc (14), NOTCH1 (31), HSF1 (11), 26S proteasome (10, 25, 26), and SIAH2 (15, 31), have brought the kinase to the forefront of oncology research. For the past 2 decades, multiple studies have reported an overarching tumor suppressor role of DYRK2 across various cancers (reviewed in Yoshida and Yoshida [27]), with antitumorigenic roles including regulation of cell cycle, apoptosis, epithelial–mesenchymal transition (EMT), cancer stemness, and antimetastatic roles (reviewed in Yoshida and Yoshida [27]). On the other hand, since 2016, multiple studies report major protumorigenic roles of DYRK2 (10, 11, 25, 26), and a few studies have identified DYRK2 as a possible cancer driver (32, 33, 34). Furthermore, mRNA expression analyses from The Cancer Genome Atlas (TCGA) tumors along with matched normal controls reveal that the majority of cancers have higher median expression of DYRK2 than adjacent normal tissues (26), and a similar pattern has been shown for DYRK2 protein levels in some tumor types (11, 26). All of these data suggest that DYRK2 might be an excellent potential drug target. With such high profile, oncology-related substrates, could the function of DYRK2 differ based on cancer type or cell type? To shed some light onto this question, this review will re-examine the current literature on the role of DYRK2 in cancer and follow up with existing knowledge of small-molecule inhibitors developed to target DYRK2.</p><p>DYRK2 regulates proteostasis via a two-pronged mechanism. DYRK2 phosphorylates and upregulates the activity of the 26S proteasome, which reduces proteotoxic stress by degrading misfolded/unfolded proteins. In parallel, DYRK2 triggers phosphorylation-mediated activation of HSF1, which promotes transcriptional upregulation of chaperones that promotes folding of misfolded/unfolded proteins. Proteasome inhibitors (PIs) such as bortezomib, carfilzomib, and ixazomib inhibit the proteasome and result in enhanced proteotoxic stress because of toxic protein aggregates. Proteasome inhibition by PIs triggers indirect activation of the HSF1 pathway to compensate for the loss of proteasome activity thereby decoupling the proteasome dependence of cancer. HSF1, heat-shock factor 1.</p><!><p>In 2016, an RNAi kinase screen identified DYRK2 as a kinase-regulating 26S proteasome activity (10). The study showed that DYRK2 depletion either by si/shRNA or CRISPR/Cas9 KO led to a 30 to 40% decrease in proteasome activity (10). The mature 26S proteasome is a complex of more than 30 distinct subunits that catalyzes 80% of eukaryotic protein degradation and harbors three distinct peptidase activities in the core subunit (chymotryptic, tryptic, and caspase-like) (38, 39). Besides the core of the proteasome, the complex also consists of the 19S regulatory subunit that binds to ubiquitylated proteins, whereas a six-membered ATPase ring hydrolyzes the protein into a polypeptide chain for entry into the peptidase core for degradation (40). Interestingly, DYRK2 phosphorylates the Rpt3 subunit on the ATPase ring of the 19S subunit of the proteasome on an evolutionarily conserved Thr25 site (10). Rpt3 pT25 had been previously reported by Steve Gygi's group in their 2008 work on quantitative phosphoproteomics of mitosis (41), but the function of the phosphorylation was not known. A phospho-specific antibody generated against pT25 Rpt3 showed that the site is dynamically upregulated during G2/M stage of the cell cycle and that serum starvation leads to loss of Thr25 phosphorylation (10). Furthermore, CRISPR/Cas9 knock-in of a phospho-deficient Thr25Ala on Rpt3 mimics the DYRK2 KO phenotypes in cells wherein there is a delay in mitotic progression, slower cell proliferation rates, and inhibition of all three peptidase activities of the 26S proteasome (10). The 26S proteasome degrades nearly 80% of all eukaryotic proteins, and hence, a 30% loss in activity leads to significant proteotoxic stress and consequent cell death in breast cancer cells (10). Intriguingly, DYRK2 KO cells were significantly more sensitive to the proteasome inhibitor, bortezomib, suggesting DYRK2 could be a possible therapeutic target for treatment of cancer (10). Indeed, in an ectopic nude mouse xenograft model, DYRK2 KO and T25A Rpt3 knock-in cells were less efficient in generating a tumor as compared to parental cells (10). This study further established the DYRK2-proteasome axis as potentially tumor promoting because higher expression of DYRK2 significantly correlated with higher mortality and poorer relapse-free survival in patients with breast cancer (10). In fact, inhibition or genetic depletion of DYRK2 tipped the scales of proteostasis in TNBC and MM cells. DYRK2 mRNA levels are higher in newly diagnosed and relapsed MM than normal donors (26). In fact, mice bearing syngrafted/xenografted myeloma cells with genetic depletion of DYRK2 exhibit significantly slower myeloma disease progression and reduced bone degeneration (26). Furthermore, bortezomib-resistant RPMI8226 myeloma cells express higher protein levels of DYRK2 than nonresistant RPMI8226 (26), suggesting that DYRK2 might play a role in driving drug resistance in some myeloma cases. The potent and selective DYRK2 inhibitor, LDN192960, induces cytotoxicity in myeloma cells both in vitro and in vivo with minimal off-target effects (26). The fact that the DYRK2 inhibitor alleviates myeloma burden in vivo suggests DYRK2 could indeed be a viable in vivo target for myeloma therapeutics. Resistance to proteasome inhibitors have been reported in patients, and this is either brought about by cancer mutations in the proteasome core or via upregulation of HSF1-mediated proteotoxic response pathway.</p><!><p>The transcription factor HSF1 is the master regulator of proteotoxic stress responses and supports oncogenesis by helping cancer cells cope with the proteotoxic stress associated with both aneuploidy and oncogenic mutations. This has been demonstrated by the reduced susceptibility of Hsf1-KO mice to tumor formation driven either by Ras/p53 mutations or by chemical carcinogens (42, 43). Furthermore, high levels of HSF1 expression associate with poor outcome of various cancers (44). Upon proteotoxic stress, HSF1 is activated, translocates to the nucleus (45), and initiates the transcription of heat-shock proteins. Heat-shock proteins then function as molecular chaperones, protecting cells against proteotoxic stress by assisting in protein folding (46). HSF1 activity and stability are tightly controlled by multiple post-translational modifications (47). Among these, phosphorylation of serine 320 and serine 326 is associated with stability and nuclear accumulation followed by enhanced transcriptional activity of HSF1 (48, 49, 50). DYRK2 positively regulates HSF1 nuclear stability and activity, by phosphorylating it at Ser320 and Ser326 in TNBC cells (11). Indeed, DYRK2-depleted TNBC cells were far more sensitive to heat shock–mediated proteotoxic stress than parental cells, thus corroborating that DYRK2 plays a major role in maintaining proteostasis in TNBC cells. This link between DYRK2 and HSF1 is also observed in TNBC tumor samples, wherein a marked correlation was observed between high DYRK2 levels and high nuclear HSF1 levels.</p><p>The HSF1 pathway and the proteasome are not just two of the main pathways maintaining cell proteostasis, but they are interconnected and can compensate for each other. As mentioned before, proteasome inhibitors lead to the activation of HSF1 (Fig. 2) in an effort to protect the cell against the accumulation of toxic proteins (51, 52). The cytoprotective response mediated by HSF1 counteracts the cytotoxic effect of proteasome inhibitors (51, 52, 53), and thus, HSF1 inhibition might be effective to overcome proteasome inhibitor resistance in cancer cells. In that sense, a DYRK2 inhibitor induced cytotoxicity even in MM cells resistant to proteasome inhibitors (25, 26), suggesting that in fact DYRK2 inhibition might be targeting different complementary pathways. This observation was further echoed by a recent study showing that MM cells were extremely sensitive to increased temperatures and heat shock (54). In fact, combining heat shock with proteasome inhibitors led to higher accumulation of misfolded proteins leading to acute proteotoxic stress and apoptosis in the myeloma cells (54). Because cancer cells harbor significantly higher misfolded proteins than normal cells, targeting DYRK2 could indeed tip the scales for proteostasis in malignant cells and provide a significant therapeutic window for targeting specific cancers. This is indeed the case because normal/noncancerous cells were far more resistant to DYRK2 inhibitors (25, 26). Thus, targeting DYRK2 can significantly affect proteostasis (Fig. 2) via perturbation of both HSF1 and 26S proteasome activity leading to cancer cell death.</p><p>Hence, in the context of TNBC and MM, DYRK2 plays an overarching role as an oncogenic kinase and a potential therapeutic target.</p><!><p>A major molecular mechanism by which DYRK2 has been reported to exhibit the antitumorigenic role is via phosphorylation of tumor suppressor p53 on serine46 (Ser46). Upon genotoxic stress, energy stress, or heat shock, multiple CMGC kinases such as homeodomain-interacting protein kinase (HIPK2), mitogen-activated protein kinase p38α, and DYRK2 can phosphorylate p53 on Ser46, which triggers transcription of proapoptotic genes leading to cell death or cell senescence (reviewed in Liebl and Hofmann [55]). Upon DNA damage, DYRK2 is phosphorylated by ataxia-telangiectasia mutated kinase, which protects DYRK2 from proteasomal degradation leading to its nuclear accumulation where it phosphorylates p53 on Ser46 and promotes its transcriptional tumor suppressor activity (12, 13). Although phosphorylated Ser46 on p53 is indeed a marker for its tumor suppressor role, DYRK2 by no means is the exclusive kinase here. With multiple kinases including PKCδ, HIPK2, ataxia-telangiectasia mutated kinase, and p38α phosphorylating Ser46 upon genotoxic stress (55), it is hard to decipher to what extent DYRK2 contributes to this tumor suppressor role. Furthermore, p53 is mutated or truncated in a vast number of solid tumors and cancer patients with altered p53 exhibit significantly poorer survival (56, 57). Mutated p53 often exhibits stoichiometrically lower phosphoSer46 (58) and has been reported to trigger pro-oncogenic functions upon phosphorylation (59). This suggests that p53 phosphorylated on Ser46 serves as a tumor suppressor only in the few percentage of cancers containing WT p53 where patients exhibit better chances of survival.</p><p>Multiple publications carrying out sequencing or immunohistochemistry to study mRNA/protein levels of DYRK2 have suggested that DYRK2 is a tumor suppressor in colorectal (60, 61, 62, 63), liver (64), brain (65), and lung cancers (66, 67) and that the kinase promotes chemosensitivity in ovarian cancer (68). However, ovarian, liver, brain, lung, and colorectal cancers exhibit some of the highest mutations and variant allele frequencies in p53 compared with other cancer types (56, 57, 69). Thus, it is unclear to what extent DYRK2's phosphorylation of p53 could play as a tumor-suppressive role in these solid tumors exhibiting p53 mutation or loss. Furthermore, in endothelial cells, the pan-DYRK inhibitor, harmine (albeit with possible off-target effects), promotes p53 phosphorylation on Ser15, Ser20, and Ser37 (70), leading to higher p53 protein levels upon DNA damage (70, 71). Seemingly, in this case, DYRK2 inhibition led to tumor suppression. Hence, it is also important to decipher the molecular functions of DYRK2 in noncancer models or as a potential cancer driver. A recent unbiased deep multiomics study looking at the proteome, phosphoproteome, and transcriptome of murine high-grade brain cancer glioma model reported 41 kinases including DYRK2 exhibiting higher activity and rewired substrate signaling (34). Furthermore, the glioma murine model was generated by intracranial implantation of genetically engineered p53 null astrocytes, thus making the tumor-suppressor role of DYRK2-p53 axis highly untenable in this model.</p><!><p>Overall summary of DYRK2 in neoplasia. The figure provides a holistic view of the various reported roles of DYRK2 in different forms of cancers. For each cancer, the various interactors/substrates/effectors of DYRK2 are highlighted either in green (tumor-suppressor role) or in red (protumorigenic role). The cancer models/tools (cell-based, mouse models, patient samples, DYRK2 inhibitor) used to derive the respective conclusions are also shown. Direct DYRK2 substrates are shown with the added (P) phosphate, and conclusions based on sequencing or immunohistochemistry are also highlighted. Gray arrows indicate those cancers where controversial or conflicting reports have been documented.</p><p>DYRK2 molecular mechanism and substrates/partners listed along with reported phosphorylation sites, overarching role, and the unanswered questions raised by each study</p><p>CMGC, Cyclin-dependent kinases, Mitogen-activated protein kinases, Glycogen synthase kinases, and CDC-like kinases; DYRK, Dual-specificity tYrosine phosphorylation–Regulated Kinase; EMT, epithelial–mesenchymal transition; HSF1, heat-shock factor 1; MOAP, modulator of apoptosis protein 1; NOTCH1, neurogenic locus notch homolog protein 1; n/a, not directly reported.</p><p>Also, refer Figure 3.</p><!><p>DYRK2 has been reported to exhibit a p53-independent tumor-suppressor role by phosphorylating c-Myc on serine62 (Ser62) (14). c-Myc is a major proto-oncogenic transcription factor known to be overexpressed and mutated in various cancers (74). Post-translational modifications of c-Myc have been a topic of much debate over the past 30 years in which sequential phosphorylation of Ser62 and Threonine58 (Thr58) seems to play major roles in c-Myc transactivation (75, 76). The consensus in the field is that Thr58 is a GSK3-phosphorylation site while Ser62 seems to be the priming site for GSK3 activity, and similar to phosphorylation of p53 at Ser46, various CMGC kinases have been proposed (75), including DYRK2 to phosphorylate c-Myc. Dual phosphorylation of c-Myc on Thr58 and Ser62 triggers binding to an E3 ubiquitin ligase SCF-Fbxw7 (Skp-Cullen-F-box) and consequent proteasomal degradation of c-Myc (14), thus leading to the proposed tumor suppressor role of DYRK2. As stated previously, transcriptional upregulation of DYRK2 in CML promotes c-Myc degradation (72). In Burkitt lymphoma, nearly 60% of patients exhibit mutation of the GSK3 site, Thr58 (77), whereas primary cells exhibit lower levels of Thr58 phosphorylation (76), thus suggesting a tumor-suppressor role of GSK3 in this context. However, increased Ser62 phosphorylation has been observed in immortalized cells compared with primary cells (76), and monophosphorylation of Ser62 has been linked to c-Myc stabilization and higher transcriptional activity in multiple studies (78, 79). This indicates a similar conundrum as observed with the p53 Ser46 site wherein multiple kinases and diverse cancer subtypes exhibit altered mechanisms of action of major cancer-associated genes such as p53 and c-Myc.</p><!><p>A similar story is observed in case of c-Jun wherein two phosphorylation sites serine249 (Ser249: a bona fide GSK3 site) and Ser243 (reported to be phosphorylated by DYRK2) have been reported (14, 80). c-Jun is a transcription factor with established oncogenic roles (80). Similar to c-Myc, the E3 ubiquitin ligase SCF-Fbxw7 degrades c-Jun upon dual phosphorylation of Ser249 and Ser243 (81). Unlike c-Myc, Ser243 on c-Jun could be a DYRK-specific site because a previous study has elegantly ruled out most of the other CMGC kinase families (80). The same study did, however, observe redundancies between DYRK1A and DYRK2 for Ser243 on c-Jun in vitro (80), which is not surprising because the site is a +1P. In fact, dephosphorylation of Ser243 enhances c-Jun transcriptional activity in patients with cervical cancer exhibiting lower phosphoSer243 c-Jun in their tumors (82). Although the question of intra-DYRK redundancy remains, phosphoSer243 on c-Jun could indeed be a tumor-suppressor marker in specific cancers.</p><!><p>DYRK2 has also been reported to phosphorylate the zinc finger domain containing protein SNAIL that plays essential roles during development by triggering EMT (68, 83). DYRK2 knockdown led to upregulation of mesenchymal markers with consequent downregulation of epithelial E-cadherin mRNA in colon cancer (61) and promotion of proliferation and migration of glioma cells (65). This observation was consistent with other studies reporting downregulation of DYRK2 in metastatic colorectal secondary tumors found in the liver (62). SNAIL has been reported to be overexpressed in specific cancers and promote oncogenic progression by promoting EMT, invasion, and metastasis (83). DYRK2 phosphorylates serine104 on SNAIL that provides a priming site for GSK3, triggering the phosphorylation-mediated degradation of SNAIL (68, 83). This mechanism of antitumorigenic activity by DYRK2 is thought to promote chemosensitivity for ovarian cancer cells (68). A follow-up study reports that the DYRK2-mediated degradation of SNAIL is in fact reversed by p38α kinase (84). Although an interesting molecular mechanism, in both studies, DYRK2 ectopic overexpression has been carried out to justify the phosphorylation. Overexpression of CMGC kinases often leads to nonphysiological false-positive subcellular localizations and substrate identifications because of redundancy and high affinity for +1 P sites and hence further tools need to be used to confirm the DYRK2–SNAIL mechanism.</p><!><p>Seven In Absentia Homolog 2 or SIAH2 is an E3 ubiquitin ligase that plays a major role in targeted degradation of various proteins playing essential roles in regulating hypoxia (85). SIAH2 specifically regulates hypoxic tumor microenvironment by downregulation of key kinases in the Hippo signaling pathway (85). Furthermore, higher expression of SIAH2 is observed in lung cancer (86) and it plays oncogenic roles in castration-resistant prostate cancer (87). Interestingly, DYRK2 phosphorylates SIAH2 on 5 residues Ser16, Thr26, Ser28, Ser68, and Thr119. These modifications alter its subcellular localization thereby rewiring SIAH2 substrate specificity (15). SIAH2, in turn, is capable of degrading DYRK2 in specific cancers thereby triggering a protumorigenic hypoxic microenvironment (15, 85, 86). Some kinase redundancy has been observed wherein p38α kinase is capable of phosphorylating SIAH2 on same sites as DYRK2 (88); however, the DYRK2–SIAH2 link points to an interesting interplay between a kinase and a ubiquitin ligase regulating each other and thereby balancing protumorigenic and antitumorigenic roles.</p><!><p>As stated previously, DYRK2 forms a kinase-independent scaffold for the EDVP E3 ligase complex and a recent study has reported loss-of-function point mutations of DYRK2 in cancer, which largely alters the interactome and substrate specificity of DYRK2 (3). The recurrent mutations (Fig. 1B) are thought to alter activity and/or formation of the EDVP complex (3). As part of the EDVP complex, DYRK2 phosphorylates and triggers degradation of multiple substrates such as katanin p60 (KATNA1) (23), telomerase reverse transcriptase (TERT) (22), and centrosome protein 110 (CP110) (21). Phosphorylation-mediated degradation of these substrates are required for proper cell cycle transitions especially the G2/M stage. Cancer mutations could result in incomplete EDVP complex formation, and incomplete EDVP can exhibit oncogenic prosurvival role because DYRK2+EDD alone degrades the proapoptotic factor modulator of apoptosis protein 1 independently of DDB1 and VPRBP in ovarian cancer (89). The substrates of DYRK2–EDVP exhibit both protumorigenic and antitumorigenic roles in various cancers thus adding further complexity. Ovarian cancer patients with higher levels of KATNA1 exhibit better overall survival (90); higher CP110 can decrease breast cancer cell invasion (91), yet lung cancer tissue expresses higher CP110 than the normal lung (92), while TERT is largely oncogenic (93). Thus, DYRK2–EDVP functions are tumor specific.</p><!><p>DYRK2 has been reported to phosphorylate signal transducer and activator of transcription 3 (STAT3) in vitro (94). STAT3 is a transcription factor with both oncogenic and tumor-suppressor roles including regulation of tumor microenvironments (reviewed in Galoczova et al. [95]). STAT3 is phosphorylated on various residues upon interleukin/cytokine stimulation, and the phosphorylation on serine727 (Ser727) is thought to be an oncogenic biomarker in some subtypes of breast cancer (96). Although Ser727 is thought to promote the transcriptional activity of STAT3 (95), various kinases (CMGC family and beyond) have been reported to target Ser727 which is a +1P site (94). Thus, it is very difficult to dissect the importance of DYRK2 alone in driving phosphoSer727-mediated STAT3 activity.</p><!><p>TANK-binding kinase 1 (TBK1) is an important upstream regulator of innate immune transcription pathways triggering type I interferon (IFN) translation and signaling in response to pathogens (97). DYRK2 phosphorylates TBK1 at serine527, which leads to phosphorylation-mediated degradation of TBK1 and downregulation of type I IFN signaling upon viral infection (97). Besides infections, the elevated presence of type I IFN correlates with a favorable prognosis in patients with different cancers (98, 99). In fact, reduced IFN-related gene expression leads to an immunosuppressive tumor microenvironment resulting in immunotherapy resistance in many solid tumors (99). Thus, DYRK2-mediated downregulation of IFN signaling could play a major oncogenic role in triggering immunotherapy resistance in various cancers. However, the study reporting DYRK2 as the upstream kinase of TBK1 relies on ectopic overexpression of DYRK2 to demonstrate direct phosphorylation of a canonical +1P motif (97). There could be redundancies with other CMGC kinases at that site which needs to be addressed more thoroughly.</p><!><p>In response to chemotherapeutic agents, DYRK2 facilitates phosphorylation-mediated degradation of neurogenic locus notch homolog protein 1 (NOTCH1), which acts as an antiproliferative mechanism in breast cancer cells (31). NOTCH1 is a single transmembrane receptor and triggers intracellular signaling via binding to specific ligands (31). DYRK2 phosphorylates NOTCH1 on threonine2512 (Thr2512), which is a +1P site. However, NOTCH1 exhibits both tumor suppressor and oncogenic roles on a cancer-type basis (100). Interestingly, Thr2512 lies in the intracellular carboxy-terminal region of NOTCH1 that exhibits a PEST domain. The PEST region is the target of multiple CMGC kinases such as DYRK1A, HIPK2, CDKs, and GSK3, which triggers hyperphosphorylation and proteasomal degradation of NOTCH1 (reviewed in Lee et al. [101]). Thus, the redundancy conundrum remains to be solved to understand the function of NOTCH1's phosphorylation by DYRK2.</p><!><p>Besides modulating substrate phosphorylations, transcriptional and epigenetic mechanisms of DYRK2 regulation have also been proposed for some cancers. Specifically, the downregulation of DYRK2's gene expression has been linked to increased stemness in breast cancer (102) and CML (72) via upregulation of transcription factor Krüppel-like factor 4. DYRK2 expression was also downregulated transcriptionally by DNA methyltransferase 1 in colon cancer (103). The DYRK2 promoter region exhibited a higher level of methylation in cancer tissues than healthy tissues while treatment of cells with hypomethylating drug 5-azacytidine increased DYRK2 mRNA and protein levels (103). Furthermore, DYRK2 was reported to downregulate oncogenic miR-622 expression and reverse invasion of cancer cells (63), whereas long noncoding RNA long noncoding RNA derived from hepatocytes inhibits the proliferation of liver cancer cells by rescuing the expression of DYRK2 (104).</p><p>To reiterate, multiple molecular mechanisms have been proposed for DYRK2, and each mechanism is cancer-type or subtype specific (Fig. 3 and Table 1). The controversial role of DYRK2 is best highlighted in breast and lung cancers.</p><!><p>Various studies have focused on the role of DYRK2 in TNBC (10, 11, 25, 26). These studies revealed that both mRNA and protein levels of DYRK2 were higher in TNBC tumors than adjacent normal breast tissues (26). Complementing this information, a recent study with 715 samples of patients with breast cancer have shown that high protein levels of nuclear DYRK2 were associated with significantly reduced cancer survival and a shorter time to recurrence specifically within the TNBC subtype cohort (11). To test the potential therapeutic value of targeting DYRK2 in TNBC, three studies have compared the ability of parental and DYRK2-deficient TNBC cell lines to produce tumors in vivo (10, 25, 26). Crispr/Cas9-mediated DYRK2 deletion in MDA-MB-231 or MDA-MB-468 cells showed that tumors derived from TNBC–DYRK2–deficient cells had significantly slower growth rates and lower tumor burden than those derived from their parental cells. Importantly, two studies have shown that treatment with the DYRK2 inhibitors, curcumin and LDN192960, impaired growth of established TNBC tumors (25, 26). In contrast with these findings, other studies have used MDA-MB-231–derived xenografts and reported DYRK2 control EMT by degrading SNAIL (83) and promoting transcription factor Krüppel-like factor 4 expression (102), thereby functioning as a tumor suppressor. Both the studies used a DYRK2 overexpression system to show that higher DYRK2 decreased tumor formation. One study reported that mice xenografted with DYRK2-overexpressing MDA-MB-231 cells showed few metastatic lesions and a prolonged survival compared with those injected with control cells (83). In a second study, the authors compared the number of tumors produced by injecting increasing numbers of MDA-MB-231 cells with or without overexpressed DYRK2 (102). The authors used a sample size of n = 6 mice per condition and show that the total number of tumors derived from DYRK2-overexpressing cells was marginally lower than controls (102). This is in sharp contrast to others reporting DYRK2 depletion reduces proliferation and tumor formation potential of MDA-MB-231 cells (10, 11, 25, 26, 105). Some of these discrepancies might be due to the differential approaches used (DYRK2 knockdown/KO versus overexpression systems) or due to underpowered sample sizes. Furthermore, a phosphotyrosine proteomics study in TNBC cells reported that DYRK2 was among the top 5 phosphorylated proteins observed in aggressive basal-like TNBC cells (105). Because there is no evidence of the activation loop tyrosine exhibiting altered stoichiometric phosphorylation, the high levels of phosphorylation observed could be due to higher DYRK2 protein levels. In fact, siRNA knockdown of DYRK2 in basal-like TNBC MDA-MB-231 and HCC1395 cells lead to reduced proliferation, invasion, and colony formation potential of the cells (105).</p><!><p>Multiple studies looking at the role of DYRK2 in breast cancer have used the hormone receptor–positive and HER2-negative MCF7 cell line for xenograft studies. In the main study that supports the tumor-suppressor role of DYRK2 in breast cancer, the group identified DYRK2 as a priming kinase for c-Jun and c-Myc (14). In this study, the authors used a sample size of n = 3 mice per condition and carried out an orthotopic mammary-fat-pad breast cancer xenograft comparing MCF7 control cells and stable DYRK2 knockdown cells to investigate their ability to produce tumors (14). They found that DYRK2 knockdown cells clearly produce bigger tumors. Furthermore, DYRK2 knockdown cells showed higher invasion potential in vivo in an intracardiac injection model (n = 6 mice per condition). The same shRNA DYRK2 depleted cells were used in other studies as well to report the various tumor-suppressor roles of DYRK2 (102, 106). From the study with 715 samples of patients with breast cancer, no correlation was observed between DYRK2 expression and poor outcome in any of the receptor-positive breast cancer subtypes (11). However, TCGA data suggest that mRNA expression of DYRK2 is higher in breast invasive carcinoma and that higher DYRK2 expression correlates with poor survival in overall patients with breast cancer (8, 10, 26). Because mRNA and protein levels sometimes do not correlate, larger analysis looking at DYRK2 protein levels are needed to reach a finite conclusion.</p><p>The best way forward is to generate a conditional lox-cre mouse model for DYRK2 and generate hemizygous/homozygous deletion of DYRK2 in different subtypes of breast cancer genetically engineered mouse models (GEMMs) (107). Comparative tumor growth in the DYRK2 null versus parental GEMM over different subtypes would be a good way of addressing the pending questions on role of DYRK2 in breast cancer.</p><!><p>In 2003, the chromosome 12 region 12q13-14 was found to be amplified in adenocarcinomas of the lung and esophagus, and one of the resident genes, DYRK2, was significantly overexpressed in tumor samples as compared with normal tissues (33). In fact, DYRK2 exhibited the highest mRNA overexpression and highest copy numbers in tumors compared with normal tissue and other genes located in the 12q13-14 chromosomal region, suggesting that the overexpression of DYRK2 is the driving force behind the amplicon (33). This is reiterated in the TCGA lung adenocarcinoma and esophageal cancer cohort wherein tumor samples expressed higher DYRK2 mRNA than normal tissue (8). However, two independent studies report that higher protein or mRNA expression of DYRK2 is a favorable marker in pulmonary adenocarcinoma (66) and non–small-cell lung cancer (NSCLC) (67). In fact, pulmonary adenocarcinoma patients with higher DYRK2 expression exhibited a substantially higher 5-year survival than the group with lower DYRK2 expression. The higher DYRK2 levels associating with negative lymphatic invasion (66). Although the response rates to chemotherapy between the DYRK2-positive and DYRK2-negative patients were not different, patients with DYRK2+ tumors in recurrent NSCLC were suggested to have better outcome with chemotherapy (67). Mechanistically, in lung adenocarcinoma and squamous-cell lung cancer, E3 ubiquitin ligase SIAH2 targets DYRK2 for proteasomal degradation (86). SIAH2 protein and mRNA levels were found to be higher in samples of patients with lung cancer and exhibited a negative correlation with DYRK2 expression (86). Overall, the exact role of DYRK2 in lung neoplasia is still up for debate. Hence, using a similar strategy as suggested previously to generate conditional DYRK2 depletion in genetically engineered lung cancer mouse models for NSCLC, squamous-cell lung cancer, and other subtypes (108) could provide more clarity to this debate.</p><p>As reported previously, various global unbiased studies in various cancers have reported DYRK2 as a potential cancer driver with increased copy numbers, overexpression, and higher activity (32, 33, 34). On a similar note, a study using integrated high-resolution microarray analysis of gene copy number and expression in head and neck squamous-cell carcinoma cells reported that DYRK2 had the highest copy number and clear overexpression when compared with other genes in the 12q chromosomal amplicon (109). Furthermore, transcriptomics of blood identified DYRK2 as 1 of 10 potential prognostic biomarkers elevated in high-grade precancerous cervical lesions (110). Thus, unbiased identification of DYRK2 as a protein/kinase involved in potential protumorigenic role along with its substrates such as p53, c-Myc, and c-Jun further fuels the need to stratify cancers into subtypes before embarking on DYRK2 molecular studies. This duality of protumorigenic and antitumorigenic roles has been reported for the paralogue DYRK1A as well (111, 112) (Fig. 3 and Table 1), and hence, there is a clear precedence for such controversial roles in the DYRK family. One way of deconvoluting cancer-type and cell-type functions of a controversial kinase is by generating further tools such as potent and specific small-molecule kinase inhibitors.</p><!><p>The published DYRK2 inhibitors currently available to the scientific community</p><p>The kinase in bold format indicates the original target reported for the compound. IC50 values are provided where available.</p><!><p>A recently well-characterized DYRK2 inhibitor is the acridine analog, LDN192960 (26). Initially developed as a Haspin inhibitor (124, 125), LDN192960 inhibited DYRK2 with 13 nM IC50 and exhibited antitumor activity in mouse models of MM and TNBC (26). Interestingly, LDN192960 exhibited a 'mixed' mode of DYRK2 inhibition and cocrystal structure of LDN192960 with DYRK2 revealed that two water molecules mediated multiple hydrogen bonds between LDN192960 and DYRK2 active pocket (26). Similar water molecule–mediated interaction was also observed in the cocrystal structure of DYRK1A with inhibitor DJM2005 wherein a water molecule facilitated hydrogen bonding to further stabilize the inhibitor bound structure (2). LDN192960 reduced tumor burden of syngrafted and patient-derived xenografted mouse models of TNBC and delayed bone degeneration of allografted MM mouse models (26). In cells, LDN192960 exhibited potent cytotoxicity toward cancer lines with minimal impact on noncancerous cells (26). In fact, LDN192960 induced cytotoxicity to CD138+ primary myeloma cells of patients with significantly less impact on matched peripheral mononuclear cells (26). LDN192960 also exhibited additive effects in combination with FDA-approved proteasome inhibitor carfilzomib in inducing cytotoxicity in myeloma cells (26). LDN192960 was bioavailable in vivo and a dose of 50 mg/Kg body weight was sufficient to target neoplasia (26). Thus, LDN192960 inhibited DYRK2 in vivo and reduced 26S proteasome activity and thereby targeted proteasome-dependent neoplastic diseases such as TNBC and MM (25, 26).</p><p>Interestingly, curcumin inhibits DYRK2 with an IC50 of 5 nM and cocrystal structure revealed that curcumin binds to the active site pocket of DYRK2 via hydrophobic interactions (25). Curcumin is highly promiscuous, nonbioavailable, and unstable in the serum-free solution and has been labeled as both a pan-assay interference compound and an invalid metabolic panacea (126). Curcumin aggregates at concentrations greater than 15 μM, and most studies reporting controversial biological targets for curcumin used high 20- to 100μM concentrations leading to possible false positives (126). However, at lower 1- to 3μM concentrations, curcumin ablates DYRK2-mediated 26S proteasome phosphorylation in cells, reduces proteasome activity, and impairs cell proliferation in TNBC and MM cell lines in vitro and in vivo (25). DYRK2 KO cells exhibit no further off-target effects on proteasome activity with curcumin (25). Although neither a viable drug scaffold nor a highly potent kinase inhibitor, curcumin could serve as a decent compound for DYRK2 inhibition when used with proper controls.</p><p>In Table 2, we have listed those published inhibitors which have been tested directly on DYRK2 activity in vivo or in vitro along with a few interesting scaffolds identified in medicinal chemistry publications (https://www.kinase-screen.mrc.ac.uk/kinase-inhibitors) (127, 128). An important observation is that not a single one of these DYRK2 inhibitors exhibit protumorigenic/pro–cell-proliferation properties.</p><!><p>The controversial role of DYRK2 in cancer is evident (Fig. 3). Recent review articles mentioned the conflicting literature on DYRK2 (8, 28) and reported that mRNA expression data show that DYRK2 levels are higher in invasive breast carcinoma and lung adenocarcinoma than normal/adjacent tissue control (8, 26). However, others maintain that DYRK2 is a major tumor suppressor across all breast cancer subtypes and that DYRK2 depletion promotes proliferation (27). Indeed, being a CMGC kinase with a +1Pro active site, there are expected redundancies between DYRK2, other DYRKs, and possibly other CMGC kinases. A recent review article provides a list of substrates exhibiting overlapping DYRK kinases phosphorylating the same residues (8). Furthermore, as a member of the EDVP complex, cancer-specific mutations can alter DYRK2 substrate signature in specific cancer cells, which could be either tumor suppressive (3) or oncogenic (89). Although immunohistochemistry of tumors in patients with glioma shows lower levels of DYRK2 correlates with poorer survival (65), a receptor tyrosine kinase–transduced p53 null glioma mouse model exhibits higher DYRK2 activity and potentially altered signaling with diverse substrates (34). It is a definite possibility that we have just seen the tip of the iceberg when it comes to deconvoluting the redundancy and substrate overlap between the DYRKs and other CMGC kinases. As already shown, meticulous inhibitor screens and development of phospho-specific antibodies to the substrates could pave the way to dissect specificities and potential redundancies between the CMGC kinases (80). With the establishment of genetic engineering CRISPR KO/knock-in strategies and advanced quantitative phosphoproteomics, we might be able to dig deeper into identifying novel substrates and mechanisms in various cancer types. One major way forward would be to generate a conditional KO mouse model of DYRK2 in various cancer GEMM backgrounds to study tumor development and thereby ascertain the specific role of DYRK2 in each cancer subtype. A latest study reports that mouse embryonic fibroblasts with DYRK2 deletion exhibits significant downregulation of major cell cycle and proliferation drivers and markers such as Ki67, Aurora kinase A, PLK1, Bub1, and Bub1b (19). Although not in a cancer model, these observations are consistent with those observed in TNBC tumors where depletion of DYRK2 leads to reduced cell proliferation with greatly reduced Ki67 (10, 25, 26). However, DYRK2 could drive tumor-suppressor functions pairing with substrates reported in Figure 3 and Table 1 or other yet-undiscovered substrates in cancer type–specific mechanisms (27). In future, the roles of DYRK2 in cancer need to be approached in a more holistic way using multiple controlled models in each study including but not limited to genetic depletions and biochemical analyses, using specific inhibitors, in vivo animal models, and in vitro cell-based assays. Although overexpression of mRNA in cancers does indicate a potential oncogenic role, correlating that to corresponding increase of protein levels is important because mRNA and protein level often do not correlate in tumor samples. Before immunohistochemistry analysis on patient samples, proper antibody optimization steps are necessary while data analyses and sample size determinations need to be supported by proper statistical principles. Furthermore, ectopic overexpression of DYRK2 often results in false-positive substrate phosphorylation/binding and such experiments should always be accompanied with controls to ascertain the physiological/bona fide roles of the kinase. Like its paralogue DYRK1A (111, 112) and many other kinases, DYRK2 may indeed play both protumorigenic and antitumorigenic roles in different cancer types and subtypes, which is often determined by spatiotemporal interactions between kinases and specific substrates. Novel cancer therapeutic targets are a need of the hour, and hence, controversies delaying the establishment of a potential target or a tumor suppressor need to be objectively and quickly addressed. Deconvolution of the enigmatic roles of DYRK2 in various cancer types and subtypes should be prioritized among those in the field making our tools and expertise available for the greater scientific community in this endeavor.</p><!><p>The authors declare that they have no conflicts of interest with the contents of this article.</p>

PubMed Open Access

Mapping of Ion and Substrate Binding Sites in Human Sodium Iodide Symporter (hNIS)

The human sodium iodide symporter (hNIS) is a theranostic reporter gene which concentrates several clinically approved SPECT and PET radiotracers and plays an essential role for the synthesis of thyroid hormones as an iodide transporter in the thyroid gland. Development of hNIS mutants which could enhance translocation of the desired imaging ions is currently underway. Unfortunately, it is hindered by lack of understanding of the 3D organization of hNIS and its relation to anion transport. There are no known crystal structures of hNIS in any of its conformational states. Homology modeling can be very effective in such situations; however, the low sequence identity between hNIS and relevant secondary transporters with available experimental structures makes the choice of a template and the generation of 3D models nontrivial. Here, we report a combined application of homology modeling and molecular dynamics refining of the hNIS structure in its semioccluded state. The modeling was based on templates from the LeuT-fold protein family and was done with emphasis on the refinement of the substrate-ion binding pocket. The consensus model developed in this work is compared to available biophysical and biochemical experimental data for a number of different LeuT-fold proteins. Some functionally important residues contributing to the formation of putative binding sites and permeation pathways for the cotransported Na+ ions and I\xe2\x88\x92 substrate were identified. The model predictions were experimentally tested by generation of mutant versions of hNIS and measurement of relative (to WT hNIS) 125I\xe2\x88\x92 uptake of 35 hNIS variants.

mapping_of_ion_and_substrate_binding_sites_in_human_sodium_iodide_symporter_(hnis)

INTRODUCTION<!>Choice of Templates<!>Multiple Sequence Alignment Protocols<!>3D Threading and Refinement of Models<!>Side-Chain Relaxation with ROSETTA<!>Putative Binding Sites from GCMC and PB Mapping<!>Refinement of Binding Sites with MD Simulations<!>Experimental Protocols<!>Site-Directed Mutagenesis<!>Generation of Stable Cell Line<!>In Vitro Uptake Assay<!>Consensus 3D Structural Model of the Semioccluded State of hNIS<!>Putative Binding Sites for Cotransported Cations and Substrate in hNIS from Molecular Modeling<!>Location of the Putative Na2 Site<!>Location of the Putative Na1 Site<!>Substrate Binding Sites in hNIS<!>Location of the Putative S1 Site<!>Location of the Putative S2 Site<!>Differences between Iodide and Perchlorate Binding in hNIS<!>Insights for Permeation Pathway and Gating<!>CONCLUSIONS

<p>Sodium (Na+) iodide symporter, or NIS, is a secondary transporter from the SLC5 sodium/solute symporter family, which includes several sodium-dependent glucose transporters such as hSGLT1 and hSGLT2.1 NIS is found predominantly in the basolateral membranes of thyrocytes, where it transports the I− ions necessary for the biosynthesis of thyroid hormones against their concentration gradient by coupling to the symport of Na+ ions along their electrochemical gradient. NIS is expressed also in a variety of organs, including the stomach, intestines, salivary glands, placenta, lactating breast, lungs, and testes, where its role is not well understood.2 Iodide transport defects (ITD) in the function of human NIS (hNIS) due to missense pathological mutations and deletions result in iodide deficiency and disorders like goiter and congenital hypothyroidism, which can lead to lifelong cognitive and physical disability. Several mutations related to ITD have been reported in patients.3</p><p>Electrophysiology studies performed on rat NIS showed that it enables secondary transport for a number of monovalent anions with remarkable differences in size and molecular shape (e.g., I−, NO3−, SCN−, TcO4−, BF4−).2,4,5 The common environmental pollutant perchlorate (ClO4−), which is also a transported substrate of NIS, serves as a potent competitive inhibitor of the transport of I− and other NIS substrates.6</p><p>Since hNIS is expressed and functional in differentiated thyroid cancers, some breast cancers, and cholangiocarcinoma,3 its ability to transport selected positron emission tomography (PET) and single-photon emission computed tomography (SPECT) radiotracers (e.g., 125I−, 99mTcO4−, [18F]-BF4−) can be potentially harnessed for early detection of these cancers.7–10 The transport of 131I− by hNIS has already been used for treatment of Graves disease, toxic nodular goiter, and benign nontoxic goiter and offers potential therapeutic implementations of hNIS in differentiated thyroid tumors, breast cancer, and other malignancies where hNIS is present.3 Importantly, recent advances in gene targeting methods have paved the way for potential use of hNIS for in vitro and in vivo PET and SPECT imaging with selected radiotracers in preclinical models and human subjects.11–14 Maximizing the potential of hNIS for such applications relies on the expression of engineered mutants of hNIS which are transport efficient even in the presence of background ion blockers in the cells of interest (e.g., cancer cells). The administration of the competitive inhibitor ClO4− can effectively suppress radiotracer uptake in all healthy cells expressing wild type hNIS, while cancer cells containing engineered variants of hNIS will accumulate the radiotracer. This procedure would eliminate uptake of radiotracers and cytotoxicity in the healthy tissues expressing wild type hNIS and would prevent background signal interference in the PET/SPECT measurements. Pertechnetate, 99mTcO4−, used in SPECT imaging and tetrafluoroborate, [18F]-BF4− (TFB), used in PET imaging are among the most promising radiotracers of hNIS,15,16 given their short half-life times (6 h for 99mTcO4− and 109.8 min for [18F]-BF4−)10,17 and high quality of the produced imaging signal. Thus, it is desirable that hNIS mutants which are optimized to transport efficiently these two radiotracers in the presence of ClO4− are developed. The rational optimization of hNIS for in vivo or in vitro imaging however is hindered by lack of crystal structures of hNIS and by often-conflicting uptake mechanisms reported for NIS-mediated transport.</p><p>Rat NIS, which has a high sequence identity with human NIS, was cloned successfully in 1996 and kickstarted the molecular characterization of the structure and function of NIS.18 Although a crystal structure of NIS is yet to be resolved, a number of studies have elucidated possible structure/activity determinants. Hydropathy analyses, coupled with mutagenesis studies, have shed light on the overall secondary structure of NIS.19 It is accepted that NIS features 13 transmembrane domains (TMs), with an N-terminus in extracellular orientation and a long cytoplasmic C-terminus tail. Glycosylation and phosphorylation positions have been identified in the extracellular loops of TMs 6 and 7 and 12 and 13.19</p><p>The transport motif of NIS is composed of TMs 2–11 and likely possesses the 5 + 5 inverse repeat20 found in the crystal structures of the bacterial leucine transporter LeuT21 and various other secondary transporters, including the galactose transporter vSGLT of Vibrio parahemolyticus,22 which is regarded as a homologue of the SLC5 family members hSGLT1 and hSGLT2.1 Moreover, rat NIS is reported to exhibit either electrogenic (Na+:substrate = 2:1) or electroneutral (Na+:substrate = 1:1) transport, depending on the transported anion.2,4–6 Functional mutagenesis studies along with the known pathological ITD mutations have identified several amino acids in NIS as critical for I− transport.23–46 The overall geometry of NIS, the structure of its binding sites, and its transport mechanism, however, remain poorly understood.</p><p>Some information about NIS has been obtained from comparison to other proteins which have the same 5 + 5 LeuT-fold as NIS. Among these proteins are two homologues of hNIS, vSGLT, whose structure has been resolved with X-ray diffraction,22 and the proline transporter, PutP, which has been extensively studied with functional mutagenesis.47–61 In the absence of actual crystallographic structures, computational modeling provides an avenue for exploration of structural characteristics and their potential effect on protein function. Validated homology models of PutP, based on vSGLT, have been reported before56,57 and offer the opportunity for structure/activity analysis of PutP in analogy to vSGLT and other LeuT-fold transporters. Homology models of hNIS based on the vSGLT structure have also been reported in recent articles2,3,44–46 without extensive cross-validation. Furthermore, some of the studies did not consider the copermeant Na+ ions that are required for substrate or inhibitor binding.45</p><p>Given the potential importance of hNIS as a reporter and therapeutic gene used for early cancer detection and treatment, and the need for better understanding of the transport mechanism in hNIS for construction of engineered strains with enhanced selectivity for I−, 99mTcO4−, and [18F]-BF4−, even in the presence of competitive inhibitors like ClO4−, we report here a thoroughly validated homology model of hNIS, consistent with the available experimental data. Several hNIS models based on different combinations of LeuT-fold templates were constructed, and a combination of Grand Canonical Monte Carlo (GCMC)62 and all-atom molecular dynamics simulations was employed with the aim of identification of putative binding sites for Na+ and I− and other structural elements of relevance to transport. The molecular simulations were used to guide mutagenesis and uptake experiments by mapping residues essential for substrate and copermeant ion binding. The final validated structure of our human NIS model is available upon request.</p><!><p>It has been suggested that hNIS has a similar architecture as the vSGLT galactose transporter,2 a bacterial orthologue of hSGLT1 and hSGLT2 from the SLC5 protein family to which hNIS belongs.1 The sequence similarity, however, remains relatively low (<35%). vSGLT features the typical 5 + 5 LeuT-fold inverted repeat which is found in a number of proteins transporting a wide range of diverse substrates.20 For sequence alignment, necessary for the construction of the homology models of hNIS, we chose six LeuT-fold proteins crystallized in the occluded or semioccluded state, including vSGLT22 (Na+-coupled galactose transporter, PDB code 3DH4), LeuT21 (Na+-coupled leucine transport, PDB code 2A65), AdiC63 (Na+-independent arginine/agmantine antiport, PDB code 3L1L), Mhp164 (Na+-coupled hydantoin transport, PDB code 4D1B), BetP65 (Na+-coupled betaine transport, PDB code 4DOJ), and dDAT66 (Na+-coupled dopamine transport, PDB code 4XPA). The substrates of these templates are neutral or positively charged organic molecules of different sizes and shapes and have very different physical and chemical properties from iodide. Therefore, for construction of alternative models of hNIS and assessment of its overall 3D structure, we also used transporters involved in selective anion uptake with available high-resolution X-ray or cryo-EM structures such as the bicarbonate transporters (hAE167 and hNBCe168) from the SLC4 family and the chloride/proton antiporter CLC-ec169 from the CLC family of chloride channels and transporters. The sequence identity for hNIS and all of the selected templates is low (20%–30%). The highest sequence identity was observed with the LeuT-fold proteins, especially for the 5 + 5 motifs.</p><!><p>Three separate alignments were made for hNIS with respect to each of the available protein folds (LeuT-fold with six templates, SLC4-fold with two templates, and CLC-fold with one template). The multiple sequence alignments were performed with the align tool of Modeller 9.18,70 which overlaps the template crystal structures and uses structural information from this overlap to assign the amino acid sequences to helices and loops. In the case of the LeuT-fold-based models, to improve alignment in the ion/substrate binding domain of hNIS, the putative sequences of TMs 1, 12, and 13 of hNIS were removed, and the remaining 5 + 5 fold of hNIS (residues 50–439) was aligned with the 5 + 5 folds of the six selected templates. TMs 1 and 12 of hNIS were modeled after TMs 1 and 12 of vSGLT (from PDB codes 2XQ2 and 3DH4, respectively), since the rest of the templates showed significant differences in this area. TM13 of hNIS was excluded from modeling with the LeuT-fold templates due to poor sequence identity with all templates. The refined alignment for the LeuT-fold models is presented in Figure S1 of the Supporting Information. For the models based on SLC4 and CLC templates, we prepared alignments using the full hNIS sequence. The sequence identity in both cases was lower than 20%.</p><!><p>3D models of hNIS with the three different protein folds (CLC, SLC4, and LeuT) were constructed with Modeller 9.1870 and then used to map key transport residues known from functional mutagenesis in NIS. The three models based on the different templates are presented in Figure 1. Only the LeuT-fold model of hNIS satisfied basic considerations of protein structure and provided structural consistency with the mutagenesis data. Therefore, only LeuT-fold templates were used for further modeling of hNIS. 3D structures of hNIS were built using the automodel function in Modeller 9.1870 with alpha helical constraints imposed on residues 85–99, 279–296, 393–401, 409–425, and 428–438 for better reproduction of the helical transmembrane domains. LeuT,21 AdiC,63 BetP,65 and dDAT66 show notable differences from vSGLT22 and Mhp164 in the structure and orientation of TM11 which implies differences in the gating and overall transport mechanism. Such subtle mechanistic differences are also inferred from DEER measurements on Mhp1 and LeuT.20 After assessment of similarities in the 3D structure of the six overlapped LeuT-fold templates, they were divided in two groups: vSGLT-like (including vSGLT and Mhp1) and LeuT-like (including LeuT, AdiC, BetP, and dDAT). Three LeuT-fold models of hNIS were generated afterward: a vSGLT-like model based on vSGLT and Mhp1, a LeuT-like model based on LeuT, AdiC, BetP, and dDAT, and a mixed model based on all six templates. For each model, 500 structures were generated with Modeller 9.18, using automodel, and the quality of the structures was assessed with pdfpdb,70 DOPE,71 and GA34172 scores. Afterward, 5 to 10 structures with the lowest DOPE and highest GA341 scores were selected and overlapped with the templates for assessment of overall backbone structure and position of residues identified as important from experimental data (X-ray diffraction where possible and functional mutagenesis).</p><p>Comparison of the three different types of hNIS models revealed significant structural differences in the area of extracellular gating (Figure S2), with the LeuT-like and mixed hNIS models adopting helical organization closer to the helical organization in LeuT. Considering the homology between vSGLT and hNIS, only the vSGLT-like model was then used for further structural refinement and MD simulations. The three vSGLT-like structures with the best side-chain overlap with the corresponding residues of interest in the vSGLT and Mhp1 templates were selected for further side-chain relaxation with ROSETTA MP.73</p><!><p>The three vSGLT-like hNIS structures selected from the previous step were subjected to side-chain relaxation with the ROSETTA MP protocols73 in the presence of an implicit membrane. The span files for the membrane were prepared manually with topology based on the initial 3D structures of hNIS. 2000 decoy structures were generated for each of the three individual hNIS models and were scored with the mpframework_smooth_fa_2012 function. The resultant structures were clustered with the cluster tool provided in ROSETTA. The center of the largest lowest energy cluster was chosen as a representative hNIS model for the remaining GCMC and MD simulations and is displayed in Figure 2A.</p><!><p>Identification of binding sites for ions in the protein scaffold is challenging. This work adapts a strategy used previously to map binding sites in the GltPh transporter.74 Briefly, water molecules were used to sample polar cavities in the protein core with a combination of GSBP/GCMC algorithms allowing flexibility of the reduced protein system and representing the membrane/solvent environment implicitly.62,75 For the Grand Canonical Monte Carlo (GCMC) calculations, only the protein residues within a sphere with a 15 Å radius centered at the center of mass for the consensus hNIS model were treated explicitly, with the rest of the system represented as generalized solvent boundary potential (GSBP).75 The GCMC simulation was run for 150 cycles with 5000 MC steps at 300 K with excess chemical potential (μ) for the external water reservoir set to −6.2 kcal/mol. The produced hNIS structures were relaxed with Langevin dynamics for 2 ps per GCMC cycle. Here, 50 relaxed structures from the GCMC calculations were then randomly selected and submitted for electrostatic potential calculations using linear Poisson–Boltzmann (PB) equations and an implicit membrane default setup from the PBEQ solver tool in the CHARMM-GUI server.76,77 The resulting electrostatic density maps were overlapped with the water density map evaluated from the GCMC trajectories, and the areas of highest water density in the putative binding cavity were assigned as either positively charged (putative anion binding area) or negatively charged (putative cation binding area). The position of the putative anion and cation binding sites are shown in Figure 2B as blue or red spheres, respectively.</p><!><p>The putative binding sites identified from the GCMC/PB calculations were used for further assessment of Na+ and I− binding in hNIS with MD simulations. The MD models included the two proposed transport stoichiometries for rat NIS (1:1 Na+:substrate and 2:1 Na+:substrate)6,78 and were tested with both I− and ClO4− as substrates. In all simulations, one Na+ was placed in a negatively charged area of the protein identified from our GCMC/PB calculations which corresponds to the Na2 binding site in the crystal structure of vSGLT.22 The models with 2:1 Na+:substrate stoichiometry had a second Na+ ion bound at the site with the negative electrostatic potential overlapping with the location of the Na1 site in LeuT21 (Figure 2B). For each anionic substrate (I− or ClO4−) and each tentative transport stoichiometry (1:1 and 2:1 Na+:substrate), 13 different substrate binding positions in the large slightly positive substrate binding cavity of hNIS, inferred from our GCMC calculations, were used as the initial guess for MD simulations (Figure 2B, blue spheres). In addition, all calculations were done with either no water molecules present in the substrate binding cavity at the initial MD step or with 12 water molecules at the remaining 12 out of the predicted 13 GCMC/PB substrate positions (excluding the site for the bound anion). Thus, a total of 104 separate models were constructed for all-atom MD simulations aimed at the refinement of binding sites in hNIS.</p><p>The hNIS structures with the respective ions and (where applicable) water molecules were then positioned in a tetragonal periodic box of size 90.44 Å × 90.44 Å × 101.83 Å, containing a POPC bilayer (95 and 92 POPC molecules in upper and lower leaflet, respectively), water (20 Å layers), and 0.15 M KCl solution with the CHARMM-GUI server.76,79,80 The orientation of the hNIS protein in the lipid bilayer was based on the vSGLT membrane orientation evaluated from the OPM server.81 The prepared hNIS models were equilibrated with NAMD 2.1282 following the standard six-step protocol for constrained equilibration.83 The CHARMM36 force field was used for protein, lipids, counterions, and water molecules.84,85 Iodide parameters were taken from ref 86. CHARMM-compatible parameters were developed for ClO4− using the GAAMP protocol87 (Table S1). The 50 ns long MD trajectories for each of the developed hNIS models were performed in the NPaT ensemble at T = 310 K and P = 1 atm. The aggregate sampling time was 5.2 μs. The trajectories were then analyzed with in-house tcl scripts developed for VMD 1.9.3.88</p><!><p>For further validation of our hNIS model, selected residues mapped as important for Na+ or substrate binding or gating modification in hNIS were evaluated experimentally for their roles in transporter uptake efficiency. We focus our experimental studies on positions/substitutions without published data.</p><!><p>Targeted mutagenesis of specific residues in human WT hNIS was performed using a QuikChange Lightning Site-Directed Mutagenesis kit (Agilent Technologies, Santa Clara, CA). Nucleotide substitutions were confirmed by sequencing analysis. A human influenza hemagglutinin (HA) tag was added at the C-terminus of hNIS to enable detection of mutant hNIS protein expression using an anti-HA tag antibody. The hNIS mutants were subcloned into a lentiviral expression vector.</p><!><p>VSV.G pseudotyped lentiviral vectors encoding human NIS WT or human NIS mutants were generated by triple transfection of packaging and vector plasmids with FuGENE HD (Promega, Madison, WI) into 293T cells as previously described.89 Titers of lentiviral vectors were determined by transduction of HeLaH1 cells and quantitated as transducing units/mL using quantitative PCR. HeLaH1 cells stably expressing the various hNIS mutants were generated by transduction of cells at a multiplicity of infection (MOI) of 2. Cells were placed under puromycin selection (0.5 μg/mL). Puromycin-resistant HeLaH1-NIS cells were characterized by flow cytometry analysis with 1:50 dilution of Alexa Fluor 647 conjugated antibodies recognizing the HA Tag 6E2 (Cell Signaling, Danvers, MA). Median fluorescence intensity of Alexa Fluor 647 positive cells was analyzed using the FlowJo (Ashland, OR) software.</p><!><p>HeLaH1 cells stably expressing human WT hNIS or hNIS mutants were seeded in 12 well plates at a density of 3 × 105 cells/well for 1 day prior to experiments. Cells were washed once in Hanks' balanced salt solution (HBSS, Gibco) supplemented with 10 mM HEPES (Genesee Scientific, San Diego, CA). The cells were incubated with the 125I− radioisotope (∼100,000 cpm) in a HBSS/HEPES buffer at 37 °C for 1 h. Cells were washed twice with cold HEPES/HBSS and lysed with 1 M NaOH for 10 min. Radioactivity was quantified with a PerkinElmer Wizard-2 2470 gamma counter for 125I−. All the experiments were performed in triplicate and repeated at least three times. To allow comparison between experiments, the uptake of 125I− by the hNIS mutants was normalized with the uptake of 125I− by WT hNIS. Results for the mutants were represented as fractions of the WT hNIS uptake, where the WT hNIS uptake is equivalent to 1 (Figure 3).</p><!><p>Figure 1 displays the 3D hNIS models (cylindrical representation) built with templates from three different protein families: hAE167 and hNBCe168 from the SLC4 family of anion transporters, CLC-ec169 from the CLC family of chloride transporters and channels, and vSGLT22 and Mhp164 from the LeuT-fold transporters. The positions of basic and acidic residues in the models are indicated with blue and red colors, respectively. The hNIS models based on the SLC4 and CLC templates led to regions composed of acidic or basic residues exposed to the hydrophobic core of the membrane. Many of the functionally relevant residues (yellow spheres) mapped from mutagenesis studies appear either in the connecting loops or in the peripheral helices in contradiction to the experimental data reviewed in this work. Only the hNIS model based on the LeuT-fold templates has a structure consistent with the expected membrane protein organization, featuring charged residues concentrated in the hydrophilic loop regions, with the exception of the catalytically relevant D191 which appears to be oriented toward the ion binding cavity. The putative binding residues implied from functional mutagenesis studies are located in the protein core, where the sodium and substrate binding areas have been found in all LeuT-fold transporters with known structures.90 Therefore, hNIS almost certainly has a LeuT protein fold as suggested by previous homology and hydropathy modeling2 instead of a protein fold characteristic for known secondary transporters of anions. The final 3D model of hNIS used in this study (Figure 2) is therefore based on two LeuT-fold proteins (vSGLT and Mhp1), chosen due to their overall structural similarities in the areas of outward-facing and inward-facing gates (which implies similarity of the transport mechanism), the homologous relation to hNIS (in the case of vSGLT),1,2 and the presence of multiple structures in different conformations (in the case of Mhp1).64,91,92</p><!><p>The location of the binding sites in the LeuT-fold family has been well established in several available X-ray structures from various structurally similar proteins1,20,21,63,64,66,90–95 and a plethora of experimental mutagenesis and uptake studies (Table S2). The LeuT-fold proteins feature a variety of transport stoichiometries, where the coupling Na+ ions vary in numbers from n = 0 (e.g., the Na+-independent arginine transporter AdiC)63 to n = 3 (human betaine/GABA transporter).96 In some cases, K+- and Cl−-dependent transport is also observed.66,96–98 Most of the LeuT-fold proteins have at least one Na+ ion bound at a conserved sodium binding site, which corresponds to site Na2 in LeuT.20,99,100 LeuT features also a second sodium ion bound at its Na1 binding site, in immediate proximity to the primary substrate binding site often referred to as the S1 site.21 The hNIS homologues from the human SGLT family use a different number of cations coupled to substrate cotransport depending on the particular family member: hSGLT1 and hSGLT2 couple two or one Na+ ions to glucose transport, respectively.101 Cation binding sites labeled Na1 and Na2 as well as a primary substrate binding site S1 have been identified in several LeuT-fold proteins with available high-resolution X-ray structures.100 A recent structure of an outward-facing open state of the sialic acid transporter, SiaT, which belongs to the LeuT-fold family of proteins, features a new Na+ binding site, labeled as Na3,102 close to site Na2 in vSGLT. The location of the second sodium binding site in hSGLT1 has recently been delegated to the Na3 site.103 The vSGLT and Mhp1 templates used for our hNIS model feature a single sodium ion in the Na2 binding site.22,64 Taken together with the diverse substrates transported by the LeuT-fold proteins, it is evident that the LeuT-fold organization of the transmembrane domains supports a fairly powerful mechanism of secondary transport that can be easily adapted to the chemistry and geometry of many different substrates. The transport stoichiometry for human NIS is as of yet unknown. The available electrophysiological evidence suggests a 2:1 Na+:I− transport stoichiometry in rat NIS and is in support of two cation (Na+) binding sites which we label Na1 and Na2, respectively, in analogy to the Na+ binding sites in LeuT. We have probed both tentative sites Na1 and Na2 in our hNIS model.</p><p>While all of the known secondary transporters feature a primary binding site for a substrate (site S1), the existence of a second, presumably allosteric, substrate binding site, S2, has also been proposed in LeuT transporters based on molecular modeling, single-molecule, and functional studies.94,104–107 The LeuT-fold protein, L-Trp transporter MhsT, features two substrate binding sites labeled S1 and S2 using nomenclature introduced for substrate binding sites in LeuT.108 Two substrate binding sites have been identified in the human serotonin transporter hSERT and have been discussed in relation to therapies for mental disorders and addiction.98,109 Existence of a second substrate binding site has been suggested for the galactose transporter, vSGLT, and for the proline transporter, PutP, based on a combination of saturation, mutagenesis, and uptake data.47 vSGLT and PutP are the most likely homologues of the SLC5 protein family and its members hSGLT1, hSGLT2, and hNIS.47 Therefore, hNIS may potentially feature two substrate binding sites for its anionic substrates referred in our work as sites S1 and S2, respectively.</p><p>The first step in the elucidation of the potential locations of ion and substrate binding sites in our hNIS model was the collection of excess density maps for a polar molecular probe (water in this work) complemented by ensemble-averaged Poisson–Boltzmann electrostatic potential maps. The combined GCMC/PB data in Figure 2B highlights well-defined binding cavities attractive to cations as well as a well-defined binding pocket for I−. GCMC/PB mapping suggests a potential Na+ binding site in the vicinity of the D191 residue, which corresponds to the Na2 binding site preserved in most known LeuT transporters20,99,100 or the Na3 site in the sialic transporter SiaT.102,103 Another potential sodium binding site, corresponding to site Na1 in LeuT21 was found in the area of S66 and F67. The remaining portion of the cavity is positively charged and expected to attract negative substrates, such as I−. It can be divided into two sections corresponding to the S1 and S2 sites found in vSGLT (Figure 2B).47</p><p>To assess in more detail the sodium and substrate binding in hNIS, we ran a number of 50 ns long MD simulation of hNIS where I− or ClO4− were placed in one of the potential substrate binding locations identified from GCMC (13 positions per simulated system, Figure 2B). We constructed models to test transport stoichiometries proposed from the electrophysiological studies in the rat NIS: 2:1 Na+:substrate or 1:1 Na+:substrate.2 In all models, one Na+ ion was placed in the highly preserved Na2 binding site. In the models with the 2:1 Na+:substrate ratio, the second Na+ ion was placed in the putative Na1 binding site, as suggested by our GCMC results (Figure 2B).</p><p>The graphs in Figure S3 display the contact frequencies of I−, ClO4−, and Na+ in the Na1 and Na2 sites, with the protein matrix, evaluated from cumulative MD trajectories (e.g., the contact frequencies for iodide are added from all MD trajectories featuring an iodide ion, including those with or without water in the binding cavity at the MD onset and those with 2:1 and 1:1 Na+:I− ratio). The individual contributions for these cumulative graphs are displayed in Figure S4. The protein residues with the highest contact frequencies for each ion are listed in Table S2 of the Supporting Information. The contact frequencies indicate the percent of the MD steps in which an ion of interest can be found within 5 Å from the protein atoms and can be used for identification of protein residues which have frequent contact with the ions (i.e., potential binding sites where the ions tend to reside during the MD simulations). The sodium ion at the Na2 position consistently shows high contact frequencies with a small number of residues indicative of a well-defined binding site and stable ion binding.</p><!><p>Our findings for the Na2 site in hNIS are mostly consistent with the previously identified Na2 site in a homology model of hNIS based solely on vSGLT.46 The contact frequency patterns for Na+ in site Na2 indicate stable coordination of the Na+ ion by several key residues: S62, A65, S66, M68, Y178 and a nest of carbonyl groups of residues M184–188, T190, D191, Q194, S349, and T354 (Figure 4). Mutations at positions S66, S349, and T354 have proven detrimental to I− uptake.30,38,46 The I− uptake is impacted also by mutations at positions M68, K185, and Q194 (Figure 3).46 The available crystal structures of vSGLT, Mhp1, and LeuT show that their conserved Na+ (site Na2) sodium ion is coordinated by analogues of A65, M68, G350, S353, and T354 in hNIS (Table S2).21,22,64</p><p>Mutations of D191 in hNIS, which remove the negative charge (D191N, D191C), eliminate transport completely (Figure 3). Residue D191 in hNIS corresponds to D204 in hSGLT1, whose neutralization impairs substrate uptake and leads to the formation of a glucose-gated proton channel.110</p><p>Residue D189 in vSGLT, which corresponds to D204 in hSGLT1 and D191 in hNIS, has been shown to interact with the Na+ in site Na2 in MD simulations of the Na+ and substrate exit from the inward-facing vSGLT structure.111–113</p><p>Introduction of a positive charge in the vicinity of D191 (Q194R, Q194K, M198R, and M198K shown in Figure 3) leads to drastic decrease in the iodide uptake. It is likely that the positive charge of the substituted amino acid in these cases interferes with Na+ binding at the Na2 binding site and does not adopt the allosteric modulator role of the Na+ ion, as R262 and K158 do in the Na+-independent LeuT-fold transporters CaiT and ApcT, respectively.114,115 Taken together, these results show that a Na+ ion is likely present in hNIS at the areas of TMs 2, 6, and 9, which in our inward-facing open model of hNIS may correspond to either site Na2 in vSGLT or site Na3 in SiaT, and that this presence is determined by the strong electrostatic interaction with a negatively charged residue (conserved in the homologues of hNIS, such as vSGLT, hSGLT1, and PutP) and a number of conserved polar residues (Gln, Ser, Thr) present in many LeuT-fold transporters (Table S2).</p><!><p>Among the residues which emerge as potential Na+-coordinating centers in site Na1 are S64, S66, F67, Q72, Y259, L289, and T354, which are also involved in substrate binding (Table S2, Figures S3 and S4, Figure 5). Similar overlap between the substrate and Na1 binding sites has been observed in the crystal structure of LeuT,21 where the Na+ in site Na1 is coordinated by the carboxylate group of the bound Leu substrate and is also inferred from Kd values evaluated from electrophysiology for Na+ and glucose in hSGLT1.113 Unlike the stably bound Na+ (site Na2), the Na+ in site Na1 is fairly mobile, and in the cases of a dehydrated binding cavity often forms an ion pair with the I− anion in the polar substrate binding pocket. Similar behavior has been suggested for the Na+ and CO32− ions bound in the hNBCe1 protein of the SLC4 anion transporter family116 and implies that the Na+ ion in site Na1 may be required for strengthening the anion–protein interactions in the uncharged but slightly polar substrate binding cavity of hNIS, which lacks traditional anion binding residues such as Arg or Lys. Interestingly, the Na+ in site Na1 in LeuT is in proximity to two negative residues (E287 and E290) which are responsible for its coordination and stabilization.21 Molecular dynamics simulations and smFRET studies show that the protonation state of E290 also modulates some of the Na+-coupled conformational changes consistent with allosteric control during the secondary transport in LeuT, which in turns affects the protein affinity for Na+.117–120 The hSGLT1 system, which transports two Na+ ions, has a negatively charged residue (D294) in the vicinity of its putative Na1 site suggested from homology to vSGLT.113 Such residues are missing in vSGLT, PutP, and Mhp1, which have 1:1 Na+:substrate transport stoichiometry.22,56,64 Thus, the I− ion in the hNIS transporter may in turn adopt the role of a negatively charged amino acid residue for stabilization of a second Na+ ion in the polar core of hNIS and may aid in the ion permeation of both ions through the large polar substrate binding cavity after their dehydration. Rehydration of the binding cavity occurs naturally as the inward-facing state opens and is a required step in the ion and substrate exit from the LeuT-fold secondary transporters during the alternate access mechanism.121</p><!><p>The I− and ClO4− ions mostly dwell in the positively charged cavity determined from GCMC calculations at least for the duration of the 50 ns long MD simulations performed here. The I− and ClO4− density maps (Figure S5) point at the presence of two major binding regions in the anion-accessible area of the protein. The amino acid composition of these binding sites is presented in Table S2. Comparison to the previously identified hSGLT1 and PutP binding sites (Table S2) established that the two putative substrate binding sites in hNIS are indeed consistent with sites S1 and S2. Figure 5 displays representative coordination of I− within the hNIS core corresponding to the protein areas with the highest anion density. The residues of the binding sites are color coded: dark blue for site S1, cyan for site S2, and green for residues involved in both sites S1 and S2. Coordination of the ClO4− ion is shown in Figure S6 with the same color coding.</p><p>The amino acid residues common for both S1 and S2 sites are S66, F67, S69, Q72, Y144, Y259, S353, and T354. Functional mutagenesis and comparison to analogous residues in vSGLT, hSGLT1, PutP, Mhp1, and LeuT show that these residues are essential for substrate uptake. The T354P mutation in hNIS is found in patients with congenital ITDs.25,28 Cysteine or alanine mutations of residues S66 and S353 in rat NIS,38,46,122 as well as S69 in human NIS (Figure 3), lead to a drastic decrease in NIS function. Substitution of S69 with the negatively charged Asp residue eliminates transport. Removal of the amino group at position Q72 (Q72G, Q72S, Q72T) in human NIS is detrimental to I− uptake, even if the substituting residue features a polar OH group, suitable for binding to anions. A Q72N substitution leads to hNIS with lowered but observable I− uptake, hinting at the importance of the presence of an amino group at this position (Figure 3).</p><p>PutP mutations to cysteine of the A53-L60 residues, analogous to the A65-Q72 stretch in hNIS, have severe impact on the proline uptake.53 The Cys mutants of PutP are sensitive to inhibition with N-ethylmaleimide (NEM) and are accessible to the fluorescent dye fluorescein-5-maleimide, indicating that they are part of a solvent-accessible cavity, which is likely part of the substrate and ions permeation pathway in PutP. In addition, residues Q72, Y144, S353, Y259, and T354 in hNIS correspond to substrate binding residues found in the substrate-bound crystal structures of vSGLT, Mhp1, and LeuT (Table S2). Residue Y259 is the hNIS analogue of residues Y263 in vSGLT,22 Y290 in hSGLT1,113 Y248 in PutP,56 and F253 in LeuT,21 which have been identified structural components of both sites S1 and S2 in these systems and have been implicated in gating and allosteric control.47 Ala substitution of the Tyr at position 259 in rat NIS drastically decreases iodide uptake, while introduction of a Phe residue recovers some of the transport function, suggesting that a bulky aromatic residue is critical at this position.44 MD simulations in vSGLT reveal that rotation of residue Y263 is coupled to substrate release123 although nongated substrate release is also possible in this system.124 In hSGLT1, the aromatic nature of Y290 and its hydroxide group have been implicated both in substrate and Na+ binding.113 Importantly, Cys and Gly substitutions of Y248 alter the PutP:proline stoichiometry from 2:1 to 1:1 and lead to severe reduction of the proline uptake, regardless of Na+ binding.47</p><!><p>The S1 binding site contains in addition residues K185, N262, Q263, Q265, T357 and S358. Residue Q263 has been implicated in Na+ (site Na2) binding in a previous homology model of hNIS,46 and substitutions with Asn in rat NIS leads to a drastic decrease of radioiodine uptake and changes in Km for both Na+ and I−. K185A and Q265A/C substitutions in human NIS have significant detrimental impact on I− uptake (Figure 3). The analogue of the hNIS residue N262 in PutP is residue Q251 (Table S2). A Q251C substitution in PutP affects both substrate and Na+ kinetic parameters.49 Q251 is also connected to P252, which is another residue controlling the molar binding ratio in PutP.47 Residue T357 in hNIS corresponds to residues S368 (protein:substrate stoichiometry control) in vSGLT,113 C344 (protein:substrate stoichiometry control) in PutP,113 and N318 (substrate binding residue in the crystal structure) in Mhp1,64 which have proven critical for substrate uptake in these systems. A T357A mutation in hNIS drastically decreases the radioiodine uptake and alters the Km and Vmax parameters for both I− and Na+.38 S358 in hNIS corresponds to residue Q345 in PutP, which is implicated in substrate binding from electrophysiology measurements.50</p><!><p>The residues which can be considered unique for the S2 binding site are Q94, W255, Y144, V148, F417, and M420. Most of them have analogues in the substrate binding sites established from the available crystal structures of vSGLT, Mhp1, and LeuT (Table S2). W255A/C and F417Y/C mutations in hNIS eliminate transport and substitution of Q94 with other polar or charged residues (Arg, Lys, Asn, Glu) severely decreases I− uptake (Figure 3). Taking into account the low contact frequency of I− and the higher contact frequency for the larger, tetrahedral ClO4− ion, Q94 emerges as a residue from the second coordination sphere of the anions in the vicinity of residues F67 and Q72 which are critical for transport (see above) and are also part of the Na1 putative binding site (Table S2, Figures S3 and S4). Q94 has a higher contact frequency with the Na+ in site Na1 than with I− (Figure S3), which indicates that Na+ may be coordinated by Q94, while the I− ion is in its S2 site. In addition, Q94 is a neighboring residue of G93 (G93R is one of the known pathological mutations found in patients with congenital ITD2,125), which has been implicated in the control of Na+:I− transport stoichiometry.44 It is possible that the alteration of the Gly side chain at the somewhat peripheral position G93 interferes with residue Q94 from the S2 and Na1 binding sites of hNIS, which in turn may impact other Na1 binding residues (e.g., F67, Q72, Y259) leading to the observed changes in Na+:I− stoichiometry in NIS.</p><p>Previous homology models based on vSGLT place residue W255 at the periphery of TM7, and the orientation of W255 has been discussed in conjunction with residue G93 as a "ball and socket" structure, relevant to conformational transitions in NIS.44 Theoretical I− binding studies with the ABF method and the Drude force field in this model of hNIS in the absence of Na+ ions have identified residues F67, F87, M90, G93, Q94, W255, and Y259 as a putative binding site for I− and have revealed that the G93T mutation impacts the binding free energy in this site by displacing the W255 side chain into the binding pocket of hNIS.45 In our model, W255 points inward, toward the protein core, and is part of the binding region in hNIS as suggested by the anion density maps from our MD simulations. All transporters listed in Table S2 feature a Trp or a Phe residue at this position, and these residues are involved in substrate binding as seen from the crystal structures of vSGLT, Mhp1, and LeuT.21,22,64 The remaining LeuT-fold proteins used for the multiple sequence alignment (Figure S1) also have a Trp or Phe residue involved in substrate binding in the vicinity of W255. The presence of such aromatic residues in the binding site of anion binding proteins is not unique for the LeuT-fold architecture. Two Trp residues have been identified in the halide binding site of the water-soluble dehalogenases,126 and Tyr is a well-known binding residue in the CLC-ec1 chloride transporter,127 while chloride in the dopamine and serotonin transporters is coordinated with conserved Tyr and Phe residues.66,98 Halide binding in the fluoride ion channels of the Fluc family also involves highly conserved Phe residues.128 The SLC4 family of anion transporters features Phe residues in their putative binding sites, and some of these residues have a strong impact on transport, as suggested by mutagenesis and uptake data.68 The aromatic residues in the substrate binding pockets play a role not only in substrate binding, for which the indole group of the Trp and the hydroxide group of Tyr are well equipped, but also as part of a hydrophobic constriction zone, which aids in the dehydration of the substrate, necessary for stronger binding within the protein.22 The critical role for I− uptake of F417 in hNIS is supported by kinetic modeling in vSGLT and hSGLT1, which shows that the well-established gating residue F424 in vSGLT and its hSGLT1 analogue F453 are involved in coordination of conformational changes and coupling of Na+ and sugar transport in addition to the traditional role of an energy barrier.22,112,113,129,130</p><p>A region with small I− density immediately adjacent to sites S1 and S2 can be discerned by our MD simulations in the hNIS systems loaded with I− and 2Na+ ions. This binding site is composed of residues F67, Q72, Q94, and the W255-G260 (Figure S7). Given the small I− density in this area, this site may represent a transient shallow-binding/access region, present during the early production simulations of I− binding to hNIS.</p><!><p>For the most part, the contact frequency and anion density patterns of ClO4− and I− are qualitatively similar and involve residues from sites S1 and S2 as described above (Figures S3–S6). Most of the coordination is achieved through N–H bonds (from the peptide backbone, amino groups in Asn and Gln, or the indole ring in Trp) or O–H bonds (from the hydroxyl groups of Ser, Thr, and Tyr residues), although involvement of the C–H bond from various polar and nonpolar side chains is also possible (Figure 5, Figure S6). Due to the tetrahedral structure of the ClO4− ion and its slightly larger ionic radius (about 0.15 Å larger than the ionic radius of I−),131 it appears to overlap better with the surrounding protein residues, which results in longer duration of anion binding and, consequently, in higher contact frequencies, especially in the area of the S1 binding site (Table S2, Figure S3). The bulk of the ClO4− ion density falls within the S1 site or at the interface between sites S1 and S2 (Figure S5). Perchlorate also forms an ion pair with the Na+ (site Na1) ion more rarely than I− in the MD simulations with a Na+:substrate 2:1 ratio.</p><p>The water presence in the substrate binding cavity of hNIS at the onset of the MD simulations generally leads to destabilized protein–ion binding, rapid opening of the cytoplasmic vestibule of hNIS, and frequent exit of at least one ion from the protein, evident from the shorter than 50 ns ion–protein binding duration displayed in Table S3 and the lower percentage of ion–protein contacts in the contact frequency graphs (Figure S4). The most frequent ion exit induced by the water present in the core of the protein is observed in the Na+:ClO4− 2:1 and Na+:I− 1:1 systems, which points to higher protein–ion binding instability for these stoichiometries.</p><p>The S1 site is closer to the intracellular exit route for anions (Figure 2B). Due to the inward-facing semiopen conformation of our hNIS model, substrates in this site tend to unbind and move to the intracellular solution in several MD simulations (Table S3). The primary substrate for hNIS (I−) displays more frequent dissociation compared to ClO4− (Table S3). In fact, perchlorate unbinding was observed in only one of the studied hNIS systems, where ClO4− was placed in site S1. No ClO4− exit from site S2 was observed regardless of the protein core hydration. Enhanced protein–anion coordination may explain the less frequent ClO4− dissociation from the substrate binding sites of hNIS and the I− uptake inhibition effect of ClO4− in NIS, where perchlorate may bind stably to one of the available substrate binding sites and prevent iodide binding and translocation. Nevertheless, our MD simulations do not provide conclusive evidence for preference of a 1:1 Na+:ClO4− transport stoichiometry in human NIS. Such evidence would require detailed modeling and experimental studies involving assessment of the thermodynamics of anion binding in hNIS with different ion loads which is outside of the scope of the current work.</p><p>While multiple further studies would be necessary to identify the specific roles of the listed residues in the anion uptake function of hNIS, published studies and uptake experiments performed in our work show that the suggested residues from our MD simulations are crucial for hNIS transport and often correspond to residues involved in direct substrate binding and/or regulation of protein:substrate and Na+:substrate stoichiometry in other structurally similar proteins.</p><!><p>The governing hypothesis of secondary transport, known as the "alternating access mechanism", dictates that the binding sites for the substrate and accompanying species are consecutively uncovered to the outside and inside of the cell as the structure of the protein changes.121 The alternating access hypothesis has been backed by crystallographic and modeling studies on a number of secondary transporters, most prominently the bacterial leucine transporter LeuT, which has been crystallized in outward-facing open, inward-facing open, and occluded states in the presence and absence of bound Na+, substrates, and inhibitors.100 These studies have elucidated the function of LeuT and other proteins featuring the LeuT-fold.20,100 Generally, macroscopic motions in the gating regions of the proteins tend to create or destroy salt bridges and hydrophobic plugs which then obstruct or open water filled permeation pathways through the protein cores, necessary for the substrate and ion translocations. These motions are allosterically triggered by binding of Na+ ions and substrates.20,132</p><p>Review of the contact frequency maps (Figures S3 and S4, Table S2) shows that two key for transport residues S66 and T354 (see above) are in contact with all ions (Na+ in sites Na1 and Na2, I−, ClO4−in either S1 or S2 site) present in the hNIS binding cavity. These residues and their neighboring residues from the A65-S69 and T353-S358 stretches therefore may be part of the allosteric interaction network responding to the ion binding and triggering the major conformational changes necessary for ion translocation through the membrane. The plausible permeation pathway in our model of hNIS can be traced with water maps, generated from the MD simulations. A sample water density map for hNIS loaded with two Na+ and one I− ions is shown in Figure 6 and demonstrates a water accessible vestibule at the cytoplasmic side of hNIS. The water density maps of the remaining hNIS models with different ion loads look similar to the map in Figure 6. Such water distribution in our model of hNIS is expected since the crystal structure of vSGLT used in the homology modeling is a dynamic semioccluded or inward-facing open state,112 which relaxes further during the MD simulations. A number of residues line the cytoplasmic water vestibule: S116, T117, Y118, T134, Y137, A180, V181, G182, K185, N262–Q265, Q267, R268, T357, S358, N360, A361, A364,and V365 (Figure 6, magenta sticks). The water molecules permeate the protein to the area of the Na2 and S1 binding sites and provide an exit pathway for the ions bound to them. NMA accessibility studies show that areas of TM2 and TM9 form a hydrophilic and water accessible pore in PutP, consistent with the one observed in hNIS.50,53 MD studies of the gateless galactose exit from vSGLT identify several key residues lining the galactose exit pathway: S368, N371, S372, T375, and R273.124 They correspond to T357, N360, A361, A364, and R268 in the cytoplasmic water vestibule of hNIS, determined from our MD simulations. A Q267E mutation in hNIS, which leads to a decreased turnover rate, is found in patients with congenital ITD.37 Of interest in this area is also residue K185, which belongs to the loop between TM5 and TM6. A positive residue in this region is unique for hNIS and cannot be found in the otherwise hydrophobic corresponding loops in PutP, vSGLT, and hSGLT1. The backbone of residue 185 forms a part of the carbonyl nest, where Na+ (site Na2) is coordinated (Figure 4). The long positive side chain stretches toward the substrate binding site S1 and is involved in direct anion coordination once the anion starts its exit from the protein, implying a role in protein gating. Due to its position on a cytoplasmic loop, Lys demonstrates high flexibility during the MD simulations and may even form a short-lived salt bridge with residue D191, once Na+ is displaced from its Na2 binding site (Figure S8), which may have implications for the inward-facing to outward-facing transition of hNIS once all ions are released in the cytoplasm. A K185R mutation has a negligible effect on iodide uptake, but substitutions which omit the positive charge (K185A, K185Q, and K185E) prove more detrimental to iodide transport. Other residues in the vicinity of K185, which may play a role in the gating process are T117, Y118, Y137, M184, A264, T357, and A361. These residues have bulky hydrophobic methyl or phenyl groups and form an intracellular hydrophobic plug in our semioccluded model of hNIS which opens upon hydration as the MD simulations progress and allows exit of the ions from the protein. At the extracellular site, which in our hNIS model is occluded, residues V76, M90, and F417 (Figure 6, pink sticks) appear in the hydrophobic constriction zone made of residues M73, Y87, and F424, respectively, in the crystal structure of vSGLT.22 An aromatic residue is present at this position in PutP, hSGLT1, Mhp1, and LeuT (Table S2). Mutation of F417 to Cys eliminates I− uptake (Figure 3), proving that residue 417 is critical for anion transport and that hNIS requires the presence of a hydrophobic residue at this position. Similarly, a M90C mutation decreases drastically the I− uptake (Figure 3). The opening of the extracellular gate in vSGLT is attributed to residues P436 and G437, which allow for the necessary structural deformations in TM11.22 Similar Pro-Gly or Gly-Pro motives are found in this area of TM11 in hSGLT1, hNIS, PutP, Mhp1, and the sialic acid transporter SiaT.22,56,64,102 Comparison between the outward-facing open state of SiaT and the occluded structures Mhp1 and vSGLT lends support to the hypothesis that the Pro-Gly pair in TM11 is responsible for the conformational changes of TM11 leading to the opening and closing of the extracellular gate in these transporters. Therefore, it can be expected that the Pro-Gly motif in hNIS (G425-P426, red spheres in Figure 6) has a similar conformation relevant function.</p><!><p>A combination of homology modeling and molecular dynamics simulations with mutagenesis and uptake experiments led to the development of a cross-validated model for the human NIS transporter. Two putative Na+ binding sites (Na1 and Na2) and a large substrate binding cavity, with two substrate binding sites, S1 and S2, were mapped by Grand Canonical Monte Carlo calculations and assessed further with MD simulations. A number of residues lining the ion and water access pathways to the core of the hNIS transporters and the putative ion binding sites were selected for mutagenesis and uptake experiments, and their functional significance was confirmed by the detrimental impact on uptake following their mutation. The model provides a structural rationale for the substrate uptake and conformational dynamics of hNIS. Taken together, the comparison to a number of LeuT-fold transporters and the results from functional mutagenesis presented in previous studies on hNIS and in our current work afford an extensive validation of our homology model and provide an indispensable structural template for future optimizations of hNIS for enhanced SPECT/PET radiotracer transport.</p>

PubMed Author Manuscript

Actin filament nucleation and elongation factors \xe2\x80\x93 structure-function relationships

The spontaneous and unregulated polymerization of actin filaments is inhibited in cells by actin monomer-binding proteins such as profilin and T\xce\xb24. Eukaryotic cells and certain pathogens use filament nucleators to stabilize actin polymerization nuclei, whose formation is rate-limiting. Known filament nucleators include the Arp2/3 complex and its large family of Nucleation Promoting Factors (NPFs), formins, Spire, Cobl, VopL/VopF, TARP and Lmod. These molecules control the time and location for polymerization, and additionally influence the structures of the actin networks that they generate. Filament nucleators are generally unrelated, but with the exception of formins they all use the WASP-Homology 2 domain (WH2 or W), a small and versatile actin-binding motif, for interaction with actin. A common architecture, found in Spire, Cobl and VopL/VopF, consists of tandem W domains that bind three to four actin subunits to form a nucleus. Structural considerations suggest that NPFs-Arp2/3 complex can also be viewed as a specialized form of tandem W-based nucleator. Formins are unique in that they use the formin-homology 2 (FH2) domain for interaction with actin and promote not only nucleation, but also processive barbed end elongation. In contrast, the elongation function among W-based nucleators has been \xe2\x80\x9coutsourced\xe2\x80\x9d to a dedicated family of proteins, Eva/VASP, which are related to WASP-family NPFs.

actin_filament_nucleation_and_elongation_factors_\xe2\x80\x93_structure-function_relationships

Introduction<!>Actin filament nucleators<!>The W domain and actin filament nucleation<!>The Arp2/3 complex and Nucleation Promoting Factors (NPFs)<!>Tandem W domain-based filament nucleators<!>Role of oligomerization on the nucleation activities of W-based nucleators<!>The Arp2/3 complex and NPFs as a specialized form of tandem W nucleator<!>Model of nucleation by NPFs-Arp2/3 complex<!>Leiomodin (Lmod) and the nucleation of actin filaments in muscle cells<!>Ena/VASP proteins as dedicated elongation factors among W-based filament nucleators<!>Profilin and the Pro-rich regions of nucleation and elongation factors<!>Perspectives

<p>The actin cytoskeleton is intimately involved in most cellular functions, including cell motility, cell adhesion, endo/exocytosis, intracellular trafficking and the maintenance of cell shape and polarity (Chhabra and Higgs, 2007; Galletta and Cooper, 2009; Le Clainche and Carlier, 2008; Pollard and Borisy, 2003). In addition, many pathogens disrupt or kidnap the host cell actin cytoskeleton during infection (Bhavsar et al., 2007; Cossart and Toledo-Arana, 2008; Gouinet al., 2005). These processes are characterized by rapid oscillations of actin polymerization/depolymerization under tight temporal and spatial regulation. At its most basic level, the assembly of actin cytoskeletal networks depends on the regulated transition of cellular actin between its monomeric (G-actin) and filamentous (F-actin) states (Figure 1). Actin is an ATPase, and nucleotide hydrolysis by actin is a critical factor regulating the transition between the G- and F-actin states. Actin monomers join the fast growing barbed (or +) end of the filament primarily in the ATP state. Hydrolysis takes place in the filament, and ADP-actin monomers dissociate mainly from the pointed (or −) end. However, this simple steady state polymerization/depolymerization mechanism, known as actin filament treadmilling, cannot account for the vast variety of actin processes and actin networks observed in cells. Hundreds of G- and F-actin-binding proteins, along with signaling and scaffolding proteins, become involved in the regulation of actin dynamics (Pollard and Borisy, 2003). Actin-binding proteins (ABPs) have diverse functions, including actin monomer sequestration, filament barbed and pointed end capping, filament severing, and filament crosslinking. An important group of ABPs are those that regulate the de novo formation of actin filaments, which include actin filament nucleation and elongation factors. A number of excellent reviews have been written recently about these proteins (Chesarone and Goode, 2009; Faix and Grosse, 2006; Goley and Welch, 2006; Goode and Eck, 2007; Higgs, 2005; Paul and Pollard, 2009; Pollard, 2007; Qualmann and Kessels, 2009; Renault et al., 2008). This review differs in that it focuses on general structure-function principles of filament nucleation and elongation.</p><!><p>Actin is the most abundant protein in most eukaryotic cells, where its concentration (ranging from 100 to 500µM in non-muscle cells) is much higher than the critical concentration for monomer addition at both the barbed end (0.1µM) and the pointed end (0.7µM) of the actin filament (Pollard and Borisy, 2003). Yet, actin monomer-binding proteins such as profilin and Tβ4 inhibit the spontaneous polymerization of actin filaments. Cells use filament nucleators to stabilize actin polymerization nuclei, whose formation is rate-limiting during actin assembly (Sept and McCammon, 2001) (Figure 1). Filament nucleators constitute a fast evolving and relatively recent field of investigation. The Arp2/3 complex was first purified more than 10 years ago (Macheskyet al., 1994), but it was not until the discovery of ActA as a Nucleation Promoting Factor (NPF) at the surface of Listeria monocytogenes that the nucleation capacity of NPFs-Arp2/3 complex was fully recognized (Goley and Welch, 2006; Welchet al., 1998). Almost simultaneously, eukaryotic NPFs belonging to the WASP/WAVE-family of proteins were identified (Machesky and Insall, 1998; Machesky et al., 1999; Rohatgi et al., 1999; Winter et al., 1999; Yarar et al., 1999). Subsequently, formins were shown to catalyze not only nucleation but also processive barbed end elongation (Goode and Eck, 2007; Higgs, 2005; Pollard, 2007; Pruyne et al., 2002; Zigmond et al., 2003). Recently, a series of new filament nucleators have been discovered, both in eukaryotic cells and bacterial pathogens, including Spire (Quinlan et al., 2005), Cobl (Ahuja et al., 2007), VopL (Liverman et al., 2007), VopF (Tam et al., 2007), TARP (Jewett et al., 2006) and Lmod (Chereau et al., 2008). These proteins control not only the time and location for actin polymerization, but also the specific type of actin filament networks that they generate.</p><p>The actin filament (Figure 1A) can be described as either a single left-handed short-pitch helix, where consecutive lateral subunits are staggered with respect to one another by half a monomer length, or two right-handed long-pitch helices of head-to-tail bound actin subunits (Holmes, 2009; Holmes et al., 1990; Oda et al., 2009). As discussed below, different actin filament nucleators work by different mechanisms, stabilizing small actin oligomers (dimers, trimers and tetramers) along either the long- or the short-pitch helices of the actin filament.</p><!><p>With the exception of formins, all known actin filament nucleation and elongation factors use the WASP-Homology 2 (WH2 or W) domain for interaction with actin, though in Lmod and the Arp2/3 complex other domains contribute as well. The W domain has a short size (17–27aa) and is generally poorly conserved, making it difficult to identify based on sequence analysis alone (Dominguez, 2007). Other distinctive features of the W domain include its abundance and functional versatility. The N-terminal portion of the W domain forms a helix that binds in the hydrophobic cleft, or target-binding cleft (Dominguez, 2004), between subdomains 1 and 3 at the barbed end of the actin monomer (Chereau et al., 2005; Hertzog et al., 2004) (Figure 2). After this helix, the W domain presents an extended region that climbs toward the pointed end of the actin monomer. This region has variable length and sequence but comprises the conserved four residue motif LKKT(V). The N-terminal helix and LKKT(V) motif constitute the conserved core of the W domain. However, the W domain displays remarkable plasticity. For example, the actin monomer sequestering protein Tβ4 is a unique member of the W domain family (Paunola et al., 2002), featuring an additional helix C-terminal to the LKKT(V) motif that binds atop actin subdomains 2 and 4 and caps the pointed end of the actin monomer (Irobi et al., 2004).</p><p>The W domain often occurs in tandem repeats, which is a common architecture among filament nucleators, found in Spire (Quinlan et al., 2005), Cobl (Ahuja et al., 2007) and VopL/VopF (Liverman et al., 2007; Tam et al., 2007). The protein TARP (translocated actin recruiting phosphoprotein) contains a single W domain, but forms large oligomers (Jewett et al., 2006). The W domain also participates in filament nucleation by the Arp2/3 complex through NPFs, which can have between 1 and 3 W domains (Goley and Welch, 2006; Pollard, 2007; Zuchero et al., 2009). Finally, one of the three actin-binding sites of Lmod is also a W domain (Chereau et al., 2008).</p><p>It is important to note, however, that the presence of multiple copies of the W domain does not automatically mean that a protein is a filament nucleator. For example, amoeba actobindin (Hertzog et al., 2002), Drosophila ciboulot (Hertzog et al., 2004) and C. elegans tetrathymosin (Van Troys et al., 2004) present two-and-a-half, three and four copies of the W fold, respectively, but do not nucleate actin filaments. Evolutionarily, these three proteins share more in common with Tβ4, in particularly the presence of Tβ4-related sequences C-terminal to the LKKT(V) motif, than with classical W domains of the kind found in WASP family proteins (Chereau et al., 2005). However, ciboulot and actobindin do not sequester actin monomers like Tβ4, but rather promote filament barbed end growth in a way analogous to profilin (Carlier et al., 2007; Hertzog et al., 2002). These two proteins form 1:1 complexes with actin, suggesting that only one of their actin-binding sites is fully functional (Hertzog et al., 2004; Hertzog et al., 2002). In contrast, tetrathymosin appears to bind multiple actin monomers and has both monomer sequestering and filament-binding properties (Van Troys et al., 2004). Tβ4, actobindin, ciboulot and tetrathymosin are thus examples of how changes in the sequence of the W domain and modular structure of the proteins in which it is found give rise to diverse functions in the regulation of actin cytoskeleton dynamics (Dominguez, 2007).</p><!><p>The Arp2/3 complex consists of seven proteins, including two actin-related proteins, Arp2 and Arp3 and subunits ARPC1 to 5 (Figure 3). By itself, Arp2/3 complex has very low nucleation activity (Mullins et al., 1998). Nucleation is activated by NPFs, the best known of which are members of the WASP/WAVE family of proteins (Chesarone and Goode, 2009; Goley and Welch, 2006; Pollard, 2007). These proteins recruit one to three actin subunits and promote a conformational change within the Arp2/3 complex. NPFs are themselves regulated by various factors, in particular Rho-family GTPases. Thus, WASP and N-WASP function under the control of Cdc42, whereas WAVE forms part of a large complex that is regulated by Rac (Bompard and Caron, 2004; Eden et al., 2002; Goley and Welch, 2006; Hall, 2005; Kim et al., 2000; Ma et al., 1998). Classical NPFs such as WASP/WAVE (Goley and Welch, 2006), WASH (Linardopoulou et al., 2007), WHAMM (Campellone et al., 2008) and JMY (Zuchero et al., 2009), present a C-terminal WCA region, which constitutes the shortest polypeptide necessary for activation of nucleation with the Arp2/3 complex (Machesky et al., 1999). This region consists of three distinct segments: W, C and A. W binds the first actin subunit of the new filament (Figure 2). The C (central or connecting) and A (acidic) motifs interact with various subunits of the Arp2/3 complex, helping to stabilize the activated conformation. However, the mechanism by which CA participates in Arp2/3 complex activation, and the specific interactions with subunits of the complex remain a mystery. The actin monomer bound to the W domain, together with Arp2 and Arp3, are thought to form a trimeric seed for the nucleation of a filament branch that emerges at a 70° angle from the side of a preexisting filament (Figure 3). According to this model (Robinson et al., 2001), Arp2 and Arp3 are the first two subunits at the pointed end of the new filament branch, and are expected to adopt a short-pitch filament-like conformation.</p><p>The crystal structure of Arp2/3 complex was first determined in the absence of nucleotide and NPF (Robinson et al., 2001) (Figure 3B). In the structure, Arp2 and Arp3 are separated (i.e. not in a filament-like conformation) and the nucleotide cleft of Arp3 is wide open, whereas subdomains 1 and 2 of Arp2 are disordered. Thus, this structure was described as the inactive conformation of the complex (Robinson et al., 2001). Subsequently, Arp2/3 complex was crystallized in the presence of ATP or nucleotide analogs (Nolen and Pollard, 2007). Nucleotide binding favors closure of the nucleotide cleft of Arp3 and marginally stabilizes subdomains 1 and 2 of Arp2. However, the relative position of Arp2 and Arp3 was unchanged, indicating that, although necessary, ATP binding alone is insufficient to activate Arp2/3 complex. It is believed that the binding of nucleotide and WCA are thermodynamically coupled and that these two factors contribute together to activating Arp2/3 complex (Dayel et al., 2001; Goley et al., 2004; Le Clainche et al., 2001). Pre-existing filaments may help shift the equilibrium in favor of an activated complex (Pollard, 2007). However, Arp2/3 complex can bind to and cap filament pointed ends with high affinity outside the branch (Mullins et al., 1998). Because pointed end binding requires an activated conformation (Boczkowska et al., 2008; Robinson et al., 2001; Rouiller et al., 2008), side binding may not be necessary for activation, although it is probably favored by activation. Considering the high concentration of actin monomers in cells and typical affinities of the W-actin interaction ranging from ~0.05 to ~0.25µM (Chereau et al., 2005; Marchand et al., 2001; Mattila et al., 2003), it is likely that NPFs are actin-loaded prior to encountering the Arp2/3 complex. Thus, actin-loaded NPFs and nucleotide are probably the most important factors shifting the equilibrium in favor of an activated complex in vivo.</p><p>The structure of Arp2/3 complex in the branch and with bound WASP has been studied using electron microscopy (Egile et al., 2005; Rodal et al., 2005; Rouiller et al., 2008). These studies agree in that a major conformational change takes place upon activation, bringing Arp2 and Arp3 into a filament-like arrangement at the pointed end of the branch. Additionally, electron tomography of the branch junction reveals conformational changes in the mother filament at the interface with the Arp2/3 complex and suggests that all seven subunits of the Arp2/3 complex contact the mother filament (Rouiller et al., 2008). None of the existing structures, however, resolves the location and interactions of the CA activator region of NPFs with subunits of the Arp2/3 complex. This question has mainly been addressed by crosslinking and NMR solution studies, showing that CA can be crosslinked to Arp2, ARPC1, Arp3 and ARPC3 (Kelly et al., 2006; Kreishman-Deitrick et al., 2005; Weaver et al., 2002; Zalevsky et al., 2001). Because of the short length of the WCA polypeptide (~73 aa), and considering that both the C (Panchal et al., 2003) and W (Chereau et al., 2005) motifs comprise regions of helical structure, it is difficult to rationalize how CA can span these four subunits in the complex. A recent study attempts a different approach to address this question.</p><p>Actin has a highly reactive cysteine residue at position 374. The structures of W-actin complexes (Figure 2) revealed that the N-terminal portion of the W domain faces directly actin Cys-374 (Chereau et al., 2005). Based on this observation, a Cys residue was introduced by mutagenesis at the N-terminus of WCA, which was then crosslinked to actin Cys-374 (Boczkowska et al., 2008). [In support of this approach, a crystal structure of crosslinked W-actin is now available and is nearly undistinguishable from the uncrosslinked structures (Rebowski et al., in preparation).] Contrary to WCA alone, crosslinked WCA-actin forms a stable high affinity complex with the Arp2/3 complex, while also capping its barbed end so that the nucleus cannot elongate by addition of actin monomers. Importantly, the stoichiometry of this complex determined by various methods is precisely 1:1, and not 2:1 as it may be inferred from a recent study (Padrick et al., 2008). This approach produced a stable WCA-actin-Arp2/3 complex particle, whose structure in solution was analyzed by Small Angle X-ray Scattering (SAXS). The SAXS study indicated that the first actin subunit binds at the barbed end of Arp2, which additionally constrains the binding site of the C motif to subunit Arp2, near the interface with ARPC1 (Figure 3C). Less can be said about the location and interactions of the A region, except that it probably lies near the interface between subunits Arp3 and ARPC3, which is consistent with most of the biochemical evidence (Kelly et al., 2006; Kreishman-Deitrick et al., 2005; Weaver et al., 2002; Zalevsky et al., 2001). This study offers testable hypotheses and a new way to address the problem of activation, but because of its limited resolution it leaves unresolved the exact nature of the conformational change leading to activation and the precise role of WCA in this process.</p><!><p>The W domain often occurs in tandem repeats, which is the most common architecture found among known actin filament nucleators, observed in Spire (Quinlan et al., 2005), Cobl (Ahuja et al., 2007), and VopL/VopF (Liverman et al., 2007; Tam et al., 2007) (Figure 4). The actin monomers bound to the W repeats of these proteins are thought to come together to form an actin filament-like nucleus for polymer assembly. However, the specific nucleation mechanism of each protein appears to be different, as reflected by dramatic differences in their nucleation activities. For instance, Spire with the largest number of W domains (four) has relatively weak nucleation activity (Quinlan et al., 2005), whereas VopL/VopF with just three W domains are even more efficient nucleators than the Arp2/3 complex (our own observation). At least in part, the explanation may lie in the variable linkers between W domains, in particular linker-2 between the second and third W domains. Differences in the linkers may dictate the relative arrangement of actin subunits in the polymerization nucleus, and thereby the nucleation activities of each protein. When the linkers are short, as in Spire, only actin subunits along the long-pitch helix can be connected (Rebowski et al., 2008). However, the brain-specific nucleator Cobl has strong nucleation activity and presents a long, Pro-rich linker-2 (Ahuja et al., 2007). Shortening Cobl's linker-2 reduces dramatically its nucleation activity, whereas replacing this linker with an unrelated sequence of similar length restores most of the endogenous activity. Therefore, the length of the linker, but not necessarily its specific sequence, appears to be crucial for Cobl's activity. Because a longer linker may allow successive W domains to connect actin subunits laterally, it has been proposed that Cobl stabilizes a short-pitch actin nucleus (Ahuja et al., 2007) (Figure 4). However, the exact arrangement of actin subunits in Cobl's nucleus is unknown and two possibilities must be considered: the third actin subunit in Cobl's nucleus could be staggered toward the pointed end with respect to either the first actin subunit (most likely) or the second actin subunit. In any case, the examples of Cobl, the Arp2/3 complex and Lmod (discussed below) suggest that stabilization of a short-pitch actin trimer is a more effective way to promote nucleation than stabilization of a larger nucleus of four actin subunits along the long-pitch helix.</p><p>A recent study, additionally suggests that some inter-W linkers present actin monomer-binding activity, and can as a result boost the nucleation activity of tandem W constructs (Zuchero et al., 2009). Thus, for example, a fragment consisting of the two W domains of N-WASP had no nucleation activity, but a modest increase in nucleation was observed when the naturally occurring inter-W linker was replaced by Spire's linker-3 (Zuchero et al., 2009).</p><p>Microbial pathogens often disrupt or kidnap the host cell cytoskeleton for infection (Bhavsar et al., 2007; Gouin et al., 2005). A well known example is Listeria monocytogenes, whose surface protein ActA mimics eukaryotic NPFs and recruits both the filament elongation factor VASP and the Arp2/3 complex polymerization machineries at the surface of the parasite to propel its movement within and between cells (Cossart and Toledo-Arana, 2008). Vibrios are Gram-negative rod-shaped bacteria, comprising human pathogens that cause wound infections, gastro-intestinal disease and diarrhea, and are often associated with infection from consumption of raw seafood. Vibrio parahaemolyticus and Vibrio cholerae were nearly simultaneously shown to produce the type III secretion system (T3SS) virulence factors VopL (Liverman et al., 2007), and VopF (Tam et al., 2007), respectively. VopL and VopF display ~57% overall sequence identity. Similar proteins are also found among other Vibrio species. VopL/VopF disrupt actin homeostasis, and appear to be required for infection (Liverman et al., 2007; Tam et al., 2007). Both proteins present three W domains and Pro-rich sequences, and like Cobl, both are strong filaments nucleators. It is, therefore, tempting to propose that like Cobl these two proteins stabilize a short-pitch polymerization nucleus. However, linker-2 in VopL/VopF is significantly shorter than in Cobl (Figure 4C), and because the length of the linker is such a critical factor for Cobl's activity (4), the reasons for the strong nucleation activities of VopL/VopF remain a mystery. A potential explanation is given next.</p><!><p>In addition to the inter-W linkers, oligomerization may influence the nucleation activities of tandem W-based filament nucleators. For instance, Spire interacts with the formin Cappuccino (Quinlan et al., 2007; Quinlan and Kerkhoff, 2008; Renault et al., 2008; Rosales-Nieves et al., 2006), and the two proteins appear to synergize to assemble actin filaments both in vitro (Bosch et al., 2007) and in vivo (Rosales-Nieves et al., 2006), where they may be involved in maintaining microtubule organization (Dahlgaard et al., 2007). The interaction, which involves the kinase non-catalytic C-lobe domain (KIND) of Spire (Figure 4A) and the formin homology 2 (FH2) domain of Cappuccino, enhances the nucleation activity of Spire (Quinlan et al., 2007). It is likely that the increased activity results from Spire dimerization mediated by the FH2 dimer. Another possibility for Spire to function as a dimer is through its C-terminal FYVE zinc-finger domain (Figure 4A), which in some proteins has been shown to dimerize (Dumas et al., 2001). A recent report additionally finds that WASH, a WASP-family NPF (Linardopoulou et al., 2007), interacts directly with Spire and synergizes with both Spire and Cappuccino to control actin and microtubule dynamics during Drosophila oogenesis (Liu et al., 2009). WASH also appears to dimerize through its N-terminal WASH homology domain 1 (WHD1) (Liu et al., 2009), providing yet another potential mechanism for Spire dimerization. Whether Spire dimerizes directly through its FYVE zinc-finger domain or indirectly through interaction with Cappuccino or WASH, dimerization is likely a contributing factor in Spire's nucleation activity. Indeed, an optimally assembled Spire dimer could stabilize the formation of a nucleus consisting of eight actin subunits, four on each side of the filament (or long-pitch helix), potentially resulting in a very powerful nucleator.</p><p>Another example is the T3SS protein TARP from Chlamydia trachomatis. Despite having a single W domain, TARP nucleates actin filaments, but this activity depends on the presence of the central Pro-rich domain, which in this protein appears to mediate oligomerization (Jewett et al., 2006).</p><p>The existing relationship between NPFs-Arp2/3 complex and W-based nucleators is discussed below. In this regard, it is interesting to note that a recent study finds that forcing the dimerization of WASP by external factors increases its affinity for the Arp2/3 complex and enhances its nucleation activity (Padrick et al., 2008). It is still unknown whether WASP dimerization plays a role in vivo. However, as pointed out above, the WASP-related protein WASH appears to dimerize by itself. Although dimerization in the case does not seem to enhance Arp2/3 complex-mediated nucleation, it is likely to play a critical role in vivo, notably by mediating the bundling and crosslinking of F-actin and microtubules under the control of the GTPase Rho1 (Liu et al., 2009). Finally, sequence analysis identifies potential oligomerization domains among other nucleators and NPFs (Figure 4A). Oligomerization is thus emerging as an important factor modulating the activities of W-based nucleators, which is analogous to formins (Copeland et al., 2004). Whether oligomerization also contributes to the nucleation activities of Cobl and VopL/VopF remains to be demonstrated.</p><!><p>Structural considerations suggest that NPFs-Arp2/3 complex can be conceptually viewed as a specialized form of tandem W-based nucleator (Boczkowska et al., 2008). According to this view, the actual nucleators are the NPFs, and not the Arp2/3 complex as it has been traditionally described. The distinction is not merely semantic, but rather stems from a different structure-function understanding of how these proteins work. In isolation, neither the Arp2/3 complex nor the NPFs nucleate; they need each other for this activity. There is only one known exception to this rule, which actually reinforces the proposed relationship between NPFs and tandem W-based nucleators. It is the newly discovered NPF protein JMY, which presents three W domains N-terminal to its CA region and, in addition to activating the Arp2/3 complex, has some nucleation activity of its own (Zuchero et al., 2009). More importantly, there is the undisputable fact that the newly discovered filament nucleators (Spire, Cobl, TARP and VopL/VopF) share far more in common with NPFs than they do with the Arp2/3 complex (Figure 4), including the presence of tandem W repeats and Pro-rich regions. The number of bona fide W domains in NPFs varies from 1 to 3, whereas the newly discovered nucleators contain between 1 and 4 W domains. However, as it has been pointed out by various investigators (Aguda et al., 2005; Boczkowska et al., 2008; Chereau et al., 2005; Hertzog et al., 2004), the C motif of NPFs is also related to the W domain, a relationship that can be further extended to the F-actin-binding (FAB) motif of Ena/VASP proteins (Ferron et al., 2007) (Figure 4). Based on the location of the first actin subunit in the SAXS structure of WCA-actin-Arp2/3 complex, it was proposed that the C motif binds Arp2 (Boczkowska et al., 2008). Like the W domain, the N-terminal portion of the C motif consists of an amphiphilic helix (Panchal et al., 2003), which according to this proposal binds in the hydrophobic cleft of Arp2 (Figure 3C), somewhat analogous to the binding of W to actin (Figure 2). As for the Arp2/3 complex itself, it can be thought of as an actin dimer that upon activation adopts a short-pitch conformation. The association of the Arps with five other proteins in the Arp2/3 complex probably emerged from a need to integrate nucleation and branching within a single system. Based on these considerations, NPFs can be described as tandem W-based filament nucleators, whose function is to recruit and realign the Arp2-Arp3 short-pitch heterodimer and one to three actin monomers to form a polymerization nucleus.</p><p>Certain proteins interact with the Arp2/3 complex and modulate its activity, but have only a modest effect (if any) on its nucleation activity. These proteins also have markedly different domain organization compared to classical NPFs such as WASP/WAVE (Goley and Welch, 2006), WHAMM (Campellone et al., 2008), WASH (Linardopoulou et al., 2007), and JMY (Zuchero et al., 2009), and should probably not be considered as fully-fledged NPFs. In a recent review, some of these molecules were grouped into a separate category, identified as class II NPFs (Goley and Welch, 2006). A well-studied example is the protein cortactin (Ammer and Weed, 2008). Cortactin has only a limited effect on the nucleation activity of the Arp2/3 complex, but it plays a critical role by binding to the Arp2/3 complex at branch points, which stabilizes branch junctions and inhibits filament de-branching and network breakdown (Weaver et al., 2001).</p><!><p>The model of filament nucleation by NPFs-Arp2/3 complex proposed here takes into account the purported relationship with tandem W-based filament nucleators. With the determination of the structure of inactive Arp2/3 complex it was proposed that the complex must undergo a major conformational change during activation that would bring Arp2 and Arp3 into a short-pitch filament-like arrangement, with Arp3 staggered toward the pointed end by half a monomer length relative to Arp2 (Robinson et al., 2001). Nucleotide, the WCA region of NPFs and actin are all necessary ingredients of this conformational change (Pollard, 2007). However, the details of the activation mechanism remain a mystery, and one of the most pressing challenges in the field concerns the determination of a high-resolution structure of the activated complex. Part of the challenge is to make adequate guesses about the mechanism of activation so as to formulate strategies toward obtaining a structure of the activated complex. With this in mind, a model is proposed in Figure 5.</p><p>According to this model, the conserved Trp in the A region of NPFs, which contributes the most to the binding affinity of WCA to the Arp2/3 complex (Marchand et al., 2001; Weaver et al., 2002), works as a 'hook', linking actin-loaded NPFs to the Arp2/3 complex. The SAXS study of WCA-actin-Arp2/3 complex suggests that after this initial encounter the first actin subunit binds at the barbed end of Arp2 (Boczkowska et al., 2008). Arp2, which is partially disordered in the inactive structure (Nolen and Pollard, 2007; Robinson et al., 2001), may transition between active/inactive states, but is stabilized in the activated structure by interaction with the C motif and the first actin subunit of the branch (bound to the W domain of NPFs). This model predicts that Arp2 moves mostly alone during activation, with minimal energetic cost, such as to occupy a filament-like position next to Arp3 (Aguda et al., 2005; Boczkowska et al., 2008). This is supported by flexibility of Arp2 in the inactive structure and the fact that it can be moved with minimal steric clashes. A different model had been initially proposed (Robinson et al., 2001), predicting a more dramatic rearrangement of the complex, involving a rotation of Arp2, ARPC1, ARPC4, and ARPC5 relative to Arp3, ARPC2, and ARPC3. Although the latter model cannot be completely ruled out based on the available data, it appears less likely, because it would involve a large structural change and breakage of hydrophobic contacts along a large interface between the two halves of the complex. Yet, it is reasonable to expect that, in addition to movement of Arp2, other changes will occur in the complex during activation. Activation and branching (i.e. binding to the side of pre-existing filaments) may occur nearly simultaneously. Steric hindrance (discussed below) of the W domain with the actin subunits that begin joining the branch after activation (and possibly of the A motif with the mother filament) may help release NPFs after activation.</p><!><p>The actin "thin" filaments in cardiac and striated muscle sarcomeres display regular length and spacing and are uniformly decorated with muscle-specific proteins such as the troponin complex, tropomyosin (TM) and the barbed and pointed end capping proteins CapZ and Tmod, respectively. Toward the center of sarcomeres, the actin filaments overlap with the myosin "thick" filaments, forming a tight hexagonal lattice. The appearance is that of a rigid structure, and it is not surprising that it has been traditionally thought that the actin filaments in sarcomeres are less dynamic than in non-muscle cells. This view is evolving (Gunst and Zhang, 2008; Sanger and Sanger, 2008; Skwarek-Maruszewska et al., 2009; Wang et al., 2005). The sarcomere may undergo constant dynamic remodeling (or repair), and actin filament nucleators may play a critical role in this process.</p><p>Leiomodin (Lmod) is a tropomodulin (Tmod)-related protein expressed almost exclusively in muscle cells. mRNA expression analysis indicates there are three Lmod isoforms (Conley et al., 2001): Lmod1 expressed at low levels in most tissues and at high levels in smooth muscle, Lmod2 expressed exclusively in heart and skeletal muscles and the fetal isoform Lmod3. The first ~340 amino acids of Lmod are ~40% identical to Tmod, a pointed end capping protein in muscles (Fischer and Fowler, 2003; Fowler et al., 2003; Kostyukova et al., 2007). In Tmod, the N-terminal portion is unstructured, except for three helical segments involved in binding TM and actin. Tmod has a second actin-binding site within the C-terminal Leu-rich repeat (LRR) domain (Fowler et al., 2003; Krieger et al., 2002). Lmod shares this domain organization, except for one important difference: only one of the two TM-binding sites of Tmod appears to be conserved in Lmod. More importantly, Lmod has a ~150 amino acid C-terminal extension featuring a third actin-binding site in the form of a W domain. With three actin-binding sites, Lmod could hypothetically recruit three actin monomers to form a trimeric polymerization nucleus, which led to the identification of Lmod as a potential filament nucleator (Chereau et al., 2008). Consistent with this idea, initial characterization of Lmod revealed a powerful nucleator, whose over- or down-expression had dramatic effects on sarcomeric structure and organization (Chereau et al., 2008).</p><p>Compared to other nucleators, Lmod has one distinctive and important property, it directly interacts with TM. TM is a coiled coil dimer that associates end-to-end to form long helical strands that wind symmetrically along the two long-pitch helices of the actin filament (Holmes and Lehman, 2008). At the pointed end of the actin filament in muscle sarcomeres TM interacts with Tmod via two helical segments located within the N-terminal flexible domain of Tmod (Kostyukova et al., 2007). As mentioned above, only one of these helices is conserved in Lmod. Yet, TM not only modulates the nucleation activity of Lmod, but more importantly it appears to determine Lmod's localization to filament pointed ends. Thus, Lmod162–495, lacking the N-terminal flexible domain, retains ~1/3 of the nucleation activity of full-length Lmod in vitro, but displays nuclear localization. A basic patch located within the long (and probably flexible) linker connecting the second and third actin-binding sites of Lmod is a predicted Nuclear Localization Signal (NLS) and may be responsible for the nuclear localization of Lmod162–495. While it is unknown whether trafficking through the nucleus forms part of Lmod's endogenous function, such an activity has been reported for Tmod (Kong and Kedes, 2004).</p><p>Perhaps reflecting its uniqueness as a muscle cell nucleator, Lmod shares little resemblance with other filament nucleators. With the presence of three actin-binding sites, Lmod is predicted to stabilize a trimeric actin seed for nucleation (Figure 6). However, the actual organization of actin subunits in the Lmod nucleus is a mystery. The W domain in Lmod seems to play an auxiliary role, somewhat analogous to its role in NPFs where the W domain contributes an actin subunit to complete a trimeric nucleus with the Arps. However, it is unknown which of the two actin subunits of the short-pitch dimer in the Lmod nucleus is staggered forward. In other words, it is unknown whether the actin subunit bound to the N-terminal flexible domain of Lmod is staggered forward with respect to the one bound to the LRR domain or vise versa. This question also applies to Tmod, whose pointed end arrangement is still unknown. Lmod's linker-2 is also much longer than Cobl's linker-2, conferring significant freedom with respect to the relative positioning of the third actin subunit. Thus, the third actin subunit could be at the barbed end of either the first or the second subunit. Because Lmod contains a single TM-binding site, it can be predicted that in cells the Lmod nucleus is associated with a single TM dimer, but this has not been formally demonstrated. Finally, one of the most intriguing questions about Lmod concerns the interplay with Tmod. The two proteins are clearly related and appear to have similar localization, but despite this similarity Lmod and Tmod probably have well-separated roles.</p><!><p>A convenient way to introduce the Ena/VASP family of filament elongation factors (Drees and Gertler, 2008) is in contrast to formins. Formins have been reviewed extensively (Chesarone and Goode, 2009; Faix and Grosse, 2006; Goode and Eck, 2007; Higgs, 2005; Paul and Pollard, 2009; Pollard, 2007) and are only discussed briefly herein. Formins are the only proteins that do not use the W domain for nucleation or elongation, although there is at least one formin, INF2 (Chhabra and Higgs, 2006), that contains a W domain, but uses it for filament depolymerization and as a diaphanous autoregulatory domain (DAD) (Chhabra et al., 2009). Formins use the dimeric forming-homology 2 (FH2) domain for interaction with actin, and like W-based nucleators, formins contain Pro-rich regions positioned N-terminal to the actin-binding domain (Goode and Eck, 2007; Higgs, 2005; Paul and Pollard, 2009; Pollard, 2007). But probably the most interesting property of formins is that in addition to nucleation they also promote processive barbed end elongation (or depolymerization).</p><p>So, the obvious question to ask is why do W-based nucleators not sustain processive barbed end elongation? The answer appears to be simple; because of steric hindrance of the W domain with intersubunit contacts in the actin filament. Indeed the crystal structure of a long-pitch actin dimer stabilized by a tandem repeat of two W domains has just been determined in our lab (Rebowski et al., in preparation). The structure shows that although the two actins adopt a filament-like arrangement, they are somewhat more separated than in the actin filament. The separation occurs because the second W domain, bound in the hydrophobic cleft of the second actin subunit, interferes with filament-like contacts between the two actins. The implication is that tandem W-based nucleators cannot stay bound to filaments after nucleation, and therefore are unlikely to influence elongation. This is also likely to explain why NPFs are ejected from the Arp2/3 complex once the branch filament begins to grow.</p><p>Part of the binding interface of the W-domain remains exposed in F-actin. So, tandem W domains may weakly (and non-specifically) co-sediment with F-actin and bind to (or cap) filament barbed ends, as discussed in a recent review (Renault et al., 2008). But this also means that minor changes in the sequence of the W domain may give rise to an F-actin-binding domain. This appears to be the solution that nature has found to produce a filament elongation factor that is compatible with the W domain. Indeed, while formins participate in both nucleation and elongation, the elongation function among W-based filament nucleators has been "outsourced" to a dedicated family of proteins, Eva/VASP, which have a similar domain organization and may be evolutionarily related to WASP-family NPFs (see Figure 4 and legend for details). Eva/VASP and WASP/N-WASP both contain an N-terminal EVH1 (or WH1) domain, a central Pro-rich region and W-related sequences. A trace of their relationship can still be found in the acidic region C-terminal to the W-related sequences (albeit in Ena/VASP this region is less acidic than in NPFs and lacks the important tryptophan involved in binding to the Arp2/3 complex). The G-actin-binding (GAB) domain of Ena/VASP is not only related to the W domain of NPFs, but has also been shown to interact with actin in a similar manner (Ferron et al., 2007) (compare Figure 2 and Figure 7). Immediately C-terminal to the GAB domain is the F-actin-binding (FAB) domain, which is also related to the W domain. However, the FAB domain is more closely related to the C region of NPFs (Figure 4). Indeed, the FAB domain has evolved minor differences compared to the W domain, which probably reflect adaptation to bind F-actin in a way compatible with intersubunit contacts in the filament. A similar necessity may have arisen at the interface between Arp2 and the first actin subunit of the branch (Figure 3C), which probably explains the resemblance between the C and FAB domains. Another event in Ena/VASP's adaptation for processive barbed end elongation is tetramerization, which is mediated by the C-terminal coiled-coil domain (Bachmann et al., 1999; Kuhnel et al., 2004). Tetramerization may allow Ena/VASP to work cooperatively, by sequentially allowing each subunit of the tetramer to release and advance during monomer addition to the barbed end while the other subunits remain attached to the growing filament.</p><!><p>As mentioned above, two proteins, profilin and Tβ4, contribute to maintaining a large fraction (~50%) of the cellular actin in the unpolymerized pool. Tβ4 (Figure 7A) is a short 43-aa polypetide related to the W domain (Dominguez, 2007; Paunola et al., 2002), but contains an additional C-terminal helix that binds atop actin subdomains 2 and 4 (Irobi et al., 2004), making it an effective actin monomer sequestering protein (Safer et al., 1990). As a result, Tβ4-actin complexes cannot participate in actin filament nucleation or elongation. Instead, Tβ4 is though to function as an actin buffer, losing actin in competitive equilibrium to profilin, which has higher affinity for actin monomers (Pollard and Borisy, 2003).</p><p>Multiple properties make profilin a crucial player in filament assembly (Paul and Pollard, 2009; Pollard and Borisy, 2003; Witke, 2004). Thus, profilin catalyzes the exchange of ADP for ATP on actin (which replenishes the pool of polymerization-competent ATP-actin), and inhibits nucleation and pointed end elongation, while having almost no effect on steady-state barbed end elongation. But probably the most interesting property of profilin is that it can bind simultaneously to Pro-rich sequences and actin (Figure 7B), and it binds both with higher affinity as a ternary complex than either one separately (Ferron et al., 2007). This probably provides selectivity, so as to avoid non-productive interactions of profilin with Pro-rich regions in cells.</p><p>Indeed, a common feature of actin filament nucleation and elongation proteins is that they typically contain Pro-rich regions (magenta regions in Figure 4A). Pro-rich regions are most commonly positioned immediately N-terminal to the actin binding sites. These regions bind signaling and regulatory proteins, whose interactions are mediated by any of the existing Pro-rich binding modules, including SH3, WW, and EVH1 domains (Zarrinpar et al., 2003). In addition, these regions frequently bind profilin-actin complexes, and could serve to increase the local concentration of actin monomers ready for assembly. As discussed above, W-based nucleators and NPFs are not expected to remain bound to newly formed filaments. So, any gain resulting from the initial recruitment of profilin-actin will be short-lived and probably difficult to detect using classical polymerization assays. The situation is far more interesting with processive barbed end elongation factors. The clearest evidence so far is for formins, whose elongation rates are increased by recruitment of profilin-actin (Kovar et al., 2006). Perhaps not surprisingly the situation remains confusing with Ena/VASP. Various groups have demonstrated that profilin-actin stimulates Ena/VASP-dependent Listeria motility (Auerbuch et al., 2003; Dickinson et al., 2002; Geese et al., 2000; Grenklo et al., 2003; Kang et al., 1997; Machner et al., 2001). But, while a recent study established that Ena/VASP proteins accelerate barbed end elongation in a processive manner, it also concluded that profilin had no effect in this process (Breitsprecher et al., 2008). This contrasts with the results of another group, which found that VASP accelerates filament elongation from a profilin-actin pool (Pasic et al., 2008). Given the presence of various profilin-binding sites in Ena/VASP proteins, and the stimulatory effect that profilin has on formin-mediated assembly (Kovar et al., 2006), this question deserves to be revisited. At least based on their domain architecture, Ena/VASP proteins appear to have evolved all the necessary adaptations to 'process' profilin-actin complexes from the cellular pool onto the barbed ends of elongating filaments. Thus, the ternary structure of profilin-actin with a fragment of Ena/VASP consisting of the last Pro-rich segment and the GAB domain (Figure 7B) suggests that profilin-actin complexes may be delivered directly from the Pro-rich region to the barbed end of the growing filament (Ferron et al., 2007). A similar monomer transfer mechanism may exist in formins, although this remains to be demonstrated.</p><!><p>Despite notable progress, many question remain about filament nucleation and elongation. New Arp2/3 complex NPFs are being continuously discovered and each NPF redirects the Arp2/3 complex polymerization machinery into a different subcellular location and function. The exact function of Cobl and its role in neuronal morphology is still poorly understood and much remains to be learned about the Spire-Cappuccino-WASH connection. Concerning Lmod, an important question is the interplay with Tmod; which filaments are associated with Tmod in muscle cells and which are associated with Lmod? An issue that is catching attention is the crosstalk between different nucleation and elongation factors in the assembly of different actin cytoskeletal structures (Chesarone and Goode, 2009; Schirenbeck et al., 2005). However, more directly connected with this review is the substantial gap of understanding of the structural aspects of filament nucleation/elongation, including the lack of a structure of activated Arp2/3 complex, and complexes of tandem W-based nucleators and Lmod with actin. Some of these questions are likely to dominate the research within the actin cytoskeleton community during the following few years.</p>

PubMed Author Manuscript

GOLGI IN COPPER HOMEOSTASIS: A VIEW FROM THE MEMBRANE TRAFFICKING FIELD

Copper is essential for a variety of important biological processes as a cofactor and regulator of many enzymes. Incorporation of copper into the secreted and plasma membrane-targeted cuproenzymes takes place in Golgi, a compartment central for normal copper homeostasis. The Golgi complex harbors copper-transporting ATPases, ATP7A and ATP7B, that transfer copper from the cytosol into Golgi lumen for incorporation into copper-dependent enzymes. The Golgi complex also sends these ATPases to appropriate post-Golgi destinations to ensure correct Cu fluxes in the body and to avoid potentially toxic copper accumulation. Mutations in ATP7A or ATP7B or in the proteins that regulate their trafficking affect their exit from Golgi or subsequent retrieval to this organelle. This, in turn, disrupts the homeostatic Cu balance, resulting in copper deficiency (Menkes disease) or copper overload (Wilson disease). Research over the last decade has yielded significant insights into the enzymatic properties and cell biology of the copper-ATPases. However, the mechanisms through which the Golgi regulates trafficking of ATP7A/7B and, therefore, maintain Cu homeostasis remain unclear. This review summarizes current data on the role of the Golgi in Cu metabolism and outlines questions and challenges that should be addressed to understand ATP7A and ATP7B trafficking mechanisms in health and disease.

golgi_in_copper_homeostasis:_a_view_from_the_membrane_trafficking_field

Introduction<!>Localization and function of Cu ATPases in the Golgi<!>Trafficking of Cu-ATPase from the Golgi<!>Retrieval of ATP7A and ATP7B to the Golgi<!>Concluding remarks

<p>Golgi complex operates as a central compartment in the biosynthetic pathway: it receives newly synthetized proteins from the endoplasmic reticulum (ER), provides machinery for their post-translational modifications (e.g. glycosylation, proteolytic cleavage) and sorts modified proteins towards appropriate post-Golgi destinations (Polishchuk and Mironov 2004; Wilson et al. 2011). Over the years novel functions of the Golgi have gradually emerged. This organelle appears to serve as a hub for initiation of several signaling pathways as well as an important compartment involved in cell fate decision and lipid metabolism (Wilson et al. 2011). Studies over the last decade also revealed that the Golgi is a central station for cellular copper (Cu) homeostasis (La Fontaine and Mercer 2007; Lutsenko et al. 2007).</p><p>Cu is an essential but toxic metal. Its ability to undergo reversible oxidation from Cu(I) to Cu(II) under physiologic conditions is utilized by enzymes for critical biochemical processes that include cellular respiration, free radical detoxification, pigmentation, neuropeptide processing, cross-linking of collagen and elastin, and iron transport (Lutsenko 2010; Nevitt et al. 2012). However, oxidation of Cu(I) produces reactive oxygen species and unchelated Cu can be toxic to cells. Therefore, cells evolved a complex network of regulatory mechanisms that operate to satisfy the metabolic demand for Cu and to control Cu levels both at the cellular and systemic levels. High affinity transporter CTR1 imports Cu from extracellular space into the cytosol, where the metal is captured by cytosolic copper chaperones and shuttled towards different intracellular destinations (Fig. 1). The copper chaperone ATOX1 carries copper to the trans-Golgi network (TGN), where it is transferred to ATP7A/7B, which in turn load Cu into newly synthesized cuproenzymes that move through the secretory pathway (La Fontaine and Mercer 2007; Lutsenko et al. 2007).</p><p>ATP7A and 7B are complex multispan membrane proteins that belong to a P1B type ATPase family. The N-terminal portion of both proteins contains 6 metal binding domains followed by 8 transmembrane domains that form a copper translocation pathway to move Cu from the cytosol at the expense of ATP hydrolysis. Autophosphatase, nucleotide binding and phosphorylation domains in 2nd and 3rd cytosolic loops coordinate the catalytic activity of ATP7A/7B. Apart from their biosynthetic function in the TGN, both Cu-ATPases exhibit a unique property to traffic out of the Golgi in response to the increasing cytosolic Cu to post-Golgi structures and then to a the plasma membrane. This regulated trafficking serves multiple functions. In intestinal cells, it allows copper export from the enterocytes into the bloodstream for further distribution to other tissues (La Fontaine and Mercer 2007; Lutsenko et al. 2007). In hepatocytes, movement of ATP7B to post-Golgi Cu excretion sites serves to remove excessive Cu, which is toxic for the cell due to its high redox potential (La Fontaine and Mercer 2007; Lutsenko et al. 2007). In melanocytes, trafficking to melanosome is necessary to maintain the activity of tyrosinase in this specialized compartment (Setty et al. 2008). The inability of Cu-ATPases to traffic in response to changing Cu levels and/or to pump Cu across the membrane results in severe aberrations of Cu metabolism, which are especially apparent in Menkes and Wilson disease caused by mutations in ATP7A and ATP7B genes, respectively (de Bie et al. 2007; La Fontaine and Mercer 2007; Lutsenko et al. 2007).</p><p>Despite the importance of ATP7A/B trafficking in the maintenance of Cu homeostasis, the mechanisms of Cu-ATPase targeting to the Golgi and their transport from this organelle remain poorly understood. In this review, we focus on outstanding mechanistic questions related to the interplay between Cu ATPases localization and function. Answering these questions, in our view, has the potential to open new avenues in ATP7A/B trafficking studies.</p><!><p>The main function of ATP7A/B in the Golgi is to supply Cu to newly synthetized cuproenzymes that move through the secretory pathway. ATP7A/B receive Cu from the chaperone ATOX1 through direct interactions (Lutsenko et al. 2007; Lutsenko 2010; Nevitt et al. 2012). Then the Cu pumps transport the metal across the membrane into the Golgi lumen, where Cu is loaded on a number of enzymes with important functions in the central nervous system (dopamine β-hydroxylase, peptidylglycine α-amidating monooxygenase), connective tissue and blood vessel development (lysyl oxidase, superoxide-dismutase 3), pigmentation (tyrosinase) as well as in iron and Cu transport (ceruloplasmin, hephaestin). The lack of biosynthetic ATP7A/B function in the Golgi leads to a number of severe symptoms that are manifested in Menkes or Wilson diseases (Table 1) (de Bie et al. 2007; Tumer and Moller 2010).</p><p>Within the Golgi stack Cu-ATPases reside mainly in the TGN compartment, as revealed by co-localization studies with different markers (Cobbold et al. 2002; La Fontaine et al. 1998; La Fontaine et al. 2001) and immuno-EM labeling (La Fontaine et al. 1998; La Fontaine et al. 2001, Hasan, 2012 #391). Why ATP7A/B prefer the TGN to the earlier Golgi compartments is not entirely clear. It is possible that the lower pH environment in the TGN lumen facilitates the release of Cu from ATP7A/B to the cuproenzymes (Safaei et al. 2008). On the other hand, the TGN operates as the Golgi exit site where the cargo proteins undergo sorting and packaging into the post-Golgi transport carriers (De Matteis and Luini 2008; Luini et al. 2008; Polishchuk and Mironov 2004). Thus, in the case of intracellular Cu increase, TGN localization allows ATP7A/B to leave the Golgi rapidly and without the need to cross the entire Golgi stack in the cis-to-trans direction.</p><p>Several studies have been done to identify signals within the protein structure that retain Cu pumps in the Golgi. A 38 amino acid sequence containing the transmembrane domain 3 appears to be sufficient to support ATP7A localization in the Golgi complex (Francis et al. 1998). It is clear, though, that several other determinants contribute to TGN compartmentalization of ATP7A/B at basal Cu conditions. The di-leucine and tri-leucine endocytic motifs in the carboxyl-tails of ATP7A and ATP7B, respectively, are required for their efficient retrieval to the Golgi from the peripheral Cu excretion sites (Cater et al. 2006; Francis et al. 1999; Petris et al. 1998).</p><p>Given that copper stimulates Cu-ATPase activity and also triggers the redistribution of ATP7A/7B from the TGN to vesicles, it is interesting to consider the role of copper transport in TGN targeting and retention. The catalytically inactive ATP7B-D1027A mutant (with a replacement of invariant aspartate in the conserved phosphorylation DKTG motif; see Fig. 2) remains within the TGN even at elevated Cu levels, similar to the E1064A Wilson's disease causing mutation that prevents exit from Golgi (Dmitriev et al. 2011). In contrast, studies of the disease-associated ATP7A isoforms identified mutations in the phosphatase A-domain that disrupted retention of the pump in the Golgi and induced its redistribution to the PM (Petris et al. 2002). These observations led to the suggestion that phosphorylation/dephosphorylation of aspartate during the catalytic cycle is directly coupled to trafficking events. Given a very different time scale for the enzyme turnover (hunderds per second [Lewis et al. 2012]) and the TGN exit (which is usually complete within hours), the direct causative link seems unlikely. However, the conformations that copper-ATPases adopt in the absence of copper (and hence low transport activity) and in a copper-saturated state could be markedly different and could be easily distinguished by cellular retention and trafficking machineries.</p><p>The idea that protein conformations of Cu-ATPases contribute significantly to ATP7A/7B targeting and trafficking received further support in recent studies analyzing the role of kinase-mediated (non-catalytic) phosphorylation in the Golgi targeting of ATP7A/B. Serine cluster at positions 340–341 favors Golgi localization of ATP7B as substitution at these positions to alanine (or any other residue) induces ATP7B redistribution to peripheral vesicles (Hasan et al. 2012). In contrast, the mutations of S1469 within the C-terminal part of ATP7A and S478/481/1121/1453 in ATP7B have been reported to keep the protein in the Golgi (Pilankatta et al. 2011; Veldhuis et al. 2009). The important role of the precise inter-domain contacts is also evident from studies on the N-terminal region of ATP7B, where mutation of a single residue Y44A not only diminishes the retention of ATP7B in the TGN, but also causes mis-sorting of ATP7B to a basolateral membrane (the normal destination of ATP7B in polarized cells is apical membrane) (Braiterman et al. 2009).</p><p>The above studies have raised questions about the specific role of a kinase-mediated phosphorylation in ATP7A/7B compartmentalization. Earlier reports have shown that both Cu-ATPases have a basal level of phosphorylation and became hyperphosphorylated in response to copper elevation (Vanderwerf et al. 2001). This phosphorylation is distinct from the catalytic phosphorylation of aspartate and involves serine residues. Variation of phosphorylation status may change conformation/inter-domain interactions in Cu ATPases and therefore enable or prevent their interaction with the components of the post-Golgi trafficking machinery. Unfortunately, the kinases that execute phosphorylation at the traffic-relevant sites of either ATP7a or ATP7B are yet to be identified. The TGN associated PKD1 represents an attractive candidate, as its activity is required for ATP7B export from the Golgi (Pilankatta et al. 2011). On the other hand, PKD inhibitors do not impact trafficking of ATP7A (Cobbold et al. 2002).</p><p>Finally, to reach the Golgi destination, newly synthetized ATP7A/B proteins must be properly folded. Investigation of Wilson disease-causing ATP7B mutants (including the most clinically frequent H1069Q and R778L) revealed their strong retention in the ER (Forbes and Cox 2000; Payne et al. 1998). Despite residual Cu-translocating activity, retention and degradation of these mutants in the ER does not allow them to load Cu on the enzymes and to remove excess metal from the cell. Thus, identification of the molecules that improve the ER-to-Golgi export of these ATP7B mutants represents an important task as its application could benefit a large group of Wilson disease patients.</p><!><p>Apparently the ability to exit from the Golgi towards Cu excretion sites was developed by ATP7A and ATP7B during the evolution of high eukaryotes. Yeast regulate their intracellular Cu levels mainly through the metal-responsive expression of genes operating in Cu influx, efflux, and storage rather than through relocation of the proteins between different compartments (Nevitt et al. 2012). Indeed, the yeast homolog of ATP7A/B, Ccc2, never moves away from the Golgi where it performs, almost exclusively, biosynthetic functions (Yuan et al. 1997). Compared to mammalian orthologues, Ccc2 is a smaller protein: it contains only 2 N-terminal metal binding domains instead of 6 and lacks targeting signals for the apical or basolateral delivery in polarized cells. This observation suggests that domains, missing in yeast, may determine the ability of ATP7A/B to traffic in response to Cu.</p><p>An important challenge faced by mammals in maintaining whole body Cu homeostasis is associated with the development of Cu sensing and Cu transport mechanisms beyond the level of individual cells. At the organismal level, these mechanisms have to accommodate not only multiple cell types in various organs with different metabolic demands for Cu, but also the special and temporal separation of major Cu consumption sites. The post-Golgi trafficking of Cu ATPases to distinct apical/basolateral membranes in polarized cells and the differential expression of Cu ATPases in tissues and organs allow higher vertebrates to deal with systemic changes in Cu levels (Lutsenko 2010; Nevitt et al. 2012).</p><p>ATP7A is expressed in most tissues (except adult hepatocytes), but its most important function is in the small intestine and the choroid plexus. In enterocytes, ATP7A receives adsorbed dietary Cu and moves from the Golgi to the basolateral surface of the cells to release Cu towards portal circulation (Fig. 3). Mutations, which result in a loss of the ATP7A protein, its transport activity, or a failure to traffic out of the Golgi, do not allow Cu to move beyond the intestinal barrier. As a result, severe Cu deficiency is observed in most tissues of Menkes disease patients, with a notable exception of intestine and kidneys (de Bie et al. 2007; Lutsenko et al. 2007; Lutsenko 2010). In contrast to ATP7A, ATP7B is highly expressed in the liver and present at lower levels in many other organs. In the liver, ATP7B receives Cu from the portal circulation and utilizes it in the Golgi for metallation of ceruloplasmin, which uses copper for regulation of iron balance. Cu elevation beyond a certain threshold activates ATP7B export from the Golgi to the "vesicular compartment" and biliary surface in the apical part of hepatocytes (Fig. 3). There, ATP7B supports the efflux of excess Cu into the biliary flow for further elimination of the metal from the body. In Wilson disease, the lack of ATP7B activity and/or delivery of the pump to the apical domain of hepatocytes induces a marked accumulation of Cu in hepatocytes, development of morphologic and metabolic abnormalities, culminating in liver failure (de Bie et al. 2007; Lutsenko et al. 2007; Lutsenko 2010).</p><p>Despite the fundamental importance of ATP7A/B post-Golgi trafficking for the regulation of Cu balance in the body, many mechanistic questions are yet to be answered in full detail. These include the molecular basis of sorting within Golgi, the characteristics of post-Golgi routes for ATP7A or ATP7B, and the machinery involved in the anterograde, retrograde trafficking and recycling.</p><p>Several transport pathways, which are directed to the cell surface and/or endo-lysosomal system, originate from the TGN. The complexity of post-Golgi trafficking increases in polarized cells, as the cell surface-destined proteins have to be delivered to distinct domains (apical or basolateral) of cell membranes (De Matteis and Luini 2008; Muth and Caplan 2003; Rodriguez-Boulan et al. 2005). Earlier studies examined whether ATP7A reaches the PM via the post-Golgi pathway utilized by the constitutively secreted proteins. Insensitivity of ATP7A exit from the Golgi to inhibitors of constitutive secretion, such as a dominant negative PKD mutant, led to the suggestion that ATP7A is sorted from the TGN into the specific route regulated by Cu (Cobbold et al. 2002). On the other hand, ATP7A trafficking was suppressed by Cdc42 and PKA inhibitors, which were equally effective against constitutive cargo proteins (Cobbold et al. 2002). Thus, to firmly identify the ATP7A export route from the Golgi, the approach using molecular inhibitors should be complemented with other methods (see below).</p><p>While some aspects of ATP7A post-Golgi trafficking are understood, the mechanisms of ATP7B export from the Golgi remain poorly studied. With the exception of the role of PKD in ATP7B trafficking from the Golgi (Pilankatta et al. 2011), there are no studies identifying molecular players that support ATP7B export from the Golgi. Thus, the existing body of evidence does not allow us to assign ATP7B to any well-studied pathway or identify its itinerary as a newly-identified post-Golgi route.</p><p>In this context, it will be important to understand whether ATP7A or ATP7B can be packed into the post-Golgi carriers containing cargoes moving through the constitutive pathway (such as VSVG, CD8, GPI-anchored proteins, Na/K-ATPase, etc.) (De Matteis and Luini 2008; Polishchuk and Mironov 2004). The biogenesis and morphology of such carriers has been extensively characterized (Polishchuk et al. 2003; Polishchuk et al. 2000) as well as the methodologies that allow efficient analysis of their composition (Polishchuk et al. 2004). If ATP7A or ATP7B are found within the post-Golgi structures carrying known constitutive cargo proteins, then the role of Cu in inducing the TGN exit would be to allow the Cu ATPases to adopt the conformation necessary for incorporation into well-known constitutive TGN-to-PM route(s). The lack of co-localization with the conventional markers of the post-Golgi pathways, on the other hand, would indicate that ATP7A or ATP7B are transported along a specific route regulated by Cu. If the latter scenario turns out to be the case, it will be important to understand (i) how Cu mechanistically triggers the formation of Cu-ATPase containing carriers from the TGN membranes and (ii) whether any other cargo protein(s) utilize this ATP7A/B-specific post-Golgi pathway.</p><p>It is worth noting that, although both ATP7A and ATP7B are targeted to the TGN, a closer examination of their localization reveals significant segregation from conventional TGN markers, such as TGN38, TGN46 or Golgin 97 (Guo et al. 2005; Holloway et al. 2007; Nyasae et al. 2007). Therefore, the TGN regions containing Cu ATPases could constitute specific TGN sub-compartments from where ATP7A or ATP7B exit towards distinct post-Golgi destinations (Guo et al. 2005; Holloway et al. 2007; Nyasae et al. 2007). A recent study revealed Arf1 to be involved in both the maintenance and biogenesis of ATP7A containing TGN membranes (Holloway et al. 2007). This finding may have an interesting implication, as Arf1 recruits clathrin adaptor proteins (mainly AP1, AP3, AP4 and GGAs) to the TGN, where these adaptors drive both sorting and trafficking events (Robinson and Bonifacino 2001). ATP7A and ATP7B possess, respectively, di-leucine and tri-leucine C-terminal motifs that can be recognized by adaptor proteins. However, recent studies have reported that the suppression of AP1, clathrin or GGA, while causing intracellular ATP7A redistribution, does not block ATP7A or ATP7B exit from the Golgi (Hirst et al. 2012; Holloway et al. 2013; Martinelli et al. 2013). The role of AP3 and AP4 in the TGN exit of ATP7A/7B as well as the biogenesis of the TGN subcompartment, where Cu ATPases reside, remains to be determined.</p><p>The other important issue that has to be addressed is whether ATP7A and ATP7B occupy the same TGN sub-compartment or are sorted within the TGN into independent domains. Although ATP7A and ATP7B localization was studied in cell types where both proteins were expressed (La Fontaine and Mercer 2007), a detailed comparison of their localizations in the Golgi has not been done. Generation of high resolution maps of ATP7A and ATP7B distribution over the TGN membranes (using advanced light microscopy and immuno-EM) would allow new insights into the mechanisms of Cu-ATPase trafficking and sorting at the TGN level.</p><p>Polarized cells utilize ATP7A or ATP7B function at the basolateral or apical surface, respectively. The need to deliver copper transporters to these distinct domains constitutes additional complexity in the trafficking mechanisms. Site-directed mutagenesis revealed the requirement for a dileucine (1487LL1488) motif and the PDZ target (1497DTAL1500) domain in the C-terminus of ATP7A for localization at the basolateral membrane (Greenough et al. 2004). In contrast, the apical targeting of ATP7B has been shown to rely on a novel N-terminal (37FAFDNVGYE45) sequence that is absent from the corresponding region of ATP7A (Braiterman et al. 2009). However, the mechanistic details of how these signals are utilized by the trafficking machinery and where the sorting of ATP7A and ATP7B is executed remain unclear.</p><p>In general, the TGN operates as a main sorting station along the secretory pathway from where cargo proteins are delivered to their target compartments and surface domains (De Matteis and Luini 2008; Polishchuk and Mironov 2004; Rodriguez-Boulan et al. 2005). Several apical and several basolateral routes emerge from the TGN and their number may vary significantly in different cell types or even during polarization of the same cell (Muth and Caplan 2003; Rodriguez-Boulan et al. 2005). In addition to the TGN, several endocytic compartments are thought to perform sorting functions along the secretory pathway (Mellman and Nelson 2008; Rodriguez-Boulan et al. 2005). Specific apical or basolateral cargo proteins, which follow either direct TGN-to-PM or "through-endosome" exocytic routes, were identified as well as selective molecular tools that allow the interception of trafficking along the individual routes (Mellman and Nelson 2008; Rodriguez-Boulan et al. 2005). The analysis of ATP7A or ATP7B colocalization with such cargo markers in the post-Golgi carriers and the treatments with specific molecular inhibitors should allow us to determine (i) which pathway is utilized by Cu-ATPase to get from the TGN to the correct surface domain and (ii) whether this pathway involves an endocytic intermediate.</p><p>Among the sorting endocytic stations, the so called "AP1-B recycling compartment" plays an extensive role in basolateral targeting of several membrane proteins such as VSVG, transferrin receptor and LDL receptor (Folsch 2008; Gonzalez and Rodriguez-Boulan 2009) and could be involved in the basolateral sorting of ATP7A. Clathrin adaptor complex AP-1B (containing epithelial-specific mu1b subunit) drives sorting and transport events in this compartment and, therefore, defines its identity (Folsch 2008; Gonzalez and Rodriguez-Boulan 2009). Interestingly, high expression levels of mu1b subunit of AP-1B were found in a number of tissues (kidney, intestine, placenta, mammary gland) (Ohno et al. 1999) where both ATP7A and ATP7B were detected (La Fontaine and Mercer 2007; Lutsenko et al. 2007). In these tissues Cu pumps undergo delivery to the opposite membrane domains where they execute specific functions (La Fontaine and Mercer 2007). However, whether AP-1B is required for correct delivery of either ATP7A or ATP7B remains unclear. ATP7B trafficking and sorting is unlikely to involve AP-1B because hepatocytes do not express this adaptor (Ohno et al. 1999), yet target ATP7B to the apical (canalicular) surface (Guo et al. 2005; Roelofsen et al. 2000). Whether ATP7A trafficking requires AP-1B activity also remains to be tested. Some membrane proteins with similar milti-span topology (like Na/K ATPase) bypass the AP-1B compartment on their route from the Golgi to the basolateral surface (Farr et al. 2009). On the other hand, expression of inactive Rab22 arrests ATP7A within the post-Golgi recycling station (Holloway et al. 2013). Rab22 is known to cooperate in the regulation of recycling with Arf6 (Weigert et al. 2004), which in turn recruits AP-1B to the membranes (Shteyn et al. 2011). Therefore, the engagement of post-Golgi recycling AP-1B station in ATP7A basolateral delivery cannot be ruled out but has to be verified using ablation of mu1b in polarized kidney or intestinal cells.</p><p>In response to copper elevation, Cu-ATPase ATP7B moves from the TGN to large cytosolic vesicles (Cater et al. 2006; Roelofsen et al. 2000). This observation suggests that the endocytic intermediate is almost certainly involved in the post-Golgi trafficking of ATP7B towards apical surface of hepatic cells. Although the ATP7B containing vesicles remind endosomes, the earlier immuno-EM microscopy did not detect a significant overlap of these ATP7B-containing structures with the conventional endo-lysosomal markers (La Fontaine et al. 2001). It has been proposed (although not yet shown) that ATP7B pumps Cu inside these vesicles and that vesicles fuse with the canalicular surface of hepatocytes to expel Cu from their lumen directly into the bile (Cater et al. 2006). Whether ATP7B is delivered to the apical canalicular membrane of hepatocytes during vesicle exocytosis remains an issue of ongoing debate (Cater et al. 2006; Roelofsen et al. 2000). Several studies fail to detect ATP7B at the canalicular membrane of polarized hepatic cells (Cater et al. 2006), while others provided compelling evidence that ATP7B reaches the apical surface of hepatocytes (Guo et al. 2005; Roelofsen et al. 2000). Determining whether ATP7B is present at the apical membrane (either transiently during vesicle fusion or for a longer time to mediate copper export) is essential in order to understand copper homeostasis in the liver.</p><p>From the technical point of view, it is difficult to investigate a coupling between ATP7B trafficking and Cu excretion and to determine whether these events are coordinated. The lack of comprehensive data on the molecular composition and, therefore, identity of the ATP7B-containing vesicles does not allow to judge, which molecular players may be involved in the delivery of ATP7B from the Golgi to vesicles and then from vesicles to the plasma membrane. The isolation of ATP7B vesicles combined with the characterization of their proteome would circumvent this obstacle as it would provide testable targets to investigate the ATP7B trafficking to and from the vesicles.</p><p>Several features of ATP7B vesicles are similar to secretory granules, or specific lysosome-related organelles that release their content in response to stimuli (Raposo et al. 2007). Like the above organelles, ATP7B vesicles are employed in storage (as they accumulate Cu) and undergo exocytosis upon specific stimulus (increase in Cu concentration) (Cater et al. 2006). Given these behavioral similarities, it would be interesting to understand (i) whether ATP7B vesicles share some components of the molecular machinery with the exocytic storage organelles and (ii) what is the specific signaling mechanism that links changes in intracellular copper with vesicle exocytosis. Interestingly, ATP7A also undergoes redistribution from the TGN to the specific post-Golgi vesicular structures upon exposure of enterocytes to Cu (Nyasae et al. 2007). Therefore, the existence of a Cu-sensitive post-Golgi storage station may represent a common feature in trafficking of both ATP7A and ATP7B. This vesicular pool may provide additional sorting of ATP7A/7B towards the plasma membrane (when copper is elevated), back to the TGN (when copper becomes depleted) and/or forward Cu-ATPases for lysosomal degradation at the end of their life span. Differential phosphorylation by kinases (reported for both ATP7A and ATP7B) may play a key role in such sorting along with specialized adaptor proteins such as COMMD1. The existence of a specialized vesicular compartment may be especially useful in neurons where a robust response at the synaptic cleft may require rapid vesicular fusion rather than slow trafficking of ATP7A/7B from the TGN. Participation of the ATP7A-containing vesicles in rapid fusion (independent of trafficking from the TGN) is seen in response to activation of the NMDA receptor (Schlief et al. 2005).</p><p>Despite numerous gaps in our understanding of the mechanisms responsible for ATP7A and ATP7B trafficking, the regulatory role of Cu in these processes has been firmly established by many studies (Hung et al. 1997; Petris et al. 1996). It is thought that Cu stabilizes ATP7A and ATP7B in a conformation favorable for interaction with the components of membrane trafficking machinery; this allows for the export of Cu-ATPases from the Golgi and their delivery towards post-Golgi destinations. Several Cu-dependent binding partners of ATP7A and 7B were predicted using yeast two hybrid screen (La Fontaine and Mercer 2007); none of them belongs to conventional traffic machineries and so far their role in mammalian cells has not been explored in detail. One potentially interesting candidate is p62 subunit of dynactin-dynein microtubule motor complex. p62 interacts with ATP7B in the presence of high Cu (Lim et al. 2006) and, therefore, being in complex with dynein motor, can potentially pull ATP7B-enriched membranes along the MTs away from the bulk of the TGN. Whether and how other components of membrane budding/fission machinery can be triggered by changes in Cu levels remains unclear. One unexplored possibility is that Cu-induced structural changes in ATP7A/7B open binding sites for lipids, such as cholesterol or sphingomyelin, that may influence Cu-ATPase sorting within the TGN subdomains or initiate assembly of trafficking machinery. The ability of COMMD1 protein (a protein with a known role in hepatic copper balance) to specifically detect phosphotidyl inositols (PIPs) in vitro has been experimentally demonstrated (Burkhead et al. 2009), but the role of PIPs in ATP7A trafficking from Golgi and along the secretory pathway remains unexplored.</p><p>Recent bioinformatics analyses suggest that about 1% of the entire eukaryotic proteome is composed of putative Cu binding proteins (Andreini et al. 2008), suggesting that the list of Cu-dependent regulators of ATP7A and B trafficking is likely to be expanded. To this end, generation of ATP7A and ATP7B interactomes in low and high Cu would provide a valuable tool for the identification of new molecules involved in Cu-ATPase trafficking.</p><!><p>After elimination of excess Cu at the cell surface, ATP7A and ATP7B return to the Golgi where they switch their activities towards metallation of the newly synthetized proteins. It has been convincingly demonstrated that the C-terminal di-leucine or tri-leucine signatures in ATP7A and ATP7B, respectively, are essential for the retrieval of proteins back to the Golgi (Braiterman et al. 2011; Francis et al. 1999; Petris et al. 1998). The ability of the leucine-based motifs to interact with clathrin adaptors (Robinson and Bonifacino 2001) led to the hypothesis that ATP7A and ATP7B undergo internalization through the clathrin-dependent pathway (Francis et al. 1999; Petris et al. 1998). However, the initial attempt to test this hypothesis experimentally led to the opposite conclusion (Cobbold et al. 2003) and only recently calthrin down-regulation with RNAi indicated that ATP7A endocytosed from the cell surface via the clathrin-mediated pathway (Holloway et al. 2013). The importance of the di-leucine motif for recognition by endocytic machinery was recently confirmed in cells from patients with a new disorder of Cu metabolism, ATP7A-related distal motor neuropathy (Yi et al. 2012). The disease-causing mutation in ATP7A, P1386S, is located in the vicinity of the di-leucine motif and leads to a shift in steady-state ATP7A localization from the TGN to the cell surface.</p><p>Reduced retrieval of ATP7A and ATP7B to the Golgi was also observed when AP-1 function was suppressed (Hirst et al. 2012; Holloway et al. 2013). In a recent, very elegant study, Hirst and colleagues demonstrated that the inactivation of AP-1 components results in a depletion of both ATP7A and ATP7B from the clathrin coated vesicles, which are likely directed from the endocytic compartment(s) to the Golgi (Hirst et al. 2012). The role of AP-1 in the endocytic trafficking of Cu ATPases to the Golgi was further confirmed by the study of mechanisms involved in the pathogenesis of MEDNIK syndrome, a congenital disorder with alterations of Cu homeostasis (Martinelli et al. 2013). This disease is caused by mutations in the AP1S1 gene encoding σ1A, the small subunit of AP-1. Fibroblasts from MEDNIK patients exhibit ATP7A mostly at the cell surface, even in the presence of Cu chelator, therefore indicating that the mutation severely compromised the retrieval of the Cu pump to the Golgi (Martinelli et al. 2013). The reduced amount of ATP7A in the Golgi could impair metallation of several Cu-dependent enzymes (see Table 1) and produce neurologic, metabolic, pigmentation and skin symptoms observed in MEDNIK patients and in a model system, such as Ap1s1 zebra fish (Martinelli et al. 2013; Montpetit et al. 2008).</p><p>Surprisingly, mutation of di- or tri-leucine motifs, as well as the suppression of clathrin/AP-1 functions, does not impact the export of ATP7A/B from the Golgi (Hirst et al. 2012; Holloway et al. 2013). This indicates that Cu ATPases take a clathrin-independent exit route(s) from the TGN. On the other hand, the above studies provide new insights into the nature of the pathway that carries Cu-ATPases back to Golgi. Apart from ATP7A and ATP7B, AP-1 suppression eliminates from clahtrin-coated vesicles several well-studied proteins, such as M6PR, Sortilin-1 and furin (Hirst et al. 2012), which recycle from the endosomes to the Golgi (Bonifacino and Rojas 2006). Therefore, it is likely that Cu ATPases get delivered to the Golgi from sorting endosomes through the pathway utilized by M6PR, Sortilin-1 and furin.</p><p>Overall, Cu-ATPases exhibit interesting and distinct trafficking properties. Their export from the Golgi to the sites of Cu excretion seems to require a specific exocytic route, tightly regulated by Cu, whereas their retrieval back to the Golgi seems to proceed through a more common endocytic pathway.</p><!><p>The well-known role of ATP7A and ATP7B in the maintenance of Cu homeostasis has been recently expanded to their involvement in other processes, such as modulation of the Alzheimer disease phenotype, pathogen defense, and anti-cancer drug resistance (Gupta and Lutsenko 2009; Wang et al. 2011). The list of these new functions continues to grow. Therefore, the Golgi complex, which operates as a hub for Cu-ATPases, will remain an important focus of cell biology research on Cu metabolism. In this review we pointed to the main challenges and questions related to Cu-ATPase trafficking and hope that the need to answer these many intriguing questions will gain attention from both the copper and Golgi communities.</p><p>We believe that the advent of modern system biology approaches could help achieve real breakthroughs in the understanding of ATP7A and ATP7B trafficking pathways and mechanisms. Next generation sequencing is likely to reveal new ATP7A and ATP7B mutants/variants as well as new genes associated with cell responses to Cu toxicity and deficiency (Fuchs et al. 2012); studies of new regulators will determine their impact on the localization and trafficking of Cu-ATPases and expand the network of proteins involved in the regulation of copper metabolism. Similarly, proteomics/mass-spectrometry approaches may help in the discovery of new ATP7A and ATP7B binding partners that regulate trafficking of Cu-ATPases to and from the Golgi, as well as establish the role of posttranslational modifications in Cu-ATPase targeting and sorting. Combining this approach with immunoisolation of ATP7A and ATP7B-containing membranes would help characterize the post-Golgi compartments along the exocytic route(s) of both proteins. Finally, high content microscopy screening of the siRNA or chemical libraries is expected to further expand or confirm a list of new molecular players that regulate ATP7A and/or ATP7B localization and activity. Taken together such studies will provide new insights into the role of the Golgi in Cu homeostasis and will uncover new molecular targets for the development of next generation therapeutic approaches to treat disorders associated with Cu imbalance.</p>

PubMed Author Manuscript

Effect of Photobiomodulation Therapy on Oxidative Stress Markers in Healing Dynamics of Diabetic Neuropathic Wounds in Wistar Rats

BackgroundProlonged and overlapping phases of wound healing in diabetes are mainly due to the redox imbalance resulting in the chronicity of the wound. Photobiomodulation therapy works on the principle of absorption of photon energy and its transduction into a biological response in the living tissue. It alleviates the cellular responses, thereby improving the mechanism of wound healing in diabetes.ObjectiveTo find out the effect of photobiomodulation therapy of dosage 4 J/cm2 in the healing dynamics of diabetic neuropathic wounds in Wistar rats and its relation with oxidative stress markers.MethodologyDiabetes was induced using Streptozotocin of 60 mg/kg of body weight to eighteen female Wistar rats. Neuropathy was induced by the sciatic nerve crush injury followed by an excisional wound of 2 cm2 on the back of the animal. Experimental group animals were treated with dosage 4 J/cm2 of wavelength 655 and 808 nm, and control group animals were kept unirradiated. The biomechanical, histopathological, and biochemical changes were analysed in both groups.ResultsThere was a reduction in mean wound healing time and an increased rate of wound contraction in the experimental group animals compared to its control group. The experimental group showed improved redox status, and histopathological findings revealed better proliferative cells, keratinisation, and epithelialization than un-irradiated controls.ConclusionsPhotobiomodulation therapy of dosage 4 J/cm2 enhanced the overall wound healing dynamics of the diabetes-induced neuropathic wound and optimised the oxidative status of the wound, thereby facilitating a faster healing process.

effect_of_photobiomodulation_therapy_on_oxidative_stress_markers_in_healing_dynamics_of_diabetic_neu

Introduction<!>Animal Selection and Care<!>Induction of Diabetes<!>Confirmation of Diabetes<!>Induction of Neuropathy by Sciatic Nerve Injury<!>Confirmation of Neuropathy<!>5.07 monofilament (10g) test<!>Hind paw withdrawal test for hot stimulus<!>Hind paw withdrawal test for cold stimulus using acetone<!>Excisional Wound Model<!><!>Wound area measurement<!>Rate of contraction of wound<!>Histopathological analysis<!>Immunohistochemical analysis<!>Estimation of antioxidant catalase assay<!>Lipid peroxidation by malondialdehyde (MDA) assay<!>Statistical Analysis<!><!>Discussion<!>Strength and Limitations<!>Conclusion<!>Future Direction<!>

<p>Diabetes mellitus (DM) is a leading non-communicable disease in India and Globally. Individuals with DM suffer from microvascular and macrovascular complications that predispose them to develop conditions like cardiovascular diseases, peripheral vascular diseases, retinopathy, nephropathy, and peripheral neuropathy [1]. Diabetic foot ulcer (DFU) is the most complex yet neglected complication of diabetic peripheral neuropathy (DPN). The intrinsic and extrinsic factors of DPN makes the feet vulnerable to injuries and tissue damage, causing chronic wound or ulcer [2]. The global burden of DM affected nearly 463 million people in 2019 and is anticipated to rise up to 700 million by the end of 2045. During the 2019 survey, India reported the prevalence of 72 million people affected by DM and an anticipated increase of up to 153 million by 2045. Globally, foot ulcers occur in 15–25% of people with DM and between 5–7.5% of those with neuropathy [3].</p><p>Delayed wound healing in DM is a major secondary complication that often ends with the loss of limb and disability. Regular wound repair progresses through the discrete phases of inflammatory, proliferative, and remodelling. In DM, these phases are prolonged and overlapped due to underlying excessive tissue damage, tissue perfusion, and lack of oxygenation that delay the wound healing process [4]. The tissue renewal process depends on the level of oxygen (O2) at the wound bed. The fundamental cellular process where oxygen is involved in the oxidative phosphorylation in mitochondria, mainly yields adenosine triphosphates (ATPs). The derivatives of O2, known as reactive oxygen species (ROS) produced during mitochondrial ATP production, act as signalling molecules in tissue repair. The low-level ROS results in cell cycle arrest, increased levels result in oxidative damage and cell apoptosis. Therefore, the optimum levels of ROS molecules are required for the normal tissue repair process [5]. The homoeostasis of ROS molecules generated during wound healing is by the special group of molecules known as antioxidants. These antioxidants have the capacity to significantly delay or prevent oxidative action of ROS and mitigate its deleterious effects. The well-known antioxidants are glutathione, superoxide dismutase, and catalase etc. [6].</p><p>Catalases are the haem containing antioxidant enzyme mainly produced in nuclear peroxisomes and mitochondria that catalytically decompose Hydrogen peroxide (H2O2) into oxygen and water using iron and manganese as co-factors. It can reduce the concentration of H2O2 at a rapid rate without consuming cellular energy, thereby providing cell defence against oxidative damage. However, in diabetic wound healing, catalase activity is decreased, leading to increased H2O2 concentrations in the blood and tissues [7, 8]. The failure of redox balance at the wound site results in delayed cellular responses, failing the wound recovery.</p><p>Photobiomodulation therapy (PBMT) is a promising conservative non-pharmacological approach in the management of diabetic wound healing. PBMT works on the principle of low energy biostimulation, causing photochemical reactions in the cells/tissue [9]. Studies have explored the effectiveness of PBMT in diabetic wound healing in vitro, in vivo, where PBMT of dosage 2–6 J/cm2 had shown a beneficial effect in wound healing dynamics [10–12]. PBMT is believed to alter enzyme activation and cell cycle progression, thereby affecting cells' redox-sensitive transcription factors. PBMT enhances antioxidants production by enzymatic reactions, which are the fundamental mechanisms involved in wound healing [13].</p><p>Although there are studies involving the mechanism of PBMT in wound healing, very few studies have elucidated the dose-dependent mechanism of PBMT and their relation with oxidative markers in the healing dynamics of diabetic neuropathic wounds. Therefore the present study aims to find out the effect of PBMT of dosage 4 J/cm2 in the healing dynamics of diabetes-induced neuropathic wounds in Wistar rats, with biomechanical, histopathological, and biochemical finding which may add substantial evidence to the existing literature of wound healing in diabetes.</p><!><p>We obtained the institutional animal ethics committee approval before the study's initiation (IAEC/KMC/95/2018). In house bread, eighteen female Albino Wistar strain rats of weight (220.78 ± 11.87) g and age (5.43 ± 0.11) months were procured from the institutional central animal research facility. The animals were housed in propylene cages (29 × 22 × 14 cm) with paddy husk bedding. Their living environment was maintained with temperature 28 ± 1 °C, humidity 55 ± 5%, and 12 h of light and dark cycle. The animals were provided with sufficient food once in the day and water ad libitum.</p><!><p>Baseline blood glucose levels were measured using calibrated Accu-check Performa Nano glucometer (Roche diagnostics India). Diabetes was induced to all the animals by intraperitoneal injection of Streptozotocin (STZ) (MP BiomedicalTM India) of dosage 60 mg/kg body weight dissolved in 0.1 M citrate buffer of pH 4.5 (1 mL) [14]. Food and water were provided after 30 min of injection. Further, the animals were placed in individual metabolic cages to observe clinical and behavioural changes of DM for seven days.</p><!><p>The animals were observed for polydipsia, polyphagia, and polyuria, and changes were recorded. Besides, the weight of the animals was recorded on alternative days. On the 7th day, blood was drawn in the fasting state from the intra-orbital plexus by inserting a capillary. Blood (1.5 mL) was drawn into the test tube containing sodium fluoride. The sample was mixed thoroughly and allowed to clot for 20 min. The sample was then centrifuged at 2000 × g for 10 min, and the supernatant serum was collected. Blood glucose levels were estimated using the glucose-oxidase and peroxidase method (GOD-POD kit, Coral clinical systems India). Animals with fasting blood glucose levels of ≥200 mg/dL were included in the study. Blood glucose levels were monitored periodically using calibrated Accu-check Performa Nano glucometer.</p><!><p>We followed previously established protocols developed for the neuropathic model in all the animals [15, 16]. Under expert supervision, the animals were anaesthetised with intravenous Ketamine of dosage 2 mg/kg body weight. The right/left hind leg from the knee to the hip was shaved using an electric shaver. An incision was made ~0.5 cm parallel to the femur and about 1.5 mm anterior to the femur. Under sterile conditions, the sciatic nerve was revealed until mid-thigh level and crushed for 20 s with the tip of watchman forceps (2 × 15 s) to induce neuropathy.</p><!><p>The animals were kept for three weeks of the stabilisation period. Their clinical and behavioural changes for neuropathy were recorded. The walking, scratching, grooming, resting/sleeping, eating, and freezing activities were observed. The following tests were performed to confirm the neuropathy in the feet.</p><!><p>Semmes Weinstein 10 gm monofilament is used in human participants for the detection of loss of protective sensation. A slightly modified procedure was performed on the animal's feet [17]. Here, the animals were placed on a metal mesh grid. Monofilament was applied perpendicularly with a force of 10 g from the lower surface of the mesh to the plantar areas of the hind paws. The time taken in seconds by the animals to withdraw paw in response to monofilament touch was noted. In the normal feet, The time taken for the paw withdrawal was between 1–4 s, whereas the neuropathy induced feet showed a delayed response.</p><!><p>We followed the previously established procedure for hind paw withdrawal to confirm the neuropathy [16]. The animals were placed on a metal hot plate (45 ± 0.5° C) of dimensions 15 × 15 cm for an evaluation period of 3 min. The total duration for responding to the hot stimulus by each animal to lift neuropathy induced and non-induced leg was noted. In the non-induced feet, The time taken for the paw withdrawal was between 1 and 4 s, whereas the neuropathy induced feet showed a delayed response.</p><!><p>The slightly modified procedure from the previous study was conducted [18]. The animal was placed upon a metal mesh grid with open access to the paws from below. A cotton swab was dipped in acetone and was brought into the plantar surface of the hind feet. The time taken to withdraw the leg from the supporting mesh after the exposure to acetone was noted. Both normal and neuropathy-induced legs were exposed to acetone, and the test was performed for 3 min. The response within 4 s was considered a normal, brisk response to differentiate from the altered lifting.</p><p>On the final day of PBMT intervention, the tests were repeated to evaluate to determine the changes in their neuropathic state.</p><!><p>An excisional wound model was created on the back of the animals near the neuropathy induced leg. Animals were anaesthetised using intravenous Ketamine of 2 mg/kg body weight. The wound area to be created was marked using a circular rubber stamp of area 2 cm2 dipped in methylene blue. The dorsal fur was shaved, and an excisional wound was created along the markings using toothed forceps using number 21 surgical blade and pointed scissors [11]. The area of the wound was recorded on transparent sheets. After the procedure, each animal was kept in a separate cage, and all wounds were exposed to air.</p><!><p>Control group N = 9 (diabetic neuropathic wounded-unirradiated)</p><p>Experimental group N = 9 (diabetic neuropathic wounded-irradiated with 4 J/cm2)</p><p>Characteristics of PBMT used in the study</p><!><p>The wound areas were recorded on a transparent sheet. The area was measured using ImageJ software (Downloaded from https://imagej.nih.gov/ij/download.html Windows Download ImageJ bundled with 64-bit Java 1.8.0_172). The total area was represented in cm2.</p><p>Mean wound healing time was calculated by taking the sum of the number of days taken for complete wound closure by each animal in the control and experimental group. Data was reported as Mean ± SD (N = 9 in each group). Serial photographs were taken for observation of wound closure on day-1, 7 and 16.</p><!><p>The rate of contraction of the wound was calculated using the formula [initial area (I)-final area (F)/number of days (N)] per day cm2 [11].</p><!><p>The tissues were processed on day 16th from both the experimental and the control group and</p><p>fixed in formalin (10%) for further histopathological analysis. As per standard laboratory protocol, the tissue samples were dehydrated with ascending grades of alcohol. The samples were then transferred to Xylene to clear the remaining alcohol. The tissue was embedded and moulded in paraffin wax. The 5-μm thick sections were obtained using a microtome, and clean sections were fixed to slides. The slides were stained with H&E staining to observe epithelialization and distribution of collagen fibres [19]. Each section was then viewed under the light microscope (Lx 300, Labomed, USA).</p><!><p>Immunohistochemistry was performed for the qualitative analysis of proliferative cells using Anti-Ki-67 antigen, Proliferating Cell [BioGenex, California], to compare the cell proliferation in the control and the experimental group. The previously established protocol was followed for immunohistochemical staining for Ki-67 antigen-specific proliferative cells [20]. Each section was then viewed under the light microscope (Lx 300, Labomed, USA).</p><!><p>Catalase activity was determined according to Beers and Sizer by spectrophotometric monitoring of hydrogen peroxide decomposition by hemolysates at 240 nm at room temperature [21]. Results were expressed in units/g of haemoglobin</p><!><p>MDA is a highly reactive three-carbon di-aldehyde produced as a by-product of polyunsaturated fatty acid peroxidation and arachidonic acid metabolism. Its level indicates the oxidative status of the cells. Serum MDA was measured using Kei Satoh's Method of spectrophotometric determination [22]. Results were expressed in nmol/g haemoglobin.</p><!><p>Descriptive statistics were measured using EZR (R version 3.4.1 (C) 2017 The R Foundation for Statistical Computing), and data were expressed in (Mean ± SD). For the normally distributed data, Paired sample t-test was used. When the data were not normally distributed, the non-parametric Wilcoxon signed-rank test was used. P < 0.05 was considered statistically significant.</p><!><p>Demographics and details of metabolic cage analysis of STZ induced animals (N = 18) (p value < 0.05 significant*). 1. Food intake (g) 2. Water intake (mL) 3. Urine output (mL) 4. Body weight (g) 5. blood glucose (mg/dl).</p><p>Response to neuropathy confirmation tests (N = 18)</p><p>Values expressed in (Mean ± SD)</p><p>*,**,*** represents the significant p values 0.17−3, 0.18−3, 0.15−3 respectively</p><p>Representative image of wound healing in experimental and control group animals. 1) Representative photographs of wound healing in experimental group in pictures 1A (Day-1), 1B (Day-7) and 1C (Day-16) group. 2) Representative photographs of wound healing in control group in pictures 2A (Day-1), 2B (Day-7) and 2C (Day-16) group</p><p>Rate of wound contraction in the animals (N = 9 in each group).</p><p>Values expressed in (Mean ± SD).</p><p>*, ** represents the significant p values 0.013, 0.014, respectively</p><p>Post neuropathy assessments (N = 9 in each group).</p><p>Values expressed in (Mean ± SD)</p><p>*,**,*** represents the significant p values 0.80, 0.12, 0.012, respectively</p><p>Representative photomicrographs of H&E staining of skin tissue of the experimental and control group. (1) Representative photomicrographs of H&E stained tissues from the wound area of experimental group treated with 4 J/cm2 (3A) showing better epithelialization than non-irradiated control group (3B) (marked with black arrows) under 10× magnification. (2) Experimental group (3C) and control group (3D) show the collagen fibres in the dermis. 3C represents more organizely arranged collagen fibres when copared to 3D (marked with black circle) under 40× maginification</p><p>Representative photomicrographs of Immunohistochemical analysis for Ki-67 antibodies for proliferative cells. Representative photomicrographs of immunohistochemical analysis of the tissues from the wound stained for IHC-Ki67 antibody for proliferative cells. 4A Experimental group treated with 4 J/cm2; 4B Control group. It can be noted that the experimental group showed well-stained nuclei indicating well proliferative cells especially near the stratum basale of epidermis when compared to control group (marked with black circle)</p><p>Catalase and MDA levels in the experimental and control group (N = 9 in each group).</p><!><p>DM is characterised by persistent hyperglycaemia that builds up the blood vessels' sugar complexes, resulting in poor circulation and deranged nerve functions [23]. In the present study, the STZ-induced animals showed high blood glucose levels from day seven. The sciatic nerve injury to the animals demonstrated the degenerated nerve functions. The control and experimental group animals showed the typical polydipsia, polyphagia, polyuria, and gradual weight loss symptoms of DM. Akbarzadeh et al. demonstrated these characteristics of STZ induced animals [24]. The induced STZ destructs the islets of Langerhans beta cells, and as a result, clinical diabetes emerges in the animals.</p><p>Previous studies have described the different types of mechanical peripheral nerve injury and evaluations of neuropathic pain in human and animal models using the stimuli that give the degree of mechanical, chemical, and temperature-induced changes [17, 25, 26]. In the present study, we evaluated the neuropathic changes by 5.07 Monofilament (10 g) test, Hind paw withdrawal test for hot and cold stimulus where the neuropathy induced leg showed the delayed response. In Dowdall et al. study, the partial sciatic nerve ligation model produced a similar and significant change in lifting behaviour on the hot plate from minimal to the absence of lifting in the neuropathy-induced leg; however, the response depends on the type of sciatic injury [18].</p><p>There is evidence for the therapeutic role of PBMT in pain management, muscle repair, and nerve damage [27–29]. We also observed moderate improvements in the neuropathy-induced leg in the post-PBMT treated group compared with its control group, but the recovery was not complete. Bertolini et al. treated the animals' sciatic nerve with low-level laser therapy (LLLT) of dosage 4, 6 J/cm2, and 830 nm wavelength and found a significant change in paw-withdrawal time on the fifth day of therapy [30]. Anju et al. suggest that low-level laser therapy of dosage 3.1 J/cm2 significantly improved in painful diabetic neuropathy conditions [17]. PBMT can enhance the microcirculation of nerves, thereby improving the blood supply.</p><p>In the present study, the diabetic neuropathic wound models were treated with 4 J/cm2 of wavelength 655 and 808 nm for 6 and 3 min. The experimental group and the control group animals were kept unirradiated. We found that PBMT effectively accelerated the mean wound healing time and rate of contraction of wounds. Therefore, this significant improvement compared to the un-irradiated control group appears to be the direct effect of PBMT. Studies demonstrate the effect of PBMT of different dosages and wavelengths in chronic wound conditions like diabetic foot ulcers [31, 32]. Maiya et al. found that LLLT of dosage 3–6 J/cm2 and wavelength 655 nm accelerated the wound healing with significant improvement in wound contraction rate, whereas 7–9 J/cm2 decelerated the wound healing [11]. Lau et al. treated the diabetic excisional wound with LLLT of dosage 5 J/cm2 and 808 nm but of different power densities; found that PBMT significantly enhanced the percentage of wound closure [33]. The possible mechanism could be that, the photon energy delivered to the wound tissue is absorbed and produces biostimulation, thereby enhancing cell proliferation. Cell proliferation is characterised by neovascularization, epithelialization, and granulation tissue formation in post-inflammatory phases [9, 11].</p><p>We have analysed the histopathological changes in the experimental and control groups to justify these results. The experimental group's H&E stained skin tissues showed better epithelium, collagen formation, and distribution than the control group. The immunohistochemical findings revealed that the proliferative cells were present abundantly in the experimental group than the control group, representing better and faster-wound healing. Therefore, it is evident that PBMT enhanced the formation of epithelial cells, fibroblastic cells, and proliferative cells, thereby enhancing epithelial tissue, granulation tissue, and collagen. Similar findings were also observed in Kilík et al. and Lau et al., who treated the diabetic mouse skin with PBMT [34, 35]. However, the present study did not obtain the skin at different stages of wounding. The mechanism of PBMT in the acceleration of wound healing could be by the mild inflammatory reaction induced by the PBMT, which enhances or upregulates neo-angiogenesis and increases the blood flow around the wound. On the other hand, the capacity of the PBMT to prevent harmful reactions during the inflammatory phase, facilitating collagen formation, might have improved wound healing [36].</p><p>Another possible mechanism is that in diabetic wound conditions, cells undergo phagocytosis and produce a large number of proteinases and ROS molecules. Even though ROS molecules contribute to wound healing, in diabetic wound conditions, there will be an uncontrolled production of ROS by auto-oxidation of glucose and glycosylation of scavenging enzymes and depletion of low molecular antioxidant, resulting in oxidative stress [14]. The MDA level is the indicator for ROS generation because it represents the redox reaction level in the cell. In contrast, catalase levels represent the cellular defence, as it is a natural scavenging molecule of H2O2.</p><p>In the present study, we observed elevated MDA levels in the control group and the lower catalase activity. Whereas in the experimental group, we observed optimised catalase and MDA levels. Similar findings were seen in Denadai et al. who treated the diabetic skin wounds of PBMT 6 J/cm2 and 660 nm, and Tatmatsu-Rocha et al. who irradiated the super pulsed 904 nm laser of dosage 2.39 J/cm2 [37, 38]. It should be noted that these changes were observed in wounded tissues. In contrast, our study estimated the serum levels of catalase and MDA. However, the changes observed were significant.</p><p>In diabetic wound conditions, the primary source of ROS is mitochondria. The non-thermal, photochemical reactions of the PBMT optimizes the mitochondrial redox potential of the electron transport chain, which is sensed and transmitted to the cytosol to regulate catalase activity and other enzyme activations. Therefore, PBMT can mediate cell signalling to produce antioxidant molecules to nullify the effect of ROS. Hence, it is evident that the net amount of ROS and antioxidants required for wound healing is facilitated by PBMT, a key for fibroblastic proliferation and angiogenesis in wound healing [39].</p><p>In the present study, we observed that, after the irradiation of PBMT of dosage 4 J/cm2 to the experimental group, catalase levels were significantly improved, representing a better antioxidant status. In contrast, MDA levels were significantly decreased in post-PBMT to diabetes-induced neuropathic wounds. Therefore, these may serve as adjuvant markers for oxidative status in diabetic wound conditions.</p><!><p>The present study evaluates the serum levels of oxidative stress markers whose evidence can be translated into clinical research. However, the analysis of these markers in each phase of the wound healing process is not done. The histological findings were taken during the final stages of wound healing. However, it can be considered at inflammatory, proliferative, and remodelling phases for further evidence.</p><!><p>PBMT of dosage 4 J/cm2 enhanced the overall wound healing dynamics in the diabetes-induced neuropathic excisional wound model. PBMT has improved the oxidative status of the wound with optimal changes in catalase and MDA levels and facilitated a faster healing process. We observed an accelerated rate of wound contraction and reduction in the mean healing time required for wound closure compared to the un-irradiated control group. In addition, PBMT also enhanced the skin epithelialization, keratinisation, and proliferation of diabetes-induced neuropathic wounds.</p><!><p>At present, in clinical practice, the management of diabetic foot ulcer and their associated complications are challenging to clinicians. Based on the present study findings, we recommend that PBMT can be one of the promising adjuvant modality in clinical practice. Based on our findings with markers of oxidative stress, may give a clear direction for the wound condition and its response to the PBMT. However, we recommend conducting further research with different doses and wavelengths.</p><!><p>adenosine triphosphate</p><p>diabetic foot ulcer</p><p>diabetes mellitus</p><p>diabetic peripheral neuropathy</p><p>low-level laser therapy</p><p>malondialdehyde</p><p>photobiomodulation Therapy</p><p>reactive oxygen species</p><p>streptozotocin</p><p>Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.</p>

PubMed Open Access

Selenium-dependent metabolic reprogramming during inflammation and resolution

Trace element selenium (Se) is incorporated as the 21st amino acid, selenocysteine, into selenoproteins through tRNA[Ser]Sec. Selenoproteins act as gatekeepers of redox homeostasis and modulate immune function to effect anti-inflammation and resolution. However, mechanistic underpinnings involving metabolic reprogramming during inflammation and resolution remain poorly understood. Bacterial endotoxin lipopolysaccharide (LPS) activation of murine bone marrow–derived macrophages cultured in the presence or absence of Se (as selenite) was used to examine temporal changes in the proteome and metabolome by multiplexed tandem mass tag–quantitative proteomics, metabolomics, and machine-learning approaches. Kinetic deltagram and clustering analysis indicated that addition of Se led to extensive reprogramming of cellular metabolism upon stimulation with LPS enhancing the pentose phosphate pathway, tricarboxylic acid cycle, and oxidative phosphorylation, to aid in the phenotypic transition toward alternatively activated macrophages, synonymous with resolution of inflammation. Remodeling of metabolic pathways and consequent metabolic adaptation toward proresolving phenotypes began with Se treatment at 0 h and became most prominent around 8 h after LPS stimulation that included succinate dehydrogenase complex, pyruvate kinase, and sedoheptulokinase. Se-dependent modulation of these pathways predisposed bone marrow–derived macrophages to preferentially increase oxidative phosphorylation to efficiently regulate inflammation and its timely resolution. The use of macrophages lacking selenoproteins indicated that all three metabolic nodes were sensitive to selenoproteome expression. Furthermore, inhibition of succinate dehydrogenase complex with dimethylmalonate affected the proresolving effects of Se by increasing the resolution interval in a murine peritonitis model. In summary, our studies provide novel insights into the role of cellular Se via metabolic reprograming to facilitate anti-inflammation and proresolution.

selenium-dependent_metabolic_reprogramming_during_inflammation_and_resolution

<!>Activation of macrophages and effect of Se supplementation on global proteome<!>Se-dependent modulation of enzymes and metabolites in the TCA cycle, glycolysis, and PPP<!><!>Discussion<!><!>Materials<!>Mice and genotyping<!>BMDM culture: LPS stimulation<!>Murine peritonitis and flow cytometry<!>Sample preparation for proteomics, data acquisition, and analysis<!>Sample preparation for metabolite acquisition and data analysis<!>Western immunoblot<!>Data availability<!>Supporting information<!>Conflict of interest<!>Supplementary File 1

<p>Edited by John Denu</p><p>Trace element selenium (Se) is incorporated as the 21st amino acid, selenocysteine, via tRNA[Ser]Sec (encoded by Trsp) dependent decoding of the UGA codon in 24 murine (25 in humans) selenoproteins (1, 2). Selenoproteins function as redox gatekeepers and catalyze reactions involving reduction of disulfides, methionine sulfoxide, and peroxides (1, 3). In addition, selenoproteins also modulate immune functions through oxidative homeostasis, prevention of iron-induced cellular ferroptosis, regeneration of reduced thioredoxin, regulation of actin repolymerization during innate immune response, and cellular calcium homeostasis (4). Although a highly debated topic, Se supplementation and its benefits in severe systemic inflammatory response syndrome, sepsis, and septic shock assumes a J-shaped curve relationship, suggesting supplementation therapies primarily benefit Se-deficient patients (5, 6). However, their mechanistic role in the modulation of pathways associated with metabolic reprograming during inflammation is poorly understood (7, 8, 9, 10, 11).</p><p>Phenotypic plasticity of macrophages as seen in the form of classically activated M1 proinflammatory phenotype, upon treatment with bacterial endotoxin lipopolysaccharide (LPS), or an alternatively activated (M2) anti-inflammatory phenotype, a trait synonymous with cellular mechanisms of resolution, upon treatment with interleukin (IL)-4 or IL-13, represents two ends of the polarization spectrum (12). The resolution program involves a highly coordinated and systemic response, involving transmigration, phagocytosis, antigen presentation, and expression of proresolving genes such as arginase-1 and Mrc-1 (CD206) that impinge on redox-dependent signaling and cellular metabolism (12, 13, 14, 15, 16). Monocyte/macrophage-specific deletion of selenoproteins has led us to recognize the importance of the selenoproteome in the transition of M1- to M2-like phenotype and resolution as seen in experimental models of gut inflammation and hematologic malignancies (17, 18, 19, 20, 21, 22, 23, 24, 25). Recent studies from our laboratory using tandem mass tag (TMT)-labeling method of nonselenocysteine peptides in murine bone marrow–derived macrophages (BMDMs) indicated a temporal regulation with a general increase in the selenoproteome upon LPS stimulation in a Se-dependent manner (26). Selenow, Gpx1, Msrb1, and Selenom were highly upregulated upon stimulation with LPS when compared with other selenoproteins. Together, it appears that selenoprotein-dependent pathways of anti-inflammation and proresolution likely impinge on metabolic reprogramming, which is not well understood.</p><p>It is evident that changes in intracellular metabolic pathways, such as glycolysis, pentose phosphate pathway (PPP), tricarboxylic acid (TCA) cycle, fatty acid oxidation, fatty acid synthesis, and amino acid metabolism, impact cellular functions in immune cells (27, 28). Macrophages stimulated with LPS predominantly display a glycolytic metabolic phenotype (14) and decreased oxidative phosphorylation (OXPHOS), like the "Warburg effect" in cancer cells (14, 17, 27, 29). LPS-activated macrophages inhibit expression of several enzymes of the TCA cycle, including succinate dehydrogenase (Sdh) complex (14) leading to accumulation of succinate and increased activation of hypoxia inducible factor-1α (Hif-1α) and IL-1β culminating in high glycolytic rates to favor M1 polarization (30). In fact, two distinct breaks have been identified in the TCA cycle in M1 macrophages that include the isocitrate dehydrogenase (Idh1) and Sdh complex (14, 31). Interruption of Idh1 and Sdh results in accumulation of citrate and itaconate (32, 33), and succinate, respectively. Conversely, M2 macrophages predominantly exhibit an OXPHOS metabolic phenotype (27, 28), relying on oxidation of glucose and fatty acids to sustain OXPHOS-mediated generation of ATP. Furthermore, LPS activation of macrophages downregulates pyruvate kinase 2 (Pkm2), whereas Pkm2 activation inhibits IL-1α production and Hif-1α–dependent genes, promoting an M2-like phenotype (34). In addition, the activity of sedoheptulokinase (Shpk) or carbohydrate-like kinase (Carkl), which catalyzes an orphan reaction in the PPP, involving the conversion of sedoheptulose to sedoheptulose 7-phosphate was shown to refocus cellular metabolism favoring an M2-like phenotype and a high-redox state (35).</p><p>Using TMT-labeled quantitative proteomics, targeted metabolomics, and unsupervised learning algorithms, we report here that Se in the form of selenoproteins modulates metabolic pathways involving the Pkm2, Shpk, and Sdh complex to predispose activated macrophages toward OXPHOS to efficiently regulate inflammation and timely resolution. Inhibition of the Sdh complex with dimethylmalonate (DMM) (36) affected the proresolving effects of Se in a zymosan model of peritonitis. Therefore, cellular Se, through its incorporation into selenoproteins, serves as a diet-derived regulator of metabolic reprograming to facilitate anti-inflammation and proresolution.</p><p>Experimental design and data acquisition. Bone marrow–derived macrophages (BMDMs) were harvested and cultured for 7 days either in the presence or absence of Se (as sodium selenite, Na2SeO3, 250 nM). BMDMs were stimulated with 100 ng of LPS for indicated time periods up to a maximum of 20 h. Cells were harvested after two washes with PBS and lysed to extract the proteome. Proteome was extracted and subjected to tryptic digestion followed by TMT labeling. Datasets were acquired on Lumos Fusion MS and analyzed by using Proteome Discoverer. Metabolites were extracted in methanol, dried, and reconstituted in the mobile phase with chlorpropamide as the internal standard. The dataset was acquired on Exactive Plus MS. Corresponding metabolite ion chromatograms were extracted using Xcalibur and inspected manually. LPS, lipopolysaccharide; TMT, tandem mass tag.</p><p>Supplementation of macrophages with Se effects global proteomic changes upon treatment with LPS. BMDMs were stimulated with LPS for indicated time periods up to a maximum of 20 h. Cells were harvested and lysed to extract the proteome. Proteomic analysis is described in Experimental procedures. BMDMs cultured in Se-deficient media and not subjected to any stimulation were used for comparison. Data were normalized to β-actin, and abundance values are expressed as relative to zero-hour zero-Se cells (naïve cells). A, scatterplots show biological replicates after LPS, which are well correlated (Pearson's r of 0.96). B, the bar chart shows the overall proteome changes during after LPS stimulation. Values are represented as log2 fold changes. C, the heat map showing the modulation of global BMDM proteome after LPS stimulation. D, the t-SNE analysis and k-means clustering of proteomic changes after LPS show seven distinct clusters, respectively. BMDMs, bone marrow–derived macrophages; LPS, lipopolysaccharide; Se, selenium; t-SNE, t-distributed Stochastic neighbor embedding.</p><!><p>By applying clustering analysis of proteins with similar kinetics and magnitude of induction (Fig. 2C), we identified 121 proteins that were differentially regulated at 0 h in Se-treated cells compared with unstimulated Se-deficient BMDMs ("naïve" cells). Although majority of the selenoproteins were increased as reported earlier upon exogenous Se addition (26), nuclear factor-erythroid 2-related factor 2 target genes, such as thioredoxin reductase1 (Txnrd1) and Txnrd2, and hemeoxygenase 1 (Hmox1) were also increased by Se supplementation followed by LPS stimulation for 8 h. In contrast, 510 proteins were differentially regulated upon stimulation with LPS for 20 h (Fig. 2C). Interestingly, clustering analysis indicated heterogeneity within LPS-responsive proteins altered upon Se supplementation that separated into seven distinct clusters (Fig. 2D; Fig. S2). Specifically, clusters 1, 3, 4, 5, and 6 were enriched with 289, 631, 1292, 1496, and 536 proteins, respectively, upon LPS treatment (Fig. S2). Cluster 1 represented early induced proteins impacted between 4 and 8 h after LPS treatment. The magnitude of protein induction was higher in the presence of Se than 0 Se, which included proteins involved in the TCA cycle and electron-transport chain that were upregulated, which was further confirmed through pathway analysis. Significant enrichment of the TCA cycle and OXPHOS pathways in clusters 1, 5, and 6 showed induction of pathway-associated proteins from 4 h after LPS stimulation. Interestingly, we observed maximally expressed proteins in cluster 1, which corroborated the upregulation of proteins associated with the TCA cycle and OXPHOS (Fig. S2). Our analysis indicated that remodeling of metabolic pathways and consequent metabolic adaptation toward proresolving phenotypes started with Se treatment of Se-deficient BMDMs in the absence of any stimulation and became most prominent around 8 h after LPS stimulation.</p><!><p>Initiation and resolution of inflammation are energy-demanding processes that require timely adaptation of cellular metabolism for support. Because proteomic analysis suggested a plausible metabolic alteration in response to Se treatment, we examined the metabolic consequences of Se supplementation on macrophage activation.</p><!><p>Differential modulation of key metabolites via Se in LPS-stimulated BMDMs indicate a Se-dependent reduction in glycolytic pathway and increased the TCA cycle. Metabolites were extracted in methanol, dried, and reconstituted in the mobile phase aqueous methanol containing 100-μM chlorpropamide as an internal standard. All samples were acquired in biological triplicate and randomized before 10 μl was acquired in LC-MS using reverse-phase UHPLC coupled to an Exactive Plus orbitrap MS. A total of 47 metabolites were profiled in biological triplicates. Individual metabolite extracted-ion chromatograms were used, with the peak area determined. All peak areas were normalized to internal standard chlorpropamide, and abundance was calculated relative to that of zero-hour zero-selenium cells (naïve cells). A, the heat map profiling of inflamed BMDMs show log2 fold change values of all 47 metabolites. All displayed log2 fold changes are represented in biological triplicates. B, the t-SNE method applied to all samples show distinct clusters separating the samples by time after LPS stimulation. BMDMs, bone marrow–derived macrophages; LPS, lipopolysaccharide; Se, selenium; TCA, tricarboxylic acid; t-SNE, t-distributed Stochastic neighbor embedding.</p><p>Selenoprotein-dependent changes in the TCA cycle metabolites in WT and Trspfl/fLysMCreBMDMs after LPS stimulation. Metabolites were extracted from BMDM cells isolated from WT and Trspfl/flLysMCre mice as described earlier and analyzed by LC-MS. BMDMs from n = 3 mice per genotype were cultured as described earlier and stimulated for various time points 0 to 20 h with LPS. Metabolite peak areas were normalized to internal standard chlorpropamide, and abundance was calculated relative to that of zero-hour zero-selenium cells (naïve cells). Se supplementation of WT BMDMs with 250-nM Se (as selenite) showed dramatic regulation of succinate levels after LPS stimulation when compared with the BMDMs isolated from Trspfl/flLysMCre mice. BMDMs, bone marrow–derived macrophages; LPS, lipopolysaccharide; Se, selenium; TCA, tricarboxylic acid. ∗p < 0.05; ∗∗p < 0.005; ∗∗∗p < 0.0005; ∗∗∗∗p < 0.00005.</p><p>Selenoprotein-dependent changes in the TCA cycle proteins.A, the heat map showing temporal regulation of the TCA cycle proteins as determined by proteomic studies. B, the Western blot of Sdha and Sdhb. BMDMs were isolated from WT and Trspfl/flLysMCre mice, respectively, and incubated with L929 containing DMEM without Na2SeO3 (250-nM Se) for 8 days. Cells were stimulated with 100-ng LPS on day eight for 1 h, and cells were harvested at 4, 8, and 20 h, followed by protein extraction and Western blot analysis of BMDMs in biological triplicate per genotype. Densitometric analysis of Sdha and Sdhb expression was performed using ImageJ and normalized to that of β-actin. C and D, densitometric evaluation of Western immunoblots for the expression of Sdha and Sdhb in BMDM incubated with Se compared with those cultured without Se before and after LPS stimulation. Representative data shown are the mean ± SEM of n =3 per genotype at each time point after LPS treatment. ∗p < 0.05; ∗∗p < 0.005; ∗∗∗p < 0.0005. BMDMs, bone marrow–derived macrophages; DMEM, Dulbecco's Modified Eagle Medium; LPS, lipopolysaccharide; Se, selenium.</p><p>Selenoprotein-dependent changes in glycolytic cycle proteins.A, the heat map showing temporal changes in glycolytic pathway proteins as determined by proteomic studies. B and C, the Western blot of pPkm2 and Pkm2 (total). BMDMs were isolated from WT and Trspfl/flLysMCre mice. Cells were stimulated with 100-ng LPS on day eight for 1 h and harvested at 4, 8, and 20 h, followed by protein extraction and Western blot analysis. Densitometric analysis of Western immunoblots using ImageJ program provided pPkm2:Pkm2 ratio that was normalized to β-actin and compared with cells cultured without Se. Representative data shown are the mean ± SEM of n =3 per genotype at each time point after LPS stimulation. ∗p < 0.05; ∗∗p < 0.005. BMDMs, bone marrow–derived macrophages; LPS, lipopolysaccharide; Se, selenium.</p><p>Selenoprotein-dependent changes in Shpk (Carkl).A and B, western immunoblot of Carkl expression upon stimulation of BMDMs with LPS in the presence or absence of Se. BMDMs were isolated from WT and Trspfl/flLysMCre mice, respectively, and cells were stimulated with 100-ng LPS on day eight for 1 h and harvested at 4, 8, and 20 h, followed by protein extraction and Western blot analysis. C and D, Densitometric analysis using ImageJ of Shpk (Carkl) expression that was normalized to the expression of β-actin. Shpk fold change in BMDMs incubated with Se were compared with those without Se. E, fold change in relative expression of Shpk in WT and LysM KO over time post LPS stimulation. Representative data shown are the mean ± SEM of n =3 per genotype at each time point after LPS stimulation. ∗p < 0.05. BMDMs, bone marrow–derived macrophages; LPS, lipopolysaccharide; Se, selenium.</p><p>DMM treatment of Se-Supplemented mice decreases the resolution index in a Zymosan-induced peritonitis model. Se-supplemented mice were treated intraperitoneally with or without DMM in PBS (160 mg/kg of body weight) or PBS 3 h before administration of Zymosan A (10 mg/kg body weight) intraperitoneally. Mice were euthanized at 0, 12, 24, and 48 h after Zymosan A injection, and peritoneal exudates were collected, centrifuged at 400g for 5 min at 4 °C, washed with the flow buffer (PBS containing 2% FBS and 100 I.U. penicillin and 100 μg/ml streptomycin), resuspended in 1-ml flow buffer, and counted by trypan blue, and 100,000 viable cells were analyzed by flow cytometry. A, the resolution index (Ri) calculation based on the percentage of neutrophils (Ly6-G+ cells) in the peritoneal lavage fluid at each time point. Ψmax and Tmax represent the highest percentage of cells after Zymosan treatment and the time point at which it occurs, respectively. Data shown are representative of biological triplicates. B, the percentage of total macrophages (F4/80+) and M2-like macrophages (F4/80+ CD206+) in peritoneal lavage fluid as a function of time after Zymosan A treatment. C, the ratio of the total macrophages to M2-like macrophages in the peritoneal lavage fluid as a function of time after Zymosan A treatment. Data shown are the mean ± SEM of n = 3 to 4 per treatment group; ∗∗∗p < 0.0005. ∗∗∗∗p < 0.0001. DMM, dimethylmalonate; FBS, fetal bovine serum; Se, selenium.</p><!><p>Resolution of inflammation is an active and highly orchestrated process that restores normal functional tissue homeostasis (37), where macrophages play an integral role (11, 12). Plasticity of these cells arises from their ability to skew metabolism from glycolysis and fatty acid synthesis in M1 macrophages to OXPHOS and fatty acid β-oxidation, in M2 macrophages, to meet their energy demands for survival and functionality (27). Here, using multiomics platforms, we examined the effect of micronutrient Se and selenoproteins on the reprogramming of metabolic pathways involving the Pkm, Shpk (Carkl), and Sdh complex to predispose macrophages toward OXPHOS and eventually facilitate resolution of LPS-induced inflammation.</p><p>In addition to increased mitochondrial ROS levels, significant reduction in both citrate and aconitate, along with increased Idh expression, decreased α-ketoglutarate, succinate, fumarate, malate, and oxaloacetate by exogenous Se suggested increased mobilization of the TCA cycle, when compared to Se-deficient cells. Temporal increase in succinate after LPS, which moonlights as a key mediator of inflammation (30), was reversed by Se with an increase in fumarate and malate. Inhibition of the Sdh complex with DMM (36) affected the proresolving effects of Se in Zymosan-induced peritonitis in mice. Although the exact mechanisms of regulation of the Sdh complex by selenoproteins are unclear, inhibition of succination of reactive cysteine in fumarate hydratase (38) by Se could serve as a potential mechanism to help restore the broken TCA cycle.</p><p>Surprisingly, despite the increase in succinate that is known to increase Hif-1α, the expression of Hif-1α or its characteristic downstream genes, including IL-1β, glycolytic enzymes, and glucose transporters (30), was not observed in Se-deficient cells. However, we observed a decrease in glycolytic enzymes in macrophages supplemented with Se, suggesting considerable reduction in glycolytic rates. Pkm, a key regulatory node, was found to be significantly downregulated in macrophages supplemented with Se. Pkm is also regulated by phosphorylation status by upstream kinases that is indicative of dimer/tetramer formation (39). The dimeric and phosphorylated forms of Pkm2, which are enzymatically inactive, also positively regulate their expression and that of Hif-1α–dependent genes (40). The effects we observed were not similar to that reported with DASA-58, a selective activator of Pkm2, which inhibited the Hif-1α–dependent induction of glycolytic proteins by LPS (34), suggesting Hif-1α–independent control of metabolic effects in the presence of Se. Along with PKM, there was a decrease in the expression of Pfkm and Pfkl, but an increased AMP:ADP ratio in BMDMs cultured with Se, along with other nucleotides that persisted even after LPS stimulation (Fig. 3A). Together, such changes indicated functional glycolysis, albeit at low levels, necessitating auxiliary pathways such as the PPP that utilize glycolytic intermediates to produce precursors of NADPH, along with increased selenoprotein expression, to maintain redox homeostasis and help buffer ROS to ultimately influence macrophage polarization (28).</p><p>Previous studies suggested that loss of Shpk (Carkl) led to a significant drop of NADH levels resulting in a redox shift (35). Compared to the Se-deficient cells, Se supplementation of BMDMs held the NAD+:NADH ratio up to 4 h after LPS treatment, whereas NADP+:NADPH ratios were unchanged. The GSH:GSSG ratio decreased temporally in Se-supplemented macrophages upon LPS stimulation but tightly maintained in Se-supplemented cells, suggesting restoration of the redox status in favor of resolution. Increased levels α-ketoglutarate, aspartate, and glutamine in Se-supplemented cells corroborated with restoration of GSH/GSSG balance and the TCA cycle through increased α-ketoglutarate. Surprisingly, Txrnd1 and Hmox1 were mostly seen to increase transiently with Se, suggesting a fine control by nuclear factor erythroid 2–related factor 2–dependent mechanisms.</p><p>Se treatment of BMDMs increased cellular levels of amino acids before and after LPS (up to 2 h), perhaps to support many functions of macrophages in addition to metabolic reprogramming. Although not exactly identical, a similar trend was reported earlier in the whole liver (41). In addition, Se also increased monocarboxylate transporters, Slc16a1 and Slc16a3, and a high-affinity copper transporter Slc31a1, to perhaps support increased pyruvate metabolism, TCA cycle, and OXPHOS (data not shown). Macronutrients such as myo-inositol, a precursor for synthesis of phosphoinositides critical for signal transduction via second messengers IP3 and Ca2+, were consistently higher in the presence of Se and both in the absence or presence of LPS. Thus, a broader understanding of Se on myo-inositol metabolism in regulating macrophage functions may benefit from further studies.</p><p>Given the importance of fatty acid oxidation and OXPHOS, which drives the (M2) macrophage-mediated resolution and tissue repair (42), higher expression of Acsl3 and Acsl4, which are involved in fatty acid metabolism, along with Acadl, Acsl5, Adam17, Cyb5r3, Gcdh, Sgpp1, Sptlc1, Sptlc2, and the acetylCoA transporter slc33a1 (proteins represented by gene names), was temporally upregulated by Se (data not shown). In agreement with Se-dependent endogenous activation of PPARγ as reported previously (19), which is implicated in regulation of M2 genes, including Arg-1 and Mrc-1, fatty acid oxidation, adipocyte differentiation, and glucose homeostasis were upregulated upon LPS treatment (19, 43, 44, 45). Pathway network analysis suggested Ppar signaling–associated proteins such as Csf2rb, Cbl, Akt3, Pik3r5, Stat3, and Irf9 were upregulated by Se supplementation. Stat3 mediates the anti-inflammatory effects of IL-10 (46), further reinforcing the anti-inflammatory role of Se via multiple mechanisms that promote the proresolution phenotype of BMDMs.</p><!><p>Schematic representation illustrating the effect of Se on BMDMs before and after LPS stimulation. Key metabolites and pathways that were significantly upregulated (red) or downregulated (green) in LPS-treated (100 ng/ml) BMDMs ± Se (250 nM) are shown. BMDMs, bone marrow–derived macrophages; LPS, lipopolysaccharide; Se, selenium.</p><!><p>The TMT-labeling reagent kit and Dulbecco's Modified Eagle Medium were procured from Thermo Fisher Scientific. Sodium selenite and LPS from Escherichia coli (Serotype 0111:B4) were procured from Sigma Aldrich. Fetal bovine serum and L929 fibroblasts were purchased from Atlanta Biologicals and the American Type Culture Collection, respectively. The basal level of Se in the culture media was 7 nM as determined by inductively coupled plasma mass spectrometry–based methods. All other chemicals and reagents were of MS grade.</p><!><p>Three-week-old C57Bl/6 male mice were purchased from Taconic Biosciences, Inc and maintained on either an AIN-76–based semipurified Se-deficient diet (<0.01 ppm) or Se-supplemented diet (0.4 ppm) from Harlan-Teklad for at least 4 weeks before being used in the experiments as described previously (21). A transgenic C57Bl/6 line carrying a lysozyme M Cre (LysMCre) transgene was crossed to floxed Trsp (Trspfl/fl) allele, both generously provided by Dr Dolph Hatfield (the NIH, Bethesda, MD), as described previously (47). Genotyping was performed as described previously (21). All studies were preapproved by the Institutional Animal Care and Use Committee and the Institutional Biosafety Committee at the Pennsylvania State University (University Park, PA).</p><!><p>Three-week-old male C57Bl/6 mice were purchased from Taconic Biosciences, Inc and maintained on either an AIN-76–based semipurified Se-deficient diet (<0.01 ppm; Se-D) or Se-supplemented diet (0.4 ppm as selenite; Se-S) purchased from Harlan Teklad, for 7 weeks before being used in the experiments. Femoral bone marrow from Se-D and Se-S mice were used as a source of BMDMs upon culture with no exogenous or 250-nM Se (as sodium selenite) addition, respectively, as described (26). Cells were cultured in biological triplicates for 7 days with or without Se (as sodium selenite; 250 nM) as described previously (19, 26). On the eighth day, cells were stimulated with LPS (100 ng/ml; Sigma), where cells were treated with LPS for an hour followed by replacement with fresh culture media until harvest at 0, 4, 8, and 20 h after stimulation. Cells were washed with ice-cold sterile PBS and scraped, and cell pellets were stored at −80 °C until further processing.</p><!><p>Peritoneal inflammation was induced as described previously (48, 49). Se-S mice were treated intraperitoneally with DMM in PBS (160 mg/kg of body weight) or PBS alone 3 h before administration of Zymosan A (10 mg/kg body weight) intraperitoneally. Mice were euthanized at 0, 12, 24, and 48 h after Zymosan A injection, and peritoneal exudates were collected, centrifuged at 400g for 5 min at 4 °C, washed, and resuspended in the flow buffer (PBS containing 2% fetal bovine serum; 1 ml), and 100,000 viable cells were analyzed by flow cytometry. Pellets were suspended in 100 μl of the flow buffer and blocked with 0.25-μg Fc block (BD Pharmingen) for 10 min followed by staining with antibodies: 1 μl of phycoerythrin-conjugated anti-mouse Ly-6G (BD Pharmingen), 1 μl of FITC-conjugated anti-mouse CD206 (BioLegend), and 10 μl of APC-conjugated anti-mouse F4/80 (Miltenyi Biotec) for 30 min at 4 °C in the dark. Cells were washed twice with the flow buffer and centrifuged at 400g for 5 min at 4 °C and analyzed on a BD Accuri C6 Flow Cytometer and analyzed with FlowJo, version 10, software (FlowJo, LLC). The resolution of inflammation defined as the Ri is the time taken to reduce the number of neutrophils at the site of inflammation by 50% (50). All animal protocols were preapproved by the Institutional Animal Care and Use Committee at the Penn State University.</p><!><p>The proteome was extracted from frozen cell pellets in RIPA buffer (Thermo Fisher Scientific), and the protein concentration was determined by the bicinchoninic acid assay (Thermo Fisher Scientific). Equal amounts of protein (100 μg) were diluted with 100-mM triethyl ammonium bicarbonate buffer and reduced with dithiothreitol (10 mM) at 60 °C for 30 min, followed by alkylation with 2-chloroacetamide (65 mM) at room temperature in the dark for 30 min. The proteome was digested with proteomic grade tosyl phenylalanyl chloromethyl ketone–trypsin at 1:50 (enzyme-to-substrate) ratio overnight at 37 °C. The peptide digest was labeled with TMT and fractionated (lot number SH253273, and sample to TMT channel information is provided in Table S1). The proteomic dataset was acquired on samples in biological triplicate using an UltiMate 3000 RSLCnano system coupled online to an Orbitrap Fusion MS (Thermo Fisher Scientific) as described previously (26). Briefly, peptides were resolved on an Acclaim PepMap C18 column (2 microns, 75 μm i.d. × 50 cm) with a flow rate of 300 nl/min using a 0.1% formic acid/acetonitrile gradient. For MultiNotch MS3, the top ten precursors from each MS2 scan were fragmented by higher-energy collisional dissociation followed by Orbitrap analysis (NCE 55; 60,000 resolution; AGC 5 × 104; max IT 120 ms, 100–500 m/z scan range). Mass spectral datasets were analyzed with Proteome Discoverer (v2.2, Thermo Fisher Scientific) using SEQUEST-HT algorithm for database search for peptide identification and queried against the UniProt Mus Musculus database (17,424 proteins, downloaded on June 14, 2018) using the following parameters: peptide and fragment mass tolerance were 10 ppm and 0.6 Da, respectively, with two miscleavages, as described previously (26). The oxidation of methionine (15.995 Da) and deamidation of asparagine and glutamine (0.984 Da) were considered as variable modifications. TMT labeling of the N-termini of peptides as well as lysine (229.163 Da) and cysteine carbamidomethylation (57.021 Da) were considered as static modifications. Relative quantitation using TMT reporter ions was performed using high-quality MS3 spectra. A percolator algorithm was used to determine the false discovery rate, and only peptides with a false discovery rate  ≤0.01 were considered for further analysis.</p><p>Log-transformed fold-change data were lowess-normalized using StatsModels (statsmodels.org), and the correlations were measured using the ordinary least squares linear regression method from SciPy (scipy.org). The sorted heat map, stacked bar charts of the log2 fold changes, and related visualization were generated by using Matplotlib (matplotlib.org). Proteins were clustered using the k-means algorithm, and the separation analysis was performed using t-distributed Stochastic Neighbor Embedding algorithm from the scikit-learn machine learning library (scikit-learn.org). Cluster maps were generated using seaborn (seaborn.pydata.org). Clustering analysis compared protein's abundance relative to naïve (untreated) 0-h, 0-Se sample and 4, 8, and 20 h after stimulation with LPS. In each of the biological triplicates, two data points per protein were used for clustering. Values from the two replicates were averaged for drawing the curves in Figure 2A. Proteins identified in at least two replicates and with at least one unique peptide were included in this analysis. Kinetic deltagram algorithm was developed by Salis Lab at the Penn State University (salislab.net) to visualize fold-change trends across different time points, as well as identify the window of most notable change, and was implemented in Python. Functional data analysis was carried out using ingenuity pathway analysis (51), database for annotation, visualization and integrated discovery (52), and gene set enrichment analysis (53) for pathway enrichment.</p><!><p>Metabolites were extracted from harvested BMDMs in prechilled 80% (v/v) methanol. Samples were snap-frozen in liquid nitrogen, vortexed, and centrifuged for 20 min at 20,000g at 4 °C. The supernatant was dried using a SpeedVac and resuspended in 3% aqueous methanol containing 100-μM chlorpropamide as an internal standard. All samples were acquired in biological triplicate with a randomized sample order. 10 μl aliquots of each sample were analyzed using LC-MS.</p><p>Metabolites were analyzed using reverse-phase UHPLC (C18 Hydro-RP column; Phenomenex) coupled to an Exactive Plus Orbitrap MS (Thermo Fisher Scientific). A linear gradient from 3 to 100 (v/v) % methanol for 25 min was achieved with 3% methanol, 10-mM tributylamine, 15-mM acetic acid (solvent A) and 100% methanol (solvent B) (54). Mass spectra were acquired in a negative-ion mode, with a scan range of 85 to 1000 m/z with a resolution of 140,000 at m/z 200. A total of 47 metabolites were profiled across five time points of the experiment (0, 2, 4, 8, and 20 h) in biological triplicate. Individual metabolite extracted-ion chromatograms were used to determine the integrated peak areas. All metabolite peak areas were normalized to the chlorpropamide, and metabolite abundance was calculated relative to that derived from naive zero-hour samples. An in-house library generated from 292 authentic metabolite standards with experimentally observed accurate m/z values and retention times aided in the identification of metabolites (55).</p><!><p>Harvested cell pellets were resuspended in the mammalian protein extraction reagent (Thermo Fisher Scientific) containing protease inhibitor mixture (Roche Applied Science) and 5-mM sodium orthovanadate (Sigma), incubated on ice for 20 min, and vortexed for 10 min followed by centrifugation for 25 min at 20,000g at 4 °C. The protein concentration was determined using the bicinchoninic acid protein assay kit (Thermo Fisher Scientific). Proteins were resolved using a 12.5% (% T) SDS-PAGE and transferred onto a nitrocellulose membrane. The membranes were blocked with 7% (w/v) skim milk and probed with antibodies: anti-Sdha (1:2000; Cell Signaling Technology), anti-SDHB (1:4000; Proteintech), anti-Pkm (1:10,000; Cell Signaling Technology), anti-pPkm (1: 10,000; Cell Signaling Technology), anti-Shpk (Carkl) (1:2500; MyBioSource), and anti-β-actin (1: 25,000; Fitzgerald). Data from three independent experiments were used in densitometric analysis, whereas only a representative Western blot is shown for brevity.</p><!><p>The mass spectrometry proteomics data generated in this study have been deposited at the ProteomeXchange Consortium via the PRIDE (56) partner repository with the dataset identifier PXD023005.</p><!><p>This article contains supporting information.</p><!><p>The authors declare that they have no conflicts of interest with the contents of this article.</p><!><p>Database search and protein identification. Database search results of protein identification with the UniProt accession numbers, protein description, sum pep score (obtained from Proteome Discoverer, version 2.2), sequence coverage, number of peptides, PSMs, unique peptides, amino acids, molecular weight, calculated pI, gene symbol, and quantification measurements of each proteins are provided.</p>

PubMed Open Access

Rosetta Protein Structure Prediction from Hydroxyl Radical Protein Footprinting Mass Spectrometry Data

In recent years mass spectrometry-based covalent labeling techniques such as hydroxyl radical footprinting (HRF) have emerged as valuable structural biology techniques yielding information on protein tertiary structure. This data, however, is not sufficient to predict protein structure unambiguously, as it only provides information on the relative solvent exposure of certain residues. Despite some recent advances, no software currently exists that can utilize covalent labeling mass spectrometry data to predict protein tertiary structure. We have developed the first such tool, which incorporates mass spectrometry derived protection factors from HRF labeling as a new centroid score term for the Rosetta scoring function to improve the prediction of protein tertiary structure. We tested our method on a set of four soluble benchmark proteins with known crystal structures and either published HRF experimental results or internally acquired data. Using the HRF labeling data, we rescored large decoy sets of structures predicted with Rosetta for each of the four benchmark proteins. As a result, the model quality improved for all benchmark proteins, as compared to when scored with Rosetta alone. For two of the four proteins, we were even able to identify atomic resolution models with the addition of HRF data.

rosetta_protein_structure_prediction_from_hydroxyl_radical_protein_footprinting_mass_spectrometry_da

Introduction<!>Benchmark Dataset and Experimental Protection Factors<!>Rosetta ab initio Folding<!>Residue Exposure Metric<!>hrf_ms_labeling Score Term<!>Rescoring of Rosetta Structures<!>Model Evaluation<!>Generation of Control ab initio Model Set for Benchmark Proteins using Rosetta<!>Rescoring Model Sets using hrf_ms_labeling<!>Conclusion

<p>Historically, mass spectrometry has been used as a tool to quantify the mass and oligomeric distribution of proteins and protein assemblies.1, 2 More recently, advances have been made that allow mass spectrometry experiments to yield three-dimensional structural information on proteins and their complexes. By itself, there is no one mass spectrometry technique that can unambiguously elucidate atomic-resolution tertiary structure of a protein or protein complex. Hence, a combination of multiple different techniques is generally required.3–5 Several techniques have been particularly successful in probing the tertiary structure of proteins and their complexes. Hydrogen-deuterium exchange (HD/X) is based upon measuring the extent of isotopic exchange of backbone amide hydrogens.6, 7 Chemical cross-linking involves studying the structurally defined distances by covalently pairing functional groups within a protein.8, 9 Non-covalent interactions between lysine residues and 18-crown-6 ether (a cyclic organic compound) can provide lysine solvent accessibility within proteins.10 Finally, covalent labeling (sometimes referred to as "protein footprinting") involves exposing a protein in solution to a small labeling reagent that will covalently bond to select amino acid sidechains that are exposed to solvent, whereas sidechains buried within the core of the protein or occluded by interacting protein subunits will not get labeled.11–13 This provides information about the relative location of certain amino acids with respect to the solvent (either on the surface and solvent exposed or buried within the protein or protein complex structure). A variety of different labeling reagents exist and some are highly specific as to which amino acid(s) can react with the reagent and others have a much broader range of potential target residues. These techniques have been successfully employed with mass spectrometry to analyze protein structures.14–22</p><p>One covalent labeling method which recently has been increasingly widely used is hydroxyl radical footprinting (HRF).23, 24 This method involves exposing a solvated protein of interest to hydroxyl radicals generated from one of a variety of sources. Initially, oxidative labeling was performed using a synchrotron that ionized water to form the hydroxyl radicals.25 With recent advancements, a new method of hydroxyl radical labeling, fast photochemical oxidation of proteins (FPOP), has been developed.26, 27 With FPOP, a pulsed laser is used to photolyze hydrogen peroxide on a microsecond timescale, which is faster than the unfolding of a protein. This ensures that the labeling process does not disrupt the native state of the protein. In conjunction with mass spectrometry, FPOP provides important insight into the structure of proteins. This labeling method is quite broad in that it can label 19 of the 20 different amino acids, yielding extensive structural information. Despite the wealth of information provided by FPOP, the data itself is sparse, meaning that the solvent exposure information of a set of protein residues cannot provide unambiguous determination of protein structure. There remains a critical need for computational methods that can facilitate and compliment the structural interpretation of mass spectrometry FPOP labeling data.</p><p>Over the years, numerous experimental techniques have been successfully combined with computational methods to predict protein structures. Some examples of this are sparse experimental data from site-directed spin labeling electron paramagnetic resonance (SDSL-EPR) in conjunction with Rosetta to improve protein structure predictions,28, 29 nuclear magnetic resonance spectroscopy (NMR),30, 31 small-angle X-ray scattering (SAXS),32–35 and cyro-electron microscopy (cryo-EM).36–43 Mass spectrometry techniques have also been utilized in conjunction with computational methods. Malmström and coworkers have made significant contributions by incorporating data from MS chemical cross-linking experiments as inputs into computational methods for protein structure prediction.15, 44–47 The work of Sali and coworkers has contributed greatly to the field with the development of the Integrative Modeling Platform (IMP), an open source platform that integrates experimental data into computational methods.19, 35, 48–52 IMP is designed as a set of self-contained modules that can be mixed and matched based upon a user's preference. Models are generated and scored based upon spatial restraints that are derived from multiple sources of experimental data. Currently IMP supports the use of experimental data gathered from sources such as SAXS profiles, EM images and density maps, NMR, chemical cross linking, HD/X, and chromosome conformation capture. With IMP, both monomeric and multi-unit protein structures can be studied. Finally, Yang and coworkers have developed an integrative method, iSPOT, to determine protein-protein complexes that combines SAXS, hydroxyl radical footprinting, and computational docking of either rigid-body or molecular dynamics models.32</p><p>Computational modeling using FPOP data is still in its early stages. Recently, an integrated workflow was developed by Xie and coworkers that successfully demonstrated correlation between experimental high-resolution hydroxyl radical footprinting data and residue solvent exposure (as measured by absolute average solvent accessible surface area) as well as differentiated between low and high RMSD models for the soluble proteins myoglobin and lysozyme.53 This elegant work demonstrated that there is strong potential for successfully incorporating HRF or FPOP experimental data into computational methods in order to improve protein structure prediction. Despite the many advances and successes with using sparse data from various experimental methods for structure prediction, the use of covalent labeling mass spectrometry as the data source had yet to be accomplished.</p><p>In this work, we incorporated mass spectrometry derived protection factors from FPOP and synchrotron-based HRF labeling as a new score term for the Rosetta scoring function to improve the prediction of protein tertiary structure. Rosetta is one of the primary computational tools used for protein structure prediction.54 To accomplish our goal, we compiled a set of four soluble benchmark proteins with known crystal structures and either published HRF/FPOP experimental results or internally acquired data. We developed an efficient metric to quantify residue-specific burial that correlated linearly to the natural logarithm of experimental protection factors derived from the labeling rates. A new Rosetta centroid score term, that utilizes residue-resolved protection factors as inputs, was developed. This score term was used in conjunction with the standard Rosetta scoring function to rescore large decoy sets of predicted structures for each of the four benchmark proteins. In this process of rescoring, the quality of all models improved such that after rescoring the structures with the best score correlated more closely to the native structures. For two of the four proteins, we were even able to identify atomic resolution models using the HRF/FPOP data.</p><!><p>For this work, we used protection factor (PF) which was first described by Chance and coworkers and is derived from a labeling rate constant as a metric for residue labeling.55 PF is defined as the relative intrinsic reactivity of a given residue to hydroxyl radicals divided by the rate constant. The intrinsic reactivities of each amino acid type are well defined in the literature.24 The PF, as expressed on a natural logarithmic scale, has been shown to correlate with the solvent exposure of a given residue.16, 55, 56 Within the literature, the PF has been defined multiple ways, but for our purposes we have defined the protection factor for residue i, where Ri is the intrinsic reactivity for residue i and ki is the experimentally determined labeling rate constant, as defined by eq 1: (1)PFi=Riki</p><p>As a benchmark set, four different proteins with available FPOP or HRF labeling data were utilized. These proteins were calmodulin (PDB ID: 1PWR), myoglobin (PDB ID: 1DWR), lysozyme (PDB ID: 1DPX), and cytochrome c (PDB ID: 2B4Z). The experimentally determined PFs for calmodulin were extracted from the published work of Kaur et al who generated radicals via a millisecond timescale synchrotron radiation method.16 For myoglobin, the PFs were calculated from the reported labeling rate constants by Xie et al.53 using the reactivities reported in the literature.24 For this study, radicals were generated using sub-microsecond FPOP with a dosimeter to provide varying doses of radicals. Finally, the experimental PFs for both lysozyme and cytochrome c were oxidatively modified by FPOP at a single radical dose as described in the Supporting Information.</p><p>For incorporation of the data into the newly developed score term, input files were created for each protein consisting of a heading line followed by two columns comprising the residue number and the natural logarithm of the protection factor, with each labeled residue on a new line. FPOP/HRF can label 19 of the 20 amino acids, however data from the following residue types were omitted due to either too low/high reactivity or unclear products: M, C, D, N, Q, T, S, A, G, R, K, and V. Of this list of omitted residues, it has been previously suggested by Xie et al. that the sequence context plays a role in whether or not these amino acid types are labeled. This is a complex issue and has not been examined in this current work. As a result, only eight of the twenty amino acids were considered in the analysis: I, L, P, F, W, Y, E, and H. These residues have intermediate reactivities and correspond with the residue types utilized in similar studies.16, 53</p><!><p>In the absence of any experimental labeling data, decoy sets of 20,000 structures were generated for each of the four benchmark proteins using the AbinitioRelax application within Rosetta.57–59 The AbinitioRelax protocol consists of two main steps: 1) a coarse-grained fragment-based search of conformational space that uses a low-resolution "centroid"-based (treating each residue with backbone atoms defined explicitly and the side-chain represented as a single sphere) scoring function and 2) a high-resolution refinement using the full-atom Rosetta score function.</p><p>The generated decoy sets act as benchmarks to compare the structure prediction capabilities of Rosetta in the absence of FPOP/HRF labeling data. Specifics of the protocol have been detailed extensively in the literature.60 The fragment libraries for this work were generated using the Robetta online server.61 The required FASTA formatted sequences and native protein structures were extracted from each protein's respective PDB file. The fragment libraries, FASTA sequences, and native PDB structures (used solely for determining the deviation of the generated models from the native) were used as inputs for Rosetta's AbinitioRelax application. For lysozyme, disulfide bonds were present between the following residues: 6 and 127, 30 and 115, 64 and 80, and 76 and 94. An additional input file was provided to specify the residues that are a part of the disulfide bonds. The generated structures were scored using the Rosetta energy function (Ref15), where the score is an approximation of the energy of the protein or complex.62 The scores and respective root mean square deviation (RMSD) to the native crystal structure were extracted from the output score file. Structures were ranked based upon their scores with lower scores anticipated to correspond to models closer in structure to the native. Rosetta score versus RMSD to the native protein were generated to demonstrate this correlation.</p><p>For each of the benchmark proteins, two small sets of representative structures were generated. The first set represented ten native-like conformations of each protein which were obtained by relaxing each crystal PDB in the Rosetta force field using the relax application.63, 64 We will refer to these structures as the ten native-like models or the native-like model set. The second set contained models that scored well with the Rosetta energy function, but had high RMSDs compared to the crystal native structures. These were obtained by extracting the top ten scoring models with RMSD > 10 Å for each protein from the initial ab initio calculations. We will refer to these structures as the good scoring/high RMSD model set. Together, these sets represented the two extremes of potential models that we desired to efficiently differentiate between using our new score term.</p><!><p>To compare the protection factors extracted from the FPOP/HRF labeling data to residue exposure in protein models, a corresponding residue exposure measure was developed which enabled calculation of the level of exposure of every labeled residue in a protein model. The PF has been shown to correlate to a residue-level solvent accessible surface area (SASA).16, 53, 56 Because residue-level SASAs are expensive to calculate,65, 66 we explored other metrics, aside from SASA, that were less computationally expensive and provided even stronger correlation to the natural logarithm of the experimental FPOP/HRF PFs. Assuming solvent exposed residues are preferentially labeled, we sought to find a residue burial/exposure metric that showed correlation to the natural logarithm of the PFs. Several methods, such as weighted neighbor count and SASA,65, 67 were investigated. For reference, the correlation between SASA and the natural logarithm of the PFs can be found in Supplemental Figure S-1. However the burial measure found to give the strongest correlation to the experimental data was a neighbor count determined for each labeled residue. A residue with a high neighbor count can be thought of as buried whereas a residue with a low neighbor count can be considered solvent exposed. For this analysis, a low-resolution model of the protein was used where all of the backbone atoms were represented explicitly and the side-chain was represented as a single sphere called a centroid. To calculate a residue's neighbor count, the distances between the labeled residue's centroid (residue i) and all other residues' centroids (residues j ≠ i) were measured. The distance, rij, was then used in a sigmoid function that defined a value between 0 and 0.7, as shown in Supplemental Figure S-2, representing the amount of contribution of a neighboring residue j to the total neighbor count of the target residue i. The closer a residue j's centroid is to labeled residue i's centroid, the more it contributed to the overall neighbor count; conversely, the further away it is, the less it contributed. The total neighbor count for each labeled residue i was then defined as the sum of every residue's contribution to the neighbor count: (2)neighborcounti=∑j≠itotal#residues1.01.0+e0.1(rj-9.0)</p><p>We developed a new Rosetta application, burial_measure_centroid, which calculated the neighbor counts (as defined in eq 2) for arbitrary protein structures. For each of the eighty models comprising the native-like and good score/high RMSD model sets, the neighbor counts were calculated using the burial_measure_centroid Rosetta application. The neighbor counts for the ten native-like structures of calmodulin (1PRW) were used to perform a linear regression with the corresponding experimental lnPF values. The linear fit obtained was then used as a prediction function to predict the neighbor count for all 80 representative models with their respective experimental lnPF values as inputs.</p><!><p>A new score term, hrf_ms_labeling, was developed to be incorporated into Rosetta to assess the agreement of Rosetta models with experimental FPOP/HRF labeling data. This score term was defined as a centroid score term that rewards protein conformations that show agreement with the experimental labeling data. By treating the score term in a Bayesian fashion, the total Rosetta score was derived (as shown explicitly in the Supporting Information) to be the sum of the weighed score term and the current Rosetta score: (3)TotalScore=whrf∗hrf_ms_labeling+RosettaScore</p><p>The score term, hrf_ms_labeling, was implemented using the linear prediction function obtained by correlating the observed neighbor counts and experimental lnPF for the benchmark protein calmodulin (see the previous section, Residue Exposure Metric). A value for hrf_ms_labeling was calculated by summing the per-residue neighbor scores over the set of labeled residues and was defined as: (4)hrf_ms_labeling=∑i#labeledresidus-1.01.0+e2.0(∣diff∣i-7.5) where |diff|i is the absolute value of the difference between the observed neighbor count (calculated using eq 2 for the modeled protein) and the predicted neighbor count (calculated using the linear prediction function) for labeled residue i. Using the definition in eq 4, each labeled residue contributed a per-residue score ranging from −1 to 0 with a value of −1 in case of strong agreement with the experiment and a value of 0 in case of complete disagreement. If the value of |diff|i fell between 5 and 10 (which corresponded to the same cutoffs as the delta lines used in analyzing the prediction function), the residue received a logistically increasing value ranging from −1 to 0. The per-residue score (function found within the summation in eq 4) is depicted in Figure 1 with all relevant points highlighted.</p><!><p>To test the capability of our new score term to improve Rosetta model quality, the 20,000 Rosetta models initially generated as part of the ab initio folding for each benchmark protein were rescored with the hrf_ms_labeling score term. The calculated hrf_ms_labeling score was weighted by a value of 6.0 and added to the Rosetta score calculated using Rosetta's Ref15 energy function: (5)TotalRosettaScore=Ref15RosettaScore+6.0∗hef_ms_labeling</p><p>A weight of 6.0 was the lowest possible value that showed the greatest improvement. We iterated through all integer values from 1–36 and determined the top scoring models' RMSDs at each weight. The results of this analysis are shown in Supplemental Figure S-3. To calculate the hrf_ms_labeling contribution for each model, the score Rosetta application was run on each of the 80,000 models using the output structures from the initial ab initio model generation as input. For each of the 80,000 rescored models, the total Rosetta scores, the RMSD to the native structure, and the hrf_ms_labeling scores were extracted.</p><!><p>Several different metrics were used to evaluate the performance of both Rosetta and the score term. Those metrics were based upon the concept of an energy funnel, i.e. that within the overall energy landscape, low RMSD models can be distinguished from high RMSD models due to their lower energy (Rosetta score).68 The first metric used was a simple determination of the top scoring model's RMSD to the native structure. In practice, the Rosetta model with the lowest (most favorable) Rosetta score is assumed to be closest in structure to the native. Because all the benchmark proteins chosen for this study had crystal PDB structures available, an RMSD for that model can be calculated.</p><p>The second metric used was the goodness-of-energy-funnel metric Pnear, as defined by Bhardwaj et al.69 A value of Pnear was calculated for each Rosetta score versus RMSD distribution using the following equation: (6)Pnear=∑m=1Nexp(-rmsdm2λ2)exp(-EmkBT)∑m=1Nexp(-EmkBT) where N is the total number of models and Em and rmsdm are the Rosetta score and RMSD of model m. The parameter λ was given a value of 2.0 and controlled how similar a model needed to be to the native to be considered native-like. The final parameter, kBT, was set to 1.0 and governed the shallowness or depth of the funnel affects Pnear. Values of Pnear can range from 0 (very non-funnel like) to 1 (funnel-like).</p><p>The final metric used was a comparison of the number of top 100 scoring models with RMSD's below a 10.0 Å. By comparing this metric between different Rosetta score versus RMSD distributions we were able to investigate how well (or poorly) the addition of hrf_ms_labeling was at improving model quality.</p><!><p>To establish the baseline performance of Rosetta's Ref15 scoring function at predicting protein structures without any additional experimental knowledge, decoy sets consisting of 20,000 models were generated for each of four benchmark proteins. The four proteins selected for the benchmark were calmodulin (PDB ID: 1PWR), myoglobin (PDB ID: 1DWR), lysozyme (PDB ID: 1DPX), and cytochrome c (PDB ID: 2B4Z). Table 1 summarizes the benchmark proteins. These proteins ranged in size from 104 to 153 amino acids in length. Contact orders (CO) were calculated for each of the proteins.70 The contact orders for all four proteins were low, ranging from 10.7 to 13.7. The secondary structure content for the four proteins were relatively high, ranging from 41% to 74%. Because these proteins were all relatively small (approx. fewer than 150 amino acids), had high secondary structure content and low contact orders, we concluded that they were amendable to Rosetta ab initio protein structure prediction.</p><p>Using Rosetta to generate 20,000 models for each of the four proteins resulted in the selection of best-scoring structures with RMSDs ranging from 5.0 Å to 15.2 Å, as summarized in Table 2 and indicated on the Rosetta score versus RMSD to native structure plots in panel A of Figure 2 by stars. The two proteins with top scoring structures that were closest to their respective native structures were myoglobin (RMSD = 5.0 Å) and cytochrome c (RMSD = 5.5 Å). The predictions for the remaining two proteins, calmodulin and lysozyme, were poor, yielding top scoring models with RMSD's of 11.8 and 15.2 Å, respectively. Considering the size of the benchmark proteins, none of these best-scoring models were high-quality, near-atomic resolution models. For two of the proteins, even an incorrect topology was identified. However, as can be seen in Figure 2A, models with significantly lower RMSDs to the native structure were built for all four proteins. For calmodulin, the RMSDs for the generated models ranged from 2.9 Å to 21.5 Å. Similar ranges were sampled for cytochrome c and myoglobin, with RMSDs ranging from 1.4 Å to 21.3 Å and 1.5 Å to 27.3 Å, respectively. Lysozyme had the poorest sampling, where model RMSDs ranged from 6.0 Å to 18.7 Å. This indicated that better, and in some cases even near-atomic resolution models, were in fact generated for all proteins, but they were generally not identified by the lowest score.</p><p>The goodness-of-energy-funnel metric, Pnear, was used to evaluate the funnel quality of each of the distributions. As can be seen in Table 2, none of the distributions had Pnear values greater than 0.1, strongly suggesting that none of the ensembles of models exhibited funnel-like score distributions. This lack of a funnel in the Rosetta score versus RMSD to native structure plots made structure prediction and particularly native structure identification challenging. Based upon these ab initio structure prediction results, we concluded that incorporation of experimental data, such as HRF/FPOP labeling data, had the potential to improve identification of low RMSD models by score.</p><!><p>The overall goal of this work was to utilize experimental HRF/FPOP labeling data in order to improve models predicted by Rosetta. To accomplish this, a new Rosetta score term, hrf_ms_labeling, was developed that incorporated experimental HRF/FPOP protection factors (PFs). After developing hrf_ms_labeling, we confirmed that incorporation of HRF/FPOP labeling data did enable discrimination of near-native and high RMSD models and that combination of this score with the total Rosetta Ref15 score did improve the quality of the models selected from the structure ensembles.</p><p>The first step in this process was to demonstrate that a correlation existed between the experimental labeling data (the PFs) and a residue solvent exposure metric derived within Rosetta. The metric that demonstrated the best correlation was the per-residue neighbor count, as defined in the Methods section. The calculated neighbor count for every labeled residue within calmodulin (1PRW), one of our benchmark proteins, was plotted against the natural logarithm of the respective PF values. The positive correlation, as seen in Figure 3, had an R2 of 0.48 and p-value of 1.36E-36. The observed trend matched our expectation where residues with a low lnPF also showed a low neighbor count (suggesting a higher solvent exposure) and residues with a high lnPF showed a high neighbor count (suggesting a lower solvent exposure). The derived relationship between PFs and neighbor count was used to predict neighbor counts for all four benchmark proteins based on the experimental HRF/FPOP protection factors. For comparison, observed neighbor counts for two small sets of representative structures (the native-like model sets and the good scoring/high RMSD model sets) were calculated from each pdb structure using burial_measure_centroid. The predicted neighbor counts have been plotted against the observed neighbor counts (calculated directly from representative structures of the four benchmark proteins) in Figure 4. In order to quantify the accuracy of the prediction, two delta lines were defined (d1 = 5.0 and d2 = 10.0). These delta lines represent how close the predicted neighbor counts were to the actual observed values. Using the native-like model sets for all four proteins, an average of 81% and 59% of the labeled residues fell within d2 and d1, respectively, whereas only 67% and 38% of those belonging to the good scoring/high RMSD model sets did. This demonstrated that we predicted the majority of the labeled residues in native-like models within the delta lines and simultaneously excluded the majority of residues in the high RMSD models from within the delta lines. This suggested that agreement between a model's residue exposure and the neighbor count metric derived from experimental FPOP/HRF mass spectrometry data can indeed distinguish between low and high RMSD models and can thus be used to rescore protein models built in the absence of experimental FPOP/HRF labeling data. To be able to rescore protein models, a hrf_ms_labeling score term was developed for incorporation into Rosetta.</p><p>We next demonstrated that the new score term was effective in improving model prediction. The 20,000 model decoy sets generated for each of the four benchmark proteins were rescored with the hrf_ms_labeling term added to the Ref15 Rosetta score. For each set of models, Rosetta score + hrf_ms_labeling versus RMSD plots were generated. Based upon the rescored structures, new top scoring models were selected. As shown in Table 3, the RMSDs of the top scoring models improved for all four proteins, while for two of the proteins near-atomic resolution models were identified. The biggest increases in top scoring model quality were observed for lysozyme. Addition of HRF/FPOP labeling data improved the RMSD of the top scoring lysozyme model from 15.2 Å to 7.2 Å, a significant improvement in the model's quality. Although a model with an RMSD of 7.2 Å is not usually considered high quality, considering that the best lysozyme ab initio model had an RMSD of 6.0 Å, one of the best existing models was identified. Both myoglobin and cytochrome c showed decreases in their RMSDs to near-atomic resolution models (2.2 and 1.8 Å respectively), also identifying models with RMSDs close to the best existing models within the 20,000 structures. Calmodulin had the least improvement with a change in RMSD from only 11.8 to 10.2 Å. When superimposing the top scoring models onto their respective native structures, as depicted in panels B and D of Figure 2, a significant increase in model quality could be observed as a result of the addition of hrf_ms_labeling. All top scoring models now identify the correct protein topology.</p><p>In addition to analyzing the RMSD of the top scoring models, the overall energy landscape of the structures was analyzed. Values of Pnear were calculated for each score versus RMSD distribution, identical to what was done without the addition of hrf_ms_labeling (see Table 2). With the addition of the hrf_ms_labeling term to the scoring function, there was an increase in Pnear, i.e. an increase in funnel quality of the score vs RMSD plots, for all four proteins. As can be seen in panel C of Figure 2, the distributions appear more funnel like with lower RMSD models receiving lower scores. Interestingly, the Pnear values of the two proteins for which near-atomic resolution models were identified (myoglobin and cytochrome c) were several orders of magnitude higher than those of the other proteins. We thus speculated that Pnear might be used as a confidence measure to identify cases for which near-atomic resolution models were identified. To explore this idea we recalculated score vs RMSD plots with respect to the lowest scoring structure (to obviate the necessity for knowledge of the native structure) and measured Pnear values for these distributions as shown in the last column of Table 2. While the trend was not as pronounced as before, this Pnear value still served as a confidence measure in that the Pnear values of the two proteins for myoglobin and cytochrome c were more than two orders of magnitude higher than those of the other proteins. Upon rescoring with hrf_ms_labeling, the overall distribution of structures did not shift to a lower RMSD, because hrf_ms_labeling was simply used to rescore previously generated models. Plots of hrf_ms_labeling versus RMSD are shown in Supplemental Figure S-4. For all four proteins, models with poor (i.e. high, closer to 0) hrf_ms_labeling scores also had a higher RMSD. Likewise, some of the models with a better hrf_ms_labeling score tended to have a lower RMSD. There were a fair number of models however that had good hrf_ms_labeling scores but a high RMSD. This trend is not concerning, because the information obtained from the HRF/FPOP labeling experiments are not all encompassing of a proteins structure. Individual score terms within Rosetta generally do not exhibit the exact trend of low score/low RMSD and high score/high RMSD. Combination of this score term with the Rosetta scoring function however generated the desired trend.</p><p>We finally investigated whether a larger set of top scoring models after the rescoring were of increased quality. Histograms were generated showing the RMSD frequency of the top 100 scoring models for the distributions pre- and post-addition of hrf_ms_labeling. Based upon these histograms shown in Figure 5, there was a definite shift in the model quality for calmodulin and myoglobin, with more models scoring well with low RMSDs. The percentage of the top 100 scoring models that had a RMSD < 10 Å increased from 35% to 68% for calmodulin with the addition of hrf_ms_labeling. This illustrates that despite not identifying a near-atomic resolution model for calmodulin, addition of the labeling information significantly improved the model quality. Myoglobin demonstrated an increase in the percentage of models in the top scoring 100 with RMSD < 5 Å from 47% to 70%. A shift in model quality of the top 100 scoring models was also seen with for lysozyme and cytochrome c, albeit much less significant.</p><p>The hrf_ms_labeling score term has shown great success in rescoring structures based on experimental HRF/FPOP labeling data and has been designed efficiently. A centroid form of the score term was chosen for two reasons. First, this implementation showed the highest correlation between the centroid based neighbor count and experimental lnPFs. Second, a centroid-based score function is crucial in predicting structures within Rosetta's AbinitioRelax protocol. Within this protocol, the main sampling of conformational space occurs during the centroid scoring phase. Thus hrf_ms_labeling would have maximal impact on predicting structures ab initio if it was utilized during the centroid scoring phase. Future work will focus on developing these ab initio capabilities.</p><!><p>In this work, a new Rosetta score term, hrf_ms_labeling, was developed. This score term utilizes residue-resolved protection factors from hydroxyl radical labeling (HRF/FPOP) mass spectrometry data and assesses agreement of protein model with the experimental data. Four proteins (calmodulin, cytochrome c, myoglobin, and lysozyme) which had both available experimental data and known crystal structures were used to benchmark the performance of the score term. Using the linear correlation between the natural logarithm of the experimental protection factors and calculated neighbor counts for one of the benchmark proteins, calmodulin, a prediction function was generated to predict the neighbor counts for the other proteins using their respective lnPFs. This prediction function was used as the basis of the new score term hrf_ms_labeling. The new score term was used to rescore sets of 20,000 models for each protein generated using Rosetta's AbinitioRelax application. As a result, the top scoring model increased in quality for all four proteins. The method used for radical generation did not adversely affect the modeling. For two of the four proteins, we were even able to identify atomic resolution models using the HRF/FPOP data. In addition, the overall distribution of models moved more towards a funnel-like energy landscape, indicating that good scoring models were closer in structure to their respective natives. Finally, we were able to identify a confidence measure that has the potential to identify successful models without having to know the native structure.</p><p>To our knowledge, we are reporting the first method to incorporate experimental HFR/FPOP labeling data in protein structure prediction. This marks an important first step in fully utilizing mass-spectrometry-based covalent labeling techniques in quantitative structure predictions, rather than just qualitative explanations. By demonstrating the potential of covalent labeling in conjunction with the protein structure prediction capabilities of Rosetta, these techniques will be elevated to be comparable in utility to other structural biology techniques such as EPR or FRET. The scoring term and applications discussed in this paper are freely available and easily accessible through Rosetta. We have added a tutorial, including a summary of necessary files and command lines to the supporting information.</p><p>Future work will focus on extending this methodology to other labeling techniques. While this particular scoring term is specific to HRF, we plan to implement the capability to use labeling data from other mass-spectrometry-based covalent labeling experiments in the future. A second direction of our future efforts will be to develop covalent labeling-guided ab initio structure prediction, where the labeling data is used as part of the actual structure generation as opposed to rescoring structures generated in the absence of the experimental data.</p>

PubMed Author Manuscript

Behavior of Hydrated Lipid Bilayers at Cryogenic Temperatures

Neutron diffraction was used to study the behavior of water present in phospholipid multilamellar stacks from 1,2-dimyristoyl-sn-glycero-3-phosphatidylcholine (DMPC) at cryogenic temperatures. Evidence was found for the existence of a highly viscous phase of water that exists between 180 and 220 K based on the observation that water can leave the intermembrane space at these low temperatures. Similar measurements are described in the literature for purple membrane (PM) samples. From a comparison with results from this natural membrane by using the same flash-cooling protocol, it is found that in the case of pure lipid samples, less water is trapped and the water flows out at lower temperatures. This suggests that the water is less hindered in its movements than in the PM case. It is shown that at least the Lβ′-phase of DMPC can be trapped likely by flash cooling; upon heating to about 260 K, it transforms to another phase that was not fully characterized.

behavior_of_hydrated_lipid_bilayers_at_cryogenic_temperatures

Introduction<!>Sample Preparation for Neutron Diffraction<!>Neutron Diffraction, Data Processing, and Analysis<!><!>Lamellar Spacing of the Lipid Stack as a Function of Temperature<!>Ice Formation in the Interlamellar Space as a Function of Temperature<!>Changes in the Lipid Phase as a Function of Temperature<!>Discussion<!>Conclusions<!>Data Availability Statement<!>Author Contributions<!>Conflict of Interest

<p>The liquid and gel phases of phospholipids and phosphatidylcholines (PCs) have received much attention since the work of (Chapman and Quinn, 1976). Four lamellar phases have been recognized in saturated PCs: a liquid–crystalline phase, Lα, and phases with ordered hydrocarbon chain arrangements, designated ripple phase, Pβ′; gel phase, Lβ′; and "subgel" or "crystal" phase, Lc.</p><p>In the liquid Lα phase, molecules exhibit rotational and lateral diffusion; the transition to the more ordered gel phase is accompanied by a dramatic loss of lateral diffusion and restriction of a variety of bond motions. In this phase, chains are liquid-like and do not present a regular in-plane structure. The Pβ′ phase has been shown to form a periodic surface ripple and exists for all saturated PCs at water contents higher than 20%, in a temperature range just below the main phase transition temperature, Tc (Janiak et al., 1976). In this phase, chains are frozen and tilted with wavelength modulation with a period of ~120–200 Å. In the Lβ′ phase, the chains form a closely packed lattice in which the bilayer planes are flat and tilted with respect to the normal. In the case of bilayers with saturated chains of 16 carbon atoms [1,2-dipalmitoyl-sn-glycero-3-phosphatidylcholine (DPPC)], two types of crystal phase have been characterized in which the acyl chains are packed in different subcells, and each phase exhibits a different thermal stability (Ruocco and Shipley, 1982). Differential scanning calorimetry (DSC) heating curves of hydrated DPPC following rapid cooling from 20°C and equilibration at −2°C for increasing periods of time are reported. Immediate reheating shows a small but detectable sub-transition endotherm with transition onset at Tc = −1.6°C. From diffraction and DSC measurements, the authors conclude that that the Lβ′-Lc conversion involves dehydration and hydrocarbon chain ordering (Ruocco and Shipley, 1982).</p><p>Neutron and X-ray diffraction from stacks of lipid bilayers, that is, lipid multilayers, are capable of providing key structural information on lipid membranes (Büldt et al., 1979; Zaccai and Gilmore, 1979; Nagle and Tristram-Nagle, 2000; Kučerka et al., 2008). In these experiments, the diffraction of neutrons and X-rays from aligned lipid multilayers deposited on a solid substrate is used to determine the number and nature of the lipid phases present in the sample as well as structural information about the membrane thickness and internal structure (Foglia et al., 2015). In neutron scattering experiments, as well as some NMR or IR spectroscopy experiments, sample deuteration enables highlighting specific membrane components. In neutron scattering experiments, it can also improve the signal-to-noise ratio (Pabst et al., 2010; Luchini et al., 2018).</p><p>1,2-Dimyristoyl-sn-glycero-3-phosphatidylcholine (DMPC) has been extensively used with neutron techniques as a model lipid for structural and dynamical bilayer studies, the reason being that it is in the fluid phase at room temperature and the deuterated form is easily obtainable in large quantities. The lipid presents a gel-to-gel pre-transition (Lβ′ > Pβ′) at 284.15 K and a gel-to-liquid crystalline main acyl chain crystallization transition (Pβ′ > Lα) at 296.5 K upon temperature increase (Janiak et al., 1979). In heavy water, D2O, the transition temperatures are slightly different, being 288.15 K for Lβ′ > Pβ′ and 297.15 K for Pβ′ > Lα (Faure et al., 1997). The d-spacing of fully hydrated DMPC samples as determined by X-ray diffraction is reported to be 59.9 Å (Tristram-Nagle et al., 2002). At a relative humidity content <100%, Smith and co-workers (Smith et al., 1990) found that the Lβ′ phase consists in fact of three distinct two-dimensional phases differentiated by the direction of chain tilt with respect to the in-plane lattice. Janiak et al. (1979) discussed also a lamellar structure occurring at low temperature (<0°C) and water content (5–20% w/w) with a similar diffraction pattern and lamellar repeat distance to those found at higher temperatures, suggesting that no major structural alterations occur at the transition from the Lβ′ to this low-temperature phase. DMPC crystallizes from water-containing solution with two water molecules (5% w/w) of hydration, and data have been collected at 10–15°C (Pascher and Pearson, 1979). The crystals are monoclinic (space group P21; unit cell parameters a = 8.72, b = 8.92, c = 55.4 Å and β = 97.40°) with two molecules in the asymmetric unit.</p><p>A study of the literature on phospholipid phases at low temperatures indicates a strong dependence on the conditions of hydration and temperature of equilibration of the bilayers so that various phases have been identified like the Pcc phase (so called according to its characteristic convex–concave bilayer curvature observed by freeze-fracture electron microscopy) appearing in DMPC and DPPC samples, when hydration and storage occur at 4°C (Meyer et al., 2000); this phase is not observed when hydration occurs at 50°C and storage at 4°C.</p><p>Kiselev et al. (2000) have studied, by small-angle X-ray scattering (SAXS), and wide-angle X-ray scattering (WAXS), ice formation in the presence of cryo-protectants, in model biological membranes formed by phosphocholine lipids including DMPC. Ice formation in fully hydrated DMPC samples occurred at 255.6 K, leading to a decrease in the membrane repeat distance from 59.8 to 53.5 Å. For samples with different degrees of hydration, calorimetric studies (Grabielle-Madelmont and Perron, 1983) allowed the determination of the temperature dependence of the freezing of water and the melting of ice for different water contents in DPPC–water mixtures. The lower the hydration, the lower the temperatures of freezing and melting. More generally, lipid and water dynamics in lipid stacks at subzero temperatures have been studied by temperature-modulated DSC (Svanberg et al., 2009), anelastic spectroscopy (Castellano et al., 2006), Raman scattering (Surovtsev and Dzuba, 2009), and neutron spectroscopy (Swenson et al., 2008). Here, we present a temperature-dependent neutron diffraction study on flash-cooled hydrated DMPC stacks at subzero temperatures and infer about characteristic changes in water and lipid dynamics.</p><!><p>The sample was prepared from DMPC lipid powder purchased from Avanti Polar Lipids (Alabaster, AL, USA), used without further purification. In 6 ml trifluoroethanol and 2 ml chloroform was dissolved 400 mg of the powder, so that a clear homogenous solution was obtained, which was kept overnight in a freezer at −20°C. The solution was carefully placed into the base of a 4 × 3 cm2 flat aluminum sample holder with a pipette, under a glass bowl as protection against dust and other impurities, and left to stand until the solvent was completely evaporated. The evaporation of the solvent and the flat surface of the sample holder favor the formation of multilamellar bilayer stacks (Nagle and Tristram-Nagle, 2000). The sample was then placed in a desiccator and dehydrated under vacuum for 12 h. Finally, it was equilibrated in D2O vapor for several days. After equilibration, the sample was tightly sealed with an indium wire placed between the base and the lid, which was then screwed shut. The sample displayed a very high mosaicity, which means that there is not just one single stack of bilayers, all with their bilayer-normal oriented in the same direction, but a number of stacks with an angle of up to 15° between them. This is due to the way in which the sample was prepared.</p><!><p>Neutron diffraction experiments were carried out in 2003 on the D16 diffractometer (Leonard et al., 2001; Cristiglio et al., 2015) at the Institut Laue-Langevin (Grenoble, France), with a neutron wavelength λ = 4.53 Å (1% spread, FWHM). The diffracted neutron beam is detected with a square Multi-Wire-Proportional-Chamber 3He detector with a resolution of 2 mm in the horizontal and vertical directions. With the detector placed at 1 m from the sample, this corresponds to a resolution of 0.115° (0.0028 Å−1 in Q, at a wavelength of 4.53 Å) and an angle of 14.7° for the whole detector. Measurements were carried out with a detector–sample distance of 1 m. To ensure that all measurements on one sample were made in the same orientation, the samples were aligned using the rocking-curve method, every time the sample was remounted, for example, after each flash cooling. First, the sample is placed approximately orthogonal to the neutron beam. Then a simple ω-scan, where ω is the sample angle from ω = 70° to 110°, is carried out by observing the intensity of the first lamellar peak, at a detector angle of γ = 5°, corresponding to the lamellar spacing d of the membrane stack. This curve is ideally reminiscent of a Gaussian broken by two characteristic minima due to high absorption where the sample is oriented parallel to the incident or diffracted beam. We choose the first minimum to set ω = 90° and measured the position of the lamellar peak at ω = 92°, where the intensity of the rocking curve is maximum. The 2D images were integrated and plotted vs. Q, the scattering vector perpendicular to the membrane stacks [Q = (4π/λ)sinθ with 2θ being the scattering angle]. Temperature control was carried out in an "Orange" ILL Cryostat on D16 (Cristiglio et al., 2015). Three temperature-dependent data collection protocols were applied sequentially in the following order: (1) slow heating after flash cooling (SHFC protocol), (2) slow cooling (SC protocol), and (3) slow heating (SH protocol).</p><p>In the SHFC protocol, the sample was equilibrated for several hours in a cold room at 8°C in order to ensure that the lipids were in the Lβ′ phase. Flash cooling was then achieved by plunging the sample holder into liquid nitrogen in the cold room. Subsequently, the sample holder was transported in liquid nitrogen to D16 and rapidly transferred into the cryostat that had been precooled to 100 K. The temperature was raised from 100 to about 300 K in 5 K steps and was kept constant after each step.</p><p>In the SC protocol, the temperature was lowered from 300 to 100 K in 5 K steps and was kept constant after each step. In the SH protocol, the temperature was raised from 100 to 300 K in 5 K steps and was kept constant after each step. The time interval between successive data points was 30 min.</p><p>Neutron diffraction data were recorded during the constant-temperature phases. The sample was weighted before and after each neutron diffraction experiment in order to ascertain that the sample was properly sealed and did not lose hydration water in the vacuum of the cryostat during diffraction experiments.</p><p>The lamellar spacing, d, was determined from the position of the first-order Bragg peak of the lipid stack (Figure 1A) in an experimental and reversible series consisting of sequential executions of the SHFC, SC, and SH protocols (Figures 2, 3). To this end, the sample had been aligned with an angle of 92° between the normal of the sample and the incoming neutron beam and data recorded for 10 min at each temperature plateau. In a second series of sequential executions of the SHFC and SC protocols, a diffraction peak (called lipid peak hereafter; Figure 1B) originating from the intra-planar molecular structure of the phase Lβ′ of the lipids (scattering vector Q = 1.49 Å−1 at 300 K) and ice formation (Figure 1C) were monitored simultaneously. Transmission geometry was used for these measurements. The sample was aligned with an angle of 32.5° and 35.5° between the normal of the sample and the incoming neutron beam to measure the lipid peak and the ice peaks, respectively, and data were recorded for 7 min at each angle and temperature plateau. Ice formation was monitored (Figure 4) by integrating the intensities of part of the most prominent powder ring, which corresponds to spacing of 3.91 Å [Q = 1.61 Å−1; (1 0 0) reflection of hexagonal ice; called ice peak 1 hereafter] and 3.67 Å [Q = 1.71 Å−1; originating from amorphous ice, the (1 1 1) reflection of cubic ice, or the (0 0 2) reflection of hexagonal ice; called ice peak 2 hereafter; Figure 1C]. Figure 5 shows the scattering vector Q of the lipid peak (Figure 1B) as a function of temperature.</p><!><p>Neutron diffraction intensity of hydrated DMPC lipid stacks at 260 K as a function of the scattering vector Q in three different Q-ranges. Diffraction peaks originate from (A) the lamellar spacing, (B) in-plane lipid ordering, and (C) crystalline water ice. In (C), the peaks at Q = 1.61 Å−1 and Q = 1.71 Å−1 are called ice peak 1 and ice peak 2, respectively. Diffraction intensities in all three q-ranges have been recorded after flash cooling upon heating from 100 to 300 K (SHFC protocol). Error bars in (B,C) correspond to sqrt(N(x, y))/t, where N(x, y) are the total counts in pixels x, y of the detector and t is the acquisition time. In (A), the error bars are too small to be seen.</p><p>Lamellar spacing of stacks of hydrated DMPC lipids as a function of temperature as determined by neutron diffraction on D16. Closed black circles show the lamellar spacing after flash cooling upon heating from 100 to 300 K (SHFC protocol), closed gray circles represent the ones during subsequent slow cooling from 300 to 100 K (SC protocol), and open circles show the ones during subsequent slow heating from 100 to 300 K (SH protocol). The time interval between successive data points was 30 min. The experimental sequence of heating and cooling steps started with flash cooling, followed by SHFC, then SC, and finally SH.</p><p>Lamellar spacing of stacks of hydrated DMPC lipids as a function of temperature as determined by neutron diffraction on D16 after flash cooling upon heating from 100 to 300 K (SHFC protocol). Closed circles correspond to the data presented in Figure 2, open triangles to those resulting from a repeated experiment carried out with the same sample. The similarity of the two curve profiles informs about the reproducibility of the experiment.</p><p>Integrated intensity of the diffraction peaks at Q = 1.61 Å−1 (ice peak 1, circles) and Q = 1.71 Å−1 (ice peak 2, diamonds) after flash cooling upon heating from 100 to 300 K (SHFC protocol, black data points) and during subsequent slow cooling from 300 to 100 K (SC protocol, gray data points). The time interval between successive data points was 30 min.</p><p>Scattering vector Q of the lipid peak as a function of temperature during the SHFC protocol (black circles) and the subsequent SC protocol (gray circles).</p><!><p>After flash cooling, at 100 K, the lamellar spacing d is 50.2 Å. In the literature, the bilayer thickness is given to be 44.5 Å in the gel phase (Janiak et al., 1976) and 39 Å in the fluid phase (Perino-Gallice et al., 2002). A lipid contraction of about 1–2% for membranes between 300 K and liquid nitrogen temperature may be deduced from the literature (Weik et al., 2005; Mehra et al., 2020) which, in our case, leads to a bilayer thickness of 43.8 ± 0.2 Å and therefore a water layer with a thickness of 6.4 ± 0.2 Å separating adjacent bilayers in the stack. The reason for using values from the literature for the bilayer size is that it is not possible to perform contrast variation studies on the same sample in these experiments that would allow the mathematical calculation of the bilayer thickness. Furthermore, at least three lamellar peaks would be needed for such a determination (Leonard et al., 2001), and only two were observed in this work.</p><p>Upon heating (SHFC protocol) to 180 K, d increases to 50.5 Å and then decreases to 49.8 Å at 220 K. Further heating leads again to an increase in d, with a plateau between 255 and 260 K, a maximum of 55.7 Å at 280 K, and a sudden drop at 290 K. During the following slow cooling (SC protocol), the evolution of d globally follows the one during the preceding SHFC protocol down to 265 K and then continues to decrease to reach a d value of 48.7 Å at 100 K. During subsequent heating (SH protocol) to 300 K, d evolves almost identically to the preceding SC protocol with the exception of a hysteric behavior between 270 and 280 K.</p><!><p>After flash cooling, at 100 K, the absence of ice peak 1 and the presence of ice peak 2 indicate the absence of hexagonal and the presence of cubic and/or amorphous ice, respectively. Upon heating (SHFC protocol), the integrated intensity of ice peak 2 remains more or less constant up to 190 K (Figure 4). Between 190 and 240 K, the integrated intensity of ice peak 2 increases, and ice peak 1 appears and increases in intensity. Upon further heating from 240 to 300 K, the integrated intensity of ice peak 2 decreases to zero, and that of ice peak 1 continues to increase, reaches a maximum at 260 K, and then decreases to zero. Upon subsequent slow cooling (SC protocol), ice peak 1 appears and increases in intensity at a high rate between 270 and 250 K and at a lower rate between 250 and 220 K and then fluctuates upon further cooling to 100 K. Ice peak 2 appears and increases in intensity in parallel to ice peak 1 down to 245 K and then slightly decreases in intensity during further cooling to 100 K. When comparing results from the SHFC and the SC protocols, it is evident that ice peak 1 intensities are higher than ice peak 2 intensities during cooling and lower during heating.</p><!><p>After flash cooling, the position of the lipid peak (Figure 1B) slightly shifts to lower Q values during heating from 100 to 240 K (SHFC protocol, Figure 5). A sudden shift to even lower Q values is observed between 240 and 265 K, a temperature interval that includes the plateau observed in the evolution of the lamellar spacing d in the SHFC experiment (Figure 2). Upon further heating to 290 K, the lipid peak position shifts to higher Q values (Figure 5). During the subsequent slow cooling (SC protocol), the shift in the lipid peak position follows down to 260 K that of the previous SHFC experiment. Upon further cooling from 260 to 100 K, the Q value of the lipid peak position increases again and reaches 100 K, a similar value to that after flash cooling.</p><!><p>Based on their temperature-dependent behavior and their position, the three observed Bragg peaks at high angles (for all measurements) were attributed to the Lβ-phase of the lipid (Figure 1B) and to the hexagonal and the cubic forms of the ice in the sample (Figure 1C). The growth of hexagonal ice from cubic ice could be observed starting from around 230 K where ice peak 1 grows considerably, while ice peak 2 remains constant.</p><p>Based on a value of 44.5 Å for the bilayer thickness (Janiak et al., 1976), we can calculate an approximate value for the thickness of the water layer by subtracting this value from the d-spacing. For 100 K after SC, we have about 5.7 Å, which corresponds to roughly two layers of water (Lechner et al., 1998). These two layers are the first hydration layers of the lipids that do not crystallize at subzero temperatures (Castellano et al., 2006). Because of the high affinity of the water molecules to the lipid head groups, the diffusion of water in these layers is considerably slowed down. At 280 K, the water layer thickness is ≈12 Å.</p><p>The increase in d-spacing when heating from 100 to 180 K could have been caused by an increase in bilayer thickness or an increase in the volume taken up by the water between the bilayers or by both. We observe no change in the position of ice peak 2 and only a slight change in the lipid peak position. It is thus reasonable to assume that the lipid bilayer is the cause for the increase and that the water density remains constant. It has been reported that the volume per lipid increases with rising temperature in a linear manner within one phase (Nagle and Tristram-Nagle, 2000). Even though the reported measurements were carried out above 273 K, there is no reason to suspect a different behavior at cryo-temperatures.</p><p>Now we turn to the decrease in d-spacing that we observe in the temperature window from 180 to 220 K on heating a flash-cooled sample (Figure 2), the central part of our study. Several scenarios are possible to explain the decrease.</p><p>First, the lipid bilayer undergoes a transition that diminishes its thickness. Secondly, the water density increases, without any change in the state of the lipids. Thirdly, neither lipid bilayer nor the water density changes, and a portion of the water leaves the inter-bilayer space. From the raw diffraction data, it is observed that in the range 180–220 K, the lipid peak (Figure 1B) does not change abruptly, neither in shape nor in intensity and, most importantly, not in position. Since the lipid peak position and the area per lipid molecule and the bilayer thickness are correlated, we conclude that the bilayer does not become thinner and the first scenario can be discarded. The second seems very unlikely, since we observe no change in the position of the ice peaks that would be expected if ice density increased. In addition, such a change would correspond to a transition to another non-equilibrium state on heating. This is not likely, since the molecules' ability to return to equilibrium is increased when increasing thermal energy. The third scenario is supported by the observation of a growth in ice peak intensity (Figure 4). It is widely accepted that crystalline ice does not form within the lipid bilayers (Gleeson et al., 1994); therefore, the observed increase in ice peak intensity must be due to ice forming outside of the membranes. All water not being confined by the membranes must surely be frozen at these temperatures, so the water forming the newly observed crystalline ice is likely to be the interlamellar water that has started to leave the bilayers at 180 K. The water leaving the bilayer space at 180 K is reminiscent of the ultra-viscous water in the no-man's land reported earlier (Mishima and Stanley, 1998). We note that the glass transition of water confined in dioleoylphosphatidylcholine (DOPC) membranes has been reported to take place at a similar temperature, that is, at 170 K (Castellano et al., 2006). A decrease in d-spacing on heating a flash-cooled sample to above 180 K has occurred reproducibly in the same sample in three different experiments (see, for example, Figure 3). A comparison with the decrease in d observed for a similar experiment with purple membranes (PMs) (Weik et al., 2005) shows that the decrease in d is considerably lower here, which suggests that by using the same flash-cooling procedure, that is, the same cooling rate, a significantly lower amount of water can be trapped in the process. This, together with the fact that the water in DMPC leaves the intermembrane space at a temperature (180 K) that is 20 K lower than in PM (200 K), suggests that the interlamellar water is less hindered in its movements by the pure-lipid bilayers than by the PM. Water flowing out of the intermembrane space above 180 K (Figure 2), forming crystalline ice (Figure 4) is reminiscent of cold crystallization (Tanaka et al., 2000) of intermediate water (Tanaka et al., 2015) associated with polymers.</p><p>Coinciding with the increase of d above 230 K (Figure 2), a dramatic decrease in the intensity of ice peak 2 is observed (Figure 4). We interpret this decrease as melting of cubic ice that most probably had formed during flash cooling outside of the interlamellar space (Weik et al., 2005). At the same time that the intensity of ice peak 2 decreases, that of ice peak 1 increases (Figure 4), indicating the formation of hexagonal ice. Above 230 K, we thus suggest that the melting cubic ice partially recrystallizes into hexagonal ice outside the membrane stack and partially rehydrates the membrane stack, leading to the increase in d (Figure 2). In contrast to the decrease in d from 180 to 220 K, the plateau from 250 to 265 K is accompanied by a drastic change in the position of the in-plane lipid peak (Figure 5). This suggests that a transition Lβ′-phase appears to be trapped by flash cooling since the position in Q of the lipid peak is similar before and after flash cooling (see Figure 5). The trapped Lβ′-phase then probably transforms into another lipid phase at around 260 K. The fact that this plateau occurs only when examining a flash-cooled sample could mean that in addition to the water, a lipid phase was trapped during flash cooling that normally does not exist at these temperatures. In this interpretation, the plateau and the subsequent identical evolution of d in SHFC and SH experiments mean the lipid phase trapped by flash cooling returns to equilibrium at around 260 K. The temperatures where trapped water (180 K) and trapped lipids (260 K) gain mobility are different, in line with observations in single supported DMPC bilayers that suggested a decoupling of hydration water and membrane freezing (Toppozini et al., 2012).</p><p>Regarding the SH experiment (Figure 2), the increase in d during heating from 100 to 225 K is again likely to be due to a change in the lipid organization. Above around 230 K, d increases at a higher rate, again suggesting membrane rehydration. The hysteresis in d observed when comparing SH and SC experiments is very similar to the one observed by Gleeson et al. (1994), indicating the presence of super-cooled water during SC.</p><p>In order to rationalize the draining of water from the intermembrane space upon SC and water influx upon heating, one may consider that the chemical potential of water confined by the bilayers depends strongly on the thickness of the water layer (Gleeson et al., 1994). When the thickness of the water layer dw is smaller than 10 Å, which is the case at subzero temperatures considering a DMPC bilayer thickness of around 44.5 Å (Janiak et al., 1976) (cf. Figure 2), this dependence is well-approximated by an exponential law so that the difference of the chemical potential is Δμ = C * exp(–df/Λ), where C is a scale factor and Λ the hydration repulsion decay length (Gleeson et al., 1994). The chemical potential of the sample is given by the known chemical potential μT of ice existing in equilibrium after formation in pockets and defects of the sample when the temperature is about 265–270 K and is then imposed on the interlamellar water. μT changes with temperature and bilayer thickness changes, which means water flows in or out of the membrane, so that μ inside and outside of the membranes match again.</p><p>In the SHFC experiment, the sample is in a metastable state after flash cooling. The water in the flash-cooled system did not form hexagonal ice. From the fact that the intensity of ice peak 2 in SHFC is almost double than that of ice peak 2 in the SC experiment, we can infer that a high amount of cubic ice has been formed during flash cooling. Most likely, all ice present at 100 K after flash cooling is cubic ice, present in defects where it is confined enough to have not been able to form hexagonal ice. The (1 1 1) reflection of cubic ice is of a higher order than the (1 0 0) reflection from hexagonal ice. This might explain why the intensity of ice peak 2 in SHFC is much lower than that of ice peak 1 in SC, even though both peaks represent about the same amount of ice.</p><!><p>We studied the behavior of a bilayer stack of synthetic phospholipids (DMPC) in a hydrated environment, at cryo-temperatures, and present evidence for the existence of a highly viscous phase of water above 180 K in flash-cooled samples. In comparison with PMs (Weik et al., 2005), less water is trapped using the same flash-cooling protocol, and the water flows out at lower temperatures, which suggests that the water is less hindered in its movements than in PMs. Our results suggest it is likely that at least the Lβ′-phase of DMPC can be trapped by flash cooling and that it transforms at around 260 K to a different phase, which was not fully characterized.</p><!><p>All datasets generated for this study are included in the article/supplementary material.</p><!><p>JM, GF, and MW designed and performed the experiments. JM and MW did the data analysis. GF, GZ, and MW wrote the manuscript. All authors contributed to the article and approved the submitted version.</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

Designing interchain and intrachain properties of conjugated polymers for latent optical information encoding

Building molecular-design insights for controlling both the intrachain and the interchain properties of conjugated polymers (CPs) is essential to determine their characteristics and to optimize their performance in applications. However, most CP designs have focused on the conjugated main chain to control the intrachain properties, while the design of side chains is usually used to render CPs soluble, even though the side chains critically affect the interchain packing. Here, we present a straightforward and effective design strategy for modifying the optical and electrochemical properties of diketopyrrolopyrrole-based CPs by controlling both the intrachain and interchain properties in a single system. The synthesized polymers, P1, P2 and P3, show almost identical optical absorption spectra in solution, manifesting essentially the same intrachain properties of the three CPs having restricted effective conjugation along the main chain.However, the absorption spectra of CP films are gradually tuned by controlling the interchain packing through the side-chain design. Based on the tailored optical properties, we demonstrate the encoding of latent optical information utilizing the CPs as security inks on a silica substrate, which reveals and conceals hidden information upon the reversible aggregation/deaggregation of CPs.

designing_interchain_and_intrachain_properties_of_conjugated_polymers_for_latent_optical_information

Introduction<!>Results and discussion<!>Conclusions

<p>Conjugated polymers (CPs) are key materials for various optoelectronic applications, including chemical and bio sensors [1][2][3][4][5][6][7][8][9] that detect from ions, biomolecules to dangerous chemical gases, and security inks [10][11][12] for hidden information delivery or anti-counterfeiting. The optoelectronic properties of CPs are determined by both the intrachain (intrinsic) and the interchain (packing) properties of the polymer chains. While the intrachain properties are mainly determined by the chemical structure of the main chain, the interchain properties are driven by the polymer packing, which is inuenced by various design factors. Therefore, the rational design of both the intrachain and the interchain properties of CPs is important for realizing CP applications with optimal characteristics.</p><p>Various strategies for regulating the intrachain properties of CPs through the design of the conjugated main chain have been investigated, including atom substitution, [13][14][15] addition of functional groups, 16,17 and the modication of CP building blocks. 18,19 These main chain design strategies also partly affect the interchain properties of CPs: for example, introducing planar building blocks and linkages, such as thiophene linked diketopyrrolopyrrole (DPP), promote close co-facial packing of CPs' main chain, which results in enhanced hole mobility in CP lms. 13 However, unlike the widely investigated main chain design of CPs, the effects of exible side chains on the interpolymer assembly and packing are rather unexplored. Flexible side chains are oen introduced mainly to provide solubility to the rigid-rod like CPs. For this purpose, long and branched side chains are usually adapted into CPs without a thorough consideration of their effects on the interchain packing. However, recent work has emphasized the role of the side chain and discusses it as an important design parameter to tune the properties of CPs. For instance, by investigating different side chains on the same conjugated main chain backbone, the side chain design has proven to be critical to the interchain properties. [20][21][22][23] Our research group recently showed an increase in hole mobility of three orders of magnitude in aligned lyotropic liquid crystalline CPs by rationally designing the molecular architecture of side chains together with the main chain conformation and form factor, which rendered directed selfassembly and packing of CPs. 24 Therefore, the optimization of CP characteristics requires a careful control for both intrachain and interchain properties: for this purpose, simultaneous design of conjugated main chain as well as exible side chain is essential.</p><p>With recent development of high-end scanning and copying technologies, counterfeiting has become a serious problem due to easy accessibility as well as its high quality. 10 In this trend, security inks for anti-counterfeiting or hidden information delivery have become a research eld of great interest. [10][11][12][25][26][27][28] Security inks provide invisible information in the initial stage which is not possible to read or photocopy; the hidden information such as color, pattern or letters are programed to be revealed under certain circumstances. Various functional organic molecules have been developed for security inks. [25][26][27][28] In these molecular systems the hidden information was encoded based on uorescence or color changes upon specic treatment (heat, chemical, or photo irradiation), and the concealed information was readable under irradiation of UV light. On the other hand, in case of CPs, polydiacetylenes that are prepared from unique molecular assembly of diacetylene monomers have mostly been investigated for security ink application, based on colorimetric change upon heat treatment. [10][11][12] However, CPs with aromatic conjugated main chains have not been fully investigated for security ink application despite its great potential based on the readily tunable optimal properties through systematic molecular design.</p><p>In this work, we demonstrate the modication of the optical and electrochemical properties of diketopyrrolopyrrole (DPP)-based CPs by controlling both the interchain and intrachain properties with effective but simple polymer-design strategies. We synthesized three CPs having an alternating donor-acceptor structure (P1, P2 and P3 in Fig. 1) using a phenyl linker connecting the electron-withdrawing DPP unit and an electron-donating building block. Regardless of three different electron-donating building blocks, all three CPs have almost identical absorption in diluted solutions and deep highest occupied molecular orbitals (HOMOs). This is caused by the large dihedral angle of the phenyl linker, which ensures a restricted effective conjugation length. In contrast, the interchain packing of the three CPs in the aggregated-lm state was gradually adjusted by changing the bulkiness and linking directions of the side chains (Fig. 1) to control the optical absorbance and HOMO levels. The color gradation of CPs in the aggregated state, in contrast to their identical color in solution, enables us to demonstrate the encoding of covert information utilizing CP solutions as security inks on a silica substrate, which immediately develop hidden information by water dipping.</p><!><p>The three designed CPs in Fig. 1 have distinct side chains, which differ in bulkiness and linking direction. Each CP also has a phenyl linker between the electron-donor and electronacceptor building blocks. In comparison to CPs having the commonly used planar thienyl linkers for the DPP unit, 18,29 the phenyl linker induces a large dihedral angle between building blocks, and hence the twisted conformation is expected to reduce the effective conjugation length. As a result, we anticipate largely blue-shied absorption and deeper HOMO levels due to the enlarged band gap of the phenyl-linked CPs. This is similar to the cases of poly(p-phenylene) or polyuorene, which exhibit a large band gap with bluish emission even in the highmolecular-weight regime because of the non-planar conformation of the phenyl linkage. [30][31][32] The optical properties of the three CPs are characterized by UV-vis spectrophotometry in diluted polymer solutions (10 À5 M based on the repeating unit). Compared to the analogous CPs having the thienyl linker, whose optical band edge is around 800-950 nm, 18,29 the optical band edge of the diluted solution of all three CPs is shied to ca. 600 nm. This is a result of the modication of the intrachain properties by the phenyl linker incorporation (Fig. 2a and Table S1 †). Interestingly, even though the three CPs are composed of different electrondonating building blocks (Fig. 1), the color of each CP solution is almost identical (Fig. 2c). This implies that the phenyl linker insertion effectively reduces the intramolecular charge transfer effect between the electron-donating and electron-withdrawing building blocks as well as the effective conjugation length. The identical intrachain properties, independent of the choice of electron-donating blocks, enables us to systematically investigate the effects of side chains on the interchain packing of CPs and, consequently, the alteration of their optoelectronic properties of P1, P2 and P3.</p><p>To further probe the role of the phenyl versus thienyl linkers on the structural and optical properties of the synthesized CPs, we performed structural-relaxation and electronic-structure calculations using Guassian09. 33 The molecules were relaxed using density functional theory (DFT) with the B3LYP hybrid functional and the 6-31G** basis set. We started by investigating the dihedral angle of the monomer unit (Fig. 3). In the case of the thienyl linker, the dihedral angle between the electron-donating/withdrawing building blocks and the thienyl linker varies between 0.45 to 8.02 , which implies an almost planar chain conformation (Fig. 3b and Table S2 †). On the other hand, the phenyl linker shows a much larger dihedral angle, between 20.42 to 23.65 , which is expected by the CP design (Fig. 3a and Table S2 †). Using the optimized conformation of the monomeric units with phenyl or thienyl linker, we investigated the band gaps of the monomer, dimer, and trimer structures. It is revealed that the phenyl linked CPs do exhibit a larger band gap than the thienyl-linked CPs (Table 1). For the P1, P2 and P3 analogues, trimers bearing a phenyl linker have band gaps of 2.28, 2.27, 2.22 eV respectively, which larger than those of thienyl-linked trimers (1.90, 1.94, 1.84 eV). Furthermore, the band gap decreases faster in thienyl-linker analogues as the chain length increases from the monomer to the trimer. This difference in band gaps supports the idea that the effective conjugation is more restricted in twisted phenyl linker analogues compared to planar thienyl linker analogues. In addition, the phenyl linker insertion shis the calculated HOMO levels of the CPs deeper (Table S3 †). The deep HOMO level is highly plausible to be advantageous when CPs are utilized in energy device application in terms of open circuit voltage (V oc ).</p><p>Even though the three CPs show almost identical optical properties in dilute solutions, their solid-state optical properties differ signicantly (Fig. 2b and d). The variation of the solidstate properties is originated from the molecular design, as each CPs has a different side chain with distinct bulkiness and orientation, which affects the interchain packing. P1 has two branched 2-ethylhexyl side chains connected through a bulky thienyl ring to the CP main chain. Our calculations show that the thienyl rings of the side chains are not in the same plane as the electron-donating benzo[2,1-b:3,4-b 0 ]dithiophene that they are connected to, but have instead adopted a largely twisted conformation (ca. 69 ) to minimize steric hindrance (P1 in Fig. 3a). Because of the branched bulky 2-ethylhexyl chains together with the twisted thienyl rings, the main chains of P1 are anticipated to have minimal co-facial p-p stacking. On the other hand, P2 has two linear octyl side chains connected on the same side of the benzo[2,1-b:3,4-b 0 ]dithiophene unit but without the thienyl linkage, which is expected to have a mediocre propensity for interchain packing. The n-octyl side chains of P2 might stretch out to the opposite out-of-plane directions of the main chain plane due to steric hindrance as shown in Fig. S1. † In case of P3, the two 2-ethylhexyloxy side chains are not bulky enough to prevent P3 from aggregating. 24 dithiophene donor unit such that the side chains do not have steric repulsion either. As a result, P3 is expected to have strong p-p stacking between the main chains. Indeed, the colors and the UV-vis absorption spectra of the thermally annealed CP lms show distinct gradation despite the similar absorption spectra in solution (Fig. 2). The absorption l max of the P3 lm is red-shied by 36 nm compared to that of the P3 solution and a new aggregation band appears at 620 nm (Fig. 2b and Table S1 †). The red-shied new absorption band in P3 lm may also be ascribed to an extended effective conjugation length caused by planarization of the conjugated main chain in the aggregated lm state. In contrast, P1 essentially shows almost identical absorption in the solid lm and in solution due to weak interchain interactions, while the absorption l max of the P2 lm is red-shied by 9 nm compared to the P2 solution (Fig. 2b and Table S1 †). The grazing incidence X-ray diffraction (GIXRD) analysis also revealed the same trend in the interpolymer packing: P3 shows a distinct peak at about 5 corresponding to the developed interchain packing, while P1 shows amorphous packing due to the out-of-plane bulky side chains (Fig. S2 †).</p><p>We further demonstrate the encoding of covert optical information on a silica substrate using the three CPs as security inks by exploiting the difference among their interpolymer packing. A highly polar silica gel is likely to interact strongly with the CPs in the dry-state due to polar-polar interactions, which can prevent the aggregation of CPs. When dilute solutions of the CPs (ca. 5 Â 10 À4 M based on the repeating unit) were painted on polar silica gel substrates, the CPs retained identical solution-like colors even aer evaporation of the chloroform solvent (Fig. 4a). However, because water molecules are highly polar and favorable for hydrogen bonding, they interact strongly with the silica substrates and break the CP-substrate interaction. The different aggregation tendencies of the CPs caused by the side chains were manifested by a vivid color difference among the CP lms when the substrate was dipped into water: P1 is red, P2 is purple, and P3 is bluish purple (Fig. 4a). By using P1 and P3 as inks, we demonstrated the encoding of latent patterns and messages. An array of 5 Â 5 dots of P1 and P3 inks with identical colors were patterned on a silica substrate (Fig. 4b). When the dot array was immersed into water, the latent arrow image was developed immediately (Movie S1 †). By increasing the array size to 11 Â 7, we could demonstrate the covert letter "M" (Fig. 4b and Movie S2 †). Moreover, this phenomenon is completely reversible: the patterns developed by water are completely erased by chloroform dipping or under chloroform vapor treatment (Fig. 4a and Movie S2 †).</p><p>The design strategies to control the intrachain and interchain properties of CPs also essentially link to the electrochemical properties of the CP lms. Each CP shares the same phenyl linker so as to has a deeper HOMO than the conventional analogous CPs with thienyl linkers. Cyclic voltammetry (CV) measurements were conducted to measure the energy levels of the three CPs. CP lms for the measurements were spin cast onto ITO substrates, followed by thermal annealing. The energy levels of regioregular poly(3-hexylthiophene) (P3HT) and [6,6]-phenyl-C 61 -butyric acid methyl ester (PCBM) are presented together with the CPs for comparison (Fig. 5b). The three CPs indeed have deep HOMO energy levels (À5.34 to À5.69 eV) (Fig. 5), which are deeper by about 0.3 eV than the HOMO levels of the analogous CPs with thienyl linkers reported in the literature. 18,34 Furthermore, the interpolymer packing propensity difference of the three CPs also affects the electrochemical properties of CP lms. Similar to the variation of the optical properties, the different interchain properties of the CPs result in a gradual change of the HOMO levels: the restricted interchain interaction via the side chain design results in deeper HOMO levels. As a result, P1 with an amorphous nature has the deepest HOMO level (À5.69 eV), while P3 with a relatively high interpolymer packing propensity exhibits the shallowest HOMO level (À5.34 eV) among the three CPs. A similar trend of the HOMO levels were reported in recent publications, emphasizing the role of side chains in designing the electrochemical properties. 20,21,35</p><!><p>In summary, we present an effective but simple design strategy for controlling the optical and electrochemical properties of CPs. Controlling the dihedral angle along the CP main chain with a phenyl linker combined with side chains of varying bulkiness and orientation can modulate both the intrachain and interchain properties of CPs. Regardless of the difference in electron-donating building blocks, the synthesized CPs, P1, P2 and P3, exhibit almost identical absorption with deep HOMO energy levels, due to a restricted effective conjugation length resulting from the large dihedral angle of the phenyl linker. The side chain design decisively affects the interpolymer packing of the CPs and renders gradual shis in optical properties and HOMO energy levels in the aggregated state. We further demonstrate the encoding of latent optical information on a silica substrate by exploiting the interpolymer packing difference of the three CPs. Hidden patterns recorded in the dry state are immediately revealed upon water treatment in a completely reversible manner.</p>

Royal Society of Chemistry (RSC)

A New Function of Nell-1 Protein in Repressing Adipogenic Differentiation

A theoretical inverse relationship has long been postulated for osteogenic and adipogenic differentiation (bone versus adipose tissue differentiation). This inverse relationship in theory at least partially underlies the clinical entity of osteoporosis, in which marrow mesenchymal stem cells (MSCs) have a predilection for adipose differentiation that increases with age. In the present study, we assayed the potential anti-adipogenic effects of Nell-1 protein (an osteoinductive molecule). Using 3T3-L1 (a human preadipocyte cell line) cells and human adipose-derived stromal cells (ASCs), we observed that adenoviral delivered (Ad)-Nell-1 or recombinant NELL-1 protein significantly reduced adipose differentiation across all markers examined (Oil red O staining, adipogenic gene expression [Pparg, Lpl, Ap2]). In a prospective fashion, Hedgehog signaling was assayed as potentially downstream of Nell-1 signaling in regulating osteogenic over adipogenic differentiation. In comparison to Ad-LacZ control, Ad-Nell-1 increased expression of hedgehog signaling markers (Ihh, Gli1, Ptc1). These studies suggest that Nell-1 is a potent anti-adipogenic agent. Moreover, Nell-1 signaling may inhibit adipogenic differentiation via a Hedgehog dependent mechanism.

a_new_function_of_nell-1_protein_in_repressing_adipogenic_differentiation

Introduction<!>Isolation of Cells<!>In vitro adipogenic differentiation<!>Adipogenic Assessments<!>Bromodeoxyuridine Incorporation<!>Osteogenic Differentiation and Assessments<!>Statistical analysis<!>Nell-1 Effects on 3T3-L1 Adipogenic Differentiation<!>Nell-1 Effects on 3T3-L1 Adipogenic Gene Expression<!>Nell-1 Effects on 3T3-L1 Proliferation<!>Nell-1 Effects on Human ASC Differentiation<!>Nell-1 Effects on Hedgehog Signaling<!>Discussion<!><!>Supplemental Figure 2. Cellular Proliferation in 3T3-L1 Cells<!>Supplemental Figure 3. Schematic Representation of Nell-1 Regulation of MSC Differentiation

<p>Numerous studies have proposed an inverse relationship between the differentiation of multipotent mesenchymal stem cells (MSCs) toward osteoblastic and adipocytic cell fates [1,2,3,4]. A number of signaling cascades have been implicated in this cell fate decision, including bone morphogenetic protein (BMP), Hedgehog, and Wnt (Wingless Protein) signaling, to name only a few [5,6,7,8,9,10]. In cases such as Hedgehog and Canonical (or β-catenin dependent) Wnt Signaling, a general enhancement of osteogenic over adipocytic differentiation has been repeatedly observed – or a 'shift' in lineage differentiation [8,9,10,11,12,13,14]. Thus, such concepts as 'loss of bone is a gain in fat' is a commonly held conception [1].</p><p>The growth factor Nell-1 has long been recognized to have osteoinductive properties. Nell-1 was first discovered as overexpressed in prematurely fusing (ossifying) calvarial sutures of human patients [15]. Nell-1 overexpressing mice demonstrate calvarial bone overgrowth and inappropriate cranial suture fusion [16]. Conversely, animals deficient in functional Nell-1 demonstrate major skeletal defects, including defects in the cranial, axial and appendicular skeleton [17,18]. Moreover, Nell-1 has been observed to induce bone formation both in small and large mammalian models [19,20,21,22,23], and in vitro across numerous cell types [16,24,25]. The effects of Nell-1 on adipogenic differentiation have yet to be systematically evaluated. However, recent evidence has found that Nell-1 can reverse the pro-adipogenic effects of high-dose BMP2 (bone morphogenetic protein2) both in vitro and in vivo (data in submission).</p><p>In this present study, we expose an immortalized pre-adipocyte cell line 3T3-L1 and human primary adipose-derived stromal cells (hASCs) to enhanced Nell-1 signaling via an adenoviral vector. As Nell-1 signaling is a potent pro-osteogenic inducer, we hypothesized that Nell-1 would negatively effect adipogenesis. We then in a candidate fashion examined Nell-1 mediated changes to the Hedgehog signaling pathway, to potentially explain the molecular mechanisms whereby Nell-1 may deleteriously impact the process of fat formation.</p><!><p>3T3-L1 cells were a kind gift of the Tontonoz Laboratory, University of California, Los Angeles. ASCs were isolated from human adult lipoaspirate. No individually identifiable information about the donor was received, therefore IRB approval was not requested. Fat tissues was obtained from N=3 healthy patients. After standard liposuction procedures, the lipoaspirate (100 ml) was washed with an equal volume of PBS and centrifuged at 1800 RPM for 10 min. The tissue fraction on the top layer was collected and transferred into a new tube. An equal volume of digestion solution (RPMI, 3.5% BSA, 10ug/ml DNAse, 1mg/ml Collagenase II) was added and transferred into a shaking incubator for 30-45 min at 37°C 250 RPM. Following the digestion, the solution was filtered using a 70 μm cell strainer, and centrifuged at 1800 RPM for 10 min. The fatty top layer containing the adipocytes was discarded and the pellet resuspended in PBS 5mM EDTA. After two washes in PBS 5mM EDTA, the pellet was resuspended in red blood cell lysis solution for 10 min at RT. After adding 3 volumes of PBS 5mM EDTA and centrifuging at 1500 RPM for 5 min, the pellet was resuspended in PBS 5mM EDTA prior to counting the number of viable stromal vascular fraction (SVF) cells via Trypan blue staining. Cells were expanded in 100 mm dishes thereafter.</p><!><p>For adipogenic differentiation, 3T3-L1 or hASCs were seeded in 12-well plates at a density of 50,000 cells per well. All assays were performed in triplicate wells. After attachment, cells were treated with adipogenic differentiation media (ADM). For 3T3-L1 cells, ADM consisted of Dulbecco's Modified Eagle Medium (DMEM), 10% FBS, 1 uM dexamethasone, 0.5 mM IBMX, 5 ug/mL insulin, 1% penicillin / streptomycin. For hASCs, propriety ADM was purchased (Stem Cell Technologies, Inc.). Cells were maintained for 3 days in ADM, and thereafter refreshed with DMEM, 10% FBS, and insulin only for 3T3-L1 cells, or propriety ADM for hASCs. All experiments were performed in biological triplicate (cells derived from 3 separate patients), and each assay was performed in triplicate wells (N=3 wells per cell type per experiment, N=3*3=9 samples per condition). Cell medium was supplemented in all cases with recombinant human Nell-1 protein at various dosages, or adenoviral delivered Nell-1 (Ad-Nell-1). Control virus was used at equal concentrations (Ad-LacZ). Ad-Nell-1 and Ad-LacZ were used as previously published, at a concentration of 20 and 40pfu/cell [16].</p><!><p>Assessments of adipogenic differentiation were performed as previously described [10,26,27]. Oil red O staining was performed using a 0.2% solution in 60% isopropanol / 40% deionized water. Real time PCR was performed as previously described, performed in triplicate wells per RNA isolate [28]. Specific mouse and human genes and primer sequences are listed in Table 1.</p><!><p>Proliferation was assessed by bromodeoxyuridine (BrdU) incorporation assays as previously described [29]. Cells were grown in 96 well plates seeded at a density of 1,000 cells / well. After 2, 4 and 6 days growth, BrdU labeling was performed for a period of 8 hours. Next, BrdU incorporation was quantified using a photometric ELISA (Roche Applied Science, Indianapolis, Ind.). N=6 per treatment group, experiments were performed on three separate occasions.</p><!><p>Osteogenic differentiation and assessments were performed using hASCs as previously described. To assess early to intermediate osteogenesis, alkaline phosphatase staining and quantification was performed as previously described; quantification was in each case normalized to total protein content in sister wells [30]. To assess bone nodule formation Alizarin red staining and quantification was performed as previously described; quantification was performed by CPC leaching and photometric quantification, normalized to total protein content [31].</p><!><p>Means and standard deviations were calculated from numerical data, as presented in the text, figures and figure legends. In figures, bar graphs represent means, whereas error bars represent one standard deviation. Statistical analysis was performed using the appropriate ANOVA when more than two groups were compared, followed by a post-hoc Student's t-test to directly compare two groups. The exact statistical analysis for each dataset is described in the figure legends. Inequality of standard deviations was excluded by employing the Levene's test. *P ≤ 0.01 was considered to be significant.</p><!><p>3T3-L1 preadipocytes are one of the most commonly studied adipoprogenitor cell lines, and represent a unipotent cell type. 3T3-L1 cells were exposed to standard adipogenic differentiation media with adenoviral (Ad)-Nell-1 or Ad-LacZ as a control. After seven days, while Ad-LacZ treated cells showed robust lipid accumulation (Fig. 1A), Ad-Nell-1 showed only sparse lipid droplets (Fig. 1B). Two viral titers were assessed, with both showing the same trend. Real time quantitative PCR demonstrated an over 500% increase in Nell-1 expression with Ad-Nell-1 treatments with both viral titers (Supp. Fig. 1). Oil red O staining results were quantified by leaching with isopropanol and photometric quantification, normalized to total protein content of sister wells (Fig. 1C). Quantification demonstrated an approximate 80% reduction in Oil red O staining intensity. Importantly Trypan blue staining showed no increase in cytotoxicity with Ad-Nell-1 treatment (data not shown).</p><!><p>To confirm the anti-adipogenic effects of Nell-1 signaling on 3T3-L1 pre-adipocytes, cells were cultured in the presence of Nell-1 overexpressing adenovirus (Ad-Nell-1) or control virus (Ad-LacZ); RNA was harvested at stratified timepoints post induction (two, four and six days). As expected, all markers gradually increased overtime in adipogenic medium under control conditions (including Peroxisome proliferating factor gamma [Ppar], Lipoprotein lipase [Lpl] and Adipocyte protein 2 [Ap2]) (Fig. 2, blue bars). Significantly Ad-Nell-1 addition to adipogenic medium resulted in a significant reduction in all markers (Fig. 2, red bars). These data confirmed the significant anti-adipogenic effects of Nell-1 signaling in 3T3-L1 cells.</p><!><p>Theoretically a reduction in the proliferation of pre-adipocytes by Nell-1 signaling could result in a reduction in Oil red O staining (see again Fig. 1). To confirm or reject this hypothesis, bromodeoxyuridine (BrdU) incorporation assays were performed in the presence of control (Ad-LacZ) or Nell-1 overexpressing adenovirus (Ad-Nell-1) (Supp. Fig. 2). Interestingly, no significant change in BrdU incorporation was observed (Supp. Fig. 2, compare red and blue bars). Thus, we reasoned that Nell-1 signaling likely inhibits adipoprogenitor cell differentiation rather than pre-adipocyte expansion in vitro.</p><!><p>Mesenchymal Stem Cells (MSCs) clearly retain important biological differences from 3T3-L1 pre-adipocytes. We next inquired as to whether Nell-1 signaling would have the same negative effects on adipogenic differentiation a human MSC population: adipose-derived stromal cells or human (h)ASCs. We opted to utilize recombinant protein instead as a more defined and quantitative delivery of Nell-1 alternative source. Recombinant human (rh)NELL-1 from CHO cells was supplemented to standard adipogenic or osteogenic ASC differentiation medium (Fig. 3). As expected, rhNELL-1 had a negative effect on adipogenic differentiation, as demonstrated by Oil red O staining of lipid accumulation (ORO, appearing red) (Fig. 3A). In stark contrast, rhNELL-1 increased markers of osteogenic differentiation by biochemical staining, including alkaline phosphatase (ALP, an intermediate marker of osteodifferentiation appearing purple) and Alizarin red (AR, a stain for bone nodule formation appearing red) (Fig. 3B,C). Quantification of each stain confirmed a significant upregulation of osteogenic differentiation and converse and negative effect on adipocytic differentiation (Fig. 3D). Thus, Nell-1 signaling inhibited adipogenic differentiation in both a unipotent pre-adipocyte cell line, and a primary MSC.</p><!><p>Hedgehog signaling has been known to positively regulate osteogenic differentiation at the expense of adipogenic differentiation in multiple cell types, including ASCs [10,32,33,34]. We therefore next inquired as to the effects of Nell-1 on hedgehog signaling. 3T3-L1 cells were culture in the presence of control virus (Ad-LacZ) or Nell-1 overexpression virus (Ad-Nell-1) (Fig. 4). Indian Hedgehog (Ihh) ligand showed upregulation after both 4 and 6 days differentiation (Fig. 4A). Likewise two markers of hedgehog signaling activation showed similar upregulations, Patched1 (Ptc1) and Gli1 (Fig. 4B,C). Conversely, Gli3, generally considered a negative regulator of Hedgehog pathway activity, showed a reduction in transcript abundance (Fig. 4D). Thus and in summary Nell-1 appears to reduce adipogenic differentiation of 3T3-L1 cells with concomitant induction of Hedgehog ligand expression and Hedgehog signaling activity.</p><!><p>Adipocytes and osteoblasts arise from a common mesenchymal progenitor cells, and a theoretical inverse relationship exists between these two cell types (Supp. Fig. 3). Numerous bipotent or multipotent cell types exist for future tissue engineering applications, including traditional bone marrow MSCs (BMSCs), adipose-derived stromal cells (ASCs) and umbilical cord-derived (UC-) MSCs. Accumulating data suggests that this bipartite destiny of MSCs (bone versus fat) can be shifted given appropriate cytokine stimulation [1]. Here, we sought to examine whether Nell-1 would be an appropriate signal to shift the balance from adipocytic to osteogenic differentiation.</p><p>In fact we found this to be the case: namely that Nell-1 definitively represses adipogenic differentiation in vitro. First, Nell-1 represses adipogenesis in a unipotent cell (a pre-adipocyte – or 3T3-L1 cell). Secondarily, Nell-1 likewise serves to repress adipogenic differentiation in a multipotent ASC. This suggests the effect of Nell-1 on adipogenesis are direct repression of lipogenesis, rather than an indirect repression via enhancement of the converse lineage (osteogenic) differentiation.</p><p>No doubt, Nell-1 is one of many cytokines known to 'shift' this MSC differentiation away from fat and towards bone. One particular cytokine pathway of interest is Hedgehog signaling, which in multiple cell types including ASCs has been shown to cause a similar 'shift' in differentiation [10,32,33,34]. Three mammalian Hedgehog ligands exist, including Sonic Hedghog (Shh), Indian Hedgehog (Ihh) and Desert Hedgehog (Dhh) – and as Dhh appears limited to the mammalian testes, both Shh and Ihh have been studied in greater detail. Both ligands appear to have a similar and equivalent pro-osteogenic effect (for example the Ihh null has severe skeletal defects) – however Shh has been studied in greater detail in vitro, where the N-terminal domain is responsible for signal propagation [34,35,36]. In our previous manuscript, we detailed the extent to which Shh-N positively regulated osteogenesis and inhibited adipogenesis, both in vitro and in vivo in mouse ASCs [10].</p><p>Given the striking similarities in effect between Shh and Nell-1, we sought here to define the extent to which Nell-1 may positively regulate hedgehog signaling. In fact, a positive relationship was observed. These studies suggest highly that at least some portion of the differentiation effects of Nell-1 are via activation and/or potentiation of Hedegehog signaling. Whether Nell-1 directly increases Hedgehog ligand (Shh or Ihh) expression, directly increases Gli transcription or Gli activation, or inactivates a Hedgehog repressor (Ptc1, Gli3, Rab23 among numerous others) remains as yet unknown. Certainly there exist other potentially signaling networks that may work to transduce a Nell-1 signal into a 'shift' in lineage differentiation. For example, MAPK signaling in known to modulate both osteogenic and adipogenic differentiation and has been previously shown to be up-regulated in certain contexts by Nell-1 [37]. In addition, Canonical Wnt signaling is well known to overall produce a similar 'shift' in differentiation and again has been shown to be up-regulated both in vitro and in vivo by Nell-1 protein (in submission). These studies suggest the multifactorial and interwoven nature of cell fate decisions, and suggest Nell-1 to be integral to the balance of osteo- and adipogenic MSC differentiation.</p><p>In summary, in addition to the well-recognized role of Nell-1 is promotion of osteogenic differentiation and bone formation, Nell-1 significantly represses adipogenic differentiation. This anti-adipogenic effect of Nell-1 may be through potentiation of Hedgehog signaling among other signaling cascades. These studies suggest the future therapeutic utility of Nell-1 for either skeletal or soft-tissue regenerative medicine.</p><!><p>Supplemental Figure 1: Nell-1 gene expression was evaluated by quantitative RT-PCR after 2 days treatment with Ad-Nell-1 or Ad-LacZ, showing upregulation of Nell-1 expression among Ad-Nell-1 treated samples at both low and high viral titers as depicted in Figure 1. Ad-Nell-1 and Ad-LacZ were used as previously published, at a concentration of 20 and 40 pfu/cell [16].</p><!><p>BrdU Incorporation assays were performed after 2, 4 and 6 days growth with control virus (Ad-LacZ, blue bars) or Nell-1 overexpressing adenovirus (Ad-Nell-1). Data are normalized to Ad-LacZ 2 day BrdU uptake levels. Ad-Nell-1 and Ad-LacZ were used as previously published, at a concentration of 20 pfu/cell [16]. No difference was observed across all timepoints.</p><!><p>Shown in blue, Nell-1 positively regulates MSC differentiation into pre-osteoblasts as well as further bone tissue ossification. Shown in pink, Nell-1 conversely negatively regulates MSC differentiation into pre-adipocytes as well as further adipocyte maturation.</p>

PubMed Author Manuscript

Morphological Analysis of White Cement Clinker Minerals: Discussion on the Crystallization-Related Defects

The paper deals with a formation of artificial rock (clinker). Temperature plays the capital role in the manufacturing process. So, it is useful to analyze a poor clinker to identify the different phases and defects associated with their crystallization. X-ray fluorescence spectroscopy was used to determine the clinker's chemical composition. The amounts of the mineralogical phases are measured by quantitative XRD analysis (Rietveld). Scanning electron microscopy (SEM) was used to characterize the main phases of white Portland cement clinker and the defects associated with the formation of clinker mineral elements. The results of a study which focused on the identification of white clinker minerals and defects detected in these noncomplying clinkers such as fluctuation of the amount of the main phases (alite (C3S) and belite (C2S)), excess of the free lime, occurrence of C3S polymorphs, and occurrence of moderately-crystallized structures are presented in this paper.

morphological_analysis_of_white_cement_clinker_minerals:_discussion_on_the_crystallization-related_d

1. Introduction<!>2.1. Materials<!>2.2.1. X-Ray Fluorescence (XRF)<!>2.2.2. X-Ray Diffraction (XRD)<!>2.2.3. Scanning Electron Microscopy<!>3. Results and Discussion<!>3.1. Characterization of Clinker Phases by SEM<!><!>3.2. Defects Associated with the Crystallization<!><!>3.2.2. Poor Crystallization<!>3.2.3. The Presence of Lamellar Structures<!>

<p>Portland cement is a mixture of clinker (artificial rock from cooking a vintage mixture of limestone and clay) and ground gypsum (controller plug). The morphology and composition of the phases in a clinker can vary significantly depending on the manufacturing process and raw materials used [1]. In the clinker, the following predominate chemicals elements are Ca, Si, Al, Fe, Mg, Na, and K. These elements are expressed as a percentage of oxides. These elements are expressed as a percentage of oxides and the Bogue [2, 3] notation is used to refer to them: CaO = C, Al2O3 = A, SiO2 = S, Fe2O3 = F, and MgO = M. Clinker is a multiphase mixture and, so far, more than 30 constituent phases have been identified [4]. Despite the wide variety of clinker phases, only four of them are, in practice, of real importance: silicates including alite (C3S = Ca3SiO5 = 3CaO-SiO2) give the hydrated cement short-term resistance; belite (C2S = Ca2SiO4 = 2CaO-SiO2) which confers long-term resistance to the finished product [5]; aluminates consisting of tricalcium aluminate (C3A = Ca3Al2O6 = 3CaO-Al2O3) and aluminoferrites (C4AF = Ca4Al2 Fe2O10 = 4CaO-Fe2O3-Al2O3) [6]. Consider(1)C3S=Ca3SiO5,C2S=Ca2SiO4,C3A=Ca3Al2O6,C4AF=Ca4Al2Fe2O10.</p><p>Depending on the temperature and impurities, C3S has seven polymorphic phases: three triclinic, three monoclinic, and a rhombohedral one [6, 7]. C2S has five polymorphic forms denoted as γ, β, α′L, α′H, and α in the temperature range between room temperature and 1500°C [7]. The C2S phase usually found in clinker is the β monoclinic one [8]. For aluminates, the C3A is frequently found with cubic and orthorhombic forms [9]. The ferrite crystallizes into an orthorhombic form [10]. It is difficult to distinguish between the different interstitial aluminate phases (C3A and C4AF) [11]. Nevertheless, the clinkers usually contain some amount of free lime (up to 1-2%) and free calcium sulphate [12].</p><p>In white cement, the whiteness index is a very important parameter to control. However, white cement contains a very low amount of iron since it decreases the whiteness, which implies insignificant content of C4AF. Therefore, white clinker contains three major constituents: C3S, C2S, and C3A [11].</p><p>The minerals formed at low temperature (ca 1200–1300°C) are ill crystallized and contain large amounts of admixtures [13]. Their size is usually very small, less than 5 μm, since the clinker crystals are primarily formed by a solid state reaction [14].</p><p>Most of the SEM studies to date have dealt with the general characterization of clinker phases, including their compositional variations [15]. Among the articles describing use of the SEM in analysis of clinker microstructure is the work of [16, 17] in which the various clinker phases are identified and described in order to interpret the manufacturing process. Other publications describe the use of SEM in the clinkers minerals analysis [15, 18–27].</p><p>With the development of research, we can now ensure that X-ray powder diffraction, combined with the Rietveld [28] method, is the most recent and most accurate way of quantifying the mineralogical composition of Portland clinker [4, 8, 11, 29–37].</p><p>However, in some cases, the industry does not comply with the regulations in force when manufacturing clinker. For economic and environmental reasons, the clinker should be recycled back into the original raw material. So any defect in the clinker will be added to the energy cost. This paper uses several analytical techniques (SEM, XRD, and XRF) for the analysis of some poor cement clinkers, aiming to identify different phases and defects within the clinkers.</p><!><p>Clinker samples (SO1–SO6) and raw materials samples were taken from SOTACIB (Tunisian-Andalusian White Cement Company). They were subjected to the following analysis: a chemical analysis by X-ray fluorescence (ARL type XP 9800) to determine their compositions expressed as oxides in wt.%. Scanning electron microscopy (JEOL JSM-5400) was carried out to identify the mineralogical phases of the clinkers and the associated defects. Finally, a quantitative phase analysis was performed using X-ray powder diffraction (Bruker D8 ADVANCE Diffractometer) and the Rietveld method.</p><!><p>The same ground samples were used to determine the geochemical compositions. We started by measuring the loss on ignition (LOI) for each sample. Then, in a platinum crucible, 1 g of decarbonated material was added to 6 g lithium tetraborate. The crucible was placed in a furnace at 1100°C for 20 min while melts were stirred every 5 minutes. A mould was placed in the furnace for 5 min and the melts in the crucible were then poured into it. The resulting pearls were analyzed by X-ray fluorescence (XRF).</p><!><p>All samples were finely ground (down to ~10 μm) for the powder diffraction measurements. XRD data were collected at room temperature using Cu-Kα radiation (λ = 1.5406 Å) operated in the reflection geometry (θ/2θ). Data were recorded from 10° to 60° (2θ) with a step-size of 0.02. The X-ray tube was operated at 40 kV and 40 mA.</p><!><p>The clinker compounds studied were examined by scanning electron microscopy (SEM). The same samples as those for XRD were analyzed by SEM, but in the latter case they were not ground. Instead, grains of the order of several millimeters in diameter were used for SEM analysis.</p><!><p>The raw material is represented by the chalky limestone of Jebel Feriana (Tunisia) of Abiod from Campanian-Maastrichtian age, silica sands of Beglia (Tunisia) from Miocene age and kaolin of Turkey (imported).</p><p>Limestone. This is a white limestone from the Campanian-Maastrichtian age, low hardness, which has a high content of CaCO3 (>95%) and minimal amounts of colorants metal oxides, mainly Fe2O3, Ti2O3. This creates a fairly high degree of whiteness (β ≈ 85.34) (Table 1).</p><p>Purity. Compared to the calcite containing 56% of CaO, the limestone of Jebel Feriana has a mean content of 98.16% CaCO3.</p><p>Kaolin. The kaolin used by SOTACIB is imported from Turkey. It meets the quality requirements expressed in a specified set of specifications. Imported kaolin is rich in silica; it contains a mean of 58.52%. While it is poor in alumina, it contains a 26.07% mean content of Al2O3. The main metal oxides are present in minimal concentrations of 0.85% for Fe2O3 and 0.79% for titanium (TiO2) (Table 1).</p><p>Sand. The Sands of Beglia are extrasilica sand (SiO2 > 90%). Coloring oxides are presented with very low concentrations as shown in Table 1.</p><p>Generally, in the white cement, limestone acts to a weight of about 4 times higher than that of sand and kaolin (about 80% limestone, 10% kaolin, and 10% sand). But it is unstable because of the variation of the geochemical composition of raw materials. In light of crude's control parameters, namely, LSF (between 96 and 97%) and MS (in the order of 5.3%); and alumina (2.6 to 2.7%), the flow rate of material was modified to achieve a geochemical composition of regular crude in time and complying with the standards.</p><p>The XRD data showed the main mineralogical components of different raw material used for white cement clinker production (Figure 1).</p><p>The values for the weight percentage of oxides are listed in Table 2. The compositions of clinkers studied were distributed as follows: lime and silica (CaO ≈ 70% and SiO2 ≈ 23-24%), alumina (Al2O3 ≈ 3-4%), and other minor elements (Fe2O3, K2O, SO3, TiO2, and MgO, all present in small quantities).</p><p>The XRD data showed the effect of clinker defects when analyzing the X-ray diffraction peaks. The comparison of the full width at half maximum (FWHM) values for different peaks of the (SO1) sample diffraction pattern and that of a standard clinker shows this effect (Figure 2). In fact, peak broadening was observed in the sample diffraction. Furthermore, the DRX pattern of sample SO1 presents peaks less intense than those of standard clinker, which indicates the incomplete crystallization of the calcium silicate. This is an indication of the poor quality of the clinker as well as the poor formation of the mineralogical phases. The scanning electron micrograph (Figure 3) shows that this clinker presents crystals with hexagonal and circular shapes for alite and belite, respectively [43]. The SEM observations of the samples SO1 (Figure 3) showed likewise, globular shapes with truncations and invisible crystal edges. It is also important to note the presence of voids on either side of the crystal, often employed by the uncombined lime, as well as the irregular and interconnected pores.</p><p>The X-ray data was refined by Rietveld analysis using the X'Pert HighScore Plus program (from PANalytical). The Rietveld analysis always gives the sum of the phases present normalized to 100%. Therefore, all the phases present must be entered into the analysis with their known crystal structure.</p><p>Crystal structures of alite [38], belite [39], cubic C3A [9], free lime [40], portlandite [41], and periclase [42] taken from the literature were used for refinement and quantification.</p><p>XRD profiles were fitted using the pseudo-Voigt function.</p><p>The Rietveld refinement is carried out in the following order:</p><p>Background (polynomial of 5 coefficients combined with 1/X term), zero shift (2θ), scale factor, and cell parameter were refined first, followed by the phase profile parameters (W, U, and V).</p><p>An example of the Rietveld fitting (for sample SO5) is displayed in Figure 4.</p><p>The calculated wt.% of the phases and the Rietveld conventional agreement indexes are listed in Table 3.</p><p>The quantification of poor white Portland clinkers by the Rietveld method showed the abundance of two major phases (C3S-monoclinic and C2S-monoclinic (beta)). These two phases showed significant variation from one sample to another. The abundance of lime in sample (SO3) was also noted Table 3. The fluctuations in the contents of C3S and C2S (C3S: 58.5–85.8%; C2S: 2.8–27.6%) can be explained by changes in the kiln temperature (cold kiln). The levels variability of β-C2S (belite-monoclinic) also indicates the instability of the kiln temperature and noncontrolled cooling. The occurrence of high percentage of portlandite (SO3 and SO4; Table 3) suggested noncontrolled cooling.</p><!><p>SEM analysis was carried out in order to recognize different phases (crystalline and amorphous phases) within the clinker. This analysis compared the cooking state between clinkers. SEM also helped to interpret the burning condition.</p><!><p>The crystals had well-defined geometric shapes. The presence of blunt, cracked, ovoid, and anhedral shapes was also observed [20]. In addition, large flattened crystals were also detected.</p><p>Belite was also present in many forms. This is often associated with large, flat alite crystals.</p><p>Lime, with a whish appearance and flower-like structure, existed near the newly formed crystals and in the voids.</p><p>In addition, the presence of anhedral particles can be due to silica gel which formed isolated crystals.</p><p>Alite (C3S): the alite crystal is the preferred clinker phase in Portland cement clinker. The formation of alite is very complex. Alite is only formed in the presence of excess CaO, that is, at CaO/SiO2 > 1 and is thermodynamically stable above 1250°C [44]. This mineral is found in all the samples in a distinctive crystalline form (Figure 5). The crystallized phases of alite present different morphologies and crystal sizes [43]. However, if the size of the crystals is taken into account, one can easily distinguish the polymorphism of this mineral, (cf. Sample SO4 and SO5) (Figure 5). More precisely, one can observe large, elongated, and well-crystallized crystals which often appear with truncations (SO4 06 and SO4 09 in Figure 5). Image SO5 01 in Figure 5 shows the crystal distribution in the clinker grains. It is a homogeneous distribution where the crystal edges are very clear with an amorphous phase. There were also other well-crystallized crystals (Euhedral alite crystals in clinker void) with smooth sides and clear edges similar to hexagonal shapes (SO5 04 and SO6 04, Figure 5).</p><p>Belite (C2S): another most common and well-known component of clinker is the belite β-C2S (monoclinic). In industrial cement production belite is formed in the rotary kiln at 900–1250°C [45]. The basis of belite formation is the reaction between solid CaO and SiO2 particles. In our case, this mineral had ovoid or globular shape (SO5 03 and SO6 02, Figure 6). It was not possible to distinguish an individual crystalline structure. The belite formation is closely related to alite formation, that is, if one increases, automatically, the other will be decreased (samples SO2 and SO4. Table 3).</p><p>The tricalcium aluminate (C3A): it was not possible to distinguish well-defined shapes. This mineral appeared as an interstitial phase [46] within the major phases (C3S and C2S) (SO3 01, Figure 6) or vitreous (SO3 04, Figure 6). C3A plays a crucial role in clinkerisation formation, as this phase forms a molten phase at lower temperatures, which influence the development of the chemical and physical properties of the final clinker product greatly.</p><!><p>Critical crystallization-associated defects which are visible under the SEM are as follows.</p><!><p>high refusal (grinding problem);</p><p>insufficient treatment of raw materials (cold oven);</p><p>dissociation of C3S in C2S due to uncontrolled cooling and the production of CaO (C3S → CaO + C2S);</p><p>gas flow and raw material.</p><!><p>Poor crystallization is reflected by the presence of incomplete crystalline phases (SO6 01, Figure 8) and gels (SO4 04, Figure 8). The phases are rather rough without well-defined contours. The growth of alite and belite crystals was observed in the melt phase (Figure 9) [46]. This phenomenon is linked primarily to fluctuations in the temperature of the oven (low temperature ≈ 1200–1300°C). This poor crystallization is also evidenced by the appearance of rounded to subrounded shapes with diffuse boundaries between the phases.</p><!><p>Leafy and lamellar structures were observed (SO4 11, Figure 8) and accumulated around the alumina silicates or the belite crystals (C2S) which crystallize into lamellar shapes.</p><!><p>It is possible to distinguish different morphologies (euhedral and anhedral shapes) and sizes of alite within the same sample. This indicates that baking is inhomogeneous in the oven and that the temperature fluctuates at least at the chamber level relative to the formation of alite. It is known from the literature that, at room temperature, only the monoclinic phase exists.</p><p>The presence of free lime, the vitreous phase, and lamellar structures proves the instability of the oven temperature during baking or cooling.</p><p>Excess lime and silica gels also seem to be related to the formulation and even the homogenization of the raw material. In fact, the kaolin is exceptionally rich in silica which results in baking being affected by the uncombined silica gel.</p><p>SEM observation provides information on the degrees of cooking of clinker.</p>

PubMed Open Access

Redox Properties of the Disulfide Bond of Human Cu,Zn Superoxide Dismutase and the Effects of Human Glutaredoxin 1

The intramolecular disulfide bond in human Cu,Zn superoxide dismutase 1 (hSOD1) plays a key role in maintaining the protein\xe2\x80\x99s stability and quaternary structure. In mutant forms of SOD1 that cause familial amyotrophic lateral sclerosis (ALS), this disulfide bond is more susceptible to chemical reduction, which may lead to destabilization of the dimer and aggregation. During hSOD1 maturation, disulfide formation is catalyzed by the copper chaperone CCS1. Previous studies in yeast demonstrate that the yeast glutathione (GSH)/glutaredoxin redox system promotes reduction of the hSOD1 disulfide in the absence of CCS1. Herein, we further probe the interaction between hSOD1, GSH, and glutaredoxins to provide mechanistic insight into the redox kinetics and thermodynamics of the hSOD1 disulfide. We demonstrate that human glutaredoxin 1 (hGrx1) uses a monothiol mechanism to reduce the hSOD1 disulfide, and the GSH/hGrx1 system reduces ALS mutant SOD1 at a faster rate than WT hSOD1. However, redox potential measurements demonstrate that the thermodynamic stability of the disulfide is not consistently lower in ALS mutants compared to WT hSOD1. Furthermore, the presence of the metal cofactors does not influence the disulfide redox potential. Overall, these studies suggest that differences in the GSH/hGrx1 reaction rate with WT vs. ALS mutant hSOD1 and not the inherent thermodynamic stability of the hSOD1 disulfide bond may contribute to the greater pathogenicity of ALS mutant hSOD1.

redox_properties_of_the_disulfide_bond_of_human_cu,zn_superoxide_dismutase_and_the_effects_of_human_

INTRODUCTION<!>Yeast strains and growth conditions<!>Plasmids<!>Recombinant hGrx1 and hSOD1 purification<!>Non-reducing SDS-PAGE and immunoblotting techniques<!>GRX activity assays in yeast cell extracts<!>In vitro hSOD1 reduction assays<!>Measurement of hSOD1 disulfide redox potential<!>Expression of human Grx1 correlates with increased reduction of the disulfide bond in WT and ALS mutant human SOD1<!>Human Grx1 reduces the disulfide bond of human SOD1 via a monothiol mechanism<!>Measurement of the disulfide bond redox potential for WT and ALS mutant hSOD1<!>DISCUSSION<!>

<p>Cu- and Zn-containing superoxide dismutase 1 (SOD1) is a highly conserved, ubiquitous enzyme that detoxifies superoxide radicals in both the cytosol and mitochondrial intermembrane space of eukaryotic cells. Nascent SOD1 polypeptides must undergo four post-translational modifications to reach the mature form: insertion of zinc, insertion of copper, formation of an intramolecular disulfide bond between Cys57 and Cys146, and dimerization [1, 2]. The mechanism for insertion of the zinc ion is unknown; however, copper is inserted by the copper chaperone for SOD1 (CCS1) [3]. In addition to transferring copper, CCS1 also catalyzes formation of the SOD1 disulfide bond [4]. Human SOD1 can also acquire copper and form the intramolecular disulfide through a CCS1-independent pathway that requires glutathione (GSH), although this process is not completely understood [5]. Biophysical studies on purified hSOD1 demonstrate that both metal occupancy and disulfide redox state play major roles in influencing the structure and stability of the hSOD1 polypeptide. Holo, disulfide-oxidized hSOD1 forms unusually stable, active homodimers, while the inactive apo, reduced form is found predominantly as a monomer at physiological concentrations [6–8].</p><p>Despite its role as a protective enzyme, over 100 mutations in hSOD1 have been causally linked to the familial form of the common neurodegenerative disease amyotrophic lateral sclerosis (ALS) [9]. The molecular mechanism for SOD1-linked ALS pathology is still unclear; however, in vivo studies in mouse models and human cell culture have demonstrated that ALS mutant forms of hSOD1 are prone to form insoluble aggregates [10–14], which may be toxic to motor neurons. The biophysical/biochemical properties of WT and ALS mutants of hSOD1 have been compared in a number of studies in order to explain the greater aggregation propensity of ALS mutants [8, 15–20]. Collectively, these studies demonstrate that ALS mutants are more prone to disulfide reduction, unfolding/misfolding, and metal loss than WT hSOD1. These structural aberrations are proposed to promote non-native interactions between SOD1 monomers or other cellular components that lead to aggregation [15, 21]. The apo, reduced forms of ALS mutant hSOD1 are especially vulnerable to destabilization and readily aggregate under mild oxidative stress conditions [20].</p><p>Given the importance of the intramolecular disulfide in maintaining the structure and stability of hSOD1, we sought to identify factors that influence the hSOD1 disulfide redox state in vivo. Previous studies have demonstrated that the disulfide bond in human SOD1 is reduced by yeast glutaredoxin 2 (yGrx2) when expressed heterologously in a Saccharomyces cerevisiae ccs1Δ strain. Furthermore, in vitro enzyme assays indicated that hSOD1 ALS mutants are more vulnerable to disulfide reduction by yeast Grx2 than WT hSOD1, suggesting a possible role for Grxs in redox-dependent destabilization of ALS mutants [22]. Since members of the Grx family exhibit subtle structural and mechanistic differences [23, 24], we extended these studies to characterize the in vivo and in vitro interactions between human SOD1 and the human homologue of yGrx2, namely human Grx1. Human Grx1 and hSOD1 both co-localize to the cytosol and intermembrane space [25, 26], thus hGrx1 is poised to have a direct impact on the redox state of hSOD1 under physiological conditions. However, the molecular interactions between human Grx1 and WT and mutant forms of human SOD1 have not previously been addressed. The studies reported herein demonstrate that hGrx1 facilitates reduction of the disulfide bond of hSOD1 in a similar manner to yGrx2. Furthermore, we determined that hGrx1 uses a monothiol mechanism to reduce the disulfide bond in vivo and in vitro, suggesting that at least one of the disulfide cysteines of hSOD1 forms a transient mixed disulfide with GSH prior to reduction by hGrx1. We compared the reactivity and thermodynamic stability of the disulfide bond in holo and apo WT and ALS mutants A4V and G93A in vitro. These studies demonstrate that hGrx1 displays higher reactivity towards mutant hSOD1 than WT hSOD1. However, the thermodynamic stabilities of the disulfide bond in ALS mutant forms are not consistently lower than WT hSOD1. Overall, these studies suggest that hGrx1 may play a significant role in destabilization of ALS mutant hSOD1 in vivo by selectively reducing the kinetic barrier for disulfide reduction.</p><!><p>The yeast S. cerevisiae strains used in these studies are derived from the parental strain CY4 (MATa ura3-52 leu2-3 trp1-1 ade2-1 his3-11 can1-100). Strains Y117 (grx1Δ::LEU2 grx2Δ::HIS3), MC108 (ccs1Δ::ADE2), and MC120 (grx1Δ::LEU2 grx2Δ::HIS3 ccs1Δ::ADE2) were described previously [22]. Strains were maintained at 30 °C on synthetic defined medium (SD) supplemented with 2% glucose and the appropriate amino acids (US Biological). Anaerobic cultures were maintained by growth in an O2-depleted culture jar (BBL Gas Pak).</p><!><p>Human SOD1 was expressed in yeast cells under the control of the PGK1 promoter using plasmids pLC1 (WT hSOD1), pLC2 (A4V hSOD1), and pLC3 (G41D hSOD1) (2µ URA3) described previously [27]. Site-directed mutagenesis of WT hSOD1 was carried out to generate a G93A hSOD1 expression plasmid (pMD100) using the QuikChange Site-Directed Mutagenesis kit (Stratagene). Yeast expression vectors for human Grx1 (hGrx1) were created by PCR amplification of hGrx1 cDNA (Open Biosystems) and insertion of the PCR product into the NdeI and SnaBI sites in pLS108 (CEN LEU2) [28, 29], allowing for insertion of hGrx1 between the promoter and terminator for ySOD1 to create pCO202. Digestion of pCO202 with SalI and BamHI allowed for insertion of the ySOD1 promoter, the hGrx1coding sequence, and the ySOD1 terminator into pRS414 (CEN TRP1), yielding pCO204. A plasmid for overexpression of recombinant hGrx1 in Escherichia coli was created by PCR amplification of the hGrx1 cDNA with primers that introduced NcoI and EcoRI sites at the start and stop sites, respectively. The hGrx1 coding sequence was cloned into the overexpression vector pET24d to create pSB100. Site-directed mutagenesis of hGrx1 expression plasmids pCO204 and pSB100 was conducted according to the QuikChange Site-Directed Mutagenesis kit (Stratagene).</p><!><p>For production of recombinant hGrx1, pSB100 was transformed into the Escherichia coli strain BL21-CodonPlus®(DE3)-RIL (Stratagene). The cells were grown in 500 ml of ZYP-5052 autoinduction media at 37 °C for approximately five hours followed by growth overnight at 30 °C. Following centrifugation, the cell pellet was subjected to 3 freeze-thaw cycles and soluble protein was extracted with 20 mM Tris-HCl, pH 8.0, 5 mM dithiothreitol (DTT). Human Grx1 was precipitated with 40–85% ammonium sulfate and the pellet resuspended in 20 mM Tris-HCl pH 8.0 and subsequently loaded onto a desalting column followed by a HiPrep 16/10 DEAE FF column (GE Healthcare) both equilibrated with 50 mM Tris-HCl, pH 7.5, 5 mM DTT. Since human Grx1 does not bind to the DEAE column, the flow-through was collected and concentrated using an Amicon ultrafiltration apparatus. Mutant forms of hGrx1 were purified using the same protocol. The activity of purified hGrx1 was tested using a coupled enzyme assay monitoring reduction of the model substrate 2-hydroxyethyl disulfide (HED) as previously described [30].</p><p>WT and ALS mutant forms of recombinant hSOD1 were expressed in S. cerevisiae and purified as previously described [8]. WT, A4V, and G93A protein prepared by this method typically contained ~ 50% Zn and 50% Cu in the Cu binding site and 100% Zn in the Zn binding site. Preparation of apo-SOD1 with the intramolecular disulfide intact was conducted according to published methods [8]. The apo forms prepared by this method typically contained < 0.05 Cu and Zn per monomer as determined by inductively coupled plasma atomic emission spectroscopy (ICP-AES).</p><!><p>Protein concentrations were determined by the Bradford assay using bovine serum albumin as the standard. Disulfide oxidation of hSOD1 in yeast cells was monitored by non-reducing SDS-PAGE and immunoblotting using iodoacetimide (IAM) or N-ethylmaleimide (NEM) to alkylate free thiols as previously described (also known as a redox western blot) [31–33]. Briefly, yeast strains were grown at 30 °C to mid-log phase in glucose SD medium without shaking. For IAM alkylation, the cells are washed and lysed with glass beads in 600 mM sorbitol, 10 mM HEPES, pH 7.5, 100 mM IAM, containing 1:100 dilution of protease inhibitor cocktail (Sigma). Oxidized and reduced proteins were separated using non-reducing SDS-PAGE by loading 50–100 µg cell extracts on pre-cast 14% Tris-glycine gels (Invitrogen). For NEM alkylation, 5 OD600 units of cells were acid-quenched with trichloroacetic acid (TCA) (15% w/v final concentration) and incubated on ice for 20 minutes. Pellets were resuspended in 1 ml 10% TCA and lysed by glass beads. The supernatant was aspirated off and the pellet was resuspended in 500 µl 1 SDS loading buffer containing 40 mM NEM and incubated on ice for 10 minutes prior to non-reducing SDS-PAGE. We found that NEM and IAM alkylation produced similar results. After electrophoresis, separated proteins were subjected to in-gel reduction prior to transfer to nitrocellulose membranes as previously described [33]. For analysis of hGrx1 levels in yeast cells, glass bead extracts were prepared as described above for hSOD1 in the absence of IAM. Total cell lysate (50 – 100 µg) was separated by reducing SDS-PAGE on 14% Tris-glycine gels (Invitrogen) and transferred to nitrocellulose for western blotting. Human SOD1 was detected using an anti-hSOD1 antibody (1:5000) kindly provided by Valeria Culotta [33], while hGrx1 was detected using an anti-Grx1 antibody (1:10000) from Abcam. For both hGrx1 and hSOD1, a secondary anti-rabbit IRDye 800CW antibody diluted to 1:30000 was used for detection via the Odyssey Infrared Imaging System according to the manufacturer's instructions (LI-COR, Lincoln, NE).</p><!><p>GRX activity is yeast cell lysates was measured using published protocols [22]. Briefly, yeast cell lysates were prepared by glass bead lysis and heated to 85 °C to inactivate glutathione reductase and thioredoxin reductase. GRX activity in the heat-treated lysates was then measured using a coupled enzyme-HED assay.</p><!><p>An in vitro assay to monitor reduction of hSOD1 by hGrx1 was adapted from previous methods [22]. Briefly, disulfide-oxidized hSOD1 was diluted to 3 µM in a 300-µl reaction mixture that included 100 mM Tris-HCl, pH 8.0, 2.0 mM EDTA, 1 mM GSH, 0.2 mM NADPH, and 6 µg/mL glutathione reductase. Varying concentrations of GSH were also tested from 0 to 5.0 mM GSH. Human Grx1 (WT or mutant form) was added (50 nM final concentration) to the reaction mixture and incubated at 30 °C for different time periods. At each allotted time, 15-µl aliquots are removed from the stock and 15 µl of 2X sample buffer (160 mM Tris-HCl pH 6.8, 20% glycerol, 5% SDS, 1.0 mg bromophenol blue) with 100 mM IAM are added and incubated at 37 °C for 30 minutes prior to non-reducing SDS-PAGE. Reduced and oxidized forms of hSOD1 were analyzed by quantitative immunoblot with anti-hSOD1 antibodies or Coomassie staining using an Odyssey Infrared Imaging System.</p><!><p>The redox potential of the disulfide bond of WT and ALS mutant hSOD1 was determined by incubating the disulfide-oxidized protein in buffers poised at defined redox potential values as previously described [34]. Briefly, 10 µg/ml apo, oxidized SOD1 was incubated with mixtures of reduced DTT and the oxidized form (trans-4,5-dihydroxy-1,2-dithiane) at a total concentration of 2 mM in 50 mM MOPS pH 7.0 (purged with nitrogen) in a Coy anaerobic chamber at 30 °C. The samples were incubated for 12-60 hrs in order for the redox state of the protein to reach equilibrium with the buffer. After incubation, 20% TCA was added for a final concentration of 10% and the samples precipitated on ice for 30 minutes under anaerobic conditions. Following centrifugation, the pellet was resuspended in 2X SDS sample buffer with 100 mM IAM. The samples were incubated at room temperature (25 °C) for 30 minutes, diluted to 1X with water, and separated by non-reducing SDS-PAGE. The redox potential values were calculated by fitting the data to the Nernst equation for a two-electron process using non-linear least squares fitting [35].</p><!><p>To determine whether hGrx1 influences the disulfide bond of hSOD1 in vivo, we used a yeast expression system to manipulate intracellular redox factors by gene deletion and heterologous protein expression. The disulfide-oxidized and reduced forms of hSOD1 have different mobilities by non-reducing SDS-PAGE, allowing efficient separation and quantification of the two forms (Figure 1). In WT yeast cells expressing the yeast copper chaperone for SOD1 (CCS1) and both yeast dithiol glutaredoxins (GRX1 and GRX2), the intramolecular disulfide in WT hSOD1 is primarily oxidized (Figure 1A, lane 2). Deletion of GRX1 and GRX2 has little effect on the redox state of hSOD1 (lane 3). However, in a ccs1Δ strain, the pool of WT hSOD1 is shifted to the reduced form (Figure 1A, lane 4), confirming that yCCS1 plays an important role in controlling the redox state of hSOD1 as previously reported [22, 32] (Fig 1B). In the grx1Δ grx2Δ ccs1Δ triple mutant strain (Fig 1A, lane 5), hSOD1 is slightly more oxidized than in the ccs1Δ strain (lane 2), which is consistent with previous reports that yCCS1 and the dithiol yGrxs play opposing roles in controlling the redox state of hSOD1 [22]. To test whether hGrx1 influences the redox state of hSOD1, grx1Δ grx2Δ ccs1Δ cells were transformed with a low copy plasmid expressing hGrx1. We confirmed that recombinant WT hGrx1 is active in yeast extracts using a standard assay for Grx activity (Figure 2B). Our data show that expression of hGrx1 (Figure 1A, lane 6) correlates with a shift towards the reduced form of WT hSOD1 as demonstrated for yGrx2 [22]. However, this effect is only observed in a ccs1Δ strain since expression of hGrx1 had no effect on the steady-state redox state of hSOD1 in WT and grx1Δ grx2Δ yeast strains (data not shown).</p><p>We next tested how hGrx1 expression influenced the disulfide bond in ALS mutant forms of hSOD1. A4V, G41D, and G93A hSOD1 mutants are all enzymatically active "β-barrel mutants," and A4V and G93A have been shown to produce an ALS phenotype in transgenic rodent models [36–38]. Unlike WT hSOD1, the disulfide of A4V hSOD1 is largely reduced in WT yeast cells (Figure 1A,B). Interestingly, deletion of the dithiol Grxs has an obvious effect on the redox state of this disulfide, shifting it to the oxidized form. In contrast, G41D and G93A hSOD1 are largely oxidized in WT cells and deletion of GRX1/2 has little effect on the redox state of the disulfide. The three mutant forms, A4V, G41D, and G93A hSOD1, display a similar pattern of redox state changes as WT hSOD1 in response to deletion of ccs1 alone and in combination with grx1/2. However, G93A hSOD1 is more reduced under these conditions in comparison to A4V and G41D (Figure 1A, lanes 4,5). Furthermore, expression of hGrx1 in the triple ccs1Δ grx1Δ grx2Δ mutant was found to shift A4V, G41D, and G93A hSOD1 to more reduced states (Figure 1A, lane 6). These data suggest that hGrx1 acts similarly to yGrx2 by promoting reduction of the disulfide in both WT and ALS mutant hSOD1 in the absence of CCS1 activation.</p><!><p>After determining that expression of hGrx1 impacts the in vivo redox status of the hSOD1 intramolecular disulfide, we tested whether one or both active site cysteines in hGrx1 were essential for reduction. Dithiol Grxs have two cysteines in the CXXC active site that display different reactivities. Grxs that utilize a dithiol mechanism to reduce intramolecular disulfide bonds require both cysteines, while Grxs that catalyze glutathionylation/deglutathionylation reactions via a monothiol mechanism require only one active site cysteine [39]. Using a variety of model substrates, human Grx1 was shown to favor reduction of glutathione-containing mixed disulfides via a monothiol mechanism [40]. The N-terminal Cys23 is essential for this catalytic function while the C-terminal Cys26 is dispensable [41]. Human Grx1 has no intramolecular disulfide targets identified thus far, and hSOD1 was proposed to be one of those targets [22]. To test whether a dithiol or monothiol mechanism is required for reduction of hSOD1 in vivo, we expressed active site hGrx1 Cys mutants in yeast and monitored the Grx activity and redox state of WT and ALS mutant hSOD1 (Figure 2). The redox state of both WT and ALS mutant hSOD1 is more reduced in strains expressing hGrx1 constructs (WT and C26S) that are active in the enzyme coupled-HED Grx assay (Figure 2A, lanes 2,4). These data indicate that only the N-terminal active site cysteine of hGrx1 (Cys23) is required for reduction of hSOD1, consistent with a monothiol mechanism.</p><p>To more directly demonstrate that hGrx1 reduces the disulfide bond of hSOD1 via a monothiol mechanism, we used an in vitro assay to monitor the redox state of purified hSOD1 upon incubation with purified hGrx1 and GSH. Previous studies have shown that yGrx2 specifically reduces the disulfide bond in the apo form of A4V hSOD1, but not holo A4V hSOD1 or apo or holo WT hSOD1 [22]. We have extended these studies to examine the effects of WT and Cys mutants of hGrx1 on WT, A4V, and G93A hSOD1. Recombinant WT and active site mutants of hGrx1 were overexpressed and purified from E. coli and their respective activities verified (Supplemental Figure S1). The assay mixture containing GSH, purified hGrx1, and WT or ALS mutants of hSOD1 was removed at different time points and the hSOD1 disulfide redox state assessed by non-reducing SDS-PAGE (Figure 3A). These results demonstrate that GSH/hGrx1 exhibits no activity towards the Cu/Zn forms of WT and G93A mutant hSOD1 (Figure 3B–D, top), and low activity toward Cu/Zn A4V hSOD1, as previously demonstrated for yeast Grx2 [22]. In contrast, the apo forms of WT, A4V, and G93A all displayed increased reduction by GSH/hGrx1 (Figure 3B–D, bottom) in comparison to the holo forms. In each case we found that WT and C26S hGrx1 exhibited similar reactivity towards the hSOD1 disulfide, while the C23S and C23,26S hGrx1 mutants were similar to the control with no hGrx1 added. Thus, hGrx1 uses a monothiol mechanism involving Cys23 to reduce the disulfide in both WT and ALS mutants of hSOD1.</p><p>By comparing the rate of reduction of the apo forms of WT vs. ALS mutant SOD1 (Figure 3B–D, bottom), it is clear that the ALS mutants are reduced at a faster rate than WT hSOD1. These differences are apparent in the first 15 minutes of the assay in which WT hSOD1 is only ~10% reduced, while A4V and G93A are 50% and 65% reduced, respectively. Even with inactive hGrx1 (C23S, C23,26S) and no hGrx1 added (control), the intramolecular disulfide in A4V and G93A hSOD1 is more susceptible to reduction by GSH alone, which is consistent with previous reports [17]. The greater susceptibility of apo mutant hSOD1 to GSH reduction may reflect higher reactivity or greater accessibility of the hSOD1 disulfide to nucleophilic attack by GSH (see Discussion). Here we demonstrate that hGrx1 with a single active site Cys accelerates reduction of hSOD1 in the presence of GSH. To confirm the dependence of disulfide reduction on GSH, the assay for apo G93A hSOD1 was also performed with varying GSH concentrations (Supplemental Figure S2). The results indicate a hyperbolic dependence of hSOD1 reduction on [GSH] as reported for other redox reactions catalyzed by hGrx1 [24, 41].</p><!><p>In addition to kinetic factors, the favorability of a redox reaction is also dictated by the reaction thermodynamics [42]. After determining that WT and ALS mutant SOD1 exhibit differential rates of reduction by GSH/hGrx1, we next investigated the thermodynamic stability of the intramolecular disulfide bond of apo and holo WT, A4V, and G93A hSOD1. We measured the redox potential of the disulfide bond by incubating the proteins in redox buffers poised at defined redox potential values under anaerobic conditions [32, 34]. Surprisingly, the redox potential of WT hSOD1 (−301 mV) is virtually identical with and without the metal cofactors as shown in Figure 4 and Table 1. A similar pattern is observed for the apo and holo forms of A4V and G93A hSOD1 (Figure 4B, Table 1). However, we noted that the disulfide redox potentials for the two ALS mutant hSOD1 proteins are significantly different from WT hSOD1, but not similar to each other. The redox potentials of the apo and holo forms of G93A hSOD1 (−315 mV) are lower than WT hSOD1, while the redox potentials of the A4V hSOD1 forms (−282 mV) are higher (Table 1). Thus, the disulfide bond in G93A hSOD1 is more thermodynamically stable than WT hSOD1, while the A4V disulfide is less stable. Since A4V and G93A hSOD1 are both implicated in ALS development, these data suggest that differences in the redox potential of the hSOD1 disulfide may not be a significant factor contributing to the toxicity of hSOD1 mutants. Rather it appears that the more favorable redox kinetics of the mutant forms (i.e. the higher reactivity of the disulfide bond) may be more influential.</p><!><p>The redox state of the disulfide bond in hSOD1 plays a critical role in the stability and oligomeric state of the protein. Increased levels of apo, disulfide-reduced SOD1 correlates with increased misfolding and aggregation of SOD1 in both in vitro and in vivo studies [8, 17, 18, 20, 22, 43–45]. Thus we sought to identify intracellular factors that influence oxidation/reduction of this disulfide bond. A previous study demonstrated that overexpression of Grx1 in mouse cells lines increases the solubility of ALS mutant hSOD1, presumably via reduction of intermolecular disulfides found within SOD1 aggregates [46]. Although in that case, Cys111 is implicated in formation of these disulfide-linked oligomers rather than the two cysteines that form the intramolecular disulfide bond (Cys57 and Cys146). Here we provide in vivo and in vitro evidence that the SOD1 intramolecular disulfide is a substrate for human Grx1. Our results demonstrate that hGrx1 specifically reduces the intramolecular disulfide bond of WT, A4V, G41D, and G93A ALS mutants in a yeast expression system. For WT, G41D, and G93A hSOD1, this effect requires the absence of CCS1, confirming that CCS1 and Grxs play opposing roles in redox control of the SOD1 disulfide [22]. However, control of the hSOD1 redox state is dominated by CCS1 for these forms since Grx expression has little effect on steady-state levels of the hSOD1 disulfide in the presence of CCS1 in vivo. This result is consistent with our in vitro redox kinetics experiments demonstrating that holo-hSOD1 is more resistant to reduction by hGrx1 than apo-hSOD1. In CCS1+ cells, a higher percentage of hSOD1 is Cu-loaded [32, 47], since CCS1 catalyzes both Cu insertion and disulfide bond formation [48]. Thus, a larger pool of holo-hSOD1 will be kinetically inert to reduction by GSH/hGrx1 in CCS1+ cells. However, we do note that A4V hSOD1 is predominantly reduced in the presence of CCS1 and the dithiol Grxs, suggesting that the dithiol Grxs can successfully compete with CCS1 in controlling the A4V hSOD1 redox state in vivo. Interestingly, this mutant is the only one tested that showed some reactivity with hGrx1 in the holo form in vitro (Figure 4C), providing a possible explanation for its more reduced state in vivo, even in the presence of Cu-loading by CCS1. In addition, the in vivo redox state differences between WT hSOD1 and the various ALS mutants tested likely also reflect differences in the ability of CCS1 to associate and facilitate copper insertion and disulfide formation in the variant forms of hSOD1 [9].</p><p>Interestingly, we find that hGrx1 has similar apparent effects on the steady-state redox status of both WT and mutant hSOD1 in vivo, since all four hSOD1 forms tested are shifted to a more reduced state with the addition of hGrx1 in ccs1Δ grx1/2Δ cells (Figure 1). However, the in vitro redox kinetic assays clearly demonstrate that GSH/hGrx1 reduces WT hSOD1 at a much slower rate than the ALS mutant forms (Figure 3). These seemingly conflicting results may be explained by that fact that the in vivo redox state of hSOD1 captured by the thiol-trapping method is a snapshot of the steady-state redox status of the protein. In addition to the rate of disulfide reduction, a number of other factors may influence the steady state redox status of hSOD1. These include the rate of CCS1 oxidation and Cu insertion, the rate of metal loss from the active sites, and the rate of protein turnover for oxidized vs. reduced forms. Previous studies have demonstrated that A4V and G41D hSOD1 proteins are more susceptible to degradation when expressed in the yeast model system than WT hSOD1 [22]. In particular, the reduced form of A4V hSOD1 is more rapidly degraded than the oxidized form, while both the oxidized and reduced forms of WT hSOD1 exhibit much slower turnover rates in vivo [22]. Thus, the pool of reduced mutant hSOD1 generated via GSH/hGrx1 reduction or deficient CCS1-catalyzed maturation may be rapidly degraded in vivo and thus underrepresented in the steady-state redox state measurements. Nevertheless, these results clearly demonstrate that GSH and hGrx1 are additional intracellular factors that have a significant impact on the disulfide redox state of hSOD1 in vivo.</p><p>As mentioned above, the in vitro studies strongly indicate that the ALS mutant forms of hSOD1 are more susceptible to reduction by the GSH/hGrx1 system than WT hSOD1, and furthermore that loss of the metal cofactors greatly accelerates this reaction. In general, Grx proteins catalyze thiol-disulfide exchange reactions that utilize one or both cysteines in the active site CXXC motif. These reactions can proceed via a dithiol mechanism, which reduces intramolecular disulfides, or a monothiol mechanism reducing a mixed disulfide bond with GSH [39]. Our results indicate that hGrx1 reduces the disulfide bond of hSOD1 via a monothiol mechanism that only requires the N-terminal active site cysteine of hGrx1. Thus, reduction of the hSOD1 disulfide first proceeds via attack of the hSOD1 disulfide by GSH followed by deglutathionylation catalyzed by hGrx1 (Scheme 1). We note that GSH alone can also facilitate complete reduction of apo A4V and G93A hSOD1 (Figure 3C,D). In the absence of hGrx1, glutathionylated hSOD1 may be fully reduced via reaction with a second GSH molecule, or the hSOD1 disulfide may re-form via nucleophilic attack by the free cysteine in hSOD1. Since reduction of hSOD1 is faster in the presence of hGrx1, our data suggest that hGrx1 is more efficient at reducing the hSOD1-SSG mixed disulfide than GSH itself. Glutathionylation of the hSOD1 disulfide prior to reduction by GSH or hGrx1 is likely a transient modification since a stable GSH adduct with either of the disulfide cysteines (Cys57 and Cys146) has not been detected in vivo [22, 49]. However, we do note that Cys111, which is solvent exposed in WT and mutant holo hSOD1, has been shown to form stable GSH mixed disulfides [49]. In addition, all four cysteines in hSOD1 have been implicated in formation of intermolecular disulfide bonds in higher order hSOD1 aggregates [11, 37, 45, 46, 50, 51].</p><p>The accessibility of the disulfide bond may be a significant factor controlling the differences in disulfide reactivity of apo vs. holo, WT vs. ALS mutant hSOD1. Both disulfide formation and metal insertion drive dimerization of hSOD1 [6–8, 52]. Once formed, the disulfide bond in WT hSOD1 is partially buried near the dimer interface [43]. In the case of the holo proteins, only A4V hSOD1 showed some reactivity with GSH/hGrx1 in the experimental time frame. This mutation is located near the dimer interface and is thus suspected to weaken the dimer interaction [38, 53]. Increased monomerization will likely expose the disulfide to attack by GSH and subsequent deglutathionylation by hGrx1. In addition, A4V hSOD1 has a 30-fold lower Zn(II) affinity that WT hSOD1 and is thus more prone to demetallation [54, 55]. Loss of zinc binding is another factor that promotes monomerization and has been shown to increase the lability of the disulfide bond [56]. Thus it is not surprising that all three apo proteins tested showed some reactivity towards GSH alone that was accelerated with the addition of hGrx1. For WT hSOD1, removal of the metals loosens the structure by increasing the mobility of the electrostatic and zinc-binding loops [57, 58], but does not disrupt dimer formation if the disulfide is intact [6, 8]. This greater structural flexibility may allow limited access to the disulfide located near the dimer interface. However, unlike apo, oxidized WT hSOD1, apo, oxidized A4V and G93A ALS mutants are prone to monomerization [38], which may explain the even greater reactivity of the disulfide bond in these demetallated, mutant forms.</p><p>In addition to cysteine reactivity, another important factor that influences thiol-disulfide redox regulation is thermodynamic stability. How does the thermodynamic stability of the hSOD1 disulfide differ for apo and holo forms of WT and ALS mutant hSOD1? To answer this question, we measured the midpoint redox potentials for apo and holo forms of recombinant WT, A4V and G93A hSOD1. We found that apo and holo WT hSOD1 had virtually identical redox potentials (−301 and −302 mV), demonstrating that the metallation state of hSOD1 does not influence the redox potential. Interestingly, this value is significantly lower than the reported value for WT hSOD1 purified from E. coli (−248 mV) [32], indicating higher thermodynamic stability for the disulfide in yeast-purified hSOD1. There is one notable structural difference between hSOD1 purified from yeast vs. E. coli that may account for these different potentials: recombinant hSOD1 expressed in yeast is acetylated at the N-terminus similar to native hSOD1, while hSOD1 expressed in E. coli is not [59]. Although the acetylation site is ~ 20 Å from the disulfide bond, it is possible that this modification has a subtle effect on the protein structure that significantly influences the redox potential; for example, by altering the local electrostatic environment of the disulfide, reducing strain within the disulfide bond, or reducing strain within the overall protein structure. In addition, the N-terminus is located at the dimer interface and thus N-terminal acetylation may impact the quaternary structure, which in turn may affect the strength of the disulfide bond.</p><p>Similar to WT hSOD1, our results also demonstrate that the metallation state of ALS mutants does not significantly alter the redox potential for these proteins. However, the measured redox potentials for A4V hSOD1 (−282 mV) and G93A hSOD1 (−315 mV) were higher and lower than WT hSOD1, respectively (Table 1). Thus, the presence of ALS mutations in hSOD1 does not uniformly destabilize the disulfide bond. Our results investigating the kinetic and thermodynamic properties of the disulfide bond in WT and ALS mutant hSOD1 parallel recent studies comparing the thermodynamic stability and folding kinetics of the overall protein structure. Not all ALS mutants are more unstable than WT hSOD1 in the apo state, since some mutants exhibit similar or even higher thermodynamic stability than WT hSOD1 [60]. However, the folding kinetics for ALS mutants are consistently slower than WT hSOD1 [61]. A similar pattern emerges for the redox state of the disulfide bond. We find that the thermodynamic stability of the disulfide in ALS mutants is not consistently lower that WT hSOD1. However, the reduction rate of the disulfide is consistently faster for ALS mutants. Although we have only tested two ALS mutants in this study, our results are in line with previous reports demonstrating that a large variety of ALS mutants are more susceptible to disulfide reduction by GSH than WT hSOD1 [17]. We also demonstrate that hGrx1 plays a significant role in accelerating this reduction via a monothiol catalytic mechanism. Overall, these results suggest that the reaction kinetics between hSOD1 and GSH/Grxs may be a critical factor influencing the pathogenicity of ALS mutant hSOD1. However, in order for the reduction of hSOD1 to proceed via the GSH/Grx system in vivo, the reaction must be thermodynamically favorable. In vivo redox measurements in yeast and mammalian cells confirm that the redox potential of the cytosolic GSH:GSSG pool (−290 to −300 mV) [31, 62, 63] is very similar to the hSOD1 disulfide redox potential, and thus poised to facilitate at least partial reduction of both WT and ALS mutant hSOD1 under physiological conditions.</p><!><p> AUTHOR CONTRIBUTION </p><p>Samantha Bouldin and Maxwell Darch both designed the research and performed the experimental work. Samantha Bouldin wrote the first draft of the paper and Maxwell Darch worked on subsequent drafts. P. John Hart provided the purified hSOD1 proteins and edited the manuscript. Caryn Outten supervised the research, analyzed the data, and prepared the manuscript.</p>

PubMed Author Manuscript

Ribosome Subunit Stapling for Orthogonal Translation in E. coli

The creation of orthogonal large and small ribosomal subunits, which interact with each other but not with endogenous ribosomal subunits, would extend our capacity to create new functions in the ribosome by making the large subunit evolvable. To this end, we rationally designed a ribosomal RNA that covalently links the ribosome subunits via an RNA staple. The stapled ribosome is directed to an orthogonal mRNA, allowing the introduction of mutations into the large subunit that reduce orthogonal translation, but have minimal effects on cell growth. Our approach provides a promising route towards orthogonal subunit association, which may enable the evolution of key functional centers in the large subunit, including the peptidyl-transferase center, for unnatural polymer synthesis in cells.

ribosome_subunit_stapling_for_orthogonal_translation_in_e._coli

<p>The ribosome is a large molecular machine, universally composed of two subunits, that decodes non-overlapping triplet codons in mRNAs for the encoded polymerization of amino acids into proteins.1 The small subunit, containing 16S rRNA, binds mRNA and decodes the interaction between codons on mRNAs and their cognate tRNA anticodons, and the large subunit, containing 23S rRNA, facilitates many functions, including peptide bond formation. While natural translation encodes the polymerization of the canonical 20 amino acids, extensions of translation for the polymerization of unnatural building blocks will unlock routes to encode and evolve new classes of polymers. However, because the ribosome is essential for proteome synthesis and many mutations in the ribosome are dominant-negative or lethal in the cell,2 it is challenging to alter and evolve the natural ribosome for unnatural polymer synthesis in cells.</p><p>To address the challenge of creating an evolvable ribosome, we have previously created orthogonal (O)-ribosome-O-mRNA pairs (Figure 1) in E. coli.3 The O-ribosome contains a mutated anti-Shine–Dalgarno (ASD) sequence within its O-16S rRNA, enabling O-ribosomes to selectively and efficiently translate O-mRNAs bearing the orthogonal Shine–Dalgarno (O-SD) sequences. Likewise, O-mRNAs are not translated by endogenous ribosomes. Because the orthogonal ribosome, unlike the natural ribosome, is not responsible for synthesizing the proteome, its O-16S rRNA may be evolved to perform new functions. We have previously evolved ribo-X in which the decoding center, within the O-16S rRNA of the orthogonal ribosome, no longer recognizes release factor 1, thereby enabling efficient incorporation of unnatural amino acids in response to the amber stop codon.4 We have also evolved ribo-Q, which uses extended anticodon tRNAs to efficiently incorporate unnatural amino acids in response to diverse quadruplet codons, enabling the site-specific incorporation of multiple distinct unnatural amino acids into recombinant proteins.5</p><p>An orthogonal 16S (O-16S) rRNA contains an orthogonal anti-Shine–Dalgarno (O-ASD) sequence, which confers specificity to the small subunits that contain an O-16S rRNA to translate orthogonal mRNAs (O-mRNA) that bear orthogonal Shine–Dalgarno (O-SD) sequences. With no further components added to the cell, this O-16S rRNA shares endogenous 23S rRNA with endogenous 16S rRNA (left). Creating an orthogonal 23S rRNA that specifically functions with an orthogonal 16S rRNA (that does not function with endogenous 23S rRNA), will enable an altered 23S rRNA to be insulated from cellular translation and selectively used in orthogonal translation (right).</p><p>Many key ribosomal functions, including interactions with tRNAs and elongation factors, peptide bond formation in the peptidyl-transferase center (PTC), and the folding and release of the nascent chain through the exit tunnel,1, 6 are mediated by 23S rRNA within the large subunit. These functional centers cannot be evolved in the current orthogonal ribosome that uses the endogenous pool of large subunits, containing 23S rRNA, in combination with the orthogonal small subunit, containing O-16S rRNA, to translate the O-mRNA (Figure 1). Creating an O-23S rRNA that assembles into an orthogonal large subunit and is specifically coupled to the orthogonal small subunit, containing O-16S rRNA, will enable the creation of orthogonal ribosomes in which both subunits are selectively recruited to an orthogonal message (Figure 1). This will facilitate alteration and evolution of functional centers in the O-23S rRNA not possible on the endogenous 23S rRNA.</p><p>The large and small ribosomal subunits interact through non-covalent RNA–RNA interactions between 16S rRNA and 23S rRNA that bury approximately 6000 Å2, and these interactions are dynamically regulated through the translation cycle.7 Efforts to control non-covalent subunit interactions through rRNA mutagenesis have proved unsuccessful thus far. Here we investigate the creation of an orthogonal ribosome in which the O-16S rRNA is covalently attached to a 23S rRNA to create a fused rRNA (Figure 2 A). The fused rRNA assembles into a new orthogonal ribosome that translates an orthogonal message and permits mutagenesis of the 23S rRNA.</p><p>Ribosome subunit stapling. A) Illustration of the stapled ribosome. B) 1-D representation of the designed rDNA operon. The 16S gene is separated into two halves (16S-5′ and 16S-3′) by the insertion of 23S rDNA. The 23S gene is circularly permuted such that a 3′-terminal segment (23S-3′) precedes its 5′-terminal segment (23S-5′). This transcript (ca. 4500 nt) is flanked by the native 16S processing sites (flanks). The 23S processing sites (which would normally liberate the 23S rRNA into a separate RNA molecule) have been deleted. C) 2-D representation of the stapled ribosome, illustrating the linker sequence used to staple helix 44 (h44) of 16S rRNA to Helix 101 (H101) of 23S rRNA into a single rRNA molecule.</p><p>We envisioned joining the two subunits by reorganizing the rrnB operon such that a 23S rRNA would be nested within the 16S rRNA as a large insertion (Figure 2 B). We were encouraged by previous observations that in various organisms 16S rRNAs can exist in multiple fragments or with long insertions.8 Moreover, the 23S rRNA is tolerant to circular permutation,9 indicating that it might be possible to circularly permute the 23S rRNA to open up new 5′ and 3′ termini at positions proximate to surface exposed features of the 16S rRNA, and then insert this permuted 23S at that site on the 16S, connected on both ends by an RNA linker (Figure 2 C).</p><p>We used high-resolution structures of E. coli ribosomes7a, 10 and phylogenetic variation11 in rRNA sequence to identify regions of 23S rRNA and 16S rRNA that come close in space, and may be tolerant to insertion (Supporting Information, Figure S1 A). This analysis identified Helix 101 (H101) on the 23S and helix 44 (h44) on the 16S as an excellent pair of sites to test our strategy (Figure S1 B). These helices come into close contact (3 nm) in ribosome structures,7a, 10 and are tolerant to insertions as judged by their natural phylogenetic variation11b and previous genetic engineering.8b Moreover, these helices are distal from the corridor through which tRNAs transit and elongation factors dock (Figure S1 B). Taking a rational structure-based approach, we opted to circularly permute 23S at H101 and insert it within 16S, at the terminal loop of h44 (Figure 2 C). We linked the 16S and 23S sequences via the J5/J5a region from the Tetrahymena group I self-splicing intron (Figure 2 C), an RNA hinge that can toggle between an extended and "U-turning" form.12 This "stapled" ribosome rDNA was synthesized by overlap extension PCR (Figure S2, Table S1), cloned into a pRSF plasmid following an inducible Ptac promoter, and given an orthogonal ASD (O-ASD) via site-directed mutagenesis.3 We refer to the resulting construct as pRSF-O-ribo(h44H101).</p><p>Because the unusual topology of the O-ribo(h44H101) rRNA could complicate ribosome folding and assembly pathways,13 it was critical to ascertain the extent to which pRSF-O-ribo(h44H101) produces a full-length rRNA that persists in vivo. To address this question we probed RNA extracted from E. coli expressing O-ribo(h44H101) by northern blot using a biotinylated probe specific to the O-ASD sequence of the orthogonal ribosome (Figure 3 A). We detected a single band at 4500 nt, demonstrating that the major species bearing an O-ASD, in cells transformed with pRSF-O-ribo(h44H101), is the full length O-ribo(h44H101) rRNA. These data suggest that translation of O-mRNAs in cells bearing pRSF-O-ribo(h44H101) results from the activity of the stapled ribosome. In control experiments, RNA extracted from cells expressing the orthogonal ribosome from pRSF-O-Ribo (a plasmid with the same copy number, encoding orthogonal ribosomes under the same promoter, but with wild-type operon topology) was probed in a northern blot with the O-ASD-specific probe. In this experiment we detected a band at 1500 nt, as expected for the 16S rRNA (Figure 3 A), and the intensity of this band was approximately four times that of the band detected for O-ribo(h44H101) rRNA (Figure 3 B). These data suggest that either the O-ribo(h44H101) rRNA is not transcribed as efficiently as the rrnB operon with native topology and/or a fraction of the transcript does not assemble correctly and is ultimately degraded.</p><p>rRNA derived from pRSF-O-ribo(h44H101) is assembled into functional ribosomes in vivo. A) Northern blot using a probe specific to the O-ASD sequence detects RNAs from total RNA extracts that bear O-ASDs. Wild-type (WT) E. coli, which does not possess orthogonal ribosomes, generates no band. O-ribo possesses an O-ASD on the 16S rRNA, and generates a band at 1500 nt (the length of 16S rRNA). pRSF-O-ribo(h44H101) generates an O-ASD-containing band near 4500 nt (nucleotides). B) Increasing isopropyl β-d-thiogalactopyranoside (IPTG) concentrations up to 0.1 mm increases the expression of orthogonal ribosomes (blue and black traces). Stable O-ribo(h44H101) rRNA is generated to about 25 % the levels of O-ribo (red trace). C) Growth curves of E. coli bearing an O-cat reporter, with or without O-ribo(h44H101), at 37 °C in liquid LB media supplemented with IPTG and chloramphenicol (Cm).</p><p>To test the activity of O-ribo(h44H101) in protein translation we co-transformed pRSF O-ribo(h44H101) and an O-cat reporter in which a chloramphenicol acetyltransferase gene (cat) is downstream of an O-SD site for ribosome binding.3, 4 Following induction of rRNA synthesis with IPTG, we followed the growth of cells in different concentrations of chloramphenicol (Cm) to assess the activity of O-ribo(h44H101).</p><p>Cells bearing the O-cat reporter alone, or provided with pRSF-O-ribo but not induced with IPTG, do not grow on 10 μg mL−1 Cm (Figure 3 C; Supporting Information, Figure S3, Tables S2, S3). In contrast, when cells are provided with pRSF-O-ribo(h44H101) and O-cat they grow robustly on Cm concentrations up to 70 μg mL−1 (Figure 3 C, Table S4), indicating that pRSF-O-ribo(h44H101) directs the synthesis of ribosomes that specifically translate the orthogonal message. The activity of O-ribo(h44H101) in the assay is lower than that of O-ribosomes with independent subunits produced from a standard operon, which confer Cm resistance up to 200 μg mL−1, but not 300 μg mL−1, in our assay (Figure S4, Table S5). We further demonstrated the activity of O-ribo(h44H101) in an independent assay by measuring its ability to translate O-luciferase (a luciferase gene expressed from an orthogonal ribosome binding site),4 as measured by a luciferase activity assay (Figure S5). This led to results that are quantitatively consistent with our observations in the chloramphenicol resistance assay.</p><p>To investigate whether the activity of the O-ribo(h44H101) is dependent on the stapled 23S rRNA (Figure 4A, B) we introduced two mutations (G2252A and G2553C) into the 23S portion of O-ribo(h44H101), creating O-ribo(h44H101(G2252A)) and O-ribo(h44H101(G2553C)). The guanosines targeted for mutation base pair with the universally-conserved 3′-CCA ends of tRNAs and their mutation is reported to severely hinder protein synthesis.14 When O-ribo(h44H101(G2252A)) and O-ribo(h44H101(G2553C)) were co-transformed with O-cat, cells failed to grow on 30 μg mL−1 Cm after 20 h (Figure 4 A; Supporting Information, Tables S6, S7), while O-ribo(h44H101) grew robustly on 30 μg mL−1 Cm (Figure 3, Figure 4 A) and continued to survive on Cm concentrations up to 70 μg mL−1 (Figure 3 C). These data are consistent with the un-mutated large subunit of O-ribo(h44H101) being functional and important in orthogonal translation.</p><p>Investigating the orthogonality of the small and large subunit portions of O-ribo(h44H101) with respect to endogenous ribosome subunits in translation. A) Growth curves of E. coli bearing an O-cat reporter and O-ribo(h44H101), with or without mutations that inactivate the large subunit portions for translation, at 37 °C in liquid LB media supplemented with IPTG and Cm. Inactivating the stapled large subunit results in a decrease in Cm resistance. B) For orthogonally associating ribosome subunits, the small subunit portion of O-ribo(h44H101), after forming an initiation complex with an O-cat mRNA, would be primarily captured by the stapled large subunit (solid arrow) and not the endogenous large subunit (dashed arrow). C) Growth curves of E. coli bearing O-ribo(h44H101), with or without mutations in the large subunit portion that confer a dominant-lethal phenotype in cells, at 37 °C in liquid LB. Red curves +IPTG; black curves −IPTG. D) For orthogonally associating ribosome subunits, dominant-negative mutations in the large subunit portion would not substantially affect cell fitness, and endogenous small subunit-mRNA complexes would minimally capture the large subunit (dashed arrow).</p><p>Although G2252A and G2553C in 23S rRNA are reported to be dominant-negative when expressed in cells,14 they were readily introduced into O-ribo(h44H101) by site-directed mutagenesis. Moreover, the reduction in growth imposed by these mutant ribosomes (O-ribo(h44H101(G2252A)) and O-ribo(h44H101(G2553C))), with respect to O-ribo(h44H101), was small, even with maximum IPTG induction of rRNA expression (Figure 4 C, Supporting Information, Tables S8–S10).</p><p>These data indicate that the mutations do not have a substantial dominant-negative effect on cellular translation in the stapled ribosome, consistent with the large subunit of O-ribo(h44H101) being functionally insulated from the endogenous small subunit (Figure 4 D).</p><p>In conclusion, we have described the rational, structure-based design of a stapled orthogonal ribosome. Our design inserts a circularly permuted 23S rDNA into the 16S rDNA at sites determined by structural and phylogenetic analysis, and uses an RNA hinge to staple the two subunits and facilitate subunit association and disassembly. Our results indicate that the stapled orthogonal ribosome allows the effects of mutations in 23S rRNA to be specifically coupled to translation of an orthogonal message and insulated from endogenous translation. Future work will focus on optimizing the activity of our rationally designed stapled ribosome, and fully characterizing the extent to which orthogonality in subunit association (Figure 1) may be achieved through the stapling of ribosome subunits. We anticipate that the development of stapled orthogonal ribosomes may further extend orthogonal translation, and enable further progress on the genetically encoded synthesis of unnatural polymers in cells.</p>

PubMed Open Access

Detection of antimicrobial resistance-associated proteins by titanium dioxide-facilitated intact bacteria mass spectrometry

Titanium dioxide-modified target plates were developed to enhance intact bacteria analysis by matrixassisted laser desorption/ionization time-of-flight mass spectrometry. The plates were designed to photocatalytically destroy the bacterial envelope structure and improve the ionization efficiency of intracellular components, thereby promoting the measurable mass range and the achievable detection sensitivity. Accordingly, a method for rapid detection of antimicrobial resistance-associated proteins, conferring bacterial resistance against antimicrobial drugs, was established by mass spectrometric fingerprinting of intact bacteria without the need for any sample pre-treatment. With this method, the variations in resistance proteins' expression levels within bacteria were quickly measured from the relative peak intensities. This approach of resistance protein detection directly from intact bacteria by mass spectrometry is useful for fast discrimination of antimicrobial-resistant bacteria from their nonresistant counterparts whilst performing species identification. Also, it could be used as a rapid and convenient way for initial determination of the underlying resistance mechanisms.

detection_of_antimicrobial_resistance-associated_proteins_by_titanium_dioxide-facilitated_intact_bac

Introduction<!>TiO 2 -facilitated intact bacteria MALDI-TOF MS ngerprinting<!>Detection of antimicrobial resistance-associated proteins from intact bacteria<!>Simultaneous bacteria identication and antimicrobial resistance-associated protein detection with clinical pathogens<!>Conclusions<!>Conflicts of interest

<p>Infectious diseases caused by pathogenic bacteria are serious threats to human health. Misuse and overuse of antimicrobial drugs over many years have led to the emergence of antimicrobial resistance among microbes worldwide. 1 For fast diagnosis and efficient treatment, it is crucial to perform pathogen identication and a rapid analysis of their antimicrobial resistance phenotypes. With the ability to generate characteristic mass spectral ngerprints directly from intact bacteria, matrixassisted laser desorption/ionization time-of-ight mass spectrometry (MALDI-TOF MS) provides a rapid method for bacteria identication (e.g. $30 min for 48 samples) and has received clearance from the US Food and Drug Administration (FDA). 2,3 Commercial systems, including Vitek MS (bioMérieux) and MALDI Biotyper (Bruker Daltonics), have been installed and constantly used in many hospitals. Meanwhile, antimicrobial resistance detection in hospitals still mainly relies on antimicrobial drug culture approaches like the broth (or agar) microdilution method and disk diffusion method, which need several hours or even several days. 4 Therefore, performing a complete clinical diagnosis remains a lengthy process.</p><p>In addition to the classical culture-based methods, several new strategies have been proposed for antimicrobial resistance detection. Examples include nucleic acid-based resistance gene detection, 5 single-cell morphological analysis, 6 surfaceenhanced Raman spectroscopic biomarker detection, 7 atomic force microscope cantilever-based nanomechanical sensors, 8 etc.</p><p>Recently, continuous efforts have been made to explore the potential of MALDI-TOF MS for rapid antimicrobial resistance analysis. Related studies were mainly carried out with three approaches. The rst one is an indirect evaluation by detection of resistance enzyme activity, such as the degradation of b-lactam antibiotics through hydrolysis (mass increased by 18 Da) by b-lactamases 9 and the alternation of rRNA through methylation (mass increased by 14 Da) by rRNA methyltransferase. 10 This is a fast method for resistance mechanism investigation, but is limited to certain enzyme-related resistance types. The second approach is an isotope labelling-bacteria culture method, with the appearance of peak mass shis in bacterial ngerprinting patterns if a resistant strain is incubated with culture medium supplemented with stable (non-radioactive) isotope labelled amino acids and corresponding antibiotics. 11,12 Based on the machinery of protein biosynthesis, this approach is applicable to determining bacterial resistance or susceptibility to a wide range of antibiotics, but limited by the need for special isotope labelled culture medium. The third one is also a culture-based method, in which semi-quantitative MALDI-TOF MS using an internal standard is employed to investigate bacterial growth status with the presence or absence of antibiotics by measuring the quantity of biomass within a spectrum. 13,14 This method has been demonstrated to be feasible for different antibiotic classes/bacterial species combinations. In addition to the above three approaches, bacteria subtyping assays have also been conducted to study the correlation between antibioticsusceptible and resistant strains by comparing their ngerprint patterns. For example, it has been used to discriminate major methicillin-resistant Staphylococcus aureus lineages 15 and to identify vancomycin-resistant Enterococcus spp. 16 Proteins encoded by antimicrobial resistance genes are directly involved in bacterial resistance process against antimicrobial drugs. 17 Antimicrobial resistance can be analysed by tracing these resistance-associated proteins within bacterial cells. Ideally, they should be read out directly from MALDI-TOF MS ngerprint patterns of intact bacteria without any sample pre-treatment, a useful procedure that would be comparable to fast bacteria identication. But many of those proteins are large ones (>15 000 Da) expressed in low abundance, and are difficult to detect directly from intact cells by classic MALDI-TOF MS measurements, which typically focus on smaller proteins (<15 000 Da) expressed in high abundance. 18 Until now, to the best of our knowledge no studies have reported the success of this procedure, as pointed out in a review by Walkova et al. 19 In order to detect these resistance-associated proteins, preparatory extraction and enrichment processes are required prior to their identication by MS, which is labour-intensive and timeconsuming. [20][21][22] Very recently, a surrogate marker around 11 kDa was detected from carbapenem-resistant bacteria strains containing bla KPC -harboring plasmids by a MALDI-TOF MS ngerprinting approach. But it is a particularly small protein and an additional step of protein extraction was required prior to MS analysis. 23 Herein, we have developed a MALDI-TOF MS ngerprinting approach for intact bacteria analysis using photo-reactive titanium dioxide (TiO 2 )-modied target plates, providing access to a high mass range with enhanced detection sensitivity. TiO 2 -modied target plates or more generally metal oxide-modied plates have been used for many different MALDI-TOF MS applications. 24,25 In the current work, the rationale of the proposed approach is to take advantage of the photo-reactivity of TiO 2 to destroy bacterial cell membranes and to facilitate inner component desorption/ionization. Such experimental improvement makes feasible a direct fast read out of resistanceassociated proteins from intact bacteria cells without any sample pre-treatment.</p><!><p>One important factor affecting MALDI-TOF MS measurements is the choice of matrix. Sinapinic acid was utilized as a matrix throughout this work, as it provides satisfying reproducibility and facilitates the detection of large proteins (Part S1, ESI †). Based on our experience in designing photo-reactive TiO 2 -modied target plates for inducing in-source electrochemical reactions, 26,27 we have developed here a plate able not only to absorb bacteria on a porous structure but also to lyse them by photocatalytic oxidation, improving intact bacteria ngerprinting in a broad mass range as demonstrated below.</p><p>This target plate was prepared by depositing an aqueous suspension of TiO 2 nanoparticles (NPs) on the spots (3 mm diameter) of a classic bare stainless steel target plate, or by dropping TiO 2 suspension as an array of spots on a stainless steel foil (20 mm thick), which was aerwards affixed onto a bare target plate by an adhesive tape (Fig. 1a). The TiO 2 NPs were subsequently thermally or photonically sintered. The sintered NPs exhibited strong adherence to the steel substrate and provided a stable support layer ($3 mm thick) for the bacteria and matrix, with small particles (of 20-25 nm size) densely covering the bottom and large particles (of 0.5-3 mm size) observed on the surface (Fig. 1b). The TiO 2 used is a commercial P25 nanopowder, a mixture of anatase (80%) and rutile (20%) crystalline phases. The anatase is more photo-reactive than the rutile, but the latter is more thermodynamically stable. The crystalline phases of TiO 2 were not changed aer sintering, and the corresponding X-ray powder diffraction patterns are shown in Fig. 1c. Compared to bare steel target spots, the spots with TiO 2 NPs had rough and mesoporous surfaces (see the surface roughness proles in Part S2, ESI †), with a larger surface area and lower water contact angle (decreased from 70 to 38 , Part S3, ESI †). As the surface of the TiO 2 spots are more hydrophilic than the steel substrate, this kind of TiO 2 -modied plate can be used as "AnchorChip" targets. Upon deposition, bacteria cells (mostly of 0.2-2 mm size) entered into the porous TiO 2 NPs structure. Due to the high affinity between the bacterial membrane and TiO 2 , 28 the cells tended to be absorbed on the surface of TiO 2 . Matrix drop casting consequently led to the formation of ne and well dispersed bacteria/matrix crystals, highly favourable for an efficient desorption/ionization process (Fig. 1d). MALDI-TOF MS analysis of intact Escherichia coli (E. coli, strain DH5a) yielded much higher quality ngerprint patterns using a TiO 2 -modied target plate in comparison with a bare steel target plate (Fig. 1e). Such signicant improvement, especially in the mass range m/z ¼ 15 000-60 000, could not be solely caused by the high quality of bacteria/matrix co-crystals resulting from the mesoporous spots' surface. It could also be explained by the ability of TiO 2 to destroy the bacterial cell membrane and to improve analyte desorption/ionization due to its well-known photo-reactivity. 28,29 As proof of cell membrane disruption, the morphological changes of E. coli were visualized by scanning electron microscopy (Fig. 2). E. coli cells showed a straight, rod-like shape when they were deposited on the spots of a bare and a TiO 2 -modied target plate with no matrix covering and no MALDI laser irradiation (Fig. 2a and b). The spots with E. coli were then covered with matrix and underwent the running of a typical MS measurement (500 nitrogen laser shots on each sample spot, 20 Hz laser frequency). On the spots of the bare target plate, most E. coli cells (>95% according to microscopic observation) generally maintained their rod-like shape (Fig. 2c). The diameter of the laser beam used in the MALDI-TOF MS instrument (Bruker Microex) is about 100 mm, 30 times smaller than the sample spot size (3 mm diameter). Thus, a typical MS measurement is accomplished with many "blind shots", and only the cells exactly shot by the laser could be lysed. However, the situation was different on the spots with TiO 2 : most cells were seriously damaged with apparent deformation and membrane rupture, and the "melted" cells were embedded into the mesoporous spot surface (Fig. 2d). Interestingly, it was found that the crystal shape of the matrix on the spots with TiO 2 was quite different from that on the bare spots (Fig. 2c and d). The disruption of more bacteria cells facilitated the detection of barely accessible inner cell components. The importance of oxidative cell disruption for enhancing bacteria MS analysis was conrmed by sample treatment with scavengers of these reactive oxygen species. Bacteria aqueous solutions containing different scavengers, i.e., sodium oxalate, isopropanol and ferrocenemethanol, were deposited onto TiO 2 -modied target plates for MALDI-TOF MS measurements. The concentration of each scavenger was set to an appropriate value to eliminate its possible inuence on bacterial cells (initial pH 7.44): 2 mM sodium oxalate (pH 7.87), 2 mM isopropanol (pH 7.11) and 0.4 mM ferrocenemethanol (pH 7.90). 32,33 The MALDI-TOF MS analysis in the presence of these scavengers showed low quality bacterial ngerprint patterns throughout the mass range m/z ¼ 2000-80 000 (Part S6, ESI †). In addition to disrupting bacteria cells, the high photo-reactivity of TiO 2 favours efficient energy absorption from the laser source and transfer of this energy to matrix/analyte. 34 This process occurs in addition to laser energy absorption directly by the matrix, and thus can facilitate analyte desorption/ionization even further. This was demonstrated by the analysis of bacterial protein extracts and standard protein mixtures. For the bacterial protein extracts, prepared according to the oen-used ethanol/formic acid/acetonitrile extraction protocol, much higher quality MS patterns were observed using a TiO 2 -modied target plate than when using a classic bare one (Part S7, ESI †). For the standard protein mixtures, containing cytochrome c ($12 kDa), myoglobin ($17 kDa), bovine serum albumin (BSA, $66 kDa) and lactoferrin ($82 kDa), the MS peak intensity of each protein was increased by the presence of TiO 2 . Consequently, the detection sensitivity was also improved, especially for the two hardly-ionized large proteins BSA and lactoferrin (Part S8, ESI †).</p><p>For comparison, we have tested the performance of nonphoto-reactive nanomaterials like Al 2 O 3 NPs (<50 nm in particle size) and SiO 2 NPs (200 nm in particle size). They were shown to have detrimental effects on MS results for both standard protein mixtures and intact bacteria (Part S9 and S10, ESI †). TiO 2 -facilitated intact bacteria MALDI-TOF MS ngerprinting was further tested on different bacteria species. In addition to E. coli (strain DH5a), two more species, i.e., Pseudomonas aeruginosa (P. aeruginosa, strain ATCC 27853) and Bacillus subtilis (B. subtilis, strain 168), were chosen as model analytes. All bacteria were measured in their intact whole state, without any preparatory protein extraction. Corresponding ngerprint patterns generated with a classic bare target plate and a TiO 2 -modied one with exactly the same measurement parameters are compared in Fig. 3a-c Overall, MALDI-TOF MS analysis of bacteria was promoted by TiO 2 , with signicant improvement in both peak numbers and peak intensities of bacterial ngerprint patterns within During the peak number counting process, the threshold value of signal-to-noise ratio (S/N) of counted peaks was set as 3, and the threshold value of relative peak intensity was set as 2.0% for the mass range m/z ¼ 2000-15 000 and 0.1% for the mass range m/z ¼ 15 000-80 000. The number of bacterial cells on each sample spot was around 5 Â 10 5 . All measurements were conducted under the exact same instrumental parameters.</p><p>a broad mass range. Attributed to its high photo-reactivity and photocatalytic bacteria disruption ability, TiO 2 helped not only to break the cellular envelope structure, but also to enhance the desorption/ionization efficiency of intracellular components. By generating high quality bacterial ngerprint patterns, the TiO 2 -modied target plate could greatly boost the reliability of bacteria identication, which is based on ngerprint pattern matching. More importantly, it facilitates the extraction of more bacterial cellular information, and enables the detection of large molecular weight and low abundance bacterial components, especially those related to antimicrobial drug resistance, as discussed further.</p><!><p>The possibility of antimicrobial resistance-associated protein detection by an intact bacteria MS ngerprinting approach was investigated with the TiO 2 -modied target plate. The detection was rstly conducted with bacteria samples that were modied by gene transfer. Corresponding plasmid DNAs, carrying specic resistance genes, were articially transformed into recipient bacteria using recombinant techniques. 37 Following this strategy, dened non-resistant (or antibiotic-susceptible) E. coli strains were equipped with the desired antimicrobial resistance, i.e., resistance against ampicillin, kanamycin, gentamicin and chloramphenicol, respectively. MALDI-TOF MS ngerprint patterns of the resistant strains were measured within the mass range of m/z ¼ 2000-80 000 and compared with those of non-resistant strains. To ensure result reliability, each type of resistance was repeatedly developed within two E. coli strains, i.e., two DH5a, XL1-Blue or BL21. The MS results showed that resistance-associated proteins were successfully detected from all of these resistant strains (Fig. 4), as explained in further detail below.</p><p>Gene bla TEM-1, encoding a TEM-1 b-lactamase, conferred resistance against ampicillin. With a molecular weight around 29 kDa, TEM-1 b-lactamase inactivates ampicillin by hydrolysis of the b-lactam ring in the ampicillin molecule. 38 Compared to the ampicillin-susceptible E. coli, the ampicillin-resistant ones exhibited almost the same MALDI-TOF MS ngerprint patterns except for an additional peak at m/z ¼ 28 972 AE 5 for strain DH5a and m/z ¼ 28 972 AE 3 for strain XL1-Blue (Fig. 4a). This result coincides with a previous study, in which a special preparatory protein extraction was conducted before MS measurement. 38 The resistance against kanamycin resulted from the expression of neomycin-kanamycin phosphotransferase type II (29 048 Da, UniProtKB-P00552), which inactivates kanamycin by phosphoryl transfer at its 3 0 -hydroxyl group. 39 Using TiO 2 -modied target plates, this phosphotransferase was successfully detected in two kanamycin-resistant E. coli strains (at m/z ¼ 29 046 AE 2 for strain DH5a, and m/z ¼ 29 047 AE 2 for strain BL21), but not in their non-resistant counterparts (Fig. 4b). The resistance against gentamicin was conferred by gene aacC1, encoding gentamicin acetyltransferase I (19 442 Da, UniProtKB-P23181), which inactivates gentamicin by acetylating its 3-amino deoxystreptamine moiety. 40 This protein was detected exclusively in the gentamicin-resistant E. coli at m/z ¼ 19 442 AE 1 for both strains DH5a and XL1-Blue (Fig. 4c). The resistance against chloramphenicol was caused by the synthesis of chloramphenicol acetyltransferase (CAT, 24-26 kDa), 41 which catalyses the transfer of an acetyl moiety from bacterial coenzyme A to the chloramphenicol molecules, and, therefore, results in antibiotic inactivation. Here, in contrast with non-resistant E. coli, a peak around m/z ¼ 24 820 was clearly detected for the chloramphenicol-resistant E. coli (24 820 AE 4 for strain DH5a, 24 819 AE 3 for strain XL1-Blue), conrming the expression of CAT (Fig. 4d). For all of the above measurements, detection of each resistance protein showed high reproducibility for both tested E. coli strains. The allowed tolerance between the measured and the theoretical masses was 300 ppm, due to the limited resolving power of the MALDI-TOF MS instrument used.</p><p>To investigate the expression of the same resistance gene within different bacteria species, Enterobacter cloacae ssp. cloacae (E. cloacae s. C.) and Enterobacter aerogenes (E. aerogenes) were articially transformed with an ampC gene encoding AmpC type b-lactamase ($39.5 kDa) 42,43 and measured with MALDI-TOF MS. Aer the gene transfer, both E. cloacae s. C. and E. aerogenes acquired resistance against 10 different b-lactam antibiotics, becoming multidrug resistant. Their detailed antimicrobial susceptibility proles (measured with a bio-Mérieux VITEK 2 automated AST system based on an antimicrobial drug culture method) before and aer the gene transfer are shown in Part S16, Tables S1-S4, ESI. † For the multidrugresistant E. cloacae s. C. and E. aerogenes, the minimum inhibitory concentrations (MICs) of the 10 antibiotics varied from 16 to 128 mg mL À1 . Compared to their non-resistant counterparts, the two resistant strains both exhibited an additional peak around m/z ¼ 39 500 (m/z ¼ 39 496 AE 5 and 39 505 AE 4, respectively) (Fig. 4e), conrming the expression of the AmpC type b-lactamase.</p><p>It should be mentioned that none of the above resistanceassociated proteins were detectable when classic bare stainless steel plates were used (Part S14, ESI †), showing the importance of TiO 2 -modied target plates in bacteria analysis.</p><p>To further conrm the identity of the detected resistanceassociated proteins, the antibiotic-resistant or non-resistant strains were analysed with a widely used proteomic approach. Bacteria cells were lysed in sodium dodecyl sulphate loading buffer, and the extracted proteins were separated by sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE) and subsequently identied by liquid chromatographytandem mass spectrometry (LC-MS/MS). Taking gentamicinresistant E. coli DH5a and kanamycin-resistant E. coli BL21 as examples, a protein band around 19 kDa or 29 kDa was clearly observed on their corresponding SDS-PAGE gel running lanes, but not observed for their non-resistant counterparts (Part S15, ESI †). Excision of the $19 kDa band from both gentamicinresistant and non-resistant E. coli DH5a (as control) lanes, followed by digestion in trypsin, revealed the presence of 60 gentamicin acetyltransferase I exclusive unique peptides in the gentamicin-resistant strain, with 100% protein identication probability and 93% (164/177) amino acid coverage (Table S1 in Part S15, ESI †). The $29 kDa bands from kanamycin-resistant and non-resistant E. coli BL21 (as control) lanes were analysed in the same way, revealing the presence of 72 neomycinkanamycin phosphotransferase type II exclusive unique peptides in the kanamycin-resistant strain, with 100% protein identication probability and 93% (246/264) amino acid coverage (Table S2 in Part S15, ESI †). The above results coincide with the MALDI-TOF MS intact bacteria ngerprinting results in Fig. 4, conrming the expression and identity of antibiotic resistance-associated proteins in corresponding resistant strains.</p><p>The described TiO 2 -facilitated MALDI-TOF MS approach can also quickly sense the variations in resistance genes' expression levels within bacteria. To demonstrate this, antibiotic-resistant bacteria were cultured in Luria-Bertani (LB) medium that contained different concentrations of corresponding antibiotics. A gradual increase of a given antibiotic concentration brings a proportionally higher selection pressure to the bacterial cells. As a response, bacterial cells modulate the resistance genes' expression level to increase the synthesis of resistance proteins for survival. 44 Such kinds of change were measured for ampicillin-, kanamycin-and chloramphenicol-resistant E. coli DH5a by comparing the relative peak intensities (r.t.) of the corresponding resistance proteins in the MALDI-TOF MS ngerprint patterns (Fig. 5a-c). The r.t. of the resistance proteins (i.e., TEM-1 b-lactamase at m/z ¼ 28 972 AE 5, neomycin-kanamycin phosphotransferase type II at m/z ¼ 29 046 AE 2 and CAT at m/ z ¼ 24 820 AE 4) were calculated using signals from E. coli DH5a D-ribose-binding periplasmic protein (RbsB, $28.5 kDa) 45 as an internal intensity standard (r.t. RbsB ¼ 1). For all three proteins, their r.t. increased with an increase in corresponding antibiotic concentration. These data conrm that higher levels of antibiotic resistance would accompany higher expression levels of resistance proteins and consequently higher r.t. values of the corresponding MS peaks. For the chloramphenicol-resistant strain, however, the r.t. of CAT decreased when the chloramphenicol concentration reached 120 mg mL À1 (Fig. 5c). Probably, this concentration was already too high and started to negatively affect the bacterial physiological state. In addition to the antibiotics present, the type of culture medium can also affect the expression level of resistance proteins. Synthesis of resistance proteins to ght against antibiotics is an energyconsuming process, which can be positively inuenced by the use of nutritionally rich growth medium. 46 To observe this effect, ampicillin-resistant E. coli DH5a was cultured in different growth media containing a xed concentration (60 mg mL À1 ) of ampicillin. The corresponding MALDI-TOF MS ngerprint patterns indicated that 2xYT medium, specically rich in amino acids and peptides, favoured the up-regulation of gene bla TEM-1 expression. In particular, when the growth medium was changed from LB to 2xYT, the averaged r.t. of TEM-1 b-lactamase increased from 0.60 to 6.66 (r.t. RbsB ¼ 1) (Fig. 5d).</p><!><p>The proposed method can be used for antimicrobial resistance protein detection whilst performing bacteria species identication. The feasibility was explored with three clinical pathogens: extended-spectrum b-lactamase-producing E. coli (ESBL-E. coli), multidrug-resistant Pseudomonas aeruginosa (MDR-P. aeruginosa) and methicillin-resistant Staphylococcus aureus (MRSA).</p><p>ESBL, rst reported in Germany in 1983, confers resistance to a broad spectrum of b-lactam antibiotics. 47 Worldwide emergence of ESBL-E. coli raises serious therapeutic problems. Resistance in the ESBL-E. coli tested here was conferred by the expression of CTX-M type b-lactamase ($28 kDa). 48 E. coli ATCC25922, a strain without such kind of resistance, was used as the reference for species identication and resistance protein detection of the testing strain (ESBL-E. coli). Detailed antimicrobial susceptibility proles of the two strains, measured with a bioMérieux VITEK 2 automated AST system, are shown in Part S16, Tables S5 and S6, ESI. † For the testing strain, the MICs of the corresponding antibiotics were 4-320 mg mL À1 (Table S6 †). The averaged MALDI-TOF MS ngerprint patterns of the two strains are displayed in Fig. 6a. The similarity score between the two patterns was calculated using a public bacteria identication platform, BacteriaMS, with a cosine correlation algorithm (http://bacteriams.fudan.edu.cn/#/). This algorithm gives the maximum score as 1.0. Here, with the pattern similarity as high as 0.9427, the testing strain was identied to be the same species as the reference one, i.e.,E. coli. The two strains shared almost all MS peaks (S/N > 3, r.i. > 0.1%) in the mass range of m/ z ¼ 10 000-80 000, except for a peak at m/z ¼ 28 074 AE 4 only detected for the testing strain (Fig. 6a, zoom-in). The appearance of this peak most probably results from the expression of CTX-M type b-lactamase. Therefore, together with the species identication, the CTX-M type ESBL resistance was recognized in the testing strain. Similarly, the MDR-P. aeruginosa and MRSA were also iden-tied at the species level by comparison of their ngerprint patterns with those of corresponding reference strains (i.e., P. aeruginosa ATCC 27853 and S. aureus ATCC 29213), resulting in pattern similarity scores of 0.8546 and 0.9578, respectively (Fig. 6b and c). Simultaneously, two mass spectral peaks, at m/ z ¼ 38 080 AE 6 and m/z ¼ 40 900 AE 8, were exclusively observed from the MDR-P. aeruginosa (Fig. 6b, zoom-in). They most likely come from efflux pump proteins MexA ($38 kDa) 49,50 and MexX (40.9 kDa, UniProtKB-Q9ZNG9), which confer the multidrug resistance of the MDR-P. aeruginosa. These two proteins are involved in the extrusion of b-lactam antibiotics (e.g., tazobactam, ceazidime, and cefepime) and aminoglycosides (e.g., amikacin, gentamicin, netilmicin, and tobramycin) from within bacteria cells into the external environment. 51 Meanwhile, antimicrobial resistance in MRSA, one of the most common multidrug resistant pathogens, arises from the expression of gene mecA, which causes the alteration of penicillin binding protein (PBP) and triggers the expression of its alternative, i.e., PBP 2a ($78 kDa). PBP 2a has a low affinity for most b-lactam antibiotics including methicillin, thereby making bacteria resistant against them. 52 According to previous studies, a characteristic fragment of PBP 2a ($13 kDa) can be detected for MRSA by a proteomics-based method. [53][54][55] In the present work, a peak at m/z ¼ 13 080 AE 2 was exclusively detected for the MRSA strain (Fig. 6c, zoom-in), which could come from the PBP 2a fragment. To conrm this assumption, two more MRSA strains were tested and the peaks around 13 kDa (at m/z ¼ 13 083 AE 3 and m/z ¼ 13 081 AE 4, respectively) were repeatedly detected, as shown in Part S17, ESI. † The detailed antimicrobial susceptibility proles of MDR-P. aeruginosa, MRSA and their reference strains are shown in Part S16, Tables S7-S12, ESI. † As shown in these proles, antibiotic MICs for the resistant strains were measured as 0.5-8 mg mL À1 .</p><p>Herein, antimicrobial resistance-associated proteins were successfully detected directly in intact bacteria without any sample pre-treatment, by TiO 2 -facilitated MALDI-TOF MS. The developed approach showed feasibility for both Gram-negative and Gram-positive bacteria species bearing different types of antimicrobial resistance. Each of the resistance proteins were specically detected from the corresponding antibioticresistant strains, not from the non-resistant reference strains or the strains resistant to other antibiotics. For certain resistance types tested in this work, it was also shown that higher levels of antibiotic resistance could accompany higher expression levels of the resistance proteins. Sensing of the expression level variations was proven possible through direct readout of the relative intensities of the corresponding MS peaks.</p><p>The described method performed resistance protein recognition according to their m/z values. Due to the limited resolving power of current MALDI-TOF MS instruments, it would be difficult to distinguish closely-related protein isoforms with quite similar molecular weights like TEM-1, TEM-2 and TEM-3 b-lactamases that differ only in a few amino acid substitutions. This is a drawback for MALDI-TOF MS-based analysis of proteins in comparison with nucleic acid-based molecular detection of the related genes or proteomics-based approaches.</p><p>In this work, it has been conrmed that the expression levels of resistance proteins directly determine their MS peak appearances in bacterial ngerprint patterns. As intact bacteria are analysed directly without preparatory protein extraction, enrichment or selective separation, the proposed method could lack some sensitivity when the resistance proteins are expressed at a very low level. Here, the method was shown to be sensitive enough for resistant strains with antibiotic MICs as low as a few microgram per millilitre when 1 uL of the bacteria sample ($5 Â 10 5 cells) was measured.</p><p>Nonetheless, compared to existing methods for resistance gene or protein detection such as nucleic acid-based molecular techniques or proteomics-based approaches, the proposed MALDI-TOF MS-based method has clear advantages of simplicity and rapidity of sample preparation, measurement protocol and data analysis. It is a useful procedure for quick discrimination of antimicrobial-resistant bacteria strains from their non-resistant counterparts, as well as a fast method for the initial determination of resistance mechanisms and prediction of antibiotic types or classes that the strains could be resistant to.</p><!><p>In this work, intact bacteria MALDI-TOF MS analysis was improved by TiO 2 due to its ability to photo-catalytically destroy bacterial envelopes and to facilitate analyte desorption/ ionization. Impressive improvement in detection sensitivity and working mass range was achieved, pushing the current limits of the bacteria MALDI-TOF MS ngerprinting approach. Accordingly, antimicrobial resistance-associated proteins, especially those larger than 15 kDa, were successfully detected from intact bacteria by the direct readout of the corresponding MS peaks from the ngerprint patterns, together with a rapid sensing of their expression level variations. With the potential of simultaneous species identication and antimicrobial resistance analysis, the TiO 2 -facilitated MALDI-TOF MS opens new avenues for bacteria analysis.</p><!><p>There are no conicts to declare.</p>

Royal Society of Chemistry (RSC)

Aromatic secondary amine-functionalized fluorescent NO probes: improved detection sensitivity for NO and potential applications in cancer immunotherapy studies

Tumor-associated macrophages (TAMs), constituting up to 50% of the solid tumor mass and commonly having a pro-tumoral M2 phenotype, are closely associated with decreased survival in patients. Based on the highly dynamic properties of macrophages, in recent years the repolarization of TAMs from protumoral M2 phenotype to anti-tumoral M1 phenotype by various strategies has emerged as a promising cancer immunotherapy approach for improving cancer therapy. Herein, we present an aromatic secondary amine-functionalized Bodipy dye 1 and its mitochondria-targetable derivative Mito1 as fluorescent NO probes for discriminating M1 macrophages from M2 macrophages in terms of their difference in inducible NO synthase (iNOS) levels. The two probes possess the unique ability to simultaneously respond to two secondary oxides of NO, i.e., N 2 O 3 and ONOO À , thus being more sensitive and reliable for reflecting intracellular NO than most of the existing fluorescent NO probes that usually respond to N 2 O 3 only. With 1 as a representative, the discrimination between M1 and M2 macrophages, evaluation of the repolarization of TAMs from pro-tumoral M2 phenotype to anti-tumoral M1 phenotype, and visualization of NO communication during the immune-mediated phagocytosis of cancer cells by M1 macrophages have been realized. These results indicate that our probes should hold great potential for imaging applications in cancer immunotherapy studies and relevant anti-cancer drug screening.

aromatic_secondary_amine-functionalized_fluorescent_no_probes:_improved_detection_sensitivity_for_no

Introduction<!>Results and discussion<!>Conclusions<!>Conflicts of interest

<p>Macrophages are specialized immune cells found all over the body that exist primarily to engulf and digest cellular debris, foreign substances, microbes, and cancer cells in a process called phagocytosis. Macrophages are particularly active in inammation and infection, under which conditions, blood monocytes are recruited into the tissue where they differentiate into macrophages. 1 Notably, macrophages are the most wellcharacterized type of tumor-inltrating immune cells, and play crucial roles from anti-tumor to tumor progression and metastasis. Accumulating evidence reveals that the dual functions of macrophages can be attributed to their ability to adapt to the macroenvironment that leads to two main polarized phenotypes, i.e., classically activated M1 macrophages and alternatively activated M2 macrophages. [2][3][4] M1 macrophages, characterized by the expression of high-level inducible nitric oxide (NO) synthase (iNOS) as well as some pro-inammatory cytokines, such as IL-12, [5][6][7] have high bactericidal and tumoricidal activity partially due to their ability to secrete high levels of reactive oxygen/nitrogen species (ROS/RNS), such as hydrogen peroxide (H 2 O 2 ), superoxide (O 2 À c), and NO and its secondary metabolites dinitrogen trioxide (N 2 O 3 ) and peroxynitrite (ONOO À ). In contrast, M2 macrophages, characterized by the expression of arginase 1 (Arg-1) and anti-inammatory cytokines, such as IL-10, commonly have a low level of iNOS [5][6][7] and can assist tumor development by inducing angiogenesis, remodeling the extracellular matrix, stimulating cancer cell proliferation, and inhibiting adaptive immunity. In fact, under cancer-initiating conditions, the inltrated macrophages have an M1 phenotype and are anti-tumoral; however, their continued presence in a tumor microenvironment polarizes them to tumor-associated macrophages (TAMs), which commonly have an M2 phenotype and are closely associated with decreased survival in patients due to their pro-tumoral role. [2][3][4] Of note, the polarization of macrophages is a highly dynamic process and the phenotype of M1-or M2-polarized macrophages can be reversed depending on the microenvironmental cues they receive. 8 For instance, the reversion of macrophages from M2 phenotype to M1 phenotype and reduction of immunosuppressive effects from the M2 population have been observed when TAMs were treated with interferon-g/lipopolysaccharide (IFN-g/LPS); 9,10 in patients with extended survival, the M1 macrophages account for the majority of macrophages present within tumors, 11 distinct from the cases in tumor development and metastasis, where macrophages predominantly exhibit a pro-tumoral M2 phenotype. [2][3][4][5][6][7] Based on these discoveries, the repolarization of TAMs from M2 phenotype to M1 phenotype to activate their anti-tumoral potential by various strategies has emerged as an attractive and promising approach in cancer immunotherapy in recent years. 2,[12][13][14][15][16][17] In this context, the development of efficient methods that can discriminate M1 macrophages from M2 macrophages is of crucial guiding signicance for cancer immunotherapy studies and relevant anti-cancer drug screening. Although immunohistological quantication, enzyme linked immunosorbent assay (ELISA), and Western blot analysis of various biomarkers have routinely been used to distinguish between M1 and M2 macrophages, [5][6][7]17 these methods are complex, time-consuming, and especially incompatible with living systems. By comparison, uorescent probe-based techniques, which have become the gold standard for detection and imaging of various biological species in living systems, are the most promising to overcome these limitations due to their simplicity, convenience, sensitivity, noninvasiveness, and realtime spatial imaging capacity. 18,19 However, the attractive techniques have never been exploited to identify M1 or M2 macrophages to date. Given that M1 macrophages express higher levels of iNOS than M2 macrophages, we envisioned that uorescent NO probes when properly designed should have the potential to distinguish between M1 and M2 macrophages in terms of their difference in the iNOS level and thus the NO level.</p><p>Among various uorescent NO probes, [20][21][22][23][24][25][26][27][28] o-diamine-based ones, pioneered by Nagano's group, are by far the most oen studied and applied uorescent NO probes. [23][24][25][26][27][28] The corresponding sensing mechanism is based on the reaction of the odiamine group with the autoxidation product of NO, i.e., dinitrogen trioxide (N 2 O 3 ), 29 to form the benzotriazole derivative, thereby triggering a uorescence off-on response by inhibiting the photoinduced electron transfer (PeT) process. Although uorescent NO probes of this kind have widely been applied in biological systems, some limitations still remain, such as possible interference by dehydroascorbic acid (DHA)/ascorbic acid (AA)/methylglyoxal (MGO) [30][31][32][33] and a relatively long response time (commonly more than 5 min). To overcome these limitations, in recent years some new strategies have been actively developed, such as diazo ring formation, 34,35 reductive deamination, 36,37 monoprotection of vicinal diamine groups, [38][39][40][41] aromatization of Hantzsch ester, 42,43 and N-nitrosation of aromatic secondary amines. [44][45][46][47] Yet despite the remarkable progress that has already been achieved, a widely overlooked issue is that almost all of these probes, including odiamine-based ones, can reect intracellular NO only by reacting with its autooxidation product N 2 O 3 , which would inevitably decrease the detection sensitivity for NO given that NO could also rapidly react with O 2 À c to generate ONOO À at near diffusion control ($10 10 M À1 S À1 ), 48 and that these probes usually fail to give a uorescence response toward ONOO À . The situation may be especially serious during the immune response of macrophages, where large amounts of NO and O 2 À c were simultaneously produced and coexisted with O 2 . 49 Thus, the development of new uorescent NO probes that can sensitively sense both N 2 O 3 and ONOO À is highly desired for improving not only the detection sensitivity of NO but also the reliability in distinguishing between M1 and M2 macrophages.</p><p>Recently, we reported for the rst time that aromatic secondary amines could function as both the reaction group and PeT donor to construct uorescent NO probes. 44 However, like most of the previous reports, the as-obtained probe, i.e., Nbenzyl-4-hydroxyaniline-functionalized Bodipy, only exhibited a selective uorescence off-on response toward N 2 O 3 but not ONOO À . Further studies revealed that although not giving a uorescence response, the probe could react with ONOO À to lead to a nonuorescent debenzylation product. This means that in practical bioimaging assays, the probe would probably suffer from the risk of being consumed by coexisting ONOO À , thereby resulting in decreased sensitivity for NO. However, to our delight, when N-benzyl-4-methoxyaniline was employed as the reaction group instead of the N-benzyl-4-hydroxyaniline group mentioned above, the newly developed uorescent probe, i.e., N-benzyl-4-methoxyaniline-functionalized Bodipy 1 and its mitochondria-targetable derivative Mito1 (Scheme 1), not only overcame the shortcomings of classic o-diamine-type probes, such as possible interference by DHA/AA/MGO and a long response time, but also displayed a signicant uorescence off-on response for both N 2 O 3 and ONOO À . The unique sensing properties endow the probes with high sensitivity for reecting intracellular NO as indicated by their ability to image basal and endogenous NO in living cells. With 1 as a representative, we have successfully realized the discrimination between M1 and M2 macrophages and visualization of the repolarization of TAMs from M2 phenotype to M1 phenotype induced by IFN-g/LPS. Also, we conrmed that during the immune-mediated phagocytosis of cancer cells by M1 macrophages, NO secreted by M1 macrophages could diffuse across the cancer cell membrane to exert its tumoricidal action by producing cytotoxic N 2 O 3 and ONOO À . These ndings strongly indicate that our probes should hold great potential for imaging applications in cancer immunotherapy studies and relevant drug screening.</p><!><p>Synthesis and spectral response of 1 and Mito1 for N 2 O 3 and ONOO À Probes 1 and Mito1 could be easily synthesized by a simple three-step procedure starting from commercially available 2methoxy-5-nitrobenzaldehyde, including the initial synthesis of Bodipy dye, subsequent reduction of the nitro group to the amino group, and nal reductive amination with corresponding benzaldehydes. The detailed synthesis and characterization data of 1 and Mito1 are presented in the ESI. ‡ With the two probes in hand, we rst evaluated the spectral response of 1 for N 2 O 3 and ONOO À in PBS buffer (50 mM, pH 7.4, containing 20% CH 3 CN). As shown in Fig. 1A, the solution of 1 itself had an extremely poor uorescence at 518 nm (F ¼ 0.016), presumably due to PeT from the electron-rich N-benzyl-4-methoxyaniline unit to the excited Bodipy core; however, upon treatment with excess NO solution under aerobic conditions (conditions for producing N 2 O 3 ), [23][24][25][26][27][28] a signicant uorescence enhancement (880-fold) was observed from the dark background, indicating that the reaction of 1 with N 2 O 3 could efficiently block the PeT process and thereby turn-on the uorescence. Notably, as revealed by the kinetics study (Fig. 1A, inset), the uorescence response of 1 for N 2 O 3 was fairly fast and could be completed within 10 s, indicative of the potential of 1 for real-time imaging of endogenous N 2 O 3 in biosystems. The uorescence titration assay was further performed to evaluate the sensitivity of 1 for N 2 O 3 . As shown in Fig. S1 (ESI ‡), a good linearity between the uorescence intensities at 518 nm and the concentrations of added NO (0-16 mM) was observed, and the detection limit (DL) for N 2 O 3 (4NO + O 2 ¼ 2N 2 O 3 ) was calculated to be as low as 0.4 nM based on 3s/k. When compared with the N-benzyl-4hydroxyaniline-functionalized Bodipy probe reported by us previously, 44 1 displays a bigger uorescence off-on response and higher detection sensitivity for N 2 O 3 , indicating that the Nbenzyl-4-methoxyaniline group of 1 should be a more excellent reaction group for N 2 O 3 than the N-benzyl-4-hydroxyaniline group. Importantly, when 1 was treated with excess ONOO À under the same conditions, a rapid and great uorescence offon response was also observed, which is almost consistent with the case of N 2 O 3 in either uorescence intensity or response kinetics (Fig. 1B). Moreover, a good linear correlation between the uorescence intensities and the concentrations of ONOO À in the range of 0-2.5 mM was also found (Fig. S2, ESI ‡), and the DL for ONOO À was calculated to be 0.14 nM based on 3s/k. The results reveal that 1 is extremely sensitive not only for N 2 O 3 but also for ONOO À , thus being very promising as a more sensitive indicator to reect intracellular NO.</p><p>To establish the selectivity, we tested the uorescence response of 1 toward various biologically relevant species, including reactive oxygen/nitrogen species (ROS/RNS:</p><p>À , NO, and ONOO À ), DHA/AA/MGO, metal ions (K + , Ca 2+ , Na + , Mg 2+ , Al 3+ , Zn 2+ , Fe 2+ , Fe 3+ , and Cu 2+ ), and biothiols (Cys and GSH). As shown in Fig. 1C, the treatment of 1 with either N 2 O 3 (NO/O 2 ) or ONOO À could induce a signicant uorescence off-on response, while the other competitive species failed to give any obvious uorescence alteration of 1, indicating that the probe is highly specic for N 2 O 3 and ONOO À . In addition, 1 was almost nonuorescent in the pH range of 5-9, but displayed the best uorescence response for N 2 O 3 and ONOO À at 7.4 (Fig. S3, ESI ‡), thus being suitable for imaging application at physiological pH. Encouraged by the above results, we further tested the uorescence sensing performances of Mito1 for both N 2 O 3 and ONOO À under the same conditions. Indeed, the probe was designed as a mitochondria-targetable uorescent NO probe by installing a mitochondria-targeted triphenylphosphonium (TPP) cation 50,51 to the molecular skeleton of 1. Interestingly, as shown in Fig. S4-8 (ESI ‡), Mito1 displayed almost the same sensing performances for N 2 O 3 and ONOO À as 1, such as the signicant and rapid uorescence off-on response, high selectivity and sensitivity, and excellent uorescence response at Overall, as revealed by the above assays, 1 and Mito1 displayed high sensitivity, excellent selectivity, and fast response ability for both N 2 O 3 and ONOO À under the simulated physiological conditions, thus holding great potential for probing NO-related physiology and pathology.</p><p>Sensing mechanisms of 1 and Mito1 for N 2 O 3 and ONOO À With 1 as a representative, we subsequently studied the sensing mechanisms of the probe for both N 2 O 3 and ONOO À by HPLC-HRMS assays. As shown in Fig. S9 (ESI ‡), the HPLC analysis showed that the reaction of 1 with N 2 O 3 mainly produced a new peak, which could be assigned to N-nitroso product 1-NO in terms of HRMS data (m/z calcd for [M + H + ] 489.2273, found 489.2263) (Scheme 2A). This is consistent with the previous report that aromatic secondary amines can react with NO under aerobic conditions to give the N-nitroso product. 44 However, in the case of ONOO À , in addition to the major N-nitroso product 1-NO (m/z: calcd for [M + H + ] 489.2273, found 489.2263), two unknown new products, one major and the other minor, were observed as well in HPLC analysis (Fig. S10, ESI ‡). Considering that secondary amines can react with ONOO À to produce both N-nitroso and N-nitro products, 52 we proposed a possible reaction mechanism as follows (Scheme 2B): rst, the reaction of 1 with ONOO À produced the N-nitroso product 1-NO and N-nitro product 1-NO 2 ; due to the strong push-pull electronic interaction between the -OMe group and -NO 2 group, 1-NO 2 was unstable and underwent an intramolecular two-electron transfer to give p-benzoquinone imine intermediate B1; the intermediate was also unstable and could be attached by the H 2 O molecule via a Michael addition-like reaction to generate intermediate B2; the oxidative dehydrogenation of B2 afforded o-benzoquinone imine BQI as the nal product of the reaction pathway. According to the proposed mechanism, the abovementioned two unknown products could reasonably be assigned to B1 (minor) and BQI (major) in terms of the excellent matching of calculated and observed m/z values (for B1, calcd for [M + ] 458.2215, found 458.2201; for BQI, calcd for [M + H + ] 474.2164, found 474.2154) (Fig. S10, ESI ‡). Thus, the HPLC-HRMS assays nicely support our proposed reaction mechanisms of 1 for N 2 O 3 and ONOO À .</p><p>Further, the uorescence off-on response of 1 for N 2 O 3 and ONOO À by inhibiting the PeT process was rationalized by the Frontier orbital energy diagrams of 1, 1-NO, and BQI, obtained by Becke's three-parameter hybrid exchange function with the Lee-Yang-Parr gradient-corrected correlation functional (B3LYP functional) and 6-31+G* basis set (Fig. S11, ESI ‡). To support the conclusion, we studied the uorescence changes of 1 in mixed water-glycerol systems (0-100% of glycerol) with varied viscosity. As shown in Fig. S12 (ESI ‡), in these cases 1 still displayed negligible uorescence, strongly indicating that no uorescence of 1 is indeed due to the PeT process, rather than rotation or vibration-relevant nonradiative processes. 53 Basic imaging ability of 1 and Mito1 for N 2 O 3 and ONOO À in living cells as well as their subcellular distribution Prior to biological imaging applications, the cytotoxicity of 1 and Mito1 was rst tested in HeLa cells by MTT assays. As shown in Fig. S13 (ESI ‡), aer 24 h of cellular internalization of less than 8 mM of 1 or Mito1, >90% of the cells remained viable, indicative of the good biocompatibility of the two probes. Notably, 1 displayed an obviously lower cytotoxicity than Mito1, presumably due to its uncharged property reducing its interaction with either the negatively charged DNA or the mitochondrial membrane with highly negative potential. Even so, in order to reduce the interference to cell proliferation and physiology, a low concentration of 1 or Mito1 (2 mM), survival rates close to 100% in the case, was used in the subsequent bioimaging assays. Subsequently, we evaluated the selectivity of 1 or Mito1 for N 2 O 3 and ONOO À in human cervical cancer HeLa cells. As shown in Fig. 2, HeLa cells loaded with 1 or Mito1 showed negligible background uorescence; when the 1or Mito1-loaded HeLa cells were treated with NOC-9 (a commercial NO donor) or SIN-1 (a commercial ONOO À donor), a strong intracellular green uorescence was observed for both cases; when 1or Mito1-loaded HeLa cells were treated with representative ROS, such as H 2 O 2 and ClO À , almost no any intracellular green uorescence was found. The results suggest that 1 and Mito1 still possess high specicity for N 2 O 3 and ONOO À in a cell environment.</p><p>Encouraged by the above results, we further tested the ability of 1 or Mito1 for imaging endogenous N 2 O 3 and ONOO À in mouse RAW264.7 macrophages that are known to express highlevel iNOS upon stimulation by LPS/IFN-g. 54 As shown in Fig. 3, RAW264.7 cells themselves were nonuorescent; upon incubation with 1 or Mito1, the cells displayed a weak yet clear intracellular green uorescence; when the cells were pretreated with NO synthase inhibitor aminoguanidine (AG) 55 and then treated with 1 or Mito1, the intracellular green uorescence was greatly inhibited. results indicate that the two probes are sensitive enough to determine the basal level of intracellular NO by responding to N 2 O 3 or ONOO À . Further, when the cells were stimulated with LPS/IFN-g and then treated with 1 or Mito1, a bright intracellular green uorescence was clearly observed; when the cells were stimulated with LPS/IFN-g in the presence of AG and then treated with 1 or Mito1, the intracellular green uorescence was greatly inhibited. Thus, the two probes can be used to sense the LPS/IFN-g-triggered outburst of endogenous NO, indicating their potential for studying various NO-related pathophysiological events. Also, we tested the subcellular distribution of 1 and Mito1 in HeLa cells by costaining assays. In the assays, NOC-9 was used to light up the two probes in cells, and Pearson's correlation coefficient (R) was used to analyze the linear correlation of uorescence signals between the green channel (for probes) and red channel (for commercial trackers). As shown in Fig. 4A, when HeLa cells were co-incubated with 1/MitoTracker or 1/LysoTracker followed by NOC-9 treatment, a poor overlapping image was observed for both cases (R ¼ 0.35 and 0.31, respectively), indicating that 1 is not specic for either mitochondria or lysosomes. However, when HeLa cells were co-incubated with 1, MitoTracker, and LysoTracker followed by NOC-9 treatment, we observed an excellent overlapping image from the green channel and red channel (R ¼ 0.88), indicating that 1 was indeed distributed over both mitochondria and lysosomes. However, as shown in Fig. 4B, when HeLa cells were coincubated with Mito1/MitoTracker or Mito1/LysoTracker followed by NOC-9 treatment, a good overlapping image along with a high Pearson's correlation coefficient was only observed for the former (R ¼ 0.89) but not the latter (R ¼ 0.10), indicating that Mito1 could preferably localize in mitochondria rather than lysosomes. The excellent localization of Mito1 in mitochondria could indeed be attributed to its lipophilic TPP cation that directs the probe into mitochondria by the highly negative potential of the mitochondrial membrane (about À180 mV). 50,51 Potential applications in cancer immunotherapy studies Having established their excellent imaging ability for NO in chemical systems and living cells by responding to both N 2 O 3 and ONOO À , we envisioned that 1 and Mito1 should have the Fig. 3 Confocal images of the basic and stimulator-induced NO in RAW 264.7 macrophages using (A) and Mito1 (B). For imaging of intracellular basal NO, cells were treated directly with 1 or Mito1 (2 mM, 20 min) in PBS; for imaging of the stimulator-induced NO, cells were pretreated with stimulators LPS (20 mg mL À1 )/INF-g (150 units per mL) for 6 h in PBS and then treated with 1 or Mito1 (2 mM, 20 min); for inhibition assays, cells were pretreated with LPS (20 mg mL À1 )/INF-g (150 units per mL) for 6 h in the presence of AG (0.5 mM) and then treated with 1 or Mito1 (2 mM, 20 min). Emission was collected at 493-600 nm (l ex ¼ 488 nm). Scale bar: 10 mm. potential to distinguish between M1 M2 phenotypes in terms of their difference in the iNOS level and thus the NO level. [5][6][7] To this end, we set up a model of human macrophage polarization according to a previously reported method. 3 Briey, human monocytic THP-1 cells were rst differentiated into macrophages by 24 h incubation with phorbol 12-myristate 13acetate (PMA) followed by 24 h incubation in RPMI medium; then, the macrophages were polarized in M1 macrophages by incubation with IFN-g/LPS, and in M2 macrophages by incubation with interleukin 4 (IL-4) and interleukin 13 (IL-13) (Fig. 5A). Aer a thorough wash to remove all stimuli, the M1and M2-polarized macrophages were treated with 1 (as a representative) and then imaged under a confocal laser scanning microscope. As shown in Fig. 5B, a bright intracellular green uorescence could be observed in 1-loaded M1 macrophages, but not in 1-loaded M2 macrophages; moreover, when M1 macrophages were pretreated with NO synthase inhibitor AG and then treated with 1, the intracellular green uorescence was greatly inhibited. The results indicate that 1 could discriminate M1 macrophages from M2 macrophages in terms of their difference in the iNOS level and thus the NO level. Of note, when M2 macrophages were pretreated with IFN-g/LPS and then treated with 1, the intracellular green uorescence was greatly recovered, consistent with the report that M2 macrophages could be repolarized to M1 macrophages by IFN-g/LPS treatment. 9,10 Finally, we tested the ability of 1 to image NO communication during the phagocytosis of cancer cells by macrophages in a co-culture system containing M1-or M2-polarized macrophages and SKOV-3 human ovarian cancer cells. Briey, the M1or M2-polarized macrophages were rst stained with commercial blue-uorescent nucleus dye DAPI, and then co-cultured with 1-loaded SKOV-3 cells for 12 h, followed by imaging under a confocal laser scanning microscope. Note that the use of DAPI dye is in order to distinguish M1 or M2 macrophages from SKOV-3 cells in the co-culture system. As shown in Fig. 6A, the SKOV-3 cells cultured alone displayed almost no intracellular uorescence in the green channel when treated with 1, indicating that the cancer cells express negligible intracellular NO. Upon capture by M1 macrophages in the co-culture system, the 1-loaded SKOV-3 cells exhibited a strong intracellular green uorescence (Fig. 6B), indicating that during the immunemediated phagocytosis of cancer cells, NO secreted by M1 macrophages could diffuse across the cell membrane of cancer cells to exert its tumoricidal inuence by producing cytotoxic N 2 O 3 and ONOO À . The result was further supported by an inhibition assay, where the 1-loaded SKOV-3 cells captured by the AG-pretreated M1 macrophages displayed an obviously decreased intracellular green uorescence (Fig. 6C). In sharp contrast, when the 1-loaded SKOV-3 cells were captured by M2 macrophages in the co-culture system, almost no intracellular green uorescence could be observed in the former (Fig. 6D), in line with the low iNOS level in M2 macrophages. However, when M2 macrophages were pretreated with IFN-g/LPS and then For the probe channel, emission was collected at 493-600 nm (l ex ¼ 488 nm); for the DAPI channel, emission was collected at 410-485 nm (l ex ¼ 405 nm). Scale bar: 5 mm. co-cultured 1-loaded SKOV-3 cells, the captured SKOV-3 cells displayed a dramatically increased intracellular green uorescence (Fig. 6E), conrming that M2 macrophages can be polarized to M1 macrophages by IFN-g/LPS. 9,10 Overall, the above results conrm that 1 could be used to discriminate M1 macrophages from M2 macrophages, evaluate the repolarization of TAMs from pro-tumoral M2 phenotype to anti-tumoral M1 phenotype, and image NO communication between macrophages and cancer cells during the immunemediated phagocytosis process, thus being very promising for imaging applications in cancer immunotherapy studies and relevant anti-cancer drug screening.</p><!><p>In summary, we in this work presented an aromatic secondary amine-functionalized Bodipy dye 1 and its mitochondriatargetable derivative Mito1 as highly sensitive uorescent NO probes for discriminating M1 macrophages from M2 macrophages in terms of their difference in the iNOS level and thus the NO level. The high sensitivity of the two probes for NO originates from their unique ability to simultaneously respond to N 2 O 3 and ONOO À . In this regard, the two probes should be superior to most of the existing uorescent NO probes that can only respond to N 2 O 3 . Although an o-diamine-locked rhodamine lactam derivative has been reported to be able to give a similar uorescence off-on response for both N 2 O 3 and ONOO À , 38 the later studies found that this type of probe could suffer from serious interference by intracellular abundant Cys to lead to decreased sensitivity for NO. 56 Importantly, with 1 as a representative, we have successfully realized the discrimination of M1 macrophages from M2 macrophages, evaluation of the repolarization of TAMs from pro-tumor M2 phenotype to anti-tumor M1 phenotype, and visualization of NO communication during the immune-mediated phagocytosis of cancer cells by M1 macrophages. Thus, our probes should hold great potential for NO-related physiological and pathological studies as well as anticancer drug screening in cancer immunotherapy.</p><!><p>There are no conicts to declare.</p>

Royal Society of Chemistry (RSC)

Effect of the duty cycle of the ultrasonic processor on the efficiency of extraction of phenolic compounds from Sorbus intermedia

This paper studies the effect of different ultrasonic parameters on the yield of extraction and antioxidant activity of selected phenolic compounds from Sorbus intermedia berries. The sonication was carried out in two modes: continuous and pulse. In the pulse mode, the samples were sonicated with the following processor settings: 1 s on-2 s off. The effective ultrasonic processor times were 5, 10, and 15 min, and the total extraction times were 15, 30, and 45 min. The results showed that the duty cycle significantly affected the antioxidant activity of the extracts and the yield of chlorogenic acid, rutin, and total flavonoids. Compared to the continuous mode, the pulse ultrasound increased the extraction yield of rutin by 5-27%, chlorogenic acid by 12-29%, and total flavonoids by 8-42%. The effect of the duty cycle on the extraction yield was dependent on the intensity and duration of the ultrasound treatment. The mechanism of the influence of the pulsed ultrasound field on the extraction process has been elucidated. This research clearly demonstrated the superiority of pulsed ultrasoundassisted extraction for production of antioxidants from Sorbus intermedia berries.Given the rising costs of energy and the need to reduce greenhouse gas emissions, the food, pharmaceutical, and natural pharmaceutical industries are focused on the use of more efficient and environmentally friendly techniques for extraction of bioactive substances 1 . Ultrasound-Assisted Extraction (UAE) may be one of such methods. Compared to other techniques, this type of separation has many advantages, e.g. higher efficiency, lower process temperature, low consumption of solvents, and relatively low costs of equipment 2 . Moreover, this economical method shortens the process time and lowers energy consumption 3,4 .The mechanism of the ultrasound-assisted extraction process consists in intensification of the extraction of bioactive compounds from plant raw materials with the use of high-intensity sound waves. Sonication contributes to partial disintegration of plant tissue through acoustic cavitation and accelerates the release of extracted components into the solvent by increasing the mass transport 5 .The efficiency of ultrasound-assisted extraction is influenced by the following factors: the ultrasound intensity and frequency, the process time and temperature, the liquid-to-solid ratio, the type, concentration, and pH of the solvent, and the sonicator duty cycle (continuous or pulse) 6,7 .The UEA efficiency increases with increasing ultrasound intensity and then decreases after reaching a certain critical value specific for a given bioactive substance 8,9 . The higher efficiency of the process is related to increased disintegration of plant tissue and increased solvent penetration related to a rapid collapse of cavitation bubbles. However, high-intensity ultrasound may degrade the bioactive compound, thereby reducing the extraction yield 10 .Another determinant of the extraction efficiency is the ultrasound frequency. The number of reports on this issue is limited, and most research has been carried out at a constant frequency of ultrasound. Researchers usually use low ultrasound frequencies due to the possibility of induction of the cavitation phenomenon at a lower level of ultrasound intensity [11][12][13][14][15][16] .

effect_of_the_duty_cycle_of_the_ultrasonic_processor_on_the_efficiency_of_extraction_of_phenolic_com

<!>Materials and methods<!>Statistical analyses.<!>Results<!>Discussion<!>Conclusions

<p>Another important parameter of the extraction process is the selection of an appropriate solvent. Ethanol and its solutions are used most commonly. It has been found that ethyl alcohol has the highest affinity for phenolic compounds in many systems and is therefore the first choice for the extraction of phenolic compounds from plant materials 17 . An increase in the concentration of ethanol in an aqueous solution initially increases the yield of phenolic compound extraction and then reduces the efficiency 10 . The efficiency of ultrasound-assisted extraction also rises with the increasing liquid-to-solid ratio. A greater difference in solute concentrations increases the diffusivity and dissolution of the solute in the solvent, thereby supporting the extraction process. The cavitation intensity at a high liquid-to-solid ratio is higher and causes greater fragmentation, erosion, and pore-formation, thus increasing the extraction efficiency. However, the bioactive substance may undergo degradation after a certain level of cavitation intensity has been exceeded 18,19 .</p><p>The number of reports on the impact of solvent pH on the extraction of phenolic compounds is limited, and only Rodrigues et al. 20,21 determined the optimal solvent pH to be in the range of 1-3.</p><p>The temperature is an important determinant of the extraction efficiency. It improves the desorption properties and solubility of the substance dissolved in the solvent. Additionally, it reduces the viscosity of the solvent, which increases the diffusivity of the solvent and the solute. In turn, an excessive temperature rise weakens the cavitation effect 10,22 .</p><p>The effect of the ultrasound-assisted extraction time on the content of bioactive compounds has been extensively studied. It is known that a longer sonication time initially increases but then reduces the efficiency. The initial extension of the sonication time accelerates the processes of raw material swelling and hydration as well as fragmentation of plant cells in which the soluble substance is dissolved. An extremely long ultrasonic extraction time results in solute degradation and reduces the extraction efficiency. In the case of the extraction of phenolic compounds, the optimal extraction time has been estimated at 10-90 min, depending on the type of raw material [10][11][12][13][14][15][16] .</p><p>A less known issue is the effect of the ultrasonic processor duty cycle on the efficiency of extraction of bioactive substances. Pan et al. 23 reported no significant differences in the extraction yield and duration related to the ultrasound duty cycle, but approximately 50% lower electricity consumption was observed in the pulse mode. Similar results were obtained by Patience et al. 24 during the extraction of pectins from orange peels. There was no significant difference in the yield between the continuous and pulse modes despite the large difference in power consumption (190 kJ vs. 80 kJ).</p><p>In contrast, Christou et al. 25 recovered polyphenols from mature and unripe carob (carob flour) and achieved slightly higher yields with the use of pulse-mode UAE. In turn, Kobus et al. 26 showed that the use of the pulsed ultrasonic field resulted in a 1.14-34% increase in the yield of extracted polyphenols and anthocyanins in comparison with the continuous ultrasonic field variant. Xu et al. 19 showed that, compared to CHE (conventional heating extraction), ultrasound-assisted extraction in the pulse mode (2 s on-2 s off) improved the efficiency of the process by 26.74% and shortened the extraction time (51.79 min).</p><p>Currently, there is increasing interest in extraction methods and research of bioactive substances contained in less known fruit species due to their potential health benefits in the prevention of chronic diseases 27 . Sorbus L. species are widely distributed in the northern hemisphere 28 . Sorbus intermedia is native to north-western Europe but now naturally found in other parts of Europe 29 . Extracts of various Sorbus L. species are used as a traditional remedy for various digestive disorders, treating respiratory tract infections, flu, fever, cold, rheumatism, anaemia, gout, oedema, dyspepsia 30 . The chemical composition of fruits of various species of rowan is similar 31 . Rowan fruits contain organic acids: sorbic, parasorbic, citric, malic, tartaric and succinic, phenolic acids: vanillin, coffee, gallic, p-coumaric, ellagic, syringic, ferulic, benzoic and protocatechic 31,32 . It has also been found flavonoids: quercetin, kaempferol, isorhamnetin, rutin, isoquercetin and jaceosidin, anthocyanins, vitamins and bioelements 31 . Phytochemical studies on Sorbus intermedia are limited. Articles refer to the presence of flavonoids and phenolic acids [33][34][35][36] . Sołtys et al. 37 analyzed of Sorbus intermedia fruits with regard to their triterpenoid composition.</p><p>The promising results of the research on the extraction of polyphenols and anthocyanins from hawthorn with the ultrasound-assisted method have prompted us to extend the scope of the research to include other raw material and other bioactive substances. Hence, the aim of the study was to compare the effect of the ultrasonic processor duty cycle on the yield of bioactive compounds and the antioxidant activity of the extracts.</p><!><p>Study material. Plant material (fruits) of Sorbus intermedia (Ehrh.) Pers was collected in October of 2020 in Lublin Upland in Poland. DSc Wojciech Durlak identified it as Sorbus intermedia. The specimen vouchers (S. intermedia No. 12/2021) were deposited by DSc Wojciech Durlak in the herbarium of the Horticultural Production Institute, Faculty of Horticulture and Landscape Architecture, University of Life Sciences in Lublin. The collection of plant material and experiment had been conducted in compliance with relevant guidelines and regulations. Immediately after harvesting, the raw material was freeze-dried in a lyophilizer. Next, the material was crushed and divided into 5 fractions. Three fractions were selected for further analyses. They were obtained by sieving through sieves with different mesh diameters: fraction 1-0.25-0.5 mm mesh diam., fraction 2-0.5-1.0 mm mesh diam., and fraction 3-1.0-2.0 mm mesh diam. Ultrasonic treatment. Four grams of raw material from each fraction were placed in the extraction cell and covered with an aqueous solution of 60% ethyl alcohol. Next, the extraction cell was placed in a cooling jacket connected to an ultra-thermostat to stabilize the temperature. The extraction cell was closed with a 19-mm diameter. ultrasonic probe on the top. The experimental samples were sonicated with a VC750 Sonics processor (Sonics and Materials Inc., USA) operating at a frequency of 20 kHz. The sonication was performed at Chemical analyses. Determination of flavonoid content (TFC). The content of flavonoids was determined with the spectrophotometric method using quercetin as a reference standard 38 . First, the sample extract (1.0 mL) was mixed with 1 mL of a 2% AlCl 3 × 6 H 2 O solution (in methanol), and the mixture was made up to 10 mL with distilled water. After incubation of the mixture for 10 min at room temperature in the dark, the absorbance was measured at 430 nm. A calibration curve was prepared with quercetin and the results were expressed as mg quercetin equivalent per 1 g of dry matter (mg QE g −1 dry matter).</p><p>Determination of chlorogenic acid and rutin by HPLC. Chlorogenic acid and rutin in the extracts were quantified using a modified version of the HPLC (High Performance Liquid Chromatography) method proposed by Alberti et al. 39 . The analysis was conducted using HPLC System S 600 Series equipment (Sykam GmbH, Eresing, Germany) coupled to a photodiode arrangement detector (PDA S 3345, Sykam). All extract were filtered through a 0.20 µm syringe filter (Nylon, Macherey-Nagel, Düren, Germany), and 20 µL of the sample was injected in the system in triplicate. The separation was carried out using a Bionacom Velocity STR (3.0 mm × 250 mm, 5.0 µm) column at 25 °C40 .</p><p>The mobile phases consisted of 25 mL•L −1 of acetic acid (solvent A) and acetonitrile (solvent B). The system was run with at a flow rate of 1.0 mL•min −1 in the following gradient program: 3-9% B (0-5 min), 9-11% B (5-13 min), 13-30% B (13-20 min), 30% B (20-25 min), and 30-3% B (25-27 min). The peaks of the compounds were identified and quantified by comparison of the retention times and spectra of the samples with calibration curves that had been previously prepared with standards. The runs were monitored at 260 nm 40 .</p><p>Determination of antioxidant activity. Determination of free radical scavenging activity with the DPPH assay. The antioxidant activity of the extracts was evaluated using the DPPH (2,2-diphenyl-1-picrylhydrazyl) assay 41 . For this analysis, 0.2 mL of the extract was mixed with an aliquot of 5.8 mL of freshly prepared 6•10 −5 M DPPH radical in methanol. The mixture was allowed to stand at room temperature for 30 min, and the spectrophotometric absorbance was measured at 516 nm using methanol as a blank. 40 The measurement was performed in three replicates for each sample. Antioxidant activity was expressed as a Trolox equivalent in µg per g of dry matter.</p><p>Ferric reducing antioxidant power assay (FRAP). The ferric reducing antioxidant power of the extract was evaluated according to the method described by Benzie & Strain 42 with some modifications. First, the working reagent was prepared as a mixture of 20 mM FeCl 3 , 300 mM acetic acid, and 10 mM of TPTZ in 10:1:1 (v/v/v) proportion in 40 mMHCl. Next, 30-μL of the extracts were mixed with 5 mL of the FRAP solution. After incubation at 37 °C for 10 min, absorbance of the mixture was measured at 593 nm. The FRPA values of the extracts were expressed as a Trolox equivalent in µg per g of dry matter.</p><!><p>The results were statistically analyzed with Statistica software via analysis of variance (ANOVA). The significance of differences between the evaluated mean values (in figures) was analyzed with the Tukey test at a significance level of p < 0.05 26 .</p><p>The tables present the mean values with standard deviations, while the graphs present the mean values and whiskers representing standard deviations. The results of the chemical properties are presented as average values of three measurements in each sample.</p><!><p>Flavonoids. Flavonoids are bioactive compounds with multiple health-enhancing properties, e.g. antioxidant, anti-inflammatory, anti-cancer, and detoxifying activity. Figure 1 shows the effect of the type of treatment (continuous, pulse) on the content of flavonoids in Sorbus intermedia alcoholic extracts. The highest content of flavonoids, i.e. 0.502 mg QE/g, was detected in extracts obtained in the pulse mode at the 36 µm amplitude and 15 min treatment time. The lowest amounts of flavonoids, i.e. 0.132 mg QE/g, were extracted in the continuous mode (12 µm amplitude and 5-min time). The increase in the pulse amplitude from 12 µm to 36 µm resulted in an increase in the flavonoid content from 105 to 216%. Another important determinant of the yield was the extraction time. The differences in the extraction between the shortest and longest extraction time ranged from 12 to 45%. The duty cycle of the ultrasonic processor also had a positive effect on the content of flavonoids. The difference in the flavonoid content between the pulse and continuous mode variants ranged from 8 to 42%. In the case of the 12 µm amplitude, the processor duty cycle did not exert a statistically significant effect on the content of flavonoids, whereas the differences in the case of the 24 and 36 µm amplitudes turned out to be statistically significant. The antioxidant activity of S. intermedia berry extracts was assessed using the FRAP and DPPH methods. DPPH. Table 1 shows the effect of the sonication parameters on the antioxidant activity of Sorbus intermedia extracts determined with the DPPH method. The highest values were obtained in the case of the pulse mode with the following processing parameters: 24 µm amplitude and 15 min time. The lowest values were noted in the continuous mode with 12 µm amplitude and 5 min time. The difference in between the variants with these processing parameters was 129%. The extracts obtained with the use of the pulsed ultrasound field exhibited higher antioxidant activity than the extracts from the continuous mode treatment. These differences turned out to be statistically significant for the 12 µm amplitude and 15 min time and for the 24 µm amplitude and the time of 5, 10, and 15 min. The greatest impact on the antioxidant activity was exerted by the ultrasonic vibration amplitude. The increase in the vibration amplitude from 12 µm to 36 µm resulted in an increase in the DPPH value from 66 to 120%. Another important parameter was the ultrasonic treatment time. An increase in the treatment time from 5 to 15 min resulted in an increase in the DPPH value from 1.5% to 37%. 2 shows the effect of the treatment type (continuous, pulse) on the antioxidant activity of the Sorbus intermedia extracts determined with the FRAP method. The highest antioxidant activity was detected in extracts obtained in the pulse-mode treatment with the following parameters: 36 µm amplitude and 15 min time. The lowest values were noted in the continuous-mode treatment with the 12 µm amplitude and 5 min time.</p><p>The difference between the extreme values reached 287%. The amplitude had the greatest effect on the ability to scavenge free radicals. The increase in the vibration amplitude from 12 µm to 36 µm resulted in an increase in the FRAP value from 96 to 196%. Another important determinant of the antioxidant capacity in the Sorbus intermedia extracts was the extraction time. The increase in the sonication time from 5 to 15 min resulted in an increase in the FRAP value from 12 to 48%. The difference in the FRAP values between the pulse and continuous modes ranged from 11 to 26%. The processor duty cycle was found to exert a statistically significant effect on the reducing power.</p><p>Chromatographic analysis. Chlorogenic acid. Chlorogenic acid is a phenolic compound from the hydroxycinnamic acid family and it is one of the most common compounds among the polyphenol group contained in Sorbus intermedia raw material. Chlorogenic acid possesses many health-promoting properties, most of them related to the treatment of metabolic syndrome, including anti-oxidant, antilipidemic, antidiabetic, anti-inflammatory, and antihypertensive activities. This polyphenol has shown antimicrobial activity against a wide range of organisms 43 . Figure 2 shows the effect of the treatment mode (continuous, pulse) on the content of chlorogenic acid in the Sorbus intermedia alcoholic extracts. In the case of the 12 and 24 µm amplitudes, there was a statistically significant effect of the processor duty cycle on the chlorogenic acid content. The highest values were found in the case of the pulse-mode treatment with the following processing parameters: 36 µm amplitude and 10 min time. The lowest values were obtained in the continuous-mode sonication at the 12 µm amplitude and 15 min time. The difference between these values was estimated at 176%. The ultrasonic vibration amplitude exerted the greatest impact on the chlorogenic acid content. The increase in the vibration amplitude from 12 µm to 36 µm resulted in an increase in the chlorogenic acid content from 84 to 146%. The increase in the treatment time from 5 to 15 min contributed to an increase in the chlorogenic acid content in the range from 1 to 23%. The differences in the content of this acid between the variants with the different processor duty cycles ranged from 12 to 29%. www.nature.com/scientificreports/ Rutin. Rutin is an organic chemical compound from the group of flavonoid glycosides. Rutin demonstrated a lot of medical activities, including anticarcinogenic, antioxidant, chemotherapeutic, cytoprotective, neuroprotective, vasoprotective, and cardioprotective. Rutin has an effect on the prevention of neuroinflammation, promotion of neural crest cell survival, sedative activity, anticonvulsant activity, anti-Alzheimer activity, and treatment of hyperkinetic movement disorder, antidepressant, analgesic and antinociceptive, antiarthritic, antidiabetic, anti-hypercholesterolemic, antiplatelet aggregatory, antiulcer antiasthmatic activity and other associated, antiosteoporotic and antiosteopenic, anticataract and ophthalmic, diuretic, anticancer 44 . Figure 3 shows</p><!><p>The content of bioactive compounds is strongly determined by the genotype and habitat conditions. This indicates that, in addition to the species, such factors as plant height, light, temperature, content of nutrients available in the soil, the sampling site, and the harvesting time (maturity stage) play an important role 45 . Moreover, the content of bioactive compounds depends on the extraction method.</p><p>There are no literature reports on the analysis of the TFC content in Sorbus intermedia. The number of available literature reports on the content of chlorogenic acids, rutin and values of DPPH and FRAP in Sorbus intermedia is limited, and the comparison of the present results poses many problems. Therefore, the discussion refers to other Sorbus species.</p><p>The values of TFC obtained in the present experiment ranged from 0.132 to 0.502 mg QE/g DM. Jin et al. 46 evaluated the chemical composition of Sorbus commixta fruit extracts prepared with different ethanol concentrations. The total flavonoid content ranged from 2.49 to 8.65 μg QE/mg of extract. Turumtay et al. 32 showed that the values of TFC for Sorbus aucuparia was 12.336 mg QE/g dried extract and for endemic Sorbus caucasica var. yaltirikii was 17.128 mg QE/g dried extract. Majić et al. 47 assess service tree (Sorbus domestica L.) bark, fruit exocarp and mesocarp, and seeds. The value of TFC ranged from 6.8 to 37.0 mg QE/g DM.</p><p>Several methods have been developed to determine antioxidant activity of plant materials. In our study, two methods (DPPH and FRAP) were used to evaluate the antioxidant potential of Sorbus intermedia fruit. FRAP is based on ferric reducing power and DPPH is based on the ability to scavenge DPPH radical. The antioxidant activity of the extracts analyzed in the present study determined with the DPPH method ranged from 32.79 to 74.38 µmol TE/g DM, while the extracts determined with the FRAP method ranged from 11.72 to 45.31 µmol TE/g DM. There are very little data in the literature with which our results could be easily compared, given the methods of extraction, analysis, units and plant species. Mrkonjić et al. 48 determined DPPH and FRAP values in methanolic extracts from the fruits of Sorbus intermedia, but they express their results in ascorbic acid equivalents. They obtained the following data: DPPH 0.20 ± 0.01 mg of AAEc/g DW and FRAP 4.47 ± 0.39 mg of AAEc/g DW. Olszewska and Michel 36 evaluated the antioxidant potential of 70% methanolic extracts from the fruits of Sorbus aucuparia, Sorbus aria and Sorbus intermedia. According to their research the fruits of Sorbus intermedia showed antiradical efficiency of 86.9 µmol TE/g DM for DPPH method and 221.1 µmol TE/g DM for FRAP method. The results of the present experiment show that the levels of DPPH values coincide with the values reported by Olszewska and Michel 36 , but FRAP values are much lower. Probably the main reason for the differences in antioxidant activity determined by the FRAP method was the different incubation time of the samples. Both Olszewska and Mitchel 36 and Stratil et al. 49 proved that DPPH and FRAP tests are strongly depended on incubation time and the reactivity of various phenolic standards and plant extracts. Antioxidant activity of extracts are also related with the polarity of solvent. Bobinaite et al. 50 demonstrated that antioxidant capacity of Sorbus aucuparia for water extracts in DPPH and FRAP system were 309 µmol TE/g and 323 µmol TE/g, while ethanol extracts were 103 µmol TE/g and 118 µg TE/g, respectively.</p><p>The results of the present experiment show that the levels of chlorogenic acid ranged from 436.43 µg/g DM to 1202.66 µg/g DM. Olszewska and Michel 36 evaluated the content of chlorogenic acid isomers quantified by HPLC in 70% methanolic extracts from the fruits of Sorbus intermedia and the value of chlorogenic acid was 0.23 ± 0.01%. Šavikin et al. 51 showed that the levels of chlorogenic acid in Sorbus aucuparia extracts ranged from 0.35 to 10.01 mg/g DW, while in Sorbus aria ranged from 0.22 to 2.30 mg/g DW. Kylli et al. 52 indicated that the value of chlorogenic acid in wild rowanberries (Sorbus aucuparia) was 5.36 ± 0.10 mg/g D. Mrkonjić et al. 53 determined chlorogenic acid in water and methanol extracts, and in the jam of Sorbus aucuparia. The chlorogenic acid content was were 5.69, 5.80 and 2.60 mg/g DW, respectively. Bobinaitė et al. 50 evaluated acetone, ethanol and water extracts of rowanberry (Sorbus aucuparia L.) pomace. The value of chlorogenic acid was the highest in ethanol extract (3970 µg/g extract). Jin et al. 46 showed that the concentration of chlorogenic acid ranged from 111.81 to 344.7 µg/g extract.</p><p>Olszewska 35 showed that rutin and isoquercitrin were predominant components of the extracts. In the present study, the rutin levels ranged from 37.22 to 173.76 µg/g DM. The concentrations of rutin in water and methanol extracts of Sorbus aucuparia were 82.3 and 80.4 mg/g DW, respectively 48 . Šavikin et al. 51 indicated that rutin content ranged from 138.4 to 892.0 μg/g DW in Sorbus aria and 40.1 to 598.3 μg/g DW in Sorbus aucuparia extracts. Bobinaitė et al. 50 showed that the value of rutin in ethanol extract was 123.9 µg/g extract. Rutin content in ethanolic extracts of Sorbus commixta ranged from 0.28 to 0.6 µg/g extract 46 .</p><p>The growing interest in green chemistry and the possibility of using ultrasounds have contributed to the increasing application of this method in the extraction of bioactive compounds. Ultrasounds increase mass transfer and contribute to disintegration of plant tissues and cells, which results in a shorter extraction time and higher yields. The available literature provides many examples of the application of ultrasounds to intensify the polyphenol extraction process. Phenolic compounds and antioxidant compounds have been extracted from mango peel 12 , coconut shell powder 20 , pomegranate peel 23 , jabuticaba peel 21 , grape by-product 16 , black chokeberry by-products 17 and used coffee grounds 22 .</p><p>Pulsed ultrasonic field is increasingly being used to support the extraction process. However, the number of studies focused on the direct comparison of the continuous and pulse ultrasonic processor modes in the extraction of phenolic compounds is highly limited. Pan et al. 23 found no significant differences between the processor duty cycles during extraction of polyphenols from pomegranate peel. However, the pulsed ultrasonic field was found to save approximately 50% of energy in comparison with the continuous sonication mode.</p><p>Kobus et al. 26 demonstrated a statistically significant effect of the ultrasonic processor duty cycle on the extraction of bioactive substances. The increase in the yield of extracted total polyphenols and anthocyanins ranged from 1.14% to 34% in the case of the pulse mode. The study also showed a reduction in energy consumption from 20 to 51% in the process of extraction assisted with the pulsed ultrasonic field. As reported by Ravanfar 54 , this mode (1 s on-1 s off) had an approx. 1.5-fold higher efficiency than the continuous mode in the case of anthocyanin extraction.</p><p>There are also ambiguous reports on the impact of the duty cycle length on the efficiency of pulsed fieldassisted extraction. Xu et al. 19 observed an increase in the yield of pectin extraction from grapefruit peel along with the increase in the pulse duration/pulse interval ratio from 30 to 50% followed by a decrease after exceeding this threshold. More and Arya 55 reported a significant effect of the interactions between the ultrasonic power and the duty cycle on the efficiency of extraction of polyphenolic compounds from pomegranate peel. Maximum extraction efficiency was achieved with an 80% duty cycle. A similar statistically significant effect of the duty cycle on the yield was obtained by Li et al. 56 during extraction of curcuminoids from dried turmeric and by Kazemi 57 during extraction of phenolics from pomegranate peel. Jain et al. 58 studied the effect of various parameters such as extraction time (min), ultrasonic amplitude (%) and pulse interval (s) with 2 M concentration of hydrotropes and showed that the highest composition of patchoulol i.e. 70.06% was obtained at a 5-min extraction time, 40% ultrasonic amplitude, and a 30:30 s pulse interval. Rakshit et al. 59 showed that an increase in the duty cycle from 60 to 90% was accompanied by an increase in the concentration of punicalagin (98-152 mg/g). A study conducted by Kaderides et al. 60 showed that the combination of 10-s intervals with 10-s pauses in the ultrasonic processor work was the best scheme for extraction of polyphenols from pomegranate peel and ensured low energy consumption. The authors demonstrated that the ultrasonic pulse/pause duration ratio of 1.2/1 should be applied in the process of pomegranate peel extraction. Appropriate selection of the pulse duration and intervals between pulses is highly important for reduction of the total extraction time and energy consumption. The extraction yields increased with the decreasing value of the pulse duration/pause ratio 61 . A further reduction in this ratio may cause degradation of plant material 60 . This indicates that an adequate use of the duty cycle may have a beneficial effect on the extraction efficiency. In contrast, Luque-Garcia and De Castro 62 reported that the duty cycle exerted an insignificant effect on fat recovery from oleaginous seeds.</p><p>In the present study, there was a positive and statistically significant effect of the ultrasonic processor duty cycle on the extraction efficiency of all substances, i.e. total flavonoids, rutin, and chlorogenic acid, and on the antioxidant activity of the extracts determined with the FRAP and DPPH methods. The increase in the extraction yield in the pulse mode variant ranged from 8 to 42% for TFC, from 11 to 26% for FRAP, from 7 to 23% for DPPH, from 12 to 29% for chlorogenic acid, and from 5 to 27% for rutin and depended on the amplitude of ultrasonic vibrations and the processing time.</p><p>The higher extraction efficiency can be explained by the slight differences in the mechanism of the pulsed acoustic field compared to the continuous field. The short duration of ultrasonic pulses results in the formation of fewer cavitation bubbles than in the case of the continuous field treatment. The collapse of such bubbles generates a stronger shock wave, which results in greater fragmentation of plant tissue. The pause following the collapse of the cavitation bubbles allows soluble substances to diffuse into the surrounding solvent. The implosion of cavitation bubbles in the subsequent duty cycle of the ultrasonic processor induces defragmentation of further parts of the plant tissue and contributes to release of another portion of soluble substances. This mode of interaction causes gradual destruction of plant tissue and a lower concentration of the soluble substance in the boundary layer than in the case of the continuous field treatment. Additionally, the smaller number of cavitation bubbles at the surface of the ultrasonic probe reduces the likelihood of the so-called the saturation effect, which reduces the transmission of acoustic energy to the solvent surrounding the probe. The lower concentration of the soluble substance within the solid matrix accelerates the diffusion of the extracted component into the surrounding solvent.</p><p>The extraction temperature is an important parameter that may be responsible for the differences in the efficiency of the ultrasonic processor duty cycle. The increase in temperature results in an increase in the yield on the one hand but substantially reduces the cavitation effect of ultrasounds on the other hand. This has been confirmed in many studies 63,64 .</p><p>It should also be noted that despite the use of a temperature stabilizing system directly under the ultrasonic probe, the solvent temperature in the continuous mode may temporarily reach considerably higher values than in the case of the pulse mode. In the pulse-mode treatment, the cooling system has substantially more time to dissipate successively generated portions of heat from the surroundings of the ultrasonic probe. The increasing temperature under the ultrasonic probe may be another element that reduces the efficiency of extraction in the continuous field mode.</p><p>Summing up, taking into account the type of the processor duty cycle, at least the same or higher extraction yield of each bioactive substance was achieved using the pulse mode. The phenomenon of the different effects of the processor duty cycle on the efficiency of extraction of bioactive components from Sorbus intermedia requires further research.</p><!><p>The present study has shown that the efficiency of the extraction process is strongly determined by the ultrasonic processor duty cycle. This analysis is the first that compares the two operating modes of an ultrasonic processor during extraction of polyphenols from Sorbus intermedia berries.</p><p>A higher statistically significant value of the process efficiency was obtained in the pulse-mode variant. The largest difference, i.e. in the range from 8 to 42%, was observed during the extraction of total flavonoids analyzed by spectrophotometry. Smaller differences were found for individual polyphenols analyzed with the chromatographic method. In the case of chlorogenic acid, the differences were from 12 to 29% and in the case of rutin from 5 to 27%. The effect of the processor operating mode on individual polyphenolic compounds depended on the amplitude of ultrasonic vibrations.</p>

Scientific Reports - Nature

Bioinspired Methodology for Artificial Olfaction

Artificial olfaction is a potential tool for noninvasive chemical monitoring. Application of \xe2\x80\x9celectronic noses\xe2\x80\x9d typically involves recognition of \xe2\x80\x9cpretrained\xe2\x80\x9d chemicals, while long-term operation and generalization of training to allow chemical classification of \xe2\x80\x9cunknown\xe2\x80\x9d analytes remain challenges. The latter analytical capability is critically important, as it is unfeasible to pre-expose the sensor to every analyte it might encounter. Here, we demonstrate a biologically inspired approach where the recognition and generalization problems are decoupled and resolved in a hierarchical fashion. Analyte composition is refined in a progression from general (e.g., target is a hydrocarbon) to precise (e.g., target is ethane), using highly optimized response features for each step. We validate this approach using a MEMS-based chemiresistive microsensor array. We show that this approach, a unique departure from existing methodologies in artificial olfaction, allows the recognition module to better mitigate sensor-aging effects and to better classify unknowns, enhancing the utility of chemical sensors for real-world applications.

bioinspired_methodology_for_artificial_olfaction

<!>EXPERIMENTAL SECTION<!>Conventional, All-at-Once Approaches<!>Divide-and-Conquer Approach<!>CONCLUSIONS<!>

<p>Biological olfactory systems have the extraordinary ability to demonstrate reliable recognition of odorants over long time scales and to classify new odorants based upon chemical similarity to those that have been previously learned. Inspired by the olfactory system, artificial devices that combine arrays of chemical sensors with pattern recognition techniques, commonly termed "electronic noses", have been explored for use as inexpensive chemical point detectors.1–5 There is increasing demand for such devices to fill vital roles in medical diagnostics, homeland security, industrial process control, and other areas. In spite of their high potential for use in a variety of applications, these devices have achieved only limited success, largely due to variability in sensor output, which affects recognition of odorants over time, and a lack of methods to allow generalization of training to new analytes. These two issues are at the core of chemical sensor research.</p><p>Correct recognition of a specific chemical requires detection, in some way, of an aspect of the molecular features unique to that analyte. On the other hand, generalization to unknown chemical species requires detection of features that are common across a desired class of analytes. The opposing nature of these constraints suggests that achieving both capabilities requires a multistep chemical identity resolution process. This is in opposition to the conventional, one step ("all-at-once") procedure that has been used in engineered chemical sensors to date.6–11 Furthermore, for practical use, chemical sensors must be able to cope with variability in sensor responses due to various sources of aging.12 Such aging potentially leads to misalignment between sensor measurements during operation and the chemical finger-prints registered during training, thereby impeding recognition of analytes.</p><p>How does the biological olfactory system deal with these conflicting analytical tasks? Odorant molecules that enter nostrils bind with olfactory receptor neurons in the olfactory epithelium that transduce the chemical stimuli into an electrical neural signal. The biological system incorporates both redundancy (multiple copies of each type of sensory neuron) and diversity (multiple types of sensory neurons depending on the gene they express) to detect chemicals.13 A given odorant, whether it is a pure analyte or a complex mixture, evokes a combinatorial response across a large population of sensory neurons.14 The high-dimensional inputs from the sensory neurons are subsequently transformed by neural circuits such that initially coarse odor representation is increasingly refined over time to become more odor-specific.15 The segmentation of odor class and identity information is done in a hierarchical fashion.</p><p>Here, we adapt this elegant divide-and-conquer biological approach to solve a parallel engineering problem in artificial olfaction and demonstrate its viability using a microelectromechanical system (MEMS)-based microhotplate array with metal oxide chemiresistors. We note, however, that the approach is applicable to any sensor array that yields analytically rich data. The response of each microelement at a select temperature is treated as a separate "pseudo"16 or "virtual" sensor17 to generate a high-dimensional chemical representation (5600 sensor–temperature pairs) that qualitatively mimics the combinatorial nature of the input from sensory neurons. By selecting (using the training data) a subset of sensor responses from these analytically rich data sets to initially perform discrimination of broad chemical classes, and progressively refining the selected data subset to allow finer discrimination between members within a class, we demonstrate robust recognition and generalization capabilities during validation testing. We also show that the proposed bioinspired approach is less sensitive to sensor aging compared to traditional approaches, thereby allowing sensing devices to remain functional for extended periods of operation.</p><!><p>We used a 16-element MEMS microhotplate chemical sensor array18,19 (see Figure 1) for our experiments. Each microhotplate includes an independently controlled, polysilicon electrical heating line and a thin-film, metal oxide semiconductor sensor material. SnO2 and TiO2 sensor materials (some pure, some layered, and some with ruthenium dopant) were placed on individual elements by a thermal, self-lithographic chemical vapor deposition technique.18 In addition, WO3 sensors were produced by micropipetting and then thermally decomposing a peroxytungstate solution on two of the microhotplates. The resulting films populate a 16-element array as shown in Supporting Information, Figure S-1.</p><p>To generate the required analytical information, we also enriched the outputs of each microelement along a key chemical dimension, i.e., temperature, to probe thermodynamics and kinetics of interactions between the sensor and its environment. During the training and testing stages, sensors were operated using a ramp from 50 to 500 °C, with a ramp rate of ≈5 °C/s. Conductance measurements were made for each sensor in 1 °C increments. Immediately before the beginning of each ramp, a sensor was held at 500 °C for 2 s to provide a "burnoff" period. The measurements between 50 and 150 °C were often noisy and thus ignored. A total of 16 sensors and 350 temperatures per ramp resulted in an extremely complex and feature-rich data set.</p><p>The background of all measurements was flowing, zero-grade dry air. Added to this background were the analytes, added one at a time and in random order. Each analyte except water was diluted to 3 μmol/mol. Water was presented at 100 μmol/mol. Methanol and propane were additionally presented during training at concentrations ranging from 30 nmol/mol to 10 μmol/mol. The total flow rate across the sensor was a constant 1 standard L/min.</p><p>Analytes used during the training were as follows: three simple oxides (water, carbon monoxide, carbon dioxide); two alcohols (methanol, ethanol); two ketones (acetone, methyl ethyl ketone); two alkanes (ethane, propane) and two aromatics (benzene, toluene). The molecular structures of these analytes are given in Supporting Information, Figure S-2. Between each analyte, the sensor was returned to dry air. The output of the sensor was relatively constant after a transition of two to three measurement cycles following the introduction or removal of an analyte. These initial, transitional measurements were removed from the training set.</p><p>Between the training and testing stages, the sensor was subjected to an accelerated aging protocol. For 24 h, the sensors were intermittently held at 500 °C, rapidly thermally cycled, and exposed to methanol and propane. After this, the embedded heaters were recalibrated. These processes were intended to cause drift in the sensors through various means.</p><p>During the test phase, the sensor was operated as before, exposing it to some of the analytes seen in the training set: methanol, acetone, ethane, and benzene. In addition, four new analytes not included in the training set were presented to the sensor array: two alcohols (1-propanol, 2-propanol) and two ketones (methyl isobutyl ketone, cyclohexanone). Note that the new analytes are more complex and, other than the defining functional group presence, structurally quite different from the model analytes used in the training set (see Supporting Information, Figure S-2). These aspects were specifically selected as a challenge to our classification approach. Replicating the training-phase protocol, the sensor was returned to dry air between different analytes; however, the task of detecting a chemical event from a nonevent (dry air) was also assigned to the recognition module to simulate a real-time test.</p><p>All data analyses were done using custom routines written in a commercial, matrix-based mathematics software package. A differential approach was used to mitigate drift in an online, unsupervised manner. In this approach, the relative change in conductance, G, with respect to the conductance at 150 °C, G0, was considered as the drift-corrected sensor response, G′</p><p>Since the baseline response was determined separately for each measurement and for each sensing material, this simple approach accounted for the variable amount of drift for different materials and different analytes, as seen in Supporting Information, Figure S-3.</p><!><p>We first analyzed the training data using two all-at-once approaches that have been extensively used for analyte recognition in previous sensing studies: principal components analysis (PCA) and hierarchical clustering analysis (HCA).8,9,20–22 We used PCA to visualize the high-dimensional sensor array data (16 sensors × 350 temperatures).23 The multidimensional sensor responses were projected onto the first two or three dimensions, defined by the first few eigenvectors (sorted starting with the largest eigenvalue) of the response covariance matrix that captures most of the variance in the data set. We normalized each sensor response to its maximum conductance value in order to obtain PCA projections that favor separation of the analytes independent of their concentrations. Figure 2a shows the projection of the 5600-dimensional sensor array response to different training analytes along the first three principal components. Each point represents the sensor response to a particular analyte. As can be clearly seen, sensor responses to a given analyte are well clustered. However, analytes that have overlapping chemical features (e.g., the same functional group) do not necessarily group together to form superclusters. Further, the dry air responses do not form a single, well-defined cluster, indicating drift in sensor response over time even within the training period. This variability of sensor response to dry air samples makes reliable detection of some analytes like CO, benzene, and ethane extremely difficult.</p><p>To quantitatively evaluate the relationship between different training analytes, we performed an HCA using the mean response of the sensor array [G′S1 T1,…, G′S1 T350,…, G′S16 T1,…, G′S16 T350] to each analyte, where the subscript refers to the sensor and temperature indices. A Euclidean distance measure was used to assess similarity of samples. Average pairwise distances between all samples in two different clusters were used to evaluate their similarity. This analysis is shown in Figure 2b. A more detailed analysis employing every measurement and not just the mean measurement in each analyte is presented in Supporting Information, Figure S-4; this includes response scatter and thus may be more indicative of separation or overlap between different analytes. Each introduction of dry air between two analytes was averaged separately to study variability over time. Similar to the PCA results, analytes that have common chemical features were not necessarily similar to each other. For example, the mean ethanol response was more similar to that of the two ketones (methyl ethyl ketone, acetone) than to the methanol response. Benzene and toluene were grouped with ethane and propane, respectively. At higher concentrations, the methanol response was similar to that of acetone, whereas at lower concentrations, it resembled the propane response. Taken together, these results from the PCA and the HCA analyses illustrate that the training data were not ordered in any fashion based upon their constituent chemical features. More importantly, these results also suggest that predicting the chemical composition of a novel analyte based on these training measurements would be highly improbable using the traditional recognition approaches.</p><p>Another problem that critically affects the performance of the all-at-once approaches is the sensor response drift. Figure 2c shows the sensor array response during the training and testing phases projected along the principal directions of variance of the training data. Despite using a differential sensor response readout to compensate for a linear drift component, the effect of sensor aging is catastrophic. The test samples are far offset from the training samples, and recognition during the test phase of even those analytes repeated from the training set is highly unlikely using the PCA approach.</p><!><p>Unlike the conventional approaches, we exploit the inherent, hierarchical structure of the chemical space by breaking the sensing problem into a series of simpler subtasks. For this application, each of the subtasks is based on the chemical structure and functionalization of the analytes, since these play a critical role in sensor–analyte interactions. The hierarchy thus constructed is shown in Figure 3a. The scheme echoes the inspiration of biological systems, as the initial level defines broad chemical classes (which allows generalization) before progressively moving to more specific classifications (which is required for identification). A requirement for such an approach is that the sensors must show repeatable behaviors that correlate not just to a specific analyte but also to specific compositional features that are common across whole classes of analytes. Figure 3b shows an example of such behavior found with the metal oxide sensors used in this study: a region of raw data with a similar conductance versus temperature profile for all of the alkanes and aromatics, a different profile for all of the ketones, and a third profile for all of the alcohols.</p><p>To resolve chemical identity, we begin with a simple event detection task that differentiates a nonevent (dry air) from two types of chemical events: presence of an organic or presence of a simple oxide. To perform this event detection task, we choose a subset of sensor features that provides maximum discrimination between samples of these three different categories and at the same time shows low variance across members within the same group. The feature selection is based on their modified t-statistic:23</p><p> (2)t=∣μi−μj∣σi+σj where μi and σi are the mean and standard deviation of all of the measurements from a particular sensor at a particular temperature within analytes of chemical class i, and μj and σj are the corresponding values within analytes of chemical class j. For each categorization problem within the divide-and-conquer approach, only those data features with maximal t-statistic were selected (for categorizations including more than two chemical classes, the average pairwise separability was used).</p><p>Using such a discrimination method, we found that species like benzene, which were difficult to detect using the all-at-once approaches (see Figure 2a), can be easily identified. In addition, each category has a large and diverse membership, which, as we will show later, allows generalization beyond the analytes of the training set. For building the subsequent levels of the hierarchy, the subcategories are further divided based upon more specific chemical features. Each time, only the training data collected within the analyte(s) belonging to that specific subcategory are used for categorization, excluding from consideration all of the data collected for other analytes. Within this reduced set of training data, the sensor features with maximal t-statistic for making the desired distinction are determined.</p><p>As an example, the 56 drift-corrected (G′) data points with maximal t-statistic, just 1% of the total collected, for grouping the organics into those that either do or do not contain oxygen are shown in Figure 4a; the best data points for grouping those that do contain oxygen into either alcohol or ketone are shown in Figure 4b; and, finally, the best data points for grouping the alcohols into either methanol or ethanol are shown in Figure 4c. Note that, for each discrimination level, the data from limited subsets of operating temperatures within a few select sensors are used. Supporting Information Figure S-5 shows the drift-corrected conductance data collected during the training phase for each sensor at each temperature and indicates, for each question in the hierarchy, whether that datum is used to answer that question.</p><p>The number of features selected for each question is an important parameter that affects the success of the technique. It is important to note that the distribution of t-statistics for sensor features critically depends on the complexity of the classification problem. Supporting Information Figure S-6 shows the distribution of t-statistics for each question in the hierarchy. For more general problems at higher levels of the hierarchy, the groupings of chemical species are rather loose. For example, the members of simple oxides group, CO, CO2, and H2O, do not have particularly similar chemistry. Hence, the probability of finding features that allow these chemicals to cluster together and at the same time distinguish them from the other organics group is low. On the other hand, for more specific discrimination tasks at the bottom levels of hierarchy (e.g., methanol vs ethanol), the class memberships are very well defined and therefore these individual analytes can be easily discriminated with very few optimal features. Thus, as detailed in Supporting Information, Table S-1, we have used in our approach more features for the initial questions and fewer features for later questions.</p><p>As a simplified strategy, we have also analyzed our data using a constant number of features across all levels of the hierarchy. By systematically varying the number of features selected, we find a peak performance level when using ≈10% of the total features for each step in the hierarchy. When using fewer than 10% of the features, we observed that there were many misclassifications for the very first task (simple oxide vs dry air vs organics). This was expected as the chemical groups at this level are not well defined compared to the lower levels and more features are required for addressing this problem. When using more than 10% of the features, the classification performance again dropped as less relevant features were included (especially at the lower levels). This is detailed in Supporting Information Figure S-7.</p><p>The multistep approach uses only those features that provide optimal separation between subsets of analytes. By limiting the analysis to sensor responses that are highly repeatable within an analyte group and that provide maximum discriminability between different analyte groups, robustness against drift is implicitly incorporated into the approach. Figure 5a shows the sensor features selected from the training data used for categorization of the alcohols versus ketones, as measured during the training and testing phases. Figure 5b shows the corresponding sensor features for categorization of ethanol versus methanol. Figure 5a shows the generalizability of this approach, and Figure 5b further demonstrates that reliable recognition of individual analytes is also possible. The stability in the sensor features selected from the training data using the modified t-statistic demonstrates how our feature selection approach mitigates drift, even if the overall sensor response profile has changed significantly.</p><p>The drift-corrected data from the test phase, including the new analytes and those repeated from the training set, were categorized using the hierarchical scheme built from the training data. The data were classified using a k-nearest neighbor classifier23 based on a Euclidean distance measure on the selected features (k = 3; k values of 1–9 exhibited nearly identical results). Figure 6a schematically shows the results, with dot color indicating the chemical group. Figure 6b shows the success or failure in classification of each measurement in the test phase. Other than ethane, the only misclassifications were during apparent lag times in recognition during the first one or two measurements after introduction or removal of some analytes. Unlike the training data, where introduction and removal of analytes were explicitly known, nothing was assumed of the test data. Thus, though in the training data the transitional measurements were manually removed, this was not done to the test data and is the likely cause of these transitional misclassifications. Ethane could not be distinguished from dry air during the test phase. It is likely that the sensors were not sufficiently sensitive to ethane, a relatively nonreactive molecule, to overcome the drift induced during the aging process. Despite these difficulties, the overall success rate in categorization of the test phase, including the new analytes and those repeated from the training phase, was 87%. The new alcohols and ketones were successfully categorized nearly every time.</p><p>How might unknown analytes be handled in the scheme? In our evaluations, we stopped at the levels of determining the functional group of the test analytes. In Figure 7a, we add to Figure 5b the testing data collected when presenting the new alcohols. While the "training" and "testing" measurements of methanol are extremely close to each other in nearest-neighbor space, the two newer alcohols (1-propanol, 2-propanol) lie between the two known alcohols (methanol, ethanol) and, with some misclassifications, can be recognized as unknowns (i.e., the propanols would be classified as unknown alcohols, and not misidentified as methanol or ethanol) by setting a threshold on the distance measure.</p><p>Similarly, how might different analyte concentrations be handled in the scheme? Concentration information may be obtained by adding to the hierarchy a final, lowest level for this purpose. Figure 7b shows sensor features selected from the training data used for categorization of the 30 nmol/mol methanol versus 3 μmol/mol methanol versus 10 μmol/mol methanol, as measured during the training (solid lines) and testing phases (dashed lines). The concentration during the testing phase was 3 μmol/mol methanol, and as can be seen, the test concentration is correctly identified. No normalization was necessary as both concentration-invariant and concentration-dependent features are available in the temperature spectra. The concentration-invariant features were used at higher levels and concentration-dependent features were used to quantify concentrations.</p><p>The ability to recognize pretrained or behaviorally important odors and to generalize to new odors based upon chemical similarity with known odorants are important challenges faced by both biological and artificial chemical sensing systems. Even though the analytical requirements of these tasks are at odds with each other, the biological system demonstrates an amazing ability to deal with these issues. What signal-processing strategy followed by the olfactory system allows it to address these analytical problems reliably? Physiological15 and behavioral24 studies reveal that biological olfaction uses time as an additional coding dimension to decouple these tasks: initial coarse, generalizable representation gets refined over time by adding additional features that allow finer discrimination. This divide-and-conquer approach contrasts the conventional monolithic approaches followed in artificial olfaction where no distinction is made between the generalization and the recognition problems. While appropriate for highly orthogonal data, where analytes are ordered based upon their constituent chemical features,9 the performance of such approaches is significantly affected when dealing with sensor responses with low signal-to-noise ratios and high variability due to sensor aging (see Figure 2c). More importantly, these approaches lack the ability to examine the presence of a desired organization of the chemical space and therefore lack the flexibility to be tuned for a specific application.</p><p>We found that much of the success of the divide-and-conquer approach to mitigate errors due to drift was due to the following: (a) subdivision of the problem into a series of recognition tasks, each involving only a subset of analytes, and (b) independent selection of high signal-to-noise response features for each task. We observed that using all of the sensor response features for every question leads to a reduced overall successful classification rate of 31% for the test data (the great majority of the successes were dry air). Supporting Information, Figure S-8 shows the test data classification using the hierarchy without feature selection.</p><p>How flexible is the choice of groupings within the hierarchy? To examine whether discrimination between any two groups can be a subtask in the hierarchy, we designated two groups: A (ethanol, acetone, propane, toluene) and B (methanol, methyl ethyl ketone, ethane, benzene). These groupings were made such that there are no chemical features that distinguish one group from the other. Figure 8 shows that even the best features selected from the training data for these groupings show considerable spread within each group and no separation between the groups. This can be compared to Figure 4a, which is a more chemically logical categorization for the same set of analytes. This test implies that the data features selected during the more sensible categorizations relate to repeatable chemical interactions between the sensors and specific chemical features of the analytes. Furthermore, the ability to find these repeatable interactions for multiple levels of the hierarchy shows that chemical information at multiple levels of abstraction is available from the temperature spectra of metal oxide chemiresistors. Though it does not seem possible to use this technique to classify analytes according to arbitrary groupings, the use of alternate or multiple hierarchies is not precluded. For example, a carbon chain length-based scheme may be useful in classifying the analytes studied here.</p><!><p>In this work, we have shown that chemical sensors can be used within a biologically inspired, hierarchical categorization scheme to provide a reliable, highly drift-resistant mechanism for useful classification of both previously trained chemicals and those never seen before. This ability is possible because of the discovery that temperature-controlled, semiconductor chemical sensors are responsive to multiple levels of chemical information. The methodology developed is robust against drift in the signal, since it relies upon finding where analyte–sensor chemical interactions occur with high consistency. The method does not rely upon complex and delicate mathematical formulations that can overfit the training data. Key aspects to the approach are an analytically rich database, a sensible hierarchy scheme, and a selection method to determine which data are most useful for a particular categorization. Temperature-modulated sensor arrays built on MEMS microhotplates were found to be an excellent platform, although not uniquely so, with which to accomplish this task. The methods described may enable the use of chemical sensor arrays as miniature, inexpensive analytical tools.</p><!><p>SUPPORTING INFORMATION AVAILABLE</p><p>Additional information as noted in text. This material is available free of charge via the Internet at http://pubs.acs.org.</p><p>Optical micrographs of the microhotplate sensors. A 16-element array is shown in the upper portion, and a magnified view of one of the elements is shown in the lower portion. Bright features on the microhotplates are the platinum interdigitated electrodes that electrically contact the sensing films. Coloration of the microhotplates comes from optical interference caused by the sensing films. Dark regions indicate the micromachined etch pit that provides thermal isolation.</p><p>Conventional, all-at-once approaches to classification of the training data. (a) PCA, projecting the training data along the three principal directions that accounted for the maximum variance. Each sphere represents one measurement during the training phase. The color uniquely identifies the analyte to which the sensor array was exposed during the measurement: dark green, toluene; light green, benzene; dark yellow, propane; light yellow, ethane; dark blue, methyl ethyl ketone; light blue, acetone; dark red, methanol; light red, ethanol light gray, dry air, cyan, water; dark gray, CO; black, CO2. (b) HCA of the same data, using the mean value of the response in each analyte. (c) PCA scatter plot showing both the training and testing data along the three principal directions of variance in the training data.</p><p>Divide-and-conquer approach to classification of the training data. (a) The hierarchy used to categorize the analytes present in the training data. (b) A region of raw data from sensor 1 with a similar conductance vs temperature profile for all four alkanes and aromatics, a separate profile for both ketones, and a third profile for both alcohols. Line color represents the analyte present during measurement, using the same color code as in Figure 2a. Though difficult to distinguish, between 7 and 10 overlapping lines of each color are plotted, representing repeated measurements. The presented data represent less than 2% of the overall microarray data set.</p><p>Drift-corrected raw data used for various categorization tasks. For each task, the data represent the 1% of data points with maximal t-statistic. Line color represents the analyte present during measurement, using the same color code as in Figure 2a. (a) Categorization of organics into those that do or do not contain oxygen. For simplicity, only data collected at 3 μmol/mol analyte concentrations are shown. (b) Categorization of oxygen-containing organics into alcohols or ketones. (c) Categorization of alcohols into methanol or ethanol.</p><p>(a) Response of sensor–temperature features that were selected from the training data to discriminate alcohols from ketones. Both training and test measurements are shown. (b) Similar plot showing response of sensor–temperature features selected to discriminate ethanol from methanol (ethanol test data are not present because ethanol was not a member of the test set).</p><p>Hierarchical categorization of the test data. (a) Graphical view of the traversal of each measurement through the hierarchy. The chemical family of the analyte present during measurement is color-coded: gray, dry air; cyan, simple oxide; red, alcohol; blue, ketone; yellow, alkane; and green, aromatic (note that no simple oxides were used in the test phase). (b) Chart of the accuracy of placement of each measurement into its proper category: green box, correct placement; red, incorrect placement. Analytes not included during the training phase are indicated with an asterisk. The order of analyte exposure during the test phase is as shown, progressing from left to right and then top to bottom.</p><p>(a) Response of sensor–temperature features that were selected from the training data to discriminate ethanol from methanol. Both training and test measurements (including all three alcohols measured in the test phase: methanol, 1-propanol, and 2-propanol) are shown. (b) Response of sensor–temperature features that were selected from the training data to discriminate 30 nmol/mol methanol vs 3 μmol/mol methanol vs 10 μmol/mol methanol. Both training and test measurements (taken at 3 μmol/mol) are shown.</p><p>The 1% of drift corrected raw data points with maximal t-statistic for categorization of the organics into group A (ethanol, acetone, propane, toluene) or B (methanol, methyl ethyl ketone, ethane, benzene). Line color represents the analyte present during measurement, using the same color code as in Figure 2a.</p>

PubMed Author Manuscript

Thiol-reactive amphiphilic block copolymer for coating gold nanoparticles with neutral and functionable surfaces

Nanoparticles designed for biomedical applications are often coated with polymers containing reactive functional groups, such as \xe2\x80\x93COOH and \xe2\x80\x93NH2, to conjugate targeting ligands or drugs. However, introducing highly charged surfaces promotes binding of the nanoparticles to biomolecules in biological systems through ionic interactions, causing the nanoparticles to aggregate in biological environments and consequently undergo strong non-specific binding to off-target cells and tissues. Developing a unique polymer with neutral surfaces that can be further functionalized directly would be critical to develop suitable nanomaterials for nanomedicine. Here, we report a thiol-reactive amphiphilic block copolymer poly(ethylene oxide)-block-poly(pyridyldisulfide ethylmeth acrylate) (PEO-b-PPDSM) for coating gold nanoparticles (AuNPs). The resultant polymer-coated AuNPs have almost neutral surfaces with slightly negative zeta potentials from -10 to 0 mV over a wide pH range from 2 to 12. Although the zeta potential is close to zero we show that the PEO-b-PPDSM copolymer-coated AuNPs have both good stability in various physiological conditions and reduced non-specific adsorption of proteins/biomolecules. Because of the multiple pyridyldisulfide groups on the PPDSM block, these individually dispersed nanocomplexes with an overall hydrodynamic size around 43.8 nm can be directly functionalized via disulfide-thiol exchange chemistry.

thiol-reactive_amphiphilic_block_copolymer_for_coating_gold_nanoparticles_with_neutral_and_functiona

Introduction<!>Materials and methods<!>Synthesis of block copolymer PEO-b-PPDSM<!>Encapsulation of AuNPs using PEO-b-PPDSM<!>Stability of AuNPs in different biological media<!>Thiol-modified FITC<!>Functionalization of AuNPs coated with PEO-b-PPDSM<!>Results and discussion<!>Conclusions

<p>Gold nanoparticles (AuNPs) have been used in various applications as imaging contrast agents,1–3 therapeutic agents,4,5 biological sensors,6 and cell-targeting vectors.7 For both in vitro and in vivo applications, AuNPs are usually coated with a polymeric layer to protect them from aggregation in physiological conditions or to further conjugate targeting ligands to generate targeted nanoparticles.8–14 Traditionally, these nanoparticles are coated with polymers containing reactive functional groups, such as –COOH and –NH2, which facilitate the conjugation of targeting ligands.2,15–17 However, in biological environments nanoparticles with highly charged surfaces will bind to biomolecules through ionic interactions, causing them to aggregate18 and thus experience non-specific uptake by healthy tissues or cells of the immune system.19,20 To reduce non-specific binding in biological environments, nanoparticles with a neutral coating are favorable. A common approach is to conjugate multiple poly(ethylene oxide) (PEO) molecules without polar groups onto the nanoparticle surface.21,22 However, most of them are not functional for further ligand conjugation. In order to functionalize the nanoparticles carboxyl- or amine-modified PEO has to be used, which simultaneously increases the surface charge of PEO stabilized nanoparticles.17 While PEGylation reduces aggregation of AuNPs, if the zeta potential is close to zero they become unstable during the subsequent conjugation process for functionalization of the AuNP surfaces and during in vivo applications.23 This is still one of the major challenges for successful applications of AuNPs. Developing a unique polymer to coat AuNPs with neutral surfaces that can be further functionalized directly would be critical to overcome these obstacles and design AuNPs suitable for biomedical applications.</p><p>Recently, polymers having functional pendant groups like pyridyldisulfide (PDS) exhibiting selective reactivity toward thiols have attracted great interest for applications in nanomedicine and nanobiotechnology.24–31 The versatility of the PDS group in preparation of biofunctional constructs of polymers for drug delivery applications has been proven by a number of studies.24,29 Using a random copolymer that contains oligoethyleneglycol (OEG) and PDS units, Thayumannavan's group studied these self-cross-linked polymer nanogels as a versatile nanoscopic drug delivery platform.24 By altering the degree of cross-linking, the same group also studied the leakage characteristics of nanocontainers using a FRET-based method and confirmed that the dynamics of encapsulated guest interchange (i.e., pre-mature release) can be controlled and that release can be externally triggered.25 Bulmus' group also reported the synthesis of versatile thiol-reactive polymer scaffolds having PDS pendant groups via RAFT polymerization.26,27 The formation of reversible disulfide linkages upon the thiol-PDS reaction makes PDS a very attractive functional group for preparation of reversible bioconjugates and release of protein/siRNA and drugs.28,32</p><p>Here we report an amphiphilic block copolymer poly(ethylene oxide)-block-poly (pyridyldisulfide ethylmethacrylate) (PEO-b-PPDSM) having the PDS pendant group to coat AuNPs. To the best of our knowledge, it is the first report in which this polymer is applied to coat AuNPs to confer properties desirable for biomedical applications. Our polymer-coated gold nanoparticles are individually dispersed with uniform particle size and are highly stable under physiological condition. Most importantly, they have neutral but functionable surfaces suited for developing targeted nanoparticles for in vitro and in vivo applications with enhanced potential to minimize non-specific binding and uptake by healthy cells and tissues.</p><!><p>The initiator, 2,2-Azobis(isobutyronitrile) (AIBN, 98%, Aldrich) was purified by recrystallization twice from ethanol. Fluorescein isothiocyanate (FITC) and tris(2-carboxyehtyl)phosphine hydrochloride (TCEP) were obtained from Pierce Biotechnology. Aldrithiol-2, mercaptoethanol, glacial acetic acid, methacryloyl chloride, sodium methoxide (25% solution in methanol), anhydrous methanol, elememtal sulfur, benzyl chloride, potassium ferricyanide (III), 4,4′-azobis(4-cyanopentanoic acid), cystamine dihydro-chloride, 4-dimethylaminopridine (DMAP), 1,3-dicyclo-hexylcarbodiimide (DCC), silica gel (60 Å, 230-400 mesh) and the solvents used for monomer synthesis and polymerization were purchased from Sigma-Aldrich and used directly as received. Poly(ethylene oxide) methyl ester (PEO) (Mn=5000 g/mol, Mw/Mn=1.10) were purchased from Polysciences Inc. Traditional PEGylated gold nanoparticles (methoxy-PEG5000-SH) and non-PEGylated gold nanoparticles (CG-15-100) were supplied by Cytodiagnostics (Ontario, Canada). Mouse plasma (Catalog #: IMS-C57BL6-N) was ordered from Innovative Research (MI, USA). 1H and 13C NMR were taken in Varian 400 MHz NMR spectrometer, UV visible spectra were recorded in a BioTek micro plate reader (Synergy 2) for aqueous solutions and UV-3600 (Shimadzu) for organic solutions. Molecular weight and molecular weight distribution of the copolymer was estimated by gel permeation chromatography (GPC) with THF as the eluent (flow rate = 1.0 mL/min) using PS standard and UV detector. A series of three linear Styragel columns: HR0.5, HR1, and HR4 and a column temperature of 40 °C were used. The nanoparticles hydrodynamic size and zeta potential were measured using a dynamic light scattering (DLS) instrument (Malvern Zeta Sizer Nano S-90) equipped with a 22 mW He-Ne laser operating at λ = 632.8 nm. The AuNPs were made by femtosecond laser ablation and were viewed by transmission electron microscopy (TEM) (Philips CM-100 60 kV). The polymer coating was viewed through negative staining with OsO4.</p><!><p>Synthesis of monomer pyridyldisulfide ethymethacrylate (PDSM) and PEO macro-RAFT agent was provided in supporting information (Figs. S1-S5). The RAFT polymerization was performed in a schlenk flask with a magnetic stirring bar. The polymerization procedure is as follows. PDSM (1.03 g, 4 mmol), PEO-CTA (0.80 g, 0.16 mmol), and AIBN (6.3 mg, 0.04 mmol) were dissolved in DMAc (10 mL). The homogenized reaction mixture was subjected to four freeze-pump-thaw cycles to remove oxygen. The flask was then immersed into an oil bath preheated to 70 °C to start the polymerization. After 12 h, the reaction flask was quenched into the mixture of dry ice/2-propanol to stop the polymerization. After thawing, the solution was precipitated three times in diethyl ether and then dried in vacuo.</p><!><p>AuNPs were generated by laser ablation using a femtosecond laser system delivering 700 fs laser pulses width centered at a wavelength of 1.045 μm (maximum energy, 10 μJ per pulse; beam diameter, 50 μm) on a gold metal plate, which was placed on the bottom of a glass vessel filled with 20 mL of acetone. After a couple of days aging, the top clear red solution was transferred and mixed with 2 mL of dimethylformamide (DMF). Acetone was evaporated under reduced pressure to form a concentrate gold solution in DMF. One mL of gold solution (20 μM in DMF) was mixed with 1 mL of PEO-b-PPDSM solution (50 mg/mL in DMF) in a 15 mL flask equipped with a magnetic stirring bar with gentle stirring at room temperature overnight. Then the temperature was increased to corresponding temperatures in an oil bath for pre-set time points. After cooling to room temperature slowly, the resultant mixture was added dropwise to 20 mL of deionized water under magnetic stirring. The block copolymer encapsulated AuNPs were isolated through three times centrifugation using an Eppendorf 5424 centrifuge at 15,000 rpm for 30 minutes. Supernatant was removed by careful pipetting, and the AuNP was resuspended in deionized water.</p><!><p>To explore the stability in biological media, both AuNPs coated with PEO-b-PPDSM and commercial PEGylated gold nanoparticles were incubated with Dulbecco's Modified Eagle Medium (DMEM). UV-Vis spectra of gold nanoparticle solutions in different media (phenol-red free) at 0, 1, 4, and 24 h were acquired using Plate Reader. The stability of AuNPs in 10% mouse plasma was evaluated by monitoring their hydrodynamic size by DLS.</p><!><p>A mixture of FITC (20 mg, 0.052 mmol), cystamine dihydrochloride (6.0 mg, 0.026 mmol) and triethylamine (26.0 mg, 0.26 mmol) was dissolved in DMSO (800 μl) and stirred for 4 h. To this reaction mixture was added tris(2-carboxyehtyl)phosphine hydrochloride (17.6 mg, 0.062 mmol) and stirred for 1 h. The resultant mixture was precipitated in ethyl ether and washed with water. The crude product was used for polymer coated AuNPs surface modification without further purification.</p><!><p>One mg of FITC or thiol-modified FITC was dissolved in 100 μL of DMF and then 1 mL of polymer coated AuNPs (4.8 nM) in water was added. 0.1 M NaOH was used to adjust the pH until the solution is clear. The mixture solution was stirred overnight at room temperature. Non-conjugated dye molecules were removed by ultrafiltration and re-suspended using 1.0 mM sodium carbonate until there is no detectable dye in the filtrated solution (five times) using a nanosep® filter (Pall Corp.) with a molecular weight cutoff of 30,000 g mol-1. The concentration of an AuNP solution without FITC modification was adjusted to match the same optical density at 535 nm as FITC modified one to show the FITC signal after subtraction. A calibration curve of AuNPs and FITC in 1.0 mM sodium carbonate was created to estimate the number of FITC conjugated on each AuNP.</p><!><p>Block copolymer PEO-b-PPDSM that contains PDS functional groups is synthesized by reversible addition fragmentation chain transfer (RAFT) polymerization using PEO (Mn 5000 g/mol) macro-RAFT agent (Supporting Information, Schemes S1-S3, Figs. S1-S5).26 The block copolymer structure is confirmed by the 1H NMR spectrum as shown in Fig. 1a. The spectrum shows the characteristic peaks from both the PEO block (peak a) and PDSM block (peaks b, c, d, e, and f). The proton number of each peak on the spectrum for the PDSM block matches well with the expected structure, revealing the absence of any significant transfer reaction to the PDS containing side groups.27 It is estimated that the block copolymer contains ∼20 PDSM units based on the integration of peak f and peak a. The block copolymer structure is also confirmed by gel permeation chromatography (GPC). The expected elution peak shifts to a higher molecular weight in the elution profile (Mn 11,600 g/mol) and exhibits a low polydispersity index (PDI, 1.16) as shown in Fig. 1b.</p><p>For the first time, we investigate the application of PEO-b-PPDSM as a coating material for inorganic nanoparticles like gold colloids, as shown in Scheme 1. We expect that coating AuNPs with this polymer will confer three desirable characteristics. First, the multiple PDS groups on the PPDSM block will interact with AuNPs through multiple Au-S binding sites so that stable and individually dispersed AuNPs in aqueous solution could be formed. Second, the resultant nanoparticles will have neutral surfaces since there are no charged groups on polymer. Third, the PDS bonds can be functionalized through thiol-disulfide exchange reactions.</p><p>For the preparation of individually dispersed nanoparticles coated with amphiphilic polymers, it is generally believed that polymers have to bind to nanoparticles originally made in organic solvent before they are transferred from a non-aqueous to an aqueous solution.15 Consequently, the bound amphiphilic polymer will collapse in situ on the surface of the nanoparticles with a hydrophobic inner layer and hydrophilic outer layer.33 Here, we find that heat treatment of the AuNPs and polymer mixture at 130 °C can enhance the Au-polymer binding.34 The enhanced binding is probably attributed to the exposure of thiol groups on polymer chains by partially reducing the PDS bonds after heat treatment. The exposure of thiol groups is revealed by the emergence of the absorption peak at ∼374 nm in the UV-absorption spectra of the mixture of polymer and AuNPs after heat treatment (Figs. S6a and S6b), indicating the release of pyridine-2-thione upon reducing the PDS bonds.27 Based on the extinction coefficient of pyridine-2-thione in DMF (ε374nm = 5440 M-1cm-1),27 it is estimated that on average 0.8% of all the PDS bonds on polymer chains were reduced after heat treatment for 2 h.</p><p>Once the mixture of polymer and AuNPs is transferred into water after heat treatment and successfully purified to remove unbound polymer using centrifugation, transmission electron microscopy (TEM) is applied to visualize the core-shell structure of polymer-coated AuNPs as shown in Fig. 2. Fig. 2a (no negative staining) shows that the AuNPs are individually dispersed; statistical analysis by ImageJ in Fig. 2b reveals an average core size of ∼12 nm. The polymer coating around each gold nanoparticle is clearly visible by the negative staining as shown in Fig. 2c. It shows the polymer shell around the AuNPs is ∼8 nm thick on average. This polymer shell is composed of a hydrophilic PEO outer layer and a collapsed hydrophobic PPDSM inner layer, which has the potential to encapsulate hydrophobic therapeutic drugs.35 To demonstrate this concept, we show that the composite nanoparticles have at least 20% loading efficiency (based on polymer mass) of the neutral anti-cancer drug doxorubicin (Fig. S7). As a potential drug carrier, the polymer layer around the AuNPs could be crosslinked to the drug; release could be triggered by glutathione (GSH), which has a higher concentration inside cells than in the bloodstream.24 Fig. 2d compares the average hydrodynamic size of polymeric micelles only and polymer-encapsulated AuNPs measured by DLS. The data reveal that the hydrodynamic size increases from ∼26 for the pure micelles to 44 nm after encapsulation of the AuNPs, which is similar to the overall size of the composite nanoparticles revealed by TEM negative staining. Monodispersed amphiphilic polymer-coated AuNPs with a smaller overall size from (5–40 nm) are favorable for in vivo applications due to a longer mean blood circulation time and better tissue penetration.36</p><p>We expect our coated AuNPs to have neutral surfaces since there are no charged groups on the copolymer. We measure the zeta potential as shown in Fig. 3a, which shows that these polymer-coated AuNPs have slightly negative zeta potentials (-10–0 mV) over a wide pH range from 2 to 12. Although the zeta potential is close to zero we show that the polymer-coated AuNPs have good stability in physiological conditions and various pH conditions, which is a prerequisite for in vivo applications. The stability of our polymer-coated AuNPs in PBS is demonstrated by monitoring the absorption spectrum over three days and detecting no obvious decrease in absorption as shown in Fig. 3b. When compared to the stability of typical PEGylated AuNPs,37 AuNPs coated with PEO-b-PPDSM show better stability in physiological conditions; this suggests protection from aggregation in vivo. Fig. 3c shows the stability of both AuNPs coated with PEO-b-PPDSM and commercial AuNPs, either with our without a PEG coating, in cell culture media. The data clearly show the great stability of AuNPs coated with PEO-b-PPDSM in media (DMEM) over 24 h. As expected, only the surfactant stabilized non-PEGylated AuNP solution becomes purple immediately in DMEM, revealing the aggregates of AuNPs. For commercial PEGylated AuNPs, more than 50% of their plasma signal is lost over 24 h in DMEM. The improved stability of our copolymer-coated AuNPs in varying physiological conditions is attributed to the formation of an insulated hydrophobic layer around the gold surface and a hydrophilic PEG layer, which is different from, for example, a thiol-PEG ligand based coating.</p><p>This property of formulating nanoparticles containing neutral surfaces has potential advantages to enhance stability and reduce non-specific binding to tissues or other biological components in both in vitro and in vivo applications.38 Although AuNPs are generally believed to be biocompatible, severe sickness had been reported from administration of citrate-capped AuNPs.39 After simple PEGylation, AuNPs did not induce any acute side effects as reported previously.40 However, AuNP surfaces cannot be fully covered with simple PEGylation, and thus their non-specific binding with biomolecules remains high for potential in vitro and in vivo applications.41 In Fig. 3d we investigate the potential of the PEO-b-PPDSM polymer coating to improve the stability, and potentially the toxicity profile, of AuNPs compared to PEO alone. We incubate PEO-b-PPDSM-coated AuNPs and commercially available PEO-coated AuNPs (control) with mouse plasma for two hours. Their hydrodynamic sizes are monitored by DLS. Our data in Fig. 3d show that PEO-b-PPDSM-coated AuNPs have a very limited size increase (gained by ∼10 nm) after incubation with mouse plasma compared to commercial PEO-coated AuNPs (increased by ∼110 nm), indicating that PEO-b-PPDSM-coated AuNPs show reduced non-specific adsorption of proteins/biomolecules. We expect that the antifouling property of reducing opsnonization through an effective polymer coating will hinder uptake of administered nanoparticles by immune cells in the reticuloendothelial system (RES) and thus enhance the effective accumulation of the PEO-PPDSM AuNPs at target sites, consistent with our previous studies.42,43</p><p>We further confirm the stability of our polymer-coated AuNPs by recovering more than 90% soluble AuNPs after as least four centrifugation processes (Fig. S8). This stability after repeated centrifugation will provide a significant advantage for further modification compared to AuNPs with other coatings. For instance, citrate-stabilized AuNPs cannot tolerate two centrifugation-dispersion processes, as revealed by significant loss of absorption from AuNPs due to aggregation. It is worth noting that the layer of amphiphilic block copolymer PEO-b-PPDSM around each gold nanoparticle contains multiple disulfide bonds and very likely multiple Au-S interactions, which provide potential stability against possible dilution (e.g., in the blood stream).</p><p>One of the most important advantages of our polymer-coated AuNPs is that the resultant nanoparticles have neutral surfaces and can undergo conjugation without additional modification. Our hypothesis is that the surface functionalization is achieved through thiol-disulfide exchange reactions with the PDS groups.30 The existence of free PDS groups on polymer coating layer is first proven by using GSH. UV-vis absorption is used to monitor the release of pyridine-2-thione, the by-product of the thiol-disulfide exchange reaction, which has a maximum absorption at 343 nm in aqueous solution as shown in Fig. 4. The increase in GSH-mediated release of pyridine-2-thione from the polymer-coated AuNPs over time reveals that PDS groups are available for further functionalization using thiol-chemistry.</p><p>To further demonstrate the ability to directly conjugate to the polymer-coated AuNPs, we treat them with thiol-modified FITC. Fig. 5 compares the absorption peak of polymer-coated AuNPs before and after treatment with FITC followed by washing to remove free dye. The appearance of a specific absorption peak at 494 nm from FITC after treatment indicates the polymer-coated AuNPs are covalently functionalized with thiol-modified FITC by disulfide linkages. This is also confirmed by the absorption spectrum of AuNPs treated with FITC lacking a thiol modification, for which a signal from FITC is absent after washing for purification (Fig. S9). Upon subtraction of the absorption spectrum of the unmodified AuNPs from the AuNPs treated with FITC, the FITC absorption spectrum is clearly seen as a result of its conjugation to the AuNPs (Fig. S10). It is estimated that ∼1200 FITC molecules were conjugated to each polymer-coated gold nanoparticle based on the calibration curves of both FITC and AuNPs (Fig. S11).</p><!><p>In summary, we report a thiol-reactive amphiphilic block copolymer poly(ethylene oxide)-block-poly(pyridyldisulfide ethylmethacrylate) (PEO-b-PPDSM) for coating AuNPs. The resultant copolymer encapsulates individually dispersed AuNPs and creates neutral surfaces available for direct, facile functionalization of molecules for biomedical applications (e.g., targeting ligands, fluorescent dyes). Furthermore, these composite nanoparticles are very stable in varying physiological conditions and can be formulated to encapsulate hydrophobic drugs such as doxorubicin. These unique properties that our PEO-b-PPDSM polymer confers to commercially available AuNPs could enable the facile development of the next generation of stable theranostic nanoparticles with minimized non-specific binding and uptake by healthy tissues or cells of the immune system.</p>

PubMed Author Manuscript

Assessing capacity loss remediation methods for asymmetric redox flow battery chemistries using levelized cost of storage

Redox flow batteries, a promising grid-scale energy storage solution, have an open architecture that can facilitate a broad range of redox electrolytes. Vanadium is the most mature chemistry, which is largely due to its symmetry, where all active species are based on a single parent compound, that allows for inexpensive crossover remediation via rebalancing; however, the industry has increasingly sought chemistries with lower-cost and higher-abundance redox couples.Most chemistries cannot be configured symmetrically, though, necessitating research into capacity-recovery methods for asymmetric chemistries. In this work, we adapt our previously developed levelized cost of storage model, which tracks capacity fade and recovery and evaluates the costs across the battery's lifetime, to analyze two classes of asymmetric chemistries, those with active species of finite or infinite lifetimes, and their respective remediation options. For finitelifetime chemistries, we explore active-species replacement to counter decay. For infinite-lifetime chemistries, we consider two methods for addressing crossover: imposition of pseudo-symmetry via the spectator strategy and elimination of crossover via membranes with perfect selectivity. We anticipate this framework will help guide the evaluation and design of new redox chemistries, balancing the desire for low capital costs with the need to remediate capacity repeatedly and inexpensively.

assessing_capacity_loss_remediation_methods_for_asymmetric_redox_flow_battery_chemistries_using_leve

Introduction<!>Methods<!>Results and discussion<!>Remediating capacity loss for asymmetric chemistries with active species of finite lifetime: active-species replacement<!>Remediating capacity loss for asymmetric chemistries with active species of infinite lifetime: the spectator strategy<!>Eliminating crossover with ceramic membranes<!>Conclusions

<p>High levels of renewable energy penetration in the grid (>60%) are likely to be impractical without the development of complementary strategies to combat intermittency and meet demand, such as integration of energy storage [1,2]. No single technology can economically perform the vast array of grid services that, among other factors, vary in response and discharge timescales as well as total capacity requirements, necessitating a diverse portfolio of solutions [3,4]. Redox flow batteries (RFBs) store charge in liquid electrolytes that are pumped from external reservoirs through separate half-cells of an electrochemical reactor [5]. This open architecture enables independent scaling of power and energy components (i.e., the electrochemical stack and the electrolyte tanks, respectively) [6], which allows for economic benefits including the decrease of capital costs on a per unit energy basis at longer durations [7,8] and long-term cost savings through component-specific maintenance to extend the battery lifetime and remediate decay [9].</p><p>The unique architecture of RFBs enables excellent resiliency for maintaining their energy capacities. However, one must minimize the transport of active species through the semipermeable membranes separating the positive and negative electrodes, which are designed to allow for transport of supporting ions to maintain charge balance [10]. If these membranes are not perfectly selective for the desired charge-carrier species, then RFBs experience capacity reductions via undesired permeation of active species, often referred to as crossover [11]. While crossover is not the sole cause of capacity decline within RFBs, it is often the largest contributor and may halve the accessible capacity within 100-200 cycles [12]. One strategy for mitigating the effects of crossover is the use of a "symmetric" redox chemistry, where all active species are based on a single parent compound [13]. In this case, crossover does not lead to cross-contamination and associated capacity losses are recoverable via periodic electrolyte rebalancing: the transfer and mixing of partial or full volumes of electrolyte between the two reservoirs to balance the concentrations of active species. Rebalancing is a powerful capacity-remediation tool, as it allows the electrolyte to be used indefinitely, assuming other non-crossover capacity losses can be managed and/or remediated as well, which significantly reduces maintenance costs [14]. A number of symmetric chemistries have been contemplated for RFBs leveraging inorganic [15], organic [16][17][18][19], and organometallic [20][21][22] active species, but vanadium remains the canonical example.</p><p>Vanadium RFBs (VRFBs) are the most researched and commercialized RFB technology, primarily because vanadium has four stable and soluble oxidation states accessible within the electrochemical stability window of aqueous acidic electrolytes on carbon electrodes. This, in turn, allows for a symmetric chemistry (V 2+ /V 3+ in the negative half-cell and V 4+ /V 5+ in the positive half-cell) and continual recovery of crossover capacity losses via rebalancing. While the VRFB system benefits from reduced maintenance costs, it suffers from a high upfront cost, due, in part, to the price of the active species [23,24]. Vanadium prices have been relatively volatile since the 1980's, and especially so in the last four years with a late 2018 peak of over ten-fold the price at the start of 2016 [25]. The volatility has been attributed to: 1) the limited geographic locations of vanadium mines that leave the few countries that contain them (mainly China, Russia, and South Africa) with a strong control over the global vanadium supply [26], 2) new steel rebar standards that require increased vanadium content (already, 90% of current vanadium demand is steel [27]), and 3) decreases in supply due to mine closures [28]. The magnitude and uncertainty of vanadium prices is considered a key impediment to broad deployment, which has motivated research into alternative chemistries based on lower-cost and widely-available materials [29][30][31].</p><p>In recent years, the literature has seen a surge of new, potentially-inexpensive, and usually asymmetric RFB chemistries, necessitating consideration of how one might execute asymmetric electrolyte maintenance. Since asymmetric chemistries utilize different active species in the positive and negative half-cells, active species crossover results in cross-contamination. With these chemistries, capacity-loss remediation is expected to be a more challenging and chemistrydependent problem whose technical and economic consequences remain largely unarticulated in the open literature. A recent perspective by Perry et al. describes potential approaches for mitigating and remediating capacity losses due to crossover [14]. The options for crossover remediation depend on the fate of the active species upon entering the opposing half-cell; crossover can either be "destructive," where the active species are unstable in the chemical and electrochemical environment of the opposing half-cell and thus results in non-recoverable losses, or "non-destructive," where the active species remain intact in the opposing half-cell [14].</p><p>Destructive crossover remediation requires actives species replacement, but the authors note that there are no known RFB chemistries that experience destructive crossover in the published literature and thus do not explore this technique. What is more common, however, is timedependent (i.e., not crossover-dependent) active species decay in either half-cell, which would also necessitate active-species replacement [30]. In this work we divide the asymmetric chemistries and their remediation methods not by the stability of the active species upon crossover, but rather by the general stability or lifetime of the active species in their intended chemical and electrochemical environment, of which we note two classes: infinite and finite lifetime.</p><p>Finite-lifetime species experience decay and thus may require periodic replenishment or replacement. These species are generally organic compounds, which, despite uncertainty in their long-term stability, are attractive for RFBs as they are expected to be low-cost [32] and their properties can be tuned through molecular functionalization; for example, increasing molecular size lowers crossover rates due to increased steric hindrance [33]. Recent studies have shown that molecular decay rates of organics in RFB electrolytes are time-dependent and a function of the chemical, thermal, and electrochemical environment [34]. In particular, decay is generally accelerated when the molecule is in the "energized" state (for RFBs, this is the oxidized and reduced states for the positive and negative electrolytes, respectively) [30,35]. There is ongoing research into organic active species with longer lifetimes and techniques to optimize operating conditions to mitigate decay [36][37][38]. However, to the best of our knowledge, methodologies for removing and replacing decayed active species have not yet been systematically explored, likely due to the nascence of this particular class of chemistries. Note that the addition of active species without concomitant removal of the decay products is likely to be unsustainable in most cases, as it will lead to increases in solute concentration or total volume. Though largely chemistry-specific, removal would require targeted separation processes, which are likely to be complex, energyintensive, and costly for concentrated multicomponent solutions, unless active species are intentionally designed to be easily separable from their decay products (e.g., if decay products are gases or easily precipitable). Recovered decomposition products could potentially be regenerated or repurposed, either for fresh electrolyte or as feedstock for other chemical processes, though the technical and economic feasibility of such strategies will again depend on the underlying chemistry. These complications should be factored into techno-economic assessments; although inexpensive active species may reduce upfront capital costs, operating and maintenance costs may ultimately challenge the viability of such systems [34].</p><p>Infinite-lifetime species experience minimal degradation and primarily lose capacity via crossover, which allows for a range of capacity recovery strategies [14]. In general, these redox couples consist of inorganic materials that are ideally low-cost, abundant, and soluble in aqueous electrolytes, often in the form of redox-active salts with the cation (e.g., Fe, Cr, Zn) [39,40], anion (e.g., Br, I, FeCN) [41,42], or in some cases both [43], storing charge. The stability of these species usually translates to a non-destructive crossover scenario, meaning they are stable in the chemical and electrochemical environment of both their original half-cell as well as the opposing half-cell.</p><p>Thus, these chemistries can often be employed with the spectator strategy, where the electrolytes are mixed (i.e., contain both active species) to make the chemistry pseudo-symmetric and enable electrolyte rebalancing [14]. However, the spectator strategy decreases energy density and increases electrolyte cost by reducing the active species solubility and adding inactive chemicals, respectively, but if employed by suitable chemistries for stationary applications, these drawbacks may not be critical. An alternative approach to preventing capacity fade due to crossover is the use of a perfectly-selective membrane, such as a non-porous single-ion conductor (e.g., a ceramic). This strategy has received limited attention in the RFB field as experimental campaigns have been hampered by the cost, robustness, and increased resistance of available ceramics, as compared to polymeric membranes, all of which are anticipated to limit cell performance and system cost [44][45][46][47][48]. Note that both the spectator and perfect separation strategies are also viable approaches for finite-lifetime chemistries as well, but do not address active species degradation (unless the degradation primarily results from crossover), which limits their value to these systems.</p><p>Here, we use a simple levelized cost of storage (LCOS) model to evaluate the techno-economic benefits and limitations of low-cost, asymmetric chemistries with active species of finite and infinite lifetimes. Previously, we developed an LCOS model for VRFBs to assess the value of capacity recovery, and used the framework to explore practical operating considerations, such as sizing, rebalancing schedule, and electrolyte leasing [49]. While LCOS analyses consider the lifetime costs of the system for the optimal long-term solution, short-term metrics like the capital cost are also important in evaluating considerations around project investment and financing. Indeed, capital cost targets are a key metric cited when contemplating the economic viability of different energy storage solutions [50,51]. Recognizing the need for RFBs with low capital costs, we extend our LCOS model to explore the methods and associated costs for capacity-loss remediation for asymmetric chemistries using active species of finite and infinite lifetimes. For the former, we explore the logistics and costs of the active-species replacement process. For the latter, we explore the spectator strategy, using iron-chromium as a case study, as well as the use of zero-crossover membranes as capacity remediation and elimination techniques, respectively. These systems are compared to the VRFB system, the incumbent solution (i.e., an RFB with higher capital costs and the ability to recover capacity at low costs) to determine the conditions under which the reduced upfront cost of less expensive, asymmetric chemistries offsets the more complex and, in some cases, more expensive maintenance required to recover capacity losses.</p><!><p>The methodology for this work is informed by the economic and physical models developed in Rodby et al. to assess the LCOS of VRFBs [49], with key modifications to the operating and maintenance costs based on the chemistries considered. As such, repetitive details are omitted from the main text but can be found in the Supporting Information (SI). In brief, we employ the following equation for LCOS ($ MWh -1 ), defined generally as the ratio of the discounted costs to the discounted energy stored over a project lifetime: where It ($) is the investment expenditures, Lt ($) and Tt ($) are the loans and the taxes on those expenditures, respectively, OMt ($) is the operating and maintenance costs, Ct ($) is the charging costs, and Et (MWh) is the energy stored. These terms are tracked and summed across time (t), which, depending on the term, is either on a yearly (t</p><p>basis, where n and k = 365n are the number of years and days of battery operation, respectively.</p><p>The summed costs are discounted with a periodic rate, r, which is also applied yearly (ry) or daily (rd), depending on the cost period. Capacity loss is encompassed in the dynamic Et term, while the costs to remediate capacity loss are captured in the OMt and Ct terms. The full set of equations used to calculate the terms in Equation 1, as well as the inputs used for various parameters, are provided in Section S1 of the SI. This approach for modeling LCOS has been used before to assess energy-storage technologies, and we draw input values from those published reports [52,53].</p><p>The dynamic capacity model, which incorporates fade and recovery, is similar to that used in our earlier work [49], although here it is presented in a more generalized fashion that we subsequently adapt for each chemistry. We assume linear overall capacity fade, which is the sum of a constant rate of crossover losses (rCO, in units of % capacity loss per cycle), which can be remediated by rebalancing, and a constant rate of electrolyte decay losses (rED, also in units of % capacity loss per cycle), which must be remediated using alternate servicing methods. We encompass all noncrossover losses in the electrolyte decay rate, which refers to side reactions and/or active species decay (the latter only applying to finite-lifetime chemistries). We elect to ignore any nonelectrolyte losses, such as membrane fouling or electrode decay, because these components generally degrade on timescales longer than those anticipated for electrolyte crossover and decay (i.e., require replacement every five to ten years [7,54]) and their degradation is assumed to be independent of the symmetry or lifetime of the chemistry, the focus of this work. The capacity accessed in a given cycle is equal to the product of the nominal battery size and fcap, the fraction of original capacity accessible at that time. This fraction changes as the battery experiences electrolyte decay/crossover and subsequent remediation:</p><p>where 𝑛 cyc R and 𝑛 cyc S are the number of cycles passed since the last rebalancing event (i.e., t = R)</p><p>and the number of cycles passed since the last servicing event (i.e., t = S), respectively. These counters increase each cycle and reset once rebalancing or servicing occurs and capacity is regained. These terms are further defined in Equation 3. We note that servicing also resets the rebalancing counter, as we assume servicing achieves total capacity recovery.</p><p>/ R/S cyc cyc 0, 0 or upon rebalancing (R) or servicing (S) ()</p><p>( 1) 1, 0 and not rebalancing (R) or servicing (S)</p><p>To determine when to service or rebalance the system, we define a lower capacity limit (caplim); once the accessible capacity declines to the caplim, capacity remediation is performed. In the case of symmetric and pseudo-symmetric chemistries, rebalancing will occur, which regains the capacity lost to crossover but not that lost to electrolyte decay. This process repeats until the total accessible capacity upon rebalancing has decayed to the caplim (i.e., 𝑐𝑎𝑝 lim ≤ 100% − 𝑟 ED * 𝑛 cyc S (𝑡)), at which point electrolyte servicing is performed. Where rebalancing is not feasible (i.e., asymmetric chemistries), rebalancing is not employed and instead a servicing event occurs each time the accessible capacity decays to the caplim. This iterative capacity fade and recovery process is illustrated in Figure 1, which exemplifies a simulation of the capacities of a generic symmetric and asymmetric RFB as a function of time. The nominal capacity rating is maintained via a combination of the choice of caplim and oversizing of the capacity. Under a given scenario, the tradeoff between the increase in upfront cost to oversizing (i.e., higher oversizing factor, OF) and the ability to remediate capacity losses less frequently (i.e., lower caplim) is optimized to find the combination of caplim and OF that minimizes the LCOS. The optimal caplim/OF balance changes as the conditions or chemistries vary; Figure 1 exemplifies how symmetric chemistries would likely have higher optimal caplim values, as compared to asymmetric systems, because their ability to regain capacity loss with low-cost maintenance makes more frequent remediation preferable (i.e., results in a lower LCOS) to more significant capacity oversizing.</p><p>The costs of performing electrolyte maintenance are chemistry-specific and can be generally summarized as: (1) the cost of electricity to perform rebalancing to account for the energetic losses of crossover and self-discharge, which is only applicable to symmetric and pseudo-symmetric chemistries, and (2) the cost to perform servicing, which requires "oxidative maintenance" (e.g., reductant chemicals, for VRFB, or a rebalancing cell, for iron-chromium) for infinite-lifetime asymmetric chemistries or "decay maintenance" (i.e., separation and replacement or recovery of the decayed species, or total electrolyte replacement) for finite-lifetime asymmetric chemistries.</p><p>The specific methods for modeling these costs are explained in greater qualitative detail in the</p><p>Results & Discussion and in greater quantitative detail in Section S1 of the SI. Input values for relevant parameters that are used for the various chemistries contemplated here are summarized in Table 1.</p><p>Figure 1 -Example of the simulated capacity retention for generic symmetric (left, blue; e.g., VRFB) and asymmetric (right, red) RFB chemistries (for illustrative purposes).</p><p>Table 1 -Symbols, names, assumptions, values, and sources for the chemistry-specific model parameters. Values with asterisks (*) are baseline values but are subsequently varied in sensitivity analyses. For the "generic infinite-lifetime chemistry and/or Fe-Cr" column, the values given for each parameter apply to both cases except for the asterisked variables, of which the baseline values correspond to the Fe-Cr case but are varied for the generic case (see Figure 3).</p><!><p>We apply our LCOS model to the two asymmetric chemistry classesthose with active species of either finite or infinite lifetimesand compare these results to a VRFB baseline to evaluate their ability to compete with the state-of-the-art. Regarding finite-lifetime chemistries, we model different options for active-species replacement and explore the feasibility of each, while also estimating the sensitivity of LCOS to electrolyte decay rate, electrolyte cost, and reactor cost for each remediation scheme. For infinite-lifetime chemistries, we consider two methods for addressing crossover losses: 1) remediation of crossover by applying the spectator strategy to make the chemistry pseudo-symmetric and allow for rebalancing, or 2) elimination of crossover altogether via use of a perfectly selective separator. In examining the spectator strategy, we focus on the ability of a chemistry to competitively employ this scheme through calculation of the electrolyte cost, and use iron-chromium as a case study. In examining the perfectly selective separator case, we determine the bounds of separator cost, cell potential, and cell resistance needed to approach viability.</p><!><p>RFB chemistries with finite lifetimes inevitably require periodic active-species replacement. To our knowledge, the logistics of such processes have yet to be considered in the published literature but likely possess technical and economic challenges specific to the underlying chemistry. In these finite-lifetime systems, capacity fade is due to a combination of crossover and active species decay, although, for most embodiments to date, active species decay rates are generally one or more orders of magnitude greater than the rates of loss due to crossover [30]. We propose two potential options for capacity remediation for such systems, each employed periodically: (1) separation and removal of the "contaminants" (i.e., species that have decayed or crossed over), followed by either (1a) replacement with fresh active species or (1b) recovery and reuse of the active species from these contaminants (which may require methods to reverse decay), or (2) total replacement of the electrolytes. Targeted removal of contaminants eliminates the waste of replacing non-decayed electrolyte and may even allow for reuse of the recovered species, but the selectivity, energy requirements, and cost of chemistry-specific separation processes are unknown and may be prohibitive. Conversely, one could eliminate the need for any separation processes, at least on-site, by replacing the entire electrolyte upon reaching a capacity loss threshold (i.e., caplim), but this requires the exchange of large volumes and potentially sacrifices a significant quantity of valuable material. For the first option, we model only the separate/recover/reuse scenario (1b), as the separate/replace scheme (1a) lies between the lower bound of full reuse and the upper bound of total electrolyte replacement options in terms of resources required. Again, we note that simple addition of more active species or electrolyte without concomitant removal of contaminants or contaminated electrolyte is likely to be an unsustainable solution in most cases, as it will lead to increases in solute concentration or total volume, respectively. These two remediation schemes are modeled differently, though both are fairly simple to represent.</p><p>Total electrolyte replacement cost is intuitively modeled as the product of the electrolyte cost (Celectrolyte, $ kWh -1 ) and the nominal capacity rating of the battery, plus an operational servicing fee. In this case, the fee should cover the labor to execute the replacement, the cost to transport electrolyte to and from the battery site, and perhaps the post-processing or disposal of the spent electrolyte. We estimate the costs for the labor and transport are ~4 $ kWh -1 , so we use this as a lower bound for the servicing fees used with the finite-lifetime cases (see SI Section S1.3 for details). Conversely, the separate/recover/reuse scenario is difficult to rigorously model, as there is chemistry-specificity regarding the exact methods and, by extension, associated costs needed to separate out decayed and crossed-over contaminants, reverse any decay, and finally reintroduce these species to their original half-cell. Accordingly, we elect to encompass all of these material and energy costs, in addition to the cost of the labor required to execute these actions, in the bulk operational servicing fee term. By varying the magnitude of the service fee (here, we show results using 4 and 20 $ kWh -1 ), it is possible to estimate what additional servicing costs are allowable if the RFB chemistry is to be cost-competitive with the VRFB, on a LCOS basis. To further facilitate the modeling of these chemistries, we set the rCO (the capacity fade rate that is recoverable upon rebalancing) to 0% per cycle, encompassing all fade in the rED term, as all fade experienced in these asymmetric chemistries must be recovered via servicing (i.e., rebalancing to remediate crossover losses is not an option). Thus, crossover is treated as a mode of electrolyte decay, because it requires the same general remediation mechanisms as active species decay (i.e., options 1 and 2, explained above). We note that crossover complicates the chemistry-specific separations needed by adding more species to separate on top of those that have decayed, particularly because the fate of crossed-over species (i.e., whether they stay intact or decay to any number of products) could be uncertain or variable [62]. Finally, as discussed in the Introduction, we assume these finite-lifetime species are organic compounds, with an average equivalent weight of 150 grams per mole, in aqueous electrolytes (we also assume that, in the case of two electron transfer, both transfer events occur at the same electrode potential) [55]. Quantitative representations of the chemistry-specific servicing costs (which includes both the servicing fee as well as other chemical costs, which are treated separately) can be found in Section S1.2 of the SI.</p><p>In addition to the operational service fee, the achievable lower bounds of electrolyte decay rate and electrolyte cost for finite-lifetime chemistries remain open questions and are the focus of active fundamental and applied research. To date, at-scale demonstrations of finite-lifetime chemistries in RFBs have been limited to a few start-up companies [63-67], and details on their specific redox chemistries, system configurations, performance abilities, and operational and maintenance approaches are not reported. Consequently, we perform sensitivity analyses on electrolyte cost and decay rate, along with reactor cost, to determine cost and performance targets. Figure 2 below shows the LCOS sensitivity as a function of these three variables for the two remediation schemes (separate/recover/reuse in green and total electrolyte replacement in red) for an asymmetric, finitelifetime chemistry. Values for VRFBs are provided for comparison (plotted in blue), which are treated as constant base cases because they have a relatively developed market and established body of research such that there is greater certainty around the present techno-economic parameter values. The VRFB case assumes a service fee of 0 $ kWh -1 , as it must only encompass the labor of adding the chemical reductant, which was determined to be negligible (see SI Section S1.3), while two higher service fees (4 and 20 $ kWh -1 ) are used for the asymmetric cases. These different operational service fees are represented by varying line styles. There are two immediate insights gained from Figure 2: (1) LCOS is highly sensitive to electrolyte decay rate and electrolyte cost, and (2) the separate/recover/reuse scheme appears more likely to be competitive with VRFBs than the total replacement scheme. Indeed, at a service fee of 4 $ kWh -1 , total electrolyte replacement requires very low decay rates (≤0.02 % capacity loss per day at the baseline electrolyte cost of ~50 $ kWh -1 ) and/or electrolyte costs (≤13 $ kWh -1 at the baseline decay rate of 0.1 % capacity loss per day), or some optimal combination between these baselines and targets for both parameters, to achieve a lower LCOS than a VRFB. Whereas, even with a higher service fee of 20 $ kWh -1 , the separate/recover/reuse scheme enables more lenient targets for the decay rate (≤0.06 % capacity loss per day at the baseline electrolyte cost of ~50 $ kWh -1 ) and the electrolyte cost (≤30 $ kWh -1 at the baseline decay rate of 0.1 % capacity loss per day). However, these cost and performance targets are highly dependent on the service fee, particularly for the separate/recover/reuse scheme.</p><p>To contextualize these electrolyte cost targets, we can look to the limited techno-economic studies on aqueous organic electrolytes (note: all studies assume an average cell voltage of 1.5 V). Darling et al. estimated the electrolyte cost for an aqueous organic RFB to be ~235 $ kWh -1 in 2014, and between 45 and 90 $ kWh -1 in the "future" [55]. The 2018 work by Dieterich et al. modeled the production cost of AQDS (~157 grams per mole electron, assuming a two-electron transfer), a well-known finite-lifetime active species for RFBs that is relatively easy and low-cost to manufacture [23,68], and estimated the total electrolyte cost for an AQDS chemistry (assuming the cost of the negative and positive electrolytes are approximately equal) to be 50 and 65 $ kWh -1 at production scales ~100 and ~200 MWh of flow battery capacity deployed per year, respectively [32]. Based on their estimates of materials costs alone, it is difficult to envision reducing electrolyte costs below 30 $ kWh -1 while utilizing existing production methods (regardless of production scale). Furthermore, a recent study by Gregory et al. estimates that reducing the electrolyte price of an aqueous RFB system using AQDS on the negative side or a ferrocyanide-based positive electrolyte to our baseline electrolyte cost of 25 $ kWh -1 per side (i.e., 50 $ kWh -1 overall) would require a production scale equivalent to producing 10 GWh of flow batteries per year [60].</p><p>Currently, there only ~100 MWh of RFBs deployed globally, with another ~1 GWh contracted, announced, or under construction [69]. These studies clearly demonstrate that low-cost (i.e., ≤50 $ kWh -1 ) electrolytes for finite-lifetime chemistries will require one or more of the following factors: the use of previously unstudied active molecules, development of new production pathways for existing active molecules (e.g., AQDS), internal production of the active molecules by the RFB company (to minimize markups by suppliers), and/or drastic increases to production scale (either by growth of the RFB market utilizing these chemistries and/or other markets for these active species). Therefore, the more promising pathways to viable asymmetric chemistries with finite lifetimes are those that can enable low service fees or low decay rates.</p><p>With respect to electrolyte decay rates, those reported in the literature range five orders of magnitude (from as low as order 0.001 to as high as order 10, in units of % capacity loss per day), which challenges a feasibility judgement on the decay-rate baseline of 0.1 % capacity loss per day [30,37,38]. This variability can be attributed to differences in the choices of active species, experimental apparatus, testing protocols, and other experimental conditions. Further, the protocols and the conditions used to measure these decay rates may be less aggressive than those of a deployed system, potentially making these conservative estimates. That said, several publications have reported active-species stability in the desired range of ≤0.1 % capacity loss per day [30,37,38,70,71], including many quinone-derivatives. In working to reduce these key variables, note that electrolyte price and decay rate may not be independent, as functionalization of organic molecules often improves stability [30,72] and, likely, simultaneously complicates the manufacturing process and thus adds to the chemical cost. Relative to those two variables, reactor cost has a lesser effect on LCOS, though it is important to consider as organic molecules are generally larger than their aqueous supporting salts, providing the opportunity to leverage sizeexclusion membranes as opposed to more expensive ion-exchange membranes [33]. There is also the potential to employ lower-cost membranes ill-suited for VRFBs, either via the use of electrolytes with milder pH [37,70] or a less oxidizing active species [9].</p><p>Based on our evaluations of the feasibility of achieving the relevant targets for electrolyte cost and decay rate, the separate/recover/reuse remediation process appears capable of making asymmetric chemistries of finite lifetimes competitive with VRFBs if the service fees to employ it can be kept sufficiently low. The 20 $ kWh -1 service fee already sets seemingly difficult targets; for example, an electrolyte cost of 30 $ kWh -1 seems infeasible with our current solutions, even assuming the possibility of vast scale-up, as previously discussed. From these observations, we propose that the costs to separate/recover/reuse should be limited to ≤10 $ kWh -1 to allow for viability at the expected lower limits for electrolyte cost and decay rate. At this juncture, it is difficult to assess the feasibility of this target, due to the absence of discussion or study of these methods in the open literature. A notable exception is the work of Goulet et al., which explored decay reversibility for an aqueous chemistry with a finite-lifetime species on the negative side (i.e., a quinone derivative) [36]. The authors were able to reverse 70% of the decay, which had already been mitigated to 0.14 % per day by limiting the state of charge, via aeration of the electrolyte. They estimated that the modifications to RFB operation needed to facilitate this capacity loss reduction/remediation would add ~20 $ kWh -1 to the capital cost, which corresponds to an annual operating and maintenance cost of just over 2 $ kWh -1 per year if calculated assuming a 20 year lifetime and 8% discount rate [73]; while this is within our desired range, the remaining 30% of the decayed species that cannot be rejuvenated via aeration alone would eventually require separations.</p><p>Three complicating features of this system to consider regarding separations are (1) the high overall electrolyte concentrations, (2) the need to keep the main stream of electrolyte (i.e., what remains after separating out the decayed species) almost entirely uncontaminated from the separations process and the further desire to recover the decayed species intact as well for reuse in the system, and (3) the likely similar characteristics of the decayed species being targeted for separations and the active species that must remain in the electrolyte. Methods for separating organics via exploitation of differences in the physical, chemical, and/or electrochemical properties exist and can even separate similar compounds (e.g., isomers) [74,75]. It seems reasonable to assume that expertise in separating decay product lies within the process industry given the requirements to create products of sufficient purity. Consequently, chemical manufacturers may be uniquely positioned to design new redox active compounds with decay reversal or recovery as a key design parameter, or else offer technical solutions for separation of pristine species from decay products either on-site or at a centralized facility. Next, we consider remediation techniques for infinite-lifetime species. We note that while separations processes could be used for crossover remediation in these cases as welland there are, in fact, relatively developed and low-cost methods for separating mono-and multi-valent ions from multicomponent mixtures (e.g., wastewater treatment) [76] the other techniques available for these chemistries, discussed below, are anticipated to be economically preferable to total electrolyte replacement or complicated separations.</p><p>Figure 2 -LCOS as a function of electrolyte decay rate (a, top), electrolyte cost (b, middle), and reactor cost (c, bottom) for a generic, asymmetric chemistry employing the separation/recovery/reuse remediation method (green lines) or total electrolyte replacement remediation method (red lines). These are compared against baselines for a VRFB (blue lines), which do not change with the x-axis variables. The vertical lines represent the baseline value of each x-axis variable for the asymmetric case (grey) and the vanadium case (blue) (note that vanadium does not decay and its electrolyte cost, ~122 $ kWh -1 , exceeds the x-axis scale in 2b, thus the vanadium x-axis baseline is only visible in 2c). The line styles correspond to varying operational service fees.</p><!><p>Crossoverdriven by electroosmotic drag and gradients of concentration, potential, and/or pressure across the half-cellsis a primary cause of capacity loss for asymmetric RFB chemistries with active species of infinite lifetimes. However, the stability of these active species in their home and opposing cell environments enables use of the spectator strategy, where both active species are dissolved in each half-cell electrolyte. A mixed electrolyte is typically prepared with equal concentrations of active species in their discharged forms and used in both reservoirs. During operation, the active species for the positive half-cell reaction serves as the charge storage species on the positive side of the cell and as a spectator on the negative side of the cell. The opposite is true for the active species for the negative half-cell reaction. This transforms asymmetric chemistries into pseudo-symmetric chemistries and allows for utilization of the same VRFB rebalancing methods [12,49,[77][78][79][80] to remediate losses due to crossover. Further, crossover is actually mitigated by the spectator methodology itself [14,81], as diffusive fluxes between the two electrolytes that drive crossover are also significantly decreased with this strategy [62].</p><p>As mentioned before, implementation of this strategy requires that both active species are chemically and electrochemically stable in the opposing half-cell, but there are also important techno-economic considerations. The spectator strategy typically decreases energy density and increases electrolyte cost, limiting the chemistries to which this approach can be applied costeffectively. The addition of the spectator species lowers the solubility of the active species [14], limiting the energy density, as well as doubles the active material required for the same energy output, increasing electrolyte costs. Lower energy densities are arguably less concerning for stationary energy storage applications where the size and mass constraints for the battery are more lenient as compared to mobile ones, but the increased electrolyte cost could be prohibitive for grid applications where lower cost solutions (e.g., fossil fuels or cheaper battery chemistries/technologies) are readily available. At the very least, the savings from employing a low-cost asymmetric material set may be lost if the spectator strategy increases the energy costs to the point they exceed that of the incumbent VRFB chemistry. Thus, we can calculate the total electrolyte cost for a spectator chemistry as a function of three key variablesactive species costs, active species equivalent weight, and cell potentialto estimate the available design space for these chemistries to be competitive the VRFB (Figure 3). Other variables that affect electrolyte costsuch as accessible depth of discharge, cost of solvent, solubility, etc.are generally more consistent across aqueous chemistries, as compared to these three highly chemistry-dependent variables [55]. We vary equivalent weight and cell potential within the bounds typically seen in RFB chemistries (50-150 grams per mole and 1.0-1.6 V, respectively [55]), and use a range of active species costs that would generally classify low-cost, high-abundance materials (≤20 $ kg -1 ). Across the range of values studied, the combination of the active species equivalent weight and chemical cost significantly impacts the final electrolyte cost and thus the economic viability, whereas cell potential has a less pronounced effect, particularly with increasing active species equivalent weight (though we note that cell potential also impacts the power costs, which is not accounted for in this simple electrolyte-cost comparison). We see the majority of the design space for these spectator strategy chemistries is competitive with the VRFB, and this space seems reasonable: many new chemistries being studied use abundant active materials, such as iron, zinc, sulfur, etc., all of which have been cited in RFB literature to cost ≤10 $ kg -1 and have equivalent weights and cell voltages in the middle of these ranges [55,59]. However, using this strategy with larger active species (~150 grams per mole), such as organics or ligand-modified transition metals, requires very low costs to be competitive with VRFB (≤5 $ kg -1 ). Several RFB chemistries leverage the spectator strategy [82,83], with perhaps the most notable being iron-chromium (Fe-Cr).</p><p>Figure 3 -Electrolyte cost (y-axis) as a function of active species cost (x-axis), active species equivalent weight (line colors), and cell potential (line styles) for chemistries utilizing the spectator strategy. For context, the baseline electrolyte cost of the VRFB (122 $ kWh -1 , assuming an equivalent weight of 51 grams per mole, active species cost of ~ 30 $ kg -1 , and potential of 1.4 V) is plotted as a blue solid line.</p><p>The pseudo-symmetric Fe-Cr RFB offers several benefits as compared to the VRFB. Use of the spectator method for the Fe-Cr chemistry has been shown to significantly reduce net crossover rates [84], which is important as iron and chromium ions are ~20× more permeable than vanadium ions in Nafion membranes [85]. The chemistry uses charge-storage species of high crustal abundance, as iron is the most abundant element in the Earth (by mass) and there is nearly 1000× more chromium resources than vanadium [26]. These active materials are also low-cost: from late 2019 through early 2020, the price of ferrochromium was ~2 $ kg -1 of chromium content [58].</p><p>Further, ferrochromium contains forms of both active species, which can facilitate cost savings by minimizing waste and reducing the processing steps needed to generate to electrochemical grade electrolyte if employed in an Fe-Cr system utilizing spectator strategy [84]. However, the open circuit voltage of Fe-Cr is ~0.98 V at typical operating temperatures (i.e., ~65 °C) [84] and the active species solubility in the spectator configuration are ~1 M [56,81], limiting energy and power densities. Assuming a four-hour duration, we estimate the capital cost of the Fe-Cr RFB to be lower than that of the VRFB, at ~211 and ~268 $ kWh -1 (including optimal oversizing to minimize LCOS, as explained previously), respectively, where the electrolyte costs of the Fe-Cr are about a fifth of the VRFB electrolyte costs (~23 and ~122 $ kWh -1 , respectively, not including oversizing) and the reactor costs of the Fe-Cr are about double that of the VRFB (~323 and ~158 $ kW -1 , respectively). These numbers align with other techno-economic assessments of these systems [7,24]. Despite also facing capacity loss due to hydrogen evolution at a rate ~20× that seen in VRFBs (~1 % vs 0.055 % of capacity loss to hydrogen evolution per cycle) [24,54,86], the Fe-Cr system also shows improvement over VRFBs in terms of LCOS. We estimate the LCOS of Fe-Cr to be ~260 $ MWh -1 , a moderate reduction from the LCOS of the VRFB (~290 $ MWh -1 ). Even artificially increasing the hydrogen evolution-induced capacity fade rate in the Fe-Cr system to as much as 10% capacity loss per cycle does not raise the LCOS of the Fe-Cr system above 270 $ MWh -1 . Modeling details used to derive these numbers can be found throughout the SI. The techno-economic promise for Fe-Cr is evident, however, the capital cost of the system still exceeds the Department of Energy target of ≤150 $ kWh -1 for viable grid storage [50,51]. Reductions in the power costs (i.e., beyond the chemistry choices probed in this work, perhaps by increasing the duration or using lower-cost reactor materials) are likely needed. There is, however, the potential for further cost reductions for the Fe-Cr system with any significant improvements to performance.</p><p>Most of the Fe-Cr research was executed in the 1970's and 1980's when NASA first introduced this chemistry as the first true RFB while exploring energy storage solutions for deep-space missions [87]. Research into the Fe-Cr system has been limited relative to that for VRFBs, and it is likely that many of the significant improvements to the VRFB system seen over the past 5-10 years can be applied to the Fe-Cr system to increase performance and reduce costs. Some of this has already been demonstrated; for example, recent studies have shown the benefits of advanced cell engineering and optimized electrolyte composition for the Fe-Cr system [39,57].</p><!><p>Membranes with perfect selectivity for the desired charge-carrier species, such as non-porous, single ion-conducting (SIC) materials, could eliminate crossover losses experienced by stable RFB chemistries, obviating the need for symmetric and pseudo-symmetric electrolytes. Research into SIC membranes, mainly ceramics and ceramic-polymer composites, for RFBs has been limited, with most studies employing them for energy dense semi-solid/hybrid redox chemistries or systems utilizing two electrolytes of different pH [44][45][46][47][48]88]. More extensive exploration has likely been hampered by the absence of broadly-available commercial materials, the lack of crossdisciplinary expertise between the fields, and the experimental challenges of integrating a ceramic into a contemporary flow-cell architecture. There is an inherent tradeoff between improving (i.e., increasing) selectivity and worsening (i.e., increasing) resistivity, and thus high selectivity typically results in large ohmic resistance; for example, Allcorn et al. measured a resistance of ~90 Ω-cm 2 for their 1.1 mm thick ceramic membrane (using a symmetric ferro-/ferri-cyanide chemistry) [44]. The total area-specific resistance (ASR) of a state-of-the-art RFB cell is mostly ohmic resistances, plus some minor contributions from kinetic transport losses. Therefore, these large resistances not only represent significant performance losses, but they also have major economic consequences since they substantially impact power density, efficiency, and ultimately the cost of the reactor (the power delivered per unit area of reactor is inversely proportional to ASR [55]). The price of these membranes is also uncertain; although there is a sizable body of literature on ceramic membranes for water-treatment applications, a subset of which focus on the development of low-cost options ranging from as low as 25 $ m -2 to 500 $ m -2 [89][90][91][92][93]. At present, it is not clear how relevant these estimates are to the material sets conducive to use in RFBs, as application-specific design criteria vary (e.g., flexibility, conductivity, chemical compatibility, etc.).</p><p>While the lower bounds of the cost and ASR for SIC membranes have yet to be determined, one may use techno-economic analyses to estimate what values these parameters would need to be in order to present a competitive solution for RFBs. We estimate the LCOS as a function of ASR (Figure 4), which, as mentioned previously, linearly scales the reactor cost. We use the inputs for the infinite-lifetime species in Table 1, as SIC membranes are most effective for chemistries which experience losses that are dominated by crossover, and accordingly assume zero capacity fade (i.e., negligible capacity loss from non-crossover sources). We then vary the SIC membrane price between 100 and 300 $/m 2 , which is within the range seen in water treatment literature and comparable to the present-day cost of Nafion [7,94]. We also vary two critical chemistrydependent parameters: active species cost (Figure 4a) and cell potential (Figure 4b). We find that LCOS is not particularly sensitive to active species cost and is more sensitive to membrane price, which makes sense as the reactor cost dominates the capital cost at high ASRs such that changes to electrolyte cost have a relatively small impact on both the capital cost and the LCOS. This is also why we find a strong LCOS sensitivity to cell potential (U), which scales both reactor costs (∝ U -2 ) and energy costs (∝ U -1 ). Thus, higher cell potentials can, at least partially, offset elevated cell ASRs. Reducing the measured ASR to ~15 Ω-cm 2 in RFB cells employing SIC membranes, a 6x reduction from published reports [44], may be technically achievable, but a viable implementation also requires a moderate cost for the SIC membrane (i.e., ~100 $ m -2 ) and a high voltage redox chemistry (i.e., ≥3 V). Such requirements likely necessitate the use of non-aqueous electrolytes, which is still a nascent branch of RFB systems. Furthermore, such high ASRs create physical complications: a cell with an ASR of 15 Ω-cm 2 would require 30× the area of a conventional cell configuration with a polymeric membrane like Nafion (~0.5 Ω-cm 2 , for aqueous systems) for a given power output and chemistry, which is a substantial increase to the reactor footprint that must be considered. This may be less of an issue for long-duration applications, as the energy-capacity components increasingly dominate the system configuration. While the design space presented here suggests what a successful SIC membrane might look like, further studies of SIC membranes for RFBs, with particular focus on understanding tradeoffs between resistance, cost, and mechanical stability, are needed to more completely assess the viability of this approach. Figure 4 -LCOS as a function of the cell ASR, with membrane price (line colors), and active species cost (a, left) or potential (b, right) (line styles) as variable parameters. The left plot assumes a cell potential of 1.5 V, and the right plot assumes an active species cost of 5 $ kg -1 . The baseline LCOS of vanadium (assuming a potential of 1.4 V, ASR of 0.5 ohm-cm 2 , active species cost of ~30 $ kg -1 , and membrane cost of 300 $ m -2 ) is plotted in blue.</p><!><p>The desire for and research of potentially-inexpensive RFB chemistries has been growing in response to financial concerns around the cost of vanadium active species in the most mature RFB chemistry, the VRFB. However, these options present more challenges to consider beyond the alteration in cost of the electrolyte or even the kinetic, thermodynamic, and mass transport challenges of these new chemistries. As many of these new chemistries are asymmetric, the methods and associated costs of asymmetric capacity-loss remediation must be explored to determine if the complications that arise from cross-contamination outweigh the reduced capital cost relative to the more easily remediable VRFB. Accordingly, we have adapted our LCOS model, used in previous work for VRFBs, to evaluate two classes of asymmetric chemistries: those using active species of finite and infinite lifetimes. Finite-lifetime chemistries, often employing organic active species, primarily suffer capacity losses from active species decay, necessitating their periodic replacement. This can be achieved by performing total electrolyte replacement, or by selectively separating out and replacing or reusing the decayed species. We found that the separations route is substantially more economically effective, but only if such processes can be executed with low enough costs, and the LCOS of these systems is highly sensitive to the electrolyte cost and decay rate. We estimate that the cost to separate/recover/reuse should be limited to ≤10 $ kWh -1 , and future work should explore electrolyte separation and recovery methods to better assess the feasibility of this target. This analysis has revealed an opportunity for chemical-manufacturing companies who may be uniquely positioned to design organic redox active species with decay remediation in mind as a key design criterion. Infinite-lifetime species primarily suffer capacity losses from crossover, which can be remediated by making the chemistry pseudo-symmetric via the spectator strategy or avoided altogether with the use of perfectly selective separators. The spectator strategy is only effective if the species are stable in their opposing half-cell's environments and if the resulting decrease in energy density and the increase in required active material do not increase the electrolyte cost above that of other potential symmetric chemistries like the VRFB; this requires active species with relatively low active species costs and/or equivalent weights (≤15 $ kg -1 for active species ~50 grams per mole, or ≤5 $ kg -1 for active species ~150 grams per mole). We found that the Fe-Cr system, which has not been as widely studied or improved upon as compared to VRFBs, is a promising candidate chemistry for effective use of the spectator strategy to reduce the capital cost and the LCOS, as compared to the VRFB system. However, this case study highlights that, in order to reduce capital costs below 150 $ kWh -1 , reductions in power costs (i.e., beyond choice of chemistry) are likely needed.</p><p>Perfectly selective separators, likely ceramic-based materials, eliminate crossover at the expense of high cell resistances and power costs. The reduction in cell resistance needed to make this solution competitive seems feasible if the separator can be produced at sufficiently low costs and employed with high potential chemistries, as cell potential counters the effect of resistance on power costs and also reduces energy costs. This approach appears to be most suitable for nonaqueous electrolytes, where cell potentials ≥3V are viable.</p><p>Looking forward, this LCOS model can be used as a framework to determine cost and performance targets for evaluating the techno-economic promise of new RFB chemistries and their potential capacity-loss remediation strategies. Future work will focus on expanding our treatment of capacity fade to encompass its dynamic nature, both in its mechanisms and rates, as a function of cell operating conditions (e.g., the application-informed duty cycle or temperature), cell components (e.g., choice of membrane or flow field), and electrolyte composition (e.g., choice of active or supporting species). This can be done by building on existing crossover and decay models to incorporate more fade mechanisms and power-dependence, and subsequently determining the inputs to these models for various chemistries. This will allow for a more accurate understanding of chemistry-dependent crossover, as well as evaluation of the efficacy of crossover remediation approaches (e.g., rebalancing or the spectator strategy).</p>

ChemRxiv

Corymbulosins I\xe2\x88\x92W, Cytotoxic Clerodane Diterpenes from the Bark of Laetia corymbulosa

The isolation studies of a crude MeOH/ CH2Cl2 (1:1) extract (N005829) of the bark of Laetia corymbulosa yielded 15 new clerodane diterpenes, designated corymbulosins I\xe2\x88\x92W (1\xe2\x88\x9215), as well as four known diterpenes, 16\xe2\x88\x9219. The structures of 1\xe2\x88\x9215 were characterized on the basis of extensive 1D and 2D NMR and HRMS analyses. The absolute configurations of newly isolated compounds 1\xe2\x88\x9215, as well as known 16\xe2\x88\x9219, which were reported previously with only relative configurations, were determined through ECD experiments, X-ray analysis, chemical methods, including Mosher esterification, and comparison of their spectroscopic data. The isolated compounds were evaluated for cytotoxicity against human cancer cell lines. Flow cytometric and immunocytochemical observations of cells treated with cytotoxic clerodanes demonstrated that the chromatin was fragmented and dispersed with formation of apoptotic microtubules.

corymbulosins_i\xe2\x88\x92w,_cytotoxic_clerodane_diterpenes_from_the_bark_of_laetia_corymbulosa

INTRODUCTION<!>RESULTS AND DISCUSSION<!>CONCLUSIONS<!>General Experimental Procedures.<!>Plant Material.<!>Extraction and Isolation.<!>Corymbulosin I (1).<!>Corymbulosin J (2).<!>Corymbulosin K (3).<!>Corymbulosin L (4).<!>Corymbulosin M (5).<!>Corymbulosin N (6).<!>Corymbulosin O (7).<!>Corymbulosin P (8).<!>Corymbulosin Q (9).<!>Corymbulosin R (10).<!>Corymbulosin R (11).<!>Corymbulosin T (12).<!>Corymbulosin U (13).<!>Corymbulosin V (14).<!>Corymbulosin W (15).<!>General Procedure for Esterification with (S/R)-MTPA-Cl.<!>Assay for Antiproliferative Activity.<!>Cell Cycle Analysis.<!>In Vitro Hemolytic Assay.<!>Immunocytochemistry.

<p>The family Salicaceae, formerly placed in Flacourtiaceae, is known to produce isozuelanin-type clerodane diterpenes.1 In particular, Zuelania and Casearia genera typically yield such diterpenes, including esculentin A and zuelaguidins A−D and F from Z. guidonia,2 caseanigrescens A−D from C. nigrescens,3 argutins A−H from C. arguta,4 caseamembrins A−U from C. membranacea,5–9 caseabalansin A−G and balanspenes A−H from C. balansae,10,11 and caseagrewiifolins A and B from C. grewiifolia.12 In these compounds, the tricyclic system normally is formed from a cis-fused decalin with a disubstituted tetrahydrofuran connected at C-4−C-5. Further characteristic features are two trans-configured methyls at C-8 and C-9 and a 3-methylenepent-4-en-1-yl side chain at C-9. In general, the individual compounds differ in the substituents and stereochemistries at C-2, C-6, and C-7. Although the relative configurations of the eight stereocenters in isolated compounds were usually determined by 2D-NMR experiments, such as NOE, the absolute configurations were not determined in most cases. In many reported studies, isozuelanin-type clerodanes have exhibited potent cytotoxicity.</p><p>The genus Laetia belongs to the family Salicaceae, but only limited phytochemical studies have been reported so far on plants from this genus.13–15 Previous studies of L. corymbulosa reported the isolation of eight isozuelanin-type clerodanes, corymbulosins A−H.16,17 In our continuing phytochemical investigation of rainforest plants, 15 new corymbulosins I−W (1−15) and four known 2-esterified tricyclic clerodane diterpenes were isolated from L. corymbulosa collected in Peru. Herein, the determination of the absolute configurations of all isolated diterpenes and their biological activities is described.</p><!><p>The MeOH/CH2Cl2 (1:1) extract of L. corymbulosa (N005829) was provided by the U.S. National Cancer Institute (NCI, Frederick, MD, USA). The EtOAc soluble part was separated through a combination of column chromatography, preparative HPLC, and preparative TLC using silica gel and octadecylsilane (ODS).</p><p>Compound 1 (Figure 1) was obtained as an optically active colorless amorphous solid: [α]D25 +2.3 (c 0.38, CHCl3). The HRFABMS data revealed a molecular formula of C28H40O8 from the ion at m/z 527.2631 [M + Na]+. The 1H NMR spectrum (Table 1) contained signals attributable to two methyls [ δH 0.93 (3H, d, J = 7.2 Hz, H-17), 0.92 (3H, s, H-20)], two vinyl protons [ δH 5.98 (1H, dd, J = 4.2, 1.8 Hz, H-3), 6.44 (1H, dd, J = 17.4, 10.2 Hz, H-14)], four methylidene protons [ δH 5.17 (1H, d, J = 17.4 Hz, H-15a), 5.03 (1H, d, J = 10.2 Hz, H-15b), 5.06 and 4.94 (each, 1H, s, H2-16)], two acetoxy methyls [ δH 2.08 (3H, s), 1.90 (3H, s)] at C-18 and C-19, two acetal-acyloxy methine protons [ δH 6.75 (1H, t, J = 1.8 Hz, H-18), 6.47 (1H, s, H-19)], and a methine proton [ δH 2.33 (1H, dd, J = 12.0, 5.4 Hz, H-10)], which suggested the presence of an 18,19-di-O-acetyl-18,19-epoxycleroda-3,13-(16),14-triene, the parent skeleton of corymbulosins A−H isolated from the same plant.16,17 Further observation of the NMR spectra (Tables 1 and 2) revealed an oxymethine [ δH 3.80 (1H, m)/ δc 73.0, C-6], an aliphatic moiety [ δH 2.63 (1H, sep, J = 7.2 Hz), 1.22 and 1.20 (each, 3H, d, J = 7.2 Hz)/ δc 34.7, 19.2 and 18.7], and an additional carbonyl carbon ( δc 176.4, C-1′) in the structure of 1. COSY and HMBC experiments (Figure 2) supported the presence of hydroxy and isobutanoyloxy groups at C-6 and C-2, respectively, on the parent skeleton of 1. The observation of NOESY correlations between H-1/H-6, H-6/H-8, H-7/H-19, H-10/ H-12, and H-11/H-19 as well as a ROESY correlation for H-18/H-19 suggested the relative configuration as shown in Figure 3. The absolute configuration was established by the modified Mosher ester method. The hydroxy group at the chiral C-6 was esterified using (R)- and (S)-α-methoxy-α-trifluoromethylphenylacetyl chloride (MTPA-Cl) to generate the related esters. The distribution of the positive and negative Δ δH (S−R) values of the MTPA esters indicated an S-configuration of the C-6 chiral center (Figure 4). The experimental electronic circular dichroism (ECD) spectra are shown in Figure 5. Therefore, compound 1,18 given the trivial name corymbulosin I, was assigned as (2R,5S,6S,8R,9R,10S,−18R,19S)-18,19-di-O-acetyl-18,19-epoxy-6-hydroxy-2-isobuta-noyloxycleroda-3,13(16),14-triene.</p><p>Compound 2 has the identical molecular formula, C28H40O8, as 1, based on the peak at m/z 527.2631 [M + Na]+ in the HRFABMS. The 1H and 13C NMR spectra of 2 were very similar to those of 1, except for H-1 at δH 2.15/1.71 (each 1H, m), H-2 at δH 5.59 (1H, m), and C-2 at δc 70.4. The NOESY correlation between H-2 and H-10 (Figure 3) indicated that compound 2 was the C-2 epimer of 1. The ECD spectrum (Figure 5) of 2 also exhibited a different Cotton effect from that of 1 and a similar effect to that of 18, in which H-2 has α-orientation. Therefore, compound 2 (corymbulosin J) was assigned as (2S,5S,6S,8R,9R,10S,18R,−19S)-18,19-di-O-acetyl-18,19-epoxy-6-hydroxy-2-isobutanoy-loxycleroda-3,13(16),14-triene.</p><p>Compound 3 was obtained as an optically active colorless oil: [α]D25 +10.4 (c  0.23, CHCl3). The HRFABMS data showed a molecular formula of C29H42O8 from the peak at m/ z 541.2777 [M + Na]+. The 1H and 13C NMR spectra of 3 were similar to those of 1 (Tables 1 and 2), but the differences in chemical shifts for H/C-6 and H/C-7 between 3 and 1, as well as additional peaks at δH 3.30 (3H, s) and δc 57.5 in 3, indicated the presence of a methoxy group at C-6. These results implied that compound 3 is a methoxy analogue of 1, which was verified from COSY and HMBC experiments (Figures S94 and S95). The NOESY correlations of 3 were similar to those of 1, with an additional correlation found between H-18 and the C-6 methoxy protons in 3 (Figure 3). Comparison of all NMR data and the optical rotations of 3 and 1 showed that 3 and 1 have the same absolute configuration, which was confirmed from ECD spectra as shown in Figure 5. On the basis of the above findings, compound 3 (corymbulosin K) was determined as (2R,5S,6S,8R,9R,10S,18R,19S)-18,19-di-O-acetyl-18,19-epoxy-6-methoxy-2-isobutanoyloxycleroda-3,13(16),14-triene.</p><p>Compound 4, optically active colorless oil with [α]D25 +12.3 (c 0.05, CHCl3), gave a sodiated peak [M + Na]+ at m/z 527.2608 in the HRFABMS, in agreement with a molecular formula of C28H40O8. The 1H and 13C NMR spectra (Tables 1 and 2) indicated that compound 4 is an analog of 3 with a different ester group at C-2. The combination of a three-proton triplet (J = 7.2 Hz) at δH 1.19 and a two-proton quartet (J = 7.2 Hz) at δH 2.41 suggested the presence of a propionyl group at C-2 in 4 rather than the isobutanoyl ester found in 3. The final structure assignment and relative configuration were confirmed by various 2D NMR analyses of 4 (Figures S94 and S95). The absolute configuration was determined from an ECD spectrum (Figure 5), and optical rotation is the same as that of 3. Thus, compound 4 (corymbulosin L) was assigned as (2R,5S,6S,8R,9R,10S,18R,−19S)-18,19-di-O-acetyl-18,19-epoxy-6-methoxy-2-propionyloxy-cleroda-3,13(16),14-triene.</p><p>Compound 5 was obtained as an optically active colorless oil: [α]D25 +28.4 (c 0.02, CHCl3). The HRFABMS data supported a molecular formula of C27H38O8 from the peak at m/z 513.2479 [M + Na]+, which indicated the loss of a methylene unit from 4. The 13C and 1H NMR spectra of 5 (Tables 1 and 2) were mostly identical to those of 4, except for the appearance of signals for an acetyl group at δH 2.14 (3H, s), δc 170.7 (C=O) and 21.5 (CH3), and disappearance of signals for the ethyl group at δH 1.91 (t, 3H)/ δc 9.3 and 2.41 (q, 2H)/ δc 28.0 found in the spectra of 4. The 2D NMR (Figures S94 and S95), optical resolution, and ECD analyses (Figure 5) suggested that all chiral centers in 5 have the identical stereochemistry found in 1, 3, and 4. Therefore, compound 5 (corymbulosin M) was assigned as (2R,5S,6S,−8R,9R,10S,18R,19S)-2,18,19-tri-O-acetyl-18,19-epoxy-6-me-thoxycleroda-3,13(16),14-triene.</p><p>Compound 6 has the molecular formula C30H44O8 based on the peak at m/z 555.2947 [M + Na]+ in the HRFABMS. The 1H and 13C NMR data of 6 were mostly identical with the reported data of 18,19-di-O-acetyl-18,19-epoxy-6-methoxy-2-(2′-methylbutanoyloxy)-cleroda-3,13(16),14-triene with the s-cis form of the C-9 side chain.15 However, the experimental optical rotation, [α]D25 +8.4 (c 0.20,  CHCl3), was different from the reported value, [α]D −84 (c 0.11, CHCl3). After a thorough examination of all spectroscopic data including NOESY and ECD experiments (Figures S94, S95, and 5), we concluded that compound 6 (corymbulosin N) is (2R,5S,6S,−8R,9R,10S,18R,19S)-18,19-di-O-acetyl-18,19-epoxy-6-methoxy-2-(2′-methylbutanoyloxy)-cleroda-3,13(16),14-triene with the s-trans form of the C-9 side chain. The absolute configuration of the isobutyl side chain was not determined, since no differences were calculated in the ECD spectra of the 2′R and 2′S isomers.</p><p>Compound 7 was isolated as an optically active colorless oil [α]D25 −64 (c 0.07, MeOH) . HRFABMS analysis of 7 afforded a molecular ion at m/z 555.2921 [M + Na]+ in agreement with a molecular formula of C30H44O8, which was identical to that of 6. The 1H and 13C NMR spectra of 7 were similar to those of 6 (Tables 1 and 2) and identical to those of casearlucin E.19 The β-orientation of the ester group at C-2 was supported by a 2D NOESY correlation between H-2 and H-10 in 7 (Figures S95), which indicated that 7 was the C-2 epimer of 6. The ECD spectrum of 7 exhibited a different Cotton effect from that of 6 but similar to that of 18 (Figure 5). These observations suggested that the stereochemistry of 7 was same as that of 18; accordingly, it was an enantiomer of casearlucin E.19 This was also supported by the different sign of the optical resolution from that of casearlucin E.19 Thus, compound 7 (corymbulosin O) was assigned as (2S,5S,6S,−8R,9R,10S,18R,19S)-18,19-di-O-acetyl-18,19-epoxy-6-methoxy-2-(2′-methylbutanoyloxy)-cleroda-3,13(16),14-triene.</p><p>On the basis of a sodiated molecular-related ion at m/z 625.3340 [M + Na]+ in the HRFABMS, compound 8 has a molecular formula of C34H50O9. The 1H and 13C NMR spectra of 8 were comparable with those of 5, except for the loss of a methoxy group at C-6 and the presence of an octanoyloxy group. Seven sp3 carbons were present between δc 14.1 and δc 34.7, and a carbonyl carbon was found at δc 173.2. The 1H NMR of 8 showed 15 aliphatic protons between δH 4.95 and δH 2.31, and the oxygenated methine at C-6 was downfield shifted ( δH 4.95, 1H, dd, J = 11.3, 4.6 Hz) with respect to that of 5 ( δH 3.27, 1H, dd, J = 12.0, 4.2 Hz). These results strongly suggested the location of the octanoyloxy group at C-6. The absolute configuration was determined by analysis of NOESY (Figure 3) and ECD spectra (Figure 5). Thus, compound 8 (corymbulosin P) was determined as (2R,5S,6S,8R,9R,10S,18R,19S)-2,18,19-tri-O-acetyl-18,19-epoxy-6-octanoyloxycleroda-cleroda-3,13(16),14-triene.</p><p>Compound 9 was obtained as an optically active colorless Oil [α]D25 +19.6 (c 0.15, CHCl3). The HRFABMS data showed a molecular formula of C36H54O9 from the peak at m/ z 653.3651 [M + Na]+, which indicated two additional methylene units compared with 8. The 1H and 13C NMR spectrum of 9 (Tables 1 and 2) closely resembled those of 8, except for the additional four protons at δH 1.25 (m) and two carbons at δc 29.19 and 29.12. Together with HMBC and COSY analyses (Figures S94), compound 9 has a decanoyloxy substituent at C-6 instead of the octanoyloxy unit in 8. The absolute configuration of 9 was defined by comparison of various spectroscopic data, including optical rotation, NMR spectra, and calculated and experimental ECD spectra (Figure 5). Finally, compound 9 (corymbulosin Q) was assigned as (2R,5S,6S,8R,9R,10S,18R,19S)-2,18,19-tri-O-acetyl-18,19-epoxy-6-decanoyloxycleroda-cleroda-3,13(16),14-triene.</p><p>Compound 10 has the molecular formula C27H38O7 based on the sodiated molecular-related ion at m/z 497.2511 [M + Na]+ in the HRFABMS. The 1H and 13C NMR spectra of 10 (Tables 1 and 2) were similar to those of 4. Important differences included the absence of signals for an oxygenated methine and a methoxy at C-6. Instead, an unsubstituted methylene group was suggested by the presence of a multiplet centered at δH 1.50/1.90 and a signal at δc 29.4. An unsubstituted methylene group at C-6 was further supported by the HMBC correlations between H-19 and H-6, as well as the COSY correlations between H-6 and H-7 (Figure S94). The relative configuration of 10 was determined from a NOESY spectrum (Figure S95). The optical rotation, [α]D25 −27.1 (c 0.08, CHCl3), of 10 was similar to that of 13. The absolute configuration of the latter compound was confirmed by a modified Mosher ester method as described later. On the basis of the above data together with the ECD spectrum (Figure 5), compound 10 (corymbulosin R) was defined as (2R,5S, 8R,9R,10S,18R,19S)-18,19-di-O-acetyl-18,19-epoxy-2-propionyloxycleroda-3,13(16),14-triene.</p><p>Compound 11 was isolated as an optically active colorless oil, [α]25D −10.0 (c 0.45, CHCl3). The HRFABMS data indicated a molecular formula of C28H40O7 from the peak at m/z 511.2681 [M + Na]+. The 1H and 13C NMR spectra of 11 (Tables 1 and 2) were very similar to those of 1, except for the presence of signals for an unsubstituted methylene group rather than an oxygenated methine C-6, as also seen with the NMR comparison of compounds 10 and 4. The final structure allocation and relative configuration were determined by 2D NMR analyses of 11 (Figures S94 and S95). On the basis of an ECD spectrum (Figure 5) and optical rotation, compounds 11 and 10 have the same absolute configuration. Therefore, compound 11 (corymbulosin S) was concluded to be (2R,5S,8R,9R,10S,18R,19S)-18,19-tri-O-acetyl-18,19-epoxy-2-isobutanoyloxycleroda-3,13(16),14-triene.</p><p>Compound 12 was obtained as an optically active colorless oil, [α]25D +0.7 (c 0.35, CHCl3). HRFABMS analysis of 12 afforded a sodiated molecular ion at m/z 539.2960 [M + Na]+, which established a molecular formula of C30H44O7. The 1H and 13C NMR spectra of 12 (Tables 1 and 2) were closely consistent to those of 10 but suggested the presence of a hexanoyl group at C-2 rather than the propionyl in 10. This assignment was also supported by COSY and HMBC experiments (Figure S94). The absolute configuration of 12 was clarified from its ECD spectrum (Figure 5), which showed a similar shape to that of 11. On the basis of all spectroscopic data, compound 12 (corymbulosin T) was thus characterized as (2R,5S,8R,9R,10S,18R,19S)-18,19-di-O-acetyl-18,19-epoxy-2-hexanoyloxy-cleroda-3,13(16),14-triene.</p><p>Compound 13, optically active colorless oil with [α]25D −28.7 (c 0.07, CHCl3), displayed a peak at m/z 441.2218 [M + Na]+ in the HRFABMS analysis, showing good agreement with a molecular formula of C24H34O6. The 1H and 13C NMR spectra of 13 (Tables 1 and 2) also closely resembled those of 10. The lack of propionyl signals implied the presence of a hydroxy group at C-2, which was confirmed by the COSY correlations of H-2 to H-1 and H-3 (Figure S94). Applying the modified Mosher ester method clarified the absolute configuration at the C-2 position as R (Figure 4). Therefore, compound 13 (corymbulosin U) was established as (2R,5S,8R,9R,10S,18R,19S)-18,19-di-O-acetyl-18,19-epoxy-2-hydroxy-cleroda-3,13(16),14-triene.</p><p>HRFABMS of compound 14 showed a molecular formula C30H44O8 with a sodiated molecular ion at m/z 555.2925 [M + Na]+. The 1H and 13C NMR spectra of 14 (Tables 1 and 2) were mostly identical to those of corymbulosin D17 except for the absence of two methylene groups in the ester at C-6, which was confirmed by HMBC and COSY data (Figure S94). The absolute configuration of 14 was determined by comparisons of 2D NMR data (Figures S94 and S95) and experimental ECD spectra (Figure 5). Accordingly, this compound (corymbulosin V) was defined as (2R,5S,6S,8R,−9R,10S,18R,19S)-18,19-di-O-acetyl-18,19-epoxy-2-hydroxy-6-hexanoyloxycleroda-3,13(16),14-triene.</p><p>Compound 15 was obtained as an optically active colorless oil, [α]D25 −62.4 (c 0.075, CHCl3). The HRFABMS data supported a molecular formula of C32H46O8 from the peak at m/z 581.3075 [M + Na]+. 1D NMR spectra and MS data indicated the loss of two methylene units from the structure of a 2-oxo-clerodane, corymbulosin H,17 suggesting the presence of an octanoyl substituent at C-6 rather than the decanoyl ester in corymbulosin H. The HMBC and COSY spectra (Figure S94) also supported this conclusion. The NOESY (Figure S95) and ECD experiments (Figure 4) suggested that all stereochemical centers in 15 were identical with those in corymbulosin H. Thus, compound 15 (corymbulosin W) was assigned as (5S,6S,8R,9R,10S,18R,- 19S)-18,19-di-O-acetyl-6-octanoyloxy-18,19-epoxycleroda-3,13-(16),14-trien-2-one.</p><p>The structures of the four known compounds were identified as caseamembrin S (16),9 18,19-di-O-acetyl-18,19-epoxy-6-hydroxy-2-(2′-methylbutanoyloxy)cleroda-3,13-(16),14-triene (17),15 caseamembrin E (18),5 and corymbulosin A (19)16 from comparison of their spectroscopic data with reported values. However, the absolute configurations of 16−19 and even the relative configuration of 19 were not determined.</p><p>Compound 16 was characterized as caseamembrin S with identical HRMS and optical rotation data to the reported values;9 however, the 1H and 13C NMR assignments for H-18, H-19, H-3′, C-7, C-8, C-9, and C-10 deviated slightly from the literature values.9 Careful investigations using COSY, HMBC, HMQC, as well as DEPT methods supported the structure of 16 as caseamembrin S. The revised NMR data are shown in the Supporting Information (Table S1). The absolute configurations (2R, 5S, 6S, 8R, 9R, 10S, 18R, 19S) in 16 were established convincingly by ECD (Figure 5) and the comparison of optical rotation and NMR data with the other compounds, including 6.</p><p>Compound 17 was assigned as the reported compound 18,19-di-O-acetyl-18,19-epoxy-6-hydroxy-2-(2′-methyl-butanoyloxy)cleroda-3,13(16),14-triene. The absolute configuration of 17 was first determined by the modified Mosher method and X-ray crystal structural analysis. After MTPA esterification of the C-6 alcohol, the distribution of calculated Δ δH (S−R) values indicated the S-form at C-6 (Figure 4); therefore, the absolute configuration of 17 was established as 2R, 5S, 6S, 8R, 9R, 10S, 18R, and 19S. The S configuration of the C-2′ stereocenter was further determined by X-ray crystal structural analysis as shown in Figure 6 (CCDC 1529690, Figure S94).20</p><p>Similarly, compound 18 was identified as caseamembrin E from the agreement of optical rotation and various spectroscopic data with the reported values, except for revised 1H NMR assignments as shown in the Supporting Information. Its absolute configuration was assigned as 2S, 5S, 6S, 8R, 9R, 10S, 18R, and 19S from the ECD experiment.</p><p>The structure elucidation of 19 was reported previously;16 however, no stereochemistry, even relative configuration, was determined. The HRMS, 1H/13C NMR spectra, and optical activity matched well with the reported values. In the 2D NOESY experiment (Figure 7), correlations between H-2/H-10, H-6/H-8, H-10/H-12, H-11/H-19, H-7/H-19, and H-18/H-19 were observed. From the comparison of the ECD spectrum (Figure 5) and optical rotation values with those of 2 and 18, the absolute configuration of 19 was concluded as 2S, 5S, 6S, 8R, 9R, 10S, 18R, and 19S.</p><p>The antiproliferative activities of the isolated compounds were evaluated using five human tumor cell lines, A549 (lung carcinoma), MCF-7 (estrogen receptor-positive and HER2-negative breast cancer), MDA-MB-231 (triple negative breast cancer), KB (originally isolated from epidermoid carcinoma of the nasopharynx), and KB-VIN (a KB-subline showing MDR phenotype with overexpression of P-gp) (Table 3). Except for compound 3 with a 2-isobutanoyloxy ester, the compounds with a 6-methoxy group generally exhibited lower antiproliferative activity than compounds with a 6-hydroxy group [compare 4 (6-OMe) vs 16 (6-OH) and 6 (6-OMe) vs 17 (6-OH)]. The stereochemistry at C-2 seemed to affect the activity against A549 and KB-VIN (17 vs 18). The related compounds isolated and evaluated in our prior study17 generally exhibited similar potencies to 4 and 6 in the present study. The substitution patterns in the prior corymbulosins17 included a C-2 hydroxy group and various C-6 ester moieties, such as octanoyl-, decanoyl-, and dodecanoyl-oxy. On the basis of the present data, the combination of C-2 ester and C-6 hydroxy groups, such as in 1, 16, and 17, led to more potent antiproliferative activity against human tumor cell lines than the combination of C-2 hydroxy and C-6 ester groups.</p><p>The effects of the potent compounds 1, 3, 16, 17 and 19 on cell cycle progression in the triple-negative breast cancer (TNBC) cell line MDA-MB-231 were further investigated by flow cytometry (Figure 8). All compounds showed no effect on cell cycle progression, while significant accumulations of sub-G1 cells were observed in a dose-dependent manner, except with compound 19. When cells were treated with higher concentrations of compound 1, 16, or 17, propidium iodide (PI) signals decreased unusually and dramatically in all cells. These observations suggested that the test compounds efficiently induced either cytolysis or apoptosis. Occasionally, the cytolytic activity of cytotoxic compounds can induce hemolysis in vivo. Accordingly, we tested the hemolytic activity of selected compounds using horse's blood (Table 4). Among them, compound 19 with a long aliphatic chain at C-2 showed the most potent activity but only at the unusually high concentration of 500 μM, suggesting that all tested compounds have no hemolytic activity.</p><p>The effects of compounds on cell morphology by using immunocytochemistry were next evaluated. MDA-MB-231 cells were treated with compounds for 24 h followed by staining with antibody to α-tubulin for microtubules and 4′,6-diamidino-2-phenylindole (DAPI) for chromatin (Figure 9). Cells treated with compound 3 at a lower concentration (1.5 μM) showed normal morphology with intact nuclei and a clear microtubule network, as also found in the control cells. However, most cells treated with compound 3 at 4.9 μM exhibited fragmented and dispersed chromatin (arrowhead) with an apoptotic microtubule array (arrow)21 (Figure 9A). These observations implied that compound 3 induced apoptosis in a dose-dependent manner. Similar phenotypes were observed in cells treated with compounds 1, 16, and 17. These data were quite consistent with the results from flow cytometric analysis. In conclusion, newly isolated corymbulosins efficiently induce chromatin fragmentation with formation of apoptotic microtubules in TNBC MDA-MB-231 cells.</p><!><p>Fifteen new clerodane diterpenes, designated corymbulosins I−W (1−15), together with four known diterpenes, 16−19, were isolated from a crude MeOH/CH2Cl2 (1:1) extract (N005829) of the bark of L. corymbulosa. The structures of 1−15 were elucidated on the basis of extensive 1D and 2D NMR and HRMS analyses. The absolute configurations of newly isolated compounds 1−15 as well as known 16−19, which were reported previously with only relative configurations, were determined through ECD experiments, X-ray analysis, modified Mosher ester method, and comparison of their spectroscopic data. The isolated compounds were evaluated for antiproliferative activity against human cancer cell lines. Among all, compounds 1, 3, 16, 18, and 19 showed potent activity against all tested human tumor cell lines, including MDR subline, with IC50 values of 0.4−1.0 μM. A structure−activity relationship study revealed that the combination of C-2 ester and C-6 hydroxy groups led to potent antiproliferative activity. Flow cytometric and immunocytochemical observations of cells treated with cytotoxic clerodanes demonstrated that the chromatin was fragmented and dispersed with formation of apoptotic microtubules.</p><!><p>Optical rotations were measured on a JASCO P-2200 digital polarimeter. CD spectra were recorded on JASCO J-820 spectrometer. Infrared spectra (IR) were obtained with a Thermo Fisher Scientific NICOLET iS5 FT-TR spectrometer from samples in CH2Cl2. NMR spectra were measured on JEOL JMN-ECA600 and JMN-ECS400 spectrometers with tetramethylsilane as an internal standard, and chemical shifts are stated as δ values. HRMS data were recorded on a JMS-700 MStation quadrupole (FAB) or JMS-T100TD TOF (DART) mass spectrometer. A crystal of the compound was measured on a R-AXIS RAPID II (Rigaku). Analytical and preparative TLC were carried out on precoated silica gel 60F254 and RP-18F254 plates (0.25 or 0.50 mm thickness; Merck). MPLC was performed on Combiflash Rf (Teledyne Isco) with silica gel and C18 cartridges (Biotage, Uppsala Sweden). Preparative HPLC was operated on a GL Science recycling system using an InertSustain C18 column (5 μm, 20 × 250 mm).</p><!><p>The crude MeOH/CH2Cl2 (1:1) extract (N005829) of bark of L. corymbulosa collected in Peru was provided by NCI/NIH (Frederick, MD, U.S). A voucher specimen (voucher no. QT65T0390) was deposited at the Smithsonian Institution (Washington, DC, and voucher extracts were deposited at the NCI (Frederick, MD) and Kanazawa University (Kanazawa, Japan). The crude organic extract of N005829 was evaluated for cytotoxicity by NCI with an in vitro 60-cell tumor screening panel as reported previously.17,22</p><!><p>The crude extract N005829 (12.8 g) was partitioned between H2O and EtOAc to obtain H2O-soluble (2.2 g) and EtOAc-soluble portions (7.4 g). The latter portion was subjected to silica gel MPLC (RediSep Rf GOLD High Performance 120 g) with a gradient system [n-hexane−EtOAc 90:10 (600 mL) → 80:20 (40 mL) → 75:25 (120 mL) → 65:35 (80 mL) → 60:40 (120 mL) → 55:45 (80 mL) → 45:55 (240 mL) → 30:70 (520 mL) → 10:90 (440 mL) → EtOAc−MeOH 90:10 (120 mL) → MeOH (1400 mL)] to yield 11 fractions, F1−F11.17</p><p>F5 (1249.9 mg) was further fractionated by silica gel column chromatography (CC) eluted with n-hexane/EtOAc (5:1 to 0:1) to afford 12 subfractions 5a−l. Subfraction 5g (384.9 mg) was purified by MPLC on ODS-25 (YMC-DispoPack AT 12 g) with H2O/ MeOH (1:6), followed by repeated recycling preparative HPLC with H2O/MeOH (1:7) to yield compound 12 (2.0 mg). Subfraction 5h (279.1 mg) was purified by repeated recycling preparative HPLC with H2O/MeOH (1:9) to provide compounds 10 (2.0 mg), 11 (9.0 mg), and 15 (1.5 mg). Subfraction 5i (98.0 mg) was purified by repeated recycling preparative HPLC with H2O/MeOH (1:9) to afford compounds 7 (1.5 mg) and 9 (7.1 mg). Subfraction 5j (183.2 mg) was purified by repeated recycling preparative HPLC with H2O/MeOH (1:9) to provide compounds 10 (2.0 mg), 11 (9.0 mg), and 15 (1.5 mg). Subfraction 5i (98.0 mg) was purified by repeated recycling preparative HPLC with H2O/MeOH (1:9) followed by preparative TLC using n-hexane/EtOAc (4:1) to yield compound 8 (3.3 mg).</p><p>F6 (430.8 mg) was subjected to silica gel CC eluted with n-hexane/CH2Cl2 (1:1 to 0:1), CH2Cl2/EtOAc (19:1), EtOAc, followed by MeOH to yield eight subfractions 6a−h. Subfraction 6f (125.9 mg) was subjected to silica gel CC eluted with n-hexane/ CH2Cl2 (1:0 to 57:4) followed by MeOH to obtain seven subfractions, 6f1−7. Subfraction 6f6 (15.5 mg) was purified by ODS preparative TLC developing three times using H2O/MeOH (1:4 × 2 and 1:6) to afford compound 6 (5.7 mg). Subfraction 6g (179.5 mg) was subjected to silica gel CC eluted with n-hexane/ EtOAc (1:0 to 0:1) to obtain six subfractions 6g1−6. Subfraction 6g3 (78.0 mg) was purified by repeated recycling preparative HPLC with H2O/MeOH (1:4) to provide compounds 6 (1.7 mg), 18 (16.9 mg), and 19 (8.0 mg, 0.00062%).</p><p>F7 (3.52 g) was subjected to silica gel CC eluted with CH2Cl2/EtOAc (6:1 to 0:1) followed by MeOH to yield 10 subfractions 7a− j. Subfraction 7d (155.9 mg) was subjected to silica gel CC eluted with CH2Cl2/EtOAc (1:0 to 4:1) followed by MeOH to obtain seven subfractions, 7d1−7. Subfraction 7d4 (60.0 mg) was purified by MPLC on ODS-25 (YMC-DispoPack AT 12 g) with H2O/ MeOH (1:4), followed by repeated recycling preparative HPLC with H2O/MeOH (1:4) to provide compounds 3 (5.1 mg) and 4 (1.0 mg).</p><p>Subfraction 7e (603.7 mg) was subjected to silica gel CC eluted with n-hexane/CH2Cl2 (1:1 to 0:1) followed by CH2Cl2/EtOAc (2:1) to obtain 10 subfractions, 7e1-10. Subfraction 7e6 (385.9 mg) was purified by MPLC on ODS-25 (YMC-DispoPack AT 12 g) with H2O/MeOH (3:7 to 1:6), followed by recycle preparative HPLC with H2O/MeOH (1:6) to afford compounds 4 (0.7 mg) and 5 (0.4 mg).</p><p>Subfraction 7f (830.1 mg) was subjected to silica gel CC eluted with CH2Cl2/EtOAc (1:0 to 0:1) followed by MeOH to obtain six subfractions, 7f1−6. Subfraction 7f3 (107.0 mg) was purified by MPLC on ODS-25 (YMC-DispoPack AT 12 g) with H2O/MeOH (1:4) followed by recycle preparative HPLC with H2O/MeOH (1:4) to afford compounds 2 (1.2 mg) and 18 (10.4 mg), which was combined with 18 (16.9 mg) obtained from F6 (total 27.3 mg, 0.00213%).</p><p>n-Hexane/EtOAc (3:1) was added to subfraction 7h (192.6 mg), and the resulting insoluble material (75.7 mg) was purified by repeated recycle preparative HPLC with H2O/MeOH (1:4) to afford compounds 1 (16.9 mg), 7 (2.3 mg), and 17 (43.5 mg, 0.0033%).</p><p>F8 was subjected to silica gel CC eluted with CH2Cl2/EtOAc (1:0 to 0:1) followed by MeOH to give 14 subfractions, 8a−n. Compounds 13 (1.5 mg), and 14 (11.9 mg) were obtained by repeated recycling preparative HPLC of subfraction 8f (169.1 mg) with H2O/MeOH (1:5).</p><!><p>16.9 mg, 0.0013%; colorless amorphous solid [α]D25 +2.3 (c 0.38, CHCl3); IR νmax (CH2Cl2) cm−1 3500, 2970, 2932, 2879, 1754, 1731, 1596, 1469, 1373; 1H and 13C NMR, Tables 1 and 2; HRFABMS m/z 527.2631 [M + Na]+ (calcd for C28H40O8Na, 527.2621).</p><!><p>1.2 mg, 0.00007%; colorless amorphous solid; [α]D25 −88.0 (c 0.01, MeOH); IR νmax (CH2Cl2) cm−1 3463, 2957, 2923, 2853, 1730, 1595, 1458, 1376; 1H and 13C NMR, Tables 1 and 2; HRFABMS m/z 527.2624 [M + Na]+ (calcd for C28H40O8Na, 527.2621).</p><!><p>5.1 mg, 0.00039% ; colorless oil; [α]D25 +10.4 (c 0.23, CHCl3); IR νmax (CH2Cl2) cm−1 2970, 2936, 2879, 1755, 1731, 1596, 1469, 1373, 1227; 1H and 13C NMR, Tables 1 and 2; HRFABMS m/z 541.2777 [M + Na]+ (calcd for C29H42O8Na, 541.2777).</p><!><p>1.0 mg, 0.00007%; colorless oil; [α]D25 +12.3 (c 0.05, CHCl3); IR νmax (CH2Cl2) cm−1 2961, 2927, 2880, 1755, 1732, 1456, 1373, 1224; 1H and 13C NMR, Tables 1 and 2; HRFABMS m/z 527.2608 [M + Na]+ (calcd for C28H40O8Na, 527.2621).</p><!><p>0.4 mg, 0.00003%; colorless oil; [α]D25 +28.4 (c 0.02, CHCl3); IR νmax (CH2Cl2) cm−1 2926, 1740, 1373, 1224; 1H and 13C NMR, Tables 1 and 2; HRFABMS m/z 513.2479 [M + Na]+ (calcd for C27H38O8Na, 513.2464).</p><!><p>7.4 mg, 0.00057%; colorless oil; [α]D25 +8.4 (c 0.20, CHCl3); IR νmax (CH2Cl2) cm−1 2966, 2932, 1755, 1729, 1337, 1227; 1H and 13C NMR, Tables 1 and 2; HRFABMS m/z 555.2947 [M + Na]+ (calcd for C30H44O8Na, 555.2934).</p><!><p>3.8 mg, 0.00029%; colorless oil; [α]D25 −64.0 (c 0.07, MeOH); IR νmax (CH2Cl2) cm−1 2955, 2926, 2881, 1756, 1730, 1373, 1226; 1H and 13C NMR, Tables 1 and 2; HRFABMS m/z 555.2921 [M + Na]+ (calcd for C30H44O8Na, 555.2934).</p><!><p>3.3 mg, 0.00025%; colorless oil; [α]D25 +0.9 (c 0.20, CHCl3); IR νmax (CH2Cl2) cm−1 2964, 2928, 2860, 1751, 1733, 1373, 1224; 1H and 13C NMR, Tables 1 and 2; HRFABMS m/z 625.3340 [M + Na]+ (calcd for C34H50O9Na, 625.3353).</p><!><p>7.1 mg, 0.00055%; colorless oil; [α]D25 +19.6 (c 0.15, CHCl3); IR νmax (CH2Cl2) cm−1 2955, 2926, 2854, 1738 1373, 1220; 1H and 13C NMR, Tables 1 and 2; HRFABMS m/z 653.3651 [M + Na]+ (calcd for C36H54O9Na, 653.3666).</p><!><p>2.0 mg, 0.00015%, colorless oil; [α]D25 −27.1 (c 0.08, CHCl3); IR νmax (CH2Cl2) cm−1 2958, 2926, 2855, 1755, 1734, 1374, 1226; 1H and 13C NMR, Tables 1 and 2; HRFABMS m/z 497.2511 [M + Na]+ (calcd for C27H38O7Na, 497.2515). Corymbulosin S (11).</p><!><p>9.0 mg, 0.0007%; colorless oil; [α]25D −10.0 (c 0.45, CHCl3); IR νmax (CH2Cl2) cm−1 2967, 2941, 2870, 1749, 1730, 1373, 1226; 1H and 13C NMR, Tables 1 and 2; HRFABMS m/z 511.2681 [M + Na]+ (calcd for C28H40O7Na, 511.2672).</p><!><p>2.0 mg, 0.00015%; colorless oil; [α]D25 +0.7 (c 0.35, CHCl3); IR νmax (CH2Cl2) cm−1 2961, 2935, 2870, 1757, 1733, 1374, 1224; 1H and 13C NMR, Tables 1 and 2; HRFABMS m/z 539.2960 [M + Na]+ (calcd for C30H44O7Na, 539.2985).</p><!><p>1.5 mg, 0.00011%; colorless oil;[α]D25 −28.7 (c 0.07, CHCl3); IR νmax (CH2Cl2) cm−1 3473, 2955, 2933, 1750, 1374, 1227; 1H and 13C NMR, Tables 1 and 2; HRFABMS m/z 441.2218 [M + Na]+ (calcd for C24H34O6Na, 441.2253).</p><!><p>11.9 mg, 0.00092%; colorless oil;[α]D25 −18.8 (c 0.28, CHCl3); IR νmax (CH2Cl2) cm−1 2961, 2930, 1735, 1374, 1225; 1H and 13C NMR, Tables 1 and 2; HRFABMS m/z 555.2925 [M + Na]+ (calcd for C30H44O8Na, 555.2934).</p><!><p>1.5 mg, 0.00011%; colorless oil; [α]D25 −62.4 (c 0.28, CHCl3); IR νmax (CH2Cl2) cm−1 2961, 2927, 2855, 1761, 1374, 1220; 1H and 13C NMR, Tables 1 and 2; HRFABMS m/z 581.3075 [M + Na]+ (calcd for C32H46O8Na, 581.3090).</p><!><p>To a solution of 1 (0.7 mg, 1.4 μmol) in anhydrous CH2Cl2 (0.25 mL) were added Et3N (6.9 μL, 49.5 μmol), DMAP (3.8 mg, 31.1 μmol), and (S)-MTPACl (6.9 μL, 36.9 μmol). The mixture was stirred at rt for 3.5 h, followed by direct purification using preparative TLC with CH2Cl2 to afford the (R)-MTPA ester (0.9 mg, 90%). The corresponding (S)-MTPA ester (1.2 mg, 93%) was obtained by the same procedure using (R)-MTPACl.</p><p>(R)-MTPA ester of 1: 90% yield. 1H NMR (CHCl3, 600 MHz) δH 7.51−7.37 (5H, m, aromatic protons), 6.52 (1H, s, H-19), 6.43 (1H, dd, J = 18.0, 11.4 Hz, H-14), 5.86 (1H, brd, J = 3.6 Hz, H-3), 5.76 (1H, t, J = 1.8 Hz, H-18), 5.43 (1H, m, H-2), 5.17 (1H, d, J = 18.0 Hz, H-15), 5.06 (1H, m overlap, H-6), 5.06 (1H, s, H-16), 5.03 (1H, d, J = 11.4 Hz, H-15), 4.95 (1H, s, H-16), 3.62 (3H, s, OMe), 2.61 (1H, sep, J = 7.2 Hz, H-2′), 2.37 (1H, dd, J = 13.8, 3.6 Hz, H-10), 2.08 (2H, m, H-12), 2.03 (1H, m, H-1β), 1.94 (3H, s, 18-OAc), 1.85 (3H, s, 19-OAc), 1.75 (1H, q, J = 12.6 Hz, H-7α), 1.20 (3H, d, J = 7.2 Hz, H-3′), 1.18 (3H, d, J = 7.2 Hz, H-3′), 0.96 (3H, d, J = 7.2 Hz, H-17), 0.95 (3H, s, H-20).</p><p>(S)-MTPA ester of 1: 93% yield. 1H NMR (CHCl3, 600 MHz) δH 7.53−7.41 (5H, m, aromatic protons), 6.47 (1H, s, H-19), 6.46 (1H, dd, J = 1.8, 1.2 Hz, H-18), 6.43 (1H, dd, J = 18.0, 11.4 Hz, H-14), 5.98 (1H, brd, J = 3.6 Hz, H-3), 5.47 (1H, m, H-2), 5.16 (1H, d, J = 18.0 Hz, H-15), 5.09 (1H, dd, J = 13.2, 4.2 Hz, H-6), 5.05 (1H, s, H-16), 5.03 (1H, d, J = 11.4 Hz, H-15), 4.92 (1H, s, H-16), 3.46 (3H, s, OMe), 2.62 (1H, sep, J = 7.2 Hz, H-2′), 2.40 (1H, dd, J = 13.8, 3.0 Hz, H-10), 2.08 (2H, m, H-12), 2.03 (1H, m, H-1β), 2.02 (3H, s, 18-OAc), 1.87 (3H, s, 19-OAc), 1.80 (1H, dt, J = 13.2, 4.2 Hz, H-7β), 1.65 (1H, q, J = 13.2 Hz, H-7α), 1.22 (3H, d, J = 7.2 Hz, H-3′), 1.19 (3H, d, J = 7.2 Hz, H-3′), 0.94 (3H, s, H-20), 0.93 (3H, d, J = 6.6 Hz, H-17).</p><p>(R)-MTPA ester of 17: 99% yield. 1H NMR (CHCl3, 600 MHz) δH 7.51−7.38 (5H, m, aromatic protons), 6.52 (1H, s, H-19), 6.44 (1H, dd, J = 17.4, 10.8 Hz, H-14), 5.89 (1H, brd, J = 3.6 Hz, H-3), 5.76 (1H, dd, J = 1.8, 1.2 Hz, H-18), 5.43 (1H, m, H-2), 5.16 (1H, d, J = 17.4 Hz, H-15), 5.07 (1H, dd, J = 13.2, 4.2 Hz, H-6), 5.06 (1H, s, H-16), 5.03 (1H, d, J = 10.8 Hz, H-15), 4.95 (1H, s, H-16), 3.62 (3H, s, OMe), 2.44 (1H, m, H-2′), 2.39 (1H, dd, J = 13.8, 3.6 Hz, H-10), 2.08 (2H, m, H-12), 2.03 (1H, m, H-1), 1.93 (3H, s, 18-OAc), 1.84 (3H, s, 19-OAc), 1.75 (1H, q, J = 13.2 Hz, H-7α), 1.16 (3H, d, J = 7.2 Hz, H-5′), 0.96 (3H, d, J = 6.6 Hz, H-17), 0.95 (3H, s, H-20), 0.94 (3H, t, J = 7.2 Hz, H-4′).</p><p>(R)-MTPA ester of 13: 40% yield. 1H NMR (CHCl3, 600 MHz) δH 7.56−7.44 (5H, m, aromatic protons), 6.72 (1H, t, J = 1.7 Hz, H-18), 6.36 (1H, dd, J = 17.2, 10.6 Hz, H-14), 6.30 (1H, s, H-19), 6.02 (1H, brd, J = 4.5 Hz, H-3), 5.64 (1H, m, H-2), 5.09 (1H, d, J = 17.2 Hz, H-15), 4.99 (1H, s, H-16), 4.97 (1H, d, J = 10.6 Hz, H-15), 4.88 (1H, s, H-16), 2.33 (1H, dd, J = 14.0, 2.6 Hz, H-10), 2.05 (3H, s, 18-OAc), 2.00 (2H, m, H-12), 1.75 (1H, m, H-6), 1.73 (3H, s, 19-OAc), 1.46 (1H, m, H-6), 0.91 (3H, s, H-20), 0.88 (3H, d, J = 6.7 Hz, H-17).</p><p>(S)-MTPA ester of 13: 28% yield. 1H NMR (CHCl3, 600 MHz) δH 7.51−7.37 (5H, m, aromatic protons), 6.69 (1H, t, J = 1.7 Hz, H-18), 6.39 (1H, dd, J = 17.5, 10.8 Hz, H-14), 6.28 (1H, s, H-19), 6.02 (1H, brd, J = 3.7 Hz, H-3), 5.61 (1H, m, H-2), 5.08 (1H, d, J = 17.5 Hz, H-15), 5.01 (1H, s, H-16), 4.98 (1H, d, J = 10.8 Hz, H-15), 4.90 (1H, s, H-16), 2.31 (1H, d, J = 12.0 Hz, H-10), 2.04 (3H, s, 18-OAc), 2.02 (2H, m, H-12), 1.78 (3H, s, 19-OAc), 1.74 (1H, m, H-6), 1.47 (1H, m, H-6), 0.94 (3H, s, H-20), 0.89 (3H, d, J = 6.7 Hz, H-17).</p><p>(S)-MTPA ester of 17: 71% yield. 1H NMR (CHCl3, 600 MHz) δH 7.53−7.42 (5H, m, aromatic protons), 6.47 (1H, s, H-19), 6.46 (1H, t, J = 1.2 Hz, H-18), 6.43 (1H, dd, J = 17.4, 10.8 Hz, H-14), 6.00 (1H, brd, J = 4.2 Hz, H-3), 5.46 (1H, m, H-2), 5.16 (1H, d, J = 17.4 Hz, H-15), 5.10 (1H, dd, J = 12.0, 4.2 Hz, H-6), 5.05 (1H, s, H-16), 5.03 (1H, d, J = 10.8 Hz, H-15), 4.93 (1H, s, H-16), 3.46 (3H, s, OMe), 2.46 (1H, m, H-2′), 2.42 (1H, dd, J = 13.2, 4.2 Hz, H-10), 2.08 (2H, m, H-12), 2.01 (3H, s, 18-OAc), 1.91 (1H, m, H-8), 1.86 (3H, s, 19-OAc), 1.81 (1H, dt, J = 13.2, 4.2 Hz, H-7β), 1.18 (3H, d, J = 6.6 Hz, H-5′), 0.96 (3H, t, J = 7.2 Hz, H-4′), 0.94 (3H, s, H-20), 0.93 (3H, d, J = 6.6 Hz, H-17).</p><!><p>Antiproliferative activity of the compounds was determined by the sulforhodamine B (SRB) assay as described previously.22,23</p><!><p>Cell cycle distribution was analyzed by measurement of cellular DNA content by staining with propidium iodide (PI)/RNase (BD Biosciences) as described previously.24 Stained cells were analyzed by flow cytometry (BD LSRFortessa).</p><!><p>The assay was performed using the method reported by Vo et al.25 with some modifications. Preserved horse's blood (defibrinated) was purchased from Cosmo Bio (Japan). After the blood was centrifuged for 5 min at 232g, supernatant was discarded. Hemocytes-enriched pellets were washed three times with PBS and suspended in saline at 0.5% (v/v) concentration. A 250 μL amount of the hemocytes was mixed with a 250 μL of test compound. After 30 min incubation at 37 °C, supernatant was recovered from the mixture by centrifugation for 10 min at 232g. The absorbance of the resultant supernatant was measured at 570 nm using a microplate reader (SPARK 10M, Tecan, Switzerland). Saponin Quillaja sp. (S4521, SIGMA) was used as a positive control, for which PBS containing 50%, 20%, or 4% DMSO was used as a negative control. The percentage of hemolysis induced by compound was calculated as (absorbance of test sample − absorbance of negative control)/(absorbance of positive control − absorbance of negative control) × 100%. Experiments were conducted three times for each compound at each concentration.</p><!><p>MDA-MB-231 cells were seeded in the 8-well chamber slide (Lab-Tech) 24 h prior to treatment with compounds. After 24 h treatment, cells were fixed with 4% paraformaldehyde in PBS followed by permeabilization with 0.5% Triton-X100. Cells were then probed with antibody to α-tubulin (B5–1-2, Sigma) followed by FITC-conjugated antibody to mouse IgG (Sigma). Nuclei were labeled with 4′,6-diamidino-2-phenylindole (DAPI) (Sigma). Microtubules were detected using a confocal microscope (Zeiss, LSM700) controlled by ZEN software (Zeiss). Final images were reconstructed from 12−18 sections acquired at 0.6−1 μm intervals and merged using ZEN (black edition) software. Experiments were repeated at least twice for each compound at each concentration. Final images were prepared using Adobe Photoshop CS6.</p>

PubMed Author Manuscript

Predicting a small molecule-kinase interaction map: A machine learning approach

BackgroundWe present a machine learning approach to the problem of protein ligand interaction prediction. We focus on a set of binding data obtained from 113 different protein kinases and 20 inhibitors. It was attained through ATP site-dependent binding competition assays and constitutes the first available dataset of this kind. We extract information about the investigated molecules from various data sources to obtain an informative set of features.ResultsA Support Vector Machine (SVM) as well as a decision tree algorithm (C5/See5) is used to learn models based on the available features which in turn can be used for the classification of new kinase-inhibitor pair test instances. We evaluate our approach using different feature sets and parameter settings for the employed classifiers. Moreover, the paper introduces a new way of evaluating predictions in such a setting, where different amounts of information about the binding partners can be assumed to be available for training. Results on an external test set are also provided.ConclusionsIn most of the cases, the presented approach clearly outperforms the baseline methods used for comparison. Experimental results indicate that the applied machine learning methods are able to detect a signal in the data and predict binding affinity to some extent. For SVMs, the binding prediction can be improved significantly by using features that describe the active site of a kinase. For C5, besides diversity in the feature set, alignment scores of conserved regions turned out to be very useful.

predicting_a_small_molecule-kinase_interaction_map:_a_machine_learning_approach

Background<!><!>Background<!>Data<!><!>Data<!>General approach<!>Feature generation<!>Feature generation for kinases<!><!>Feature generation for inhibitors<!><!>Feature generation for inhibitors<!>Feature selection/reduction<!>Classification<!>Evaluation<!>The hard and the soft case<!><!>The mixed and the mixed-mixed case<!><!>Results for different feature sets<!><!>Results for different feature sets<!><!>Results for different feature sets<!><!>Results for different feature sets<!>Comparison of different kernels in SVMs<!>Performance on random feature sets<!><!>Results for the hard and the soft case evaluation strategy<!><!>Results for the hard and the soft case evaluation strategy<!><!>Results for the mixed and the mixed-mixed case<!><!>Results for the mixed and the mixed-mixed case<!><!>Results for the mixed and the mixed-mixed case<!><!>Results for the mixed and the mixed-mixed case<!>Results for an external test set<!><!>Related Work<!>Conclusion<!>Competing interests<!>Authors' contributions<!>Authors' information<!>Additional file 1<!><!>Acknowledgements

<p>The question whether two molecules (a protein and a small molecule) can interact can be addressed in several ways. On the experimental side, different kinds of assays [1] or crystallography are applied routinely. Target-ligand interaction is an important topic in the field of biochemistry and related disciplines. However, the use of experimental methods to screen databases containing millions of small molecules [2] that could match with a target protein, for instance, is often very time-consuming, expensive and error-prone due to experimental errors. Computational techniques may provide a means for speeding up this process and making it more efficient. In particular in the area of kinases, however, docking methods have been shown to have difficulties so far [3] (Apostolakis J: Personal communication, 2008). In this paper, we address the task of interaction prediction as a data mining problem in which crucial binding properties and features responsible for interactions have to be identified. Note that this paper is written in a machine learning context, hence we use the term "prediction" instead of "retrospective prediction" that would be used in a biomedical context.</p><p>In the following, we focus on protein kinases and kinase inhibitors. Protein kinases have key functions in the metabolism, signal transmission, cell growth and differentiation. Since they are directly linked to many diseases like cancer or inflammation, they constitute a first-class subject for the research community. Inhibitors are mostly small molecules that have the potential to block or slow down enzyme reactions and can therefore act as a drug. In this study we have 20 different inhibitors with partially very heterogeneous structures (see Figure 1).</p><!><p>Training set inhibitors. Structures of the 20 inhibitors that were subject of our study [7].</p><!><p>We developed a new computational approach to solve the protein-ligand binding prediction problem using machine learning and data mining methods, which are easier and faster to perform than experimental techniques from biochemistry and have proven successful for similar tasks [4-6]. In summary, the contributions of this paper are as follows: First, it uses both kinase and kinase inhibitor descriptors at the same time to address the interaction between small heterogeneous molecules and kinases from different families from a machine learning point of view. Second, it proposes a new evaluation scheme that takes into account various amounts of information known about the binding partners. Third, it provides insight into features that are particularly important to achieve a certain level of performance.</p><p>This paper is organized as follows: In the following sections, we first present the methods and datasets we used, then we give a detailed description of variants of leave-one-out cross-validation to measure the quality of predictions, present the experimental results and finally draw our conclusions.</p><!><p>This section introduces the Ambit Biosciences' dataset [7] that provides us with class information for our classification task. From the dataset we define a two class problem by assigning to each kinase inhibitor pair "binding" or "no binding" according to the measured affinities of interaction read out by quantitative PCR. This dataset is obtained by ATP site-dependent competition binding assays and represents the first approach to mass screening of protein kinases and inhibitors. Table 1 shows overview statistics concerning the size and the class distribution of the dataset. Table S1 in Additional File 1 shows how often an inhibitor binds to a certain group of kinases (group in a phylogenetic meaning). It can be clearly seen that nearly all inhibitors bind to several kinase groups. This means that there is generally no kinase group to which an inhibitor binds consistently. The kinase and inhibitor data, its corresponding features, and the binding matrix is available on the web [8].</p><!><p>Characteristics of the used dataset</p><!><p>For assessing the specificity of protein kinases for inhibitors, Ambit applied ATP site-dependent competition binding assays that directly and quantitatively measure the binding of inhibitors to the ATP binding site of kinases.</p><!><p>According to the SAR (Structure Activity Relationship) paradigm that activity is related to structure, we put the focus on features that describe the structure of molecules, that are leading to certain structures (sequence-based features) or that determine the chemical environment of the active site of a kinase or the molecule as a whole with respect to the inhibitors. Features from further categories may also give hints whether an interaction can take place, like information about the similarity of molecules, e.g., alignment scores or other phylogenetic features.</p><!><p>In the following, we will describe the features chosen to describe the kinases and inhibitors.</p><!><p>One class of features for the kinases represent sequence-based features. These features are derived from consecutive patterns of single amino acids. However, only frequent patterns are regarded as interesting, since only those can be informative for prediction. Active sites, for instance, are usually highly conserved and of special interest since inhibitors bind to that region. We scanned the PROSITE database (release 51.2) for PROSITE patterns that match our protein kinase sequences. These patterns are characteristic clusters of residue types occurring over a rather short section of a protein sequence. For the generation of further frequent patterns we implemented Agrawal and Srikant's APriori algorithm [9] with minor modifications. During the levelwise search [10] that enables us to find all frequent patterns, we count per example: multiple occurrences within a kinase are only counted once. As a refinement operator to generate patterns for the next level, we used pattern merging. In a pattern we allow wildcards, however, their number is retricted to two in order to reduce the search space. We excluded wildcards at the beginning and at the end of patterns, since they do not carry any significant biological information. The sequence-based features are represented in bitvectors that indicate whether a sequence is present or absent in a kinase.</p><p>Each position in a bitvector corresponds to a sequence where "1" indicates presence and "0" absence. Another class of features are phylogeny-based features. We extract available phylogenetic information about the kinases from KinBase [11] since a closer phylogenetic relationship implies a higher sequence similarity and thus also a higher similarity in the overall 3D structure - especially at conserved sites like the active center. We grouped the kinases into Serine/Threonine and Tyrosine kinases and made a finer division into kinase groups and kinase families. The phylogeny-based information is presented in nominal form, e.g. the phylogenetic feature "group" has several categories like AGC, CAMK or CK1.</p><p>Other phylogenetic features are directly derived from kinase sequence alignments. We implemented three different types of alignment procedures. First, a global alignment algorithm [12] was used that aligns two amino acid sequences over their full length. The protein kinases we investigated differ enormously in length, reaching from 275 to 1,607 amino acids, and therefore many gaps requiring to be introduced. This may obscure existing evolutionary relationships making therefore the alignment scores less useful. To overcome this problem, first we applied CLUSTALW [13] to get a multiple sequence alignment (MSA), from which we selected highly conserved sequence stretches (frames) where each frame must satisfy the following three criteria:</p><p>- At least one position in a frame must be highly conserved. A highly conserved position is a residue in which one amino acid occurs in more than 100 out of the 113 cases.</p><p>- The frame border is at most five amino acids away from a highly conserved position.</p><p>- A highly conserved position must be part of the active center.</p><p>For each kinase pair, each frame is matched with the corresponding frame from the second sequence and scored according to the scoring matrices (see below). The scores for the frame pairs are summed up to get an overall score for all conserved regions, whereas a higher score should indicate a more similar active center and more similar binding properties. To calculate the scores, we implemented two different techniques. First, we just cut out the amino acid sequences of the frames from the MSA and scored it without any further modification. Second, we realigned the cut out frames before calculating the total score. For all alignment procedures we used PAM120 and Blosum62 as substitution matrices with uniform costs for gap opening and gap extension. The alignment-based features are represented with a 113-dimension vector where each dimension or postition represents a numeric alignment score.</p><p>Additionally, we also use single residues which contribute to the active site and inhibitor binding as position specific features. These features' values are either the respective amino acid or the physico-chemical class of the amino acid. In Table 2, the features, the number of features, and the feature type in each group used in our study for describing the kinases are summarized.</p><!><p>Summary of different features of kinases used in our study</p><!><p>For the description of the inhibitors we used features based on their 2D structures, preferred binding partners (primary targets) and binding patterns. We visually clustered the inhibitors by simply looking at their 2D structure so that inhibitors with similar shapes are grouped together (see Table 3). Primary targets are kinases for which an inhibitor shows a highly preferred binding compared to other kinases. In this context, "primary target" concerns kinases in general and is not restricted to the kinases under consideration in this paper. Binding patterns represent the binding behavior of an inhibitor to a set of kinases and may serve also as features since similar properties on known targets give hints to binding properties on unknown targets. Therefore, we implemented a k-nearest-neighbor method (KNN) to detect each inhibitor's k nearest neighbors. In this study we used k = 3. The calculation is based on data from Ambit's binding matrix. Note that this calculation is only possible in the "soft case" evaluation (to be presented below) since only in that case all the information of a test kinase (respectively inhibitor) relative to all training inhibitors (respectively kinases) is given. As a distance measure, we used a function counting the number of common bindings of two inhibitors xi, xj (c), and the more complex Tanimoto coefficient (1) that counts besides (c) the number of bindings of inhibitor xi to a kinase (a) and the number of bindings of inhibitor xj to the same kinase (b):(1)</p><!><p>Summary of different features for the inhibitors used in our study</p><!><p>To describe the inhibitor structures, we applied the graph mining tool Free Tree Miner (FTM) [14]. With this tool, the 2D structure representations of the inhibitors are mined for frequently occurring acyclic substructures. Such substructures can describe, for instance, a hydrophobic group in an inhibitor important for bindings or an extended region that would exclude small active sites of kinases as binding partners due to steric hindrance. To avoid the exclusion of probably important substructures right from the beginning, we set the minimum frequency threshold rather low to 10%.</p><p>Additionally, we also calculated geometric features of the inhibitors from 2D data like their diameter, length and width that might prevent kinase binding due to steric hindrance.</p><p>Besides this, various chemical features determine whether or not a binding at the active site can take place. For the calculation of such features, we applied the cheminformatics library JOELib2 [15]. In this way, we obtained the following physicochemical features: XlogP, molecular weight, hydrogen bond acceptor/donor count, rotatable bond count, tautomer count and topological polar surface area. All these features are suitable for building basic structure-activity relationship (SAR) models [16]. We also described the inhibitors with pharmacophores. A pharmacophore is, in general, a 3D substructure of a molecule that is meaningful for its medical activity. It can be seen as an abstraction of the molecular structure to a usually small number of key features that contribute to the majority of the activity together with their geometric arrangement that is represented by pairwise distances. For the actual calculation of the pharmacophores we, only for simplicity, used the 2D information of the inhibitors. We calculated so-called 3-point pharmacophores [17] for our set of inhibitors. Such pharmacophores consist of three essential atoms (negatively or positively charged atoms, acceptor or donor atoms) and their distances in space. We calculated all 3-point pharmacophores, sorted the atoms lexicographically in order to avoid duplicates, and used the atoms as well as their (discretized) distances as features. Table 3 summarizes the features, the number of features and the feature type in each group that we used for describing the inhibitors.</p><p>The instances consisting of kinase-inhibitor pairs are represented by concatenating kinase and inhibitor feature vectors, i.e. each kinase is concatenated with each inhibitor. Formally, this can be stated as(2)</p><p>where Ki represents the feature vector of the ith kinase and Ij the feature vector of the jth inhibitor.</p><!><p>Feature selection techniques attempt to determine appropriate features that can discriminate well between classes. Feature sets that are too larger may contain many uninformative features leading to overfitting or a decrease in prediction accuracy or efficiency. On the other hand, feature sets which are too small may not contain enough information to determine the target class and may cause underfitting.</p><p>The feature sets generated by APriori usually contain many solution patterns which are redundant or less useful as they are too small (i.e., strings/trees of length one). Such elements can be removed, and the size of the complete solution set can be reduced significantly, e.g. by computing so-called border elements [18], i.e., the most specific patterns that are still solutions. We calibrated Free Tree Miner to solely output border elements. Apriori was implemented to output only features that are border elements and larger than a user defined size threshold. Finally, we used in our study 14 sequence-based apriori features and 78 free trees (see Tables 2, 3).</p><!><p>For classification, we used standard schemes like decision tree (C5) and large margin (SVM) learning methods. C5 [19] is commercial improvement of C4.5 [20] written in C and popular for its efficiency. For the SVM [21], we used Weka's [22] implementation of Sequential Minimal Optimization (SMO) [23]. We tested three kernels (linear (E1), quadratic (E2) and radial basis function (RBF)) with Weka's default parameter setting including the cost factor C 1.0. A higher C slows down the running time of the classifiers. A C of 0.1, however, renders the RBF kernel SVM to a majority class predictor. For an SVM with a linear kernel the opposite is true, but it performs in all cases on a lower level. The performance of the quadratic kernel SVM remains nearly the same on the test data, on the training data, however, a smaller C decreases the preditive power. For C5, the main task consists of finding the best pruning options to control overfitting. C5 provides the option to prune with confidence intervals and with a minimum support of training instances that must be covered by each leaf of the tree. We used C5's default settings, with a pruning confidence factor of 25% and a minimum support in each leaf of 2. Subsequently, global pruning can be used to optimize the tree's performance further.</p><p>Note that for C5, continuous or numeric features are discretized using standard procedures [20]. For SVMs, nominal features are transformed to "binary numeric" using Weka's standard filter NominalToBinary [22,24]. All features used within SVMs are normalized by the Weka workbench by default. The kernels we applied are constructed out of all these normalized features.</p><!><p>We use leave-one-out cross-validation (LOOCV) to evaluate our classification results. LOOCV may appear uncommon, at first sight, in this setting with 2260 instances since it is generally recommended (along with the bootstrap) for smaller datasets. This is because a smaller number of folds would result in an even larger variance. LOOCV is known to deliver estimates with a small bias, whereas the variance can be high. However, with more than 2000 instances, the training sets do not vary a lot; therefore, even the variance is low in this case. Usually, ten times ten-fold cross-validation is preferred on such datasets for practical reasons, to avoid the excessive running times of LOOCV. However, we wanted to test the "purest" setting and also obtain maximally unbiased error estimates. Finally, it should be clear that the proposed evaluation variants can easily be extended towards regular k-fold cross-validation, by leaving out pairs of sets of kinases and sets of inhibitors in turn. To evaluate the quality of a model, we used three established performance measures: In-/correctly classified instances, recall and precision:(3)(4)(5)(6)(7)</p><p>Note that (4) is also known as Sensitivity and True Positive Rate (TPR), (5) as Selectivity and Positive Predicted Value (PPV), (6) as Specificity and True Negative Rate (TNR), and (7) as Negative Predicted Value (NPV).</p><p>In the following, we will present a new way of evaluating classifiers in the present setting, and give an overview of four different variants of LOOCV applied here. Since we aim for predictions for pairs of kinases and inhibitors, different amounts of information may be available for the two potential binding partners.</p><!><p>Figure 2 shows the two extreme variants of our different implementations of LOOCV. The left-hand side of the figure shows the "hard case", in which no information about the test kinase and the test inhibitor is allowed in the training dataset. This would, for instance, represent the scenario in which a binding prediction is performed for a completely unknown pair of a kinase and an inhibitor. In the "soft case" (see the right-hand side of Figure 2), however, all information about the test kinase and the test inhibitor is already known in the training set - except for the pair itself to be predicted.</p><!><p>Hard and soft case of LOOCV. Illustration of the hard (left) and the soft (right) case of LOOCV.</p><!><p>The two cases between the extreme variants of our different implementations of LOOCV are shown in Figure 3. The left side of the figure illustrates the "mixed case", in which the equal percentage of information on the test kinase and the test inhibitor are put into the training set. This means that a certain random fraction from the test kinase and the same random fraction from the test inhibitor is put into the training set. To give an example, if we use 50% of the test inhibitor information, we put 10 kinase-inhibitor pairs in the training set where the inhibitor in the pair must be the inhibitor to be predicted. For the kinase the same holds, but 50% make up 57 pairs. On the right side, the "mixed-mixed case" is illustrated, in which the training dataset contains information on the test kinase and test inhibitor in an unequal proportion. This represents the situation in which experimental information concerning the binding patterns of a certain test kinase to the inhibitors is partly available. For the test inhibitor, the same holds, but in a different proportion.</p><!><p>Mixed and mixed-mixed case of LOOCV. Illustration of the mixed (left) and the mixed-mixed (right) case of LOOCV.</p><!><p>We start with the results from the soft case evaluation. In the following, we show how different feature sets (see Table 4) affect the performance of the classifiers. The overall plan of the experiments was (1.) to start with a set of base features for both kinases and inhibitors, (2.) to refine the representation of kinases in the next step, and after that (3.) to refine the representation of inhibitors. For the representation of kinases, we add alignment-based features, then position-specific features, and ultimately both alignment-and position-specific features. For the representation of inhibitors, we start with the base features and then add further descriptors (CF, GF and P in the table) in a final refinement step.</p><!><p>This table indicates which features are contained in which feature sets</p><p>This table indicates which features are contained in which feature sets. PT: Primary Targets, MS: 2D Molecular Structure, FTs: Free Trees, KNN: KNN clustering, CF: Chemical Features, GF: Geometric Features, P: Pharmacophores, STTK: Partitioning in Serine-, Threonine and Tyrosine Kinases, PC: Phylogenetic Clustering, PRO: PROSITE patterns, Apri: APriori patterns, glAli: global alignment scores, locAli: local alignment scores, PSF: Position Specific Features, abPSF: abstract Position Specific Features. The upper part of the table describes the chemical features, the lower part the biological features. In the left part of the table (from FS1 to FS6), the description of the kinases is optimized (testing combinations of alignment-based and position-specific features). In the right part (FS7), the chemical representation is further optimized by additional descriptors.</p><!><p>In preliminary experiments, we evaluated the individual performance of feature groups (Table 5). Here the features for kinases perform very similarly, all in a range between 73% and 74% (for C5), whereas the features for inhibitors differ in their performance: the predictive accuracy of CF, GF, KNN, FTs, and P range between 79% and 80% (again for C5), with the remaining two feature groups (PT and MS) lagging behind. Results for SVMs are mostly comparable (see Table 5).</p><!><p>Comparison of prediction accuracies for single feature groups</p><p>Comparison of prediction accuracies of C5 without global pruning and SVMs with the quadratic kernel for single feature groups on the test set. For a description of the abbreviations of the feature sets see Table 4.</p><!><p>Figure 4 shows the prediction accuracies for different feature sets for both SVMs and C5 that were run with different parameter settings. In all cases C5, without the option "global pruning", outperforms the other variant with global pruning. Compared to the SVM, it is extremly fast and handles large feature sets well concerning runtime and memory.</p><!><p>Performance on different feature sets (soft case). Prediction accuracies, recall and precision for different feature sets from C5 and Support Vector Machines with different parameter settings (soft case).</p><!><p>For C5, the usage of global alignment scores as features (FS2) reduced the prediction accuracy on test data significantly. This may be explained by the fact that these scores take into account the complete amino acid sequence, whereas only a small part constitutes the active center and is therefore important for the binding to an inhibitor. These non-informative sequences clearly outweigh the informative ones, and so they obscure the information and make the scores for global alignments less useful. Alignment scores of extracted conserved regions perform clearly better on the training and test set. In this case it is particularly remarkable that cut out and realigned conserved regions (FS3) perform 1.6% better than without realigning on the test set.</p><p>The most difficult prediction task for the applied classifiers was the correct prediction of a binding between an inhibitor and a kinase. For all different feature sets, C5 as well as SVMs have the lowest values for the recall of the positive class (Figure 4). Particularly conspicuous are the extremely low values for the recall of the positive class of SVMs with an RBF and a linear kernel. For C5, feature sets 3, 6 and 7 clearly show the best positive recall (and prediction accuracy) for the test data. Particular attention should be paid to FS3, which comprises, besides the basic feature set, only local alignment scores. This indicates that local alignment scores are very suitable for making predictions with C5. The negative recall is relatively constant for all feature sets. However, the tradeoff between positive and negative recall is visible.</p><p>Comparing FS1 with FS3, or FS4 with FS7, shows that a higher positive recall leads to a lower negative recall. A combination of global and local alignment scores, as well as a combination of position specific features with abstract position specific features, degrades the predictivity. Only when we combined alignment score features (FS4) with position specific features (FS5) to feature set 6, the prediction accuracy increases significantly on the training and test set. This combination is then further improved by adding chemical features leading to the best prediction accuracy we obtained from C5 on the test set. This success can largely be attributed to the use of chemical features and the diversity of the features.</p><p>With SVMs, we used global alignment scores as features (FS2) only once, since they slow down the computation enormously and when combined with other feature sets, do not contribute to an increase in the prediction accuracy on the test set. However, we tested the influence of using position specific features with (FS5) and without abstractions (FS4), where the use of abstract position specific features did not show an improvement of the prediction accuracy.</p><!><p>The differences between the kernels are clearly observable from Figure 4. The quadratic kernel performs with higher success than the linear and RBF kernel for all feature sets except for FS1 and FS2 for both kernels and FS3 for the RBF kernel on the test set. On the training set, this fact must be mainly attributed to overfitting. On the test set, the best results are obtained with feature set 4. This indicates that SVMs with a quadratic kernel work best with position specific features. This may be due to the fact that we described here for the first time the active site of a kinase with position specific features. An addition of further features does not lead to an increase in the predictive power for both training and testing.</p><p>For the linear and RBF kernel things are different. A larger amount of features increases the predictivity on the training data set but harms it on the test data set except for FS7 with the RBF kernel. From Figure 4 it is evident that the recall for the negative class normally drops from feature set 1 to feature set 7 for the linear kernel. The reason for this may be that SVMs are not able to predict the "binding" class with features that do not discriminate immediately between the classes. Hence, SVMs mostly predict the majority class "no-binding", leading to a high negative recall. But with increasing ability to discriminate between the classes, more bindings are predicted correctly, leading to an increase in the recall for the positive class. On the test set this is accompanied by a decrease in the recall for the negative class and an increase in its corresponding precision (Figure 4). For the RBF kernel we obtain the best predicitivity as well as the highest positive and negative recall with FS7. The chemical features seem to be the most decisive ones with respect to the RBF SVM. From feature set 1 to feature set 6 the negative recall and precision remain relatively constant. The positive recall and precision, however, decrease significantly.</p><!><p>We also tested the performance on features sets with random feature values. For feature set 3, we assigned random integers (FS3_ran) to all local alignment score values where a random integer must be in the range between the smallest and largest value of the true values. Results for C5 and SVMs with a quadratic kernel are shown in Figure 5. As expected, the usage of random features harms performance. For C5, it is visible that for feature set 3 the drop of the performance is larger than for feature sets 6 or 7. The same holds for SVMs with respect to feature set 4 and 7. This can be explained by the fact that the amount of random features for feature set 3 is much higher than for feature set 7, for instance, where random values are assigned to three rather small feature groups (chemical, geometrical and pharmacophoric features). Particularly, the worse performance of the random feature sets 6 and 7 indicates that the sheer size of a feature set does not neccessarily contribute to a better performance through chance correlation or overfitting, but that the diversity of a feature set is the factor positively impacting predictivity.</p><!><p>Comparison of prediction accuracies with random features. Comparison of prediction accuracy for different feature sets including random features.</p><!><p>All the results in this section are based on feature set 7. C5 shows only slight, nonsignificant differences between the hard and the soft case concerning the training data, however, on the test data there are large differences (Figure 6). In the hard case, recall and precision values for the positive class are very low, which indicates that the classifier is not good at identifying kinase and inhibitor features responsible for binding. From the hard to the soft case, these two values clearly increase. In contrast, there is only a small drop in the recall and the precision for the negative class. As for C5, the prediction accuracy on the test and the training set of SVMs (with a linear, quadratic or RBF kernel and cost factor C = 1.0) always increases from the hard to the soft case. In the hard case, there are even worse results regarding the recall for the positive class (5.0% recall). The quadratic kernel clearly overfits on the training data.</p><!><p>Performance comparison of the hard and the soft case. Comparison of the prediction accuracy and recall/precision in the hard and the soft case.</p><!><p>In some hard cases for different feature sets (not shown here), especially for small feature sets with predominantly features that are not able to discriminate between classes, C5 as well as SVMs are performing as good as a majority class classifier since they predict everything as non-binding. For more complex feature sets like in the cases shown here, C5 and the SVM classifiers with a linear or quadratic kernel are slightly worse than a majority class classifier (red line in Figure 6) that would reach 73.6%. SVMs with an RBF kernel, however, perform slightly better although the recall for the positive class is very low. However, in this case, this is compensated by a high precision for the positive class as well as a high recall for the negative class (see Figure 6).</p><p>In the soft case, we compare our prediction accuracy results with a simple baseline classifier that calculates the probabilities for a binding from the binding matrix, not taking into account any information about the molecules. This is only possible in non-hard cases, since the information how often a test kinase/inhibitor binds to a training set inhibitor/kinase is directly taken into account. More precisely, the simple baseline classifier calculates, separately, the probabilities pkin(b) and pinh(b) of a test kinase/inhibitor binding on the training set. Subsequently, these probabilities are multiplied and it is determined whether the product is greater than the threshold θ that was optimized empirically:(8)</p><p>A smaller value than θ results in a "no-binding" prediction, otherwise "binding" is predicted. This simple classifier is able to reach 78.5% prediction accuracy without any knowledge about the kinases or inhibitors except their binding patterns. It is clearly better than a majority class classifier, but worse than models that consider additional information about the molecules.</p><p>The difference between the hard and the soft case are the kinase-inhibitor pairs in the training set that contain either the test kinase or the test inhibitor. The performance improvement of the classifiers must be attributed to these pairs. Figure 7 shows the results if we remove all kinase-inhibitor pairs from the training set which do not contain the test kinase or the test inhibitor. On the training set (consisting of 131 instances or kinase-inhibitor pairs), a strong performance can be observed. Compared to the soft case (see Figure 6 soft case), however, the results on the test set, disregarding SVMs with a linear kernel, show a clear performance loss if we remove kinase-inhibitor pairs from the training set which contain neither the kinase nor the inhibitor to be predicted. The usage of test kinase and test inhibitor information, solely, is not sufficient to obtain reasonable results. This means that the applied machine learning methods require the other pairs in order to generalize well.</p><!><p>Performance using solely test kinase-inhibitor pairs. Comparison of prediction accuracy and recall/precision using solely test kinase-inhibitor pairs in the training set.</p><!><p>All the results in this section are based on predictions of C5 that was run on feature set 7 without global pruning. Figure 8 shows the prediction accuracies of three mixed cases, the soft and the hard case, a simple baseline classifier and a majority classifier, as well as the recall and precision values for our predictions. The results for the mixed cases and the simple baseline classifier are obtained by averaging the results from ten runs of C5 with identical parameter settings. Note that we took randomly a certain fraction of test kinase/inhibitor information in the training set. This means that the results in each run can be slightly different. Hence, it is necessary to conduct several, in our case 10, experiments with the same parameter setting and average the results in order to take these variations into account.</p><!><p>Performance comparison of different mixed cases (C5). Comparison of prediction accuracy and recall/precision for different mixed cases (C5 without global pruning).</p><!><p>The performance on the test set is strongly influenced by the usage of different fractions of the test kinases and the test inhibitors in the training set. The performance on the training data, however, is nearly independent of it (see Figure 8). The strong increase in performance from the hard case to the first mixed case (30/30) indicates the importance of the information about the test kinase and the test inhibitor. The usage of information from the test kinase and the test inhibitor leads to a substantial increase in the recall and the precision for the positive and the negative class. The classifier learns to discriminate better between the classes and thus to predict a correct binding more often without losing performance on the negative class.</p><p>Further results (not shown here) reveal only slight differences in the training set performance for the mixed-mixed case. Here, the classifier performs clearly independent of the number of instances from the test kinase and the test inhibitor in the training set. On the test set, however, there is great variability in the prediction accuracy of different mixed-mixed cases. First, we analyze the performance if a percentual amount of information on the test kinase and the test inhibitor is added to the training set. Note that this means that actually more information on the test kinase is added since the dataset consists of more kinases than inhibitors. Second, we analyze the case in which equal information on the test kinase and the test inhibitor is added to the training set, i.e. if 10 pairs consisting of the test kinase and 10 different training set inhibitors are added, then 10 pairs of the test inhibitor and 10 different training set kinases also have to be added. Further note that the training set without test molecule information can be seen as a reference. This reference represents the hard case. For both variants, it is tested how information on test kinases and test inhibitors in the training set can improve the performance compared to the reference. From Figure 9, the worst and the best values are obtained in the hard and the soft case, respectively. The same holds for the corresponding cases in Table 6 (0/0 and 19/19). Results with percentual and absolute test molecule information are similar. For the dataset under investigation, it can clearly be seen that the information about the kinase test molecule is more important than that of the tested inhibitor. If no information from the test inhibitor and only little information from the test kinase is in the training set, the prediction accuracy increases significantly. In contrast, no information from the test kinase and little information from the test inhibitor leads to a remarkably lower increase in the prediction accuracy (see Figure 9 and Table 6). This can be clearly seen, for instance, in Table 6 for the cases 10/0 and 0/10.</p><!><p>Performance comparison of different mixed-mixed cases (C5). Comparison of prediction accuracy for the soft, hard, mixed and mixed-mixed cases (C5 without global pruning).</p><p>Comparison of prediction accuracies for some mixed-mixed cases (on an absolute basis) (FS7)</p><p>Comparison of prediction accuracies for C5 for some mixed-mixed cases on the test set (on an absolute basis). Results are obtained with feature set 7.</p><!><p>We benchmarked our kinase inhibitor binding prediction approach for the mixed-mixed cases and the soft case with a majority class classifier as well as with a simple baseline classifier (Table 7). The more informed classifier, C5, performs in all cases, except one, better than both reference classifiers. Mostly, a clear performance improvement with respect to prediction accuracy can be observed. This means that the feature extraction from the kinases and the inhibitors is beneficial. The same holds for the mixed cases and the soft case.</p><!><p>Comparison of prediction accuracies for some mixed-mixed cases (on a percentage basis) (FS7)</p><p>Comparison of prediction accuracies of different classifiers for some mixed-mixed cases on the test set.</p><p>Results are obtained with feature set 7.</p><!><p>In summary, the best model achieved 83.8% predictive accuracy with a recall of 0.59 and a precision of 0.75 for the positive class. The most frequently used features in the learned decision tree are position-specific features, local alignment features and the JOELib2 chemical features.</p><!><p>In addition, the classifiers were tested on an external dataset consisting of 19 kinase inhibitors and 177 protein kinases [25]. It is the result of a later study from Ambit Biosciences and produced in the same way as the dataset described in the section "Data". Note that the original dataset consists of 38 inhibitors and 317 kinases. We removed inhibitors and kinases that are contained in the training set [7] and those where information is missing needed for descriptor calculation. The class distribution of this compiled dataset is similar with 26.6% bindings and 73.4% non-bindings. Testing on an external dataset corresponds to the hard case since neither information about the inhibitors nor information about the kinases is available. The best results on feature set 7 we obtained with C5 without global pruning (prediction accuracy on test set: 74.1%). SVMs with an RBF kernel are also able to outperform a majority class predictor. SVMs with a linear or quadratic kernel, however, perform slightly worse than a majority class predictor (Figure 10). These results represent a clear improvement in comparison to the hard case in the LOOCV setting. Primarily, this improved performance can be explained by structurally similar inhibitors present in the training set, which are not available in LOOCV (see Figure 1).</p><!><p>Performance on the external test set. Prediction accuracy and recall/precision on the external test set with feature set 7, for both C5 and SVMs.</p><!><p>Kinase inhibitor predictions have been investigated over the past few years. Basically, there exist two approaches. The simpler, and more established one, is to calculate a vectorial representation of both kinase inhibitor and non-kinase inhibitor molecules and using the result with standard machine learning algorithms to predict the probability of a molecule to be a kinase inhibitor. This approach was, for instance, taken by Briem and Günther [26]. In their study, they used a Schering in-house dataset of small molecules encoded by 120 fragment-based Ghose-Crippen descriptors and applied several machine learning techniques (SVMs, artifical neural networks, kNN with GA-optimized feature selection and recursive partitioning) to distinguish between kinase inhibitors and molecules with no reported activity on any protein kinase. Since the original dataset was strongly imbalanced, a ensemble-based sampling procedure was applied to ensure balanced training sets. In the end, 13 training sets were generated for model learning and applied to an independent test set. Briem and Günther analyzed to what extent machine-larning algorithms are capable of learning kinase inhibitor likeness and to compare the different classifiers. Results are reported for each of the 13 individual sample classifiers and for a consensus majority vote of all members of the ensemble. The results show that the latter generally outperforms averaging over the individual models. All four methods exhibited a reasonable discriminative power. Comparing the individual classifiers with respect to standard quality measures, SVMs seem to be the best choice. This is also true for a further compiled test set with significantly different structures.</p><p>Xia and colleagues [27] used a modified Naive Bayes classifier to model multifamily and single-target kinase inhibitors. In their study, they used Amgen's CORP datasets (around 200.000 molecules) composed of kinase inhibitors, potential kinase inhibitors, and random drug molecules. To describe the molecules, standard physicochemical features as well as a 2D structural fingerprint were used. To assess the performance of the Bayesian model, the positions of active compounds in ordered scoring lists of the test set were used. The approach was validated by first using an equal proportion of training and test instances (1:1) and second using a much smaller training set (1:9). The results suggest that only 10% of the data are enough to yield a performance nearly as good as if 50% were used. 85% of the active compounds occurred in the top 10% of the ordered molecules. This underlines the power of the Bayesian model which is also confirmed on 172 novel kinase inhibitor compounds from different structural classes that were classified with the 1:9 Bayesian split model. 70% of these new compounds were found in the top 10% and 85% in the top 20% rank-ordered compounds.</p><p>Compared to our study, Briem and Günther as well as Xia and colleagues used only information of small molecules (kinase inhibitors, potential kinase inhibitors and random drug molecuels) for predicting the probability of a molecule being a kinase inhibitor. Information about kinases is not considered.</p><p>The second, and more difficult approach, is to use features from kinase inhibitors and protein kinases in combination. Weill and Rognan [28] presented a novel low-dimensional fingerprint approach encoding ligands and target properties to mine the protein-ligand chemogenomic space. Kinase inhibitors are represented by standard descriptors, while protein transmembrane cavities are encoded by a fixed length bit string describing pharmacophoric properties of a defined number of binding site residues. Due to the complexity of the cavity, this study is restricted to G protein-coupled receptors (GPCRs) with a homogeneous cavity description. Several machine learning classifiers on two training sets of roughly 200.000 receptor-ligand fingerprints with different definitions of inactive decoys are applied for model learning. Two external test sets of 60 GPCRs were used to validate the models. Experimental results suggest that SVMs with an RBF kernel perform best with respect to a balanced accuracy measure combining true positive and true negative rate. The authors demonstrate that protein-ligand fingerprints outperform the corresponding ligand fingerprints in predicting either putative ligands for a known target or putative targets for a known ligand. They conclude that, with respect to GPCRs, predicting ligands is significantly easier than predicting targets.</p><p>Our approach resembles the one of Weill and Rognan in that both kinase and inhibitor information is used for modeling. However, a key difference in Weill and Rognan's approach is the restriction to GPCRs, whereas in this paper, we take a broad spectrum of different kinase families into account and thus are able to make predictions for a larger range of kinases and inhibitors.</p><!><p>We tackled the prediction task whether a binding between a protein kinase and an inhibitor can take place, given a set of features describing both molecules. We applied and tested a range of data mining and classification tools. Finally, we used both C5 and Support Vector Machines together with three variants of leave-one-out cross-validation to learn and validate concepts of protein kinase inhibitor bindings and the influence of information available about the potential binding partners. The approach performs well in the soft case validation and comparable with a majority class classifier in the hard case validation. On an external test set we obtained a clearly better performance than a majority class predictor with C5 and SVMs with an RBF kernel. As expected, the performance can be improved by features describing the active site of the kinases by local alignment scores for C5 or position specific features for SVMs. These features are frequently used by C5 and increase the prediction accuracy substantially. Primary chemical features and a diversity in the feature sets had a positive influence on the performance of the classifiers. Note that once a pair of a kinase and an inhibitor is classified as binding, a regression model could be applied subsequently to predict the quantity of the binding affinity.</p><p>In summary, the contributions of the paper are as follows: First, we presented a machine learning approach to modeling the binding affinity of inhibitors to kinases. In particular, it is the first time that the complete dataset of Fabian et al. [7] with information for all pairs from a set of inhibitors and a set of kinases, is used in predictive modeling (classification). Second, we proposed novel validation schemes for this kind of problem, depending on how much information is available for the inhibitor and for the kinase. Third, our experiments showed that for the decision tree learner C5 alignment-based features are very useful, but that a combination with position-specific features and certain chemical features is necessary for obtaining the best results. For SVMs, the best results are obtained with a quadratic kernel and position specific features. The best predictive accuracy of 86.1% indicates that machine learning methods are able to detect a signal in the data and predict binding affinity to some extent. However, it is clear that there is ample room for improvement for all kinds of methods and that the prediction of kinase-inhibitor binding will remain a relevant research topic for a long time to come.</p><!><p>The authors declare that they have no competing interests.</p><!><p>FB implemented the methods and conducted the experiments. FB, LR and SK made substantial contributions to the conception, design and coordination of the study. All three analyzed and interpreted the results, were involved in drafting the manuscript, revised it critically for important intellectual content, and gave the final approval of the version to be published.</p><!><p>FB is a PhD student at the computer science department of Technische Universität München. After receiving his diploma in bioinformatics from the Ludwigs Maximilians Universität München and Technische Universität München, he began to work on predictive toxicology in the scientific staff of Prof. Stefan Kramer (SK) in the Machine Learning and Data Mining in Bioinformatics group at Technische Universität München. His current research interests include predictive toxicology, machine learning, data mining, bioinformatics and cheminformatics.</p><p>LR is a Post-doc in the group of SK. He received a diploma and a Ph.D. in biology from Technische Universität München. After a year in a biotech-startup company he came back to Technische Universität München for a postgraduate study in computer science and joined SK's group upon completion. He is interested in data mining and integration of chemical and biological data.</p><p>SK is professor of bioinformatics at the computer science department of Technische Universität München. After receiving his doctoral degree from the Vienna University of Technology, he has spent a few years as an assistant professor in the Machine Learning lab of the University of Freiburg. He was the co-organizer of the Predictive Toxicology Challenge 2000-2001, an international competition in toxicity prediction. He has organized several conferences and workshops, edited special issues of journals, given invited talks and tutorials, and serves on the program committees of major data mining and machine learning conferences and on the editorial board of the Machine Learning journal. His current research interests include data mining, machine learning, and applications in chemistry, biology and medicine.</p><!><p>In this additional file we show in a table how often an inhibitor binds to a certain group of kinases (group in a phylogenetic meaning). It can be clearly seen that nearly all inhibitors bind to several kinase groups. This means that there is generally no kinase group to that an inhibitor binds consistently.</p><!><p>Click here for file</p><!><p>OpenTox - An Open Source Predictive Toxicology Framework, http://www.opentox.org/ is funded under the EU Seventh Framework Program: HEALTH-2007-1.3-3 Promotion, development, validation, acceptance and implementation of QSARs (Quantitative Structure-Activity Relationships) for toxicology, Project Reference Number Health-F5-2008-200787 (2008-2011).</p>

PubMed Open Access

Use of DT40 conditional knockout cell lines to study chromosomal passenger proteins function

The chromosomal passenger complex (CPC-INCENP, Aurora B kinase, Survivin and Borealin) is implicated in many mitotic processes. Here we describe how we generated DT40 conditional knockout cell lines for incenp1 and survivin1 to better understand the role of these CPC subunits in the control of Aurora B kinase activity. These lines enabled us to reassess current knowledge of Survivin function and to show that INCENP acts as a rheostat for Aurora B activity.

use_of_dt40_conditional_knockout_cell_lines_to_study_chromosomal_passenger_proteins_function

INTRODUCTION<!>1- Assessment of SURVIVIN function in mitosis and apoptosis<!>1-1 KNOCKOUT STRATEGY<!>1-2 THE ROLE of SURVIVIN in MITOSIS<!>1-3 THE ROLE of SURVIVIN in CELL DEATH<!>1-4 CONCLUSIONS<!>2- Functionnal analysis of INCENP-Aurora B interactions<!>2-1 KNOCKOUT STRATEGY<!>2-2 Regulation of Aurora B activity through INCENP binding<!>2-3 CONCLUSION<!>3- CONCLUSIONS

<p>Accurate chromosome segregation requires that kinetochores of sister chromatids bind microtubules that emanate from opposing spindle poles. During prometaphase, various aberrant kinetochore attachments occur which, if not corrected, can lead to improper chromosome segregation and aneuploidy. To avoid this, cells have a surveillance mechanism, called the spindle checkpoint, which delays anaphase until all sister kinetochores are properly captured and under tension. Kinetochore attachment and error correction are directly controlled by the Aurora B kinase, part of the chromosomal passenger complex (CPC).</p><p>In most organisms, the core CPC is composed of Aurora B kinase [1] and three non-enzymatic subunits, INCENP [2, 3], Survivin and Borealin/Dasra B [4-7] that control targeting, enzymatic activity and stability of Aurora B kinase [8]. The CPC controls many aspects of mitosis ranging from chromosome and spindle structure to the correction of kinetochore-microtubule attachment errors, regulation of mitotic progression and completion of cytokinesis [9]. Knockdown by RNA interference of any member of the complex disrupts mitotic progression [4, 7, 10-13]. Although INCENP has been shown to be responsible for the initial Aurora B activation through direct binding and a phosphorylation feedback loop [9, 14, 15], the mechanisms responsible for controlling the activity of this important kinase are still poorly understood.</p><p>The ability to shut down/off the expression of a gene by RNAi or knockout is widely used to study protein function in many eukaryotic systems. Whereas RNAi relies upon the destruction of the mRNA, which is continuously produced, conditional knockouts block production of the mRNA at its source. Another advantage of knockout strategies over RNAi is the absence of off-target effects [16]. Several cell types are used to generate knockouts including mouse ES cells, human RPE, Nalm-6, HT1080 and chicken DT40 cell lines [17-22].</p><p>DT40 cells are chicken B lymphoma with a very high rate of homologous recombination (up to 90% for some loci), greatly facilitating gene targeting. DT40s have been used to study many biological processes, including DNA damage pathways [23], transcriptional regulation [24], calcium signalling [25], apoptosis [26-29] and chromatin structure [30-32].</p><p>To better understand the role of the CPC in regulating Aurora B activity we studied the involvement of INCENP and Survivin by generating DT40 conditional knockout cell lines for their genes [33, 34].</p><!><p>Survivin is a protein involved in mitotic progression and apoptotic regulation that is up-regulated in many human tumours. Survivin's expression peaks during mitosis [8, 35, 36]. The protein is required for targeting the CPC to centromeres [8, 37]. Survivin is considered a member of the inhibitor of apoptosis protein (IAP) family, despite lacking some of the key features shared by other members of the family. The exact role of Survivin in mitosis and apoptosis remains unclear.</p><p>Our lab generated two DT40 conditional knockouts of the survivin gene. We used this system to study the role of Survivin both in mitosis and apoptosis and to assess the role of key residues of the proteins in these biological processes.</p><!><p>Survivin−/− cells were generated by deleting the entire survivin1 gene (Acc. N°: ENSGALG00000008713, Figure 1A). Briefly, after targeting and replacement of the first allele by a resistance marker, the survivin cDNA, regulated by a Tet off system, was introduced in the heterozygous cells. The second allele was then targeted, generating a conditional Survivin knockout cell line which, upon addition of doxycycline exhibits the null phenotype. Stable transfection of this cell line with various Survivin cDNA-bearing mutations of interest allowed us to study their phenotypes in a null background.</p><!><p>Using our unique survivin−/− model, we confirmed some previous findings on Survivin function and revealed important features of key residues of the protein. For instance, we showed that Survivin is essential for both chromosome segregation and cytokinesis, concordant with observations from RNAi studies [12, 13]. A mutant previously shown to abolish interaction between Survivin and Aurora B (human SurvivinD70A/D71A, chicken SurvivinD72A/D73A) [38] introduced into the Survivin knockout cells was unable to concentrate normally at centromeres. This induced delocalization of the other passenger proteins in early mitosis, surprisingly without affecting subsequent transfer of the CPC to the spindle midzone. Furthermore, this mutant showed a compromised spindle checkpoint.</p><p>Unexpectedly, despite the presence of these defects, this mutant could progress through mitosis and sustain cell growth, suggesting that Survivin binding to Aurora B is not absolutely required for cell survival [34]. In contrast with RNAi studies, we didn't observe any significant increase in defects in early phases of mitosis such as chromosome misalignment or abnormal spindles in the absence of survivin [12, 13, 39-41]. Survivin−/− cells exhibited mitotic arrest at high doses of taxol whereas at lower doses they could override the spindle checkpoint and progress through mitosis. This override was concomitant with increased apoptosis suggesting abnormal mitotic exit and subsequent cell death occurring in interphase [34].</p><!><p>Similarly to what we observed for mitosis, survivin−/− cells enabled us to confirm some previously described results regarding Survivin's role in apoptosis and to contradict others. For instance we could confirm that some mutations in specific domains of Survivin did induce a loss of function phenotype, such as point mutations in the BIR domain (SurvivinC59A or SurvivinC86A) [42]. On the other hand, loss of Survivin did not induce an increased sensitivity to pro-apoptotic stimuli. Furthermore, abolition of Survivin binding to Smac (SurvivinD55A), an antagonist of Survivin involved in apoptosis, did not prevent cell growth nor did it display a pro-apoptotic phenotype, contradicting previous reports about the protective role of Survivin in apoptosis [42-46].</p><!><p>Generation of DT40 survivin−/− cells expressing various mutants, allowed us to reassess Survivin functions both in mitosis and in apoptosis in complete absence of the endogenous wild-type protein. The null background proved to be essential to study Survivin function. Discrepancies observed between our study and previous reports demonstrate the importance of a genetically-clean system to dissect the protein functions. Our study ultimately proved that cell death induced by loss of Survivin is linked to cell cycle defects (primarily a failure in cytokinesis) rather than disregulation of apoptosis.</p><!><p>INCENP is the scaffolding protein of the CPC [2, 3]. Its N-terminus is required for centromere targeting of the complex whereas its C-terminus contains the IN box where Aurora B binds and gets activated. To study in more detail the interactions between INCENP and Aurora B, we generated a conditional DT40 knockout for incenp1 (Acc. N°: ENSGALG00000007537), expressing INCENP bearing specific point mutations in the IN box. We chose residues predicted to be important for activation (W766) or binding of Aurora B (F802) to INCENP [15].</p><!><p>To generate an incenp1 knockout, we used a promoter hi-jack strategy (Figure 1B) [33, 47]. Here, the first allele was placed under the control of a Tet off system, which was achieved by replacing the endogenous promoter with a minimal CMV/TetO promoter, together with stable expression of the tTA/VP16 transactivator under the control of the kif4A promoter. The other incenp1 allele had its ORF disrupted, potentially allowing expression of only the 28 first amino acids of the protein. In these cells, the addition of doxycycline shuts off expression from the endogenous incenp1 allele (Figure 2A, B). This strategy enabled the conditional expression of the multiple spliced isoforms of INCENP (class I and class II) [48].</p><!><p>We used this system to study how interactions between INCENP and Aurora B affect the activity of the kinase along with CPC function and localisation. The two mutants generated had different binding capacity to Aurora B leading to different levels of kinase activity. Expression of INCENPW766G in absence of wild-type INCENP abolished Aurora B interaction with INCENP, reducing Aurora B activity by 70%. INCENPF802A mutant bound Aurora B in a normal fashion but kinase activity was reduced by 50% contrasting with predictions from previous structural analysis [15]. Expression of either INCENPW766G or INCENPF802A in a null background induced cytokinesis failure and rapid cell death [33].</p><p>Both INCENPW766G and INCENPF802A were properly localised at centromeres in early mitosis but failed to transfer to the spindle midzone in anaphase. We could thus demonstrate that mitotic entry and sister chromatid separation are independent of CPC formation or Aurora B activity whereas both are required for proper transfer of the CPC to the spindle midzone and completion of cytokinesis [33]. Despite reports from RNAi experiments showing that Aurora B activity is required for outer kinetochore assembly [49, 50], we were able to detect CENP-A, -H, -O, -T, -E and the Hec1/Ndc80 complex at kinetochores of prometaphase incenp1−/− cells suggesting that kinetochore assembly is not affected by the lack of CPC at centromeres [33]. Moreover, CENP-E and PRC-1 localisation to the spindle midzone of incenp1−/− anaphase cells demonstrated that a spindle midzone was present despite the absence of CPC [33].</p><p>Expressing INCENP mutants in a null background enabled us to analyse the relationship between Aurora B kinase activity and the spindle checkpoint response induced by taxol. We established that various Aurora B activity levels yielded different spindle checkpoint responses when cells were treated with low doses of taxol. A minimal Aurora B activity (~50%, INCENPF802A) was sufficient to trigger a checkpoint response whereas lower activity (~30%, INCENPW766G or 15% with Aurora B inhibitor) only induced a weak checkpoint response [33]. This probably reflects the ability of the kinase to correct kinetochore-microtubules mis-attachments.</p><!><p>We mutated residues of INCENP that were predicted by crystallography studies to affect Aurora B binding or activity. Using incenp1−/− cells, we could assess the in vivo effects of mutating these residues. We were able to show that modulation of Aurora B activity regulates CPC localisation and function during mitosis. Ultimately, the use of DT40 incenp1−/− cells expressing various mutants allowed us to strengthen the model in which INCENP acts as a rheostat for Aurora B activity [33].</p><!><p>The DT40 cell line is a powerful system to study protein function. The ability to perform conditional knockouts of essential, multiply-spliced cell cycle-regulated genes makes it an ideal system to screen for and characterise functional domains or residues through the analysis of specific mutants against a null background.</p>

PubMed Author Manuscript

Application of Biochar Derived From Pyrolysis of Waste Fiberboard on Tetracycline Adsorption in Aqueous Solution

In this study, biochars derived from waste fiberboard biomass were applied in tetracycline (TC) removal in aqueous solution. Biochar samples were prepared by slow pyrolysis at 300, 500, and 800°C, and were characterized by ultimate analysis, Fourier transform infrared (FTIR), scanning electron microscopy (SEM), X-ray photoelectron spectroscopy (XPS), Brunauer–Emmett–Teller (BET), etc. The effects of ionic strength (0–1.0 mol/L of NaCl), initial TC concentration (2.5–60 ppm), biochar dosage (1.5–2.5 g/L), and initial pH (2–10) were systemically determined. The results present that biochar prepared at 800°C (BC800) generally possesses the highest aromatization degree and surface area with abundant pyridinic N (N-6) and accordingly shows a better removal efficiency (68.6%) than the other two biochar samples. Adsorption isotherm data were better fitted by the Freundlich model (R2 is 0.94) than the Langmuir model (R2 is 0.85). Thermodynamic study showed that the adsorption process is endothermic and mainly physical in nature with the values of ΔH0 being 48.0 kJ/mol, ΔS0 being 157.1 J/mol/K, and ΔG0 varying from 1.02 to −2.14 kJ/mol. The graphite-like structure in biochar enables the π-π interactions with a ring structure in the TC molecule, which, together with the N-6 acting as electron donor, is the main driving force of the adsorption process.

application_of_biochar_derived_from_pyrolysis_of_waste_fiberboard_on_tetracycline_adsorption_in_aque

Introduction<!>Materials<!>Preparation and Characterization of Biochar<!>Batch Adsorption Experiments<!>Characteristics of Biochar<!><!>Characteristics of Biochar<!><!>Characteristics of Biochar<!><!>Characteristics of Biochar<!><!>Adsorption Ability of BC300, BC500, and BC800<!><!>Effect of Ionic Strength<!>Effect of Biochar Dosage<!>Effect of Initial TC Concentration<!>Effect of Initial pH<!>Adsorption Isotherms and Thermodynamics Analysis<!><!>Adsorption Isotherms and Thermodynamics Analysis<!><!>Further Discussion<!><!>Conclusions<!>Data Availability Statement<!>Author Contributions<!>Conflict of Interest<!><!>Supplementary Material<!>

<p>Aquatic ecosystem pollution by antibiotics has received rising concerns due to their potential hazards on the aquatic biota and even human beings (Liu et al., 2012; Jang et al., 2018; Premarathna et al., 2019). The tetracycline (TC) group is one of the most widely applied antibiotics in human therapy and the livestock farming globally thanks to their comparatively low prices and efficient treatment (Liu et al., 2012; Nguyen et al., 2019). TC is hard to metabolize and was extensively excreted (up to 90%) in urine and feces from human and animals (Jang et al., 2018; Li et al., 2018; Selmi et al., 2018; Jang and Kan, 2019). It has been reported that TC was widely found in various water bodies, e.g., surface water, groundwater, and even drinking water (Jeong et al., 2010; Cao et al., 2019; Zhang et al., 2019). Antibiotics removal from water bodies is highly interesting for many researchers worldwide. Various technologies were adopted to remove TC, including biological, electrochemical, membrane, advanced oxidation process, and adsorption process (Jeong et al., 2010; Cao et al., 2019; Smyk et al., 2019). In comparison with other approaches, adsorption is outstanding for its unique advantages like easy operation, low toxicity, low energy cost, as well as high removal efficiency at low concentrations (Han et al., 2019; Regkouzas and Diamadopoulos, 2019; Shaheen et al., 2019). There are numerous effective adsorbents such as activated carbon (Xiang et al., 2019), carbon nanotubes (Xiang et al., 2019), zeolite (Wang et al., 2019), chitosan (Ahamad et al., 2019), etc. However, those adsorbents are relatively unaffordable when dealing with large scale of wastewater. Therefore, alternative low-cost adsorbents are highly demanded for TC removal. Biochar, a by-product of biomass pyrolysis, has been attracting huge attention as a promising cost-effective adsorbent due to its properties of abundant functional groups, porous structure, rich aromatic structures, and environmental friendly nature (Ahmad et al., 2014; Patra et al., 2017; Maleki et al., 2018).</p><p>On the other hand, fiberboard as a major wood composite possesses a huge gross production capacity (~55.54 million m3 in 2013 in China), which leads to a large quantity of waste fiberboard (Gan et al., 2004; Wu et al., 2012; Liu et al., 2014). Nowadays, waste fiberboard is generally disposed of by burning, which is uneconomical and extremely harmful to environment as well as human health (Liu et al., 2014; Zhang et al., 2014). As a result, dealing with such amount of waste fiberboard is of critical importance for both economic and environmental concerns. One effective and economical method to better utilize those waste fiberboard is to prepare biochar via pyrolysis. Previous studies have proved that biochar derived from waste fiberboard biomass is a promising adsorbent for removal of heavy metal and organics, with or even without further modification (Wu et al., 2012, 2014; Liu et al., 2014; Pan et al., 2016, 2018). It has been reported that biochar from fiberboard biomass usually possesses high content of N-containing groups, which is beneficial to the adsorption process due to the enhanced alkaline property and wettability (Wu et al., 2012, 2014; Zhang et al., 2014; Zhan et al., 2019). This also addresses the issue of high cost adsorbents along with TC removal.</p><p>Therefore, this study aims to prepare and characterize biochar derived from pyrolysis of waste fiberboard, which was then utilized for TC removal in aqueous. The properties of TC solutions (i.e., ionic strength, initial concentration, and pH) were varied to examine the changes in adsorption behavior and capacities by the waste-derived chars, while the adsorption isotherms and thermodynamics analysis were also conducted to explore the fundamental mechanism. Overall, this study has made appreciable progress for the understanding TC adsorption on biochar derived from waste fiberboard.</p><!><p>TC was purchased from Macklin Biochemical (Shanghai, China). HCl was sourced from Nanjing Chemical Reagent (Nanjing, China) while both NaOH and NaCl were supplied by Sinopharm Chemical Reagent (Shanghai, China). The above chemical reagents are all of analytical grades. All solutions used in this study were prepared with deionized water.</p><!><p>The fiberboard biomass samples used in this study were sourced from Dare Wood-Based Panels Group (Jiangsu, China). Then, the biomass samples were cut, ground, and sieved to a size of 74–200 μm. The biochar samples were produced by slow pyrolysis of fiberboard biomass using a laboratory-scale tube furnace (YGDL-1200, Shanghai yuzhi electromechanical equipment Co., Ltd, China) at 300, 500, and 800°C, respectively. All pyrolysis experiments were performed at a heating rate of 10°C/min, a holding time of 1 h, and a N2 flow rate of 0.5 L/min. Hereafter, the biochar samples are referred to as BCXXX, where the prefix "BC" denotes biochar while the suffix "XXX" represents the pyrolysis temperature (in degrees Celsius).</p><p>The BC samples were subjected to the following characterizations. The elemental analysis of biochar was performed via an elemental analyzer (Perkin-Elmer 2400 Series II, USA). Scanning electron microscopy (SEM; JSM-7600F, Japan Electronics, Japan) and X-ray diffraction (XRD; Ultima IV, Rigaku, Japan) were adopted to observe the surface morphology and the crystalline structures of biochar samples. The functional groups of biochar were studied via FTIR spectrometry (Vertex 80V, Bruker, Germany). The surface elemental composition of biochar was determined by XPS (AXIS Ultra DLD, Shimadzu, UK). BET surface area and pore size distribution of the biochar samples were measured on an autosorb specific surface area analyzer (Quantachrome, USA) with N2 as the adsorbate at 77 K.</p><p>The pH of zero point charges (pHPZC) of selected biochar samples were recorded based on a method detailed elsewhere (Liu et al., 2011; Jang et al., 2018). In brief, 50 ml of NaCl solutions (0.01 M, to maintain the ion strength of the solution) with pH ranging from 2 to 10 (adjusted by adding appropriate amount of 0.5 M HCl or NaOH solution) was placed in 250-ml conical flasks. Then 0.1 g of biochar was added to the solution followed by purging N2 gas to eliminate the effect of CO2. The final pH of solution was recorded after stirring for 48 h at ~25°C in a sealed flask. Finally, the pHpzc was calculated based on the ΔpH = 0, where ΔpH is equal to the final pH minus the initial pH. All experiments were performed at least twice.</p><!><p>A stock TC solution of 500 ppm was prepared by dissolving 0.05 g of TC together with 0.025 g of NaOH (to increase its solubility) in 100 ml of water. Then, the stock solution was diluted to desired concentrations with pH adjusted to the desired value for adsorption experiments. It should be noted that all TC solutions contain 0.1 M of NaCl to maintain the ionic strength, if not specified. For the adsorption experiments, 0.25 g of biochar sample was added into 100 ml of TC solution (20 ppm, pH = 7) in a 250-ml conical flask and stirred for 96 h at ~25°C. At certain time intervals, ~5 ml of solution was withdrawn and filtered by a 0.22-μm Millipore filter. The concentrations of TC were determined by a UV-Vis spectrophotometer (WFZ UV-2000, Unico, USA) at 360 nm. It should be noted that to better determine the concentration of TC solution, different standard curves at different pH values (i.e., 2–10) were adopted. The pH of the filtrates was also recorded. Since the adsorption equilibrium is achieved after 48 h, all following adsorption experiments were conducted for 48 h. For studying the influence of ionic strength, the concentration of NaCl was adjusted from 0.0 to 1.0 M. To study the effect of biochar dosage, the biochar dose ranged from 1.5 to 3.5 g/L. When examining the role of pH values, the initial pH of TC solution was set from 2 to 10. The initial TC concentration was varied from 2.5 to 60 ppm for obtaining the adsorption equilibrium isotherm. The removal efficiency was calculated based on the following formula:</p><p>where R is removal efficiency and C0 and Ce are the initial and equilibrium concentration of TC (mg/L).</p><p>The adsorption capacity qe was calculated according to the following formula:</p><p>where C0 and Ce are the initial and equilibrium concentration of TC (mg/L), V is the volume of TC solution (L), m is the weight of biochar (g), and Qe is the adsorption capacity (mg/g).</p><p>The Langmuir and Freundlich isotherm modes were adopted to evaluate the reaction behavior between TC and biochar, which can be expressed by the following equations:</p><p>where Ce is equilibrium concentration of TC (mg/L), qe is the adsorption capacity (mg/g), qmax is the maximum adsorption capacity (mg/g), KL (L/mmol) and KF [(mmol/g)·(L/mmol)1/n] are Langmuir and Freundlich constants, respectively, and n is another Freundlich constant that is related to adsorption intensity.</p><p>A thermodynamic study was completed by using temperatures of 25, 35, and 45°C. Equations (5)–(7) were applied to determine the change in the standard Gibbs free energy (ΔG0), enthalpy (ΔH0), and entropy (ΔS0), respectively:</p><p>where KC (dimensionless) means the apparent equilibrium constant; R is the gas constant, which is 8.314 J/mol/K; and T is the absolute temperature (K). When plotting lnKC against 1/T, a straight line can be found with ΔH0 and ΔS0 being the intercept and slope, respectively. All experiments were again repeated at least twice.</p><!><p>Table 1 shows that the ash content of biochar samples gradually increases from 5.8% of BC300 to 7.9% of BC800. The higher inorganic content for BC800 is due to the enhanced decomposition of organic components in fiberboard biomass at a higher pyrolysis temperature. The elemental compositions of biochar samples derived from fiberboard biomass are also shown in Table 1. As can be seen, the C content (83.5–87.6%) in biochar samples prepared at different temperatures are relatively high, while both the H (0.9–3.0%) and O content (8.0–10.0%) are very low. It should be noted that the N contents of biochar samples prepared from fiberboard biomass in this study are 3.2–3.5%, which are apparently higher than biochar samples derived from other biomass feedstock. This can be attributed to the presence of glue containing urea in fiberboard. Another phenomenon is that with increasing pyrolysis temperature, C content slightly rises while both H and O contents decrease. This indicates a higher carbonization degree of biochar with less hydrophilic surfaces (Wang et al., 2017). As a result, the H/C reduces from 3.6 to 1.1 and the O/C gradually reduces from 12.0 to 9.2, indicating a higher aromatization degree and the graphite-like structure of higher temperature biochar (Selmi et al., 2018; de Jesus et al., 2019).</p><!><p>Proximate and elemental analysis of biochar samples.</p><p>As received basis.</p><p>Dry and ash-free basis.</p><p>By difference.</p><p>Atomic ratio.</p><p>BCXXX stands for biochar prepared at XXX°C; XXX can be 300, 500, and 800.</p><!><p>FTIR spectra in Figure 1 also prove the existence of a graphite-like structure as well as oxygen-content functional groups. The peaks at wavenumbers of 1,114 and 1,400 cm−1 are attributed to alcoholic C–O and C–N stretching, respectively (Pan et al., 2018). The strong peak at 1,630 cm−1 is assigned to ether the aromatic stretching of C–C groups or the stretching of C=O, while the peaks at wavenumbers 2,923 and 2,854 cm−1 are due to the stretching vibrations of C–H groups (Liu et al., 2012; Ahsan et al., 2018; Pan et al., 2018). The broad band at 3,438 cm−1 could be assigned to the overlapping of –OH and –NH stretching (Pan et al., 2018; de Jesus et al., 2019).</p><!><p>FTIR spectra of biochar samples prepared at 300, 500, and 800°C. BCXXX stands for biochar prepared at XXX°C; XXX can be 300, 500, and 800.</p><!><p>To further probe the surface elemental composition and the contents of both O-containing and N-containing functional groups, XPS analysis on biochar samples were carried out (see Table 2 and Figure S2). The surface C content increases from 76.8% of BC300 to 86.4% of BC800, while the surface O content decreases from 20.4 to 11.3%. These results suggest a higher aromatization degree under higher pyrolysis temperature, which is consistent with elemental analysis aforementioned. It is obvious that the total content of O-containing functional groups (i.e., C–O and C=O groups) becomes lower from 35.5 to 19.4%. According to previous studies (Liu et al., 2011; Wang et al., 2017; Jang et al., 2018; de Jesus et al., 2019), both the graphite-like structure and the O-containing functional groups contribute to the adsorption performance of biochar. Moreover, there are four types of N-containing functional groups: pyridinic nitrogen (N-6), pyrollic nitrogen (N-5), quaternary nitrogen (N-Q), and oxidized nitrogen (N-O). All N-containing groups are located at the edges of the graphene structure, except N-Q (Wu et al., 2013). It can be seen in Table 2 that the dominant type is N-5 followed by N-6 and N-Q in sequence. As reported, N-containing groups may contribute to the adsorption performance of biochar (Wu et al., 2012, 2013, 2014; Zhang et al., 2014; Zhan et al., 2019).</p><!><p>Elemental composition, oxygen-containing functional groups, and nitrogen-containing functional groups of biochar surface from XPS analysis.</p><p>BCXXX stands for biochar prepared at XXX°C; XXX can be 300, 500, and 800.</p><!><p>The surface morphology of biochar samples is presented in Figure S1 and the surface area is listed in Table 3. As can be observed, the surfaces of biochar samples are not very rough with its surface areas being very low (32.2 and 34.9 m2/g for BC300 and BC500, respectively). BC800 seems to show more coarse surface texture than others, which is consistent with the results of BET analysis. The surface area of BC800 significantly improves to 135.1 m2/g due to the server decomposition of organic components in biomass feedstock at elevated temperature (Lian et al., 2014). A similar trend is also found for the pore volume, and both trends remain consistent with previous reports (Aller, 2016; Wang et al., 2017). The higher surface area and pore volume are particularly favorable when using biochar as adsorbents (de Jesus et al., 2019). Taking all above characterizations into consideration, biochars prepared from fiberboard biomass could be potential adsorbents, especially BC800.</p><!><p>Pore structure of biochar samples.</p><p>BCXXX stands for biochar prepared at XXX°C; XXX can be 300, 500, and 800.</p><!><p>Figure 2A presents the adsorption capacity of biochar samples prepared at different pyrolysis temperatures on TC adsorption. It clearly shows that biochar prepared at a higher pyrolysis temperature performs a better adsorption capacity on TC. As aforementioned (see section Characteristics of Biochar), BC800 has a higher aromatization degree and a bigger surface area, compared to BC300 and BC500. A previous study has also proved a positive influence of the surface area of biochar on its adsorption capacity (Wang et al., 2017). However, the differences in adsorption capacity among three biochars are vastly different from the difference in their surface areas, implying that the measured BET surface area may not be the main factor in determining the adsorbing ability. Based on the properties of biochar, possible adsorption mechanisms include pore-filling, hydrogen bonds, hydrophobic effect, electrostatic interactions, and π-π interactions (Wang et al., 2017; de Jesus et al., 2019). It has been reported that π-π interactions are one of the major mechanisms governing TC adsorption by biochar (Wang et al., 2017). Biochar serves as a π-electron donor, which could be attributed to its graphite-like structure while TC acts a π-electron acceptor due to its aromatic ring structure (Wang et al., 2017; de Jesus et al., 2019). A higher pyrolysis leads to a higher degree of graphitization of biochar samples. The crystallinity of obtained biochar samples was characterized by XRD analysis, as is shown in Figure S3. Three biochar samples exhibit a broad diffraction peak at 23.5° with the intensity of BC800 being the highest, which is attributed to the (002) crystal plain of graphitic structure. BC800 also exhibit a board diffraction peak at 44.3° with low intensity, which is assigned to the (100) crystal plain of graphitic structure. XRD analysis confirms the high graphitization degree of BC800. Furthermore, the abundant N-6 in BC800 (see Table 2) also contributes to its high absorbing ability as N-6 possesses an unshared pair of electrons, thus enhancing its performance as electron donor when interacting with TC. It has been reported that the adsorption energy of morpholine with N-6 is much higher than those with other types of N-containing functional groups (e.g., N-5, Q-N, etc.) (Li et al., 2019). Therefore, BC800 shows the best adsorption capacity with removal efficiency being 68.6%. It is also reported that O-containing functional groups can act as hydrogen-bond acceptors, thus increasing the adsorption capacity. However, in this study, BC300 possesses the highest content of O-containing functional groups while having the lowest adsorption capacity. This indicates that the hydrogen bond interactions do not dominate the adsorption behavior. Hereafter, only BC800 is chosen to use as adsorbent. It is also worth noting that when the adsorption time is longer than 48 h, the increase of removal efficiency becomes extremely slow, which means that the adsorption equilibrium is approached after 48 h. The final pH of the solution also remains stable after 48 h (see Figure 3B). Hence, an adsorption time of 48 h is adopted for the following studies.</p><!><p>Adsorption performance of (A) BC300, BC500, and BC800 on TC; (B) BC800 on TC with different concentrations of NaCl; (C) BC800 on TC with different biochar dosage; (D) BC800 on TC with different initial concentration. BCXXX stands for biochar prepared at XXX°C; XXX can be 300, 500, and 800.</p><p>(A) pHPZC of BC800; (B) the pH of TC solution with adsorption time; (C) adsorption performance of BC800 on TC solution with different initial pH; and (D) the final pH of TC solution after adsorption benchmarking against its initial pH. pHPZC stands for the pH of zero point charges; BC800 stands for biochar prepared at 800°C.</p><!><p>The real wastewater system is complex and often contains salts besides organic pollutants (Liu et al., 2019). Besides, in this study, NaOH and HCl are used to adjust the pH of TC solution, so the TC solution contains NaCl. The existence of salts may influence the removal of pollutants. Therefore, it is necessary to investigate the effect of ionic strength, and the results are shown in Figure 2B. It is obvious that the removal efficiency is similar (59.9–63.6%) with the concentration of NaCl ranging from 0 to 1.0 mol/L. This suggests that the existence of salts does not obviously affect the removal process. To further eliminate the influence of ionic strength, 0.1 mol/L of NaCl is added to TC solution for the following studies.</p><!><p>Figure 2C presents that increasing biochar dosage from 1.5 to 2.5 g/L greatly improves removal efficiency from 44.2 to 62.3%, while further increasing biochar dosage to 3.5 g/L only slightly improves removal efficiency to 67.6%. This improvement can be attributed to the more adsorption sites from the enlarged surface area of adsorbent at a larger dosage level (Ahsan et al., 2018; Alidadi et al., 2018). The slow increase of removal efficiency after 2.5 g/L may be due to the agglomeration of biochar particles that reduces the total effective surface area and thereby reduces total sorption sites (Ahsan et al., 2018; Alidadi et al., 2018). As a result, 2.5 g/L of BC800 dosage is employed in this study.</p><!><p>A negative correlation between TC removal efficiency and its initial concentration is observed in Figure 2D. When the TC initial concentration is lower than 10 ppm, more than 95% TC is removed. Increasing TC initial concentration to 20 ppm leads to a significant reduction of removal efficiency to ~60%. The removal efficiency slowly reduced from 44.2 to 27.6% with TC initial concentration growing from 30 to 60 ppm. This could be attributed to the restriction of the adsorption process at a high antibiotic loading level caused by the limited number of effective adsorption sites (Marzbali et al., 2016; Ahsan et al., 2018; Alidadi et al., 2018). The alleviated negative effect of initial TC concentration on the removal efficiency at high concentrations (20–60 ppm) may come from the increased TC concentration gradients between liquid phase and solid surfaces (Ahsan et al., 2018; Alidadi et al., 2018). An initial TC concentration of 20 ppm is used in this study.</p><!><p>Solution pH is also an important parameter to consider for the effective adsorption process, which affects the properties of both adsorbents and pollutants (Liu et al., 2012; Marzbali et al., 2016; Ahsan et al., 2018; Alidadi et al., 2018; Jang et al., 2018; Selmi et al., 2018; Jang and Kan, 2019; Nguyen et al., 2019; Premarathna et al., 2019). The pHPZC of BC800 was 8.27 (see Figure 3A), which is relatively high. This may be due to the presence of N-containing groups that lead to BC800 being more alkaline (Wu et al., 2012). The surface of BC800 is negatively charged at pH lower than 8.27 while positively charged at pH higher than 8.27. Based on the pKa values studied previously (Marzbali et al., 2016; Selmi et al., 2018), TC molecules are positively charged when pH < 3.3, neutrally charged when pH ranges from 3.3 to 7.8, and negatively charged when pH > 7.8. It should be noted that the pH value dramatically rises to 10.7 within 10 min and then significantly reduces to ~8.35, which is similar with the pHPZC (8.27) after 48 h and finally stable at this pH value (see Figure 3B). A similar trend was also found when the initial pH is 3–10. When the initial pH is 2, its final pH slightly increases to 2.2, as can be seen in Figure 3D. This suggests that the biochar samples serve as a buffer that releases some acid matters or alkali matters to react with NaOH or HCl. Consequently, both the TC molecules and biochar particles are charged ether positively (initial pH is 2) or negatively (initial pH is 3–10). It has been reported that if both the adsorbent and the adsorbate are negatively or positively charged, the adsorption process will be repelled due to the electrostatic repulsion (Li et al., 2018). However, in this case, the maximum removal efficiency (90.9%) was achieved at a pH value of 2, as can be seen in Figure 3C. The removal efficiency dramatically decreases to 83.0% (pH = 3) and then gradually reduces to 62.3% with pH increasing to 7, followed by a gradually increase to 67.1% with pH increasing. The results indicate that electrostatic attraction does not play an important role in the TC adsorption. Similar phenomenon also found in other studies and other mechanisms should be mainly responsible for TC adsorption, such as graphene-like structure in adsorbents (Li et al., 2018). Overall, natural environmental pH is appropriate for TC adsorption by BC800.</p><!><p>Both the Langmuir and Freundlich models were adopted to study the adsorption isotherms with results shown in Figure 4. As can be seen, the R2 value of the Langmuir model is 0.85 and is lower than that of the Freundlich model, which is 0.94. This suggests that both models are suitable to predict the experimental data although TC adsorption is better explained by the Freundlich model. The Freundlich model describes both physical and chemical adsorption onto heterogeneous surfaces (McKay et al., 1982; Alidadi et al., 2018). The Freundlich consistent n is 5.14, which means adsorption is favorable and BC800 is not so heterogeneous on its surface (Alidadi et al., 2018). A KF value of 3.27, which is related to adsorption capacity, also supports the adsorption process. The Langmuir model explains monolayer adsorption onto homogeneous surfaces (Doltabadi et al., 2016; Alidadi et al., 2018). A dimensionless equilibrium parameter RL can explain the nature of the Langmuir model, which is favorable (0 < RL < 1), unfavorable (RL > 1), irreversible (RL = 0), or linear (RL = 1), which can be calculated using the following equation:</p><p>The KL value is 3.79, which makes the RL 0.16. This indicates that BC800 is in favor of TC adsorption with maximum adsorption capacity being 6.37 mg/g. Taking the above considerations together, it can be concluded that the surfaces of BC800 are neither very heterogeneous to follow Freundlich model nor entirely homogenous to follow Langmuir model.</p><!><p>Sorption isotherms of TC using BC800 (T = 298 K).</p><!><p>As is known to all, temperature significantly affect the adsorption process. TC adsorption was determined at 25°C (298 K), 35°C (308 K), and 45°C (318 K) and the results are presented in Figure 5. It is obvious that with adsorption temperature rising, the removal efficiency improves greatly from 62.3% at 25°C to 84.9% at 45°C. This implies TC adsorption process is endothermic which agrees with other reports (Wang et al., 2017; Ahsan et al., 2018; Selmi et al., 2018). It is also proved by a positive ΔH0 value which is 48.0 kJ/mol (see Table 4). This result also indicates that a higher temperature is more satisfactory for adsorption process, which may be attributed to the improvement in diffusion rate of TC (Wang et al., 2017). Similar to ΔH0, ΔS0 is also positive with its value being 157.1 J/mol/K. This reveals that randomness at the TC-biochar interface is higher compared to concentrated aqueous phase (Wang et al., 2017; Selmi et al., 2018). As to ΔG0, its value is positive but very low at 298 K (1.02 kJ/mol) and then turns negative with the magnitude of ΔG0 rising with sorption temperature becoming higher. This again suggests that the sorption process is more thermodynamically preferable at a higher temperature. Basically, the ΔG0 value of physical sorption is in the range of 0–20 kJ/mol. Therefore, the adsorption process in this study is mostly physical in nature with the value of ΔG0 varying from 1.02 to −2.14 kJ/mol.</p><!><p>(A) Adsorption performance of BC800 on TC solution at 25, 35, and 45°C; (B) 1/T vs. lnKC plot. BC800 stands for biochar prepared at 800°C.</p><p>Thermodynamic parameters for the adsorption of TC on BC800 at different temperatures.</p><!><p>Compared with other biochar samples, the adsorption capacity of BC800 on TC in this study is not very advantageous (see Table S1). For example, Wang et al. could achieve a TC adsorption capacity of 13.85 mg/g using rice straw-derived biochar, while the similar performance of biochar prepared from the sewage sludge was also reported by Yang et al. At similar pyrolysis temperatures, agricultural residue (e.g., rice straw) could be decomposed more significantly than wood-based materials, thus leading to a better porous structure. One of the disadvantages of using straws as precursor for preparing functional carbon materials is the high content of inorganics, which will cause difficulty in further modification (e.g., carbonization and/or graphitization) and also may lead to a secondary pollution to environment. The high adsorption capacity from the sludge is due to the activation process by the ferric compounds. Comparatively, this study uses waste fiberboard to prepare biochar adsorbent containing limited inorganic metals with potentials to be further modified and upgraded by simple methods. Additionally, the fiberboard featuring high contents of heteroatoms (e.g., N and O) may endow the resulting biochars with special characteristics when being further carbonized. Therefore, this work is a preliminary study on exploring the effectiveness of fiberboard-derived biochars as adsorbents for organic pollutants. Subsequently, more investigation by fine modifications would be necessary to examine its potentials as environmental remediating green materials in the near future.</p><p>As discussed above, the adsorption performance of BC800 was mainly attributed to its relatively high surface area, the π-π interactions with TC, as well as the high content of N-6 as electron donor. The adsorption mechanism could thus be proposed as shown in Figure 6. This study has clearly indicated the feasibility of volarizing waste fiberboard into valuable products and the important role of inherent heteroatoms in the biomass.</p><!><p>Proposed mechanisms of the adsorption process.</p><!><p>This study shows that the biochars derived from waste fiberboard biomass can be used as adsorbent for TC removal, especially BC800 with removal efficiency being 68.6%. This is attributed to its higher aromatization degree and a bigger surface area. The π-π interactions between the graphite-like structure of biochar and the ring structure of TC dominate the adsorption mechanism. The high content of N-containing groups (especially N-6) also contributes to the adsorption performance of biochar. The ionic strength plays an insignificant role in the adsorption process, while both the biochar dosage and the initial TC concentration significantly affect the removal efficiency. The maximum adsorption capacity is obtained at a pH of 2. The results suggest that electrostatic attraction has limited influence on the adsorption process and natural environmental pH is appropriate for TC adsorption. An isotherm study indicates that the Freundlich model fits better than the Langmuir model. Thermodynamic analysis shows that both the value of ΔH0 and ΔS0 are positive, which suggests that the adsorption process of TC on biochar is thermodynamically favorable. The adsorption process is mostly physical adsorption due to the very low values of ΔG0.</p><!><p>All datasets generated for this study are included in the article/Supplementary Material.</p><!><p>SZ, WG, and DX: conceptualization. SZ and WG: methodology and funding acquisition. YG and DX: validation. WG and DX: formal analysis. YG, KK, and ZL: investigation. SZ: resources and project administration. WG and HZ: data curation. DX and WG: writing—original draft preparation and visualization. XH, HS, KK, and SS-H: writing—review and editing. WG and SZ: supervision.</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>Funding. This work was financially supported by the National Natural Science Foundation of China (Grant No. 51876093), the National Key R&D Program of China (Grant No. 2018YFE0183600), and the Start-up Fund for Scientific Research of Nanjing Forestry University (Grant No. GXL2018033).</p><!><p>The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fchem.2019.00943/full#supplementary-material</p><!><p>Click here for additional data file.</p>

PubMed Open Access

Coupling of the Non-Amyloid-Component (NAC) Domain and the KTK(E/Q)GV Repeats Stabilize the \xce\xb1-Synuclein Fibrils

The aggregates of \xce\xb1-synuclein (\xce\xb1S) are a major pathological hallmark of Parkinson\xe2\x80\x99s disease (PD) making their structure-function relationship important for rational drug design. Yet, the atomic structure of the \xce\xb1S aggregates is unavailable, making it difficult to understand the underlying aggregation mechanism. In this work, based on available experimental data, we examined plausible molecular structures of \xce\xb1S(20/30\xe2\x80\x93110) fibrils for the first time by employing computational approaches. The optimized structure was used to investigate possible interactions with aggregation inhibitors. Our structural models characterize the essential properties of the five-layered fold of the \xce\xb1S fibril. The distribution of the \xce\xb2-strands and the topology of the five \xce\xb2-strands in the relatively stable models are in good agreement with experimental values. In particular, we find that the KTK(E/Q)GV repeat motifs significantly stabilize the \xce\xb1S fibrils. The charged residues within each repeat prefer exposure to the solvent in order to further stabilize the inter-layered interactions by salt-bridges. The organization of the repeat K58T59K60E61Q62V63 between the \xce\xb22 and \xce\xb23 layers significantly affects the stability of the non-amyloid-component (NAC) domain. The coupling between the NAC domain and the KTKEGV repeats indicates that both regions can be potential binding sites for inhibitor design. The distinct binding modes of chemical agents that alter \xce\xb1S aggregation highlight the potential of our models in inhibitor design.

coupling_of_the_non-amyloid-component_(nac)_domain_and_the_ktk(e/q)gv_repeats_stabilize_the_\xce\xb1

1. Introduction<!>2.1 The organization of charged repeats is crucial for the stability of \xce\xb1S fibril models<!>2.2 The relatively stable models characterize the essential secondary and tertiary structures of the \xce\xb1S fibril as observed experimentally<!>2.3 Effects of charged repeats on the stability of NAC<!>2.4 Aggregation inhibitors block the NAC region<!>3. Conclusions<!>4. Materials and Methods

<p>α-synuclein (αS) is a 140-residue protein that is primarily expressed at presynaptic terminals of neuronal brain cells [1, 2]. The aggregation of αS is closely related to several neurodegenerative diseases including Parkinson's disease (PD) [1, 3]. Recently it was found that αS was present in aggregates of the Krabbe disease and apparently prone to fibrillization in the presence of psychosine [4]. αS adopts different conformations under various environments. Under physiological conditions, αS is characterized as an intrinsically disordered protein, exhibiting a random coil conformation in aqueous solution [3]. In the presence of the membrane, αS displays an α-helical secondary structure as observed from solution NMR structure of micelle-bound αS [5, 6]. Full-length αS can be divided into three distinct domains, the N-terminal domain (residues 1–60) which is amphipathic and responsible for binding to phospholipids [7]; the non-amyloid-β component (NAC) domain (residues 61–95) which is highly hydrophobic and assumed to be essential for αS aggregation [3, 8], and the acidic negatively charged C-terminal domain (residues 96–140) believed to be critical for the chaperone-like activity of αS [9, 10]. The self-assembly of αS is accompanied by conformational transitions [11]; however, the molecular mechanism remains largely unknown partly due to the lack of molecular structures of αS aggregates.</p><p>The structural features of αS fibrils have been extensively studied experimentally [12–18]. These studies reveal that both the N-terminal (~20 residues) and C-terminal (~35–40 resides) are highly dynamic and unfolded, whereas ~70 residues have been identified in the ordered fibril core. The folded fibril core of αS shows similar in-register parallel β structures as those observed in the fibrillar structures of amyloid β associated with Alzheimer's disease [19]. In an earlier study based on site-directed spin labeling and electron paramagnetic resonance spectroscopy, a five-layered model of the αS fibril has been derived by the Langen group. There are five β-strands connected by four loops in this folding pattern (β1-loop-β2-loop-β3-loop-β4-loop-β5) [17]. Later, a similar fibril model was proposed by the Riek group based on NMR experimental studies of αS(30–110) [13]. The five β-strands involve residues 37–43, 52–59, 62–66, 68–77, and 90–95, respectively (Fig. 1). The assignments of β-strands in Riek's model are largely in agreement with the study of Heise et al.[12] and twisting assemblies of αS in the study of Melki group [16]. The propensity of β1 and β2 is in line with a β-hairpin of αS(35–56) trapped by a β-wrap protein where two β-strands (V37–K43, V48–T54) are connected by a turn (T44–G47) [20]. The β-strand propensity of residues 68–77 is also consistent with previous experimental results which identified residues 71–82 as essential for the αS filament assembly [21]. However, due to different experimental conditions and sample preparation as well as the polymorphic property of the αS fibril structures, an exact consensus from the different experiments cannot be achieved. For example, both studies of Heise and Melki group demonstrate that different strains display distinct distribution of β-sheets along αS sequence [12, 16]. In the NMR study of Wu et al., the C-terminal region (residues 100–140) shows a high β-strand propensity at low temperature (263 K) [18]. In addition, another NMR study at 310 K indicates that the N-terminal of αS fibrils assume a certain extent of β-sheet structures [15].</p><p>These experimental studies greatly enhanced the understanding of the structure and function of αS fibrils. A partial structural model of NAC (E61–V95) has been proposed recently by computational modeling and simulation [22]. To elucidate the molecular mechanism of αS misfolding and aggregation we aim to construct atomic models of αS fibrils using computational methods based on reported NMR secondary and tertiary structures. We focus on the fold of full-length αS fibrils not only because there are reported experimental results available, but also because there are six familial PD mutations (A30P [23], E46K [24], H50Q [25], G51D [26, 27], A53T [28], and A53E [29]) that have been identified in the N-terminal region. These point mutants can affect membrane binding and aggregation kinetics of αS [30–34]. It is thus important to include the N-terminal region in the structural model. In particular, we carefully examine the role of the imperfect repeat of the KTK(E/Q)GV motif in the stabilization of αS fibrils. There are six such repeats in both the N-terminal and NAC region, which has been suggested to be crucial for lipid binding and aggregation [35–38]. Since both NAC and these repeat are important for αS aggregation, it is also interesting to examine their coupling effect within αS fibrils. Here we propose the first atomic structural models of αS(20/30–110) fibrils, which provide insights into the polymorphic behavior of αS fibrils, the organization of these repeats in the context of fibrils, and the role of the coupling of NAC and these repeats in the stability of αS fibrils.</p><!><p>In the first five-layered αS pentamer structure (M1) (Fig. 2A), we tested the position of residues K43, K45, and E46 between β1 and β2 layers. The possibility of forming E35-K58 salt-bridges to stabilize both β1 and β2 was examined by positioning K60 between β2 and β3.</p><p>However, after 20-ns MD simulation, the structure collapsed. The layer of β3 moves far away from β4 indicating unfavorable interactions between these two layers. The turn/loop connecting β1 and β2 becomes disordered possibly due to the competing electrostatic interactions between K43 and K45 with E46. Similarly, the loop containing E57, K58, E60, and K61 also becomes disordered. As a result, no stable salt-bridge was observed during the MD simulations. This structural model has the highest GBMV energy (−8054 kcal/mol, Fig. 3) among all structures investigated, implying it is the least stable αS fibril model. Also, it was noted that this fibril model has the lowest β-strand content in its NAC domain (~17%, Fig. 4).</p><p>In the second fibril model (M2) (Fig. 2B), we check the possibility of exposing both K43 and K45 to the solvent in order to eliminate the electrostatic repulsion present in M1 but still keep E46 inside the fibril layer. To increase the interaction between different layers, E35 and K58, E61 and K80 initially form salt-bridges, as well as E57 and K60. Different from M1 where H50 was oriented to β3, to examine the interactions between β2 and β3 in this model, H50 was oriented to β1. After 20-ns MD simulation, this model is even less stable than the M1 model, although the interactions between β1 and β2, and between β3 and β4, seem more favorable than those in M1, possibly due to the introduction of inter-layered salt-bridges. The separation of β2 from β3 could indicates that H50 should be oriented to β3 instead of β1 even though this model has a relatively lower GMBV energy (−8212 kcal/mol) than M1, which suggests that inter-layered salt-bridges contribute to the stability of αS fibril structures. In addition, the β-strand in the NAC domain increased to ~25%, indicating that the interaction between E61 and K80 may stabilize NAC.</p><p>To further investigate the role of salt-bridges in the loop containing K43, K45 and E46 as observed in the M2 structure, we examined the M3 model (Fig. 2C) where K43 and E46 form a salt-bridge but not K45, which was kept exposed to the solvent. After 20-ns MD simulation, β2 also moves away from β3 as observed in M2. However, the introduction of the salt-bridge between K43 and E46 seems not to be able to stabilize this loop. The structures of β1 and β2 display a globular conformation instead of extended β-strands. Compared to M2, such results imply that the formation of the salt-bridge between K43 and E46 inside the fibril layer will not contribute to the stability of αS fibril structures. As a consequence, this model has a comparative GBMV energy (−8208 kcal/mol) to that of M2. The results are also consistent with the familial variant of E46K promoting amyloid formation [39–41]. If the M3 model contribute to αS fibril formation, the E46K mutation should destabilize the αS fibril. Interestingly, this fibril model has the highest percentage of β-strand in the NAC domain (~28%).</p><p>The above results suggest that the repeat K43T44K45E46G47V48 may not form stable salt-bridges inside the fibril layer; as such, we constructed the M4 model where K43, K45 and E46 are exposed to the solvent (Fig. 2D). The organization of the remaining repeats is the same as that in the M3 structure except that H50 was oriented to β3. This model was also used to examine the effect of another charged repeat K58T59K60E61Q62V63 on the stability of αS fibril. After 20-ns MD simulation, the five-layer fold of the αS fibril structure is retained. The repeat K43T44K45E46G47V48 adopts primarily a loop conformation; however, the β-strand of β2 disappears, which could be attributed to the unfavorable electrostatic interactions in the repeat involving E57, K58, K60, and E61. It could also be due to the interactions between β1 and β2. The calculated GBMV energy (−8159 kcal/mol) suggests that it is less stable than M2 and M3, although this model could maintain the fold of the αS fibril structures. In addition, the β-strand content (~22%) in the NAC of this model is also slightly lower than that of M2 (~25%) and M3 (~28%).</p><p>Based on M4, we further examined model M5 (Fig. 2E) where both E57 and K58 were exposed to the solvent, so that initially there is no salt-bridge between E57 and K60. Only K60 was kept inside the layer as in M4. We observed that the β-strand of β2 remains relatively stable during 20-ns MD simulation, suggesting that the presence of such salt-bridge between E57 and K60 could affect the stability of the β2 strand. However, the presence of K60 inside the fibril core has a significant effect on the fold of the αS fibril as β2 separates from β3. We speculate that the absence of counter ions results in strong electrostatic interactions between adjacent K60 in different β layers, and the interactions between β2 and β3 are not able to counteract such effect. Other possibility is that neighboring charged residues like K80 could also disrupt the interactions of β2 and β3. Consequently, the M5 model has a relatively high GBMV energy (−8095 kcal/mol, Fig. 3). Unlike M3 where E57 interacts with K60, the β-strand of NAC in this model (~19%) is lower than that of M3 (~28%), implying that electrostatic interactions have a more significant effect on the stability of the NAC domain.</p><!><p>In M5, we examined the possibility that K80 could interact repulsively with K60 and then destabilize the interactions between β2 and β3, although it initially forms a salt-bridge with E61. M6 was used to explore the possibility that K80 establishes a salt-bridge with E83 inside the fibril core (Fig. 5). To increase the hydrophobic interactions between β4 and β5, the orientation of β4 was reversed, leading T75 and A78 to have opposite orientation to that shown in M5. The fold of αS fibril was maintained after 20-ns MD simulation. We then extended the simulation to 40 ns and found that this model structure is relatively stable. However, this fibril model has a higher GBMV energy (−8172 kcal/mol) even than M2 (−8212 kcal/mol, Fig. 3) where the fibril fold collapses. Note that the percentage of β-strand in the NAC domain of M6 is comparable (~23%) to that of M2 (~25%).</p><p>Although M6 indicates that the fold of the αS fibril structures could be maintained in the presence of inter-layered salt-bridges, such a model structure is energetically less stable. Results from M1 to M3 suggest that charged residues (K43, K45, and E46) in the first loop have a strong tendency to be exposed to the solvent. Taken together, the results of M4 to M6 lead us to conclude that charged resides in all three loops prefer to be exposed to the solvent. We then examined a model M7 (Fig. 6A) that satisfies all the inter-layered distance restraints, and the stabilization factors derived from M1 to M6. Initially, K45 and E46 interact with each other in the loop connecting β1 and β2 due to electrostatic attraction. Similarly, E57 interacts with K58 in the loop connecting β2 and β3. In addition, E35 in β1 may participate in the salt-bridge to further stabilize the interactions between β1 and β2. Similarly, E61 and K80 form a salt-bridge, with K60 also possibly involved. Two independent 40-ns MD simulations were performed (referred as M7a and M7b thereafter) with the same initial conformation but different velocity distribution. Both fibril structures could maintain the αS fibrils fold after 40-ns simulation (Fig. 6B and 6C). Their GBMV energies suggest that both models converged with similar energies, with M7a (−8313kcal/mol) is slightly more favorable than M7b (−8288 kcal/mol, Fig. 3). The two models also display similar percentages of β-strand in their NAC domains (~24% and 25% for M7a and M7b, respectively).</p><p>Based on the above three relatively stable fibril structures (M6, M7a, and M7b), we examined the change of the secondary structures during the MD simulations (Fig. S1 and Fig. S2). The average β-strand content of these pentamers is about 39%, 36%, and 33% for M6, M7a, and M7b, respectively; that of the inner core is about 39%, 38%, and 37% for M6, M7a, and M7b, respectively, which is close to the β-strand population (~43%) determined by Riek's group [13]. The distribution of the β-strand motif for these models was also calculated and summarized in Fig. 7 and Fig. S2. All fibril models display similar distributions of five β-strands as derived from solid-state NMR experiments. The first β-strand comprises residues ~37–41 as indicated by NMR data, however, both M6 and M7b also show a short β-strand consisting of T44–E46, which seems too short to be identified experimentally. Solid-state NMR data suggests that the second β-strand comprises residues ~52–58, whereas all models predict that it contains residues ~49–54. The layout of β1 and β2 in our models is in excellent agreement with the β-hairpin of αS(35–56) trapped by a β-wrap protein where two β-strands (V37–K43, V48–T54) are linked by a turn (T44–G47) [20]. However, it should be noted that β2 lies in the layer involving E57 and K58. Although we have carefully examined the interactions especially of the salt-bridges in this loop, we cannot exclude other possible arrangements of these charged residues that could form stable β-strands. All three models characterize the third β-strand including residues ~62–66, which is in good agreement with the experimental result. As for the fourth β-strand, only M7a exhibits a short β-strand comprising residues 70–72, consistent with NMR data. Interestingly, residues 70–72 in our fibril M7a form a turn connecting β3 and β4. Such arrangement puts residues 72 and 73 in a loop, in agreement with the proposed five-layered β-sandwich by Riek group [13]. For the last β-strand that starts from A90 and ends at V95, a good agreement was obtained between the experimental result and both M6 and M7a.</p><p>The tertiary structure of these fibril models was characterized through calculation of the inter-layered distance and compared with available experimental results. However, because the ssNMR experiments were performed at a rather low temperature (273–283 K), and the αS fibril samples were in a dried gel-like state [13], direct comparison between our results from MD simulations performed in aqueous solution at 310 K and experimental data is not straightforward. Here, we compared the most important inter-layered residue distance and validated the consistency between our model structure with available experimental distance restraints [42]. A total of 20 unambiguous distance restraints with Cα-Cα distance within 6 Å were collected from the results of 13C-13C proton-driven spin diffusion spectra. As observed from one of the Aβ fibril structures (PDB ID: 2BEG) [19], the Cα-Cα distance between two β-layers varies from ~8.0–12.0 Å; therefore, a cutoff of 12.0 was used to evaluate the topology of our fibril models. The average Cα-Cα distance of the inner core structures was calculated and shown in Fig. 8.</p><p>In all three models, there are at least 12 inter-layered residue distances that are less than 12 Å (Fig. 8), confirming that the proposed model structures are able to maintain the αS fibril fold. Due to the formation of a short β-strand in the region comprising residues 70–72, the distance between β3 and β4 layers is larger than 12 Å in M7a (Fig. 8B). In contrast, the corresponding distances are all less than 8 Å in M7b, and the distance between G67 of β3 and A76 of β4 is even less than 6 Å (Fig. 8C). Also note that the initial distance restraints used to construct these model are all well preserved (Fig. 1), implying that such organization of the five β-strands of the αS fibril could provide favorable inter-residual interactions and such interactions could be further optimized during the MD simulations. In addition, the single-layered αS fibril has been suggested to have a diameter of ~20–35 Å, and a height of ~40 Å (vertical distance between β1 and β5) in its gel-like state. We projected the mass density along x and y direction respectively, from which we could estimate the diameter and height of the αS fibril models (Fig. S3). Using a threshold of 0.05 Da/Å3 to exclude the disordered regions, the dimension of our αS fibril structures is ~45–50 Å, and the height is ~50 Å. Note that such a geometry of αS in aqueous solution is generally larger than the experimental ssNMR data due to the different hydration effect. However, the above dimensions are close or in agreement with other atomic force microscopy (AFM) experiments where samples were prepared in solution (pH 7.5, 310 K). For example, based on observation by AFM, Carter's group suggested a dimension of 32–44 Å for αS protofilament [43], and the Anderson group obtained a height of 45–60 Å for αS protofilament [44]. In addition, the Bennati group used electron paramagnetic resonance spectroscopy to measure the height of the αS fibril, and obtained a distance of 40–50 Å between the external of β-strands [45]. (Here both protofilament and fibril terms refer to the building block of fibrils).</p><!><p>It has been shown that the NAC promotes αS aggregation, and the hydrophobic motif of V71–V82 can form filaments alone [21]. Based on our fibril models, we found that the hydrophobic NAC composed of three β layers (β3, β4, and β5) can preserve about ~23%–25% of β-strand conformation in the relatively stable structures (M6, M7, and M8). The presence of charged repeat KTK(E/Q)GV, however, has distinct effects on the stability of the NAC (Fig. 4). In particular, the arrangement of charged residues in K58T59K60E61Q62V63 connecting β2 and β3 has the most significant effect on the stability of NAC.</p><p>We also found that the repeat K21T22K23Q24G25V26 in the remote N-terminal may stabilize the NAC domain near to the C-terminal. To test if the secondary and tertiary structures of the above models are also stable in the presence of an additional K21T22K23Q24G25V26 repeat, we extended the N-terminal of M7b to construct a fibril structure of αS(20–110) (M8, Fig. S4), and performed a 40-ns MD simulations under the same conditions. M8 displayed similar secondary structures as the fibril models of αS(30–110) (Fig. 7). However, the strand β5 comprises more ordered residues (residues 85–105) than M7b (residues 94–105) (Fig. 7), suggesting that the presence of the repeat K21T22K23Q24G25V26 increases the stability of the NAC domain (residues 85–95 specifically), although the percentage of β-strand in the NAC of αS(20–110) fibril is comparable to that of αS(30–110) (M6 and M7) (Fig. 4). The tertiary structure of αS(20–110) is also similar to that of αS(30-110) (Fig. 8D), implying that the additional 10 residues have no significant effect on the five-layered topology. These results confirm that the atomic structure models presented here capture the characteristic features of αS fibrils independent of the sequence length.</p><p>Shaykhalishahi et al. found that introduction of an intramolecular disulfide bond in the double cysteine mutant G41C/V48C will inhibit amyloid formation [46]. With the G41C/V48C disulfide bond, the Cα-Cα distance range of cysteine disulfide bonds would be around 5.6 Å. However, the average G41/V48 Cα-Cα distances from last 5-ns MD trajectories of the two stable fibril models are much longer, 12.8 Å and 15.9 Å for M7a and M7b, respectively. Clearly, the disulfide bond restriction would destroy the favorite turn conformation for the K43T44K45E46G47V48 repeat and thus inhibit αS amyloid formation.</p><!><p>The design of potent inhibitors of αS aggregation provides promising therapeutic strategies. Recent experiments showed that epigallocatechin-3-gallate (EGCG), an abundant catechin in green tea, as well as small β-hairpin peptides, could prevent the formation of αS amyloids in different ways [47–49]. To understand the molecular mechanisms of the interactions between aggregation inhibitors and αS, we used the most stable structure obtained in this study to dock with several αS aggregation inhibitors. First, the EGCG molecule was blindly docked onto the most stable structural model of αS fibril (M7a). Similarly, peptides [48] including WW2 (sequence: KKLTVWIP′GKWITVSA, P′=D-Pro), and cyclo-WW2 (sequence: cyclo- KKLTVWIP′GKWITVSIP′P, P′= D-Pro), were also examined. Gasteiger charges were calculated for all structures. The Lamarckian genetic algorithm was applied to search the docking conformations of EGCG. Flexible and rigid docking experiments were performed for EGCG and β-peptides, respectively (the αS fibril structure is always kept rigid). AutoDock4 and AutoDockTools 4 were used to carry out the docking studies [50]. The distribution of the 100 top ranked conformations of EGCG and β-peptides relative to αS fibrils is shown in Fig. 9. We found that EGCG and β-peptides bind to different sites of the αS fibril. EGCG prefers binding to both N-terminal and NAC domain. A weak binding site in the C-terminal was also observed (Fig. 9A and 9B). In contrast, the β-peptides (WW2 and cyclo-WW2) interact primarily with both NAC and the C-terminal of αS fibril, and the binding of WW2 to the NAC seems particularly favorable (Fig. 9C–9F). The distinct binding modes of the EGCG and β-peptides indicate that they disrupt the formation of the αS fibril through different mechanisms, consistent with recent experimental observations [48].</p><p>In a membrane environment, αS has α-helix-rich conformation [51]. There are energy barriers for conformational transitions from α-helix-rich to β-strand-rich αS conformations. To estimate the energy difference between the α-helix-rich and the β-strand-rich conformations, we simulated a helical dimer of αS(30–110), with each monomer adopting α-helix conformation (the conformation on membrane) [51]. After 20-ns MD simulations, the dimer is still rich in α-helix structures (Fig. S5). Based on the last 5-ns trajectory we calculated the GBMV energy and compared with the value of dimer conformations extracted from trajectories of M7. The obtained conformational energy of a dimer in α-helix conformation is ~−3429 kcal/mol, lower than the dimer in a β sandwich conformation (−3152 kcal/mol and −3136 kcal/mol for dimers extracted from M7a and M7b, respectively). The above result suggests that the formation of αS fibril rich in β-strand needs to overcome a large energy barrier, and inhibitors of EGCG and β-peptides may modulate such process and redirect the aggregation pathway of αS.</p><!><p>Because the aggregates of αS are closely associated with PD, the study of the structure-function relationship of αS needs reliable structural models. With sufficient experimental data available, we probed possible fibrillar structures using computational approaches. Although many studies suggest that different αS strains have different structures, aggregation propensity, and levels of toxicity, one of these αS fibrils has been well investigated and its building block unit can adopt a topology with five-layered β-strands. However, the details of how these β-strands organize need to be elucidated, especially how the four imperfect KTK(E/Q)GV repeats arrange in the fibril structures, the interaction with membrane and the αS aggregation mechanism. Adopting a hierarchical strategy, we systematically examined the role of the salt-bridge formed between residues Lys and Glu within each repeat on the stability of model structures of αS(30–110) fibrils. The coupling of these repeats with the NAC domain was specifically investigated. Two obtained models (M6 and M7) are able to preserve the topology of β-strands, with M7a being more energetically stable. One interesting result is that all charged residues in these repeats prefer to be exposed to the solvent and form salt-bridges. The arrangement of charged residues in the repeat K58T59K60E61Q62V63 between the β2 and β3 layers is particularly important for the stability of the NAC domain. Two trajectories of M7 (M7a and M7b) were generated with the same initial conformation but different velocity distributions. Structural characterization suggests that they display somewhat different secondary and tertiary structures. The distribution of β-strand is in better agreement with experimental observations in M7a whereas the distance restraints of adjacent β layers are more consistent with ssNMR data in M7b model. Energetically, the M7a structure was very close to M7b. The stability of M7b was further validated in the context of the model structure αS(20–110) fibril, indicating it also could be a building block for the full-length αS fibrils. In summary, the proposed atomic structures of the αS fibrils not only provide valuable insights into the topology of the αS fibrils, but also offer reliable starting structures for investigations of the assembly mechanism of αS in aqueous or in membrane environments, and of the effect of familial mutations on the structure and aggregation kinetics. Our structural model can also be by examining their molecular interactions useful for rational design of molecules targeting αS aggregates.</p><!><p>The five-layered model structures of αS(30–110) pentamer were constructed using Chimera software [52]. The five β-strands were assigned according to the NMR results reported by the Riek group [13]. The loop connecting β2 and β3 contains only two residues (K60 and E61) in the original NMR assignments, which makes the turn too narrow and residues in β2 and β3 may collide. We thus extend this loop involving residues 58–61. Similarly, we also extend the loop connecting β3 and β4 that consists of residues 68–73 in order to satisfy the tertiary contact between T64/N65 of β3 and A78 of β4. Other inter-layered distance restraints used to build the model are shown in Fig. 1. These restraints include residue pair Y39-A53, H50-N65, N65-A78, and T75-I88. Note that these restraints are unambiguously assigned between residues based on solid-state NMR (ssNMR) experiment [13]. Other inter-layer distance restraints are also available but not used to build the model structures and will be checked later. We found that not all the "unambiguous" restraints can be applicable to MD simulation simultaneously. For example, ssNMR indicates that Cα of Y39 is within 6 Å of Cα of A53, V55, and E57, suggesting that the inter-layer distance (β1 and β2) is within 6 Å, if we apply these three distance restraints together, β1 and β2 will distort significantly. On the other hand, the normal Cα-Cα distance between two β-layers varies from ~8.0–12.0 Å. Therefore, we applied one distance restraints between each two adjacent layers, and check the other restraints later. Four of the six imperfect KTK(E/Q)GV motifs are present in αS(30–110). Except the K32T33K34E35G36V37 motif, the remaining three repeats (K43T44K45E46G47V48, K58T59K60E61Q62V63, and K80T81V82E83G84A85) have residues in two loops that connect β2 and β3, and β4 and β5, respectively. We thus constructed various models hierarchically in order to investigate the effect of a salt-bridge formed by Glu and Lys on the stability of the αS fibril models (Table 1). The effect of charged residues on the stability of the β-strand has been well recognized [53, 54], thus worth examining in the context of αS fibrils.</p><p>Molecular dynamics (MD) simulations have become a powerful tool in drug design and discovery [55–57]. Here we apply this approach to refine our structural models of αS fibril, which could be further used in future studies such as virtual screening of potential drugs targeting PD. All MD simulations were performed in the NPT ensemble using NAMD2 (v2.10) program [58]. Each model structure was represented by CHARMM27 force filed with CMAP correction [59, 60], and solvated in a box filled with TIP3P water molecules [61]. The minimum distance between protein and the edge of water box is at least 15 Å. Each system was neutralized by adding a number of counter-ions. The Langevin piston method and Langevin dynamics were applied to control the pressure at 1 atm and the temperature at 310 K [62]. The Particle Mesh Ewald (PME) method was used to calculate the long-range electrostatics [63]. The hydrogen bonds were constrained by using the SHAKE algorithm [64]. The van der Waals interactions were calculated using a switching function with a twin cutoff of 10 Å and 12 Å. An integration time step of 2 fs was used and trajectory was save every 10 ps. MD simulations were conducted in three successive steps. Each system was first energy minimized for 10000 steps, with the intermolecular hydrogen bonds of the β-strand harmonically restrained with a force constant of 20 kcal/mol. Then the system was heated to 310 K gradually with the same restraints. In the equilibration stage, the force constant was decreased to 10 kcal/mol, 5 kcal/mol, and eliminated entirely. In each step, the system was equilibrated for 2 ns. The production simulation for each system was run for 20–40 ns. If the fibril structures (M1–M6) are not able to maintain the five-fold β-sheet layer, no additional simulations were performed. Additional simulation as performed to verify stable structure (M7a and M7b).</p><p>The relative stability of each fibril model was evaluated in terms of the conformational energy calculated by the generalized Born using the molecular volume (GBMV) method implemented in CHARMM program (c37b2) [59, 65]. The standard parameters in the GBMV II algorithm were used. A single-point energy calculation with infinite cutoffs was performed after the structure was minimized 200 steps using the steepest decedent method. For each system, the last 5-ns trajectory (500 frames) was used to obtain the average energy and other analyses.</p>

PubMed Author Manuscript

Chromophore maturation and fluorescence fluctuation spectroscopy of fluorescent proteins in a cell-free expression system

Cell-free synthesis, a method for the rapid expression of proteins, is increasingly used to study interactions of complex biological systems. GFP and its variants have become indispensable for fluorescence studies in live cells and are equally attractive as reporters for cell-free systems. This work investigates the use of fluorescence fluctuation spectroscopy (FFS) as a tool for quantitative analysis of protein interactions in cell-free expression systems. We also explore chromophore maturation of fluorescent proteins, which is of crucial importance for fluorescence studies. A droplet sample protocol was developed that ensured sufficient oxygenation for chromophore maturation and ease of manipulation for titration studies. The kinetics of chromophore maturation of EGFP, EYFP, and mCherry were analyzed as a function of temperature. A strong increase in the rate from room temperature to 37 \xc2\xb0C was observed. We further demonstrate that all EGFP proteins fully mature in the cell-free solution and that brightness is a robust parameter specifying stoichiometry. Finally, FFS is applied to study the stoichiometry of the nuclear transport factor 2 in a cell-free system over a broad concentration range. We conclude that combining cell-free expression and FFS provides a powerful technique for quick, quantitative study of chromophore maturation and protein-protein interaction.

chromophore_maturation_and_fluorescence_fluctuation_spectroscopy_of_fluorescent_proteins_in_a_cell-f

<!>Experimental setup<!>Cell-free sample preparation<!>Cell free<!>Solution<!>Brightness analysis<!>Kinetics of chromophore maturation<!>Cell-free synthesis of EGFP<!>Kinetics of chromophore maturation<!>Brightness and titration studies<!>Discussion

<p>Cell-free protein expression is a simple and flexible method for the rapid synthesis of folded proteins. These systems represent an attractive alternative for producing difficult-to-express proteins, such as membrane proteins or proteins that seriously interfere with cell physiology [1]. Recently, cell-free systems have been used to engineer complex biological processes [2]. The inherent open nature of the cell-free system allows direct access and control of the biological system, which cannot be attained by cell experiments. Thus, the cell-free system is emerging as a versatile platform for systems biology experiments [3]. In addition, cell-free expression has been used lately to study the maturation kinetics of fluorescent proteins [3,4].</p><p>Because knowledge of protein–protein interactions is of critical importance for developing a molecular description of biological processes, we investigate the potential of fluorescence fluctuation spectroscopy (FFS)1 to directly quantify these interactions in the cell-free system. FFS utilizes the intensity fluctuations of fluorescent molecules passing through a very small optical observation volume to determine concentration, diffusion coefficient, and brightness of the molecules [5]. Brightness reflects the average fluorescence signal of a single particle [6]. Suppose two monomeric proteins that are labeled with enhanced green fluorescent protein (EGFP) associate to form a dimer. Because the dimer carries two fluorescent labels, it will appear twice as bright as the monomeric protein. This simple relationship between brightness and stoichiometry has been successfully used to quantify homo- and hetero-protein complexes inside cells [7,8].</p><p>The principle of brightness analysis should be directly applicable to FFS measurements of cell-free systems. Just as in cells, expression of EGFP-tagged proteins provides a convenient method for introducing fluorescent markers. However, there are a number of potential challenges for FFS studies in cell-free systems. The fidelity of cell-free expression and of chromophore maturation needs to be checked, because the presence of nonfluorescent or incompletely expressed protein leads to biased interpretation of FFS experiments [9]. We develop a droplet-based assay for FFS experiments in a cell-free extract. The droplet format not only ensures a small sample volume, but also provides sufficient oxygenation of the sample for efficient chromophore maturation.</p><p>The widespread use of fluorescent proteins as markers in biological and medical research rests on the unique ability of the expressed protein to form a chromophore on its own. The formation of the chromophore is referred to as maturation and only requires molecular oxygen as an external reagent [10]. Although the kinetics of chromophore formation has been the subject of numerous studies, there is significant inconsistency in the reported maturation times of individual fluorescent proteins. Here we investigate the temperature dependence of the maturation process, which has received little attention so far. Specifically, we conduct FFS experiments in batch mode on EGFP, EYFP, and mCherry using an Escherichia coli-based cell-free system. We observe a pronounced temperature dependence of the maturation rates of the fluorescent proteins, model the kinetics by transition state theory, and compare our results with other maturation studies based on de novo protein synthesis. Because EGFP exhibits a very fast maturation at 37 °C, we consider EGFP to be a suitable reporter for many kinetic studies in the cell-free system.</p><p>We further use EGFP and a tandem-dimer EGFP (EGFP2) to check the fidelity of expression and maturation of the E. coli-based cell-free system and to demonstrate the feasibility of protein titration experiments with the droplet-based assay. Nuclear transport factor 2 (NTF2) is used as a model protein to explore brightness characterization of protein-protein interactions in the cell-free system. NTF2 plays a key role in nucleocytoplasmic transport and maintains the cellular Ran gradient which drives the transport process [11]. Depletion of NTF2 affects the Ran gradient and can lead to cell death [12]. The mechanism by which NTF2 performs its function is not well understood. Because there are conflicting reports regarding the oligomeric state of NTF2 [13–15], we examine the stoichiometry and concentration dependence of NTF2 in cell-free solution and compare the result with the literature.</p><!><p>A mode-locked Ti:sapphire laser (Tsunami, Spectra-Physics, Mountain View, CA) serves as a source for two-photon excitation through a modified Zeiss Axiovert 200 microscope (Thornwood, NY) as previously described [7]. FFS measurements are taken from 30 to 120 s and use two-photon excitation of the sample at 1000 nm. Excitation light is focused through a Zeiss 63 × C-Apochromat water immersion objective (N.A = 1.2) or a 63× Plan Apochromat oil immersion objective (N.A. = 1.4). An excitation power of 0.3 mW is used to avoid saturation and photobleaching effects. Dual-color experiments are recorded onto separate detectors using a 580 nm dichroic mirror (T580LXPR, Chroma, Rockingham, VT). Photon counts are detected with an avalanche photodiode (Perkin-Elmer, SPCM-AQ-14) and recorded by a data acquisition card (ISS, Champaign, IL), which stores the complete sequence of photon counts using sampling frequencies ranging from 50 to 200 kHz. The photon counts are analyzed with programs written in IDL 6.0 (Research Systems, Boulder, CO).</p><!><p>We use the S30 T7 High-Yield Expression System (Promega, Madison, WI) as our cell-free solution system. We mix DNA into 5- to 15-μL samples of cell-free solution according to the Promega protocol, and let the synthesis reaction run at room temperature for 2–4 h. Reactions were subsequently stopped by the addition of 0.1% RNase A (Sigma-Aldrich, St. Louis, MO) and then spun down at 18,000g for 20 min to remove any large particles. Samples were transferred into a ring of vacuum grease (Dow Corning, Midland, MI) to prevent spreading inside an eight-well coverglass chamber slide (Nagle Nunc International, Rochester, NY). The eight-well chamber is sealed with a rubber stopper or an additional coverslip and vacuum grease seal. Glass surfaces are treated with SigmaCote (Sigma–Aldrich) and measurements are taken 10 μm above the coverglass to avoid surface adhesion effects.</p><!><p>EGFP was amplified from the pEGFP-C1 plasmid (Clontech, Mountain View, CA) with a 5′ primer that encodes a BamHI restriction site and a 3′ primer that encodes an XhoI site. The result was cloned into the pRSET B vector (Invitrogen, Carlsbad, CA). We refer to this plasmid as pB-G. A tandem dimeric EGFP was constructed by cloning a second EGFP between the EcoRI and HindIII restriction sites of the pB-G plasmid. NTF2 was amplified from human NTF2 (GenBank Accession Number: BC002348) and cloned into the EcoRI/HindIII site of pB-G. A monomeric NTF2 was generated by mutating the methionine residue 118 to glutamate [15].</p><!><p>EGFP was expressed in E. coli and purified according to protocol using QiaExpress Ni-NTA Fast Start kit (Qiagen, Valencia, CA). Cell-free and aqueous dilution experiments were performed using HBS-EP buffer (Biacore, Uppsala, Sweden) or Dulbecco's PBS with calcium and magnesium (Biowhittaker, Walkersville, MD).</p><!><p>The time-dependent fluorescence intensity F(t) of the FFS measurement is used to determine its autocorrelation function,</p><p>where δF(t) = F(t) − 〈F〉 is the fluctuation in fluorescence intensity about the average. A fit of the autocorrelation function to a simple diffusion model determines the autocorrelation amplitude G(0), which is related to the average number of molecules NPSF present in the excitation volume,</p><p>with γ representing the shape factor of the FFS observation volume [16]. We convert NPSF into a concentration by c = NPSF/VPSF. The value of VPSF is determined from an FFS measurement of a dye sample with known concentration [7]. The brightness is calculated from the fluctuation amplitude and the average fluorescence intensity,</p><p>and reflects the average fluorescence intensity received per particle. The brightness of a monomeric fluorescent protein, such as λEGFP, serves as reference value and is determined by a calibration measurement of a sample containing only the fluorescent protein. Association of two EGFP-labeled proteins to form a dimer results in a doubling of the brightness, because the protein complex contains two EGFP proteins. This example illustrates that brightness encodes the stoichiometry of a protein complex. It is useful to define a normalized brightness b by dividing the brightness λ of a protein sample by the brightness of EGFP,</p><p>The normalized brightness b provides an experimental measure of the stoichiometry of the protein complex. For example, b = 2 indicates a dimeric protein.</p><!><p>Protein expression is stopped with 0.1% RNase and quickly transferred to the sample chamber. The observed increase in fluorescence intensity reflects chromophore maturation of previously synthesized proteins and is fit by</p><p>where F0 is the average fluorescence intensity at the time the reaction is stopped, ΔF is the change in average fluorescence intensity from the subsequently maturing chromophores, and τ is the characteristic maturation time. The maturation rate coefficient k is the inverse of the maturation time, k = (1/τ).</p><p>The temperature dependence of the maturation rate coefficient is modeled using transition state theory. The Eyring equation relates the reaction rate coefficient k to temperature T,</p><p>where ΔH‡ is enthalpy of activation, DS‡ is entropy of activation, kB is the Boltzmann constant, h is Planck's constant, and R is the gas constant. The reaction rate data are plotted as ln(k/T) vs 1/T and fit to a straight line. The slope and intercept of the fitted line determine the activation enthalpy and entropy. Sample temperatures are adjusted using an ASI 400 air stream incubator (Nevtek, Williamsville, VA) or a VWR Polyscience chiller recirculator (Niles, IL) connected to home- built coils around the microscope objective and sample holder. The sample temperatures reported in this work were recorded using a thermocouple placed in a droplet or water reservoir adjacent to the measured sample.</p><!><p>Because cell-free samples are typically 5–50 μL, the measurement protocols must be adapted for small volumes. Our first approach was to place the cell-free solution between two coverslips. On expressing EGFP we observed that the sample was only fluorescing around the edges of the sample, which is in contact with the surrounding air. Because oxygen is required for maturation of the chromophore, our result indicated that good contact of the sample with the surrounding air is crucial. To address this challenge we adopted a configuration where a drop of cell-free solution is placed on a chambered coverglass slide. Before placing the droplet, a small circle of vacuum grease is applied on the glass surface to prevent spreading of the solution across the slide. This protocol results in a well-formed droplet with a large surface area in contact with air for efficient uptake of oxygen. Expression of EGFP in the droplet solution resulted in the appearance of fluorescence throughout the sample as confirmed by an axial scan [17] through the sample with a long-working distance objective.</p><p>Because the measurement of cell-free expression reactions may take several hours, we evaluated the long-term stability of the droplet sample. A droplet sample containing an aqueous solution of the fluorescent dye Alexa 488 was measured as a control by FFS for several hours (data not shown). We observed a continuous increase in the fluorescence intensity. FFS analysis showed that the brightness of the dye remained constant, but the dye concentration increased. The increase in concentration is consistent with the loss of sample volume due to evaporation. To circumvent this problem we placed the droplet into a sealed chamber together with an additional water droplet (Fig. 2D). The water serves to establish vapor pressure equilibrium after the top opening has been sealed with a rubber stopper, thereby preventing the evaporation of the sample. The sealed chamber and water reservoir are prepared in advance and the cell-free reaction mixture is added later via syringe through the rubber stopper. We tested the sealed chamber using an aqueous droplet sample containing the dye Alexa 488, which was measured for more than 6 h. The fluorescence intensity remained constant over the whole measurement period, demonstrating that evaporation effects are negligible. Thus, all experiments described here have been performed using the sealed chamber configuration.</p><p>We study cell-free expression of EGFP using FFS to monitor the course of protein synthesis in real time. Initial experiments established that the cell-free extract contains aggregates that interfere with later FFS analysis of the data. Fortunately, such complexes are relatively easily removed through centrifugation. Centrifuging the sample at 18,000g for 5–10 min pellets the detritus, and the clean cell-free supernatant is easily removed. The centrifugation step can be conducted before or after the protein expression reaction. Immediately after adding the DNA for EGFP to the centrifuged reaction mixture, the sample is placed on the microscope. Sixty-second measurements are taken at regular intervals over the next 5 h (Fig. 1A). The intensity increases after an initial lag phase. The reaction eventually begins to run down as the synthesis process exhausts the available energy [18]. Brightness analysis of these data allow us to extract time courses for the average number of molecules in the observation volume and the brightness of EGFP (Fig. 1B and C). The number of molecules increases overtime demonstrating that proteins are being produced as expected. More importantly, the brightness (Fig. 1C) is constant throughout the reaction which shows that there are no transitional or alternate brightness states. Unfolded and unmatured EGFP proteins are nonfluorescent and so invisible to FFS measurements. Once the protein is completed, it "turns on" at the characteristic brightness of EGFP and remains stable and independent of protein concentration. This is further confirmed by comparing FFS measurements on cell-free expressed EGFP, recombinant EGFP purified from transformed E. coli, and EGFP in living mammalian cells (Table 1). Measured under the same excitation power, all three EGFP samples have the same brightness. This robustness of EGFP brightness makes it a useful and reliable tool for quantitatively examining protein interactions. Table 1 also displays the diffusion time or average residence time within the optical volume. This information is acquired through autocorrelation analysis of the data and reflects the differing solution environments of the three methods. Purified EGFP in aqueous buffer solution has the fastest diffusion time. Both EGFP in cells and in cell-free solution have a slower diffusion time. Fluorescent dye added both to the buffer solution and to basic cell-free solution shows a similar change in diffusion time (data not shown). This indicates that the difference in diffusion time reflects the changes in viscosity of the solutions.</p><!><p>We investigate the kinetics of chromophore maturation using cell-free expression in a free-standing droplet (∼10 μL). After spinning down the cell-free solution, DNA is added and left to synthesize for 10–30 min. The reaction is stopped with 0.1% RNase and quickly transferred to the sample chamber. By adding RNase to the droplet the synthesis reaction is rapidly stopped [3]. Any subsequent increases in the fluorescence intensity can be attributed to chromophore maturation of previously synthesized proteins. Since each chromophore's maturation is independent, fluorescence intensity increase can be described by a single-exponential model. Fig. 2A shows the EGFP maturation intensity as a function of time fit with Eq. (5). At room temperature (20 °C), the characteristic maturation time, τEGFP = 15 ± 3.5 min is established by repeating the experiment several times (n = 6). In addition to EGFP, we also investigated the maturation of EYFP (n = 3) and mCherry (n = 4). We found EYFP at 19 °C to have a significantly longer maturation time, τEYFP = 78 ± 12 min and mCherry at 20 °C to be rather slow at τmch = 155 ± 10 min (Fig. 2B and C).</p><p>Oxygen diffuses within ∼100 s from the edge to the center of a 10 μ.L droplet. As such, we anticipate that oxygen availability is not a limiting factor for maturation reactions with reaction times longer than 100 s. As a control experiment we measure the maturation of EGFP in a smaller droplet (1 μL). With a diffusion time to the center of only ∼20 s the rate to replenish oxygen is much faster than for the large droplet. The maturation time of EGFP measured in the 1 μL droplet at 19 °C (τEGFP =15 min) is identical to the time observed in the 10 μ.L droplet. This result demonstrates that oxygen availability is not a limiting factor for our assay.</p><p>Literature values for chromophore maturation times of EGFP cover a broad range of values and involve a variety of experimental approaches [19–23]. To further complicate matters individual studies have been performed at different temperatures. However, a quantitative study that evaluates the influence of temperature on maturation is not readily available. We decided to expand our study in order to address the temperature dependence of chromophore maturation. Measurements of the fluorescence intensity of the maturation process were taken at selected temperatures up to 37 °C and fit with Eq. (5) to recover the characteristic maturation time τ and rate coefficient k = 1/τ The results of the temperature study are plotted as ln(k/T) vs 1/T (Fig. 3). The Eyring plot shows that maturation rates increase with temperature. Maturation of EGFP at 37 °C proceeds sufficiently fast (Fig. 2D) that we approach the practical limit of the experimental setup. Thus, the temperature dependence of EGFP maturation was performed by cooling the solution below room temperature. The straight lines through the data points of Fig. 3 represent a fit of the rate coefficients to transition state theory. The slope and y-axis intercept determine the activation enthalpy ΔH‡ and activation entropy ΔS‡ of the maturation process of each protein. The activation enthalpies for EGFP, EYFP, and mCherry are 65 ± 1.6,46 ± 4.3, and 43 ± 1.7 kJ/mol, respectively. These values fall in the range expected for solution oxidation reactions. The differences in activation enthalpy between the proteins are relatively small, particularly between EYFP and mCherry. The differences in the activation entropies (−79 ± 5, −156 ± 15, −175 ± 6J/(mol K) for EGFP, EYFP, and mCherry) follow a similar pattern. Overall, the ratios of the maturation kinetics for different fluorescent proteins are approximately independent of temperature, because of the similarity of the activation enthalpies. Calculating from the fits, the characteristic maturation times at 20 °C are τEGFP = 15min, τEYFP = 82 min, τmCh = 157 min, and at 37 °C these values become τEGFP = 3 min, τEYFP = 27 min, τmCh = 57 min. Note that the fast maturation rate measured for EGFP at 37 °C (Fig. 2D) agrees with the prediction from transition state theory (upper leftmost cross in Fig. 3).</p><!><p>After a protein is expressed and matured, its properties can be studied by FFS. Here we focus on demonstrating the feasibility of brightness measurements as a function of protein concentration with the droplet setup. For the simple addition of a ligand, enzyme, or other chemical, the rubber stopper and syringe method is very effective. However, this approach is inconvenient for performing repeated dilutions of the droplet. Thus, instead of the rubber stopper a coverslip was placed as a lid on top of the chambered coverslide. A thin layer of vacuum grease applied to the contact area of the coverslip seals the sample chamber. A 10 μL droplet of purified EGFP is placed onto the coverslide. Dilutions are performed by unsealing the coverslip, adding additional buffer with a pipette, pipetting up and down, and finally removing an equivalent volume of the mixture and replacing the coverslip. At each dilution step an FFS measurement is taken to determine brightness and concentration (Fig. 4, inset). The brightness is concentration independent and reflects that EGFP is a monomeric protein. The same result is obtained for EGFP expressed in cell-free solution (Fig. 4). An important control for brightness experiments is the demonstration of brightness doubling with the tandem-dimer EGFP2, formed by encoding two EGFPs connected with a short linker sequence [7]. After expressing EGFP2 in cell-free solution, a droplet was placed on the coverslide followed by FFS measurements in between dilution steps. The brightness of EGFP2 is concentration independent and twice the brightness of monomeric EGFP, because of the presence of two EGFP molecules. These results demonstrate that the droplet sample provides a quick and reliable method for titration experiments. It further establishes that the quality and fidelity of the cell-free expression system are suitable for brightness studies.</p><p>With the cell-free system characterized and brightness analysis validated, we apply this method to a biological system. The protein NTF2 establishes the cellular Ran gradient, which is crucial for nucleocytoplasmic transport [24]. Ran exists in the form of RanGTP at high concentrations in the nucleus and at low concentrations in the cytoplasm, where RanGTP is converted to RanGDP. This gradient provides directionality for nucleocytoplasmic transport. The transport factor NTF2 maintains the gradient by returning RanGDP to the nucleus, where it is converted back into RanGTP. NTF2 has been identified as a dimer in previous studies [13,14] with one study reporting a monomer/dimer equilibrium with a dissociation constant of ∼1 μM [15]. A recent FFS study of NTF2 from our lab finds NTF2 to be a dimer in cells [17]. However, only cells with high protein concentration were used for the cell study. We now take advantage of the cell-free system to measure the oligomeric state of NTF2 over a wide concentration range using the same FFS approach as the cell study. We perform brightness titration studies on EGFP and EGFP2 expressed in cell-free solution (Fig. 5) as control experiments. Next, we repeat the dilution experiment on EGFP-NTF2 expressed with the cell-free system. The brightness of EGFP-NTF2 aligns with EGFP2 (Fig. 5), indicating that NTF2 is dimeric over the entire measured concentration range (∼30–1400 nM). We also performed a brightness titration of the mutant EGFP-NTF2.M118E expressed in cell-free solution, because it has been previously reported that the point mutation M118E prevents dimerization of NTF2 [15]. Our brightness data confirm that the point mutation abolishes dimerization and show that EGFP-NTF2.M118E is a monomer at all measured concentrations (Fig. 5). Thus, the brightness analysis of NTF2 in the cell-free system agrees with the earlier cell data [17] and indicates that NTF2 continues to be a dimer well below 1 μM.</p><!><p>There are several factors that are crucial for FFS measurements on cell-free systems. The presence of aggregates in cell-free expression systems is an obstacle to FFS measurements. Some of the aggregates are fluorescent, which leads to spikes in the fluorescence intensity and prevents meaningful brightness analysis. In the case of two-photon excitation we occasionally trapped large fluorescent aggregates in the laser beam. Furthermore, some of the nonfluorescent aggregates are large enough to exclude sufficient volume when passing through the observation volume, so that the fluorescence signal is reduced. While this effect is less noticeable in the raw data, because no spikes are generated, it biases brightness analysis. The presence of large particulates ((x02A7E) μm) was directly confirmed by viewing the sample solution through a 63× objective under bright-field conditions. Aggregates seem to be present in most cell-free systems. It is speculated that some of these aggregates build up because cell-free solutions do not have the active processes which clean up unused or discarded reaction material. The bulk of the aggregates can be removed by an additional centrifugation step after reaction solution has been reconstituted and/or after the reaction has finished. Without this step, it is difficult to perform FFS experiments in the cell-free environment.</p><p>A unique feature of fluorescent proteins is their ability to form a chromophore by posttranslational modifications. This maturation process only requires molecular oxygen as an external reagent. The amount of oxygen dissolved in air-saturated water under normal atmospheric conditions is ∼200 μM at room temperature, which exceeds the concentration of expressed fluorescent protein by more than two orders of magnitude. Despite this apparent excess of oxygen over protein, we observe that a cell-free expression solution sandwiched between two coverslips failed to develop fluorescence except at the edges of the solution that were in contact with air. This observation indicates that other processes exhaust the oxygen reservoir of the solution. In fact, it has been previously reported that oxygen is efficiently depleted within minutes in a cell-free reaction [25]. We choose a droplet as our sample geometry to ensure efficient oxygen transfer at the solution/air interface. The volume of droplets (typically 10–15 μL) permits oxygen to diffuse from the interface to the center of the droplet in ∼100 s. Oxygen diffuses from the periphery to the center of a 1-μL droplet 5 times faster than through a 10-μL droplet. Measurement of EGFP maturation under both conditions returns the same results and demonstrates that oxygen is not a limiting factor in our experiments. In fact, we succeeded in measuring maturation rates as fast as 3 min with the droplet setup (Fig. 2D). A hydrophobic ring prevents spreading of the sample droplet across the hydrophilic coverglass. This setup is important for the titration experiments, because by preserving the droplet shape it is straightforward to add, mix, and remove solution. Measurements on droplets can be performed for many hours (Figs. 1 and 2) without loss of sample. Evaporation is prevented by sealing the measurement chamber as described earlier.</p><p>The process of chromophore formation, particularly of GFP, has been carefully studied. The green chromophore pathway (GFP, EGFP, EYFP) has three primary stages: cyclization, oxidation, and dehydration. The red chromophore pathway (DsRed, mCherry) includes an additional oxidative step for the formation of an acylimine linkage in the polypeptide backbone. Studies suggest that the primary rate-limiting step is oxidation for the green pathway. While the steps for the red pathway are less distinctly time separated, a recent report states that the final dehydration is partially rate limiting [26–28]. We model EGFP and EYFP maturation with a single exponential rate process, consistent with prior studies. The maturation kinetics of mCherry is also well-modeled by an exponential process. However, the intensity traces of chromophore maturation are affected by experimental imperfections, such as drift in the axial focus and variations in the laser power. These imperfections, although minor, currently limit our ability to investigate whether any small deviation from the single-exponential rate model exists. Nevertheless, the analyses of the maturation kinetics and its temperature dependence demonstrate that a single-exponential kinetic model provides a satisfactory description of the data.</p><p>While a coherent mechanism of chromophore formation has emerged, a consistent picture of the overall kinetics is not yet available. The maturation kinetics of fluorescent proteins has been studied by various methods, such as (i) de novo protein expression, (ii) triggered protein folding of solubilized inclusion bodies, and (iii) reoxidation after chemical reduction of the chromophore. The results obtained by each method are inconsistent with each other. For example, the maturation time of EGFP is reported as ∼60 min (by folding of solubilized inclusion bodies), as ∼130 min (by reoxidation), and as ∼8 min (by de novo protein expression). Although it is not clear why the reported rates differ that widely, each method requires unique sample conditions, which may contribute to the diversity of reported values. It seems prudent to compare studies that utilize the same experimental method. Because our study is based on cell-free expression, we focus in this paper on de novo protein expression methods.</p><p>However, even studies based on the same method report different maturation rates for the same fluorescent protein. Thus, additional experimental factors influence the maturation process. Although it has been noted that the kinetics of maturation of GFP is different at 25 and 37 °C [29], the influence of temperature on chromophore maturation has not received further attention. We performed, to the best of our knowledge, the first quantitative characterization of maturation kinetics as a function of temperature. Published maturation studies have been conducted at specific temperatures, ranging from room temperature to 37 °C Because the rate of maturation changes ∼5-fold over this temperature range, one needs to account for this change when comparing maturation experiments.</p><p>Chromophore maturation by de novo protein expression is usually performed in one of two ways. Most studies employ anaerobic expression conditions followed by the addition of oxygenated solvent to complete the maturation process. Some studies utilize aerobic expression conditions and terminate protein expression while monitoring the subsequent increase in fluorescence due to maturing protein, as was done in this case. Adding RNaseA to the solution stops gene expression almost instantly and allows the determination of maturation times of less than 1 min as has been previously demonstrated [3].</p><p>A large number of experiments probing maturation kinetics exist in the literature. To ensure a meaningful comparison we only consider studies based on de novo protein expression of the proteins EGFP, EYFP, or mCherry. We make no distinction between de novo protein expression by bacterial, cell-free, or eukaryotic cell expression systems. For EYFP a maturation time τ of 56 min measured at 25 °C has been reported, while a second study determined a time constant of 23 min measured at 37 °C [4,30]. These published data are in excellent agreement with the values determined from Fig. 3 with τ = 58 min at 25 °C and τ = 27 min at 37 °C The maturation of mCherry has been measured independently by two groups at 37 °C, who reported values of 22 and 57 min [31,32]. Our maturation time for mCherry at 37 °C is τ = 57 min, which matches the value reported by Merzlyak et al. [32]. Finally, the maturation lifetime of EGFP by cell-free expression has been reported as τ = 14 min at 37 °C and as τ = 8 min at 29 °C [3,4], while our own data yield lifetimes of τ = 3 min at 37 °C and as τ = 7 min at 29 °C Thus, our result is in close agreement with the value reported by Shin and Noireaux [3]. Surprisingly, EGFP is the fastest maturing protein among the ones tested. This observation is in contrast to its reputation of being a relatively slow maturing protein as established by previous denaturation and refolding studies. Because EGFP exhibits very fast maturation kinetics with de novo protein synthesis, it is important to ensure that a sufficiently fast rate of solvent oxygenation is guaranteed. Otherwise, the measured kinetics could be slowed down by the limited availability of oxygen.</p><p>EGFP's rapid maturation makes it an excellent reporter for studies of protein reactions and interactions conducted in a cell-free expression system, ensuring that the delay between expression and onset of fluorescence is minimal. In addition to traditional fluorescence methods, such as fluorescence lifetime and polarization, it is feasible to conduct fluorescence correlation experiments with cell-free systems. Expanding the repertoire of techniques in cell-free systems to FFS measurements of brightness requires a few precautions. The quality of protein expression of the cell-free system must be fairly good. Misfolding of the fluorescent protein or failure to mature creates a nonfluorescent component that interferes with the interpretation of brightness experiments. Similarly, the presence of long-lived dark states would bias the interpretation of brightness studies. Observing the brightness of EGFP2 provides a sensitive test of the completeness of protein maturation. For example, if 30% of EGFP molecules did not mature or existed in a long-lived dark state, a subpopulation of EGFP2 would exist with only one of the two EGFPs fluorescing. This, in turn, lowers the brightness from 2 to 1.7. A detailed discussion of the influence of dark states and flickering of fluorophores on brightness experiments is available [9]. The brightness of EGFP and EGFP2 in mammalian cells is routinely measured as a control to establish that maturation and protein expression are not compromised. The brightness doubling observed with EGFP2 (Fig. 4) ensures that virtually all EGFP molecules of the cell-free system mature and acquire fluorescence [9]. The result further confirms the absence of long-lived dark states for the EGFP-fluorophore in two-photon excitation. In addition, our brightness experiments on NTF2-EGFP establish that protein interactions can be quantitatively assessed in a cell-free system over a wide concentration range.</p><p>In summary, cell-free expression offers a quick and convenient method for performing fluorescence studies of protein–protein interactions with no need for further purification of the product. The commercial E. coli-based cell-free system has proven to be suitable for brightness studies. Combining cell-free expression with the droplet setup ensures sufficient oxygenation of the sample and provides an excellent platform for chromophore maturation studies of fluorescent proteins as a function of temperature. Maturation kinetics exhibits a pronounced temperature dependence, which is conveniently characterized by determining the activation enthalpy and entropy. This information should prove useful for future comparison of maturation measurements and hopefully provides a first step toward establishing definitive rate coefficients for individual fluorescent proteins.</p>

PubMed Author Manuscript

TiO2 Simultaneous Enrichment, On-Line Deglycosylation, and Sequential Analysis of Glyco- and Phosphopeptides

Reversible protein glycosylation and phosphorylation tightly modulate important cellular processes and are closely involved in pathological processes in a crosstalk dependent manner. Because of their significance and low abundances of glyco- and phosphopeptides, several strategies have been developed to simultaneously enrich and co-elute glyco- and phosphopeptides. However, the co-existence of deglycosylated peptides and phosphopeptides aggravates the mass spectrometry analysis. Herein we developed a novel strategy to analyze glyco- and phosphopeptides based on simultaneous enrichment with TiO2, on-line deglycosylation and collection of deglycosylated peptides, and subsequent elution of phosphopeptides. To optimize on-line deglycosylation conditions, the solution pH, buffer types and concentrations, and deglycosylation time were investigated. The application of this novel strategy to 100 μg mouse brain resulted in 355 glycopeptides and 1,975 phosphopeptides, which were 2.5 and 1.4 folds of those enriched with the reported method. This study will expand the application of TiO2 and may shed light on simultaneously monitoring protein multiple post-translational modifications.

tio2_simultaneous_enrichment,_on-line_deglycosylation,_and_sequential_analysis_of_glyco-_and_phospho

Introduction<!>Reagents and Standards<!>Instruments<!>Protein Extraction<!>Tryptic Digestion of the Protein and Sample Desalination<!>Enrichment of Glyco- and Phosphopeptides from the Mixture of Fetuin and α-Casein Digest<!>On-Line Deglycosylation of the Glycopeptides<!>Effect of the Solution pH on the On-Line Deglycosylation<!>Effect of the NH4OAc Concentration on the On-Line Deglycosylation<!>Effect of the Deglycosylation Time on the On-Line Deglycosylation<!>Optimization of the Deglycosylated Peptides Elution Conditions<!>Application to the Enrichment of Mouse Brain Actual Sample<!>Mass Spectrometry Analysis and Data Processing<!>The Workflow of the Simultaneous Enrichment, On-Line Deglycosylation, and Sequential Elution Strategy to Analyze Glyco- and Phosphopeptides<!><!>Effect of Solution pH on On-Line Deglycosylation<!><!>Effect of the NH4OAc Concentration on the On-Line Deglycosylation<!><!>Effect of the Deglycosylation Time on the On-Line Deglycosylation<!>Optimization of the Elution Conditions for the Deglycosylated Peptides<!><!>Analysis of Glyco- and Phosphopeptides from the Mouse Brain<!><!>Analysis of Glyco- and Phosphopeptides from the Mouse Brain<!><!>Analysis of Glyco- and Phosphopeptides from the Mouse Brain<!>Conclusion<!>Data Availability Statement<!>Ethics Statement<!>Author Contributions<!>Funding<!>Conflict of Interest<!>Publisher’s Note<!>Supplementary Material<!>

<p>Protein glycosylation and phosphorylation are two of the most ubiquitous and important post-translational modifications (PTMs) and they play vital roles in regulating a variety of physiological and pathological processes. These two types of PTMs rarely work alone but interplay in a crosstalk dependent manner. Increasing lines of evidence indicate that the crosstalk between protein glycosylation and phosphorylation is involved in many important biological events (Hart et al., 2011) and their abnormalities are closely associated with many serious diseases (Liu et al., 2002; Takeda et al., 2015; Ma et al., 2017; Zhang et al., 2017). For example, reciprocal protein glycosylation and phosphorylation co-regulate nutrient sensing, neural development, and cell cycle (Hart et al., 2011); the hyperphosphorylation of tau protein is triggered by its abnormal N-linked glycosylation, which is key to Alzheimer's disease (Losev et al., 2021). Thus simultaneous monitoring these two PTMs and elucidating of their crosstalk in biological samples, especially for precious and trace of biological samples, have pathological and clinical significance.</p><p>In the past decade, PTM proteomics has developed rapidly, benefiting from advances of mass spectrometry (MS) technology and improvement of enrichment strategies. However, it remains challenging to simultaneously analyze glyco- and phosphopeptides, due to their low abundances and the high complexity of biological samples. To date, several materials have been developed for simultaneous enrichment of glyco- and phosphopeptides, including metal oxide affinity chromatography- (MOAC-) based materials (Xu et al., 2016; Xu et al., 2017; Sun et al., 2019), immobilized metal ion affinity chromatography- (IMAC-) based materials (Melo-Braga et al., 2014; Zou et al., 2017; Cho et al., 2019; Wang et al., 2019), and hydrogen bond-based polymer material (Lu et al., 2020). As the representative of MOAC-based materials, TiO2 is the most commonly used for its excellent robustness (Peng et al., 2017), reproducibility (Sun et al., 2019), and commercial availability. The affinity of TiO2 to glycopeptides is based on ligand-exchange and hydrophilic interactions between TiO2 and saccharides (Sheng et al., 2013) and binding of TiO2 toward phosphopeptides is based on Lewis acid-base interaction between TiO2 and phosphate groups (Yan and Deng 2019). In classical TiO2 simultaneous enrichment cases (Scheme 1A), the captured glyco- and phosphopeptides are co-eluted (Hu et al., 2018; Palmisano et al., 2012a) and undergo an enzymatic deglycosylation treatment for glycosylation sites identification (Deeb et al., 2014). However, the co-existence of deglycosylated peptides and phosphopeptides will increase the burden of further MS analysis. To reduce the complexity of samples, the two-dimensional enrichment is often employed to address this issue (Melo-Braga et al., 2015), but additional processing steps may lead to a low recovery of targets. Besides, the co-existent phosphopeptides can be hydrolyzed under alkaline deglycosylation conditions (Thompson et al., 2003), and the desalting procedure after deglycosylation will aggravate the loss of PTM-peptides.</p><p>Herein, we developed a novel strategy to simultaneously enrich and sequentially analyze glyco- and phosphopeptides, which consists of the simultaneous enrichment of glyco- and phosphopeptides with TiO2 and the on-line deglycosylation to obtain deglycosylated peptides and sequential elution of phosphopeptides. The on-line deglycosylation is key to the success of this strategy. Thus, some key factors of the on-line deglycosylation were investigated and optimized, such as solution pH, buffer concentrations, and deglycosylation time. This work will have a great potential in the simultaneous analysis of the protein glycosylation and other multiple PTMs.</p><!><p>HPLC-grade acetonitrile (ACN), urea, ammonium hydroxide (NH3·H2O), DL-dithiothreitol (DTT), iodoacetamide (IAA), ammonium formate (HCOONH4), ammonium acetate (NH4OAc), ammonium bicarbonate (NH4HCO3), formic acid (FA), acetic acid, glycolic acid, [Glu1]-Fibrinopeptide B human (GFB) (internal standard), bovine fetuin (standard glycoprotein), α-casein (standard phosphoprotein), and trypsin were purchased from Sigma-Aldrich (St Louis, MO, United States). Standard phosphopeptide (with sequence of HS*PIAPSSPSPK) was synthesized by Qiangyao Biotechnology Co., Ltd. (Shanghai, China). Trifluoroacetic acid (TFA), acetone, and ethyl alcohol were purchased from Shanghai Macklin Biochemical Co., Ltd. (Shanghai, China). PNGase F was purchased from New England Biolabs (Ipswich, MA, United States). Radioimmunoprecipitation (RIPA) lysis buffer and bicinchoninic acid (BCA) protein assay kit were purchased from Beyotime Biotechnology (Shanghai, China). GELoader was purchased from Eppendorf (Hamburg, Germany). TiO2 was purchased from GL Sciences (Tokyo, Japan). C18HC material was purchased from ACCHROM (Wenling, China). Mouse brains were provided by Dalian Medical University (Dalian, China). Pure water was purified with a Milli-Q system (Millipore, Milford, MA, United States).</p><!><p>The peptide samples and TiO2 were mixed in a thermomixer (Qianjun, Shanghai, China). TiO2 was separated from the mixture by centrifuge (Merck, Milford, MA, United States). A Labconco CentriVap system (Labconco, Kansas, MO, United States) was applied to dry samples in specific steps. Determination of protein concentrations was by a microplate reader (Thermo Scientific, San Jose, CA, United States). The qualitative analysis of the standard protein digests was conducted on a nano electrospray ionization quadrupole time-of-flight mass spectrometer (ESI-Q-TOF MS) (Waters, Manchester, United Kingdom). The qualitative analysis of the protein digests extracted from the mouse brains was performed using an Orbitrap Eclipse Tribrid mass spectrometer and a Dionex UltiMate 3000 rapid separation liquid chromatography (RSLC) system (Thermo Scientific, San Jose, CA, United States).</p><!><p>A mouse brain tissue was cleaned and cut into pieces. Then the tissue pieces were ground into white powder in a mortar with liquid nitrogen. The tissue powder was mixed with 2 ml ice-cold RIPA lysis buffer and transferred into a 5 ml centrifuge tube. The mixture was placed in an ultrasonic crushing machine on ice for 5 min. The sonication sequential mode was 1 s on and 3 s off, in addition to 30-minute cycles. After the lysis, the mixture was centrifuged at 13,000 g for 30 min at room temperature. The supernatant was collected and a precipitant was added. This mixture was deposited overnight at -20°C. After sedimentation, the sample was centrifuged at 13,000 g for 30 min and the supernatant was removed. The precipitation was washed with 3.6 ml of acetone, then 3.6 ml of anhydrous ethanol, and redissolved in 6 M urea. The concentration of the redissolved protein solution was determined by a bicinchoninic acid (BCA) method (Hussain et al., 2014). The animal experiments were authorized by the Experimental Animal Center of Dalian Medical University.</p><!><p>The above protein solution was diluted to 1 mg/ml with 6 M urea. 1 ml of protein solution was mixed with 50 μL of DTT (200 mM) and incubated at 56°C for 45 min. Then 200 μL of IAA (200 mM) was added and the mixture was placed in dark for 30 min. Then, 7.25 ml of 50 mM NH4HCO3 aqueous solution and 250 μg trypsin were added in the mixture and incubated at 37°C overnight. Finally, 5 μL FA was added to stop the digestion. Then, the sample was desalted with C18HC packed solid phase extraction microcolumns.</p><!><p>The enrichment was performed following a reported method (Palmisano et al., 2012b) with minor modification. The tryptic digests of fetuin (5 μg) and α-casein (5 μg) were mixed in 50 μL of 80% ACN/5% TFA, with 1 M glycolic acid (loading buffer). The mixture was added with 1 mg TiO2 material and incubated for 15 min. After removal of the supernatant by centrifugation, the TiO2 material was washed twice with 50 μL of loading buffer and centrifuged to remove the supernatants. The enriched peptides were used in further experiments.</p><!><p>The TiO2 materials attached with glyco- and phosphopeptides were mixed with 45 μL of 50 mM NH4OAc and 5 μL of PNGase F (2,500 U). The resulting solution was incubated for 16 h at 37°C. After centrifugation, the supernatant was collected and desalted for the MS analysis.</p><!><p>The TiO2 materials attached with glyco- and phosphopeptides were separately suspended in four solutions with different pH values: 50 mM HCOONH4 (pH 3.0), 50 mM NH4OAc (pH 6.9), 50 mM NH4HCO3 (pH 8.3), and 0.1% NH3·H2O (v/v, pH 11.5). For each solution, after incubation for 3 h at 37°C, the supernatant was collected by centrifuge and desalted for the MS analysis.</p><!><p>The TiO2 materials attached with glyco- and phosphopeptides were, respectively, suspended in 45 μL of NH4OAc solutions at different concentrations (5, 10, 20, 25, and 50 mM). For each solution, 5 μL of PNGase F (2,500 U) was added and it was incubated at 37°C overnight. After that, the supernatants were removed by centrifugation. The deglycosylated peptides were eluted with 80 μL of 40% ACN/5% TFA. Subsequently, the phosphopeptides were eluted with 80 μL of 5% (v/v) NH3·H2O. The elution fractions were vacuum-dried and desalted in 19 μL of 50% ACN/0.1% FA. Before the MS analysis, 1 μL internal standard (1 pmol GFB) was added in each sample.</p><!><p>6 mg TiO2 material, which was attached with glyco- and phosphopeptides, was mixed with 174 μL of 50 mM NH4OAc and 5 μL of PNGase F (2,500 U). The mixture was incubated at 37°C and 30 μL of suspended sample was taken after 6, 9, 12, 24, and 48 h incubation, separately. TiO2 was isolated by centrifugation, and the deglycosylated peptides and bound phosphopeptides on TiO2 were sequentially eluted and treated as described above.</p><!><p>Three experiments were performed to investigate the elution effectivities of different eluents. The experiments had the same simultaneous enrichment and on-line deglycosylation processes described above but varied in the later procedures. After on-line deglycosylation, for the first experiment, the 1 mg TiO2 material was successively washed with 80 μL of 5 mM NH4HCO3, 80 μL of 10 mM NH4HCO3, 80 μL of 20 mM NH4HCO3, and 80 μL of 5% (v/v) NH3·H2O. For the second experiment, the 1 mg TiO2 material was washed thrice with 20 μL of 50% ACN/1% FA. For the third experiment, the 1 mg TiO2 material was successively washed with 120 μL of 40% ACN/5% TFA, 120 μL of 5% TFA, and 80 μL of 5% (v/v) NH3·H2O. The supernatants of each wash were desalted for the MS analysis.</p><!><p>Simultaneous Enrichment: 100 μg mouse brain lysate was digested by trypsin, desalted, dried, and dissolved in 50 μL of 80% ACN/5% TFA (1 M glycolic acid). The sample was mixed with 2 mg TiO2 material and incubated for 30 min. After centrifugation, the precipitate was washed twice with 50 μL of loading buffer. The supernatant was removed by centrifugation.</p><p>On-Line Deglycosylation: After the simultaneous enrichment, the TiO2 material was mixed with 49 μL of 50 mM NH4OAc and 1 μL of PNGase F (500 U). The mixture was incubated at 37°C for 12 h. The supernatant was removed by centrifugation.</p><p>Sequential Elution: The deglycosylated peptides and the bound phosphopeptides on TiO2 were sequentially eluted with 80 μL of 40% ACN/5% TFA and 80 μL of 5% (v/v) NH3·H2O. The eluates were desalted and dried for the MS analysis.</p><p>The control experiment was performed on another 100 μg mouse brain lysate according to literature (Palmisano et al., 2012b).</p><!><p>The Dionex UltiMate 3000 RSLC system for chromatographic separation included a C18 trap column (75 μm × 20 mm, 3 μm) and a C18 analytical column (75 μm × 50 mm, 2 μm). The injected volume was 9 μL at a flow rate of 300 nL/min. The mobile phase was as follows: phase A was 0.1% FA and phase B was 80% ACN/0.1% FA. The gradient elution was as follows: 1–4% B, 4 min; 4%–8% B, 2 min; 8%–32% B, 104 min; and 32–90% B, 7 min.</p><p>The Orbitrap Eclipse Tribrid mass spectrometer was set as follows. For MS1, the spray voltage was 2.1 kV and the capillary temperature of the ion transport was 320°C. The first-stage full scanning range of mass spectrometry was m/z 300–1,500 with a resolution of 120,000. The RF Lens was set at 40%, the AGQ target was set at 300%, and the maximum injection time (MaxIT) was set to 50 ms. For MS2, the resolution was set at 30,000, the AGQ target was set at 100%, the MaxIT was set to 80 ms, the dynamic exclusion was set to 45 s, the isolation window was set to 1.6 Da, the collision energy was set at 30% HCD, and the fixed first mass was fixed to m/z 110. The data-dependent MS/MS was top speed mode with a cycle time of 2 s. The number of microscans to be performed was set at 1 scan s−1.</p><p>All the MS raw data were processed by Proteome Discoverer and searched with SEQUEST against the mouse proteins in the UniProt database. The trypsin cleavage with a maximum of two leakage sites was allowed. Carbamidomethyl (C) was set as a fixed modification, and oxidation on methionine (M), acetylation of protein N terminus, phospho-modification (STY), and deamination (N) were set as the variable modifications. The false discovery rate (FDR) was set at 1%. The other conditions were set by default.</p><!><p>In this work, we developed a new strategy to simultaneously enrich and sequentially separate glyco- and phosphopeptides with high efficiency and recovery. Firstly, TiO2 is used to simultaneously enrich glyco- and phosphopeptides from a complex sample, which are bound to TiO2 with their glycan chains or phosphate groups, respectively (Scheme 1B). After the enrichment, unmodified peptides are removed, and glyco- and phosphopeptides are attached on TiO2. Secondly, the on-line deglycosylation using PNGase F is carried out to remove the glycans from the glycopeptides and produce deglycosylated peptides. Thus, the deglycosylated peptides are released from TiO2 and collected. Thirdly, the attached phosphopeptides on TiO2 are eluted. Finally, the deglycosylated peptides and the phosphopeptides are characterized with MS, respectively.</p><!><p>Workflows for capturing and treatment of glyco- and phosphopeptides prior to the mass spectrometry analysis. (A) Workflow of one of the reported methods, which consists of the simultaneous enrichment with TiO2, the co-elution, and the deglycosylation with PNGase F (hereinafter abbreviated to "reported method"). (B) Workflow of our strategy, which consists of the simultaneous enrichment with TiO2, the on-line deglycosylation with PNGase F, and the sequential elution.</p><!><p>The solution pH was an important parameter for the on-line deglycosylation. An ideal solution pH should not only work out for the deglycosylation but also have no impact on the phosphopeptide retention on TiO2. Here we investigated the effect of the solution pH on the glyco- and phosphopeptides retention on TiO2 (Figure 1). After the enrichment, the TiO2 materials bound with PTM-peptides were separately resuspended in four solutions: 50 mM HCOONH4 (pH 4.6), 50 mM NH4OAc (pH 6.9), 50 mM NH4HCO3 (pH 8.3), and 0.1% NH3·H2O (v/v, pH 11.5). After incubation for 3 h and centrifugation, the supernatants were collected, desalted, and analyzed with MS. As shown in Figures 1A,B, neither glycopeptides nor phosphopeptides can be detected in the mass spectra of the HCOONH4 or NH4OAc treated samples. On the contrary, both glycopeptides (marked with red stars) and phosphopeptides (marked with blue circles) are clearly observed in the mass spectra of the NH4HCO3 and NH3·H2O treated samples (Figures 1C,D). These results implied that HCOONH4 and NH4OAc met the basic requirement for the on-line deglycosylation. Considering that the pH value of the NH4OAc solution (pH 6.9) was much closer to the reaction condition pH (pH 7.5) of the PNGase F deglycosylation, this solution pH was chosen in later experiments.</p><!><p>Optimization of the online deglycosylation condition in different solution pH. (A) 50 mM HCOONH4 (pH 4.6), (B) 50 mM NH4OAc (pH 6.9), (C) 50 mM NH4HCO3 (pH 8.3), and (D) 0.1% (v/v) NH3·H2O (pH 11.5). The glycopeptides and phosphopeptides were marked with red stars and blue circles, respectively. The signals at m/z ranging from 1,265 to 1,615 in the mass spectra were amplified tenfold (×10), where "×10" represents the magnification times.</p><!><p>The buffer concentrations have an impact on the deglycosylation efficiency by influencing the proton exchange during the enzyme catalysis (Cheison et al., 2011; Liu et al., 2009). Thus, the effect of the NH4OAc concentration on the efficiency of the on-line deglycosylation was further investigated. We measured the relative abundances of the deglycosylated peptides and phosphopeptide in the mass spectra of the elutes after the on-line deglycosylation with NH4OAc in five different concentrations (5, 10, 20, 25, and 50 mM). The deglycosylated peptides at m/z 1,225.1926 (3+) and m/z 871.3765 (2+) and the phosphopeptide at m/z 976.4594 (2+) were chosen as targets of interest. The relative abundances of the target peptides were quantified with GFB as an internal standard. As shown in Figure 2A, the relative abundances of the deglycosylated peptide and the phosphopeptide gradually increased with the NH4OAc concentration increasing and reached the maximum in 50 mM NH4OAc during the investigated concentration range. For the deglycosylated peptide, this phenomenon might be attributed to the improved efficiency of proton exchange between PNGase F and the glycopeptides with increased concentrations of NH4OAc, which is consistent with the reported result (Cheison et al., 2011). On the other hand, the enhanced relative abundance of the phosphopeptide might result from the reduced degree of its hydrolysis with the increased NH4OAc concentration. In order to further correlate the relationship between the hydrolysis degree of the phosphopeptide and the NH4OAc concentration, the relative abundances of the standard phosphopeptide in different concentrations of the NH4OAc solution were tested (Figure 2C). When the NH4OAc concentration was low, the relative abundance of the standard phosphopeptide was low, which was ascribed to the high hydrolysis degree of the standard phosphopeptide. As the NH4OAc concentrations increased, the relative abundance of the standard phosphopeptide gradually increased, which was attributed to the decreased hydrolysis degrees of the standard phosphopeptide. It seemed that higher NH4OAc concentrations were favorable for inhibiting the phosphopeptide hydrolysis. These results were in good agreement with that of Figure 2A. Taken together, 50 mM NH4OAc was chosen for further on-line deglycosylation.</p><!><p>Optimization of the on-line deglycosylation conditions in different NH4OAc concentrations (A, C) and deglycosylation times (B, D). The relative abundances of the target deglycosylated peptide and the phosphopeptide obtained after the online deglycosylation (A) in different concentrations of NH4OAc or (B) for different deglycosylation times. The relative abundances of the standard phosphopeptide (C) in different concentrations of NH4OAc or (D) in 50 mM NH4OAc for different times. DGP, deglycosylated glycopeptide; PP, phosphopeptide.</p><!><p>Compared with the glycopeptides, the phosphopeptides are more susceptible to external influence and are unstable (Hu et al., 2020). During the deglycosylation process, the phosphopeptides hydrolyze as time goes on, and, therefore, the deglycosylation time was another important factor for the on-line deglycosylation. In order to optimize the deglycosylation time, we measured the relative abundances of the target deglycosylated peptide and the phosphopeptide after the on-line deglycosylation with different times. As shown in Figure 2B, the relative abundance of the deglycosylated peptide at m/z 1,225.1926 (3+) increased over time until 12 h was reached, in sharp contrast to the relative abundance of the phosphopeptide which gradually decreased. The latter might result from the hydrolysis of the phosphopeptide (Figure 2D). Considering both the phosphopeptide hydrolysis degree and the glycopeptide deglycosylation efficiency, 12 h was chosen as further on-line deglycosylation time.</p><!><p>After the on-line deglycosylation, the released deglycosylated peptides were re-adsorbed on TiO2, which is line with the previous studies that non-modified peptides tend to be nonspecifically adsorbed on TiO2 due to Lewis acid-base interaction between the carboxyl groups on the peptide chains and TiO2 and the hydrophobic interaction between the hydrophobic peptide chains and TiO2 (Palmisano et al., 2012b). To efficiently elute and collect the absorbed deglycosylated peptides but not the phosphopeptides, the elution efficiencies of three types of eluents were evaluated. The three types of eluents were NH4HCO3 solutions with different concentrations (5, 10, and 20 mM), 50% ACN/1% FA, and 40% ACN/5% TFA. As shown in Supplementary Figure 1, 5 and 10 mM NH4HCO3 could not elute any PTM-peptides, while 20 mM NH4HCO3 could co-elute the deglycosylated peptides and the phosphopeptides. Therefore, NH4HCO3 was not suitable for the deglycosylated peptides elution. As to the acidic conditions, after the elution with 50% ACN/1% FA, the deglycosylated peptides were rarely observed (Supplementary Figure 2), while all the targeted deglycosylated peptides but not the phosphopeptides could be detected in the eluate of 40% ACN/5% TFA (Figure 3B). Afterward, the bound phosphopeptides on TiO2 were found from the eluate of 5% (v/v) NH3·H2O (Figure 3C), which was consistent with the phosphopeptides obtained by using the reported simultaneous enrichment, the co-elution, and the deglycosylation method (Figure 3A). Thus, 40% ACN/5% TFA was used as the eluent of the deglycosylated peptides.</p><!><p>Comparison between the reported method (A) and our strategy in the separation efficiency of the deglycosylated peptides (B) and the phosphopeptides (C). (A) The mass spectra of the mixture of the deglycosylated peptides and the phosphopeptides obtained by the reported method. The mass spectra of (B) the deglycosylated peptide fraction and (C) the phosphopeptide fraction obtained by our strategy. The deglycosylated peptides and the phosphopeptides are marked with green stars and blue circles, respectively.</p><!><p>To examine the effectiveness of our strategy, we applied it to analyze glyco- and phosphopeptides from a 100 μg mouse brain. Meanwhile, the reported method (Palmisano et al., 2012b) was carried out for comparison. By using our strategy, a total of 329 glycosylation sites (Supplementary Table 1) from 355 glycopeptides (Supplementary Table 2) and 1,496 phosphorylation sites (Supplementary Table 3) from 1975 phosphopeptides (Supplementary Table 4) were identified from three technical replicates. In sharp contrast to these, only 126 glycosylation sites (Supplementary Table 5) from 141 glycopeptides (Supplementary Table 6) and 1,423 phosphorylation sites (Supplementary Table 7) from 1,408 phosphopeptides (Supplementary Table 8) were identified from the identical sample using the reported method (Figures 4A–D). The numbers of the identified glyco- and phosphopeptides with our strategy were 2.5 and 1.4 folds of those with the reported method, respectively. Moreover, our strategy demonstrated high reproducibility with 73.5 and 87.9% common glyco- and phosphopeptides among three technical replicates (Figures 4E,F), respectively. The reproducibility between two technical replicates was much higher, 77.0 ± 2.2% and 91.8 ± 0.3% for glyco- and phosphopeptides, respectively. Besides the high reproducibility, the numbers of the phosphopeptides in the deglycosylated glycopeptide fraction and the phosphopeptide fraction were 67 and 2,298, respectively, suggesting a low degree of overlap between the deglycosylated glycopeptide and the phosphopeptide fraction in our strategy.</p><!><p>Comparison of the performance for identifying glyco- and phosphopeptides from a mouse brain. Venn diagram analysis of the number of (A) glycopeptides, (B) glycosylation sites, (C) phosphopeptides, and (D) phosphorylation sites identified with the reported method and our strategy. Reproducibility for (E) glyco- and (F) phosphopeptides among three technical replicates of our strategy.</p><!><p>The property differences of the identified PTM-peptides between the reported method and our strategy were further investigated. The molecular weight (Mw) distribution of the identified PTM-peptides and the percentages of the peptides with single-PTM and multiple PTM sites were compared (Figure 5).</p><!><p>The Mw distribution of identified glyco- (A) and phosphopeptides (B) by our strategy and the reported method. The percentages of the identified PTM-peptides with single glycosylation/phosphorylation site and multiple glycosylation/phosphorylation sites by our strategy (C, E) and the reported method (D, F). 1–3 GP: glycopeptide with 1–3 glycosylation sites; 1–3 PP: phosphopeptide with 1–3 phosphorylation sites.</p><!><p>As shown in Figure 5A, the Mw distribution pattern of the identified glycopeptides is consistent between the reported method and our strategy in a higher Mw range (1,500–5,000 Da) is consistent between the reported method and our strategy. However, in the lower Mw range of 500–1,000 Da and 1,000–1,500 Da, the number of the identified glycopeptides with our strategy accounts for 5.1 and 24.5% of the total ones, respectively, in sharp contrast to that of 0 and 7.1% with the reported method. These results indicate that our strategy has advantages in the enrichment and identification of low Mw glycopeptides. This result might be ascribed to the fact that our strategy omits the desalting procedure after the deglycosylation and retains well the low Mw glycopeptides.</p><p>As to the percentages of the peptides with single-PTM and multiple PTM sites, the number of identified glycopeptides with two glycosylation sites with our strategy accounts for 24% of the total glycopeptides, compared with only 14% with the reported method (Figures 5C,D). Similarly, the number of identified phosphopeptides with two phosphorylation sites with our strategy accounts for 10% of the total phosphopeptides, in sharp contrast to that of 3% with the reported method (Figures 5E,F). These results revealed the superiority of our strategy in the identification of the peptides with multiple glycosylation/phosphorylation sites. It is possibly because our strategy reduced the complexity of the samples and increased the number of the identified low-abundance peptides with multiple glycosylation/phosphorylation sites.</p><p>The above results indicated that our strategy not only realized the sequential elution of glyco- and phosphopeptides but also significantly increased the numbers of identified glyco- and phosphopeptides. Our study provided an effective means for the simultaneous characterization of the protein glycosylation and phosphorylation.</p><!><p>In this work, we developed a strategy for the analysis of glyco- and phosphopeptides based on the simultaneous enrichment with TiO2, the on-line deglycosylation, and the sequential elution. The application of this strategy to the mouse brain tissue achieved a higher number of targeted peptides compared with the reported method. Our strategy shows some advantages in the simultaneous analysis of glyco- and phosphopeptides: 1. The on-line deglycosylation and the sequential elution can separate the deglycosylated peptides and the phosphopeptides into two different fractions, which can reduce the complexity of the samples and improve the coverage of the identified PTM-peptides under the data dependent acquisition (DDA) mode. 2. The reduction of the sample complexity can reduce the ion suppression and increase the number of low-abundance glyco- and phosphopeptides with multiple glycosylation and phosphorylation sites. 3. The elimination of the desalting procedure after the deglycosylation can reduce the loss of low Mw glycopeptides. 4. A neutral online deglycosylation condition can effectively inhibit the hydrolysis of the phosphopeptides and increase the number of identified phosphopeptides.</p><p>To sum up, this work will provide a new idea to expand the applications of TiO2 and tackle the problems in the simultaneous analysis of the protein glycosylation and other multiple PTMs.</p><!><p>The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/Supplementary Material.</p><!><p>The animal study was reviewed and approved by the Biological Research Ethics Committee of Dalian Medical University. A written informed consent was obtained from the owners for the participation of their animals in this study.</p><!><p>CC and XZ carried out the experiments and wrote the manuscript. XD participated in the optimization of the enrichment methods. HZ contributed to the manuscript revision. XLL and XML contributed to the study design and the manuscript revision.</p><!><p>This work was supported by the National Natural Science Foundation of China (Nos. 21934005, 21775148, and 21804130) and DICP Innovation Funding (DICP-1202030).</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><!><p>The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fchem.2021.703176/full#supplementary-material</p><!><p>Click here for additional data file.</p><p>Click here for additional data file.</p><p>Click here for additional data file.</p><p>Click here for additional data file.</p><p>Click here for additional data file.</p><p>Click here for additional data file.</p><p>Click here for additional data file.</p><p>Click here for additional data file.</p><p>Click here for additional data file.</p>

PubMed Open Access

Small Molecule Inhibitors of the c-Fes Protein-tyrosine Kinase

SUMMARY The c-Fes protein-tyrosine kinase modulates cellular signaling pathways governing differentiation, the innate immune response, and vasculogenesis. Here we report the identification of Type I and II kinase inhibitors with potent activity against c-Fes both in vitro and in cell-based assays. One of the most potent inhibitors is the previously described anaplastic lymphoma kinase inhibitor, TAE684. The crystal structure of TAE684 in complex with the c-Fes SH2-kinase domain showed excellent shape complementarity with the ATP-binding pocket and a key role for the gatekeeper methionine in the inhibitory mechanism. TAE684 and two pyrazolopyrimidines with nanomolar potency against c-Fes in vitro were used to establish a novel role for this kinase in osteoclastogenesis, illustrating the value of these inhibitors as tool compounds to probe the diverse biological functions associated with this unique kinase.

small_molecule_inhibitors_of_the_c-fes_protein-tyrosine_kinase

INTRODUCTION<!>Identification of c-Fes inhibitors by FRET-based chemical library screening<!>Kinome-wide selectivity of lead compounds<!>Three-dimensional structure of the c-Fes SH2-KD region in complex with TAE684<!>Analysis of inhibitors in a cell-based assay for c-Fes autophosphorylation and microtubule localization<!>Inhibition of tubulin phosphorylation by c-Fes in vitro<!>c-Fes inhibitors selectively inhibit rodent fibroblasts transformed by active c-Fes mutants but not by the Src-family kinase, Hck<!>c-Fes Inhibitors Reveal a New Role for c-Fes Kinase Activity in Differentiation of Macrophages to Osteoclasts<!>SIGNIFICANCE<!>Chemical Library Screen<!>Tubulin Phosphorylation Assay<!>Cell-based Assay for Microtubule Association<!>Rat-2 Fibroblast Transformation Assay<!>Isolation and culture of bone marrow monocytes/macrophages (BMM)<!>Osteoclast differentiation assay<!>Chemical Synthesis<!>X-ray Crystallography

<p>The c-fes/fps proto-oncogene encodes a 93 kDa protein-tyrosine kinase (c-Fes), and together with the homologous kinase Fer, defines a structurally unique kinase family [reviewed in (Greer et al., 2011; Hellwig and Smithgall, 2011)]. Sequences of c-fes and fps were first isolated as part of oncogenic Gag-Fes/Fps chimeras found in several avian and feline retroviruses (Snyder and Theilen, 1969; Wang et al., 1981), leading to subsequent identification of the corresponding mammalian and avian cellular proto-oncogenes (Huang et al., 1985; Roebroek et al., 1985). Human c-fes, which maps to chromosome 15, is expressed in embryonic tissues derived from all three germ layers (Carè et al., 1994). In adults, c-Fes is present in a variety of cell lineages, including myeloid hematopoietic, vascular endothelial, neuronal and epithelial cells (Carè et al., 1994; Delfino et al., 2006; Haigh et al., 1996).</p><p>The structural organization of c-Fes is distinct from other nonreceptor tyrosine kinases such as c-Src and c-Abl (Greer et al., 2011; Hellwig and Smithgall, 2011). The unique N-terminal region features a Fes/CIP4 homology (FCH) domain, followed by two-coiled coil motifs, a central Src-homology 2 (SH2) domain and a C-terminal kinase domain. The FCH region and first coiled-coil motif comprise an F-BAR homology domain. Other F-BAR domain proteins have been implicated in the regulation of plasma membrane curvature through phosphoinositide binding and induction of membrane tubulation (Itoh and De Camilli, 2006). A recent study demonstrated the ability of the c-Fes F-BAR domain to bind phospholipids and induce membrane tubulation in vitro, suggesting that phosphoinositides may recruit c-Fes to cellular membranes and contribute to its activation by FcεRI/Lyn complexes in mast cells (McPherson et al., 2009).</p><p>c-Fes biological activity is tightly regulated in cells, with the kinase domain adopting a catalytically repressed state (Greer et al., 1988). Downregulation of kinase activity is maintained despite the absence of negative regulatory structural elements, such as an SH3 domain or a negative regulatory tail found in the Src-family of nonreceptor tyrosine kinases (Engen et al., 2008). Cellular c-Fes kinase activity is stimulated by the experimental addition of the amino-terminal myristylation signal from c-Src (Greer et al., 1994), replacement of residues of the c-Fes kinase domain with homologous v-Fps sequences (Kim and Feldman, 2002), introduction of a point mutation predicted to disrupt the structure of the first N-terminal coiled-coil domain (Cheng et al., 2001; Shaffer et al., 2009) or substitution of the SH2 domain with that of v-Src (Rogers et al., 2000). Interestingly, insertional mutagenesis of the v-Fps SH2 domain reduced kinase and transforming functions, providing the first evidence that the SH2 domain is a positive regulator of kinase activity (Sadowski et al., 1986). Subsequent studies indicated that the same is true for c-Fes (Hjermstad et al., 1993).</p><p>A recent crystal structure of a truncated form of c-Fes, consisting of the SH2 and kinase domains, revealed the molecular mechanisms behind the positive impact of the SH2 domain on kinase activity (Filippakopoulos et al., 2008). Packing and electrostatic interactions between the SH2 and the kinase domain stabilize an active conformation of the critical αC helix found in the kinase domain N-lobe. Crystallization with a synthetic substrate peptide established that substrate binding to the phosphotyrosine site of the SH2 domain stabilizes an ordered SH2 conformation and primes the kinase for catalysis through proper orientation of the αC helix. These structures suggested a model of coordinated c-Fes activation in which substrate binding to SH2 and subsequent autophosphorylation of the activation loop on Y713 stabilize a catalytically competent kinase domain conformation.</p><p>Several lines of evidence suggest a possible role for c-Fes in oncogenesis. Kinase-active mutants of c-Fes drive focus-forming activity and soft-agar colony formation in rodent fibroblast transformation assays (Cheng et al., 2001; Li and Smithgall, 1998). More recently, c-Fes was identified as a phosphorylation target of the constitutively active D816V mutant of c-Kit, a mutation commonly found in human malignancies (Voisset et al., 2007). siRNA targeting of endogenous c-Fes in TF-1 cells expressing c-Kit D816V significantly reduced proliferation and phosphorylation of STATs and p70 S6 kinase. Active c-Fes has been observed in acute myeloid leukemia (AML), and reduction of c-fes expression by RNAi demonstrated a requirement for c-Fes in AML cell survival (Voisset et al., 2010). Downregulation of c-Fes by siRNA treatment was also shown to reduce proliferation of two human renal carcinoma cell lines (Kanda et al., 2009).</p><p>Angiogenesis is a common hallmark of tumorigenesis (Hanahan and Weinberg, 2000). A role for c-Fes in angiogenesis was first suggested by the observation that membrane-targeted c-fes expression led to hypervascularization and hemangioma formation in transgenic mice (Greer et al., 1994). Subsequently, c-Fes kinase activity was shown to contribute to FGF-2-induced chemotactic cell migration and tube formation by brain capillary endothelial cells (Kanda et al., 2000). Further studies confirmed that c-Fes is a common mediator of PI3-kinase activation by numerous angiogenic factors, including VEGF-A, Ang1 and Ang2 (Kanda et al., 2007).</p><p>Delineating a role for c-Fes in cancer is complicated by observations that c-Fes may also fulfill the role of a tumor suppressor. Large-scale sequencing of the tyrosine kinome in multiple colorectal tumor cell lines identified c-fes as a one of only a small number of consistently mutated genes (Bardelli et al., 2003). Subsequent work showed that none of the reported mutations stimulated c-Fes kinase activity, and several impaired kinase function, consistent with a tumor-suppressor role (Delfino et al., 2006; Sangrar et al., 2005). Expression of c-Fes is readily detected in normal colonic epithelium, but is frequently absent in matched tumor samples as well as in human colorectal cancer cell lines as a result of extensive promoter methylation (Delfino et al., 2006; Shaffer and Smithgall, 2009). In a mouse model of breast cancer, tumor onset was accelerated in homozygous-null c-fes mice, and this effect was rescued by a c-fes transgene (Sangrar et al., 2005). Taken together, these data point to a tumor suppressor function for c-Fes in some epithelial cancers.</p><p>Spearheaded by the clinical success of the Bcr-Abl inhibitor imatinib in chronic myelogenous leukemia, kinases have become the focus of major drug discovery efforts as targets for anti-cancer drug therapy (Zhang et al., 2009). As summarized above, mounting evidence points towards a role for c-Fes in human cancer through its involvement in cell proliferation, survival signaling, and angiogenesis, making it an attractive candidate for drug targeting (Kanda and Miyata, 2011). Selective small molecule inhibitors are urgently needed to clarify the roles of c-Fes as dominant oncogene vs. tumor suppressor depending upon the cellular context. Despite the intriguing biology associated with c-Fes, no inhibitors with a useful level of selectivity and cellular activity have been reported to date.</p><p>In this study, we report the discovery and characterization of potent c-Fes tyrosine kinase inhibitors with cellular activity. Using a recombinant c-Fes protein consisting of the SH2 and kinase domains, we first screened a kinase-biased small-molecule library using an in vitro kinase assay. 'Hit' compounds were then tested for their ability to inhibit c-Fes autophosphorylation and microtubule association in COS-7 cells and for their effect on rodent fibroblast transformation driven by constitutively active c-Fes mutants. Using these screens we identified both Type I and Type II c-Fes kinase inhibitors from diverse chemical classes, including diaminopyrimidines, pyrazolopyrimidines, pyrrolopyridines and pyrazines, with activity against c-Fes both in vitro and in vivo. Type I inhibitors bind to the ATP-binding site with the kinase assuming an 'active' conformation defined by the 'DFG-motif' of the activation loop adopting an 'in' conformation conducive to substrate binding. Type II inhibitors bind to the 'inactive' conformation with the 'DFG-motif' in an 'out' conformation blocking access to the substrate binding site (Liu and Gray, 2006).</p><p>Surprisingly, we discovered that TAE684, a compound previously identified as a potent and selective inhibitor of the anaplastic lymphoma kinase [Alk; (Galkin et al., 2007)], is also a potent inhibitor of c-Fes both in vitro and in vivo. We were able to obtain a crystal structure of the c-Fes SH2- kinase region in complex with TAE684 which will guide further modifications to enhance potency and selectivity. Finally, using several of these inhibitors as chemical probes, we were able to define a novel role for endogenous c-Fes activity in osteoclast differentiation from macrophages for the first time. Our findings represent an important first step towards the development of potent and selective inhibitors of c-Fes, which will have utility in the elucidation of the roles of this enigmatic kinase in normal cellular physiology and tumorigenesis.</p><!><p>The Z′-Lyte fluorescence resonance energy transfer (FRET)-based assay platform for high-throughput assessment of kinase activity has been successfully used in a chemical library screen to identify inhibitors of HIV Nef-induced Hck tyrosine kinase activity (Emert-Sedlak et al., 2009; Rodems et al., 2002). Here we used this assay and a recombinant catalytically active fragment of c-Fes, consisting of the SH2 and kinase domains (SH2-KD) for which the crystal structure is known (Filippakopoulos et al., 2008), to screen a kinase-biased library of small molecules. A total of 586 compounds were initially screened for inhibition of SH2-KD at a concentration of 1 μM (Figure S1). We found that 19 compounds inhibited Fes SH2-KD kinase activity by 90% or more, while an additional 13 compounds inhibited kinase activity by 80–90%. Of the inhibitors identified in the primary screen, 21 compounds representing eight diverse chemical scaffolds were chosen for further analysis. The inhibitory activities of these compounds were verified in dose-response experiments and IC50 values were determined by curve fitting (Figure S2). The IC50 values for all 21 compounds were in the sub-micromolar range (Table 1), with the lowest values observed for the pyrazolopyrimidine WZ-4-49-8 (IC50 67 nM) and the diaminopyrimidine TAE684 (IC50 118 nM). In addition to these type I inhibitors, two putative type II inhibitors, HG-7-27-01 (IC50 224 nM) and HG-7-92-01 (IC50 292 nM), were also discovered in this compound set. Structures and concentration-response curves for these four inhibitors are presented in Figure 1. Chemical syntheses and characterization data for these four compounds are presented in the Supplemental Experimental Procedures section.</p><!><p>The kinome-wide selectivity for each of the four lead compounds was assessed using KINOMEscan™ screening technology, a high-throughput method for screening kinase inhibitors against a panel of either 353 kinases (for TAE684 and HG-7-92-01) or 442 kinases (for WZ-4-49-8 and HG-7-27-01). A kinome interaction map was constructed from the resulting data for each compound (Figure S3A). This approach revealed that the diaminopyrimidine TAE684, a type I kinase inhibitor, possessed a broad selectivity profile with a selectivity score of 0.38 (Karaman et al., 2008). In contrast, the pyrazolopyrimidine type I inhibitor WZ-4-49-8, which possesses an ortho-ethoxy group, displayed an exceptional selectivity profile with a selectivity score of only 0.027. The two type II inhibitors HG-7-92-01 and HG-7-27-01 displayed intermediate selectivity profiles (0.18 and 0.16, respectively). Selectivity profiling was intentionally performed at relatively high inhibitor concentrations (10 μM) to identify the full spectrum of possible targets. A complete listing of all of the kinases profiled in the screen and the relative selectivity profiles of each compound are shown in Supplemental Table S1. While this approach provides a broad measure of kinase selectivity, it is important to note that it does not necessarily translate to inhibition of kinase activity in biochemical or cellular assays. For example, TAE684 and WZ-4-49-8 scored as strongly active for Erk2 in the KINOMEscan™ assay (displacement scores of 0.3 and 0.7, respectively, Table S1). However, neither compound showed notable inhibition of Erk2 activity in kinase assays, with IC50 values for Erk2 inhibition at least 100-fold higher than those observed for the inhibition of Fes (Figure S3B,C). Further experiments will be required to determine whether any other kinases identified in this in vitro displacement assay represent true alternative targets in cells, especially at concentrations where c-Fes is inhibited.</p><!><p>The c-Fes SH2-kinase protein used in the primary screen was crystallized in complex with TAE684 and the resulting X-ray crystal structure was refined to 1.84 Å (data collection and refinement statistics are listed under Supplemental Table S2). In the crystal structure, the regulatory αC helix assumed an active conformation as indicated by the canonical salt bridge between the conserved αC glutamate (E607) and the active site lysine (K590). However, the activation segment was largely unstructured as expected for unphosphorylated, inactive c-Fes in the absence of SH2 ligands (Filippakopoulos et al., 2008). The inhibitor was very well defined by electron density (Figure 2A) and showed good shape complementarity with the c-Fes ATP binding pocket. The inhibitor pyrimidine and aniline amines formed two hydrogen bonds with the hinge backbone of V639 (Figure 2B). In addition, the 5-chloro substituent of TAE684 packed against the gatekeeper methionine (M636), an interaction that was also observed between the related inhibitor WZ-4002 and the T790M gatekeeper mutant of the EGF receptor tyrosine kinase (Zhou et al., 2009). Binding was further stabilized by hydrophobic interactions with residues A588, L638, V620, L690, V575 and I567 (Figure 2C). The piperazine moiety of TAE684 extended over helix αD and was shielded from the solvent channel by a symmetry related protein molecule.</p><p>Superimposition of the c-Fes/TAE684 co-crystal structure with that of Alk (Bossi et al., 2010) revealed an additional polar interaction of TAE684 with E1210 located in helix αD which is not present in c-Fes (Figure S4A,B). However, the electron density for the piperidine-piperazine group was not well defined in the Alk complex suggesting that this moiety is flexible. In c-Fes, the piperidine-piperazine group forms water-mediated hydrogen bonds with residues G641 and G642 located C-terminal to the hinge region (Figure S4C). Additional water-mediated hydrogen bonds were also observed between TAE684 and the active site lysine (K590), the P-loop residues F572 and N571, and D701 (Supplemental Figure S4D).</p><!><p>We next examined whether the inhibitors identified in vitro also displayed activity against full-length c-Fes in a cell-based assay. The N-terminal region of c-Fes, which is not part of the crystal structure, contains two coiled-coil homology domains that have been implicated in the regulation of c-Fes kinase activity in cells (Hellwig and Smithgall, 2011). A leucine-to-proline point mutation in the first coiled-coil domain (L145P), which has been predicted to disrupt the coiled-coil structure, strongly activates c-Fes in vivo and results in fibroblast transformation (Cheng et al., 2001). When a GFP fusion of this active c-Fes mutant is expressed in COS-7 cells, autophosphorylation of the c-Fes kinase domain activation loop on Y713 can be readily detected by immunofluorescence along with redistribution of the protein to the prominent microtubule scaffold present in this cell line (Laurent et al., 2004). Microtubule association results from c-Fes-mediated phosphorylation of tubulin, followed by association via the c-Fes SH2 domain. Association of active c-Fes with microtubules is in striking contrast to the diffuse cytoplasmic distribution of wild-type c-Fes, which is downregulated despite the high-level over-expression achievable in this cell line (Figure 3A).</p><p>Using the COS-7 expression system, we tested all 21 lead compounds from the in vitro screen for their ability to inhibit c-Fes autophosphorylation and association with microtubules in vivo. COS-7 cells were transiently transfected with the GFP-Fes-L145P fusion protein, followed by 24-hour incubation with each compound at concentrations of 1, 3 and 10 μM. Treated cells were fixed and immunostained for autophosphorylated c-Fes using a pY713-specific antibody (Takashima et al., 2003). As shown in Figure 3A, treatment with the compound TAE684 resulted in a dramatic loss of GFP-Fes localization from microtubules and concomitant loss of pY713 immunostaining. Additional experiments revealed that loss of autophosphorylation and cellular redistribution of GFP-Fes-L145P can be observed after as little as 1 hour of inhibitor treatment (data not shown). This observation suggests that inhibition of c-Fes kinase activity both reverses and prevents microtubule association.</p><p>Nine additional compounds also inhibited c-Fes-L145P autophosphorylation and microtubule association in at least a subset of cells. To quantify the effects of these inhibitors, the percentage of cells showing loss of c-Fes-L145P pY713 immunostaining was determined in three independent experiments, and the results are shown in Figure 3B. The strongest inhibition was observed with TAE684, with ~70% (at 1 μM) to ~90% (at 10 μM) of cells showing loss of c-Fes-L145P activity and microtubule association. These experiments identify TAE684 as a potent inhibitor of active c-Fes in a cellular context. The pyrazolopyrimidines WZ-4-49-1 and WZ-4-49-8 also showed strong effects in this system, with IC50 values in the low micromolar range. In contrast to these compounds, the predicted Type II inhibitor HG-7-27-01 reduced c-Fes-L145P autophosphorylation in only 10–15% of cells when tested at concentrations below the cytotoxicity threshold, despite its apparent potency in vitro. As described in the next section, this difference may be due to a preference of this compound for the downregulated conformation of the kinase domain.</p><!><p>In addition to strong association with microtubules in vivo, purified c-Fes directly phosphorylates tubulin and catalyzes tubulin polymerization in vitro (Laurent et al., 2004). In support of the inhibitor-induced changes in c-Fes-L145P autophosphorylation and microtubule localization observed in COS-7 cells, we next performed immune-complex kinase assays using purified recombinant tubulin as substrate. Flag-tagged wild-type or L145P forms of c-Fes were expressed in COS-7 cells, and immunoprecipitates were incubated with tubulin in the presence of [γ32P]ATP over a range of inhibitor concentrations. For comparative purposes, tubulin phosphorylation assays were also performed with the recombinant SH2-KD form of c-Fes used in the initial inhibitor screen. TAE648 potently inhibited tubulin phosphorylation by both wild-type and L145P c-Fes, with average IC50 values of 15 nM and 30 nM, respectively (Figure 4 and Table 2). Interestingly, the IC50 value for inhibition of wild-type full-length c-Fes is ~3-fold lower than the IC50 for the SH2-KD protein in this assay (15 nM compared to 46 nM), suggesting that TAE684 may have increased affinity for full-length c-Fes.</p><p>For HG-7-27-01, which displayed only weak inhibition of c-Fes-L145P autophosphorylation in COS-7 cells (Figure 3), inhibition of tubulin phosphorylation by c-Fes-L145P-Flag was also weak in vitro, with an IC50 value of 5.2 μM (Figure 4 and Table 2). More potent inhibition was observed for wild-type c-Fes-Flag and SH2-KD, with average IC50 values of 1.3 and 0.5 μM, respectively. Inhibitor binding in the Type II mode typically requires the kinase domain to be in an inactive conformation, with the DFG-motif rotated to an outward orientation that allows for access to a hydrophobic pocket adjacent to the ATP-binding site (Liu and Gray, 2006). The observed differences in IC50 values for inhibition of wild-type vs. the L145P and truncated forms of c-Fes by HG-7-27-01 suggest a bias of this compound towards the inactive conformation of the kinase as expected for a type II inhibitor. In addition, these results provide indirect evidence that the unique Fes N-terminal region may influence the conformation of the inhibitor binding site in the kinase domain. Potent inhibition of tubulin phosphorylation by these c-Fes kinases was also observed for WZ-4-49-8 and HG-7-92-01 with IC50 values against wild-type c-Fes-Flag of 127 nM and 306 nM, respectively (Figure S3B,C and data not shown). No significant differences in potency against wild-type c-Fes over the L145P mutant were observed in this assay for these compounds. These results are consistent with the inhibition of c-Fes autophosphorylation and microtubule association in COS cells by these compounds (Figure 3).</p><!><p>We next investigated whether the compounds that potently inhibited c-Fes activity in vitro and in the COS cell system could also suppress oncogenic transformation of cells by active forms of c-Fes. In previous work, we established that Rat-2 fibroblasts undergo rapid transformation upon stable expression of constitutively active mutants of c-Fes, including the N-terminal coiled-coil domain mutant L145P used in the COS cell assay (Cheng et al., 2001). Using a soft-agar colony-forming assay for anchorage-independent growth, we first tested the ability of TAE684 to inhibit Rat-2 transformation by two different active forms of c-Fes. Rat-2 cells expressing GFP fusions of c-Fes-L145P or an active c-Fes/v-Fps chimera (Kim et al., 2004) were grown in soft-agar in the presence of various inhibitor concentrations and the number of transformed colonies were counted two weeks later. As shown in Figure 5, TAE684 potently inhibited soft-agar colony formation by Rat-2 cells expressing either of the transforming variants of c-Fes by more than 50% at 100 nM. Growth of control Rat-2 cells in monolayer culture was only slightly reduced at this concentration of TAE684, strongly supporting a direct effect of the inhibitor on c-Fes-mediated transformation (Figure S5). Complete inhibition of c-Fes-mediated soft-agar colony formation was observed with TAE684 at 400 nM (c-Fes/v-Fps) and 600 nM (c-Fes-L145P).</p><p>To rule out a role for Src-family kinases in the inhibitory effect of TAE684 on Rat-2 fibroblast transformation, we also performed soft-agar colony assays using cells transformed by an active mutant of the Src-family kinase, Hck (Briggs and Smithgall, 1999; Lerner and Smithgall, 2002). In contrast to the strong inhibition of c-Fes-mediated transformation by TAE684, only minimal inhibition of Hck-mediated soft-agar colony formation was observed with 200 nM TAE684 (Figure 5). Complete loss of soft-agar colony growth was observed at 1 μM TAE684 with both the c-Fes and Hck transformants. However, growth of control Rat-2 cells expressing GFP alone was reduced by 50% at 1 μM TAE684 after 72 hrs and completely abolished with a concentration of 3 μM of this compound. Together, these results indicate that the effects of TAE684 on Hck-transformed fibroblasts are largely due to non-specific suppression of cell growth as concentrations approach 1 μM or higher.</p><p>One important structural difference between the c-Fes and c-Src kinase families is the identity of the amino acid that occupies the "gatekeeper" position adjacent to the ATP binding site in the kinase domain. This residue impacts the access of some small molecule inhibitors to a hydrophobic cavity that is an important determinant of inhibitor specificity. In Hck and all other members of the Src-kinase family, threonine occupies the gatekeeper position, while a bulkier methionine residue is present in this position in c-Fes. In the X-ray crystal structure of the c-Fes kinase domain, the gatekeeper methionine (M636) comes in very close contact with the chloro group of TAE684 (Figure 2). To evaluate whether this key inhibitor specificity determinant impacts the sensitivity of Hck to TAE684, we created an active form of Hck in which the gatekeeper threonine was replaced by methionine. Remarkably, this single amino acid substitution dramatically enhanced the sensitivity of Hck-transformed fibroblasts to inhibition by TAE684 in a manner very similar to cells transformed with the c-Fes mutants (Figure 5). These results suggest that the gatekeeper residue is a critical specificity determinant for TAE684.</p><p>To correlate inhibition of transforming activity with effects on kinase activity, we next treated monolayer cultures of transformed Rat-2 cells with a range of TAE684 concentrations and assayed the autophosphorylation status of each kinase by immunoblotting the cell lysates with phosphospecific antibodies against the activation loop phosphotyrosine residues (Figure 5C,D). In agreement with the selectivity of TAE684 observed in the soft agar colony assays, both c-Fes and the Hck-TMYF gatekeeper mutant were sensitive to TAE684 treatment (IC50 values of 88 nM and 685 nM, respectively), while Hck-YF autophosphorylation remained largely unaffected (IC50 > 10 μM).</p><p>Rat-2 fibroblast transformation assays were also performed with WZ-4-49-8 as well as HG-7-92-01 (Figure S5), both of which demonstrated strong inhibition of c-Fes in vitro. WZ-4-49-8 potently inhibited c-Fes-L145P-mediated soft-agar colony growth and showed remarkable selectivity for c-Fes-L145P over several active SFKs tested. This pyrazolopyrimidine inhibited c-Fes-L145P-mediated soft-agar colony formation by more than 80% at 100 nM, while cells transformed with active Hck showed no reduction in soft-agar growth at this concentration. Interestingly, Rat-2 cells transformed by either of two active SFK mutants carrying a threonine to methionine gatekeeper mutation (Hck-TMYF and Yes-TMYF) were inhibited to 61% and 25% of control values at 300 nM, respectively. These results support an important role for the gatekeeper residue in determining inhibitor specificity as described above for TAE684. HG-7-92-01 was much less potent than the other two compounds in the transformation assay, producing only 50% inhibition of c-Fes-L145P-mediated colony forming activity at 1 μM. A similar degree of inhibition was also observed for fibroblasts transformed with activated Hck as well as the Hck-TM mutant, indicating a relative lack of specificity for c-Fes in this assay.</p><!><p>Having identified a group of structurally distinct inhibitors for c-Fes kinase activity, we next investigated the utility of these compounds as chemical probes for endogenous c-Fes function. For these experiments we turned to macrophages, one of the cell lineages where c-Fes is most strongly expressed. Macrophages in turn give rise to osteoclasts, the multinucleated cells critical for resorption of bone (Boyle et al., 2003). This process is regulated in part by the coordinated action of macrophage colony-stimulating factor (M-CSF) and RANK ligand (RANKL), where M-CSF promotes macrophage proliferation and survival while RANKL induces differentiation. Given the strong expression of c-Fes in macrophages and its role myeloid differentiation (Craig, 2012), we were interested to see if c-Fes may contribute to induction of the osteoclast phenotype. To test this hypothesis, we employed both bone-marrow derived macrophages (BMM) as well as the RAW 264.7 macrophage cell line. As shown in Figure 6, treatment of BMM with M-CSF and RANKL induced strong induction of osteoclast differentiation, assessed as the formation of multi-nucleated cells that stained positive for tartrate-resistant acid phosphatase (TRAP). Remarkably, the c-Fes inhibitors TAE684 and WZ-4-49-8 (as well as the WZ-4-49-8 analog WZ-4-49-1) dramatically inhibited differentiation in this system, with IC50 values in the submicromolar range. Very similar results were observed using the same inhibitors in RAW 264.7 cells, where osteoclast differentiation was driven by RANKL in combination with vascular endothelial growth factor instead of M-CSF (Figure 6B). Importantly, none of the c-Fes inhibitors produced toxic effects on primary macrophages or RAW 264.7 cells when cultured in the absence of the differentiation inducers for eight days (data not shown).</p><p>As an independent validation of a role for c-Fes in osteoclast differentiation indicated by the inhibitors, we targeted endogenous c-Fes by transient siRNA transfection of RAW 264.7 cells (Figure S6A–C). Transfection with three murine c-Fes-specific siRNAs reduced c-Fes mRNA and protein expression by ~50% (Figure S6A,B), a level comparable to published targeting efficiencies by transient siRNAs in RAW264.7 cells (Wang and Grainger, 2010). Targeting of c-Fes had no effect on the mRNA levels of the closely related Fer kinase (Figure S6A). Transfection with c-Fes specific siRNAs suppressed RANKL-driven formation of TRAP-positive polykaryons when compared to mock-transfected cells and cells transfected with nonspecific control siRNA (Figure S6C). In addition, mRNA analyses of siRNA transfected cells by quantitative real-time RT-PCR revealed that knockdown of c-Fes reduced the expression of the osteoclast marker Cathepsin K in the cytokine-stimulated cell population, while basal expression levels of Cathepsin K in unstimulated RAW 264.7 remained unchanged (Figure 6C). This reduction in RANKL-stimulated upregulation of Cathepsin K mRNA levels is reproduced by treatment with c-Fes inhibitor compounds at submicromolar concentrations (Figure 6D). Taken together, these results suggest that c-Fes activity is required for the differentiation of osteoclasts from macrophages, and identify c-Fes as a possible therapeutic target for the treatment of bone loss associated with cancer metastasis and aging (Roodman, 2004).</p><p>Because the kinome-selectivity profile identified Erk kinases as possible alternative targets for TAE684 and WZ-4-49-8, we performed a control osteoclast differentiation assay from RAW 264.7 macrophages in the presence of the Erk inhibitor FR180204 (Figure S6C,D). Treatment with 1 μM or 10 μM FR180204 did not influence osteoclast differentiation, ruling out the Erk pathway as a target for TAE684 and WZ-4-49-8. These results are consistent with the very low potency of TAE684 and WZ-4-49-8 towards Erk in an in vitro kinase assay (Figure S3).</p><!><p>In this report, we describe the first inhibitor discovery campaign directed at the c-Fes protein- tyrosine kinase. Identified in the context of several transforming retroviruses almost thirty years ago, c-Fes kinase activity has been implicated in diverse physiological processes, including innate immune receptor signaling, myeloid differentiation, and vasculogenesis. c-Fes has also been implicated in tumorigenesis, where it may act as a dominant oncogene or tumor suppressor depending upon the cellular context. Despite the importance of c-Fes to normal cellular function and disease, no useful inhibitors of c-Fes kinase activity have been described. Through a combination of in vitro and cell-based focused library screens, we identified eight distinct chemotypes with potent activity against c-Fes. One of the most potent compounds is TAE684, a diaminopyrimidine previously developed as an inhibitor of the Alk kinase associated with anaplastic lymphoma. The X-ray crystal structure of TAE684 bound to the c-Fes SH2-kinase region reveals the basis of the observed specificity, including a role for the gatekeeper position in the c-Fes kinase domain. The importance of the c-Fes gatekeeper methionine for TAE684 binding was confirmed by replacement of the threonine residue found in this position in several members of the Src kinase family. Substitution of Src family kinase gatekeeper residues with methionine dramatically enhanced their sensitivity to inhibition by TAE684 in a cell-based assay. In addition to TAE684, the novel pyrazolopyrimidines WZ-4-49-1 and WZ-4-49-8 also showed remarkable potency and selectivity for c-Fes. Using these three inhibitors as chemical probes, we discovered a novel role for endogenous c-Fes function in osteoclast formation from cultured macrophages, a physiological site of c-Fes expression. This observation identifies c-Fes as a possible therapeutic target in osteoporosis as well as osteolytic bone metastasis often associated with advanced cancers. In summary, our study provides a strong rationale for future development of c-Fes inhibitors with enhanced potency and specificity. Such compounds will represent important tools for elucidating the complex roles of c-Fes in innate immunity, differentiation, and cancer etiology.</p><!><p>The chemical library screen for c-Fes inhibitors was performed using the Z′-Lyte kinase assay system (Invitrogen/Life Technologies). This FRET-based assay utilizes a synthetic substrate peptide (Tyr2) labeled with the fluorophores coumarin and fluorescein on opposite peptide termini (Rodems et al., 2002). In a two-step reaction process, the substrate peptide is first incubated with the kinase to allow for phosphorylation of a single tyrosine residue. In the second step, site-specific proteolytic cleavage of nonphosphorylated but not of phosphorylated peptide occurs. The coumarin and fluorescein fluorophores constitute a FRET-pair and peptide cleavage results in loss of the FRET signal. Kinase activity is monitored by measuring the emission ratio after excitation of the donor fluorophore: EmissionRatio=CoumarinEmission(445nm)FluoresceinEmission(520nm)</p><p>High kinase activity results in a low emission ratio, while inhibition of kinase activity results in a high emission ratio.</p><p>Reactions were carried out in a 384-well plate format in a volume of 10 μL according to the manufacturer's protocols. Initial titration of kinase input verified linear reaction conditions for 25 ng of recombinant c-Fes SH2-KD with a 1 h reaction time at room temperature in the presence of 50 μM ATP. The chemical library screen was performed at 1 μM final compound concentration. IC50 values of "hit" compounds, defined as those inhibiting Fes activity by at least 80% in the primary screen, were determined using 10-point serial dilutions of compounds ranging from 0.5 nM to 10 μM. Results were graphed and IC50 values determined by curve fitting using Prism (Graph Pad).</p><!><p>COS-7 cells (5 × 105) were seeded onto 60 mm culture dishes and grown at 37 °C for 24 h. The cells were transfected with pCDNA3.1 vectors carrying c-Fes-Flag or c-Fes-L145P-Flag and grown for an additional 20 h. Cell extracts were prepared by sonication in 600 μL of Fes lysis buffer [50 mM Tris-HCl, pH 7.4, 1 mM EDTA, 50 mM NaCl, 1 mM MgCl2, 0.1% Triton X-100, protease inhibitor cocktail (Calbiochem), 2 mM Na3VO4, 20 mM NaF]. Kinases were immunoprecipitated using 2 μg of anti-Flag antibody (sc-807, Santa Cruz). The immunoprecipitates were washed twice in RIPA buffer, twice in kinase assay buffer (50 mM Hepes, pH 7.4, 10 mM MgCl2) and resuspended in kinase assay buffer to give a final volume of 100 μL. For reactions using recombinant c-Fes SH2-KD, 25 ng of kinase were diluted in 10 μL kinase assay buffer.</p><p>For kinase reactions, 10 μL of each immunoprecipitate was incubated with 2 μg of bovine tubulin (Cytoskeleton), 10 μCi of [γ32P]ATP, and inhibitor in kinase assay buffer in a total volume of 20 μL. The final DMSO concentration in all reactions was 1%. Following incubation for 10 min at 30 °C, reactions were quenched by addition of SDS-PAGE sample buffer and incubation at 95 °C for 5 min. The reaction products were analyzed by SDS gel electrophoresis followed by transfer to PVDF membranes and autoradiography. Consistent input of Fes and tubulin was verified by western blotting and Coomassie staining of the membrane, respectively. Relative tubulin phosphorylation was quantified by image analysis of the autoradiographs using ImageJ. IC50 values were determined by plotting phosphotubulin levels relative to the DMSO control against the inhibitor concentration followed by curve fitting using Prism.</p><!><p>Association of active c-Fes with microtubules can be readily monitored by expression of GFP-fusions of active c-Fes mutants in transfected COS-7 cells (Laurent et al., 2004). To determine the effect of inhibitor compounds on c-Fes autophosphorylation and subcellular localization, COS-7 cells were grown in 48-well plates. Inhibitors were added 24 h later in 0.2% DMSO. Immediately following addition of the inhibitors, the cells were transfected with expression plasmids encoding GFP-Fes-L145P or the wild-type GFP-Fes control using Fugene 6 transfection reagent (Roche). Cells were fixed and immunostained 24 h later with a pY713-specific antibody (Takashima et al., 2003) and a secondary antibody conjugated to Texas Red (Southern Biotech). Fluorescence and subcellular localization were evaluated using a Nikon TE300 inverted microscope equipped with a SPOT CCD high-resolution digital camera. For the images in Figure 3A, COS-7 cells were grown, treated with inhibitors and subsequently fixed and immunostained on glass cover slips. Cover slips were mounted on slides and images were taken using an Olympus Fluoview FV1000 confocal microscope. Image acquisition settings were unchanged between treated and control slides.</p><!><p>The construction of GFP-fusions of c-Fes-L145P and the c-Fes/v-Fps chimera in the retroviral expression vector pSRαMSVtkneo, the production of recombinant retroviruses and the retroviral infection of Rat-2 fibroblasts are described in detail elsewhere (Cheng et al., 2001). Stably transfected Rat-2 cells were maintained in DMEM supplemented with 10% FBS and 400 μg/ml G418. For the soft agar colony formation assays, 35 mm culture dishes were prepared with bottom layers of 1 mL of 0.5% Seaplaque agarose in growth medium. This cell-free layer also contained test compounds at twice the final assay concentration and 0.2% DMSO. Bottom layers were solidified at 4 °C. Retrovirally transduced Rat-2 cells were trypsinized to a single-cell suspension and resuspended at a concentration of 10,000 cells/ml in medium containing 0.33% Seaplaque agarose. Each cell suspension (1 ml) was layered onto the drug-containing bottom layer and allowed to solidify at 4 °C, after which the plates were incubated at 37 °C for 13 d. Soft agar colonies were stained for 1 h at 37 °C with thiazolyl blue tetrazolium bromide (MTT) in growth medium (1 mg/ml; 1 ml/plate). Colony counts were obtained from scanned images of the plates with QuantityOne software (BioRad). A sensitivity setting of 5 and averaging setting of 5 was applied to all counts. For the immunoblot analyses in Figure 5C and D, cells were seeded in 12-well plates at 50,000 cells/well and grown for 48 h. Inhibitors were added at the indicated concentrations and cells were cultured for an additional 16 h. Extracts were prepared using RIPA buffer containing 1X protease inhibitors (Calbiochem), 1 mM Na3VO4 and 1 mM NaF. Anti-Fes pY713, anti-Src pY418 (Invitrogen), anti-Hck sc-72 and anti-Fes sc-7670 (Santa Cruz) were used for immunodetection. IC50 values were determined using Prism.</p><!><p>Six-week-old male ddY mice were purchased from Kyudo Co. Ltd., Tosu, Japan. All animal work was performed at the Nagasaki University Graduate School of Biomedical Sciences with Institutional Animal Care and Use Committee approval in accordance with institutional guidelines. Hematopoietic cells were isolated by the perfusion of femurs and tibias with α-MEM (Sigma-Aldrich Japan) containing 10% FBS (Invitrogen) and mononuclear cells were enriched by Histopaque-1077 density gradient centrifugation (Sigma-Aldrich Japan, Tokyo, Japan). Mononuclear cells were cultured in α-MEM containing 10% FBS overnight, and nonadherent cells were collected by centrifugation and seeded into 6 cm dishes in α-MEM containing 10% FBS and 20 ng/ml recombinant human M-CSF (R&D Systems). Seven days later, adherent cells (BMM) were harvested with trypsin and seeded into 48-well plates for osteoclast differentiation in growth medium containing 20 ng/ml M-CSF at a density of 2 × 104 cells/cm2.</p><!><p>BMMs were cultured in the presence of recombinant human soluble RANK ligand (RANKL; PeproTech, Rocky Hill, NJ) and M-CSF for eight days. RAW 264.7, obtained from the American Type Culture Collection, were cultured for 6 days in α-MEM containing 10% FBS, 100 ng/mL VEGF-A (PeproTech) and 100 ng/mL RANKL. Cultures were fixed with 10% buffered formalin, followed by an ethanol:acetone (1:1) wash. Tartrate-resistant acid phosphatase (TRAP) positive cells were visualized using the TRAP staining kit (Sigma-Aldrich) and the number of TRAP-positive multinucleated cells (at least 3 nuclei per cell) was counted under the microscope. For each experiment, cells in three separate wells were counted and the mean osteoclast number/well ± SD was calculated. Statistical analysis was performed using Mann-Whitney's U-test, and differences between samples were considered significant when the P-value was < 0.05.</p><!><p>Synthetic routes to compounds TAE684, WZ-4-49-8, HG-7-92-01, and HG-7-27-01 are provided in the Supplemental Experimental Procedures.</p><!><p>The recombinant c-Fes SH2-kinase domain protein was expressed in E. coli and purified as described previously (Filippakopoulos et al., 2008). Crystallization was conducted using the sitting drop vapor diffusion method at 4 °C. Aliquots of the purified proteins were dispensed for crystallization using a Mosquito crystallization robot (TTP Labtech, Royston UK). Coarse screens were typically set up on Greiner 3-well plates using three different drop ratios of precipitant to protein per condition (2:1, 1:1 and 1:2; 150 nl final volume). c-Fes crystals with TAE684 (1 mM final concentration) were grown by mixing 100 nl of the protein (6.0 mg/ml) with a 50 nl of reservoir solution containing 0.1 M SPG pH 8.0 and 30% PEG 1000. Crystals were flash frozen in liquid nitrogen. X-Ray data were collected in-house on an FR-E Superbright source using an RAXIS IV plate detector at 1.542 Å. Indexing and integration were carried out using MOSFLM (Leslie and Powell, 2007) and scaling was performed with SCALA (Evans, 2007). Initial phases were calculated by molecular replacement with PHASER (McCoy et al., 2005) using the known model of c-Fes SH2-kinase (PDB: 3BKB) (Filippakopoulos et al., 2008). Initial models were built by ARP/wARP (Perrakis et al., 1999) and building was completed manually with COOT (Emsley and Cowtan, 2004). Refinement was carried out in REFMAC5 (Murshudov et al., 1997). Thermal motions were analyzed using TLSMD (Painter and Merritt, 2006) and hydrogen atoms were included in late refinement cycles. Crystallographic data and refinement statistics can be found in Supplemental Table S2. The crystal coordinates and structure factor file will be deposited in the RCSB Protein Data Bank prior to publication.</p>

PubMed Author Manuscript

Mutation of the eunicellane synthase Bnd4 alters its product profile and expands its prenylation ability

Bnd4 catalyzes the first committed step in the biosynthesis of the bacterial diterpenoid benditerpenoic acid and was the first eunicellane synthase identified from nature. We investigated the catalytic roles of the aromatic residues in the active site of Bnd4 through a series of mutation studies. These experiments revealed that large hydrophobic or aromatic side chains are required at F162 and Y197 for eunicellane formation and that selected mutations at W316 converted Bnd4 into a cembrane synthase. In addition, the Bnd4 Y197A mutant expanded the native prenylation ability of Bnd4 from accepting C5 and C10 prenyl donors to C15. This study supports the mechanism of eunicellane formation by Bnd4 and encourages further engineering of terpene synthases into practical and efficient prenyltransferases.

mutation_of_the_eunicellane_synthase_bnd4_alters_its_product_profile_and_expands_its_prenylation_abi

<p>Terpenoids are the most structurally diverse family of natural products. 1,2 Most terpenoids possess polycyclic carbon skeletons that arise from a family of enzymes, terpene synthases (TSs). TSs use carbocation chemistry to catalyze complex cyclization reactions using acyclic prenyl diphosphates. 3,4 After the initial carbocation is formed, either by diphosphate abstraction (type I TS) or protonation (type II TS), TSs exquisitely control product formation by chaperoning the extremely reactive cation through a series of intermediates before final cation quench. Impressively, TSs guide these reaction cascades, using their hydrophobic and aromatic active sites to stabilize the innately reactive intermediates via cation-p interactions, [5][6][7] to frequently provide a single stereoselective product. All canonical TSs share the same overall structures, 3,8 but their sequence diversities create unsolved challenges in understanding sequence-function relationships.</p><p>Recently, benditerpenoic acid, a bacterial eunicellane diterpenoid harboring a 6,10-bicyclic scaffold, was isolated from Streptomyces sp. (CL12-4). 9 In collaboration with Prof. Loesgen, we also identified its biosynthetic gene cluster and characterized the first eunicellane-forming TS, Bnd4, which produces benditerpe-2,6,15-triene (1) (Figure 1A). 9 We then investigated the mechanism of eunicellane formation through a series of labeling studies, quantum chemical calculations, and mutagenesis experiments. 10 However, key questions remained including how the aromatic residues in Bnd4 assist in controlling cyclization and which, if any, residues are directly involved in final deprotonation to form 1. We set out to answer those questions with a series of mutation experiments focused on the aromatic residues of Bnd4. In the Bnd4 model, the active site is lined with five aromatic residues, W67, F162, Y197, W316, and Y323, all of which are within 4 Å of docked geranylgeranyl diphosphate (GGPP) (Figure 1B). These residues are also strictly conserved amongst the eight other Bnd4 homologues (Fig. S1). 9,10 Based on the mechanism of Bnd4, which is supported by isotope labeling studies and quantum chemical calculations, carbocations are sequentially formed on C1, C11, C1, and C15 (Figure 1A). 10 Y323 is positioned near C1 of GGPP (3.1 Å) suggesting it may stabilize either the initial carbocation after diphosphate abstraction or the monocyclic intermediate after the 1,3hydride shift, or both. W316, W67, Y197, and F162 form one wall of the hydrophobic active site with both Y197 and F162 near the methyl groups on C15. Given the proximity of Y197 to C16 (3.5 Å), we hypothesized that Y197 may act as the base that deprotonates C16 to complete the reaction.</p><p>We first performed alanine scanning mutagenesis on the five aromatic residues in the active site. Four of the five mutants, Bnd4 F162A , Bnd4 Y197A , Bnd4 W316A , and Bnd4 Y323A , were soluble (Figure S2); Bnd4 W67A was insoluble and excluded from further study. In vitro incubation of these Bnd4 mutants with GGPP resulted in the appearance of new peaks 2-4 (Figure 2A). Bnd4 Y197A abolished production of 1 and gave a polar major peak, 3, and two minor peaks, 2 and 4; Bnd4 F162A similarly produced 3 and 4, but still produced a significant amount of 1. Bnd4 W316A produced 1 and several additional minor peaks while the product profile of Bnd4 Y323A did not change from that of native Bnd4. Considering the proposed location of Y323 near C1 of GGPP, (Figure 1B) it was surprising that Bnd4 Y323A did not affect cyclization activity. To facilitate the isolation and structural characterization of enzymatic products, we established a new GGPP overproduction system in E. coli. Previously, we employed published GGPP overproduction systems 9,11,12 with varying levels of terpene production and reproducibility. Inspired by recent reports of an artificial pathway for isoprenoid biosynthesis in E. coli that leverages two kinases to sequentially phosphorylate exogenously added isoprenol, [13][14][15] we emulated these systems to establish a reliable GGPP overproduction system. We cloned two kinases, namely hydroxyethylthiazole kinase (ThiM) from E. coli and isopentenyl phosphate kinase (IPK) from Arabidopsis thaliana, with isopentenyl diphosphate isomerase (IDI) from E. coli and a putative GGPP synthase (Bnd3) from Streptomyces sp. (CL12-4) 9 into one operon under control of a single T7 promoter with a ribosome binding site upstream of each gene (pET28a-MKI4 or pJR1064; Figure S3). To test the ability of the MKI4 system to overproduce GGPP, we transformed pET28a-MKI4 into E. coli harboring Bnd4. Under the conditions tested, 1 was consistently produced at a titer of 32 mg L -1 (Figure S3). 9 With a new GGPP overproduction system in hand, we set out to isolate and identify the new peaks produced by the aforementioned Bnd4 mutants. Large-scale (12× 1-L) cultures of E. coli strains harboring the MKI4 system with the individual Bnd4 mutants led to the identification of bspringene (2), geranylgeraniol (3, GGOH), and geranyllinalool (4) (Figures 2 and S4-11, SI). Both Bnd4 F162A and Bnd4 Y197A produced the acyclic terpenes 3 and 4; however, Bnd4 F162A retained its ability to form 1 while Bnd4 Y197A did not (Figure 2B). This suggested that Y197 is an important player in the formation of the eunicellane scaffold. To investigate the role of Y197, we additionally created Bnd4 Y197F but its cyclization activity was unaffected (Figure 2B). Considering the proximity of F162 to Y197 (Figure 1B), we tested Bnd4 F162Y and the double mutants Bnd4 Y197A/F162A and Bnd4 Y197F/F162Y . Retention of aromaticity at 162 and 197 did not negatively affect the formation of 1 while the double Ala mutant almost completely abolished all activity. We also mutated Y197 to Trp, His, Met, and Leu and each of these mutants showed similar activities to that of native Bnd4 (Figure S12), suggesting that hydrophobicity at Y197 is sufficient to support eunicellane cyclization and deprotonation. Re-inspection of the active site revealed that the backbone carbonyls of L90 and V192 are both less than 3.4 Å away from the methyls on C15. 9 One of these carbonyls may act as the base for final deprotonation of the benditerpe-2,6,15-triene skeleton. 16 The product profile of Bnd4 W316A was significantly different than that of the other mutants tested. W316 is found within the WxxxxxRY motif, which is highly conserved among bacterial diterpene synthases and proposed to guide product formation. 17 W316 was additionally mutated to Tyr, Phe, and His to assess its impact on cyclization. In vitro, Bnd4 W316Y and Bnd4 W316F showed activity similar to that of native Bnd4, although the production of 1 slightly decreased (Figure 2C). The product profiles of Bnd4 W316H paralleled that of Bnd4 W316A , showing continued production of 1 and the appearance of new peaks 5-8; products 5 and 7 were the major peaks of Bnd4 W316H and Bnd4 W316A , respectively (Figure 2C). Using the MKI4 system, we identified four cembranoids from E. coli producing Bnd4 W316A or Bnd4 W316H : nephthenol (5), cembrene C (6), cembrene A (7), and the isopropylidene isomer of cembrene C (8) (Figures 2 and S13-22, Table S4). Products 5 and 8 were identified by comparison with the known products of DtCycA (vide infra). The formation of these cembranoids clearly indicates that substitution of W316 perturbs the active site cavity enough to alter the binding orientation of GGPP such that C1 and C14 are in proximity to each other and can form the 14membered macrocycle directly after diphosphate abstraction.</p><p>Only four native bacterial TSs are known to produce the cembrane skeleton and while cembranoids are common in marine organisms, particularly coral, 18 there have been no cembranoid natural products isolated from bacteria; 2 the cyanobacterial tasihalides were speculated to arise from an oxygenated cembrane diterpenoid. 19 DtcycA and DtcycB from Streptomyces sp. SANK 60404 produce (R)-5 and 8 and (R)-5 and (R)-(-)-7, respectively; 20 the gene product of rxyl_0493 from Rubrobacter xylanophilus produces 6, 21 and cembrene A synthase from Allokutzneria albata produces (S)-(+)-7. 22 Interestingly, DtcycA possesses an A 321 xxxxxRY motif in place of the expected WxxxxxRY motif (Figure S23), which made us speculate if DtcycA could be engineered to biosynthesize polycyclic diterpenes such as the eunicellane skeleton through a single A321W mutation. We thus obtained a synthetic dtcycA gene and confirmed that DtcycA produces 5 and 8 with minor amounts of 6 and 7 (Figure 2D). Contrary to our hypothesis, the product profile of DtcycA A321W did not change from that of native DtcycA indicating that the residue at 321 is not solely responsible for cembrene formation. Additional studies are needed to identify if and how cembrene synthases can be engineered into polycyclic-forming diterpene synthases.</p><p>Finally, given our recent finding that bacterial diterpene synthases also catalyze the prenylation of small molecules using prenyl diphosphates that are shorter than their native substrates, 23 we saw an opportunity to assess the prenylation activity of Bnd4 mutants. Bnd4 and other diterpene synthases such as CotB2 were shown to use both dimethylallyl diphosphate (DMAPP, C5) and geranyl diphosphate (GPP, C10) as prenyl donors but did not show prenylation activity with farnesyl diphosphate (FPP, C15) or GGPP (C20). 23 We proposed that the efficient production of GGOH (3) by Bnd4 Y197A provided an active site that may be more amenable to using longer prenyl donors for prenylation. To test this hypothesis, we incubated Bnd4 Y197A with indole and FPP or GGPP. One major peak, which was determined to be 3-farnesylindole (9, Figure S24-27, SI), and a related minor peak was identified after incubation with FPP (Figure 3); indole was not prenylated by GGPP (Figure S28). To evaluate if the corresponding residues of Y197 in other diterpene synthases can expand their prenylation activities, we constructed the CotB2 W186A mutant. As proposed, incubation of indole and FPP with CotB2 W186A with indole and FPP also resulted in the production of 9 (Figure 3). Previous mutations of W186 in CotB2 resulted in early termination of the cyclization cascade resulting in 7 (W186L), 3,7,18dolabellatriene (W186L/H), cyclooctat-7-en-3-ol (W186F/H), and 3,7-dolabelladiene-9-ol and cyclooctat-6-en-8-ol (W186F). 17,24,25 This data, along with the formation of 3 by Bnd4 Y197A , indicates that changes in the active sites of TSs that alter the natural cyclization reactions may provide opportunities to engineer TSs into prenyltransferases. One of the grand challenges in enzymology is the ability to predict substrate, reaction, and product from protein sequence alone. 26 To realize this goal in terpene enzymology, it is important to understand the roles of amino acids in substrate binding and catalysis to shape general principles of TSs. In this study, we investigated the roles of the aromatic residues in the active site of Bnd4, a bacterial eunicellane synthase. Using mutagenesis and an improved diterpene production system in E. coli, we identified that Y197 and W316 are key players for eunicellane formation. In addition, we engineered Bnd4 and CotB2 into prenyltransferases that accept FPP as a prenyl donor, thus expanding the ability of diterpene synthases to prenylate small molecules. Future studies are targeted to better understand how bacterial diterpene synthases control eunicellane formation and stereoselectivity.</p><p>ASSOCIATED CONTENT Supporting Information. The Supporting Information is available free of charge on the ChemRxiv website. Methods; strains, plasmids, and primers used in this study (Tables S1-S3); summary of NMR data for compounds 6 and 7 (Table S4); sequence alignments of relevant diterpene synthase (Figures S1 and S23); SDS-PAGE analysis of purified proteins (Figure S2); diterpene overproduction system in E. coli (Figure S3); NMR spectra of compounds 2-4, 6, 7, and 9 (Figures S4-S11</p>

ChemRxiv

Structural and Functional Characterization of a Ruminal β-Glycosidase Defines a Novel Subfamily of Glycoside Hydrolase Family 3 with Permuted Domain Topology*

Metagenomics has opened up a vast pool of genes for putative, yet uncharacterized, enzymes. It widens our knowledge on the enzyme diversity world and discloses new families for which a clear classification is still needed, as is exemplified by glycoside hydrolase family-3 (GH3) proteins. Herein, we describe a GH3 enzyme (GlyA1) from resident microbial communities in strained ruminal fluid. The enzyme is a β-glucosidase/β-xylosidase that also shows β-galactosidase, β-fucosidase, α-arabinofuranosidase, and α-arabinopyranosidase activities. Short cello- and xylo-oligosaccharides, sophorose and gentibiose, are among the preferred substrates, with the large polysaccharide lichenan also being hydrolyzed by GlyA1. The determination of the crystal structure of the enzyme in combination with deletion and site-directed mutagenesis allowed identification of its unusual domain composition and the active site architecture. Complexes of GlyA1 with glucose, galactose, and xylose allowed picturing the catalytic pocket and illustrated the molecular basis of the substrate specificity. A hydrophobic platform defined by residues Trp-711 and Trp-106, located in a highly mobile loop, appears able to allocate differently β-linked bioses. GlyA1 includes an additional C-terminal domain previously unobserved in GH3 members, but crystallization of the full-length enzyme was unsuccessful. Therefore, small angle x-ray experiments have been performed to investigate the molecular flexibility and overall putative shape. This study provided evidence that GlyA1 defines a new subfamily of GH3 proteins with a novel permuted domain topology. Phylogenetic analysis indicates that this topology is associated with microbes inhabiting the digestive tracts of ruminants and other animals, feeding on chemically diverse plant polymeric materials.

structural_and_functional_characterization_of_a_ruminal_β-glycosidase_defines_a_novel_subfamily_of_g

Introduction<!>Library Screening<!>Biochemical Characterization of GlyA1<!><!>Biochemical Characterization of GlyA1-ΔCt<!>Crystal Structure Determination<!><!>Permuted Domain Topology of GlyA1<!><!>Permuted Domain Topology of GlyA1<!>Architecture of the Active Site<!><!>Architecture of the Active Site<!>SAXS Analysis of GlyA1<!><!>SAXS Analysis of GlyA1<!><!>GlyA1 Phylogenetic Analysis<!><!>Discussion<!>Reagents and Strains<!>Metagenomic Library Screening and Positive-insert Sequencing<!>Cloning of glyA1 and Genetic Constructs in pQE80L Plasmid<!>Site-directed Mutagenesis<!>Gene Expression and Protein Purification<!>Biochemical Assays<!>Crystallization Data Collection and Crystal Structure Determination<!>SAXS Measurements<!>Sequence Analysis and Construction of a Neighbor-Joining Tree<!>Author Contributions<!>

<p>Family 3 of glycoside hydrolases (GH3)3 contains about 11,000 entries among which are diverse enzyme activities, including β-glucosidase, β-xylosidase, exo-chitosanase, β-N-acetylglucosaminidase, glucocerebrosidase, exo-1,4-β-glucosidase, and exo-1,3/1,4-β-glucanase, that have been characterized (1). A few reported cases are the bifunctional α-l-arabinopyranosidase/β-galactosidase (2), N-acetyl-β-glucosaminidase/β-glucosidase (3), β-glucosidase/cellodextrinase (4), β-xylosidase/α-l-arabinofuranosidase (5), and β-glucosidase/β-xylosidase (6). They are retaining enzymes that remove single glycosyl residues from the non-reducing end of their substrates. Therefore, they perform catalysis by a two-step mechanism through a covalent enzyme-glycon intermediate, which is subsequently hydrolyzed via an oxocarbenium ion-like transition state.</p><p>Despite the high number of known GH3 sequences, structural knowledge on members of the GH3 family was absent until 1999, when the three-dimensional structure of the β-d-glucan exohydrolase Exo1 from Hordeum vulgare (barley) was reported (7). This study showed the core structure of most GH3 enzymes consisting of an N-terminal (α/β)8 barrel domain 1, which houses the active site pocket and the nucleophile, and a C-terminal (α/β)6-sandwich domain 2, containing the acid/base catalyst. The contribution of different domains in supplying crucial catalytic residues was a highly unusual feature of GH3 enzymes. Furthermore, in the last few years many new structural studies have shown a great variety in domain composition and arrangement of typical GH3 β-glycosidases, having up to four separate domains (8–15). Although this variety produces a shift in the sequence position of the acid/base catalyst, the known structures revealed that its structural location is well conserved among the different members. In contrast, several reported structures have revealed a more uniform pattern of the β-N-acetylglucosaminidases (NagZ) members showing that, despite a few having two domains, most Gram-negative bacteria encode single domain enzymes, and all of them have the acid/base catalyst in an unusual histidine/aspartate dyad located in a flexible loop of the (α/β)8 barrel (16). This highly mobile loop has been proved to participate in substrate distortion to a 1S3 conformation, therefore forming a productive Michaelis complex along catalysis (17). This has not been observed in other GH3 enzymes, with the substrate being in a relaxed chair conformation, although a Michaelis complex has been recently reported for the Listeria innocua β-glucosidase (18). Among all GH3 β-glycosidases with available structures, insights into the substrate specificity observed in the family has been reported for the H. vulgare Exo1 (7, 19–21) or the β-glucosidases from Thermotoga neapolitana (8) and Kluyveromyces marxianus (9). However, the high varieties in structure and composition found among the different enzymes make it difficult to extrapolate general rules explaining function, and a clear classification of different subfamilies is still needed.</p><p>A proper classification of GH3 glycosidases may require extensive biochemical and structural characterization of new enzymes. In this context, nature provides an inexhaustible reservoir from which enzymes can be isolated (22), because they are continuously changing and evolving as a consequence of natural processes of selection. Genomics and metagenomics have made accessible such an enormous reserve of uncharacterized enzymes. Thus, we and others have recently taken advantage of sequencing and extensive screening technologies to develop enzyme discovery strategies and to identify microbial enzymes with improved and unusual activities and specificities (23–25), as well as distinct active site architectures and substrate preferences relative to other structurally characterized enzymes (26). These elegant studies demonstrated that nature contains proteins with novel and/or altered sequences and protein structures, the analysis of which represents one of the major challenges in postgenomic biology (27).</p><p>Here, activity screening of a metagenomic library created from rumen fluid led us to the isolation of a novel β-glycosidase, GlyA1, which was assigned to the GH3 family. Detailed biochemical characterization of the new enzyme revealed its substrate specificity, whereas its sequence and crystal structure analysis revealed a novel permuted domain topology, defining a new subgroup within the GH3 family. The enzyme contains an additional C-terminal domain, previously unidentified, with its molecular flexibility being explored by small angle x-ray scattering (SAXS) analysis. The structural and biochemical analysis of the GlyA1 hydrolase presented in this study shed new light on comparative catalysis and evolutionary model studies as well as phylogenetic relationships.</p><!><p>A subset of 14,000 clones from resident microbial communities of strained ruminal fluid (SRF) collected from rumen-fistulated, non-lactating Holstein cows (28) was screened for its ability to hydrolyze p-nitrophenyl-β-d-glucoside (pNPβGlc) and p-nitrophenyl-β-d-cellobioside (pNPβCel). We identified a positive clone (designated SRF4) that is highly active against both substrates. The fosmid with insert SRF4 (38,710 bp; G + C 41.89%) was fully sequenced. A gene herein designated as glyA1 encoding a potential GH3 β-glycosidase (GlyA1) was identified out of the 38 distinct genes on the hit fosmid. The deduced molecular mass and estimated pI value were 101,849 Da and 4.86, respectively. This 921-amino acid-long putative protein exhibited a maximum amino acid sequence identity of 59% to a similar protein in public databases (with a top hit EDO57841.1 from Clostridium sp.). A search of oligonucleotide patterns against the GOHTAM database (29) and TBLASTX analysis revealed compositional similarities between the DNA fragment (38,710 bp) containing the gene for GlyA1 with genomic sequences of Eubacterium, Butyrivibrio, and Coprococcus spp. BLASTN revealed similarities of short DNA fragments to Prevotella and Paenibacillus spp. BLASTX (search by translated DNA sequences) showed similarity to glycosidases of unknown Clostridia (phylum Firmicutes). BLASTP search with identified protein sequences showed good matches for many of them against corresponding proteins in Eubacterium and Prevotella and members of Lachnospiraceae, Clostridium, Ruminococcus, and Bacteroides. Most likely, GlyA1 has thus its origin in the phylum Firmicutes, and the presence of a phage gene may, however, indicate a horizontal gene transfer of the carbohydrate metabolism genes from Firmicutes to Bacteroidetes. Those microbes are known to be abundant in the ruminal environment and are thought to play key roles in the breakdown of proteins and carbohydrate polymers (30, 31).</p><!><p>The gene encoding putative GH3 β-glycosidase (GlyA1) was cloned, expressed in E. coli BL21 (DE3), and purified. The hydrolytic activity was analyzed using 18 synthetic model p-nitrophenyl (pNP) derivatives with different sugars as well as a series of 11 additional oligosaccharides. Their specific activities (units/g protein) (Table 1) and the half-saturation (Michaelis) coefficient (Km), the catalytic rate constant (kcat), and the catalytic efficiency (kcat/Km) values (Table 2) were determined. As shown in Table 1, activity was confirmed for 18 substrates that revealed that GlyA1 is a GH3 member with clear β-glucosidase and β-xylosidase activities, but also possessing β-galactosidase, β-fucosidase, α-arabinofuranosidase, and α-arabinopyranosidase activities at low level in this order (Table 1). The activity toward pNP-N-acetyl-β-d-glucosaminide (pNPGlcNAc) and pNP-N-acetyl-β-d-galactosaminide (pNPGalNAc) was below detection limits, and thus the enzyme does not have β-N-acetylglucosaminidase nor β-N-acetylgalactosaminidase activity. As shown in Table 2, in terms of catalytic efficiencies, pNPβCel was the preferred substrate, mainly due to the higher affinity for this substrate as compared with other pNP sugars. The purified recombinant hydrolase was also assayed for their activities toward different polymeric substrates. Using specific activity determination, GlyA1 hydrolyzed all short cello- and xylo-oligosaccharides tested (degree of polymerization (DP) from 2 to 5), with longer substrates being slightly preferred (Table 1). The catalytic efficiencies (kcat/Km) while using the non-activated substrates cellobiose and xylobiose were lower than those found for pNPβCel and pNP-β-xylobiose (pNPβXylb), respectively, mainly due to a significant decrease of kcat values for the natural disaccharides (Table 2). A comparison of kinetic parameters using the natural substrates xylobiose and cellobiose and the synthetic pNPβXylb and pNPβCel substrates confirmed the ∼2-fold higher affinity for oligosaccharides containing β-linked glucosyl versus xylosyl substrates. In contrast, the affinities for the monosaccharides pNPβGlc and pNPβXyl were essentially similar, suggesting that affinity constraints are higher as the size of the oligosaccharides increases. However, due to the differences in kcat values, no major differences in catalytic performance were observed when comparing β-Xyl- and β-Glc-containing sugars. The catalytic performance (kcat/Km) found for other substrates is from low to very low mainly due to lower catalytic rates. The enzyme also exhibited activity against lichenan, suggesting that is able to hydrolyze substrates with mixed β-1,3/1,4 linkages. No activity was detected using avicel or filter paper, as well as toward substrates without β-1,4 linkages such as β-1,3 glucan or mixed β-1,3/1,6 linkages such as laminarin. Accordingly, the enzyme showed a clear preference for short cello-oligosaccharide substrates, which may likely be produced in natural settings from the cellulose components of plant cell walls due to the action of glucanases in the ruminal fluid. Other substrates such as gentibiose (containing d-glucoses joined by a β-1,6-linkage) and sophorose (or 2-O-β-d-glucopyranosyl-α-d-glucose) were also hydrolyzed to a similar extent as cellobiose and xylobiose. The optimum activity for GlyA1 was observed within a mesophilic range (45–65 °C) and within a neutral or slightly acid pH (6.0–7.0), being most active at 55 °C and a pH close to 6.5 (Fig. 1).</p><!><p>Substrate specificity of the purified β-glycosidase GlyA1 and truncated GlyA1-ΔCt</p><p>Kinetic parameters of the purified β-glycosidase GlyA1</p><p>Temperature (A) and pH (B) profiles of the purified β-glucosidase GlyA1. The data represent the relative percentages of specific activity (units/g) compared with the maximum activity using pNPβGlc as substrate (100% in A, 2841 units/g; 100% in B, 3056 units/g). The specific activities were calculated using 0.23 μm protein and 10 mg/ml pNPβGlc as the assay substrate. A, reactions were performed in 50 mm sodium acetate buffer, pH 5.6, at different temperatures. B, reactions were performed at different pH (50 mm BR buffer) and at 40 °C. Standard deviations of the results of assays conducted in triplicate are shown.</p><!><p>A mutant containing a missing C-terminal region, herein referred to as GlyA1-ΔCt, was created in the vector pQE80L. After purification, activity was determined for the 18 sugars being hydrolyzed by the wild-type enzyme, so the effect of the C-terminal region was tested. As shown in Table 1, the specific activity of the mutant was from 2- to 18.4-fold lower than that of the wild type, suggesting the importance of this region in the overall activity of the enzyme. The negative effect of the elimination of the C-terminal domain (compared with the full-length protein) was most notable for the hydrolysis of sugars containing β-glucose (from 11.3- to 17.1-fold activity reduction) as compared with those containing β-xylose (from 4.6- to 5.5-fold lower activity).</p><!><p>Preliminary crystals from the wild-type GlyA1 were obtained after more than 3 months with PEG3350 as the precipitant, and they were cryoprotected into 25% d-glucose to obtain the complex with this sugar. The structure was solved by molecular replacement using the domains from T. neapolitana β-glucosidase as independent search models. Refinement and analysis of electron density maps allowed modeling of the chain containing residues 3–798 but did not show any density to build the C-terminal segment 800–921, suggesting a putative cleavage of this region in the slow crystallization step. The low numbers of crystals impeded analysis of the intact protein by mass spectrometry, but SDS-PAGE analysis of protein solution samples revealed the presence of two bands after incubation at room temperature or treatment with proteases. Therefore, the sample was incubated with subtilisin previously to the crystallization step, which accelerated formation of many good quality crystals, under similar conditions and with the same space group. These crystals were cryoprotected into 20% glycerol, and this molecule was found bound at the active site. Furthermore, crystals from a truncated construct containing residues 2–799 (GlyA1-ΔCt) also grew in a week with ammonium sulfate as the precipitant and, despite having different shape, yielded the same cell and space group, which is consistent with the hypothesis that the wild-type sample was cleaved. These crystals were used to obtain the complexes with d-xylose and d-galactose. Many attempts done to crystallize the complete enzyme were unsuccessful. Also, a construct with residues 800–921, containing the isolated C-terminal region (GlyA1-Ct), failed to crystallize. Crystallographic data and refinement statistics for the four structures here presented are given in Table 3.</p><!><p>Crystallographic data of GlyA1</p><p>Values in parentheses are for the high resolution shell.</p><p>a Rmerge = Σhkl Σi|Ii(hkl) − (I(hkl))|/Σhkl ΣiIi(hkl), where Ii(hkl) is the ith measurement of reflection hkl and (Ii(hkl)) is the weighted mean of all measurements.</p><p>b Rpim = Σhkl (1/(N − 1)) 1/2 Σi|Ii(hkl) − Ii(hkl)|/Σhkl ΣiIi(hkl), where N is the redundancy for the hkl reflection.</p><p>c Rwork/Rfree = Σhkl|Fo − Fc|/Σhkl|Fo|, where Fc is the calculated and Fo is the observed structure factor amplitude of reflection hkl for the working/free (5%) set.</p><!><p>The first solved structure from barley β-d-glucan glucohydrolase (7) showed the core structure common to GH3 enzymes, composed of an N-terminal (α/β)8 barrel domain 1 linked to an (α/β)6-sandwich domain 2 (Fig. 2A); both of them provided residues that make up the active site. The later reported structures from T. neapolitana (8), Trichoderma reesei (12), Aspergillus (13, 14), and L. innocua (18) β-glucosidases, and a β-glucosidase isolated from soil compost (32), showed the presence of an additional fibronectin type III (FnIII) domain (also designated fibronectin-like domain or FLD) located at the C terminus. This three-domain arrangement is shared by other reported β-glucosidases from K. marxianus (9) and Streptomyces venezuelae (11) that also contain an additional PA14 domain inserted within the same loop of their (α/β)6-sandwich, although both are arranged in a different orientation. Moreover, the structure of the Pseudoalteromonas sp. exo-1,3/1,4-β-glucanase has been reported to have a C-terminal domain attached to the core structure, structurally related to family 30 carbohydrate-binding modules (CBM30), although its function is unknown (10). To expand even more this diverse landscape, GlyA1 presents a novel structural arrangement showing permuted sequence and topology, in which the (α/β)6 sandwich (previous domain 2) is located at the N terminus and the FnIII domain is sequentially inserted between this and the (α/β)8 barrel (Fig. 2A). Additionally, a 120-residue segment attached to the C terminus most surely folded into an additional domain.</p><!><p>Permuted domain composition of GlyA1. A, comparison of GlyA1 structure with representative members of multidomain GH3 enzymes. β-glucosidases from K. marxianus, KmβGlu (9) and T. neapolitana, TnβGlu (8), the exo-1,3/1,4-β-glucanase from Pseudoalteromonas sp., PsExoP (10) and the barley β-d-glucan exohydrolase, HvExoI (7) are shown. Domains are named as ABS: (α/β)6-sandwich; FLD fibronectin-like; ABB (α/β)8 barrel; PA14, protective antigen PA14 domain. B, folding of GlyA1. The N-terminal (α/β)6-sandwich domain (red) is followed by the FnIII domain (beige) and the (α/β)8 barrel domain (green). Two long segments connect the three domains (gray). A glucose found in the active site is represented by spheres. C, scheme of the GlyA1 domain organization (left) as compared with that of T. neapolitana β-glucosidase (right) (8). D, superimposition of GlyA1 (gold) onto T. neapolitana β-glucosidase (blue) coordinates. Both enzymes present a deviation from the canonical (α/β)8 barrel topology, with their first α-helix missing, which makes strand β2 reversed and antiparallel with the other seven strands. The main difference between both enzymes is the long arm linking the FnIII to the (α/β)8 domain in GlyA1, which is missing in T. neapolitana β-glucosidase. Also, small differences in the orientation of some helixes are observed.</p><!><p>Fig. 2B displays the 3D structure of the solved 3–798 region of GlyA1, which present overall dimensions of 85 × 65 × 45 Å. The N-terminal (α/β)6-sandwich domain (red, residues 10–219) is followed by the FnIII domain (beige, residues 278–419) and the (α/β)8 barrel domain (green, residues 468–780). Two long segments connect the three domains (Fig. 2B, gray). Linker 1 (residues 220–277) and half of linker 2 (residues 411–443) are tightly wrapped over the core structure, whereas the rest of linker 2 (444–467) forms an extended arm that clasps the (α/β)8 barrel. Finally, the regions at the beginning and the end of the chain are making a two-stranded β-sheet that laces the core structure at the top.</p><p>Comparative analysis using the Dali (33) server revealed that the GlyA1 (α/β)6-sandwich domain, containing the catalytic acid/base residue Glu-143, superimposes onto the corresponding domain from the T. neapolitana β-glucosidase, with a root-mean-square deviation (r.m.s.d.) of 1.6 Å for 202 eq Cα positions (39% sequence identity). The same comparison with the other structurally known GH3 gives deviations in the range 2–2.5 Å (20–25% sequence identity). The FnIII domain seems more structurally conserved along the GH3 family; GlyA1 is most similar to those in the β-glucosidases from T. neapolitana, with r.m.s.d. = 1.5 Å (122 residues, 39% identity), and K. marxianus, with r.m.s.d. = 1.6 Å (123 residues, 32% identity), but the same analysis gives values in the range 1.8–1.9 Å (21–28% sequence identity) against the other GH3 enzymes containing this domain. Finally, the (α/β)8 barrel, which contains the nucleophile Asp-709, is most similar to the corresponding domain in the β-glucosidases from T. neapolitana (r.m.s.d. = 1.6 Å, 285 residues, 39% identity), K. marxianus, (r.m.s.d. = 1.5 Å, 276 residues, 34% identity), S. venezuelae (r.m.s.d. = 1.7 Å, 278 residues, 32% identity), and T. reesei (r.m.s.d. = 2.0 Å, 278 residues, 27% identity). Equally to GlyA1, all these domains present a deviation from the canonical (α/β)8 barrel topology, which was first observed in the T. neapolitana β-glucosidase. Thus, their first α-helix of the eight β-α motifs is missing, which has the consequence of making strand β2 reversed and antiparallel with the other seven strands. The different deviation from the canonical topology found at this domain is consistent with the higher deviations found in the structural comparison of GlyA1 with other GH3 enzymes, in the range 2.5–3 Å (16–20% identity).</p><p>Interestingly, the GlyA1 core is structurally rather conserved with known β-glucosidases with equivalent domain architecture (Fig. 2C). The superposition of the T. neapolitana β-glucosidase onto the structure of GlyA1 reported here shows small differences in the orientation of some of the helices (Fig. 2D). The main difference is the long arm that links the FnIII to the (α/β)8 domain in GlyA1, which is missing in T. neapolitana β-glucosidase. There are also significant differences in the loops surrounding the active site both in length and orientation, which must be related to the different substrate specificity, as commented below.</p><!><p>The active site of GlyA1 is located at the molecular surface, at the interface between the (α/β)8 barrel domain, which provides the nucleophile Asp-709 and the (α/β)6-sandwich domain, contributing to the Glu-143 acid/base catalyst (Fig. 3A). The participation of Asp-709 in substrate hydrolysis was confirmed by site-directed mutagenesis (D709A) in GlyA1 and GlyA1-ΔCt, as Km and kcat values could not be determined from the data obtained due to the activity value being below the detection limit. It is a pocket of 12 Å deep with a narrow entrance 4–6 Å wide. A detailed structural comparison with the T. neapolitana β-glucosidase (Fig. 3A) reveals the main differences in loop conformation observed around the active site that are responsible for making a deeper catalytic pocket in GlyA1. First, loop β7-α7 of the (α/β)8 barrel, following the nucleophile Asp-709 (residues 711–726), has an 11-residue insertion that extends away from the pocket and interacts with the long segment linking the FnIII domain to the barrel, which is missing in the T. neapolitana β-glucosidase. Here, Arg-717 makes an ion pair with Glu-447 at the small helix located in the middle of the extended linker, which helps in stabilizing this region. An important feature of this β7-α7 loop is the presence of Trp-711, close to the nucleophile Asp-709, that protrudes from the surface and delineates a narrow catalytic pocket. Moreover, and despite loop β3-α3 (residues 536–550) being shorter in GlyA1, Arg-538 clearly bulges into the pocket contributing to constrict it even more.</p><!><p>GlyA1 active site architecture. A, detail of the loops surrounding its active site from the (α/β)8 barrel (green) and the (α/β)6-sandwich (raspberry) domains, superimposed onto the T. neapolitana β-glucosidase (8) (pale blue). Three glycerol molecules from the cryobuffer found in the GlyA1 crystals are shown in orange. Asp-709 and Glu-143 are the nucleophile and the acid/base catalyst, respectively. Main features of GlyA1 are the extended loop containing Asp-709, which includes Trp-711 and the ion pair Arg-717–Glu-447 fixing it to the unique long arm and a highly flexible loop containing Trp-106. Two different conformations found among the crystals at Trp-111 and Phe-147 are highlighted. B, detail of the atomic interactions defining subsite −1. A glucose molecule is shown in gold. Xylose binds in the same relaxed chair conformation, and only interaction of the glucose O6 hydroxyl is missing. Inset, binding mode of galactose in a semi-chair conformation by flattening of the C4 atom that has the axial hydroxyl substituent and keeping the same interaction pattern. C, thiocellobiose (cyan) and thiogentibiose (pink) modeled at the active site by structural superimposition to the previously determined β-d-glucan glucohydrolase barley complexes (PDB entries 1IEX and 3WLP (34)), delineating putative subsite +1. D, molecular surface of the GlyA1 active site, with relevant residues as sticks. Three different β-1,4/β-1,3-linked tetraglucosides have been manually docked by superposition of their non-reduced end to the experimental glucose: a cellotetraose, as found in PDB entry 2Z1S (green); a Glc-4Glc-3Glc-4Glc (purple), and a Glc-4Glc-4Glc-3Glc (yellow), as built by the on-line carbohydrate-building program GLYCAM (45) and exported in its minimum energy state. E, superposition of GlyA1-Glc structure (beige) with those reported for T. reesei β-glucosidase (purple) (12) and barley β-d-glucan glucohydrolase complexed with thiocellobiose (cyan) (34).</p><!><p>With respect to the (α/β)6-sandwich, similarly to that observed in T. neapolitana β-glucosidase, this domain is shaping the active site by means of two loops, residues 139–152 containing the acid/base catalyst Glu-143 and residues 100–113 enclosing Trp-106 that clearly projects into the catalytic pocket. Interestingly, the last loop is markedly flexible as it is deduced from the fact that it could only be fully traced in the ligand-free crystal, containing only glycerol in the active site, and in the galactose-soaked crystal of the truncated form. In contrast, the crystals of the full-length and truncated forms, soaked into glucose and xylose, respectively, showed poor density that precluded tracing residues 104–107. Furthermore, the traced loops showed significant conformational changes in the different crystals at Trp-111, coupled to a change in Phe-147 from the adjacent 139–152 loop (Fig. 3A), reinforcing its intrinsic mobility. The loop equivalent to 100–113, which is highly variable within GH3 enzymes, was proposed to be involved in recognition of large substrates from the crystal structure of T. neapolitana β-glucosidase, which showed some disorder that precluded tracing of a segment equivalent to that not observed in some GlyA1 crystals. Noteworthy, the non-visible region of T. neapolitana β-glucosidase includes Trp-420 that, consequently, may be defining additional binding subsites, similarly to Trp-106. However, the remaining sequence is not conserved, with both Phe-147 and Trp-111 being unique to GlyA1, and therefore, the substrate recognition mode presented by the two enzymes to accommodate the substrate may be different.</p><p>Soaking with xylose and glucose showed a clear density indicating that both sugars occupy the catalytic pocket subsite −1 in a relaxed chair conformation (Fig. 3B). This subsite is well conserved among known GH3 β-glucosidases and, with the exception of the acid base catalyst, is made up entirely by residues from the (α/β)8 barrel domain. Thus, residues from the loops emerging from the central β-strands are making a tight net of hydrogen bonds that accommodate the glycon with all its OH groups making at least two polar interactions. The glycon moiety is located by stacking to Trp-710, and the acid base catalyst Glu-143 and the nucleophile Asp-709 interact with the O1 and O2 hydroxyls, as is expected in GH enzymes. The other residues making subsite −1 are Asp-532, Arg-597, Lys-630, His-631, Arg-641, and Tyr-677. Xylose and glucose are bound in an identical position, and the glycerol molecules observed in the ligand-free crystals are mimicking the positions occupied by C2, C3, C4, and C5 from both sugars. The additional polar interaction made by the glucose O6 hydroxyl appears consistent with the higher affinity observed in GlyA1 toward glucosides as compared with xylosides. Thus, as shown in Table 2, the affinity for cellobiose (Km = 2.4 ± 0.3 mm) was ∼2-fold higher than that for xylobiose (Km = 4.7 ± 0.2 mm). Interestingly, soaking of crystals with galactose showed that this sugar displays a semi-chair conformation at subsite −1 by flattening of the C4 atom that has the axial hydroxyl substituent (Fig. 3B, inset). In this way, galactose is accommodated by essentially the same polar interactions observed in the glucose complex, thereby explaining the activity of the enzyme on β-galactosides. However, the energy cost of getting the substrate ring distortion is reflected by the lower β-galactosidase activity, as given in Tables 1 and 2. Accordingly, the low β-fucosidase and α-arabinosidase activities must reflect some degree of deviation from the glucose-binding pattern, through ring distortion and/or loss of polar interactions, but in any case the plasticity of the catalytic site provides a notable capacity of GlyA1 to accept different sugars (from high to low and very low specificity).</p><p>As said before, and in contrast to that observed in T. neapolitana β-glucosidase that presents an active site opened to the solvent with only subsite −1 being defined, more subsites are apparent in GlyA1. To delineate a putative +1 subsite, we modeled the position of the non-hydrolysable substrate analogs thiocellobiose and thiogentibiose by structural superimposition on the previously reported experimental barley complexes (34). As shown in Fig. 3C, Trp-106 and Trp-711 define a hydrophobic patch that may allocate the oligosaccharides at a putative subsite +1, leaving a range of possible ring orientations compatible with the observed activity of GlyA1 against differently β-linked bioses, as given in Table 1. Also, the long chain of Arg-538, protruding at the catalytic pocket as said above, is in good position to stabilize the sugar unit by making hydrogen bonds to one or possibly two of its hydroxyl groups. The important contribution of subsite +1 to GlyA1 substrate binding efficiency (both glucosides and xylosides) is manifested by the lower Km value with pNPβCel compared with pNPβGlc and by the lower Km value with pNPβXylb compared with pNPβXyl (Table 2).</p><p>Furthermore, inspection of the molecular surface of the active site cavity shown in Fig. 3D suggests the possible existence of additional subsites, which is illustrated by several β-1,4/1,3-linked oligosaccharides that have been modeled at the active site as follows: a glucotetraose (green), a Glc-4Glc-3Glc-4Glc chain (purple), and a Glc-4Glc-4Glc-3Glc (yellow). These sugars have been docked by superimposition of their non-reducing units onto the observed glucose at the GlyA1 complex. The hydrophobic patch defined by Trp-106 and Trp-711 may fit the oligosaccharides at subsites +1 and +2, and the long side chain of Lys-723 seems available to make polar interactions with the hydroxyl groups defining a possible subsite +3. The putative existence of at least three subsites in the GlyA1 active site would be in agreement with the tendency of an increased activity against longer cello- and xylo-oligosaccharides (see Table 1). Also, the tendency of increased activity against longer cello- and xylo-oligosaccharides as given in Table 1 suggests interactions at more distal positions and therefore the possibility of additional subsites. Moreover, the shape of the active site seems compatible with the mixed β-1,4/1,3-links of the modeled tetrasaccharides, thereby explaining the observed activity on the medium size polymer lichenan.</p><p>Comparison of the GlyA1-Glc structure with those reported for T. reesei β-glucosidase (12) and barley β-d-glucan glucohydrolase complexed with thiocellobiose (Fig. 3E) (7) displays the different hydrophobic platforms found at each active site. The barley β-d-glucan glucohydrolase structure showed a narrow channel with the glucose tightly arranged at subsite +1, being sandwiched between Trp-286 and Trp-434 side chains. In contrast, the GlyA1 Trp-711 is perpendicular and oriented similarly to Trp-37 found in T. reesei β-glucosidase, although both residues are provided by different loops from the (α/β)8 barrel domain. At the opposite face, GlyA1 Trp-106 is structurally equivalent to Tyr-443 and Trp-434 from the barley and T. reesei enzymes, although all of them come from different loops within the (α/β)6-sandwich domain. Interestingly, other enzymes present an aromatic residue in a position identical to Trp-106, but they are provided by the PA14 domain, Phe-508 in the case of the K. marxianus β-glucosidase, or by a long loop coming from the other subunit, Tyr-583 in the case of the L. innocua β-glucosidase dimer (data not shown) (18). This feature illustrates that these highly diverse enzymes have evolved common topology and molecular mechanisms, and yet the precise structural differences behind that regulate specificity.</p><!><p>Because of the unfeasibility in crystallizing the full-length GlyA1, we explored its overall flexibility and putative shape in solution by SAXS experiments. Thus, we compared the molecular descriptors of the complete construct with respect to the truncated construct GlyA1-ΔCt, lacking the C-terminal domain. For this purpose, several solutions with varying concentrations were measured for each sample, and their scattering curves were merged to extrapolate idealized data. Analysis of the scattering curves shows a good fit to the Guinier approximation, which indicates that the samples are not aggregated. Also, the calculated radii of gyration (Rg) are consistent across the range of measured concentrations. Then, the overall size descriptors can be properly determined for each construct.</p><p>First of all, the calculated molecular masses from both samples are close to the expected values (Table 4), indicating the presence of monomers, and also a 15-kDa higher mass in the complete protein, which excludes proteolysis of the analyzed sample in the short time of the experiment. Furthermore, the Rg and the maximum distance (Dmax) for the complete protein are only slightly higher than the truncated protein, which may be indicate that the extra C-terminal domain is not too extended from the core structure. In support of this hypothesis, the pairwise distance distribution function P(r) calculated for both constructs shows a similar unimodal pattern consistent with a single domain protein in both cases. Furthermore, the analysis of the scattering function by the Kratki plots is consistent with the expected profile for a folded protein with a clear peak, in contrast what is observed in multidomain proteins with flexible linkers that present several peaks or smoother profiles. Consequently, we do not observe in the data calculated from the complete protein any of the signs that may be indicative of molecular flexibility, i.e. large Rg and Dmax, absence of correlation in the P(r) function, or smooth Kratky plots. Therefore, SAXS analysis appears consistent with a compact overall shape of the complete GlyA1, in which the extra C-terminal region would not define a marked separate or flexible domain but rather it could be folded over the core three-domain structure.</p><!><p>SAXS data collection and derived parameters</p><!><p>To test the feasibility of this hypothesis, ab initio models were generated for complete GlyA1 from SAXS data. First, two models of the last 120 residues (GlyA1-Ct) were obtained, as explained under "Experimental Procedures," with both showing an overall β-sandwich topology. This topology is related to carbohydrate-binding domains within families CBM6 and CBM35, to which GlyA1-Ct presents 15–20% sequence identity, although the equivalent carbohydrate-binding motifs, typically clusters of conserved aromatic residues, are not evident in its surface. Then, three runs of CORAL were computed by considering the experimental structure of the truncated protein and each of the two models. The six models obtained are shown in Fig. 4. Analysis of these models reveals that all of them cluster around a reduced area that would locate the C-terminal region relatively distant from the catalytic pocket but quite near the mobile loop (residues 100–113). Overall, these models are consistent with the hypothesis proposed above, suggesting that GlyA1-Ct may be somewhat packed between the two domains making the core structure and, interestingly, with a putative linker somehow exposed to solvent. This feature might possibly explain the proteolysis observed in the complete protein.</p><!><p>SAXS analysis of GlyA1. Six ab initio models were generated for complete GlyA1 from SAXS data, using the experimental structure of the truncated protein and two different models of the last 120 residues (GlyA1-Ct). The two templates were obtained from Swiss-Model (red) (48) or CPHmodel (blue) (49) servers, which predict different lengths of the linker attaching this domain to the core protein, 32 or 5 residues, respectively. CORAL (47) modeling of this linker in each run is represented in spheres. The active site pocket is indicated by the galactose found at the crystal (yellow), and the mobile loop (residues 100–113), as observed in the galactose-soaked crystals, is highlighted in green.</p><!><p>Our structural analysis illustrated that the permuted domain architecture of GlyA1 keeps the location of the active site at the interface between the (α/β)8 barrel and the (α/β)6-sandwich domains. As mentioned above, N-acetylglucosaminidases are built by a single domain, with its (α/β)8 barrel holding both the nucleophile and acid/base catalyst. Interestingly, the Bacillus subtilis NagZ shows the two-domain composition but still keeps the catalytic residues at the (α/β)8 barrel (16). Therefore, this domain may be considered as the characteristic signature of GH3 enzymes. To examine the phylogenetic positioning of β-glucosidases with inverted topology (represented by GlyA1) within the GH3 family, we have carried out a phylogenetic analysis based on the sequence of its (α/β)8 barrel domain (ABB in this analysis). Sequences representative for each of the domain architectures found in the GH3 domain were selected (details under "Experimental Procedures"). The five topologies selected for this study are ABB, ABB-ABS, ABB-ABS-FLD, ABS-FLD-ABB, and ABB-ABS(PA14)-FLD (ABS (α/β)6-sandwich; FLD is fibronectin-like type III domain). The resulting phylogenetic tree given in Fig. 5 shows apparent correlation between ABB sequence divergence and domain architecture. Most single domain sequences (ABB) cluster together and correspond to N-acetylglucosaminidases (Fig. 5, salmon area of the tree). Insertion of the ABS module is associated with three different nodes (a, b, and c in Fig. 5). Insertion at node a was not accompanied by a significant divergence in the ABB sequence because both ABB and ABB-ABS architectures appear mixed at this node. In fact, these ABB-ABS sequences also correspond to N-acetylglucosaminidases, and crystallographic data of B. subtilis NagZ show that the two modules are quite independent from a structural point of view. ABS insertion at nodes b and c would correspond to the divergence of GH3 enzymes giving rise to other activities, mainly β-glucosidase. Within node c, other modules (FLD and PA14) were appended after ABS. At node b, fusion of C-terminal FLD seems to occur close to ABS addition because most sequences contain both modules. GlyA1 and the other GH3 enzymes with inverted topology arose within this cluster. The phylogenetic analysis shows that the inverted topology is predominantly found in Firmicutes, although it is also present in at least another phylum (Actinobacteria) and even Archaea. Furthermore, it appears clearly associated to enzymes belonging to bacteria dwelling in the digestive tract of animals.</p><!><p>GlyA1 phylogenetic analysis. The unrooted circular Neighbor-Joining tree indicating phylogenetic positions of polypeptide sequences of the GlyA1 enzyme characterized in present work (red boldface) and reference similar enzymes. GenBankTM or PDB (in boldface) accession numbers are indicated. The domain architecture (ABB, ABB_ABS, ABB_ABS_FLD, ABB-ABS(PA14)-FLD, and ABS_FLD_ABB) to which each sequence is associated is specifically indicated. Multiple protein alignment was performed using ClustalW program, built into software version 2.1. Phylogenetic analysis was conducted with the Ape package implemented for R programming language. Sequences resembling NagZ (β-N-acetyl-glucosaminidase) are highlighted with pink background. Those encoding GH3 β-glucosidases are indicated in brown; within them, those with GlyA1-like permuted domain topology are indicated in gray. ABB, (α/β)8 barrel; ABS (α/β)6-sandwich; FLD, fibronectin-like type III domain; PA14, protective antigen PA14 domain.</p><!><p>In this work, a functional metagenome library analysis was used to identify a β-glycosidase from a plant polymer-degrading microorganism populating the rumen of a dairy cow. The enzyme most likely originated from the genome of a representative of Firmicutes phylum known to be abundant in the ruminal environment (30, 31).</p><p>The structural and biochemical analysis of the GlyA1 hydrolase presented in this study sheds new light on the mechanisms of the catalysis and evolutionary patterns of the GH3 family. Our data demonstrated that GlyA1 has a permuted domain topology. It is well documented that the formation of new domain combinations is an important mechanism in protein evolution. The major molecular mechanism that leads to multidomain proteins and novel combinations is non-homologous recombination, sometimes referred to as "domain shuffling." This may cause recombination of domains to form different domain architectures. Proteins with the same series of domains or domain architecture are related by descent (i.e. evolved from one common ancestor) and tend to have the same function (35), which is rarely the case if domain order is switched. Indeed, a detailed analysis of the structures of proteins containing Rossmann fold domains demonstrated that the N- to C-terminal order of the domains is conserved because the proteins have descended from a common ancestor. For pairs of proteins in the PDB in which the order is reversed, the interface and functional relationships of the domains are altered (36). This was also proved in this study, which revealed that the altered domain architecture in GH3 mostly evolved from a distinct ecological niche, most likely from digestive tracts, including that of the ruminants. Also, the substrate specificity of the GlyA1 protein is markedly different from that of reported GH3 members. Indeed, GlyA1 is a uncommon multifunctional GH3 with β-glucosidase, β-xylosidase, β-galactosidase, β-fucosidase, α-arabinofuranosidase, α-arabinopyranosidase, and lichenase co-activities, with the ability to degrade β-1,2-, β-1,3-, β-1,4-, and β-1,6-glucobioses.</p><p>From an ecological point of view, the rumen compartment provides stable and favorable conditions for microbial growth and is also permanently exposed to plant biomass; for this reason, it contains specialized microorganisms that are permanently competing or collaborating for the degradation of the plant fibers. The data herein suggest that this factor, namely the high exposure to plant biomass, which is less common in other habitats, may be a strong force driving the establishment of gut microbiota with GH3 protein with permuted structures that may provide ecological advantages. Indeed, the permuted domain topology may confer the protein different functionalities such as the ability to expand the pool of biomass-like substrates being hydrolyzed. Overall, our results (analysis of oligonucleotide pattern and phylogenetic tree) strongly suggest that GlyA1 and related GH3 enzymes with inverted topology emerged in Firmicutes, where their presence is rather frequent, and are transferred by horizontal gene transfer to bacteria from other phyla and even to another kingdom (Archaea). It is well documented that these wide ranging gene transfer events take place at high frequency in the rumen (37, 38). Probably, GlyA1 topology arose from a sequence encoding a GH3 enzyme with ABB-ABS-FLD domain architecture by gene inversion. Although the inversion surely rendered a nonfunctional gene, further mutations that would restore some sort of glycolytic activity would be strongly favored by selective pressure.</p><p>Structural analysis illustrates the permuted domain composition of GlyA1 that is composed of an N-terminal (α/β)6-sandwich domain, followed by the FnIII domain, and the (α/β)8 barrel domain. Based on sequence data, a C-terminal domain was expected after the (α/β)8 barrel domain. However, attempts to crystallize the C-terminal region of the protein were unsuccessful, and its functional role was unclear. Biochemical characterization of the GlyA1 and GlyA1-ΔCt proteins revealed that the C-terminal domain does not affect the overall substrate profile of the protein, but rather it affects the catalytic performance, which is significantly lower in the truncated GlyA1-ΔCt protein. This suggests that most likely the C-terminal domain may not have a direct role in substrate binding, but still it might disturb the dynamics of the proximate mobile loop (residues 100–113), which seems directly involved in catalysis.</p><p>According to available structure-prediction tools, this C-terminal region is expected to adopt a lectin-like topology, related to the CBM6/CBM35 domains. However, it does not seem an obvious carbohydrate-binding domain, and in fact, binding to xylan, cellulose, and barley glucan was not observed by affinity gel electrophoresis assays (data not shown). Nevertheless, although its involvement in binding small substrates does not seem apparent, this domain might be playing a role in positioning or locating the enzyme to distal positions of a yet unknown polymeric substrate by recognizing specific but still unidentified substitutions. Alternatively, it could play a role in keeping the enzyme attached to the cell surface, facilitating the intake of its products and conferring the bacteria an advantage over competing organisms. Interestingly, the analysis of the GlyA1-Ct homologous sequences shows that these domains are attached to GH3 β-glucosidases from a ruminal environment, and this feature points to a possible function related to this ecosystem. However, its presence is not related to the permuted domain topology, as only half of the sequences included in the GlyA1 cluster (Fig. 5) contain segments equivalent to GlyA1-Ct.</p><p>In conclusion, the analysis of GlyA1 here presented uncovers new features of GH3 enzymes and provides a template for a novel subfamily, including members with permuted domain topology. It also allows picturing the GlyA1 active site architecture and the molecular basis of its substrate specificity. More work is needed to have a complete picture of the intricate molecular mechanisms that these highly diverse enzymes have evolved to tailor specificity. It will contribute to improve our knowledge about enzymatic carbohydrate degradation and open up new avenues for biocatalysis.</p><!><p>Chemicals and biochemicals were purchased from Sigma and Megazyme (Bray, Ireland) and were of pro-analysis (p.a.) quality. The oligonucleotides used for DNA amplification were synthesized by Sigma Genosys Ltd. (Pampisford, Cambs, UK). The E. coli Rosetta2 (Novagen, Darmstadt, Germany) for cloning and expression of wild-type protein and the genetic constructs in pQE80L vector were cultured and maintained according to the recommendations of the suppliers.</p><!><p>A pCC1FOS fosmid metagenomic library created from microbial communities from SRF of rumen-fistulated non-lactating Holstein cows was used. The construction and characteristics of the library were described previously (28). A subset of 14,000 clones were plated onto large (22.5 × 22.5 cm) Petri plates with Luria Bertani (LB) agar containing chloramphenicol (12.5 μg/ml) and an arabinose-containing induction solution (Epicenter Biotechnologies) at a concentration (0.01% w/v) recommended by the supplier to induce a high fosmid copy number. After overnight incubation at 37 °C, the clones were screened for the ability to hydrolyze pNPβGlc and pNPβCel. For screens, the plates (22.5 × 22.5 cm; each containing 2,304 clones) were covered with an agar-buffered substrate solution (40 ml of 50 mm sodium acetate buffer, pH 5.6, 0.4% w/v agar and 5 mg/ml of pNPβGlc and pNPβCel as substrates). Positive clones were detected by the formation of a yellow color. One positive clone, herein designated as SRF4, was selected, and its DNA insert was fully sequenced with a Roche 454 GS FLX Ti sequencer (454 Life Sciences, Branford, CT) at Life Sequencing S.L. (Valencia, Spain), and the predicted genes were identified as described previously (28).</p><!><p>The full coding sequence of GlyA1 (residues 2–921) and a deleted version (residues 2–799) lacking the C-terminal domain (GlyA1-ΔCt) were amplified by PCR with 4GF (CACGAGCTCAATATTGAAAAAGTGATACTTGATTGG) as forward oligonucleotide and 4GR1 (AGCCGTCGACTTACTGCTGCTTTTTAAACTCTATTCG) or 4GR2 (AGCCGTCGACTTACACTCTTCCTGCTATCTCAACC) as reverse oligonucleotides, respectively. The SRF4 fosmid was used as the template. The PCR conditions were as follows: 95 °C for 120 s, followed by 30 cycles of 95 °C for 30 s, 55 °C for 45 s, and 72 °C for 120 s, with a final annealing at 72 °C for 500 s. The PCR products were analyzed and agarose gel-purified using the Mini Elute gel purification kit (Qiagen, Hilden, Germany). The PCR products were digested with SacI/SalI and cloned in vector pQE80L to generate plasmids GlyA1-pQE and GlyA1ΔCt-pQE, respectively. The coding sequence of the C-terminal domain (GlyA1-Ct, residues 800–921) was amplified with oligonucleotides CT1F (CACGAGCTCATAGAAGAGGATGCATTCGATATAG) and 4GR1 and cloned in the SacI/SalI sites of pQE80L (plasmids Ct-pQE). GlyA1-pQE was used as a template to introduce the mutation D709A by PCR with primers M1 (TGGTGGGCTCAGGTTAATGACC) and M2 (GGCAGTCATCACAATACCCTTAAAGCC), as described previously (39). The coding region of the resulting plasmids was fully sequenced to check for the absence of undesired mutation. The E. coli strain Rosetta2 (Novagen, Darmstadt, Germany) was transformed with the selected plasmids; the clones were selected on LB agar supplemented with ampicillin (100 μg/ml) and chloramphenicol (68 μg/ml) and stored with 20% (v/v) glycerol at −80 °C.</p><!><p>Mutation D709A was introduced into the corresponding pQE80L plasmids containing genes encoding GlyA1 and GlyA1-ΔCt, using the QuikChange II XL mutagenesis kit from Agilent Technologies, Inc. (Santa Clara, CA), with TGGTGGGCTCAGGTTAATGACC and GGCAGTCATCACAATACCCTTAAAGCC as forward and reverse oligonucleotides, respectively. The resulting variant plasmids were transferred into E. coli strain Rosetta2 (Novagen, Darmstadt, Germany) and selected on the LB agar supplemented with the same antibiotics as parental plasmids.</p><!><p>For enzyme expression and purification of wild-type and mutant GlyA1 and GlyA1-ΔCt variants, as well as GlyA1-Ct in the pQE80L vector, a single colony (E. coli Rosetta2) was grown overnight at 37 °C with shaking at 200 rpm in 100 ml of 2× TY medium (1% yeast extract, 1.5% tryptone, 0.5% NaCl) containing ampicillin (100 μg/ml) and chloramphenicol (68 μg/ml), in a 1-liter flask. Afterward, 25 ml of this culture was used to inoculate 1 liter of 2× TY medium, which was then incubated to an A600 nm of ∼0.6 (range from 0.55 to 0.75) at 37 °C. Protein expression was induced by 0.9 mm isopropyl β-d-galactopyranoside followed by incubation for 16 h at 16 °C. The cells were harvested by centrifugation at 5000 × g for 15 min to yield 2–3 g/liter pellet (wet weight). The cell pellet was frozen at −80 °C overnight, thawed, and resuspended in 3 ml of 20 mm phosphate buffer, pH 7.4, 500 mm NaCl/g of wet cells. Lysonase bioprocessing reagent (Novagen, Darmstadt, Germany) was then added (4 μl/g wet cells) and incubated for 30 min on ice with rotated mixing. The cell suspension was then sonicated for a total of 1.2 min and centrifuged at 15,000 × g for 15 min at 4 °C; the supernatant was retained. The His6-tagged enzyme was purified at 4 °C after binding to a nickel-nitrilotriacetic acid His·Bind resin (Novagen, Darmstadt, Germany). The columns were pre-washed with 20 mm phosphate buffer, pH 7.4, 500 mm NaCl, and 50 mm imidazole, and the enzyme was eluted with the same buffer but containing 500 mm imidazole. The monitoring of the enzyme elution was performed by SDS-PAGE and/or activity measurements, using standard assays (see below). After elution, protein solution was extensively dialyzed with 20 mm Tris, pH 7.5, 50 mm NaCl by ultrafiltration through low adsorption hydrophilic 10,000 nominal molecular weight limit cutoff membranes (regenerated cellulose, Amicon), after which the protein was maintained at a concentration of 10 mg/ml; the protein stock solution was stored at −20 °C until used in assays. The purity was assessed as >95% using SDS-PAGE, which was performed with 12% (v/v) polyacrylamide gels, using a Bio-Rad Mini Protean system. Prior to crystallization assays, 2 mm dithiothreitol (DTT) was added.</p><!><p>Specific activity (units/g) and kinetic parameters (Km and kcat) were first determined using pNP sugars (read at 405 nm) in 96-well plates, as described previously (28). pNP substrates tested included those containing α-glucose (pNPαGlc), α-maltose (pNPαMal), β-glucose (pNPβGlc), β-cellobiose (pNPβCel), α-arabinofuranose (pNPαAraf), β-arabinopyranose (pNPβArap), α-xylose (pNPαXyl), β-xylose (pNPβXyl), β-xylobiose (pNPβXylb), α-fucose (pNPαFuc), α-rhamnose (pNPαRha), α-mannose (pNPαMan), β-mannose (pNPβMan), α-galactose (pNPαGal), β-galactose (pNPβGal), β-lactose (pNPβLac), N-acetyl-β-d-glucosaminide (pNPGlcNAc), and N-acetyl-β-d-galactosaminide (pNPGalNAc). For cello-oligosaccharides (DP from 2 to 5), gentibiose and sophorose, the level of released glucose was determined using a glucose oxidase kit (Sigma). The level of released xylose from xylo-oligosaccharides (DP from 2 to 5) was determined using the d-xylose assay kit from Megazyme (Bray, Ireland). Substrate specificity was investigated also using carboxymethylcellulose, lichenan, barley glucan, laminarin, and avicel (all from Sigma and filter paper (Whatman, UK). Specific activity for all these sugars was quantified by measuring release of reducing sugars according to Miller (50). For Km determinations, assay reactions were conducted by adding a protein concentration of 0.23 μm to an assay mixture containing from 0 to 30 mm sugar in 50 mm sodium acetate buffer, pH 5.6, T = 40 °C. Total reaction volume was 200 μl. For kcat determinations, under the same conditions, sugar concentration was set up to 2 times the Km value, and the protein concentration was from 0 to 0.23 μm. For specific activity determinations (units/g), a protein concentration of 0.23 μm and 10 mg/ml of the sugar or polysaccharide were used in 50 mm sodium acetate buffer, pH 5.6, T = 40 °C. The pH and temperature optima were determined in the range of pH 4.0–8.5 (50 mm Britton-Robinson buffer, BR) and 20–65 °C in assays containing a protein concentration of 0.23 μm and 10 mg/ml pNPβGlc, which was used as standard substrate. BR buffer is a "universal" pH buffer used for the range pH 2–12. It consists of a mixture of 0.04 m H3BO3, 0.04 m H3PO4, and 0.04 m CH3COOH that has been titrated to the desired pH with 0.2 m NaOH. Optimal pH was measured at 40 °C, and the optimal temperature was determined in the same buffer used in the kinetic assays. In all cases, absorbance was determined immediately after reagent and enzyme were mixed using a microplate reader every 1 min for a total time of 15 min (Synergy HT Multi-Mode Microplate Reader, BioTek). All reactions were performed in triplicate. One unit of enzyme activity was defined as the amount of enzyme required to transform 1 μmol of substrate in 1 min under the assay conditions, with extinction coefficients as in Ref. 21. All values were corrected for non-enzymatic hydrolysis (background rate). The protein concentration was determined spectrophotometrically (at 280 nm) using a BioTek EON microplate reader (Synergy HT Multi-Mode Microplate Reader, BioTek) according to extinction coefficient of the protein (108,485 m−1 cm−1) corresponding to its amino acid sequence.</p><p>Note that the detection limit, using a microplate reader with a filter for 405 nm, for the yellow chromogen is about 1·10−6 mol/liter p-nitrophenol. Because the concentration of substrate in the assay ranges from 0 to 30 mm, it is expected that detection of the activity under our assay conditions is much above the detection limit.</p><!><p>Initial crystallization conditions for the complete GlyA1 (10 mg/ml) were explored by high-throughput techniques with a NanoDrop robot (Innovadyne Technologies Inc.), using different commercial screens as follows: PACT and JCSG+ Suites from Qiagen; JBScreen Classic 1–4 from Jena Bioscience; and Index, Crystal Screen, and SaltRx packages from Hampton Research. These assays were carried out using the sitting drop vapor-diffusion method in MRC 96-well crystallization plates (Molecular Dimensions).</p><p>Elongated bars grew after 3 months in 20% polyethyleneglycol (PEG) 3350, 0.2 m ammonium sulfate, BisTris, pH 5.5. For data collection, crystals were cryoprotected in mother liquor supplemented with 25% (w/v) d-glucose before being cooled in liquid nitrogen. Diffraction data were collected at the German Electron Synchrotron (Hamburg, Germany). Diffraction images were processed with XDS (40) and scaled using Aimless from the CCP4 package (41) leading to space group P212121. The structure was solved by molecular replacement using MOLREP (42) with reflections up to 2.5 Å resolution range and a Patterson radius of 54 Å. The template model was the β-glucosidase from T. neapolitana (PDB code 2X42), but the search was made in two steps. First, the region containing residues 2–315 was used for finding a partial solution. Then, another round of molecular replacement, with the region 321–721, was computed. Preliminary rigid-body refinement was carried out using REFMAC (43). Subsequently, several rounds of extensive model building with COOT (44) combined with automatic restraint refinement with flat bulk solvent correction and using maximum likelihood target features led to a model covering residues 3–798. However, no density was found for the loop 103–108 or for the last 123 residues of the protein. At the latter stages, β-glucose, sulfate ions, and water molecules were included in the model, which, combined with more rounds of restrained refinement, led to a final R-factor of 15.7 (Rfree 17.8). The free R-factor was calculated using a subset of 5% randomly selected structure-factor amplitudes that were excluded from automated refinement. Many attempts to reproduce and improve these crystals were unsuccessful, until in situ proteolysis of the sample with subtilisin was tried. Resulting crystals grew after 15 days in the same conditions, but at pH 7.0, they were cryoprotected in 20% (v/v) glycerol and showed the same space group and cell content. Then, the truncated GlyA1-ΔCt construct (residues 1–798) was tested. Initial crystallization assays were accomplished as described above, and several hits were obtained. Best crystals were grown in 2.0 m ammonium sulfate, 0.1 m BisTris, pH 5.5, and belonged to the same space group. The asymmetric unit contains a single molecule, with a Matthews's coefficient of 2.73 and a 54% solvent content within the cell.</p><p>Soaking experiments with d-xylose or d-galactose were performed with the truncated construct in mother liquor solution implemented with 5–50 mm ligand. Then, the crystals were flash-frozen into liquid nitrogen using mother liquor plus 20% (v/v) glycerol or ethylene glycol as cryoprotectants. The ligands were manually modeled into the electron density maps and were refined similarly to that described above. Although a mixture of α- and β-anomers may exist in solution, only the β-form of the monosaccharides was observed at the active site of the different complexes. For the docked glucotetraose coordinates, not present in the Protein Data Bank, a model was built by the on-line carbohydrate-building program GLYCAM (45).</p><p>Many attempts to crystallize the C-terminal section of the protein using the available construct were unsuccessful, and therefore, a model was built as explained below. The figures were generated with PyMOL (46). The atomic coordinates have been deposited in the RCSB Protein Data Bank under the accession codes 5K6I, 5K6M, 5K6N, and 5K6O.</p><!><p>GlyA1 and GlyA1-ΔCt stock solutions (10 mg/ml) were dialyzed against the same buffer (20 mm Tris-HCl, pH 7.5, 50 mm NaCl, 2 mm DTT, and 5% glycerol) for 18 h. SAXS measurements were performed at ESRF on beamline BM29, equipped with a Pilatus 1M detector. Each sample concentration, prepared by dilution of these stock solutions, was measured in 10 frames, 1-s exposure time per frame, at 4 °C, at a sample-to-detector distance of 2.867 m, using an x-ray wavelength of 0.991 Å. No radiation damage was observed during the measurements. The SAXS curves for buffer solutions were subtracted from the protein solution curves before analysis.</p><p>The scattering curves from six gradual concentrations, from 0.3 to 5 mg/ml, were scaled and averaged to obtain the I(q) function using the ATSAS software package (47). The radius of gyration (Rg) for each protein was calculated by Guinier plot using the program PRIMUS, and the pair distribution function P(r) and the maximum particle size Dmax were obtained by the program GNOM. Then, POROD was used to calculate the excluded volume of the particle, as well as the molecular weight of each sample.</p><p>Several homology and threading modeling programs were tried to obtain a model of the last 123 residues of GlyA1. All of them predicted a topology corresponding to carbohydrate-binding domains of families CBM6/CBM35, but they differed in the length of the linker attaching this domain to the core protein. Finally, models obtained from Swiss-Model (48) and CPHmodel (49) servers were used (templates from PDB entries 2W46 and 1UYX), each predicting a loop of 32 or 5 residues, respectively. Both entries share less than 20% identity with the C-terminal region of GlyA1.</p><p>Subsequently, CORAL (47) was used for several rounds of two-domain rigid body fitting, using the GlyA1-ΔCt coordinates and both templates, alternately; linkers were built as dummy atoms. The fit of the CORAL models to the SAXS experimental data were evaluated by the χ2 value calculated from the program CRYSOL (47).</p><!><p>The positioning of the sequence of the GlyA1 (α/β)8 barrel domain was analyzed in a phylogenetic tree. The predicted protein sequences were aligned against the National Center for Biotechnology Information non-redundant (NCBI nr) database using BLASTP algorithm. We downloaded all 27,499 GH3 sequences deposited in public databases. They were grouped within five different domain architectures as follows: ABB (9, 196), ABB_ABS (3,392), ABB_ABS_FLD (11,910), ABB_ABS_PA14_FLD (2,673), and ABS_FLD_ABB (328), where ABB, ABS, FLD, and PA14 refer to (α/β)8 barrel domain, (α/β)6-sandwich, fibronectin-like type III domain, and protective antigen PA14 domain, respectively. We discarded those sequences (848) from the ABB_ABS group longer than 700 amino acids, as they represent enzymes with unidentified domains downstream from the ABS module. Subsequently, the sequence corresponding to the ABB domain was extracted from all of the five sub-groups. An additional filter was applied to remove ABB sequences with coverage lower than 60% of the consensus domain defined by Interpro or Pfam databases (i.e. with less than 200 amino acids). The final number of sequences was the following: ABB (8,109), ABB_ABS (2,312), ABB_ABS_FLD (7,335), ABB_ABS_PA14_FLD (1,664), and ABS_FLD_ABB (289). For each of the five sub-groups, redundant sequences (those sharing more than 50% identity) were eliminated to select sequences that belong to different taxonomic groups. Following this procedure, the final selected sequences were as follows: ABB (132), ABB_ABS (54), ABB_ABS_FLD (45), ABB_ABS_PA14_FLD (20), and ABS_FLD_ABB (22). Multiple protein alignment was performed using ClustalW program, built into the software version 2.1. Phylogenetic analysis was conducted with the Ape package implemented for R programming language.</p><!><p>J. S. A., M. F., and J. P. conceived and coordinated the study. M. V. P. and M. F. contributed to screening, gene cloning, and enzyme production and characterization. P. N. G. contributed to metagenomics clone resources. J. S. A., B. G. P., and M. R. E. designed the crystallographic work and the SAXS experiments and interpreted the results. M. R. E. performed all the crystallography and SAXS experiments. J. M. N. and J. P. performed the phylogenetic analysis. J. S. A. and M. F. wrote the paper, and all authors read and commented on the manuscript.</p><!><p>This work was supported by Grants BIO2013-48779-C4-2-R, BIO2013-48779-C4-3-R, and BIO2014-54494-R from the Spanish Ministry of Economy and Competitiveness; ERA Net IB2 Project MetaCat through the Spanish Ministry of Economy and Competitiveness Grant PCIN-2014-107; United Kingdom's Biotechnology and Biological Sciences Research Council Grant BB/M029085/1; European Union's Horizon 2020 Research and Innovation Program (Blue Growth, Unlocking the potential of Seas and Oceans) Grant 634486, and the European Regional Development Fund (ERDF). The authors declare that they have no conflicts of interest with the contents of this article.</p><p>The atomic coordinates and structure factors (codes 5K6L, 5K6M, 5K6N, and 5K6O) have been deposited in the Protein Data Bank (http://wwpdb.org/).</p><p>glycoside hydrolase family-3</p><p>carbohydrate-binding module</p><p>degree of polymerization</p><p>p-nitrophenyl</p><p>pNP-α-glucose</p><p>pNP-α-maltose</p><p>pNP-β-glucose</p><p>pNP-β-cellobiose</p><p>pNP-α-arabinofuranose</p><p>pNP-β-arabinopyranose</p><p>pNP-α-xylose</p><p>pNP-β-xylose</p><p>pNP-β-xylobiose</p><p>pNP-α-fucose</p><p>pNP-α-rhamnose</p><p>pNP-α-mannose</p><p>pNP-β-mannose</p><p>pNP-α-galactose</p><p>pNP-β-galactose</p><p>pNP-β-lactose</p><p>pNP-N-acetyl-β-d-glucosaminide</p><p>pNP-N-acetyl-β-d-galactosaminide</p><p>root mean square deviation</p><p>seminal rumen fluid</p><p>small-angle x-ray scattering</p><p>2-[bis(2-hydroxyethyl)amino]-2-(hydroxymethyl)propane-1,3-diol</p><p>Protein Data Bank</p><p>fibronectin-like type III domain.</p>

PubMed Open Access

Strange kinetics of bulk-mediated diffusion on lipid bilayers

Diffusion at solid-liquid interfaces is crucial in many technological and biophysical processes. Although its behavior seems deceivingly simple, recent studies showing passive superdiffusive transport suggest diffusion on surfaces may hide rich complexities. In particular, bulk-mediated diffusion occurs when molecules are transiently released from the surface to perform three-dimensional excursions into the liquid bulk. This phenomenon bears the dichotomy where a molecule always return to the surface but the mean jump length is infinite. Such behavior is associated with a breakdown of the central limit theorem and weak ergodicity breaking. Here, we use single-particle tracking to study the statistics of bulk-mediated diffusion on a supported lipid bilayer. We find that the time-averaged mean square displacement (MSD) of individual trajectories, the archetypal measure in diffusion processes, does not converge to the ensemble MSD but it remains a random variable, even in the long observation-time limit. The distribution of time averages is shown to agree with a L\xc3\xa9vy flight model. Our results also unravel intriguing anomalies in the statistics of displacements. The time averaged MSD is shown to depend on experimental time and investigations of fractional moments show a scaling \xe2\x8c\xa9|r(t)|q\xe2\x8c\xaa \xe2\x88\xbc tqv(q) with non-linear exponents, i.e. v(q) \xe2\x89\xa0 const. This type of behavior is termed strong anomalous diffusion and is rare among experimental observations.

strange_kinetics_of_bulk-mediated_diffusion_on_lipid_bilayers

1 Introduction<!>2.1 Preparation of supported lipid bilayers<!>2.2 Protein expression, purification, and labeling<!>2.3 Imaging and single-particle tracking<!>3.1 Fluctuations in time-averaged MSD<!>3.2 Stationarity and dependence on experimental time<!>3.3 Strong anomalous diffusion<!>4.1 Fluctuations in the time averages<!>4.2 Fractional moments<!>5 Numerical Simulations<!>6 Discussion<!>7 Conclusions

<p>Processes at solid-liquid interfaces play important roles across multiple fields. In particular surface diffusion and diffusion-controlled reactions have key functions in life sciences and biomedical technologies1. For example, surface reactions are of utmost importance in the development of implant biomaterials2,3, affinity chromatography methods4, and biosensors as well as in blood-contacting devices5 such as heart valves and hemodialysis membranes. In cell biology biomolecular recognition and reactions on surfaces are essential for a vast array of physiological functions. The importance of molecular films in biology has been discussed for more than a century6. In fact, most biochemical reactions in cells take place at interfaces instead of in solution. Diffusion-controlled reactions often involve a search for a reactive target with the goal of minimizing the search time7–9.</p><p>The random motion of a particle is usually characterized by the mean squared displacement (MSD). In its simplest form, diffusion processes can be described by Brownian motion, which in two dimensions (2D) manifests a linear MSD 〈r2(t)〉 = 4Dt, where D is the diffusion coefficient. However, diffusion at solid-liquid interfaces can exhibit rich complexities10–14. Systems with a nonlinear MSD 〈r2(t)〉 = Kαtα display anomalous diffusion, where a slower-than-linear growth, i.e α < 1, indicates subdiffusion; and faster-than-linear growth, α > 1, indicates superdiffusion. Most importantly, anomalous diffusion alters reaction kinetics because the diffusion properties control the rate of molecular encounters15,16.</p><p>A widespread feature of molecules diffusing at the solid-liquid interface involves the desorption of molecules from the surface into the liquid phase. Molecules will diffuse in three dimensions (3D) until they reach the interface again and readsorb. This intermittent process where molecules alternate between 2D and 3D phases is known as bulk-mediated diffusion and has been previously analyzed in terms of scaling arguments17, simulations18,19, and analytical approaches20. The first experimental study that probed bulk-mediated diffusion involved the dynamics of adsorbate molecules in porous glass studied by field-cycling NMR relaxometry21. Recently bulk-mediated diffusion was experimentally observed in systems of vastly different nature including organic molecules at chemically coated interfaces12,22, polymer-surface interactions23, and membrane-targeting domains on both supported lipid bilayers24,25 and the plasma membrane of living cells26. Diffusion as measured on the surface is strongly influenced by the statistics of excursion times. On each excursion a random distance is covered on the surface, which scales in probability as the square root of the return time (〈r2(t)〉) = 4Dbt). The first return time to the surface has interesting properties27. The most fundamental of these properties is the dichotomy between mean first return time and probability of return. On one hand, the mean first return time is infinite due to its heavy tail distribution p(t) ∼ t−1.5. On the other hand, a particle always returns to the surface, that is the probability of return is one. In terms of probability theory one would say the particle returns to the surface almost surely. To place the problem in real context, if we consider a generic protein that alternates between a lipid bilayer and a water-based solution, the probability that it returns to the surface within less than 50 ms after it reached a 10-nm height is 99.75%25. Another interesting scenario is provided by a cylindrical surface. In this case the MSD features a plateau, balancing increasingly long jumps with a decreasing return probability.28.</p><p>A diffusion process where long jumps with a heavy-tail distribution occur is known as Lévy walk29. In such a random walk, jumps are performed at a velocity that might depend on the jump distance30,31. If the long jumps take place instantaneously, the process is known as Lévy flight32. Lévy walks have traditionally received more attention than flights because instantaneous jumps are not realistic. However, in the limit where bulk diffusion is orders of magnitude faster than surface diffusion, Db ≫ Ds, a Lévy walk can be approximated as a Lévy flight, at least within short time scales. This regime is found to be the most relevant for experimental observations of bulk-mediated diffusion.</p><p>Both Lévy flights and walks are superdiffusive when the probability density of jump distances scales as p(r) ∼ r−(1+β) with β ≤ 2. We recently reported that the motion of membrane-targeting domains on lipid bilayers is superdiffusive due to bulk excursions25. In these experiments, the MSD grows faster-than-linear when it is measured over an ensemble of molecules, that is the average is performed by employing a single displacement for each trajectory at any given time. Nevertheless, when the average is performed over time, i.e. by averaging all the displacements observed along a trajectory, the MSD is linear in lag time. This observation contradicts the ergodic hypothesis, one of the cornerstones of statistical mechanics, which states that ensemble averages and longtime averages of individual trajectories are equivalent. A similar behavior is found in subdiffusive continuous time random walks (CTRWs), where the ensemble-averaged MSD follows a power law tα, but the time-averaged MSD is linear33,34. In the CTRW, the non-ergodic property is rooted in the system not being stationary. Alike, molecular crowding conditions were predicted to introduce long-tailed distributions in both the unbinding times from the surface to the bulk and the rebinding times, which cause weak ergodicity breaking35. Such strange kinetics where the random walk exhibits different scaling properties depending on whether it is averaged over time or over an ensemble poses intriguing questions regarding its statistics. Beyond the MSD, the distribution of displacements also deviates from "normal" diffusion. The central limit theorem (CLT) warrants that the displacements of Brownian motion have a Gaussian distribution. However, in some types of anomalous diffusion models, the CLT breaks down and the distribution of displacements is no longer Gaussian. For example, in a CTRW or when a particle diffuses in a fractal structure, the increments are not independent and thus the CLT does not hold. In a Lévy flight the CLT breaks down because the increments can have infinite variance17,20.</p><p>Here we investigate the kinetics of membrane-targeting C2 domains on lipid bilayers using single-particle tracking. This system exhibits superdiffusive behavior in the ensemble-averaged MSD but normal scaling in the time-averaged MSD. Weak ergodicity breaking predicts large fluctuations in the time-averaged MSD of individual trajectories. Thus we examine the fluctuations in the MSD and find that it remains a random variable even in the long time limit. In contrast to the CTRW model, the increments of bulk-mediated diffusion are shown to be stationary, but the statistics of the motion still depend on experimental time. It is found that when the MSD is averaged over both time and ensemble, it does not converge to a finite value, but it increases with experimental time. Thus, if the diffusion coefficient were estimated using the MSD slope, it would increase as the experimental time increases. The experimental results for bulk-mediated diffusion are found to agree with a Lévy flight model using both analytical approaches and numerical simulations. Interestingly we also find the system exhibits strong anomalous diffusion36, i.e., the fractional moments are not characterized by a linear scaling exponent as in most diffusion processes.</p><!><p>Lipid bilayers were prepared as described elsewhere25. In brief, chloroform-suspended 18:1 (∆9-Cis) PC (DOPC) and 18:1 PS (DOPS) were mixed at a ratio of 3:1. The phospholipid mixture was vacuum dried overnight and resuspended in imaging buffer (50 mM HEPES, 75 mM NaCl, 1 mM MgCl2, 2 mM tris(2-carboxyethyl)phosphine (TCEP), 200 μM CaCl2) to a final concentration of 3 mM followed by probe sonication to form sonicated unilamellar vesicles (SUVs)37. A solution of SUVs (1.5-mM lipid) in 0.5 M NaCl and imaging buffer was introduced into a perfusion chamber (CoverWell, Grace Bio-Labs model PC8R-1.0) and incubated for one hour at 4°C. The surface was then rinsed with imaging buffer multiple times prior to addition of protein sample.</p><!><p>An expression plasmid containing the ybbr-Synaptotagmin 7 (Syt7) C2A gene38 was transformed into Escherichia coli BL21-CodonPlus(DE3) competent cells. Cells were grown at 37°C to an OD600 of 0.6 and then induced to express protein with 0.5 mM IPTG for 6 hours at room temperature. The harvested cells were lysed at 18,000 lb/in2 in a microfluidizer in a buffer containing 50 mM Tris, pH 7.5, 400 mM NaCl and centrifuged at 17,000 rpm. The clarified lysate was loaded onto a 5-ml GSTrap FF column (GE Healthcare LifeSciences, Pittsburgh, PA) followed by gradient elution with 50 mM Tris, pH 8.0, 100 mM NaCl, and 10 mM glutathione. Fractions containing protein were pooled and diluted to reduce the salt to less than 0.1 M prior to loading onto a HiTrap Q HP column (GE Healthcare LifeSciences, Pittsburgh, PA) and eluting with a linear gradient to 1 M NaCl in 25 mM Tris, pH 8.5, 20%(vol/vol) glycerol, and 0.02%(wt/vol) NaN3. A portion of the purified protein was subjected to thrombin cleavage to remove the GST tag and then separated using a Superdex 200 gel filtration column (GE Healthcare LifeSciences, Pittsburgh, PA) equilibrated in 50 mM Tris, pH 7.5 and 100 mM NaCl.</p><p>20 mM CoASH (New England Biolabs, Ipswich, MA) in 400 mM Tris, pH 7.5 was mixed with 20 mM Atto-565 maleimide (ATTO-TEC, Siegen, Germany) in dimethylformamide and incubated at 30°C overnight to form Atto-565 CoA, then diluted 10 fold with 5 mM DTT, 10 mM Tris pH 7.5 to quench the reaction. ybbr-Syt7 C2A was labeled with the Atto-565 via SFP synthase (4′-phosphopantetheinyl transferase). Samples were dialyzed against 1 L of 50 mM HEPES, pH 7.0, 75 mM NaCl, 4 mM MgCl2 and 5% glycerol overnight at 4°C, and then concentrated to 10 μM.</p><!><p>Proteins were added to the imaging buffer to a final concentration of 75 pM. Then, the perfusion chamber was filled with the solution. The perfusion chambers were 9 mm in diameter and 0.9-mm deep, holding a volume of ≈ 60 μl. Imaging was performed at room temperature without replacing the solution, so that there was always protein present in the bulk solution and the surface concentration could reach a steady state.</p><p>All images were acquired using an objective-type total internal reflection fluorescence microscope (TIRFM) as described previously10,39. A 561 nm laser line was used as excitation source. A back-illuminated electron-multiplied charge coupled device (EM-CCD) camera (Andor iXon DU-888) liquid-cooled to -85°C, with an electronic gain of 300 was used. In order to maintain constant focus during the whole imaging time we employed an autofocus system (CRISP, Applied Scientific Instrumentation, Eugene, OR) in combination with a piezoelectric stage (Z-100, Mad City Labs, Madison, WI). Videos were acquired at a frame rate of 20 frames/s using Andor IQ 2.3 software and saved as 16-bit tiff files. The images were filtered using a Gaussian kernel with a standard deviation of 1.0 pixel in ImageJ. Single-particle tracking of Atto-C2 was performed in MATLAB using the u-track algorithm developed by Jaqaman et al.40.</p><!><p>We tracked the motion of membrane-targeting C2A domains25, fluorescently labeled with Atto-565, on a supported lipid. Imaging was done in a home-built total internal reflection (TIRF) microscope under continuous illumination at 20 frames/s. Single-particle tracking is performed under conditions where the surface density is low enough to enable connections of long jumps while avoiding misconnections due to crossover between trajectories. Figure 1 shows an example of single-molecule trajectories during 10 seconds. As a first step, we characterize the diffusion by analyzing the MSD as a function of lag time. For each individual trajectory, the time-averaged MSD (TA-MSD) is calculated as</p><p>where Δ is the lag time, t the experimental time, and r the two-dimensional position of a particle. Across the manuscript we employ brackets to denote the ensemble average of an observable 〈·〉 and an overline to denote time averages ·̄. Figure 2(a) shows that, within experimental error, the TA-MSD of individual trajectories is linear in lag-time, resembling pure Brownian motion. In two dimensions, the MSD of a Brownian particle is determined by the diffusion coefficient D via the relation δ2(Δ)¯=4DΔ, but Fig. 2(a) shows that the TA-MSD exhibits broad fluctuations. In ergodic systems, the time-averaged MSD converges to the ensemble average. In other words the time-averaged MSD can be used to consistently estimate the diffusion coefficient of a molecule. However, the large scattering seen Fig. 2(a) indicates the time-averaged diffusion coefficient of individual molecules is a random variable, with no apparent convergence. This observation suggests weak ergodicity is broken in the sense that time and ensemble averages do not converge to the same values34.</p><p>Given that the TA-MSD is linear in lag time, one is tempted to find the diffusion coefficient of individual molecules from linear regression of the MSD trace. Figure 2(b) shows the distribution of the slope of the TA-MSD, i.e. δ2¯/Δ, obtained from 5,187 trajectories. The distribution shows two different populations. A peak with very low diffusivities is apparent (sample mean 〈δ2¯/Δ〉=0.006μm2/s). This population has a narrow distribution and it is attributed to particles that are immobilized and do not exhibit any motion. A second population with high diffusivities has the characteristic large variations noted in Fig. 2(a), with a mode at 2.7±0.1 μm2/s but a sample mean 〈δ2¯/Δ〉=7.3μm2/s. When particles perform long jumps, a trajectory can be truncated and traces with higher diffusivities are lost. It is thus expected that the true distribution of MSDs is even broader because experimental tracking is biased towards lower diffusivities.</p><!><p>It is important to establish whether the diffusion process evolves with time. Further, ergodicity is defined only for stationarity processes and thus we test whether the non-ergodic motion is rooted in the increments not being stationary. One way to check stationarity of the increments is to compute the quantiles as a function of time. If the quantile lines are parallel then we can infer that the process is stationary41. Figure 3(a) shows the 10-quantile lines of the increments for lag times of 50 ms. The quantile lines appear to be parallel, suggesting the distribution of increments does not change over time. Therefore we can conclude that the process is stationary.</p><p>Even though the increments are stationary, the statistics of the diffusion process depends on experimental time. This effect is observed in the average of the time-averaged MSD, i.e., the time-and ensemble-averaged MSD (TA-EA-MSD, 〈δ2¯〉). The TA-EA-MSD is simply the cumulative moving average of the square displacements, over different trajectories and for all times up to the experimental time. Figure 3(b) shows the TA-EA-MSD for Δ = 50 ms as a function of experimental time measured for 3,130 trajectories. The MSD does not appear to converge to any given value; instead it exhibits random jumps, so that it experiences an overall increase with experimental time. In ergodic systems, the TA-EA-MSD exhibits fluctuations around the mean, which become smaller as the available experimental time becomes longer due to better statistics. That type of noise is different from the behavior observed here because ergodicity would warrant the TA-EA-MSD converges to a finite value. The observed MSD increase is not monotonic and it decreases smoothly between jumps. Nevertheless, the rate of decrease of the MSD is much smaller than the average rate of increase due to the discrete jumps and thus, in probability, the MSD increases with time. As a consequence, if the ensemble-averaged MSD were employed to estimate a diffusion coefficient, then the coefficient would not be constant, but it would increase with experimental time.</p><!><p>So far, we have characterized the dynamics of molecules using the MSD and observed that the TA-MSD δ2¯ does not converge to the ensemble-averaged MSD 〈r2(t)〉. However, one may desire to characterize the motion beyond the second moment. In particular, the fractional moments 〈|r(t)|q〉 with q > 0 provide useful insight. For Brownian motion as well as many anomalous diffusion processes 〈|r(t)|q〉 ∼ tqv. As long as v is a constant, all moments are described by a scaling exponent linear in the order q and the process is scale invariant such that the propagator at different times is P(x,t) = t−vf(x/tv)36. For example, in Brownian motion v = 1/2 and f(·) is a Gaussian function.</p><p>The process is said to exhibit strong anomalous diffusion when v is not constant13,36,</p><p>Strong anomalous diffusion has been shown theoretically and via numerical simulations in a variety of systems including the motion of tracer particles in a running sandpile model42, the occupation times of renewal processes43, and flow fields36 among others44–46. In these processes, a piecewise linear scaling is found for qy(q). Experimental observation of strong anomalous diffusion has remained rather elusive. To the best of our knowledge, so far it has only been observed in the superdiffusive transport of polymer particles inside living cancer cells47. Figure 4(a-d) shows ensemble-averaged moments of the two-dimensional displacements of C2 domains, which are computed by averaging over all available trajectories 〈|r(t) − r(0)|q〉. Two regimes are visible in all the moments. At short times, the fractional moments exhibit the behavior expected for Brownian motion, 〈|r(t)|q〉 ∼ tq/2, but at long times the moments "misbehave". Two solid lines are shown in each panel of Fig. 4(a-d): a shallow line with 〈|r(t)|q〉 ∼ tq/2 and a steeper line with 〈|r(t)|q〉 ∼ tq. For short times the agreement with a Brownian motion model (qy(q) = q/2) is evident. However, this is not the case for the long-time regime. In this regime, as the order q increases, the logarithmic slopes of the moments also increase. Figure 5 shows v(q) as a function of q for both the short and long times. We see that the scaling exponent at short times does not show significant deviations from qv(q) = q/2 but in the long-time regime v(q) is not constant. In this time regime, qv(q) = q for the lower order moments and v(q) > 1 for the higher orders, which indicates strong anomalous diffusion. In our measurements, strong anomalous diffusion is caused by rare long jumps, i.e., by bulk excursions. When the large displacements are excluded from the analysis, the fractional moments display normal behavior. Figure 4(e-h) shows the fractional moments when only the displacements below a 97% cutoff are considered. We observe that in this case 〈|r(t)|q〉 ∼ tq/2, that is, the fractional moments without the long jumps scale with time as expected from Brownian motion.</p><!><p>We have previously shown25 that membrane-targeting domains can transiently dissociate from the lipid bilayer to perform bulk excursions. During these excursions, a molecule undergoes three-dimensional diffusion until it readsorbs on the surface. Within the bulk phase, the height z is modeled as a one-dimensional random walk and thus the first return time distribution satisfies p(tb)~tb−1.5, where the first return time tb represents the time the particle spends in the bulk during a single jump. A sketch of the model is shown in Fig. 6. A simple derivation27 leads to a one-sided Lévy distribution of index 1/2, also known as a Lévy-Smirnov distribution,</p><p>where Db is the diffusion coefficient in the bulk, and z0 is a scaling constant with units of length. Then, the distances on the surface covered during bulk excursions are two-dimensional Cauchy random variables17,20,25</p><p>where γ is a constant with units of length. Interestingly, the expected values of both the first return time and the displacement diverge. Therefore, we expect that the time-averaged MSD is governed by extreme values. Namely, because the TA-MSD is determined by individual long jumps, it remains a random variable, even though observation times may be long.</p><p>Let us first derive the distribution of time averages from intuitive scaling arguments. Given that one individual long jump determines the TA-MSD of an individual trajectory, each TA-MSD scales as the longest displacement within the trajectory,</p><p>where ri are the individual measured displacements. From Eq. (4), we can calculate the probability density of squared displacements and that of TA-MSD. Defining s = r2, we obtain the distribution p(s) = 0.5γ(s + γ2)−3/2. Then, we find the distribution of δ2¯ from the cumulative distribution function of the squared displacements FS(s). Namely, FMSD(δ2¯)=[FS(tδ2¯)]t because the displacements are independent and identically distributed. Thus</p><p>where for the sake of simplicity we take time t as the number of time intervals, i.e. the number of measured displacements. In the limit of large MSDs we have tδ2¯≫γ2 and Eq. (6) simplifies to</p><p>These simple scaling arguments yield a distribution of TA-MSD that has a power law tail with an exponent 3/2.</p><p>Now, we follow the derivation by Froemberg and Barkai to find the whole distribution of TA-MSDs48. In order to simplify the analysis we focus on a one-dimensional Lévy flight but the extension to two dimensions is straightforward. Again the displacements are Cauchy distributed (Eq. 4), albeit in one dimension,</p><p>and the square displacements y = x2 are distributed according to</p><p>where y ≥ 0. The displacements after time Δ are xΔ∑iΔxi, with a characteristic function ϕ(k) =exp(−γΔ|x|). Thus p(xΔ)=γt[π(xΔ2+γ2t2)]−1, also a Cauchy distribution with a scale parameter γΔ. This behavior is due to the fact that the Cauchy distribution is stable, namely a symmetric Lévy stable distribution of index 1. Therefore, we can solve for Δ = 1 and our results are still valid for any lag time after rescaling γ → γΔ.</p><p>As in Eq. 1, the TA-MSD at a lag time Δ, measured over a time t, is48</p><p>where the approximation holds for t ≫ 1. We next define the variable ζ=tδ2¯/Δ≈d∑i=1txi2, which is a sum of independent and identically distributed (i.i.d.) random variables yi. Given that the variance of yi diverges, the central limit theorem breaks down and the distribution of ζ is found using the generalized central limit theorem49. The Laplace transform of the distribution of y = x2 (see Eq. 9) is</p><p>where erfc(·) is the complementary error function. We are concerned with large values of y and therefore we only keep the first term in the series expansion in Eq. (11), that is we only consider the small uy limit in Laplace domain. The distribution of ζ in the large t limit is found in Laplace domain</p><p>The inverse Laplace transform yields</p><p>where L1/2,1 (ξ) is again the Lévy-Smirnov distribution and we introduced the constant c=(π/2)/γt. We can then change variables to obtain the distribution of the slope of the TA-MSD. By defining ξ=ζ/t=δ2¯/Δ, Eq. 13 simplifies to</p><p>Thus we find that the probability density function of the TA-MSD is a Lévy-Smirnov distribution with scale parameter 2γ2t/π. Recall that we derived this distribution for t and Δ in number of frames. In agreement with the scaling arguments discussed above, p(ξ) ∼ ξ−3/2. Importantly, the moments of this distribution diverge, causing large variations in the TA-MSD measurements as observed in Fig. 2(a).</p><!><p>Our Lévy flight model involves a tail in the distribution of displacements that scales as p(r) ∼ r−3 at long distances. Therefore, the qth moment diverges for q ≥ 1. Explicitly,</p><p>Of course this result is not realistic. The problem arises in the approximation that bulk excursions take place instantaneously. While the approximation is good within our experimental times, it does not hold for very long jumps, thus placing a bound on the higher order moments. In fact, for bulk-mediated surface diffusion the Cauchy distribution (Eq. 4) has a natural Gaussian cutoff that emerges at longer times than those probed in our study20. Precise mathematical analysis that includes the time incurred by a bulk mediated jump would lead to the correct higher order moments13,50. However, a simple model leading to Eq. (15) yield some useful insights. In particular, we can see that there is a critical order qc = 1 below which v(q) = 1. Furthermore, for values q < qc, the fractional moments yield superdiffusive behavior, i.e v(q) > 1/2 as would be determined by Brownian motion. Above this critical value, the fractional moments increase above 1. The piecewise behavior is the fingerprint of strong anomalous diffusion as observed in Fig. 5.</p><!><p>We test the predictions of our model using numerical simulations and compare them to the experimental data. Our simulations intend to model a process where molecules diffuse on a two dimensional surface and undergo dissociation into the bulk phase. Dissociation is considered as a Poisson process and the particle goes through 3D diffusion in the bulk until it finds its way back to the surface. 5000 realizations were simulated off-lattice where tracers perform a random walk with Gaussian displacements in two dimensions, and at random times the tracer performs bulk excursions25. The sojourn times within the surface are exponentially distributed with a mean of 10 and the surface diffusion coefficient is taken to be Ds = 0.5. The return times from bulk excursions are drawn from a distribution ψ(tb)=(4πtb3)−1/2exp(−1/4tb) (see Eq. 3). Then the jump distances are Gaussian with variance σb2=2tb.</p><p>Similar to the experiments on lipid bilayers, the TA-MSD of the simulations exhibit a broad scattering. Figure 7(a) shows the distribution of TA-MSDs for the individual realizations. Overlaid on this distribution in Fig. 7(a), equation 14 shows good agreement with the MSD distribution.</p><p>In our derivation of the distribution of the TA-MSDs, we have employed the Cauchy distribution (Eq. 4) for the displacements. This equation ignores the Gaussian component in the distribution of displacements that arises due to the diffusive motion on the surface25. As seen in Fig. 7(a), this approximation does not alter the distribution, at least in the long measurement time limit. The reason is that, as discussed above, the MSD is governed by the large displacements, i.e., the tail of the distributions. Further, we would achieve the same results (Eq. 14) if we only consider the power law tail of the propagator, p(r) ∼ |r|−3 and find the Laplace transform using the Tauberian theorem49,51.</p><!><p>In a similar fashion to the numerical simulations of bulk-mediated diffusion, Eq. 14 is used to model the experimental results for membrane-targeting C2 domains (red solid line in Fig. 2(b)). Even though the agreement between our bulk-mediated diffusion model and the experimental results is satisfactory, the tail in the MSD distribution of C2 domains decreases faster than predicted by the model. This effect is caused by an artificial truncation of the distribution of displacements caused by the tracking algorithm25. Namely, if a particle experiences a very long jump, it is not possible to make frame-to-frame connections with reasonable confidence and thus trajectories are cropped missing the long displacements and in turn the large diffusivities.</p><p>The increments in the motion of C2 domains on a lipid bilayer are shown to be stationary but the MSD depends on the experimental time (Fig. 3). This behavior is also observed in our numerical simulations. The increments in the simulations are stationary (Fig. 7(b)) as the displacements are simulated with the same time-independent stochastic process. However, the TA-EA-MSD of the numerical simulations also shows a strong dependence on realization time (Fig. 7(c)). In agreement with the C2 data, the simulations MSD show discrete jumps in the time series. Also here, the MSD average increases in probability with realization time.</p><p>The discontinuities in the MSD as a function of experimental time can be conceptually understood in terms of the same mechanism that causes weak ergodicity breaking. As discussed before, the estimated diffusivities of individual trajectories are governed by extreme displacements. Recall that the reason for the lack of self-averaging is the existence of one displacement in the trajectory that is likely much larger than all others and thus the MSD depends on this individual displacement. In the same way, at a given time t, a jump may occur among all the molecules such that it is much larger than all the displacements observed thus far. When such an event takes place the TA-EA-MSD increases sharply due to the contribution of one long jump. After a very large jump occurs, the relative weight of that individual long jump diminishes because more data points become available. Thus, following a jump discontinuity the MSD decreases with experimental time. The MSD continues to decrease until the next jump discontinuity takes place.</p><p>We observed that, in probability, the sample mean of the TA-MSD increases with experimental time. This is also observed in the theoretical distribution of the TA-MSD (Eq. 14), which involves a scale parameter that explicitly depends on the experimental time. Even though the expected value of the TA-MSD diverges we can estimate how the MSD increases with experimental time by evaluating other measures of central tendency, such as the theoretical mode and median. Both of these measures scale linearly with time, namely.</p><p>and</p><p>where erfc−1 (·) is the inverse complementary error function. Thus we expect the average of the TA-MSD to increase in probability linearly with experimental time as observed in Fig. 3(b).</p><p>In this manuscript we employed the fractional moments and showed that the system exhibits strong anomalous diffusion. The fractional moments are only rarely used in the diffusion literature. Nevertheless, these moments can be very useful in the analysis of bulk mediated diffusion. When the trajectories are modeled as Lévy flights the theoretical MSD diverges and its use in the analysis of motion challenging. This is not the case for fractional moments with q<1, where the theoretical moment is finite and no discontinuities are observed. Thus low order moments become a useful tool to study phenomena such as superdiffusion.</p><p>Our data shows that the common practice of finding diffusivities from time averaged MSD in membranes should be approached with care. We find that weak ergodicity is broken and as a consequence the MSD of individual trajectories are random variables even in the long time limit. In other words, the MSD from individual trajectories are not reproducible. Furthermore, the ensemble mean of the time-averages is not a reliable measure because it depends on experimental time. Careful analysis indicates that as the available measurement time becomes longer, the apparent diffusion coefficient increases. In order to deal with this subtlety we propose that, when bulk excursions are evident in the data, parameters are extracted from the distributions instead of using either time or ensemble averages. We have previously shown that it is feasible to obtain both the surface diffusion coefficient and the scale parameter γ from the distribution of displacements when the data sample is large enough25.</p><!><p>We have shown that bulk-mediated diffusion can be accurately modeled as a Lévy flight. The Lévy flight concept yields superdiffusive dynamics with complex strange kinetics, in particular because the time averaged MSD does not converge to the ensemble average. Thus the process exhibits weak ergodicity breaking. The time-averaged MSD of individual trajectories is governed by individual long jumps and, as a consequence, it remains a random variable. We have shown that the MSD also depends on experimental time and thus it does not provide a consistent estimator of the diffusion coefficient. The long time asymptotic of the displacement fractional moments has the signature of superdiffusive behavior both for low and high orders. Moreover, the Lévy flight model predicts strong anomalous diffusion, a phenomenon that deals with non-linear scaling exponents of the fractional displacement moments. We have experimentally observed this anomalous behavior in the motion of membrane-targeting domains on supported lipid bilayers using single-particle tracking. Future work will explore the effects of temperature and macromolecular crowding on bulk-mediated dynamics. Given the broad applicability of bulk mediated diffusion, we foresee these anomalies can be observed in many complex systems.</p>

PubMed Author Manuscript

Lesion orientation of O4-alkylthymidine influences replication by human DNA polymerase η† †Electronic supplementary information (ESI) available: Materials, experimental procedures, compound characterization, and additional discussion. The atomic coordinates and structure factors (codes 5DLF, 5DLG, 5DQG, 5DQH and 5DQI) have been deposited in the Protein Data Bank (http://www.wwpdb.org/). See DOI: 10.1039/c6sc00666c

Conformation of the α-carbon of O4-alkylthymidine was shown to exert an influence on human DNA polymerase η (hPol η) bypass. Crystal structures of hPol η·DNA·dNTP ternary complexes reveal a unique conformation adopted by O4-methylthymidine, where the nucleobase resides nestled at the active site ceiling where hydrogen-bonding with the incoming nucleotide is prevented.

lesion_orientation_of_o4-alkylthymidine_influences_replication_by_human_dna_polymerase_η†_†electroni

Introduction<!>Synthesis and characterization of modified oligonucleotides<!>Steady-state kinetics<!>LC-MS/MS analysis of full-length extension products produced by hPol η<!>Pre-steady-state kinetics<!>Crystal structures of ternary hPol η·DNA·dNTP complexes with templates containing O4MedT or O4EtdT at the insertion stage<!>Crystal structure of a ternary hPol η·DNA·dCTP extension-stage complex with O4EtdT paired opposite primer dA<!>Discussion<!>Conclusions<!>

<p>DNA alkylation results from a variety of endogenous and/or exogenous agents that can interfere with vital cellular processes, i.e. replication and transcription.1 The addition of alkyl appendages on the DNA scaffold can have adverse consequences such as DNA polymerase (Pol) blockage, nucleotide misincorporation, chromosomal instability, and activation of the cellular apoptotic pathway.1,2 However, organisms have various repair pathways to restore damaged DNA. In the event that a lesion evades the process of DNA repair, translesion synthesis (TLS) by Y-family DNA Pols can occur, allowing bypass of the DNA lesion in an error-free or error-prone manner.3 Y-family DNA Pols are described as more "promiscuous" given their larger active sites when compared with replicative DNA Pols, which accounts for their ability to bypass damaged nucleotides that induce blockage. DNA Pol η in humans (hPol η) plays a pivotal role in the bypass of certain UV-induced DNA damage, which impedes DNA replication.4 hPol η activity has also been correlated with chemotherapeutic resistance to platinum-based agents such as cisplatin and the efficient bypass of the oxidative DNA lesion 7,8-dihydro-8-oxo-2′-deoxyguanosine.5,6</p><p>The O4-position of thymidine is susceptible to alkylation by agents such as N-nitroso alkylamines in certain foods, water, air, and particularly tobacco products.7,8 Albeit a minor site of alkylation, lesions such as O4-methylthymidine (O4MedT) and O4-ethylthymidine (O4EtdT) are poorly processed by mammalian repair pathways, making them persistent in the genome.9,10 O4MedT and O4EtdT hinder high fidelity replicative DNA Pol activity, resulting in misinsertion of dGTP in the daughter DNA strands.10–12 Correlations between the mutagenicity of O4MedT and cancer have been established,13,14 highlighting the importance of investigating the structural properties and biological outcomes associated with this type of DNA damage.</p><p>The current understanding of the mechanism of Y-family DNA Pol misincorporation during TLS depends on a number of factors, including the nature of the DNA damage, the DNA Pol and the incoming nucleoside triphosphate. The geometrical array of the ternary complex formed (involving DNA, protein and nucleoside triphosphate) is the key characteristic that governs efficient bypass of a DNA lesion. Structural investigation by NMR and X-ray crystallography of duplexes containing an O4MedT insert has revealed that the methyl group preferentially adopts a syn conformation around the C4–O4 bond (Fig. 1).15,16 We hypothesized that the conformation of the O4-alkyl lesion could affect the base pair geometry during the primer extension reaction catalyzed by DNA Pol η. To address this possibility, we probed hPol η processivity with thymidine analogs that link the C5 and O4 atoms by a dimethylene or trimethylene group, which limits the O4-lesion to adopt an anti-conformation (Fig. 1) to relate the structural features of O4-alkylated dT with the bypass activity of hPol η. hPol η was selected as the model Y-family DNA polymerase, given previous studies concerning bypass of O4MedT and O4EtdT.17,18 Results of these studies indicated that Pol η, from yeast or human, were most efficient in extending across and past O4MedT17 and O4EtdT,18 respectively.</p><p>We investigated bypass profiles opposite all four lesions by hPol η (steady-state single nucleotide incorporation and LC-MS/MS analysis of full-length extension products). Crystal structures of ternary hPol η·DNA·dATP and hPol η·DNA·dGTP with template strands containing O4MedT or O4EtdT reveal a distinct orientation of the former lesion that stacks atop a tryptophan residue near the ceiling of the active site instead of pairing with the incoming nucleotide. Conversely, O4EtdT pairs with both incoming dA and dG nucleotides via bifurcated H-bonds in the insertion complexes and displays the same configuration opposite primer dG in the crystal structure of an extension complex adjacent to the nascent dG:dCTP pair. The structures provide a better understanding of the different behavior of the O4MedT or O4EtdT lesions in hPol η-catalyzed error-prone bypass reactions and suggests a unique intermediate step in the bypass of O4MedT.</p><!><p>The structures of O4MedT, O4EtdT and the modified pyrimidyl nucleosides 3-(2′-deoxypentofuranosyl)-5,6-dihydrofuro[2,3-d]pyrimidin-2(3H)-one (DFP) and 3-(2′-deoxypentofuranosyl)-3,5,6,7-tetrahydro-2H-pyrano[2,3-d]pyrimidin-2-one (TPP) are shown in Fig. 1 (methods describing the preparation of nucleosides and oligonucleotides can be found in the ESI†). UV thermal denaturation studies of duplexes containing single inserts of the DFP or TPP modification revealed a comparable destabilizing effect to O4MedT and O4EtdT with complementary strands containing adenine or any mismatched base pairing partner (Fig. S38†). Circular dichroism spectra of duplexes containing the DFP or TPP inserts revealed little deviation from a B-form structure (see Fig. S39†).</p><!><p>Steady-state kinetic assays of individual nucleotide incorporations opposite O4MedT, O4EtdT, DFP, TPP and unmodified dT were carried out with the catalytic core construct of hPol η (amino acids 1-432). In all cases, these pyrimidyl modifications blocked DNA synthesis by the polymerase relative to the unmodified control (Fig. 2a, values shown in Table S1†). Incorporation of the correct dAMP nucleotide by hPol η opposite to O4MedT, O4EtdT, DFP, and TPP was reduced approximately 6.5-, 12-, 4.5-, and 5-fold, respectively, relative to dT (see Fig. 2a).</p><p>hPol η incorporated dCMP and dTMP opposite all the pyrimidyl modifications and dT with similar catalytic efficiencies (ranging from approximately 0.002–0.015 μM–1 s–1). However, a strong preference for either dAMP or dGMP incorporation opposite the modified pyrimidines was observed. Other than O4MedT, hPol η preferentially incorporated the correct dAMP nucleotide opposite all the pyrimidyl modifications. The significant incorporation of dGMP when hPol η encountered these pyrimidyl modifications, compared to the unmodified control, indicates a clear loss in substrate specificity by the polymerase (Fig. 2a). In the case of O4MedT, dGMP was slightly preferred as the nucleotide incorporated by hPol η (0.19 ± 0.01 vs. 0.18 ± 0.03 μM–1 s–1 for dGMP and dAMP, respectively).</p><!><p>Analysis of single insertions by a DNA polymerase is useful for kinetic evaluation but may not reflect incorporation fidelity in the presence of all four dNTPs across the damage and beyond this site. The fidelity of hPol η and its processivity past the damage site was addressed by the use of a full extension assay coupled with LC-MS/MS analysis.5,19,20</p><p>The optimal reaction times to observe the full extension products from the template strands containing the modifications and the unmodified control were evaluated (Fig. 1). Full extension was achieved for the unmodified control at 30 min, whereas templates containing the modifications required longer reaction times (60–90 min). The time course assay revealed that hPol η had difficulty in extending past O4MedT and O4EtdT and displayed a significant "S + 1" band at reaction times of 30 and 60 min. UPLC separation of the cleaved products and mass spectrometry analysis of their sequence identities revealed that dGMP was incorporated most efficiently opposite all the modifications (see Fig. 2b).</p><p>The incorporation frequency opposite dT, O4MedT, O4EtdT, DFP, and TPP for the full extension products was evaluated (see Fig. 2b and Table S2†). The presence of O4MedT increased the level of frameshift formation by hPol η relative to the control (9.5 vs. 3.5%). Comparable levels of frameshifts were observed opposite O4EtdT and the dT control. However, the templates containing the bicyclic pyrimidine adducts did not induce a similar increase in frameshift formation with levels that were approximately one-half, relative to the control. The correct dAMP nucleotide was incorporated by hPol η at a frequency of approximately 30, 24, 22, 16, and 93% opposite O4MedT, O4EtdT, DFP, TPP, and dT, respectively. Out of the lesions investigated, hPol η exhibited the highest fidelity opposite O4MedT and lowest opposite the TPP. Incorporation of dGMP was observed to occur in the extension products with overall frequencies of 60, 70, 72, 82, 3% opposite O4MedT, O4EtdT, DFP, TPP, and dT, respectively.</p><p>The accuracy of bypass varied for the O4-alkylthymidine modifications by approximately 2 : 1 in favor of dGMP opposite O4MedT to 5 : 1 in favor of dGMP opposite TPP. An increased adduct size, from O4MedT to O4EtdT and DFP to TPP, resulted in a 10% increase of dGMP misinsertion at the expense of a 10% decrease of the correct dAMP incorporation. Similarly, the conformationally restrained analogues (DFP and TPP) induced an increase in dGMP misinsertion (10%) by hPol η compared to O4MedT and O4EtdT, respectively.</p><!><p>The pre-steady-state kinetic assays of dATP and dGTP incorporations opposite O4MedT, O4EtdT, DFP, and TPP, and dATP incorporation opposite unmodified dT were carried out with the catalytic core of hPol η. The burst rates for dATP insertion were 3.1-, 4.2-, and 1.8-fold higher compared to dGTP opposite O4MedT, DFP, and TPP, respectively (Fig. S48 and Table S3†). The burst rates were low in the case of dATP and dGTP opposite O4EtdT. The burst amplitudes for the extensions were 15–35% opposite O4MedT, O4EtdT, DPF, and TPP, which may indicate the presence of multiple non-productive ternary complexes.</p><!><p>To visualize the O4MedT and O4EtdT lesions at the active site of hPol η trapped at the insertion stage, we determined four crystal structures of ternary complexes with the Pol bound to a 12mer template strand with the incorporated lesion and paired to an ; 8mer primer and incoming purine nucleoside triphosphate. For details regarding the crystallization, data collection and structure determination and refinement procedures please see the ESI.† Selected crystal data, data collection and refinement parameters and examples of the quality of the final electron density for all structures are summarized and depicted in the Table S4.† The two complexes with O4MedT-containing templates and incoming dATP or dGMPNPP reveal similar orientations of the lesion (Fig. 3, PDB ID codes ; 5DLF and ; 5DLG, respectively). Instead of pairing with the incoming nucleotide, O4MedT is lodged near the ceiling of the active site. Thus, its base portion is nestled against Trp-64 (base stacking interaction), Met-63 and Ser-62 (hydrophobic contacts between O4Me and both Cα and C( <svg xmlns="http://www.w3.org/2000/svg" version="1.0" width="16.000000pt" height="16.000000pt" viewBox="0 0 16.000000 16.000000" preserveAspectRatio="xMidYMid meet"><metadata> Created by potrace 1.16, written by Peter Selinger 2001-2019 </metadata><g transform="translate(1.000000,15.000000) scale(0.005147,-0.005147)" fill="currentColor" stroke="none"><path d="M0 1440 l0 -80 1360 0 1360 0 0 80 0 80 -1360 0 -1360 0 0 -80z M0 960 l0 -80 1360 0 1360 0 0 80 0 80 -1360 0 -1360 0 0 -80z"/></g></svg> O) from the two residues) from a loop region in the finger domain and Gly-46 from an adjacent β-strand that together form the roof of the active site (Fig. 3a and c). In addition, O2 of O4MedT and the amino group of Asn-38 are H-bonded (Fig. 3b and d). This position of the O4MedT lesion at the entrance of the active site places it quite far away from the incoming nucleotide triphosphate. The distance between its O4 atom and N6 of dATP is 9 Å (8.2 Å between O4 of O4MedT and O6 of dGMPNPP).</p><p>Closer inspection of the positions of the incoming nucleotides shows that adenine stacks on the adjacent t(emplate)A:p(rimer)T pair, with the side chain of Arg-61 from the finger domain hovering closely above the adenine plane and engaged in H-bonds to the α- and β-phosphates of dATP via its guanidino moiety (Fig. 3b). By comparison, guanine is shifted into the minor groove and the stacking interaction with the adjacent tA:pT pair is slightly less favorable. The shift is likely a consequence of the altered orientation of the Arg-61 side chain that is extended in the structure of the complex with dGMPNPP, resulting in formation of H-bonds between its guanidino moiety and guanine O6 and N7 (Fig. 3d). The particular orientation of the incoming dG brings it closer to Asn-38 from the finger domain, but the distance of 3.64 Å between N2 of the former and the Oε1 oxygen of asparagine is slightly too long for formation of a H-bond.</p><p>A surface rendering of the hPol η active site in the O4MedT insertion-stage complex with the incoming dATP indicates that the nucleobase moiety of the lesion fits snugly into the gap between Trp-64 and Ser-62 (Fig. S49†). It is clear that O4EtdT (with the ethyl group in the syn conformation) cannot be accommodated in the same fashion, as the longer substituent would clash with residues from the finger domain. Instead, the O4EtdT lesion has been pulled inside the active site and pairs with incoming dAMPNPP or dGMPNPP via bifurcated H-bonds in the two crystal structures of insertion-stage complexes with this lesion (Fig. 4a–d, PDB ID codes ; 5DQG and ; 5DQH, respectively). As in the complexes with O4MedT, Asn-38 forms a H-bond to O2 of O4EtdT in the minor groove. However, unlike in the O4MedT complex with incoming dGMPNPP, the side chain of Arg-61 in the corresponding complex with O4EtdT does not adopt an extended conformation to contact the major groove edge of guanine. As can be seen in Fig. 4, Arg-61 is directed toward the triphosphate moiety and forms a salt bridge with the α-phosphate group in both insertion-stage complexes. The methylene group (C1) of the O4 substituent in O4EtdT adopts an anti conformation in the complex with dAMPNPP (torsion angle C1–O4–C4–N3 = –142°) and a syn conformation in the complex with dGMPNPP (torsion angle C1–O4–C4–N3 = +62°). This is a clear difference to the structures of complexes with O4MedT, both of which show the lesion adopting a syn conformation (torsion angles C1–O4–C4–N3 of +33° and +25° in the dATP and dGMPNPP complexes, respectively). A further difference between the O4MedT and O4EtdT complexes is constituted by the orientations of template nucleotides 5′-adjacent to the lesions. In the former, A2 and T3 form a stack with Trp-42 outside the active site (Fig. 3). In the O4EtdT complexes, A2 is located outside the active site and forms a stacking interaction with Trp-42. However, T3 sits inside the active site and stacks onto O4EtdT (dAMPNPP complex; Fig. 4a). In the complex with dGMPNPP, T3 juts into the major groove (Fig. 4c). Thus, neither orientation adopted by T3 in these complexes resembles that of O4MedT, lodged near the ceiling of the active site and stacked onto Trp-64.</p><!><p>The structure of a complex with O4EtdT paired to dA at the –1 position followed by template dG opposite incoming dCTP was determined at 2.05 Å resolution (Fig. 4e and f, PDB ID codes ; 5DQI). The geometry of the O4EtdT:dA pair replicates that seen in the insertion complex with O4EtdT opposite incoming dAMPNPP (Fig. 4a, b, e and f). As in the case of the latter, the ethyl group has moved outside the thymine plane and adopts an anti orientation (torsion angle C1–O4–C4–N3 = –135°). The base pair itself adopts a Watson–Crick like geometry with formation of a single H-bond; the adjacent dG:dCTP pair displays a standard geometry with three H-bonds. Arg-61 is directed toward the phosphate moieties of the incoming nucleotide and forms two salt bridges with the α- and β-phosphate groups, and Asn-38 is engaged in two H-bonds with N3 and O4′ of template dG.</p><p>The most unusual feature of the extension-stage structure is the presence of an additional nucleotide at the 3′-end of the primer (Fig. 4f). Because the crystallization solutions contained dCTP and the residual electron density is consistent with a pyrimidine, we extended the primer by dC (Fig. S50†). We suspect that hPol η possesses weak catalytic activity with Ca2+ as the cofactor or that traces of Mg2+ present in the crystallizations led to primer extension in situ (even a very low activity could result in extension by a single nucleotide over the course of two weeks). We showed earlier that the translesion DNA polymerase Dpo4 from Sulfolobus solfataricus is able to catalyze nucleotide insertion with Ca2+, although the activity is far below that seen with Mg2+ as the prosthetic group.21 The additional dC stacks against the backbone of the template strand in the minor groove and its position is further stabilized by two H-bonds between N3 and O2 and the guanidino moiety of Arg-111 (Fig. 4f).</p><!><p>The known toxicity of alkylated adducts at the O4-position of thymidine prompted us to explore the influence of restricting orientation of the alkyl group around the C4–O4 bond to an anti conformation in translesion synthesis catalyzed by hPol η. In these studies, the bicyclic pyrimidine analogs DFP and TTP, which link the C5 and O4 atoms with a di- or trimethylene linker, were evaluated in addition to O4MedT and the bulkier O4EtdT lesion. Conformationally locked analogs of damage that can occur at the nucleobase have been previously synthesized and employed in studies which have provided insights into the requirements for DNA repair processes.22,23</p><p>UV thermal denaturation studies of oligomers containing DFP and TTP revealed similar influences on duplex stability to both complementary and mismatched nucleobases compared to O4MedT and O4EtdT. The most stable pairing of either DFP or TTP was with dG, also observed with O4MedT and O4EtdT. NMR studies of a duplex containing an O4MedT·dG pair revealed, in addition to the O4-methyl group adopting a syn-conformation, that the base formed a Watson–Crick "like" pairing with a single hydrogen bond.24 In this structure, the syn-orientation of the O4–Me group influences the hydrogen bond between the imino proton of dG and the N3 atom of O4MedT by increasing the distance between the O6 and O4 atoms of dG and O4MedT, respectively. Limiting the orientation of the methylene group at the O4-atom to the anti-conformation, in the case of the DFP and TTP modifications, appears to have a minimal impact on the interaction with dG and duplex stability. In pairing with dA, a similar drop in duplex stability compared with dT is observed for oligonucleotides containing the O4MedT, O4EtdT, DFP, and TTP modifications. The NMR structure of a duplex containing an O4MedT·dA pair indicated that the O4–Me group is syn and that the bases adopt a wobble alignment with one hydrogen bond formed between the imino nitrogen of O4MedT and the amino group of dA.15 The restricted anti-orientation of the methylene group for DFP or TPP modifications does not significantly impact duplex stability compared to O4MedT or O4EtdT.</p><p>Steady-state kinetics of individual nucleotide incorporation opposite the DFP and TTP modifications by hPol η demonstrated preferred insertion of purine nucleotides relative to the pyrimidines, similar to O4MedT and O4EtdT. The efficiency of nucleotide insertion (kcat/Km) for the correct nucleotide (dAMP) across the lesions followed the order DFP > TPP > O4MedT > O4EtdT. For dGMP, a similar efficiency of nucleotide insertion occurred for DFP, TPP, and O4MedT whereas a drop was observed for O4EtdT. In agreement with studies involving Saccharomyces cerevisiae DNA polymerase η (yPol η), a reduction in incorporation efficiency due to the presence of an O4MedT insert was observed.17 However, whereas the yeast homolog revealed a significant preference for dGMP, which was incorporated approximately 80 times more efficiently than dAMP,17 hPol η displayed almost equal selectivity at incorporating dAMP (f = 0.94) as dGMP opposite O4MedT. A comparable 80-fold preference for dGMP over dAMP was exhibited by yPol η for the bulkier O4-carboxymethylthymidine lesion.25 The rationale for the preferred incorporation of dGMP opposite O4MedT by yPol η was attributed to a dG·O4MedT wobble base pairing. Differences observed for nucleotide incorporation opposite O4MedT by the yeast and human homologs of Pol η may be influenced in part by different sequence contexts, as previously observed.5,26 In addition, homologs of Pol η have exhibited differences in nucleotide incorporation across some types of DNA damage. For example, yPol η accurately inserts dCMP across 8-oxodG whereas hPol η is less accurate, inserting some dAMP across this lesion as well.5,27 Interestingly, similar misinsertion profiles have been observed in bypass experiments of hPol η and yPol η with O6MedG, a lesion which also protrudes in the major groove of the DNA duplex.28</p><p>For the bulkier O4EtdT and conformationally restricted analogs DFP and TTP, a preference for nucleotide incorporation of dAMP over dGMP was observed. For O4EtdT, hPol η was more proficient at incorporating dAMP over dGMP with catalytic efficiencies of 0.10 and 0.06 μM–1 s–1, respectively. These values are approximately two-fold lower compared to those observed for the O4MedT-containing template, but can be rationalized by the increased bulk of the ethyl group, which may influence dNTP incorporation in the hPol η active site. Ethylation of the O4-position of dT has been shown to stall the human Y-family DNA polymerases hPol κ and hPol ι but not hPol η (although steady-state analysis was not reported for oligonucleotides containing O4EtdT).18 Bypass of O4EtdT by hPol η revealed dGMP misincorporation at 55% compared to 19% for dAMP insertion, in agreement with our data despite different sequence contexts.</p><p>For the conformationally restricted DFP and TPP modifications, incorporation efficiency was observed to be ∼1.5-fold higher for dAMP (0.27 and 0.24 μM–1 s–1, respectively) and comparable for dGMP relative to O4MedT. These results demonstrate that hPol η is more proficient at incorporating both the correct (dAMP) and incorrect (dGMP) nucleotides across from these more conformationally restricted lesions. In addition, the increase in steric bulk from the DFP to TPP slightly decreases incorporation efficiency. Exposure of the hydrogen bonding face of the DFP or TPP modifications may have a greater influence on stabilizing the wobble alignment geometry that has been suggested for the O4MedT·dA pairing. The conformational restriction of the alkyl group to an anti-orientation around the C4–O4 bond, as in the DFP and TPP modifications, would direct the O4-methylene group away from the amino group of dA, which could account for the enhancement of incorporation of the correct nucleotide (dAMP) compared to O4MedT. Incorporation of dGMP may not be as influenced by orientation of the alkyl group around the C4–O4 bond as the proposed hydrogen bonding interaction, based on the NMR structure of the duplex containing the O4MedT·dG, which occurs between the amino group of dG and O2-atom of O4MedT.24 In the case of O4EtdT, the combination of the syn-orientation and the size of the ethyl group may both contribute to the reduced efficiency of nucleotide insertion of dAMP and dGMP in this series.</p><p>Primer extension reactions in the presence of all four dNTPs for templates containing the O4MedT, O4EtdT, DFP, and TTP modifications demonstrated that hPol η was proficient at incorporating nucleotides across and past the adducted site. However, both O4MedT and O4EtdT exhibited a greater accumulation of non-full length oligonucleotide products at reduced reaction times (30 and 60 min), which was not observed for bicyclic DFP or TPP analogs. The LC-MS/MS analyses of the extension products from the in vitro primer bypass studies revealed that dGMP incorporation across the lesion was preferred over dAMP in all cases except the control (dT). The ratio of dGMP : dAMP incorporation by hPol η, assessed from the extension products, was found to decrease in the series TPP (4.6 : 1) > DFP (3.3 : 1) ≈ O4EtdT (3.2 : 1) > O4MedT (2.1 : 1). The presence of the larger alkyl group for O4EtdT or the analogs with the O4-methylene group in an anti-conformation (TPP and DFP) clearly promotes dGMP misincorporation in the presence of all four nucleotides. In addition, the DFP and TPP modification were not found to induce a significant amount of frameshifts in the products compared to O4MedT. Differences in fidelity observed between the steady-state kinetic and LC-MS/MS full-length experiments have been observed previously.17,19 The variance may be attributed to accommodation of the incoming dGTP relative to dATP for these modifications, highlighting that adduct size and the conformation of the O4-methylene group can influence interactions in the active site of hPol η. It should be noted, however, that other steric and/or stereoelectronic effects may have an impact on hPol η bypass processivity of the conformationally locked analogues relative to O4MedT or O4EtdT, respectively. In the case of the analogs investigated, hPol η continued extension of the primer in an error-free manner after incorporation of dATP or dGTP across from the damaged site on the template.</p><p>Several observations based on nucleotide incorporation profiles attest to the distinct effects on hPol η bypass synthesis exerted by the O4MedT and O4EtdT lesions. These concern (i) the more error-prone bypass caused by O4MedT, i.e. dGTP is favored relative to dATP (Fig. 2a), (ii) increased accumulation of the +1 product in the full-length extension reaction for O4MedT (Fig. 1), and (iii) significantly more frameshift products caused by the O4MedT lesion (Fig. 2b). Interestingly, the structural data for insertion-stage hPol η complexes with either O4MedT or O4EtdT in the template strand reveal starkly different orientations of the two adducted nucleotides at the active site. O4MedT is trapped in an orientation that keeps it at a considerable distance from the incoming purine nucleotide triphosphates. Conversely, O4EtdT pairs opposite both dATP and dGTP with formation of bifurcated H-bonds (whereby the latter pair features a sheared orientation of the two partners, with G being pushed toward the minor groove). The increased proclivity for insertion of dG opposite O4MedT compared to O4EtdT is not surprising if one considers the strict preference by the O4-methyl substituent for a syn conformation. The syn conformation precludes adoption of an O4MedT:dA pair with standard Watson–Crick geometry, but the sheared pairing mode seen in the case of O4EtdT:dG(MPNPP) (Fig. 4c and d), also presumably adopted by the O4MedT:dG pair, is compatible with a syn conformation of the substituent. This conclusion is borne out by the observations that the ethyl moiety in the O4EtdT:dA(MPNPP) pairs assumes an anti conformation (Fig. 4a, b, e and f), whereas its conformation is syn in the O4EtdT:dGMPNPP pair. Furthermore, the TPP adduct opposite dGMPNPP was modeled from the O4EtdT:dG(MPNPP) ternary crystal structure coordinates. The configuration of the adduct seen in the model is consistent with the enhanced incorporation of dG observed in the primer extension experiments since the constrained anti conformation of the bicyclic system does not hinder the guanine nucleobase from shifting towards the major groove and potentially form two H-bonds with TPP (Fig. S51†).</p><p>On one hand, one could argue that the higher fraction of frameshifts for O4MedT relative to O4EtdT is consistent with the structural data that show the former is not engaged opposite the incoming nucleotide but trapped adjacent to the 'entrance' of the active site. Perhaps the more pronounced accumulation of the +1 product in the case of the full-length extension reactions opposite O4MedT compared to the other O4 adducts tested here are the result of non-templated insertion. Thus, purine nucleoside triphosphates would be favored and their incorporation would not be affected by the particular conformation of the O4-methyl group, syn or anti. This scenario is certainly not inconsistent with the structural data that reveal no interaction between the O4MedT lesion and the incoming dATP or dGMPNPP. Clearly, it is intriguing that both activity and structural data show distinct consequences of the O4MedT and O4EtdT lesions for bypass by hPol η. However, it is important to note that the position of the O4MedT lesion at the active site, unique among all crystal structures of hPol η complexes analyzed to date, represents one state during bypass. Perhaps other orientations and interactions of the adducted nucleotide occur during bypass, which precludes all steps involved in the mechanism of O4-alkyl bypass synthesis by hPol η.</p><p>All results from the study (kinetic evaluation, full extension assays and crystal structures) may be integrated into one potential extension model. The crystal structure supports the notion that the O4MedT nucleobase is indeed nestled at the ceiling for an undefined period of time. The purine nucleoside triphosphates could then be imported into the active site, and would be subject to a template gap, analogous to being opposite an abasic site. According to full extension assays, there are approximately twice as many dG inserted by hPol η relative to dA. The lack of O4MedT – incoming dNTP clash in the active site may explain higher kp values observed for dATP and dGTP for O4MedT compared to O4EtdT (by 8.8 or 9.3 fold respectively). This may also aid in explaining the preference for purine insertion, in an approximate 2 : 1 dG : dA ratio across O4MedT, whereas a larger ratio is observed for all the other modifications as discussed above. Perhaps the other modifications prevent the modified nucleobases from accessing the conformation at the top of the active site. The ethyl group, albeit bulkier, can populate the other conformations (syn versus anti) as observed in the crystallographic data, which may contribute to the higher correct dA insertion relative to the bicyclic analogues which are locked in an anti conformation. Increase in bulk (from O4MedT to O4EtdT and DFP to TPP) leads to an increase of incorrect dG insertion. Eventually the O4MedT is required to move from the top of the active site back to the 0 or –1 (post-replicative) position(s). It is possible that this mobility causes the frameshift adduct formation observed in the full extension assays. As a result, we have provided intriguing insights on a potentially different bypass mechanism of O4MedT in comparison to the larger O4EtdT adduct.</p><!><p>Oligonucleotides containing DFP and TPP, designed as analogs of O4-alkylated thymidine, were synthesized to explore the influence of limiting the O4-alkyl lesion to an anti-orientation on nucleotide incorporation by hPol η. These modifications were shown to destabilize the DNA duplex, based on UV thermal denaturation studies, regardless of the base-pairing partner (A, G, T, or C), similar to O4MedT and O4EtdT. Primer extension assays demonstrated that these pyrimidyl modifications hindered nucleotide incorporation by hPol η. Single nucleotide incorporation studies revealed increased selectivity towards dAMP over dGMP that followed the order O4EtdT > DFP ≈ TPP. A slight preference for dGMP over dAMP incorporation was observed for O4MedT. LC-MS/MS analysis of primer extension studies (in the presence of all four dNTPs) revealed that hPol η incorporated dGMP over dAMP across the lesions in the order TPP > DFP ≈ O4EtdT > O4MedT. These trends suggest that limiting the orientation of the O4-alkylene group enhances the proficiency of dNTP incorporation by hPol η across O4-alkylated dT damage. In the presence of all four dNTPs, error-prone nucleotide incorporation by hPol η is enhanced by restricting the O4-lesion to an anti-orientation. This study exemplifies how restricting a lesion's conformational freedom impacts bypass profile by hPol η. Moreover, our results provide mechanistic insights into the mutagenicity of the biologically relevant O4MedT and O4EtdT DNA adducts.</p><!><p>Click here for additional data file.</p>

PubMed Open Access

Structure-activity relationship study of pyridazine derivatives as glutamate transporter EAAT2 activators

Excitatory amino acid transporter 2 (EAAT2) is the major glutamate transporter and functions to remove glutamate from synapses. A thiopyridazine derivative has been found to increase EAAT2 protein levels in astrocytes. A structure-activity relationship study revealed that several components of the molecule were required for activity, such as the thioether and pyridazine. Modification of the benzylthioether resulted in several derivatives (7\xe2\x80\x9313, 7\xe2\x80\x9315 and 7\xe2\x80\x9317) that enhanced EAAT2 levels by > 6 fold at concentrations < 5 \xce\xbcM after 24 h. In addition, one of the derivatives (7\xe2\x80\x9322) enhanced EAAT2 levels 3.5 \xe2\x80\x93 3.9 fold after 24 h with an EC50 of 0.5 \xce\xbcM.

structure-activity_relationship_study_of_pyridazine_derivatives_as_glutamate_transporter_eaat2_activ

<p>Glutamate is a major neurotransmitter in the mammalian central nervous system (CNS) and essential for normal brain function including cognition, memory, and learning. However, the extracellular concentration of glutamate must remain below excitotoxic levels (~ 1 μM) to avoid overstimulation of glutamate receptors, leading to neuronal damage or death.1 Excitotoxicity has been associated with multiple acute neurological conditions such as ischemic stroke, epilepsy, and trauma, as well as chronic adult-onset neurodegenerative disorders such as Alzheimer's disease and amyotrophic lateral sclerosis (ALS).2–6 One potential approach to preventing excitotoxicity is to enhance glutamate reuptake. Excitatory amino acid transporter 2 (EAAT2) is the major glutamate transporter and functions to remove glutamate from synapses.7 An increase in EAAT2 protein expression and function may provide a means to prevent insufficient glutamate reuptake and consequently reduce neuronal damage.</p><p> </p><p>In an effort to identify small molecules that can increase EAAT2 protein expression, a high-throughput screen of approximately 140,000 compounds was previously conducted in our laboratories using a cell-based enzyme-linked immunosorbent assay.8 The hits identified from this screening provide starting points for further optimization in order to arrive at pharmacologically useful molecules for studying the role of EAAT2 in excitotoxicity induced neuronal injury and potentially as therapeutic agents.</p><p>The thiopyridazine 1 was confirmed to show a dose-dependent increase in EAAT2 protein levels after 24 h exposure. Herein, we report the structure-activity relationship (SAR) study of 1 for elevating EAAT2 protein levels.</p><p>Many of the pyridazine analogues utilized in the SAR study were prepared using the method depicted in Scheme 1. Treatment of ketone 2 with glyoxylic acid (3) and K2CO3 gave 4, which was used directly without purification. It was allowed to react with hydrazine in acetic acid at 100 °C to yield the desired pyridazinone 5 as an off-white solid after recrystallization from ethyl acetate.9 Direct alkylation of 5 gave 8 in good yields. Intermediate 5 was also converted into pyridazinethione 6 in the presence of P2S5 in pyridine at 120 °C.10 Alkylation of 6 provided 7, which could be subsequently oxidized to sulfone 9 with 3-chloroperoxybenzoic acid (m-CPBA) in CH2Cl2.11</p><p>A series of additional analogues 14–17 was prepared using the methodology outlined in Scheme 2. Recently, 2-pyridyl N-methyliminodiacetic acid (MIDA) boronate (10) has been reported as an air-stable slow-release reagent with high cross-coupling efficiency even with heteroaryl chlorides.12 Therefore, this material was used in palladium-mediated cross-couple reactions with heteroaryl chlorides 11 – 13 to obtain 14 – 16, respectively. In the case of 13 transesterification occurred during the cross-coupling reaction to give the isopropyl ester 16, which was then allowed to react with alkylamines in ethanol at 85 °C to generate amides 17.</p><p>Finally, several additional analogues designed to evaluate replacement of the pyridazine in 1 with phenyl and pyridyl moieties were prepared using the methodology outlined in Scheme 3. Suzuki coupling of 2-bromopyridine (18) with various boronic acids afforded 19 and 20.13 Treatment of 20 with P2S5 in pyridine at 120 °C provided 21 in 71% yield. Alkylation of 21 with 2-methylbenzyl bromide in the presence of K2CO3 in DMF generated 22.</p><p>All of the derivatives of 1 were initially evaluated in PA-EAAT2 cells3 (a primary astrocyte line stably expressing EAAT2 mRNAs) following compound (10 μM) incubation for 4 and 24 h before harvesting and measuring EAAT2 levels by Western blot analysis. The fold increases in EAAT2 protein levels relative to DMSO controls are reported (Tables 1 – 4).</p><p>Replacement of the 2-pyridyl with 3-pyridyl (7–1, 7–5 and 7–8), 4-pyridyl (7–2, 7–9 and 7–11) or phenyl (7–6) resulted in equal or reduced activity (Table 1). Removing either one or both of the nitrogen atoms in the pyridazine resulted in loss of activity (Table 2). Collectively, these results suggested that both the 2-pyridyl and the pyridazine were required for enhancing EAAT2 protein expression.</p><p>Next, the sulfur linker was examined (Table 3). Replacing the sulfur with oxygen generally yielded less active derivatives. An analogue containing a NH linker (14) showed weaker activity. Likewise, oxidation of the sulfur to a sulfone (9–1, 9–2, and 9–3) was also detrimental. Finally, replacing the sulfur with an amide moiety (17–1 and 17–2) was not tolerated. Collectively, these results indicated that the sulfur linker was optimal.</p><p>Finally, the benzyl group was examined (Table 4). Compared to 1, 2-chloro, 3-chloro and 2, 6-dichloro substitutions improved the potency by two-fold (7–7, 7–18 and 7–22), but 2, 3-dichloro, 2,4-dichloro and 2, 5-dichloro analogues (7–19, 7–21 and 7–20) were equivalent to 1. In the case of fluorine substitutes, only the 2-fluoro and 2, 4-difluoro derivatives (7–23 and 7–16) demonstrated improved activity. 4-Fluoro, 2, 6-difluoro, and 2, 4, 6-trifluoro analogues (7–24, 7–25, and 2–26) did not result in significant improvement. Replacement of the halogens with methyls (7–4, 7–13, 7–17, and 7–30) gave increased potency. Compounds containing 2-methyl substituted benzyl groups significantly increased EAAT2 protein level, where as the 3- or 4-methylbenzyl derivatives (7–27 and 7–28) were less potent. Increasing the tether length between the phenyl and the sulfur to an ethylene generally yielded more potent analogues (7–15 and 7–31 verses 1 and 7–7, respectively) with one noted exception (compare 7–32 to 7–4). However, truncating the linker resulting in a diarylthioether decreased potency (7–29 verses 7–17). Interestingly, adding substitutes on the carbon linker (7–14 and 7–33) increased activity about 3-fold. Finally, other changes gave compounds with essentially the same potency as 1 (7–12, 7–36, and 6–1).</p><p>In order to further characterize twelve of the most active analogues from the initial assessment, they were evaluated in a 6-point dose-response assay (0.1, 0.3, 1.0, 5.0, 10.0, and 30.0 μM). The fold increases at 1 and 10 μM, as well as the EC50 values are summarized in Table 5. In addition, these values were determined relative to two unrelated proteins (e.g. β-actin and GAPDH). Several compounds (7–13, 7–17, 7–31 and 7–18) were able to increase EAAT2 levels > 6-fold at 10 μM compared to β-actin or GAPDH, but with EC50 values in the 2.6 – 3.3 μM range. Analog 7–22 demonstrated the lowest EC50 of 0.5 μM with a maximum increase of EAAT2 levels of 3- to 4-fold.</p><p>In summary, a series of pyridazine derivatives were synthesized and evaluated for increasing EAAT2 protein levels. Several compounds were found to significantly increase EAAT2 protein levels (> 6 fold), such as 7–13, 7–15 and 7–17. Derivative 7–22 increased EAAT2 protein levels to a lower extent (3.5 – 3.9 fold), but with a lower EC50 value (0.5 μM). These compounds will provide useful tools for further assessing the role of glutamate excitotoxicity in cellular systems and potentially in animal models of acute and chronic neurodegeneration. In addition, these probe molecules may also be beneficial in determining the biological mechanisms for regulating EAAT2 levels.</p>

PubMed Author Manuscript

Direct Observation of the Oxidation of DNA Bases by Phosphate Radical Formed under Radiation: A Model of Backbone-to-base Hole Transfer

In irradiated DNA, by base-to-base and backbone-to-base hole transfer processes, the hole (i.e., the unpaired spin) localizes on the most electropositive base, guanine. Phosphate radicals formed via ionization events in DNA-backbone must play an important role in the backbone-to-base hole transfer process. However, earlier works on irradiated hydrated DNA, on irradiated DNA-models in frozen aqueous solution and in neat dimethyl phosphate showed formation of carbon-centered radicals and not phosphate radicals. Therefore, to model backbone-to-base hole transfer process, we report picosecond pulse radiolysis studies of the reactions between H2PO4\xe2\x80\xa2 with DNA bases \xe2\x80\x93 G, A, T, and C in 6 M H3PO4 at 22 \xc2\xb0C. The time-resolved observations show that in 6 M H3PO4, H2PO4\xe2\x80\xa2 causes one-electron oxidation of adenine, guanine and thymine, by forming the cation radicals via single electron transfer (SET) process; but, the rate constant of the reaction of H2PO4\xe2\x80\xa2 with cytosine is too low (< 107 L mol\xe2\x88\x921 s\xe2\x88\x921) to be measured. The rates of these reactions are influenced by protonation states and reorganization energies of base radicals and of the phosphate radical in 6 M H3PO4.

direct_observation_of_the_oxidation_of_dna_bases_by_phosphate_radical_formed_under_radiation:_a_mode

Introduction<!>Compounds<!>Pulse radiolysis set-up<!>Data analysis method<!>Theoretical Calculations<!>Results and Discussion<!>1. Phosphate radical (H2PO4\xe2\x80\xa2) formation in 6 M H3PO4 solution<!>2. Guanine Oxidation by H2PO4\xe2\x80\xa2<!>3. Adenine Oxidation and Thymine Oxidation/Addition by H2PO4\xe2\x80\xa2<!>4. The Cytosine case<!>Conclusions<!><!>Conclusions

<p>The initial chemical events of cellular DNA damage exposed to ionizing radiation (γ- or X-ray photons, fast electrons, heavy ions, etc.) have been classified as: (i) indirect effect induced by the water-derived species, e.g., hydroxyl radical (•OH), water cation radical (H2O•+), and low-energy electrons (LEEs), generated from energy deposited in the outer water shells surrounding the DNA. (ii) Direct-type effect that corresponds to the ionization and excitation of DNA components due to direct deposition of energy onto the DNA molecule. As a result of these excitations and ionization events occurring in the water molecules of first hydration layer, H2O•+ and electrons are formed; H2O•+ and the ejected electrons undergo both charge and spin transfer to the DNA (quasi-direct effect).1–4 In addition, a considerable number of low energy electrons (LEE), are generated with excess kinetic energy of 0 to 20 eV, that damage DNA.5,6</p><p>The DNA backbone consists of alternating phosphate and sugar residues with high mass fraction (0.6). Therefore, on the basis of the number of valence electrons, the hydrated sugar-phosphate moiety can be significantly damaged either by the radiation-induced ionization-mediated one-electron oxidation events7 or via indirect one-electron oxidation due to hole transfer from H2O•+ formed in the hydration shell.4</p><p>Recent electron spin resonance (ESR) studies on radiation-mediated direct ionization of dimethyl phosphate and on γ-irradiated, as well as ion-beam irradiated, DNA showed that for sugar radical (e.g., C5′•) formation, a very rapid (< 10−12 s) deprotonation must occur from the one-electron oxidized sugar-phosphate backbone prior to the competitive backbone-to-base hole transfer.4,7–10 Laser flash photolysis studies provided evidence that ionization of organic phosphates, such as ribose-5-phosphate, could produce the oxygen-centered primary phosphate radical, O3PO•. Owing to the very oxidizing nature of O3PO•, this radical undergoes fast intramolecular H-abstraction from the C4 of the pentoxyribose ring with rate constants up to 5 ×107s−1.11 Therefore, if the lifetime of the hole in the sugar-phosphate backbone could be increased, the backbone-to-base hole transfer process could be directly probed using ESR spectroscopy. Via substitution of the phosphate by the easily oxidizable phosphorothioate group at specific phosphate sites in the DNA backbone, this goal has been achieved.8 The phosphorothioate-centered radicals have been shown to be formed in appreciable yields at the picosecond timescale, and subsequently lead to the fast one-electron oxidation of the proximate guanine base (i.e., backbone-to-base hole transfer) by temperature-activated hopping.</p><p>ESR spectroscopic studies did provide the evidence of phosphorus-centered radical by its large 31P hyperfine couplings.12,13 The phosphorous-centered radical was found to be produced in significantly reduced quantities at 77 K in hydrated (⌈(no. of water molecules per nucleotide) = 12) irradiated (temp. at irradiation = 77 K) DNA samples. These phosphorus-centered radicals were formed owing to the radiation-produced LEE-mediated fragmentation of the P-O bond in the sugar-phosphate backbone via the dissociative electron attachment (DEA) mechanism.9</p><p>From the viewpoint of backbone-to-base hole transfer and of carbon-centered sugar radical formation via the phosphate radical intermediate, it is therefore rationalized that reactivity of phosphate-centered radicals towards nucleobase sites would play an important role in radiation-induced DNA damage. However, apart from the work on S-oligomers, the role of phosphate radical in backbone-to-base hole transfer has not been sufficiently investigated to date. This might be due to the following reasons: (a) the LEE-mediated phosphorus-centered radical yield has been observed to be only about 0.1% of the total DNA-radicals present in ion-beam irradiated DNA;4,7,13 (b) phosphate radicals are difficult to produce by radiation-induced reactive species, such as •OH. In fact, pulse radiolysis studies of the reaction of •OH with alkyl phosphates (a model of sugar-phosphate backbone) showed formation of only C-centered phosphatoalkyl radicals;14 (c) a sufficient concentration of the phosphate radicals has to be achieved to carry out the investigation of the rate and the extent of the reaction between phosphate radicals and a nucleobase (i.e., backbone-to-base hole transfer).</p><p>The reasons (b) and (c) become particularly important in aqueous solutions under ambient conditions owing to various factors. First, the value of the rate constant of •OH with phosphate anions to form the phosphate radicals is found to be very low (~105 L mol−1 s−1)15,16; on the other hand, the reaction of •OH with the nucleobases being diffusion-controlled,1,2 •OH predominantly reacts with the base moieties in DNA, in preference to the phosphates in the DNA-backbone. Secondly, in contrast to the formation of an oxidizing radical, e.g., SO4•−, via reaction of the radiation-produced electron with K2S2O8, generation of phosphate radicals via electron addition to K2P2O8 is not feasible. Moreover, photo-dissociation of peroxydiphosphate (e.g., K2P2O8) in the presence of DNA-bases in aqueous solution at ambient temperature is not a suitable approach to study the reactions of phosphate radical with DNA-bases because K2P2O8 being a very strong oxidant, would lead to the decomposition of the nucleobases during experiment.</p><p>To overcome the above-mentioned challenges and to observe the time-resolved reactivities of phosphate radicals with the individual DNA bases at room temperature, the present work introduces highly concentrated phosphoric acid (6 M) as a novel medium, based on our recent picosecond pulse radiolysis studies of direct and indirect effects of radiation.17,18 In addition, our studies in highly concentrated phosphoric acid could better represent the relevant molecularly crowded environments of localized hydrated DNA in nuclei than in dilute aqueous solutions. As an example, in concentrated phosphoric acid solutions (> 2 mol L−1), the phosphate radical, H2PO4•, with an absorption maximum at 520 nm was clearly observed; its high radiolytic yield (~ 2 × 10−7 mol J−1) is attributed to two ultrafast (within 7 ps electron pulse) parallel reaction pathways – (a) direct ionization of the phosphonate solutes, and, (b) electron transfer from water hole or water cation radical (H2O•+) in close contact.19 The lifetime of H2PO4• is on the timescale of a few microsecond and its decay is mainly due to the reaction with H•.19 In addition, the precursor of hydrated electron (pre-solvated electron), which is proposed to attack DNA molecules via dissociative attachment, is completely scavenged by H3O+ ions within the pulse (7 ps) when the concentration of H3PO4 is above 4 mol L−1.20 Meanwhile, the H2O•+, precursor of •OH, is also partially quenched through electron transfer.19 The side reactions stemming from pre-solvated electron and from solvated electron towards biomolecules, for example, nucleobases, are suppressed. Therefore, during the high-energy electron pulse radiolysis of 6 M phosphoric acid, the sufficient number of phosphoric groups produces optically detectable H2PO4•, and allows us to gain new insights of the rate and extent of one-electron oxidation of DNA bases - viz., thymine, guanine, adenine and cytosine, by H2PO4• in aqueous solutions at ambient temperature. The rates of these reactions and the scaling of these rates can be accounted for by the extents of protonation of the DNA bases, as well as by the reorganization energies of the base radicals and H2PO4• in 6 M H3PO4. These experimental studies, in combination with theoretical investigations, provide valuable insight on the backbone-to-base hole transfer process in irradiated DNA.</p><!><p>The aqueous solutions were prepared using ultrapure water from an Elga system. The chemicals - guanine, thymine, cytosine, adenine, and phosphoric acid, were obtained from Sigma-Aldrich (purity > 99%) and were used as received (i.e., without further purification). The solutions were saturated with oxygen (O2, purity > 99.995% from Air Liquide) by bubbling before and during experiments. The solutions were irradiated at room temperature (21 ± 0.2°C).</p><!><p>Pulse radiolysis experiments were carried out employing the picosecond laser-triggered electron accelerator, ELYSE, coupled with a time-resolved absorption spectrophotometric detection system.21, 22 Electron pulses with 7.5 MeV energy, up to ~6 nC charge, and 15 ps duration were delivered at a repetition rate of 5 Hz. The sample was contained in a fused silica cell with a path length of 10 mm connected to a closed circulation system from a 150 mL Ar-purged stock solution used to renew the sample in the irradiation cell, after each pulse, by means of a peristaltic pump (flow rate 100 mL/min). The diameter of the electron beam was 3 mm, and the irradiated volume was less than 0.1 mL.</p><p>Absorption spectral measurements were performed using the original detection setup that we developed a few years ago. This setup is composed of a white light beam produced by a homemade xenon flash lamp, focused through the sample collinearly with the electron beam, with a diameter smaller than the electron beam. It is then directed onto a flat field spectrograph (Chromex 250IS), which disperses the light on the entrance optics of a high dynamic range streak-camera (model C-7700-01 from Hamamatsu). This setup allows us to record the variation of the intensity of the analyzing light with a resolution time of 4 ps using the 500 ps to 1 ms sweep time/full scale in the wavelength range from 250 to 850 nm. Data are collected in the form of images with 1024 pixels on the time axis and 1344 pixels on the wavelength axis.</p><p>The flash lamp was synchronized with the electron pulse; as a result, an image resolved in wavelength and time was acquired. To improve the signal-to-noise ratio, series of 600 images were acquired and were then averaged in a single image used for the calculation of the absorbance by reference to the similar image obtained on the averaging of 400 flashes of the analyzing light without the electron pulse. The kinetic data and absorption spectra were both extracted from the same series of resulting images. In this work, the transient spectra were measured from 290 to 650 nm and from 40 ps to 10 μs.</p><!><p>The data matrices images are analyzed by a multivariate curve resolution alternating least-squares (MCR-ALS) approach. The spectra of the absorbing species have been found to strongly overlap in this type of system; as a result, it becomes difficult to deconvolute individual spectra. Therefore, a global data analysis approach was used. Global matrices were built by delay-wise binding matrices for 6 M phosphoric acid with the data matrices for different time resolutions and concentrations. In this configuration, the individual spectra of various components are common to all of the experiments in the global data set, while each experiment has its own set of kinetic traces.</p><p>The number of distinguishable compounds in the global matrix was assessed by singular value decomposition (SVD). Subsequently, the MCR-ALS analysis with the corresponding number of species was performed. Positivity constraints were imposed for all spectra and kinetics. On the basis of quantum mechanical information, unimodality constraints can also be imposed to the spectra. Residuals maps were inspected to assess the absence of model inadequacy. All calculations have been made with codes developed in-house in the R environment. Our MCR-ALS code was adapted from the ALS package by K. M. Mullen.23</p><!><p>The geometries of H2PO4• complexed with guanine (as a model of purine base) and cytosine (as a model of pyrimidine base) were fully optimized using the B3LYP/6-31++G** method as well as including the complete effect of aqueous solution via the integral equation formalism of the polarized continuum model (IEF-PCM) of Tomasi et al.24,25 The complete methodology is herein designated as B3LYP-PCM/6-31++G**. All these calculations are performed employing the Gaussian 09 suite of programs (see supporting information).</p><!><p>High-energy electron pulse radiolytic measurements were performed in 6 M H3PO4 solutions in the presence of varying concentrations (0.25 to 20 mM) of DNA bases. The radiation-induced reactions taking place in these conditions are presented in Table 1. The solutions have been saturated by continuous O2 bubbling for converting H• to HO2• (R7); HO2• has been shown to be ineffective to nucleobases on the timescale of several microseconds.3</p><!><p>In the absence of DNA base, the kinetics of formation and decay of phosphate radicals as a function of H3PO4 concentrations (3 to 14 M, Table 2) are compared in Figure 1. It is evident from Figure 1 that for a given absorbed dose per pulse, the initial absorbance slightly increases with the rise of H3PO4 concentration. The slow decay observed in 6–14 M phosphoric acid is mainly attributed to the reactions of phosphate radicals with H• and HO2• (R10, R13, Table 1). At a relatively lower concentration of H3PO4 (3 M), the •OH produced via proton transfer from H2O•+ (R6) is not scavenged by H3PO4 and leads to the slight rise of phosphate radical via reactions R8 and R9. Aqueous H3PO4 at 6 M was selected as a "reference" medium for several reasons (Table 2). First, it constitutes a high solute electron fraction (fs = 0.43) environment in which direct ionization (R1) and hole transfer (R2) give rise to a significant yield (~1.5 × 10−7 mol J−1 at 100 ns) of H2PO4•.19 The yield of this radical is larger than that of the remaining OH• in the system. Secondly, the electrons either in their pre-hydrated or hydrated states are completely scavenged by H3O+ at tens of picoseconds (R3 and R4).20 Thirdly, for 6 M H3PO4, plenty of water molecules are still available for biologically relevant reactions and are sufficient for dissolving the DNA bases.</p><p>In 6 M H3PO4 solution, the phosphate radical exists in its acidic form (H2PO4•). This conclusion is based on the fact that the transient spectrum showing the absorption maximum at 520 nm in the pulse radiolysis study of phosphoric acid (see Figure 1) matches to the one already reported and assigned to H2PO4•.16,26 Previous studies by Katsumura et al.16 provided an estimate of the standard potential of this radical as high as E0 (H2PO4•, H+/H3PO4) = 2.65 V (vs. NHE) on the basis of E0 (•OH, H+/H2O) = 2.72 V. This standard potential value of H2PO4• is as high as those of other strong one-electron oxidants, e.g., sulfate anion radical (SO4•−), which has a standard potential of E0 = 2.43 V vs. NHE.27</p><p>Concerning nucleobases, nucleosides and nucleotides, they are rapidly oxidized by sulfate anion radical (SO4•−) with rate constants nearly at the diffusion-controlled limit.27 The carbonate and dibromide anion radicals, CO3•− and Br2•−, are weaker oxidants than SO4•−;3,28 CO3•− and Br2•− are known to oxidize guanine out of all four DNA bases.28 Based on these observations, H2PO4•, which is a stronger oxidizing agent is predicted to be thermodynamically capable to cause the direct one-electron oxidation of all nucleobases (E° from 2.03 to 2.5 eV) in aqueous solutions.</p><!><p>Guanine has the lowest reduction standard potential among the four nucleobases (A, T, G and C). The holes (i.e., unpaired spins) created by one-electron oxidation or by ionization events in irradiated DNA are eventually transferred to the guanine moiety, thereby producing transient guanyl radicals (guanine cation radical, G•+, or its neutral conjugate base, (G(−H+)•). Therefore, the structure and reactivity of these transient guanyl radicals, as the initial step in the oxidative damage of DNA, have been a center of interest ever since the early days of radiation chemistry and biology.1–4, 29, 30 Guanine molecule is a derivative of purine; solubility of guanine in water is intrinsically low (< 10−5 M), acting as the keto-9NH guanine tautomer moiety.31 Three pKa values of guanine have been reported by experiments and by theoretical calculations as, 12.4, 9.4, and 3.3 corresponding to N9, N1, and N7, respectively.32 In highly concentrated H3PO4 solutions (e.g., 6 M), the predominant species are monomer H3PO4 and dimer H5P2O8− ions, as well as free H3O+ ion33 (Table 2).</p><p>At such a low pH, our steady-state UV-visible spectral measurements show that the guanine molecule in phosphoric acid is in the fully protonated form and the complex, denoted (G···H3PO4), is formed via H-bonding. In addition, the presence of phosphoric acid in water allows guanine to be dissolved in solution; furthermore, phosphoric acid does not cause degradation of guanine for at least over one week as found from UV-visible spectral studies. For the other three bases, concerns of their stability in concentrated H3PO4 solutions have been similarly examined and have been found to be similar to the above-mentioned results of guanine. More information on the chemical stability and structure of DNA bases in aqueous phosphoric acid are presented in SI (Figure SI1–Figure SI5).</p><p>In case of time-resolved spectroscopic measurements of H3PO4 in the presence of guanine induced by electron pulse, the evidence of reaction of H2PO4• with guanine molecule giving rise to a new transient absorbing species is presented in Figure 2. Within the electron pulse (7 ps) the observed typical broad absorption spectrum with a maximum at 520 nm has suggested the immediate formation of H2PO4• 0.1 μs after the pulse, the formation of additional radical(s) is evident from the appearance of three additional bands with one sharp peak located at 305 nm and two less intense ones - at 370 nm and 520 nm, respectively (see Figure 2, top and middle). The kinetics of formation of this intermediate is accelerated by increasing the guanine concentration (Figure 2, bottom). Moreover, the amount of radical observed at 300 nm increases by increasing guanine concentration, meaning that phosphate radical undergoes a number of competitive reactions (R10, R11 and R12 in competition with R16) and by increasing the guanine concentration, the oxidation reaction of guanine by phosphate radical is favored. Furthermore, the transient kinetics observed at 520 nm in which H2PO4• has its maximal absorption, is found to differ much from those in the UV region (300 nm–375 nm) and is observed to level off over the 2 μs range (Figure 2, top).</p><p>The spectroscopic features of this transient species are in good agreement with that of the cation radical, G+•, formed after oxidation by SO4•− and Br2•− in aqueous solution by pulse radiolysis at room temperature, as well as by one-electron oxidation of the nucleoside dGuo in supercooled homogeneous solutions at low temperatures. 4,10,29,34,35 At neutral pH, it is well-accepted that the formation of G+• in aqueous solutions is followed by a rapid deprotonation (1.8 × 107 s−1) of the N1-H proton, i.e., the resulting radical is uncharged G (−H+)•. Optical and conductance studies at ambient temperature allowed estimation of the pKa value of G+• of dGuo to be 3.9.3,34 So, in basic medium, deprotonation of G+• is characterized by a much pronounced change in the spectrum occurring at higher wavelength range (λ > 600 nm).4,29 The rapid formation of G(−H+)• was also spectrophotometrically observed in single stranded and in duplex DNA by Kobayashi et al.10(a) In their work, absorbance at 650 nm which is referred to the typical electronic difference between G+• and G(−H+)•, is verified to be time-evolved and red-shifted. In 6 M H3PO4, guanine exists as N-7 protonated guanine as well as N-9 protonated guanine; upon one-electron oxidation, either the N-7 protonated or the N-9 protonated guanine cation radical (G(N7H)2+• or G(N9H)2+•) undergo immediate deprotonation from N-7 or N-9 thereby forming the "pristine" G+• (scheme 1 and vide supra).</p><p>From the existing literature, as well as under the highly acidic condition of our system (6 mol L−1 H3PO4), and based on our theoretical calculations, the pKa value of N7-protonated guanyl cation radical is predicted to be in the negative region.36 Our results further establish that guanyl radical formed under highly acidic conditions of our system, exists predominantly in its cation radical state with the proton at N1 (G+•) instead of its N1-deprotonated state, G(−H+)• (vide infra).</p><p>As the spectra of various radical intermediates (e.g., H2PO4•, G+•) overlap with each other (e.g., see Figures 2 (Middle), 3 (Top)), MCR-ALS analyses have been performed subsequently to clarify the number of absorbing species and their corresponding spectra as well as their reaction pattern. Figure 3 (top) presents an example of the two-dimensional transient absorption spectra measured in 1 mmol L−1 guanine/6 mol L−1 H3PO4. Only two distinguishable transient absorbing species are sorted out by MCR-ALS analysis of the matrix spectral image. We have observed similarities of UV-Visible spectra in Figure 3 with those of the "pristine" G+• that has already been reported in the literature.29, 34 – 36 This can be explained by the rapid deprotonation of G(N7H)2+• or G(N9H)2+• already mentioned. None of the spectra in Figures 2 and 3 show any absorption peak of hydroxylated guanyl radicals that are formed due to addition of •OH to guanine. For instance, the guanyl C4-OH adduct radical was characterized by a peak at 330 nm37, and the guanyl C8-OH adduct radical should show a spectral feature at 610 nm.38 Therefore, the remaining •OH, that is involved in the oxidation of G via addition-elimination process,3,34,36–38 leads to the rapid formation of the same radical, G+•. The rate of this •OH-mediated formation of G+• has to be faster than that the G•+ production via H2PO4•; •OH-mediated formation of G•+ is completed in less than 100 ns.37 The correlated dynamic features of G•+ and H2PO4• are presented in Figure 3 (bottom). The additional formation of G+• after 100 ns is clearly observed by the faster kinetics; moreover, the rate of the G+• formation is found accelerated by increasing the guanine concentration. On the other hand, the decay of H2PO4• that is proportional to the increase of the G+• confirms the direct one-electron oxidation of guanine by H2PO4•. It is worth noting that the cation radical, G+• formed under our conditions, undergoes other processes - such as, hydration or solvation4,30 at the several-microsecond timescale39. However, the transient spectrum that we obtain, show some distinct features when compared with previous reports of (G+•-H2O) radicals (see supporting information Figure SI6 and SI7). Owing to the high concentration of phosphoric acid (6 M H3PO4) in our system, it is considered that phosphates may also compete with hydration process, giving rise to phosphate adduct radicals at C8 of guanine. These and similar to •GOH-type radicals30. Benchmark UV-vis spectra of these radicals are not known albeit spectrum of •GOH has been reported 39. In future, it is of great interest for us to address this issue.</p><p>Bimolecular rate constants (6.8 × 108 L mol−1 s−1, Figure 4 and Table 3) were determined from the slopes of the linear plots of the observed pseudo-first-order rate constants against the guanine concentrations (Figure 4). It is worth noting here that one-electron reduction or H-atom abstraction of guanine by H• is not likely to take place in such acidic conditions. This is because the solutions employed here are O2-saturated; as a result, H• is rapidly quenched by O2 by producing HO2• and the rate of reaction of HO2• with G is not high enough to be competitive with that of H2PO4• with G.3 One-electron oxidation of guanine by other radicals, such as SO4•−, was previously found to be nearly diffusion-controlled with a relatively higher rate (~3.2 × 109 L mol−1 s−1).35 The rate constant of oxidation of G in 6 M H3PO4 by H2PO4• is found to be lower in case of 6 M H3PO4 (see section 3 for discussion).</p><p>Our pulse radiolysis data (Figures 2–4) show that H2PO4• causes one-electron oxidation of DNA bases. To support the experimental observations, we have performed theoretical calculations on guanine (an example of purine base) for demonstrating the effect of H2PO4• on the oxidation of G and the effect of protonation of the G base on this oxidation. In the case of guanine, the C4=C5 double bond is well-known to add radicals, such as OH•, to form the G(C4-OH•) and G(C5-OH•) adducts.25</p><p>Based on this, we have investigated the addition of H2PO4• to the C4=C5 double bond of the guanine base moiety. Employing the B3LYP-PCM/6-31++G** method, the full (i.e., complete) geometry optimizations of G(C4-H2PO4•) and G(C5-H2PO4•) adducts show that H2PO4• does not form an adduct with either C4 or C5 of guanine. Rather, H2PO4• stabilizes itself over the plane of the guanine at distances between 2.8 Å to 3.0 Å, respectively (Figures 5(A) and 5(B)).</p><p>The spin density distributions of these conformations, calculated by B3LYP-PCM/6-31++G**, show that guanine is completely oxidized (spin density = 0.97 on guanine) to form G+• and H2PO4• is converted to its anion (H2PO4−). Thus, electron transfer from guanine to H2PO4• (Figures 5(A) and 6(B)), is complete. Thus, the mechanism of reaction between G and H2PO4• is due to single electron transfer (SET). Since, in our system (6 M H3PO4), guanine is protonated at N7, we have checked whether this protonation can hinder the oxidation of guanine by H2PO4• via SET. Therefore, we fully optimized the geometries of N7-protonated guanine and the H2PO4• by employing B3LYP-PCM/6-31++G** method; the optimized structures (N7H+G(C4-----H2PO4•) and N7H+G(C5-----H2PO4•)) are shown in Figures 5(C) and 5(D). In this case, our calculations also predict that H2PO4• substantially oxidizes the protonated guanine and the spin density on N7 protonated guanine is found to be 0.63 for structure N7H+G(C4-----H2PO4•), (Figure 5(C)) and 0.72 for structure N7H+G(C5-----H2PO4•), (Figure 5(D)). Globally these results are in agreement with the oxidation of guanine by phosphate radical via SET. Thermodynamically, the reaction is favored at the C4–C5 site (Figure 5).</p><!><p>Similar to our results found in guanine solutions, the observed absorbing radicals in adenine and thymine systems are assigned to be one-electron oxidized [A(N1H)]2+• and T+•. The detailed experimental analyses and discussions are present in supporting information (Figures SI8 and SI9). The rate constant value of thymine case is surprisingly ca. 1.5 times higher than the corresponding one of guanine and is ca. 4 times higher than that of adenine under identical conditions (Figure 4). The high rate constant of T oxidation by H2PO4• indicates that standard potential values of the DNA bases are not the only factor involved in the reaction between a DNA base and H2PO4•. Previous DFT calculations 40,41 suggested the protonation occurs preferentially at O4 in thymine, N3 in cytosine, N1 in adenine, and N7 in guanine. The proton affinity of thymine remains the lowest of all the nucleobases. The presence of acid (6 M H3PO4 in our case) may lead to increases of the standard potentials of DNA-base radical cation couples, E°(Base+•/Base), in varying degrees depending on the proton affinities of the bases. As a result, the one-electron oxidation rate of thymine case can be faster than those of other bases if it is not effectively protonated. In addition, this discrepancy can be attributed to possibly higher reorganization energies (for example, a Marcus-type effect) of base radicals and of H2PO4• in 6 M H3PO4, compared to the reorganization energies of those (e.g., G(−H+)•) in dilute aqueous solutions42. A small change in reorganization energy (e.g., 0.1 eV) can increase or decrease the rate constant of electron transfer reaction by one order of magnitude. Therefore, the different protonation states as well as reorganization energies of the base radicals and of H2PO4• in 6 M H3PO4 would play a significant role in the electron transfer rates from the strongly protonated bases (G, A, C (for C, see section 4 below)) and the weakly protonated base (T) to H2PO4•.</p><p>In case of thymine, immediately after the formation of T•+, a pseudo-first-order decay of the cation radicals has been observed. This decay depends on the concentration of thymine (Figure SI9). On the basis of similar type of concentration-dependent decay of N-centered Hoechst 33258 radicals (N-centered H-258 radicals) which has been attributed to the aggregation of an N-centered H-258 radical with a parent H-258 molecule,43 this concentration-dependent decay is attributed to the formation of thymine cation radical dimer (T2•+). Moreover, following recent ESR studies,42 H2PO4• can add to the C5=C6 double bond of T•+ (see SI).</p><!><p>Figure 6 represents the transient spectra and kinetics of various amounts of cytosine in 6 M H3PO4 solutions. In our system (6 M H3PO4), cytosine might exist as the N3-protonated cytosine, (C(N3H+)).3,45 Owing to the increase of the ionization potentials of both protonated cytosine (C(N3H+)) and protonated phosphate, H2PO4•,40 it is expected that the rate of oxidation of C(N3H+) by H2PO4• should be lower than for G. It is evident from Figure 6 that the kinetics at 520 nm and 310 nm do not change significantly by varying cytosine concentration from as low as 0.25 mM to 20 mM. However, besides the absorption of H2PO4• at 520 nm, we find a new absorption band at 425 nm whose intensity increases with the cytosine concentration. Our MCR-ALS analysis of 2D spectra (Figure 7) indicates that, compared to the three other DNA bases – G, A, and T - the decay of H2PO4• in the presence of cytosine (0.25 mM to 20 mM) appears to be independent of cytosine concentration. This finding suggests that H2PO4• does not react or reacts very slowly (< 107 L mol−1 s−1) with cytosine. Thus, the H2PO4• induced cytosyl radical formation is not observed within tens of microsecond. However, formation of a new absorbing species has been observed at ca. 550 nm and it is correlated with cytosine concentration. This is due to the addition reaction of OH• to cytosine either at C5 or at C6 of the C5=C6 double bond.3, 45 – 47 The lifetime of this radical is estimated to be tens of microsecond. 3, 45 –47</p><p>Again, similar to the guanine case, our theoretical calculations clearly present no evidence of H2PO4• mediated adduct of cytosine; rather, only a complex with H2PO4• stacked over cytosine is found (Figure 5E). The spin density distribution of this cytosine-H2PO4• stacked system shows that 0.76 spin is localized on cytosine while 0.24 spin is localized on the H2PO4. Therefore, a possible oxidation of cytosine by H2PO4• via SET cannot be ruled out; however, this reaction should be very slow. Thus, these results clearly show that the H2PO4• mediated oxidation of pyrimidine nucleobases is kinetically controlled.</p><!><p>The salient features of the work reported here are the following:</p><!><p>Implication of one-electron oxidation of DNA bases by H2PO4• to the backbone-to-base hole transfer process in irradiated DNA: the phosphate radical, H2PO4•, is an oxyl radical. Oxyl radicals are strong one-electron oxidizing radicals and can cause H-atom abstraction reaction. Analysis of our UV-visible pulse radiolysis spectra of DNA-base radical formed by the reactions of individual DNA base with H2PO4• clearly presents the evidence of direct one-electron oxidation of G, A and T in 6 M H3PO4. However, the rate of oxidation of protonated cytosine by H2PO4• is too slow to be detected by our system. These results, therefore, clearly indicate that the phosphate radical formed via direct ionization events in the sugar-phosphate backbone would oxidize the DNA bases; thus, these results should be treated as the benchmarks of backbone-to-base hole transfer process.</p><p>The rate constants of the oxidation of a DNA base in acidic condition (e.g., A, T, G, C) by H2PO4• are with a decreasing order of T > G > A > C. These results point out that the formation of the DNA base cation radical (e.g., G•+) from the activated complex (e.g. G(N7H+)···H2PO4•) is much faster than the diffusion of reactants (i.e., G(N7H+), T or A and H2PO4•) and is mediated by SET. The values of the rate constant reported in Table 3 show clearly that the standard potential values of DNA bases 3, 45, 48 are not the only determining factors of oxidation reaction rate with H2PO4•. Considering Schemes 1, S1 and S2 and the different extents of protonation of the DNA-bases in 6 M H3PO4, with the proton affinity of T being the lowest 41, it is not surprising that the observed rate of one-electron oxidation of T by H2PO4• is higher than those of the other bases. Also, the reorganization energies of protonated base radicals and of H2PO4• in 6 M H3PO4 (Marcus-type effect) could be a factor in these SET reactions and in scaling the order of reactivity of these radicals towards the SET reactions. Also, the standard potentials of the radicals will be different in 6 M H3PO4.</p><p>Nature of the DNA base radicals produced in our system (6 M H3PO4): It is evident from Schemes 1, S1 and S2 that the guanyl radical, produced by direct one-electron oxidation, is in its cation radical state. Adenyl radical is observed in its protonated cation radical state. Factors affecting reactivity of these radicals are discussed above (see point 2 (Conclusion) and section 3, Results and Discussion).</p><!><p>The present study is limited to the fundamental model that shows the reactions between phosphate radicals and DNA bases in a highly concentrated medium (6 M H3PO4). In this work, we show that phosphate radicals undergo intermolecular reactions with nucleobases. However, taking into account the 3-dimentional model of charge transfer processes in DNA49 and the proximity of the phosphate group to the base, the corresponding reaction in DNA is either intramolecular through-bond electron transfer, or intermolecular through-space electron transfer 49,50. Therefore, the parameters that govern these two reactions (i.e., intermolecular and intramolecular) should be different. Nevertheless, our present work clearly demonstrates the direct observation of electron transfer from phosphate radical to the weakly protonated base DNA-base, thymine. It will be interesting, if possible, to extend investigation of the phosphate radical reactivity to the DNA-bases via an intramolecular reaction. Moreover, our earlier work7 showed that the reaction of phosphate radicals in DNA either involves electron transfer from the sugar to the phosphate moiety followed by deprotonation, or alternatively, H-atom abstraction reactions by the phosphate radical from the 2′-deoxyribose moiety. Henceforth, we are extending our research efforts to study reactions of phosphate radicals with the DNA-models that are biologically relevant. Simultaneously, we are carrying out quantitative analyses to characterize stable molecular products that are formed via these reactions. These results will shed light into the mechanisms of this important oxidative pathway of DNA-damage.</p>

PubMed Author Manuscript

Synthesis and biological evaluation of 2-arylbenzofuran derivatives as potential anti-Alzheimer’s disease agents

AbstractAlzheimer's disease (AD) is a type of progressive dementia caused by degeneration of the nervous system. A single target drug usually does not work well. Therefore, multi-target drugs are designed and developed so that one drug can specifically bind to multiple targets to ensure clinical effectiveness and reduce toxicity. We synthesised a series of 2-arylbenzofuran derivatives and evaluated their in vitro activities. 2-Arylbenzofuran compounds have good dual cholinesterase inhibitory activity and β-secretase inhibitory activity. The IC50 value of compound 20 against acetylcholinesterase inhibition (0.086 ± 0.01 µmol·L−1) is similar to donepezil (0.085 ± 0.01 µmol·L−1) and is better than baicalein (0.404 ± 0.04 µmol·L−1). And most of the compounds have good BACE1 inhibitory activity, of which 3 compounds (8, 19 and 20) show better activity than baicalein (0.087 ± 0.03 µmol·L−1). According to experimental results, 2-arylbenzofuran compounds provide an idea for drug design to develop prevention and treatment for AD.

synthesis_and_biological_evaluation_of_2-arylbenzofuran_derivatives_as_potential_anti-alzheimer’s_di

Introduction<!><!>Molecular design<!><!>Chemistry<!><!>Chemistry<!>In vitro ChE inhibitory activity<!><!>In vitro BACE1 inhibitory activity<!>Kinetic characterisation of ChEs inhibition<!><!>Cytotoxic<!><!>Inhibition of reactive oxygen species in cells<!><!>Molecular design<!><!>Molecular design<!>Materials and methods<!>General method for synthesis of compounds 2a–2v<!>General method for synthesis of compounds 3a–3v<!>General method for synthesis of compounds 1–15<!>General method for synthesis of compounds 16–22<!>2-Phenylbenzofuran (1)<!>7-Methoxy-2-phenylbenzofuran (2)<!>6-Methoxy-2-phenylbenzofuran (3)<!>5-Methoxy-2-phenylbenzofuran (4)<!>4-Methoxy-2-phenylbenzofuran (5)<!>4,6-Dimethoxy-2-phenylbenzofuran (6)<!>7-Methyl-2-phenylbenzofuran (7)<!>6-Methyl-2-phenylbenzofuran (8)<!>5-Methyl-2-phenylbenzofuran (9)<!>4,7-Dimethyl-2-phenylbenzofuran (10)<!>2-Phenylnaphtho[2,1-b]furan (11)<!>2-Phenylnaphtho[1,2-b]furan (12)<!>N, N-diethyl-2-phenylbenzofuran-6-amine (13)<!>7-Ethoxy-2-phenylbenzofuran (14)<!>6-Ethyl-2-phenylbenzofuran (15)<!>2-Phenylbenzofuran-7-ol (16)<!>2-Phenylbenzofuran-6-ol (17)<!>2-Phenylbenzofuran-5-ol (18)<!>2-Phenylbenzofuran-4-ol (19)<!>2-Phenylbenzofuran-4,6-diol (20)<!>4-Methoxy-2-phenylbenzofuran-6-ol (21)<!>6-Methoxy-2-phenylbenzofuran-4-ol (22)<!>In vitro ChEs inhibitory activity<!>In vitro BACE1 inhibitory activity<!><!>In vitro BACE1 inhibitory activity<!>Kinetic characterisation of ChEs inhibition<!>Cytotoxicity test<!>Inhibition of reactive oxygen species in cells<!>Conclusion<!><!>Disclosure statement

<p>Alzheimer's disease (AD) is a type of progressive dementia caused by degeneration of the nervous system. It has a higher incidence in the elderly, so it is called senile dementia. Clinical manifestations are mainly memory impairment, cognitive dysfunction, mental symptoms, and behavioural abnormalities1. The two core pathological features of AD are amyloid plaques and neurofibrillary tangles2. AD features include progressive memory loss and severe cognitive decline. AD is a typical multifactorial disease, and its pathogenesis involves abnormalities in the structure and function of multiple systems3. Therefore, single-target drugs often do not work well. Design and develop multi-target drugs so that one drug can specifically bind to multiple targets to ensure clinical effectiveness while reducing toxicity.</p><p>Most of the cognitive symptoms of AD are caused by the depletion of cholinergic neurons in the basal forebrain, which leads to a decrease in cholinergic neurotransmission, of which acetylcholine (ACh) is the most important neurotransmitter. Various evidences indicate that Cholinesterases (ChEs) are closely related to the aetiology and symptoms of the disease. The drugs used increase cholinergic neurotransmission by inhibiting acetylcholinesterase (AChE), such as tacrine4, donepezil5, rivastigmine6, or galantamine7. The data from clinical trials of these AChE inhibitor provides convincing evidence that they have sufficient AD efficacy and reliability, as well as acceptable side effects. In recent years, the role of butyrylcholinesterase (BuChE) inhibition in AD has received increasing attention from a medicinal chemistry and clinical perspective8. There is evidence that BuChE may be a co-regulator of the neurotransmitter ACh activity. BuChE plays an important role in cholinergic neurotransmission and is involved in other nervous system functions and neurodegenerative diseases9.</p><p>One of the main causes of the course of AD is the accumulation of β-amyloid (Aβ) in the brain10. Aβ is a short peptide produced by the hydrolysis of amyloid precursor protein (APP) by β-secretase (BACE 1)11 and γ-secretase12. Aβ is secreted outside the cell, accumulates to form aggregates and fibrils, and finally forms amyloid deposits, which are called senile plaques13. Blocking the activity of BACE1 can inhibit the production of Aβ peptide. Since there are no effective therapeutic drugs in the clinic, the study of BACE1 inhibitors for therapeutic intervention in patients with Alzheimer's disease has increasingly become a research hotspot.</p><p>Studies have found that baicalein (Scheme 1) can improve neurodegenerative diseases and has the potential to treat AD14. However, there are three adjacent phenolic hydroxyl groups in the structure of baicalein, which are easy to form intramolecular hydrogen bonds, resulting in poor lipophilicity and hydrophilicity, and it has disadvantages such as poor solubility and low bioavailability in clinical applications. Therefore, based on the structure of baicalein, 2-arylbenzofurans were designed to explore its anti-AD activity. Benzofuran compounds have attracted much attention due to their many biological pharmacological activities, such as anti-oxidant15, anti-bacterial16, anti-fungal17, anti-inflammatory18, anti-tumour19, hypoglycemic20, anti- cholinesterase21, anti-monoamine oxidase22, and other activity. The aetiology of AD is complex, and multi-targeted drugs are superior to single-targeted drugs for this disease23. Recently, various benzofuran-based compounds were reported as potent acetylcholinesterase inhibitors. The natural product benzofuran skeleton can be regarded as a mimic of the indanone part of donepezil, which has a good inhibitory activity of AChE and a moderate ability to inhibit the self-mediated aggregation of Aβ1-4224–25. The benzofuran backbone has become an important pharmacophore for the design of antiviral26 and antibacterial agents27, as well as cyclin-dependent kinases (CDKs)28 and cholinesterase inhibitors29. The current study describes the preparation and in vitro activity of 2-arylbenzofuran derivatives as ChE inhibitors and BACE1 inhibitors. The study aims to screen ChEs/BACE1 dual target inhibitors, which have the potential to treat AD and other neurodegenerative diseases. The experiments we have performed provide an idea for the development of drug designs for therapeutic or prophylactic agents for AD.</p><!><p>The structure of baicalein.</p><!><p>Using 40 benzaldehyde compounds and 28 methyl phenylacetate compounds as raw materials, 1120 2-arylbenzofuran compounds were designed by creating a small molecule library according to the reaction type. According to the analysis of drug-like properties, the compounds in the constructed small molecule library all meet the five rules of drug-like properties.</p><p>Use molecular docking technology to screen ChEs/BACE1 dual target inhibitors from small molecule libraries. The compounds were docked with AChE, BuChE and BACE 1, and the compound with higher scores in 3 docking times was selected as the target compound for synthesis. The final screening yielded 11 compounds with high docking scores for all three enzymes. Three randomly selected compounds and enzyme interaction patterns are displayed (Figures 1–3). Based on the comprehensive consideration of design results and synthesis cost, 22 2-arylbenzofurans have been synthesised.</p><!><p>Schematic presentations of the compound 612 binding modes with AChE, BuChE and BACE1.</p><p>Schematic presentations of the compound 1004 binding modes with AChE, BuChE and BACE1.</p><p>Schematic presentations of the compound 584 binding modes with AChE, BuChE and BACE1.</p><!><p>2-Arylbenzofuran were obtained from the substituted 2-hydroxybenzaldehyde in a three-step reaction. The routes for the synthesis of 2-arylbenzofuran derivatives are shown in Scheme 2 and the final products are listed in Table 1. The synthesis method of 2-arylbenzofuran compound is based on the Droździk and co-workers' method and slightly optimized30. O-Alkylation of the substituted 2-hydroxybenzaldehyde with methyl α-bromophenylacetate in the presence of potassium carbonate in dimethylformamide gave the corresponding methyl 2–(2-formylphenoxy)-2 -phenylacetates in high yields. Hydrolysis of the methyl 2–(2-formylphenoxy)-2 -phenylacetates to the 2–(2-formylphenoxy)-2-phenylacetic acids, followed by cyclisation of the acids to obtain the corresponding compounds 1–22. Details on the chemical and spectroscopic characterisations of compounds 1–22 were described in the Supporting Information.</p><!><p>General synthetic route to 2-arylbenzofuran, reagents and conditions: (a) K2CO3, DMF, 50 °C; (b) 10%KOH, CH3OH, 10%HCl, 82 °C; (c) Ac2O, AcONa, 125 °C; (d) CH3CN, I2, Al, 90 °C.</p><p>Compounds 1–22.</p><p>13, 20, 21 and 22 are new compounds.</p><!><p>Compared with the original method, the temperature of the reaction in the first stage is adjusted to 50 °C to prevent the reaction from proceeding excessively. Excessive reactions can affect compound yield and purity. The completion of the reaction is monitored by thin layer chromatography (TLC), not just the reaction time. Compounds with different substituents have different reaction times. Most 2-arylbenzofuran compounds can be purified by recrystallization from methanol, which is efficient and fast.</p><!><p>The method of Ellman et al.31–33 was used to evaluate the ChEs inhibitory activity of all compounds with donepezil as a standard sample. As shown in Table 2, most compounds have AChE inhibitory activity. Among all the compounds, compound 20 (IC50 = 0.086 ± 0.01 µmol·L−1) has the strongest activity, similar to the positive drug donepezil (IC50 = 0.079 ± 0.01 µmol·L−1), and far better than the natural product baicalein (IC50 = 0.404 ± 0.04 µmol·L−1). Nearly half of the compounds have BuChE inhibitory activity. In the same situation, compound 20 (IC50 = 16.450 ± 2.12 µmol·L−1) has the highest inhibitory activity, which is slightly worse than donepezil (IC50 = 7.100 ± 0.23 µmol·L−1), but still better than baicalein (IC50 = 31.624 ± 0.01 µmol·L−1). According to the results in Table 2, most of the compounds have dual ChEs inhibitory activity, and their effect may be better compared with selective ChE inhibitors. Compared with other compounds, R4 and R6 of compound 20 each have a hydroxyl substituent, which may be the reason for its high activity.</p><!><p>Biological evaluation in vitro.</p><!><p>A BACE1 activity assay kit was used to evaluate the BACE1 inhibitory activity of all compounds. As shown in Table 2, most of the compounds have good BACE1 inhibitory activity, of which 3 compounds (8, 19 and 20) show better activity than baicalein (IC50 = 0.087 ± 0.03 µmol·L−1). It can be seen that the 2-arylbenzofuran structure is a superior BACE1 inhibitor than baicalein. According to comprehensive activity research experiments, in addition to good ChEs inhibitory activity, compound 20 also has high BACE1 inhibitory activity (IC50 = 0.043 ± 0.01 µmol·L−1). We speculate from the results that the inclusion of two or more hydroxyl groups on the aromatic ring of the core of the 2-arylbenzofuran compound can increase its BACE1 inhibitory activity and also improve other activities.</p><!><p>To understand the mechanism of action of these derivatives on ChEs, the compound 20, which has good inhibitory activity on ChEs, was selected for kinetic study. Figure 4 shows a graphical analysis of the steady-state suppression data for compound 20 pairs of ChEs. According to this figure, it can be judged that the inhibition mode of compound 20 on ChEs is reversible inhibition.</p><!><p>Kinetic study of the mechanism of ChEs inhibition by compound 20. Overlaid Lineweaver–Burk reciprocal plots of ChEs initial velocity at increasing substrate concentration in the absence of inhibitor and in the presence of 20 are shown. (A) A double reciprocal plot of compound 20 inhibition of AChE. (B) A double reciprocal plot of compound 20 inhibition of BuChE.</p><!><p>In order to study the activity of the compound at the cellular level, firstly select compounds 20 and 21 with the better activity in all aspects for cytotoxicity experiment34. The purpose of this experiment is to study whether 2-arylbenzofurans are toxic to normal cells. The results are shown in the Figure 5, each concentration of compounds can make the survival rate of cells reach more than 90%, and it is basically non-toxic to cell growth. And there is a certain linear relationship between the cell survival rate and the concentration of the compound.</p><!><p>Effect of compound 20 and 21 on the survival rate of HEK-293 and SH-SY5Y cells.</p><!><p>The reactive oxygen species (ROS) detection kit was used to analyse the ability of compound 20 with the best activity to inhibit the level of ROS in model cells by the DCFH-DA method. As shown in Figure 6, compound 20 significantly reduces the surge of ROS levels in AD model cells, and its effect is better than that of baicalein at the same concentration. Compound 20 (5 mg·L−1) can significantly reduce the level of ROS in AD model cells, and as the concentration increases, the effect gets better and better. Compound 20 (20 mg·L−1) can reduce the level of ROS in AD model cells to be similar to normal cells. It can be seen that compound 20 can significantly improve the oxidative stress response caused by AD.</p><!><p>The inhibitory effect of compound 20 on ROS in cells. (1) is the ROS contained in normal cells. (2) is the sharp increase of ROS levels in the AD model cells. (3) is the inhibition of baicalein (20 mg·L−1) on ROS in AD model cells. (4), (5) and (6) are the inhibition of compound 20 (20, 10, 5 mg·L−1) on ROS in AD model cells.</p><!><p>Discovery Studio (DS) is a molecular simulation software in the field of life sciences, aimed at biological macromolecules and small organic molecules. A small molecule library containing a series of 2-arylbenzofuran compounds was constructed by constructing a small molecule library according to the reaction type. Draw the structure of benzaldehyde compounds and methyl phenylacetate compounds by Chemdraw. Use Chemdraw to draw the reaction formula (Scheme 3) needed to construct the small molecule library and save it in .rxn format for later use. Use Enumerate Library by Reaction in the Design and Analysis Libraries module of DS to construct a virtual combinatorial compound library by reaction type. Copy 40 benzaldehyde compounds and 28 methyl phenylacetate compounds to the same molecular window and use them as reactant options. The reaction type file selection has been saved in the .rxn format. Keep the remaining settings as default.</p><!><p>Reaction formula needed to construct the small molecule library.</p><!><p>Compounds that meet the five rules of drug-like properties have better pharmacokinetic properties, higher bioavailability in the process of metabolism in the body, and more likely to become oral drugs. Filter the constructed small molecules through Lipinski and veber rules in Prepare or Filter Ligands built in DS. According to the analysis of drug-like properties, the compounds in the constructed small molecule library all meet the five rules of drug-like properties.</p><p>Molecular docking of the compounds into protein was performed employing Libdock rapid docking technology introduced in the DS. Three-dimensional (3 D) protein structure of AChE (PDB id: 6O4W), BuChE (PDB id: 6EP4) and BACE1 (PDB id: 4IVS) were accessed from protein data bankk. The ligand is processed first, and the prepared ligand method built into DS is used to optimise the energy of the generated ligand. Then the DS built-in prepare protein method is used for the protein required for docking. The main purpose is to hydrogenate the protein, retain the water in the protein binding pocket, and remove the original ligand.</p><!><p>Melting points were determined using a Thiele tube and were uncorrected. The 1H NMR and 13 C NMR spectra were recorded with a Bruker AM-600 spectrometer (Billercia, MA, USA) with TMS as the internal standard. Chemical shifts were reported at room temperature on a scale (ppm) with CDCl3 or DMSO-d6 as the solvents and J values are given in Hertz. Mass spectra were obtained with an Agilent Trap VL LC/MS spectrometer (Santa Clara, CA, USA). The absorbance was recorded by RZ-9618 Microplate Reader. Unless otherwise noted, all solvents and reagents were commercially available and used without further purification.</p><!><p>Taking the synthesis of methyl 2–(2-formyl-6-methoxyphenoxy)-2-phenylacetate as an example. Other phenylacetates were obtained using the same procedures. A mixture of methyl 2-bromo-2-phenylacetate (1.08 g, 4.7 mmol), o-Vanillin (0.72 g, 4.7 mmol), K2CO3 (0.76 g, 5.5 mmol) and DMF (10 ml) was stirred and heated at 92–94 °C. After the TLC detection reaction was completed, the mixture was cooled, poured onto ice/H2O (150 ml) and the mixture was kept at 4 °C overnight. After that, the precipitate was filtered, washed with water and dried, and recrystallized from methanol to obtain white crystals (yield: 85.10%).</p><!><p>Taking the synthesis of 2–(2-formyl-6-methoxyphenoxy)-2-phenylacetic acid as an example. Other phenylacetic acids were obtained using the same procedures. A mixture of methyl 2–(2-formyl-6-methoxyphenoxy)-2-phenylacetate (2.10 g, 7 mmol), 10% aq KOH (30 ml) and methanol (2 ml) was stirred and heated at 80–82 °C for 2 h. After the mixture was cooled to room temperature, it was poured onto ice/H2O (150 ml), and 10% aqueous HCl was added with stirring until precipitation appeared. The precipitate was filtered off, washed several times with H2O and dried. The crude product was obtained as a solid (yield: 78.57%).</p><!><p>Taking the synthesis of 7-methoxy-2-phenylbenzofuran as an example. Compounds 2–15 were obtained using the same procedures. A mixture of 2–(2-formyl-6-methoxyphenoxy)-2-phenylacetic acid (0.94 g, 3.3 mmol), anhydrous AcONa (2.71 g, 33 mmol) and Ac2O (35 ml) was stirred and heated at 120–125 °C for 4 h. Then, the mixture was cooled and poured onto ice/H2O (200 ml) and left at refrigerator for 12 h. The precipitate was filtered off, washed several times with cooled H2O and dried. The crude product was recrystallized from n-hexane (yield: 90.91%).</p><!><p>Taking the synthesis of 2-phenylbenzofuran-7-ol as an example. Compounds 17–22 were obtained using the same procedures. Add 150 ml of acetonitrile solution to the three-necked reaction flask, add iodine (1.27 g, 5 mmol) and aluminium powder (0.50 g, 18.5 mmol), reflux for 3 h, cool to room temperature, add 0.56 g (2.5 mmoL) of compound 1, and heat to reflux for reaction, TLC monitors the progress of the reaction, and the reaction is complete after 24 h. The reaction solution was concentrated, and the concentrated solution was washed with 5% sodium bisulphite to remove excess iodine. After acidification with hydrochloric acid, the aqueous solution was extracted with ethyl acetate. The organic layer was concentrated and evaporated to dryness to obtain a yellow product (yield: 72%).</p><!><p>White solid, yield 65.14%. 1H NMR (600 MHz, CDCl3) δ 7.90 (m, 2H, Ar), 7.61 (d, J = 7.6 Hz, 1H, Ar), 7.55 (d, J = 8.1 Hz, 1H, Ar), 7.47 (t, 2H, Ar), 7.38 (t, 1H, Ar), 7.31 (dd, 1H, Ar), 7.25 (d, J = 7.3 Hz, 1H, Ar), 7.05 (s, 1H, Ar). 13 C NMR (150 MHz, CDCl3) δ 155.90 (s), 154.87 (s), 130.47 (s), 129.20 (s), 128.77 (s), 128.53 (s), 124.92 (s), 124.24 (s), 122.91 (s), 120.88 (s), 111.16 (s), 101.28 (s). MS: m/z (%) [M + Na]+ 217.6.</p><!><p>White solid, yield 90.91%. 1H NMR (600 MHz, CDCl3) δ 7.92 (m, 2H, Ar), 7.46 (t, 2H, Ar), 7.36 (t, 1H, Ar), 7.19 (m, 2H, Ar), 7.03 (s, 1H, Ar), 6.83 (d, J = 7.7 Hz, 1H, Ar), 4.07 (s, 3H, OCH3). 13 C NMR (150 MHz, CDCl3) δ 156.05 (s), 145.31 (s), 144.11 (s), 130.91 (s), 130.31 (s), 128.69 (s), 128.53 (s), 125.02 (s), 123.57 (s), 113.31 (s), 106.69 (s), 101.61 (s), 56.13 (s). MS: m/z (%) [M + Na]+ 247.5, [2M + Na]+ 471.5.</p><!><p>White solid, yield 79.4%. 1H NMR (600 MHz, CDCl3) δ 7.83 (m, 2H, Ar), 7.45 (m, 3H, Ar), 7.34 (t, 1H, Ar), 7.09 (d, J = 1.8 Hz, 1H, Ar), 6.97 (s, 1H, Ar), 6.89 (dd, J = 8.5, 2.2 Hz, 1H, Ar), 3.89 (s, 3H, OCH3). 13 C NMR (150 MHz, CDCl3) δ 158.05 (s), 155.89 (s), 155.12 (s), 130.69 (s), 128.73 (s), 128.03 (s), 124.43 (s), 122.53 (s), 120.97 (s), 111.95 (s), 101.13 (s), 95.85 (s), 55.72 (s). MS: m/z (%) [M + Na]+ 247.2, [2M + Na]+ 471.2.</p><!><p>White solid, yield 83.3%. 1H NMR (600 MHz, CDCl3) δ 7.86 (m, 2H, Ar), 7.44 (dd, 3H, Ar), 7.36 (t, 1H, Ar), 7.06 (d, J = 2.6 Hz, 1H, Ar), 6.98 (s, 1H, Ar), 6.91 (dd, J = 8.9, 2.6 Hz, 1H, Ar), 3.88 (s, 3H, OCH3). 13 C NMR (150 MHz, CDCl3) δ 156.68 (s), 156.05 (s), 149.92 (s), 130.53 (s), 129.75 (s), 128.75 (s), 128.48 (s), 124.83 (s), 112.96 (s), 111.58 (s), 103.29 (s), 101.45 (s), 55.88 (s). MS: m/z (%) [M + Na]+ 247.6, [2M + Na]+ 471.6.</p><!><p>White solid, yield 71.6%. 1H NMR (600 MHz, CDCl3) δ 7.87 (dd, J = 8.3, 1.2 Hz, 2H, Ar), 7.46 (t, 2H, Ar), 7.35 (t, 1H, Ar), 7.23 (t, 1H, Ar), 7.19 (d, J = 8.3 Hz, 1H, Ar), 7.15 (d, J = 0.7 Hz, 1H, Ar), 6.69 (d, J = 7.8 Hz, 1H, Ar), 3.98 (s, 3H, OCH3). 13 C NMR (150 MHz, CDCl3) δ 156.04 (s), 154.61 (s), 153.41 (s), 130.53 (s), 128.74 (s), 128.26 (s), 124.94 (s), 124.71 (s), 119.50 (s), 104.45 (s), 103.28 (s), 98.76 (s), 55.58 (s). MS: m/z (%) [M + Na]+ 247.7, [2M + Na]+ 471.7.</p><!><p>White solid, yield 94.1%. 1H NMR (600 MHz, CDCl3) δ 7.81 (dd, J = 8.3, 1.1 Hz, 2H, Ar), 7.43 (m, 2H, Ar), 7.31 (m, 1H, Ar), 7.05 (d, J = 0.7 Hz, 1H, Ar), 6.72 (m, 1H, Ar), 6.35 (d, J = 1.9 Hz, 1H, Ar), 3.94 (s, 3H, OCH3), 3.88 (s, 3H, OCH3). 13 C NMR (150 MHz, CDCl3) δ 159.16 (s), 156.54 (s), 153.65 (s), 153.45 (s), 130.75 (s), 128.70 (s), 127.72 (s), 124.19 (s), 113.22 (s), 98.72 (s), 94.29 (s), 88.20 (s), 55.76 (s), 55.55 (s). MS: m/z (%) [M + H]+ 255.6, [M + Na]+ 277.7, [2M + Na]+ 531.7.</p><!><p>White solid, yield 79.2%. 1H NMR (600 MHz, CDCl3) δ 7.91 (dd, J = 8.3, 1.2 Hz, 2H, Ar), 7.46 (m, 3H, Ar), 7.37 (m, 1H, Ar), 7.16 (t, 1H, Ar), 7.11 (d, J = 7.2 Hz, 1H, Ar), 7.04 (s, 1H, Ar), 2.62 (s, 3H, CH3). 13 C NMR (150 MHz, CDCl3) δ 155.51 (s), 153.89 (s), 130.67 (s), 128.69 (d, J = 11.6 Hz), 128.40 (s), 125.19 (s), 124.87 (s), 122.94 (s), 121.40 (s), 118.33 (s), 101.56 (s), 15.04 (s). MS: m/z (%) [M + H]+ 209.6, [M + Na]+ 231.1.</p><!><p>White solid, yield 85.32%. 1H NMR (600 MHz, CDCl3) δ 7.87 (dd, J = 8.4, 1.2 Hz, 2H, Ar), 7.46 (m, 3H, Ar), 7.36 (m, 2H, Ar), 7.08 (dd, J = 7.9, 0.7 Hz, 1H, Ar), 7.00 (d, J = 0.8 Hz, 1H, Ar), 2.51 (s, 3H, CH3). 13 C NMR (150 MHz, CDCl3) δ 155.33 (s), 155.31 (s), 134.54 (s), 130.67 (s), 128.73 (s), 128.25 (s), 126.67 (s), 124.73 (s), 124.33 (s), 120.34 (s), 111.41 (s), 101.16 (s), 21.74 (s). MS: m/z (%) [M + Na]+ 231.6.</p><!><p>White solid, yield 91.11%. 1H NMR (600 MHz, CDCl3) δ 7.88 (m, 2H, Ar), 7.44 (dd, 3H, Ar), 7.37 (dd, 2H, Ar), 7.12 (dd, J = 8.3, 1.3 Hz, 1H, Ar), 6.97 (s, 1H, Ar), 2.47 (s, 3H, CH3). 13 C NMR (150 MHz, CDCl3) δ 155.97 (s), 153.31 (s), 132.31 (s), 130.61 (s), 129.29 (s), 128.73 (s), 128.39 (s), 125.51 (s), 124.83 (s), 120.71 (s), 110.63 (s), 101.06 (s), 21.33 (s). MS: m/z (%) [M + H]+ 209.1, [M + Na]+ 231.1.</p><!><p>White solid, yield 58.0%. 1H NMR (600 MHz, CDCl3) δ 7.92 (dd, J = 8.3, 1.1 Hz, 2H, Ar), 7.48 (t, 2H, Ar), 7.37 (m, 1H, Ar), 7.06 (s, 1H, Ar), 7.01 (d, J = 7.4 Hz, 1H, Ar), 6.96 (d, J = 7.4 Hz, 1H, Ar), 2.58 (s, 3H, CH3), 2.54 (s, 3H, CH3). 13 C NMR (150 MHz, CDCl3) δ 155.02 (s), 153.60 (s), 130.80 (s), 128.71 (s), 128.43 (s), 128.25 (s), 128.08 (s), 125.07 (s), 124.79 (s), 123.10 (s), 118.58 (s), 100.29 (s), 18.34 (s), 14.81 (s). MS: m/z (%) [M + H]+ 223.2, [M + Na]+ 245.3, [2M + Na]+ 467.3.</p><!><p>White solid, yield 81.8%. 1H NMR (600 MHz, CDCl3) δ 8.08 (d, J = 8.2 Hz, 1H, Ar), 7.85 (m, 3H, Ar), 7.62 (m, 2H, Ar), 7.51 (m, 1H, Ar), 7.41 (m, 4H, Ar), 7.16 (s, 1H, Ar). 13 C NMR (150 MHz, CDCl3) δ 155.37 (s), 152.35 (s), 130.63 (s), 130.41 (s), 128.83–128.79 (d, J = 6.1 Hz), 128.25 (s), 127.59(s), 126.25 (s), 125.16 (s), 124.64 (s), 124.55–124.52 (d J = 4.5 Hz), 123.44 (s), 112.27 (s), 100.43 (s). MS: m/z (%) [M + H]+ 245.6, [M + Na]+ 267.5.</p><!><p>White solid, yield 83.2%. 1H NMR (600 MHz, CDCl3) δ 8.43 (d, J = 8.2 Hz, 1H, Ar), 7.97 (m, 3H, Ar), 7.64 (m, 3H, Ar), 7.51 (m, 3H, Ar), 7.38 (m, 1H, Ar), 7.17 (s, 1H, Ar). 13 C NMR (150 MHz, CDCl3) δ 155.30 (s), 150.29 (s), 131.47 (s), 130.72 (s), 128.81 (s), 128.41 (s), 128.21 (s), 126.33 (s), 125.02 (s), 124.80 (s), 124.65 (s), 123.61 (s), 121.33 (s), 120.01 (s), 119.54 (s), 102.45 (s). MS: m/z (%) [M + H]+ 245.3, [M + Na]+ 267.1.</p><!><p>White solid, yield 90.1%. 1H NMR (600 MHz, DMSO-d6) δ 7.80 (dd, J = 8.3, 1.1 Hz, 2H, Ar), 7.44 (t, 2H, Ar), 7.39 (d, J = 8.6 Hz, 1H, Ar), 7.30 (t, 1H, Ar), 7.21 (d, J = 0.6 Hz, 1H, Ar), 6.82 (d, J = 1.8 Hz, 1H, Ar), 6.69 (dd, J = 8.7, 2.2 Hz, 1H, Ar), 3.38 (q, J = 7.0 Hz, 4H, CH2), 1.11 (t, 6H, CH3). 13 C NMR (150 MHz, DMSO-d6) δ 156.73 (s), 152.18 (s), 146.29 (s), 130.48 (s), 128.90 (s), 127.54 (s), 123.63 (s), 121.24 (s), 117.69 (s), 109.64 (s), 101.96 (s), 93.46 (s), 44.19 (s), 12.39 (s). HR-MS: m/z (%) [M + H]+ 266.1215.</p><!><p>White solid, yield 79.1%. 1H NMR (600 MHz, CDCl3) δ 7.92 (dd, J = 8.3, 1.1 Hz, 2H, Ar), 7.46 (t, 2H, Ar), 7.37 (m, 1H, Ar), 7.17 (m, 2H, Ar), 7.03 (s, 1H, Ar), 6.83 (dd, J = 7.8, 0.9 Hz, 1H, Ar), 4.35 (q, J = 7.0 Hz, 2H, CH2), 1.56 (t, 3H, CH3). 13 C NMR (150 MHz, CDCl3) δ 155.93 (s), 144.55 (s), 144.37 (s), 131.02 (s), 130.37 (s), 128.68 (s), 128.48 (s), 125.03 (s), 123.55 (s), 113.23 (s), 108.18 (s), 101.62 (s), 64,74 (s), 15.00 (s). MS: m/z (%) [M + H]+ 239.3, [M + Na]+ 261.1.</p><!><p>Yellow solid, yield 78.8%. 1H NMR (600 MHz, CDCl3) δ 7.86 (dd, J = 8.2, 1.0 Hz, 2H, Ar), 7.49 (d, J = 7.9 Hz, 1H, Ar), 7.46 (dd, 2H, Ar), 7.36 (m, 2H, Ar), 7.10 (dd, J = 7.9, 1.1 Hz, 1H, Ar), 7.00 (d, J = 0.6 Hz, 1H, Ar), 2.79 (t, J = 7.6 Hz, 2H, CH2), 1.31 (t, 3H, CH3). 13 C NMR (150 MHz, CDCl3) δ 155.40 (d, J = 12.2 Hz), 141.22 (s), 130.69 (s), 128.73 (s), 128.26 (s), 126.89 (s), 124.73 (s), 123.30 (s), 120.44 (s), 110.15 (s), 101.17 (s), 29.14 (s), 15.99 (s). MS: m/z (%) [M + H]+ 223.5, [M + Na]+ 245.4.</p><!><p>Light yellow solid, yield 72%. 1H NMR (600 MHz, CDCl3) δ 7.87 (m, 2H, Ar), 7.46 (t, 2H, Ar), 7.38 (t, 1H, Ar), 7.17 (dd, J = 7.7, 0.9 Hz, 1H, Ar), 7.12 (t, 1H, Ar), 7.04 (s, 1H, Ar), 6.86 (dd, J = 7.7, 0.8 Hz, 1H, Ar). 13 C NMR (150 MHz, CDCl3) δ 156.04 (s), 143.10 (s), 140.77 (s), 130.79 (s), 130.21 (s), 128.74 (d, J = 17.8 Hz), 124.93 (s), 123.91 (s), 113.22 (s), 110.71 (s), 102.05 (s). MS: m/z (%) [M + Na]+ 233.2, [2M + Na]+ 443.2.</p><!><p>Light yellow solid, yield 76.6%. 1H NMR (600 MHz, CDCl3) δ 7.82 (dd, J = 8.3, 1.1 Hz, 2H, Ar), 7.43 (m, 3H, Ar), 7.33 (m, 1H, Ar), 7.03 (d, J = 1.9 Hz, 1H, Ar), 6.96 (d, J = 0.6 Hz, 1H, Ar), 6.80 (dd, J = 8.3, 2.2 Hz, 1H, Ar). 13 C NMR (150 MHz, CDCl3) δ 155.75 (s), 155.28 (s), 153.59 (s), 130.59 (s), 128.73 (s), 128.10 (s), 124.48 (s), 122.87 (s), 121.12 (s), 112.03 (s), 101.11 (s), 98.28 (s). MS: m/z (%) [M + Na]+ 233.3, [2M + Na]+ 443.3.</p><!><p>Light yellow solid, yield 84.3%. 1H NMR (600 MHz, DMSO-d6) δ 9.22 (s, 1H, OH), 7.87 (dd, J = 8.3, 1.2 Hz, 2H, Ar), 7.49 (t, 2H, Ar), 7.39 (ddd, 2H, Ar), 7.28 (d, J = 0.8 Hz, 1H, Ar), 6.95 (d, J = 2.4 Hz, 1H, Ar), 6.76 (dd, J = 8.8, 2.5 Hz, 1H, Ar). 13 C NMR (150 MHz, DMSO-d6) δ 155.54 (s), 153.55 (s), 148.48 (s), 129.98 (s), 129.58 (s), 129.01 (s), 128.64 (s), 124.47 (s), 113.35 (s), 111.32 (s), 105.37 (s), 101.98 (s). MS: m/z (%) [M + Na]+ 233.7, [2M + Na]+ 443.7.</p><!><p>Light yellow solid, yield 83.2%. 1H NMR (600 MHz, DMSO-d6) δ 10.03 (s, 1H, OH), 7.88 (dd, J = 8.3, 1.1 Hz, 2H, Ar), 7.48 (m, 2H, Ar), 7.41 (d, J = 0.8 Hz, 1H, Ar), 7.38 (m, 1H, Ar), 7.11 (t, 1H, Ar), 7.06 (d, J = 8.2 Hz, 1H, Ar), 6.64 (dd, J = 7.7, 0.7 Hz, 1H, Ar). 13 C NMR (150 MHz, DMSO-d6) δ 155.84 (s), 153.21 (s), 151.31 (s), 129.94 (s), 129.01 (s), 128.43 (s), 125.50 (s), 124.36 (s), 118.19 (s), 108.02 (s), 102.29 (s), 99.63 (s). MS: m/z (%) [M + Na]+ 233.1.</p><!><p>Lilac solid, yield 62.8%. 1H NMR (600 MHz, CDCl3) δ 7.78 (d, J = 7.6 Hz, 2H, Ar), 7.42 (t, 3H, Ar), 6.99 (s, 1H, Ar), 6.63 (s, 1H, Ar), 6.25 (d, J = 1.6 Hz, 1H, Ar). 13 C NMR (150 MHz, CDCl3) δ 156.88 (s), 154.38 (s), 154.00 (s), 149.28 (s), 130.49 (s), 128.73 (s), 127.98 (s), 124.33 (s), 112.33 (s), 98.03 (s), 97.88 (s), 91.52 (s). HR-MS: m/z (%) [M + H]+ 227.1260.</p><!><p>Purple solid, yield 71.1%. 1H NMR (600 MHz, CDCl3) δ 7.79 (dd, J = 8.3, 1.0 Hz, 2H, Ar), 7.42 (t, 2H, Ar), 7.31 (dd, 1H, Ar), 7.04 (d, J = 0.5 Hz, 1H, Ar), 6.65 (d, J = 0.9 Hz, 1H, Ar), 6.28 (d, J = 1.8 Hz, 1H, Ar), 3.92 (s, 3H, OCH3). 13 C NMR (150 MHz, CDCl3) δ 156.35 (s), 154.59 (s), 153.67 (d, J = 18.9 Hz), 130.65 (s), 128.70 (s), 127.80 (s), 124.24 (s), 113.25 (s), 98.69 (s), 94.17 (s), 91.20 (s), 55.62 (s). HR-MS: m/z (%) [M + H]+ 241.0574.</p><!><p>Lilac solid, yield 50.3%. 1H NMR (600 MHz, CDCl3) δ 7.79 (m, 2H, Ar), 7.42 (d, J = 8.0 Hz, 2H, Ar), 7.32 (dd, J = 10.1, 3.8 Hz, 1H, Ar), 7.02 (d, J = 0.5 Hz, 1H, Ar), 6.72 (d, J = 1.1 Hz, 1H, Ar), 6.33 (d, J = 1.9 Hz, 1H, Ar), 3.85 (s, 3H, OCH3). 13 C NMR (150 MHz, CDCl3) δ 158.79 (s), 157.02 (s), 153.95 (s), 149.15 (s), 130.55 (s), 128.72 (s), 127.92 (s), 124.28 (s), 112.35 (s), 97.95 (s), 97.76 (s), 89.02 (s), 55.79 (s). HR-MS: m/z (%) [M + H]+ 241.0575.</p><!><p>The ChEs inhibitory activity of the synthesised 2-arylbenzofuran compounds was evaluated by Ellman's method31 with slight modifications. Acetylthiocholine iodide and S-butyrylthiocholine iodide were used as substrates. 5,5′-Dithio bis-(2-nitrobenzoic acid) (DTNB) was used as a reagent. Donepezil was used as a positive control. Add 120 µL phosphate buffer solution (0.1 mol·L−1, pH = 8.0, PBS), 20 µL DTNB (3.3 mmol·L−1 in 0.1 mol·L−1 PBS, pH = 8.0), 20 µL AChE solution (0.2 U·mL−1 in 0.1 mol·L−1 PBS, pH = 8.0), and 20 µL sample solutions of different concentrations to the 96-well plate in sequence, shake well, and then incubate at 37 °C for 5 min. Then, add 20 µL of substrate (5 mmol·L−1 in 0.1 mol·L−1 PBS, pH = 8.0), shake well, and incubate at 37 °C for 20 min. Set the sample solution to 4 concentration gradients, and repeat the experiment 3 times. Use a microplate reader to measure the absorbance of the sample at 412 nm, and calculate the ChEs inhibition rate and IC50 value of each sample according to the formula. ChEs inhibitory effect (%) = [A0− (A1− A2)]/A0×100%,</p><p>where A0 is the absorbance of the blank group; A1 is the absorbance of the sample group; A2 is the absorbance of the sample blank group.</p><!><p>Set the fluorometer on well plate reader mode with excitation at 320 nm and emission at 405 nm. Bring all components (except the BACE1 Enzyme Solution) to room temperature. Add components to a fluorometer 96 well plate according to Table 3. Mix well by gentle pipetting. Add the BACE1 Enzyme Solution just before reading. Read the fluorescence immediately after adding the enzyme. This is "time zero" reading. The signal in the wells could increase between the addition of enzyme and this initial reading. Cover the plate with parafilm and incubate at 37 °C for 2 h. Read the signal at "time zero" plus 2 h. The plate should be at room temperature before reading. After the readings are made, add 40 µL of Stop Solution. The addition of the Stop Solution will stabilise the signal for at least 24 h. The final reading of each well is the "2 h" reading minus "time zero" reading.</p><!><p>Reaction scheme.</p><!><p>The Positive Control group represents 100% enzyme activity. Calculate the BACE1 inhibition rate and IC50 value of each sample according to the formula. BACE1 inhibitory effect (%) = [1 – A2/A1] × 100%,</p><p>where A1 is the reading of the Positive Control group; A2 is the reading of the Inhibition group.</p><!><p>The compound that showed better inhibitory activity against ChEs was selected for kinetic measurements. To obtain the mechanism of action of compound, reciprocal plots of velocity versus substrate were constructed at different substrate solution concentrations by using Ellman's method31. The experiment were measured at eight different substrate solution concentrations (0.039, 0.078, 0.156, 0.312, 0.625, 1.25, 2.5, 5 mM) and four different concentrations of target compound (0, 1, 10 and 100 nM). To a 96-well plate, phosphate buffer solution, 5,5-dithiobis-2-nitrobenzoic acid, AChE solution and target compound were added sequentially, shaken well, and incubated at 37 °C for 5 min. Then add the substrate, shake well, immediately use the microplate reader to measure the absorbance of the sample at 412 nm, and repeat the experiment three times. Lineweaver–Burk plot was made based on the reaction rate and substrate concentration to determine the type of inhibition of the ChEs by the compound.</p><!><p>The cytotoxicity test was carried out using the MTT assay34. The detection principle is that the succinate dehydrogenase in the mitochondria of living cells can reduce the exogenous MTT to water-insoluble blue-purple crystal formazan and deposit it in the cells, while dead cells have no such function. Dimethyl sulfoxide (DMSO) can dissolve formazan in cells, and its absorbance value (OD value) is measured at 490 nm wavelength with a microplate reader, which can indirectly reflect the number of living cells. The greater the OD value, the stronger the cell viability and the lower the toxicity of the compound.</p><p>Collect adherent cells, adjust the concentration of cell suspension, add 100 µL to each well, and pave the plate so that the density of the cells to be tested is 5000–10,000 per well. Pave the plate in the afternoon of the previous day, add compounds in the morning of the next day, set up 5 gradients, each hole is 100 µL, and set up 3 duplicate holes. Incubate in an incubator for 48 h, add 10 µL MTT solution to each well, and continue to incubate for 4 h. After terminating the culture, aspirate the culture medium in the wells, add 150ul DMSO to each well, and place on a shaker to shake at low speed for 10 min to fully dissolve the crystals. Measure the OD value of each well at 490 nm of the microplate reader, and calculate the cell survival rate according to the formula. Cell survival rate (%) = A1/A2× 100%,</p><p>where A1 is the OD value of the sample group; A2 is the OD value of the blank group.</p><!><p>Reactive oxygen species (ROS) include superoxide free radicals, hydrogen peroxide, and its downstream products peroxides and hydroxylates, etc., and participate in cell growth and proliferation, development and differentiation, ageing and apoptosis, as well as many physiological and pathological processes35. The ROS detection kit uses the fluorescent probe DCFH-DA for reactive oxygen detection. Dilute DCFH-DA with DMEM medium at a ratio of 1:1500. Remove the cell culture medium, add DCFH-DA diluted in an appropriate volume, and incubate for 20 min in a cell incubator. The cells were washed three times with DMEM medium to sufficiently remove DCFH-DA that did not enter the cells. The cells of each group were directly observed with a laser confocal microscope.</p><!><p>In summary, a series of 2-arylbenzofuran derivatives have been designed, synthesised and evaluated as multi-target anti-AD drugs, which have ChE inhibitory activity and BACE1 inhibitory activity. According to the test results of inhibition of ChEs, most of the compounds have dual ChEs inhibitory activity, and their effect may be better compared with selective ChE inhibitors. Compound 20 has superior BACE1 inhibitory activity than other compounds. Compared with other compounds, R4 and R6 of compound 20 each have a hydroxyl substituent, which may be the reason for its high activity. We selected compound 20 to study the kinetics of ChEs inhibition. According to kinetic experiments, the type of ChEs inhibitory effect of compound 20 is reversible, indicating that our design strategy is reasonable. The compound has extremely low toxicity to normal cells and has almost no effect. Compound 20, which has the best activity in all aspects, can significantly reduce the level of ROS in AD model cells. Our research shows that the number and position of hydroxyl groups on the aromatic ring are important. Among the tested compounds, the two hydroxyl substituents in R4 and R6 have the strongest inhibitory effect on ChEs. We also speculate that the meta-substitution of the hydroxymethoxy group promotes the increase of the activity of the compound. In conclusion, our experiments on 2-arylbenzofuran provide an idea for the development of drug design for treatment or prophylaxis for AD.</p><!><p>Click here for additional data file.</p><!><p>The authors declare that they have no competing interests.</p>

PubMed Open Access

Advances in the directed evolution of proteins

Natural evolution has produced a great diversity of proteins that can be harnessed for numerous applications in biotechnology and pharmaceutical science. Commonly, specific applications require proteins to be tailored by protein engineering. Directed evolution is a type of protein engineering that yields proteins with the desired properties under well-defined conditions and in a practical time frame. While directed evolution has been employed for decades, recent creative developments enable the generation of proteins with previously inaccessible properties. Novel selection strategies, faster techniques, the inclusion of unnatural amino acids or modifications, and the symbiosis of rational design approaches and directed evolution continue to advance protein engineering.

advances_in_the_directed_evolution_of_proteins

Introduction<!>Advancing selection technologies<!>Maximizing library quality<!>Refining selection steps<!>Improving speed of selection<!>Photoresponsive binders<!>Selecting unnatural peptide binders<!>Multi-subunit protein selections<!>In vitro compartmentalization selections applied to new reactions<!>Combining directed evolution with rational design<!>Conclusion

<p>Synthetic biology describes the engineering of biological parts and whole systems by either modifying natural organisms or building new biosystems from scratch. To date, most proteins used as parts in synthetic biology are taken from nature. Utilizing naturally evolved proteins has led to numerous successful applications in biotechnology. Nevertheless, these applications invariably benefit from an optimization of the original natural proteins by protein engineering [1]. In contrast, building entirely artificial proteins that do not resemble natural proteins is still a major challenge [2–4] and therefore much less common than the engineering of natural proteins for new or improved properties.</p><p>Protein engineering has developed into a multi-faceted field with hundreds of publications in the last two years alone. This field encompasses a variety of approaches for creating desired protein properties, ranging from purely computational design to selecting proteins from entirely random polypeptide libraries. Due to the incredible breadth of the field, and to enable us to focus on recent advances, we will direct the reader to excellent reviews on the fundamentals of directed evolution technologies [5–9] and computational protein design [10–12]. This review will therefore focus on the latest developments in the directed evolution of proteins (Figure 1).</p><!><p>In any directed evolution experiment, the isolation of the desired protein from a library of gene variants is the crucial step. Many efforts have been made to push the boundaries of evolution schemes, attempting to create better protein libraries, new selection systems with improved features, and faster selection procedures (Figure 2).</p><!><p>The chance of discovering desired protein variants is directly related to the quality and complexity of the starting library. For example, random mutations that destabilize a protein can be detrimental. Therefore, building libraries with a high potential of containing functional proteins is vital. 'Smarter' libraries have been pursued that are less complex but of high-quality [13]. To build those libraries, targeted mutagenesis guided by structural or phylogenetic information, the use of compensatory stabilizing mutations and other approaches have successfully been applied [14,15]. Alternatively, a library maximizing complexity while enriching for well-folded proteins was constructed based on one of nature's most common enzyme folds, the (β/α)8 barrel fold. All residues on the catalytic face of the protein scaffold were randomized and, simultaneously, the library was enriched for protease resistance by an mRNA display selection, which has been correlated with well-folded and therefore more likely functional proteins [16•].</p><!><p>In vivo directed evolution of membrane proteins has been challenging due to toxicity of either the membrane protein or the selection conditions. Liposome display is a new method that has enabled in vitro directed evolution of toxic integral membrane proteins [17••]. This approach creates giant unilamellar liposomes and encapsulates a single DNA molecule along with a cell-free translation system. Each liposome will therefore display many copies of a single variant. Coupling protein activity to a fluorescent signal enables subsequent sorting by fluorescence-activated cell sorting (FACS). This approach was applied to evolve an α-hemolysin mutant with pore-forming activity 30-fold greater than wild-type. In addition to membrane protein toxicity, selection conditions can be challenging when using lipid-based barriers, for example when selecting for stability in detergent. To overcome this issue, a cellular high-throughput encapsulation, solubilization, and screening method (CHESS) was developed to screen a library of G-protein coupled receptor (GPCR) variants [18•]. GPCRs are an important group of drug targets. A library of 108 variants was expressed in E. coli and the cells were then encapsulated in a polymer. The cells were lysed, but the "nano-container" trapped the GPCR variants along with their encoding DNA while allowing free diffusion of fluorescent ligands and thereby enabling FACS. With this technique, functional receptors were identified in the presence of the detergent of choice.</p><p>The use of bead display for directed evolution has been limited by very few copies of DNA or displayed protein [19–23]. Recently, a "megavalent" bead surface display (BeSD) system was developed to allow the display of protein and its encoding DNA in defined quantities up to a million copies per bead [24]. This method combines advantages of in vitro selections with multivalency of in vivo display systems, enabling the ranking and sorting of the output variants of an in vitro selection by flow cytometry.</p><p>Protease enzymes have a tremendous potential in medicine and biotechnology but engineering their activities via directed evolution for altered specificity, instead of simply broadening activity, has been successful until recently in only a few select cases using E. coli cell surface display of the E. coli outer membrane protease T [25]. This system is limited to the relatively few bacterial proteases that can be displayed and active on the prokaryote's cell surface. To enable the engineering of more complex mammalian proteases, yeast surface display was modified to evolve novel protease specificity. In the revised system, both the protease variants and a yeast adhesion receptor were colocalized inside the endoplasmic reticulum (ER) through attached signal sequences [26••]. Successful proteolytic cleavage of a linker region detached the ER retention signal and enabled the yeast surface display of the adhesion receptor including its FLAG tag, which was then identified by anti-FLAG antibodies. Counter-selection tags were also incorporated to improve the selectivity of resulting protease variants. This method was used to alter the specificity of tobacco etch virus protease, as well as granzyme K and hepatitis C virus protease, and was even modified to demonstrate in principle the selection of kinase activity.</p><p>A directed evolution approach was devised to improve the targeting specificity of an engineered methyltransferase. Methylation of only a single site in a target DNA was selected for by digesting with a target site-specific restriction endonuclease and a second, unusual restriction enzyme that digests DNA with two distally methylated sites [27]. This method identified methyltransferase variants that showed 80% methylation at the target site and less than 1% methylation at off-target sites.</p><p>Phage assisted continuous evolution (PACE) enables the sustained evolution of protein variants through hundreds of rounds of evolution in a week with little researcher intervention [28]. This method was used to probe evolutionary pathway independence by evolving RNA polymerases for various promoter specificities [29•]. RNA polymerases that initially recognized the T7 promoter were evolved to recognize T3 or SP6 promoters separately, and then a final hybrid promoter of T3 and SP6. The resulting RNA polymerases from the SP6 pathway were ~3–4-fold more active than those from the T3 pathway and further evolution did not diminish this gap. Sequencing at multiple steps along the evolutionary path further illuminated that the divergent populations were unable to converge to the same solution. This suggests that it may be beneficial to evolve through multiple subpopulations instead of a single large population. In additional work, PACE was improved to allow the modulation of selection stringency via engineering phage propagation to be dependent on the small molecule anhydrotetracycline [30•]. Further, the authors enabled counterselection to refine promoter specificity. The combination of these methods was used to create RNA polymerases with a 10,000-fold net change in promoter specificity. While this method now enables fast selections with advanced features such as counter-selection, it still can only be used to evolve proteins or activities that directly or indirectly involve expression, such as polymerase activity.</p><p>The use of chaperonins such as GroEL and GroES during directed evolution has been shown to allow more destabilizing mutations and mutations in the protein core to survive during evolution by stabilizing folding intermediates [31,32]. This chaperonin system was used to characterize the evolutionary pathway for a natural phosphotriesterase to a novel arylesterase [33•]. This study demonstrated for the first time on a molecular level how mutations found early during an evolutionary optimization yield larger improvements than later mutations. The results suggest that mutations seem to initially cluster near the active site and then radiate towards the rest of the enzyme to stabilize the early mutations.</p><!><p>Two different strategies have significantly expedited the directed evolution process for in vitro selections. In the first strategy, many rounds of selection were performed very quickly, followed by sequencing and characterization of relatively few output sequences. For this purpose, a modified version of mRNA display, named "TRAP display" was devised where the puromycin linker was attached simply via base pairing instead of covalent modification, enabling a round of selection in as little as 2.5 hours compared to the traditional 2–3 days [34•]. In just 14 hours and 6 rounds of selection, macrocyclic peptides with low nanomolar affinity against human serum albumin were selected.</p><p>In the second strategy, only a single round of stringent selection was performed, but then a large number of selected clones were analyzed by high-throughput sequencing to enable population-level statistical analysis. Following this approach, nanomolar affinity binders were identified after a single round of selection from the small protein scaffold 10Fn3 (10th fibronectin type III domain) with two random sequence loops, using the continuous flow magnetic separation mRNA display technology [35••]. The key to this approach was identifying the clones with the most enriched copy numbers after selection. In another example of the same strategy, multiple rounds of phage display biopanning of a heptapeptide library was performed against target cells, but high-throughput sequencing was performed at each step to assess the value of each round [36••]. Overall, a single round of screening was capable of identifying the best binders when sequenced at sufficient depth; and multiple rounds were only helpful in decreasing the background of non-binders when sequencing a small pool.</p><!><p>Biological systems engineered to use light-sensing components have attracted attention due to their vast potential applications and have been recently reviewed [37]. Similar efforts have developed photo-reactive peptide aptamers for future use with in vitro or in vivo photoregulation, immunoassays, or bio-imaging analyses. Ribosome display of peptides that contained azobenzene-modified lysine enabled the selection of UV-responsive streptavidin binders [38]. A different approach used a ribosome display scheme with a benzoxadiazole-modified phenylalanine to select for calmodulin binding peptides with single-digit micromolar affinity that fluoresce upon binding [39]. Two laboratories created cyclized peptides via azobenzene linkers to select for light-responsive peptides by phage display technology. A peptide cyclized via azobenzene-modified cysteines flanking a randomized 7 residue peptide was capable of binding to streptavidin in the dark and cease binding upon irradiation with a 22-fold discrimination [40]. Likewise, a synthesized photoswitchable azobenzene-based cyclization compound enabled the identification of peptide binders with single-digit micromolar affinity and a 3-fold change in affinity upon UV exposure [41]. Photoresponsive peptide ligands can now be selected with large UV-induced binding affinity changes. In the future, in vivo activity of evolved ligands needs to be demonstrated to further their use in optogenetics.</p><!><p>Binding peptides are valuable for a variety of purposes including detection assays and therapeutic applications. mRNA display and cDNA display have been used recently to create short peptide aptamers with unusual structure or composition, creating new diversity and thereby enabling the selection of binders with unique properties. In one approach, up to 12 unnatural amino acids were incorporated into an mRNA-displayed peptide library using a custom-mixed cell-free PURE translation system that was reconstituted from the purified components necessary for E. coli translation. The peptides were then cyclized via cysteine residues and the selection yielded unnatural peptides with nanomolar binding affinities for thrombin [42]. A related approach created mRNA-displayed lantipeptides by using a translation system where lysine was substituted with 4-selenalysine, and inducing post-translational elimination via H2O2 and dehydroalanine. This provided an alternative cyclization mechanism for a drug candidate library and yielded binders with low micromolar affinity for Sortase A [43•]. Similarly, a cDNA display library was cyclized [44] and a phage display library was bi-cyclized [45] through disulfide bonds formed in cysteine-rich peptides and used to select for peptide aptamers. Additional work on macrocyclic peptide selections has been recently reviewed [46].</p><p>Unnatural amino acids have also been used to decorate peptides and evolve multivalent glycopeptides [47]. Alkyne-containing glycine residues were incorporated into an mRNA-displayed peptide library to enable glycosylation via click chemistry. A selection against a broadly-neutralizing antibody of HIV identified glycopeptides with potential as vaccine candidates.</p><!><p>Multi-subunit proteins are common, but directed evolution of these has been limited to in vivo methods, as opposed to in vitro methods that are capable of screening much larger libraries. To address this issue, an mRNA-displayed Fab fragment was entrapped in emulsion PCR enabling the in vitro selection of heterodimeric Fab fragments [48•]. However, the emulsion step reduced the library complexity to a similar range as in vivo methods. To achieve an in vitro selection of multi-subunit proteins of potentially up to 1014 variants, ribosome display was performed where one Fab subunit was randomized while the other subunit was held constant [49•]. This was carried out with both heavy chain and light chain libraries yielding tight binders to VEGF and CEA. While cell-surface display of multi-subunit proteins has been performed before, a notable advance is the description of a mammalian cell-surface display that also features a titratable secretion of the same Fab fragments through alternate splicing of the pre-mRNA [50].</p><!><p>The performance of in vitro compartmentalization methods has been improved in recent years with reported screening speeds of 2,000 droplets per second for water-in-oil emulsion screening [51•]. A number of creative protocols have been developed for this technology and applied to evolve enzymes capable of a range of chemical reactions. This includes a generalizable screen for hydrogenase activity [52], a selection for meganuclease specificity [53], an entirely microfluidic screen for hydrolytic activity of a sulfatase [54•], and a quantitative screen for glucose oxidase activity [55]. While the above methods predominantly used custom-made microfluidic chips to sort their water-in-oil emulsions, another study described a generalizable method to produce water-in-oil-in-water emulsions that can be sorted by standard fluorescence-activated cell sorting (FACS) equipment [56•]. This protocol enabled the generation of monodisperse double emulsions at 6–12,000 droplets per second, which can be stored for months to years and manipulated to adjust their volume as needed. These droplets can be sorted in a commercial FACS-machine at 10–15,000 droplets per second, while enriching active variants by up to 100,000 fold. The downside to most IVC methods is that fluorescence must be linked to product formation.</p><!><p>The combination of rational design to create informed libraries of variants with directed evolution to refine activity and efficiency has become a recurrent theme in protein engineering. A good example is the optimization of a computationally designed Kemp eliminase by directed evolution. Through rounds of error prone PCR, DNA shuffling, and site-directed mutagenesis, this artificial protein was refined to yield an enzyme that accelerated the reaction 6 × 108-fold, approaching the efficiency of natural enzymes [57••].</p><p>In another study evolving an unrelated artificial Kemp eliminase, stabilizing consensus mutations were added during the library generation process. This stabilization facilitated the identification of a variant with >2,000-fold improved catalytic efficiency after rounds of DNA shuffling, error prone PCR and selection [58•]. Furthermore, an artificial Diels-Alderase was evolved by combining mutations from different rational design variants with rounds of error prone PCR. The modestly active original enzyme was thereby turned into a proficient biocatalyst for this abiological [4+2] cycloaddition reaction [59].</p><p>Cytochrome P450-derived enzymes can perform a variety of reactions and have been engineered to improve multiple properties [60]. A collection of cytochrome P450 mutants was screened for cyclopropanation of styrenes, and optimized through informed site-directed mutagenesis [61••]. This enzymatic activity has not been observed in nature but is very useful to synthetic chemists. The work presents a great example how catalytic promiscuity of enzymes can be exploited. Using chemical intuition, the active site of one of the collection of cytochrome P450s was rationally re-designed to change the reduction potential of the heme-bound FeII/III, which allowed the efficient NAD(P)H-driven cyclopropanation while suppressing the native monooxygenation activity [62]. The modification therefore enabled the use of the P450 variant as a whole-cell catalyst. In another case, high regio- and stereoselective hydroxylation of unactivated C-H bonds by a cytochrome P450 enzyme was achieved through a creative combination of active site mutagenesis, high-throughput "fingerprinting" to identify functionally diverse variants, and fingerprint-driven reactivity predictions [63••].</p><p>A mononuclear zinc metalloenzyme was computationally redesigned and then evolved with a combination of saturation mutagenesis and error prone PCR to create a variant that was 2,500-fold better than the initial design [64]. Structure-guided design of libraries of paraoxonase 1 and directed evolution led to variants capable of up to 340-fold higher catalytic efficiency for toxic isomers of G-type nerve agents [65]. Similarly, iterative saturation mutagenesis was used to evolve variants of phosphotriesterase capable of hydrolyzing V-type nerve agents with a 230-fold improvement of catalytic efficiency [66]. In a separate effort to create a toxin-neutralizing protein, a unique catalytic triad was computationally designed and subsequently optimized by yeast display. The resulting protein reacted with a fluorophosphonate probe at rates comparable to natural serine hydrolases, yet it was incapable of catalytic turnover [67].</p><p>Rational design has also been used to construct whole artificial protein scaffolds. In some cases, directed evolution was applied to these artificial structures to select for desired functions. Proteins from a combinatorial library of artificial four-helix bundle proteins were found to function in vivo by rescuing E. coli strains that lacked a conditionally essential gene [68]. In addition, select four-helix bundle proteins bound to heme and exhibited peroxidase activity. This activity was improved by random mutagenesis and directed evolution [69]. Using structural principles of natural repeat proteins, designed ankyrin repeat proteins (DARPins) have been built and shown to function as artificial antibody mimetics. Recently, a flexibility loop and additional randomized regions were incorporated, creating the LoopDARPin scaffold [70•]. A library based on this improved scaffold design yielded picomolar binding proteins after only a single round of selection by ribosome display.</p><!><p>With the investment of sufficient resources and determination, directed evolution generally appears to yield desired improvements of protein properties, sometimes producing remarkable results [71]. Therefore, one might be tempted to consider protein engineering a mature field. But those success stories mainly apply to improving or changing proteins that were provided by natural evolution. In contrast, the generation of novel activities without natural precedent is still in its infancy, although several examples have been reported [2,6,10,61,72,73]. Achieving the synthetic biology goal of integrating artificial proteins into biological systems will introduce additional challenges, which again can be overcome with the help of directed evolution [9,74].</p>

PubMed Author Manuscript

Omega-3 fatty acids in obesity and metabolic syndrome: a mechanistic update

Strategies to reduce obesity have become public health priorities as the prevalence of obesity has risen in the United States and around the world. While the anti-inflammatory and hypotriglyceridemic properties of long-chain omega-3 polyunsaturated fatty acids (n-3 PUFAs) are well known, their antiobesity effects and efficacy against metabolic syndrome, especially in humans, are still under debate. In animal models, evidence consistently suggests a role for n-3 PUFAs in reducing fat mass, particularly in the retroperitoneal and epididymal regions. In humans, however, published research suggests that though n-3 PUFAs may not aid weight loss, they may attenuate further weight gain and could be useful in the diet or as a supplement to help maintain weight loss. Proposed mechanisms by which n-3 PUFAs may work to improve body composition and counteract obesity-related metabolic changes include modulating lipid metabolism; regulating adipokines, such as adiponectin and leptin; alleviating adipose tissue inflammation; promoting adipogenesis and altering epigenetic mechanisms.

omega-3_fatty_acids_in_obesity_and_metabolic_syndrome:_a_mechanistic_update

Introduction<!>Adipose tissue and obesity<!>Synthesis and metabolism<!>Sources of omega-3 PUFAs<!>Dietary omega-3 PUFA intake, obesity and metabolic disorders<!>Omega-3 PUFA in animal studies<!>Omega-3 PUFA, energy intake and obesity<!>Omega-3 PUFA and insulin resistance<!>Animal study conclusions<!>Omega-3 PUFA and weight loss in humans<!>Omega-3 PUFA in combination with dietary interventions<!>Omega-3 PUFA and exercise<!>Limitations in human studies<!>Mechanisms by which n-3 PUFA improve adiposity and metabolic disorders<!>Omega-3 PUFA and adipogenesis<!>Adipose tissue inflammation<!>Adipokine secretion<!>Appetite suppression<!>Insulin resistance<!>Lipid metabolism<!>Thermogenesis<!>Lean mass<!>Epigenetics and microRNA<!>Future perspectives<!>Conclusions

<p>The American Medical Association recognizes obesity as a disease [1] and considers it a major public health problem. In the United States, 36.5% of adults are obese [2], while approximately 39% of the world's adult population is overweight and more than 13% are obese [3]. Obesity increases morbidity risks for heart disease, type 2 diabetes mellitus (T2DM) and some types of cancer [4]. Metabolic changes associated with these diseases comprise metabolic syndrome (MetS), which is diagnosed when three of the following five conditions exist: abdominal obesity, elevated triglycerides (TG), reduced high-density lipoprotein (HDL) cholesterol, high blood pressure and elevated fasting blood glucose [5].</p><p>Lipids are key macronutrients in the human diet. The type and proportion of dietary fatty acids consumed impact health and whole-body physiology [6]. Research has shown that saturated fatty acids (SFAs) are detrimental to health, while monounsaturated (MUFAs) and polyunsaturated fatty acids (PUFAs) offer health benefits [7]. In the diet, fatty fish and fish oil rich in omega-3 (n-3) PUFAs, such as eicosapentaenoic acid (EPA) or docosahexaenoic acid (DHA), have demonstrated cardioprotective, anti-inflammatory and hypotriglyceridemic properties. Hence, these fatty acids may assist in the treatment and prevention of obesity comorbidities, especially by improving individual components of the metabolic syndrome [7–9]. Therefore, the effect of n-3 PUFAs on body weight and body composition is of particular interest.</p><p>In this review, we provide an update on the effects of n-3 PUFAs on obesity and MetS in both animal and human studies, highlighting potential mechanisms for n-3 PUFAs in reducing body weight, improving body composition and counteracting the adverse metabolic consequences of obesity.</p><!><p>Body fat is primarily stored in adipose tissue, a connective tissue composed of adipocytes, preadipoctyes, vascular endothelial cells, fibroblasts and various types of immune cells, including adipose tissue macrophages [10]. Adipose tissue is an active endocrine organ that secretes numerous hormones, including leptin and adiponectin, and cytokines (adipokines) such as interleukin (lL)-6 [11]. Three major types of adipose tissue have been identified: white adipose tissue (WAT), brown adipose tissue (BAT) and beige ("brite") adipose tissue. WAT is primarily responsible for energy storage in the form of TG and the release of fatty acids during periods of fasting; it is mainly located in two distinct depots, as subcutaneous adipose tissue or visceral adipose tissue [10]. The adipocytes of visceral fat surrounding internal organs are more metabolically active than those of subcutaneous adipose tissue and thus contribute to the risks of cardiovascular disease and T2DM [12]. BAT plays a key role in thermogenesis and is mainly found above the clavicle and scapula in adults [12]. Obesity leads to adipose tissue dysfunction, which is mechanistically linked to the pathogenesis of insulin resistance in the liver and in skeletal muscle (Fig. 1) and may result in MetS [13].</p><p>While weight loss via lifestyle modification is the primary treatment in the management of obesity and its comorbidities, compliance is difficult. Adjunct treatments for the management of obesity include pharmaceuticals [14], surgery [15] and dietary supplements [16]. Despite these measures, however, the prevalence of obesity has continued to rise. Thus, alternative strategies to assist in weight loss and reduce body fat are necessary. Natural bioactives such as n-3 PUFAs present few side effects and so may be safer than other modalities for the treatment of obesity. This review summarizes current basic and clinical research and mechanistic insights regarding the effects of n-3 PUFAs on obesity.</p><!><p>The human body can synthesize many fatty acids, but not linoleic acid (LA; omega-6; C18:2 n-6) or α-linolenic acid (ALA; C18:3 n-3), which must be consumed in the diet. ALA is the precursor for EPA (C20:5 n-3),docosapentaenoic acid (DPA; C22:5 n-3) and DHA (C22:6 n-3) in the human body (Fig. 2) [6]. Many studies have found low conversion rates of ALA to EPA and DPA, and little to no DHA synthesis [17,18]; hence, any direct benefits of these very long chain fatty acids depend on dietary intake [17]. Both dietary intake and fatty acid desaturase activity determine plasma n-3 PUFA levels [19]. A balanced n-6:n-3 fatty acid ratio (1:1 to 2:1 is optimal) is important for homeostasis and normal development throughout the lifespan. High n-6 PUFA intake in the Western diet increases the n-6:n-3 ratio to a range from 10:1 to 20:1 and may play a role in the pathogenesis of obesity and related diseases [19,20].</p><!><p>Dietary sources of n-3 PUFAs are much less abundant than n-6 PUFAs. ALA is synthesized by plants from LA and can thus be found in green leafy vegetables, seeds such as flaxseed (linseed), nuts and legumes. Vegetable oils such as sunflower, corn, perilla, canola and soybean also provide ALA but are much more abundant in LA. Fish, such as salmon, tuna, trout, mackerel, anchovy, bluefish, herring, mullet, sturgeon and sardines, are also rich sources of n-3 PUFAs, particularly EPA and DHA. Lean fish, such as cod, store lipids in their liver, and for this reason, cod liver oil is a good source of n-3 PUFAs. Fatty fish, such as salmon, mackerel, sardines and tuna, store lipids throughout their bodies and are good whole sources of n-3 PUFAs [6].</p><p>The 2015 Dietary Guidelines for Americans recommend consuming about 8 oz per week of seafood, which would provide about 250 mg/day of EPA and DHA [21]. The recommended intake for n-3 PUFAs corresponds to consuming fish twice weekly, including one serving of oily fish. Even though there is ample evidence for a role of n-3 PUFAs in modulating chronic diseases, an optimal dose has not been agreed upon, and recommendations vary based on governing body. The U.S. Food and Drug Administration has stated that levels up to 3 g/day are generally recognized as safe [22], although other authorities have reported no adverse effects at up to 5–6 g/day [23]. It has been suggested that the bioavailability of n-3 PUFAs is improved by emulsification. Emulsified n-3 PUFA is more easily exposed to pancreatic lipase and colipase, enhancing its digestion. Additionally, emulsified n-3 PUFA is easily transported into enterocytes, thus increasing fatty acid absorption [24,25].</p><!><p>Dietary fish intake is considerably higher in people of the circumpolar arctic regions and relatively much lower in those living in the United States, Australia, France and the United Kingdom. Fish intake closely reflects n-3 PUFA consumption, with intakes of approximately 3 to 4 g/day by Eskimos, 5 to 6 g/day by Japanese, 0.189 g/day by Australians and 0.25 g/day by Europeans and North Americans [26,27]. After it was reported that Japanese and Eskimo populations had healthier metabolic profiles associated with elevated plasma n-3 PUFA levels attributed to high fatty fish intake [28], many prospective studies began to examine whether fish or fish oil intake prevents the development of obesity.</p><p>The Health Professional Follow-up Study suggested that men with a high level of fish consumption were less likely to be overweight [29]. In contrast, the Nurses' Health Study found that women with higher fish intake (two or more fish meals per week) had a higher risk of being overweight [30]. In China, data from the Shanghai Women's and Men's Health studies found similar indices of body mass among groups of varying fish intake [31]. Clearly, findings of prospective studies regarding the beneficial effects of fish intake on obesity are far from agreement. The evident discrepancies may have arisen due to differing or inadequate methods of data collection on fish intake (food frequency questionnaires), differences in cooking methods and other unaccounted for lifestyle practices (exercise, etc.) from study to study and among different study populations.</p><p>Plasma, erythrocyte and tissue n-3 PUFA concentrations are largely determined by consumption and thus may be taken to accurately reflect n-3 PUFA consumption. Plasma levels of fatty acids reflect recent intake, whereas tissue levels of fatty acids reflect long-term intake [32,33]. Erythrocyte fatty acid content (i.e., the omega-3 index) correlates with fatty acid intake and parallels tissue concentrations and thus is more reflective of long-term intake [34]. An increase in n-6: n-3 ratio [35] and overall lower serum phospholipid n-3 concentrations, particularly of DHA have been associated with obesity, specifically waist circumference measures [36] in obese adolescents [35] and obese adults [36]. Thus, prospective studies on the relationship between plasma and erythrocyte fatty acid content and the long-term risk of obesity are warranted to clarify this issue.</p><!><p>Animal studies performed to investigate the antiobesity effects of n-3 PUFA have used a variety of models, diet compositions, and n-3 PUFA compositions and doses. Such differences across experimental designs complicate the interpretation of their results into cohesive and conclusive findings (Table 1).</p><!><p>Studies relating effects on body weight and energy intake with n-3 PUFA supplementation are inconsistent; while some show either decreased [37,38] or increased energy intake [39], most show unchanged energy intake with the addition of n-3 PUFA or with varying n-3 PUFA doses [40–57]. Only one study performed with female mice reported decreased energy intake with no significant effect on body weight [38].</p><p>Supplementation with n-3 PUFA prevented high-fat (HF)-diet-induced weight gain in a number of rodent studies, most of which supplemented an HF diet provided from the start, concurrent with inducing obesity [41,43,45,48,49,51]. Several studies have utilized a design that investigated the effectiveness of n-3 PUFA to reverse diet-induced weight gain and related metabolic changes by adding n-3 PUFA to the HF diet at approximately midstudy [46,47,58,59]. Our lab found that mice on an EPA reversal diet (6 weeks of HF followed by 5 weeks of HF-EPA) had body weights similar to mice fed the HF-only diet [47]. In a different study, body weight decreased significantly at the beginning of the 6-week reversal period, returning to weights similar to the low-fat-fed group for the remainder of the study (18 weeks total) [58]. Taken together, these studies suggest that antiobesity effects of N-3 PUFA in mice are predominantly seen when it is fed from the start rather than introduced after obesity is already established.</p><!><p>Obesity leads to insulin resistance, which is at least in part responsible for the pathogenesis of MetS [13]. Most weight loss interventions improve insulin resistance. Similarly, most animal studies document a beneficial effect of n-3 PUFAs on insulin sensitivity [60]. Since n-3 PUFAs induce weight loss in rodent models of obesity, it is difficult to state whether there are direct effects of n-3 PUFAs on insulin sensitivity. By contrast, we have shown weight-independent benefits of EPA on insulin sensitivity in HF-diet-induced obese C57BL/6J mice. These EPA-fed mice had significantly improved homeostatic model assessment of insulin resistance (HOMA-IR) scores when compared to HF-fed mice, despite similar body weights [47].</p><!><p>Thus far, most rodent studies have shown an antiobesity effect of n-3 PUFA, while fewer studies have found no change in body weight [37,39–41,44–46,49,51–57,59,61,62]. These studies do suggest that n-3 PUFA plays a role in reducing adipose tissue mass [40], particularly in the epididymal [39,42,43,46,49,52,53,56,57,61–64] and retroperitoneal locations [37,39,41,46,49,57]. Differences in the outcomes of studies on the effects of n-3 PUFA on body weight could be due to differing animal models of obesity (genetic vs. diet-induced obesity), the content of the diet (HF vs. high sucrose), the n-3 PUFA (EPA or DHA) formulation, the form of n-3 PUFA (TG form or as ethyl ester) provided or various combinations of these factors. Differences in n-3 PUFA dosage and duration may contribute to differences in outcomes as well. Failure to assess energy expenditure also limits meaningful comparisons. The combination of calorie restriction and n-3 PUFA supplementation may be the most effective strategy for reducing weight and improving body composition [54]. Interestingly, Ruzickova et al. extrapolated findings from their animal study to humans to suggest that with a daily intake of 100 g dietary fat, 11 g of EPA/DHA would be required to limit weight gain [42]. Few animal studies have considered translation to human studies since the amounts of n-3 PUFA, as EPA, DHA or both, in animal studies far exceed amounts feasible in humans [42,45,64]. It should be noted, however, that human studies of fish oil intake among Eskimo and Japanese populations have shown beneficial effects of these fatty acids even at intake levels below 11 g/day. Since these populations consume more fish and less red meat, it is plausible that a relatively lower arachidonic acid (AA) intake leading to a decreased n-6:n-3 ratio is contributing to the beneficial effects observed.</p><!><p>There are a variety of approaches to investigating the effects of n-3 PUFA on body weight, body composition and energy intake in human interventions that use different types of fish and varying levels of fish oil content, particularly EPA and DHA (Table 2). Fish and fish oil have also been used in addition to a variety of weight loss and dietary interventions of different durations with or without an exercise regimen. Participants have ranged from healthy to obese, with a variety of obesity-associated disorders, including T2DM, hyperinsulinemia and other features of MetS. The control, or type of placebo, which consists of assorted oils containing n-6 PUFA, such as sunflower, corn, soybean and paraffin oils, also varies among studies.</p><p>With n-3 PUFA supplementation alone, studies report no change in body weight (Table 2). One study in healthy adults supplemented with fish oil diets demonstrated decreased body fat mass, basal respiratory quotient and increased basal lipid oxidation when dietary intake was controlled [65]. Another study found reduction in fat mass along with significantly increased lean mass (fat-free mass) despite no alterations in total body mass, resting metabolic rate (RMR) or respiratory exchange ratio when compared to placebo supplementation [66]. Participant-reported diet diaries indicate significant reductions in carbohydrate, fat and total caloric intake with n-3 PUFA supplementation in one study [67], but others show no change in energy intake [68–71]. Since most studies only report total caloric intake, the effect of n-3 PUFA supplementation on macronutrient and energy intake should be repeated in larger studies to conclusively determine the role of n-3 PUFA in weight loss in humans.</p><!><p>Weight loss results appear more promising when n-3 PUFA supplementation is combined with calorie restriction (Table 3), but it is difficult to draw conclusions due to the variety of calorie restriction programs in different studies. Greater improvements in metabolic parameters, such as improved insulin resistance and decreased TGs, were attained with combined n-3 PUFA supplementation and calorie restriction compared to calorie restriction alone [72–74] or replacement of SFA [71]. Interestingly, results appear to be independent of the source, form or dose, i.e., different fish species (salmon, tuna, sardines, etc.) or fish oil capsules, of n-3 PUFA supplied [72,73,75]. This was confirmed by Thorsdottir et al., who compared the effects of various fish (cod or salmon) and fish oil (DHA/EPA capsules) in conjunction with 30% calorie restriction on weight loss in young, overweight adults for 8 weeks. After 4 weeks, men receiving cod, salmon or fish oil capsules lost approximately 1 kg more than those on 30% calorie restriction alone. The fish species and fish oil capsules supplied various amounts of n-3 PUFA: 0.3 g/day from cod, 3.0 g/day from salmon and 1.5 g/day from fish oil capsules, yet the n-3 PUFA dose did not influence weight loss outcomes. This suggests that variations in weight loss benefits may not depend solely on variations in n-3 PUFA dosages from study to study [76].</p><p>Rapid weight loss, induced by a very low calorie diet, alters adipose tissue and serum fatty acid composition [77,78]. Supplementation with n-3 PUFA during rapid weight loss increases serum n-3 PUFA concentrations [79] and may help prevent unfavorable changes in fatty acid tissue composition and essential fatty acid deficiency [77]. Some studies have utilized n-3 PUFA supplementation prior to a weight loss intervention, such as dietary restriction and/or an exercise regimen, and reported significant reductions in weight [80], while others observed no changes in body mass index (BMI) or body composition, particularly in insulin-resistant individuals [81]. Nonetheless, this type of study design should be refined and pursued further due to the relationship between tissue/plasma/erythrocyte n-3 PUFA concentrations and obesity. It will be important to verify if increasing n-3 PUFA concentrations prior to interventions would aid in weight loss and ameliorate obesity related metabolic dysfunctions.</p><!><p>Others have explored the influence of n-3 PUFA in conjunction with exercise and with or without a dietary intervention to determine if the addition of n-3 PUFA leads to greater weight loss (Table 4). With the addition of n-3 PUFA to an exercise regimen and dietary intervention, only one study has shown a decrease in body weight [82]. However, only a few such studies have been conducted, and the dietary intervention consisted of nutritional counseling rather than a prescribed diet [83,84].</p><p>The combined effects of n-3 PUFA and exercise are currently unknown. Well-designed placebo-controlled randomized clinical trials are lacking [85], as they require healthy and lean participants [86]. Differences in the intensity and forms of exercise (i.e., aerobic or resistance training) employed prevent valid comparisons across studies. The addition of n-3 PUFA to aerobic training without dietary intervention has resulted in decreases in fat mass [87]. Furthermore, the addition of n-3 PUFA to resistance training without dietary intervention resulted in increases in lean mass [88] and improved muscle quality [89].</p><!><p>Overall, findings on the effects of n-3 PUFA in humans are inconclusive. Improvements in study design and analyses could help resolve apparent inconsistencies in the effects of n-3 PUFA on weight and body composition. For example, sex, metabolic phenotype and geographic location should be taken into consideration in addition to n-3 PUFA supplementation. There is also a case for evaluating translation to real-world weight-loss diets, which are complicated by the need to control for eating behavior and physical activity. Even when participants are supplied with food, outpatient studies are difficult to translate because measurements of adherence to the recommended interventions [90] generally rely on self-reporting. Hence, studies using inpatient feeding and analysis of energy utilization should be carried out [91].</p><p>Failure to assess energy expenditure in human studies limits our understanding of the associations between n-3 PUFA status and decreased adiposity, weight loss and energy balance, especially when energy intake is unchanged. This highlights the need for tightly controlled studies, similar to that of Hall et al., to validate fuel partitioning and n-3 PUFA influence on metabolism in humans. Unfortunately, work of this nature is expensive, labor intensive and generally of short duration with small sample sizes [91].</p><p>Other limitations on current human studies of n-3 PUFA supplementation include the use of less reliable anthropometric methods [92] and failure to dose according to body weight to meet the threshold of tissue membrane n-3 PUFA phospholipid enrichment [93]. Finally, it is of utmost importance to utilize a standardized method, such as the omega-3 index, to assess n-3 PUFA status, the biological effects of n-3 PUFA and n-3 PUFA related metabolites [34]. Future human studies should employ this method to verify that n-3 PUFA consumption parallels n-3 PUFA concentrations in the body.</p><!><p>There are several proposed mechanisms by which n-3 PUFA could work in reducing body weight and improving the metabolic profile (Fig. 3). These include alterations in adipose tissue gene expression; changes in adipokine release; adipokine-mediated or adipokine-related pathways; appetite suppression; alterations in carbohydrate metabolism; increases in fat oxidation; increases in energy expenditure (possibly through thermogenesis); activating mechanisms involved in muscle anabolism; and, lastly, influence on epigenetics.</p><!><p>Adipose tissue expansion in obesity occurs via adipocyte hypertrophy (enlargement of adipocytes) and hyperplasia (increase in adipocyte number due to adipogenesis). The latter is associated with smaller adipocyte size and a metabolically healthy phenotype. Both n-3 and n-6 PUFAs can bind and/or regulate transcriptional factors that control genes involved in preadipocyte differentiation. PUFAs, particularly AA and its metabolites, serve as ligands for peroxisome proliferator-activated receptors (PPAR) gamma (PPARγ) and delta (PPARδ) to induce fat cell differentiation and accelerate maturation by elevating lipoprotein lipase expression in vitro [94,95]. Elevated concentrations of n-6 and n-3 PUFA in human subcutaneous tissue correlate with reduced adipocyte size; increased SFA concentrations lead to increased fat cell size [96]. Differences in fatty acid concentrations are more strongly associated with abdominal subcutaneous than visceral adipose tissue [35].</p><p>Studies performed in clonal adipocytes (3T3-L1) also demonstrate up-regulation in PPARγ expression, adipogenesis and lipid droplet formation after the addition of n-3 PUFA [43,97]. Taken together, these studies suggest that n-3 PUFAs promote adipogenesis and a healthy expansion of adipose tissue during positive energy balance, promoting a metabolically healthy phenotype.</p><!><p>Chronic low-grade inflammation and changes in adipokine patterns are key factors in the pathogenesis of metabolic derangements in obesity (Fig. 1). Indeed, a relationship exists between BMI, body fat percentage and inflammatory markers [98]. Omega-3 PUFAs inhibit nuclear transcription factor kappa B, a key transcription factor in cytokine gene expression and inflammation [99]. In humans and in vitro, n-3 PUFAs also have a documented role in reducing cytokines, including 1L-1 [100,101], 1L-6 [102] and TNF-α [100,103], which are all elevated in obesity. (For an extensive review of mechanisms of n-3 PUFA and adipose tissue inflammation, see Kalupahana et al., 2011).</p><p>Omega-3 PUFAs act as agonists for different members of the free fatty acid receptor family (FFARs) present on a variety of cell types involved in both energy homeostasis and the inflammatory response. A number of saturated and unsaturated long-chain fatty acids can activate FFAR1 and FFAR4 [104,105]. Agonist stimulation that impedes the inflammatory response occurs through activation of the G-protein-independent signaling pathway through interaction with β-arrestin proteins, which may further interact with the transforming growth factor kinase protein (TAK1) and binding protein (TAB-1). Stimulation of FFAR4 or β-arrestin inhibits lipopolysaccharide (LPS)-mediated release of inflammatory cytokines, including TNF-α and 1L-6 in the macrophage-like cell line RAW264.7. In fact, decreased macrophage infiltration into adipose tissue has been shown in mice fed an n-3-PUFA-enriched diet, possibly via activation of FFAR4 (G-protein-coupled receptor 120). Since n-3 PUFAs are unable to reduce adipose tissue macrophage infiltration in FFAR4 knockout mice, this highlights the mechanistic importance of FFAR4 in mediating the anti-inflammatory effects of n-3 PUFA [106]. Furthermore, fish oil supplementation (4 g n-3 PUFA/day) in obese humans has been associated with decreased M1 macrophage presence in adipose tissue and subsequent decreases in proinflammatory markers, such as IL-8 [107]. Accordingly, monocytes differentiate preferentially into M1 macrophages when treated with human postprandial triglyceride-rich lipoproteins following a meal rich in saturated fatty acids, versus a meal high in MUFA or PUFA, after which they shift towards M2 macrophages [108].</p><p>Furthermore, n-3 PUFAs halt inflammatory processes by inhibiting activation of the NLRP3 (nucleotide-binding oligomerization domain-like receptor; NLR family, pyrin domain containing 3) inflammasome via an arrestin-FFAR4-dependent pathway [109], which triggers a caspase-dependent cascade, resulting in the release of proinflammatory cytokines [110]. The n-3 PUFA DHA acts through FFAR1 or FFAR4 to suppress caspase-1 activity via formation of a β-arrestin-2/NLRP3 or NLRP1b complex and thus decrease the release of proinflammatory cytokines [109].</p><p>Omega-3 PUFAs also influence lipid rafts, which are cholesterol- and sphingolipid-rich areas of the plasma membrane [111] that can form signaling platforms [112,113]. Incorporation of n-3 PUFAs into plasma membranes disrupts lipid rafts [114] and hence could mediate anti-inflammatory and antichemotactic n-3 PUFA properties.</p><!><p>Several studies have shown that n-3 PUFAs modulate adipokine secretion. Obese individuals have high plasma leptin levels [115] suggestive of leptin resistance. Conversely, weight loss leads to parallel decreases in plasma leptin levels [116]. This weight-loss-associated decrease in leptin could contribute to hunger and a lower metabolic rate and ultimately lead to weight regain [117]. EPA supplementation attenuates the decrease in blood leptin levels that occurs during weight loss in obese women, suggesting a potentially significant role of EPA in weight loss maintenance [118]. Indeed, EPA increases the production of leptin in rodents and cultured adipocytes [37,119], suggesting a direct effect of n-3 PUFA on leptin production. However, the few studies that have assessed the role of n-3 PUFA in weight maintenance have found no significant effect on weight or blood leptin concentrations between n-3-PUFA-supplemented subjects compared to other weight loss groups [120,121]. Omega-3-PUFA-mediated effects on leptin are dependent on a number of factors, such as diet composition and energy balance, which could cause conflicting results.</p><p>Independent of body weight, both animal [37,48,122] and human [123,124] studies have found significant increases in blood levels of the insulin-sensitizing adipokine, adiponectin, following n-3 PUFA consumption. EPA appears to regulate adiponectin levels at the translational or posttranslational level rather than at the transcriptional level [123]. It has been proposed that the anti-inflammatory properties of n-3 PUFA supplementation induce an increase in adipocyte adiponectin production [123] and improve leptin sensitivity [125]. This type of interplay could have a significant influence on body weight regulation. An inverse relationship between serum adiponectin concentrations and TNF-α has also been demonstrated in ob/ob mice [123] and in overweight and insulin-resistant children following n-3 PUFA supplementation [126].</p><p>Fatty acid-binding proteins are cytosolic proteins that bind long-chain fatty acids and promote transport to several organelles. Fatty acid-binding protein 4 (FABP4; adipocyte FABP, A-FABP; or aP2) is secreted from both macrophages and adipocytes and functions as an adipokine [127]. An elevated FABP4 serum concentration is associated with obesity, insulin resistance and hypertension [128]. Adipocytes are the predominant contributors of circulating FABP4. During lipolysis, FABP4 functions in a nonclassical secretion pathway [129]. Omega-3 PUFA dose-dependently reduced FABP4 secretion in 3T3-L1 adipocytes and reduced serum FABP4 concentrations in humans [130]. Omega-3-PUFA-mediated reductions in FABP4 could also be due in part to reduced expression of transcription factors involved in adipocyte differentiation, including PPARγ2 and C/EBPα [130]. Another possible mechanism by which FABP4 levels are lowered by n-3 PUFA is through the β-adrenergic receptor [129] since n-3 PUFAs reduce sympathetic nerve activity and thus may lower FABP4 serum level [130]. Taken together, n-3 PUFAs modulate adipokine secretion by exerting anti-inflammatory effects and promoting a metabolically healthy phenotype.</p><!><p>In addition to leptin, central and peripheral peptides and hormones involved in food intake and energy expenditure signaling pathways are targets for n-3-PUFA-derived endocannabinoids and thus may be implicated in the prevention and treatment of obesity. A subanalysis of the study conducted by Thorsdottir et al. reported elevated sensations of fullness in the participants who consumed higher n-3 PUFA content meals (fatty fish and fish oil) compared to those who consumed lower n-3 PUFA content meals (control and lean fish) both immediately and 2 h after consuming the meal. Feelings of hunger were consistently lower in participants who ate the meal higher in n-3 PUFA content [131]. Therefore, it is possible that increased feelings of satiety following a meal high in n-3 PUFA content could aid weight loss by reducing subsequent food intake. Appetite suppression could also be mediated through FFAR4 (GPR 120). Omega-3 PUFAs are agonists for FFAR4 [132], which elicits the secretion of cholecystokinin, a peptide hormone that is synthesized and released from the small intestine and has roles in hunger suppression [133].</p><!><p>Adipose tissue inflammation is at least in part responsible for obesity-associated insulin resistance. Since n-3 PUFAs alleviate adipose tissue inflammation as outlined above, reducing adipose tissue inflammation is a possible mechanism for n-3-PUFA-associated improvements in insulin sensitivity observed in animal models.</p><p>Hepatic insulin resistance in which both glucose production and lipogenesis are increased is characteristic of the metabolic dysregulation seen in obesity and T2DM. This dysregulation is attributed to reductions in proximal insulin signaling kinases, such as P13K and AKT, which hinder gluconeogenesis, as well as activation of mTORC1 and p70S6K, which control lipogenesis [134,135].</p><p>Fibroblast growth factor (FGF) 21, which is produced by the liver, adipose tissue and skeletal muscle, has been shown to reduce both hepatic glucose production and plasma glucose levels, while it also increases insulin sensitivity and adipocyte glucose uptake [136]. Circulating FGF21 levels are elevated in diet-induced obese mice [137] and obese and type 2 diabetic humans [138], suggesting obesity-related FGF21 resistance. Omega-3 PUFAs attenuate HF-diet-induced increases in FGF21 [139] with associated reductions in hyperglycemia, hypertriglyceridemia and plasma insulin levels [140,141]. This could be a potential mechanism by which n-3 PUFAs improve insulin resistance.</p><p>Omega-3 PUFA supplementation prevents insulin resistance in muscle of rats fed an HF diet [142], partly by improving glycogen synthesis [143]. Omega-3 PUFAs also decrease fat content in muscle and maintain normal PI3K activity and expression and transcription of GLUT 4 receptors in muscle and thus improve myotubule glucose uptake. Omega-3 PUFAs also promote inhibition of hepatic glucose production [142].</p><p>Hence, n-3 PUFAs may be a valuable nutritional tool for preventing or diminishing muscular and hepatic insulin resistance associated with obesity. However, n-3 PUFAs appear ineffective once T2DM is established [144].</p><!><p>In both animal and human studies of n-3 PUFA supplementation, reductions in weight or fat mass were not accompanied by changes in energy intake (Tables 1–4). Omega-3 PUFAs can partition dietary fuel away from storage and toward oxidation by suppressing lipogenic genes and activating genes that encode for mitochondrial and peroxisomal fatty acid oxidation in both the liver and muscle.</p><p>Given their cardioprotective properties, n-3 PUFAs can improve endothelial function in patients with varying metabolic profiles [145], possibly through increased production of nitric oxide [146]. Furthermore, during exercise, fish oil has been shown to increase arterial dilation and blood flow to skeletal muscle [147]. Hence, improved blood flow may increase the delivery of fats to be utilized as energy in skeletal muscle, especially during exercise.</p><p>Regulation of lipid metabolism may vary by n-3 PUFA type, as well as by fat depot. For example, EPA is preferentially directed towards β-oxidation, while DHA and DPA are spared from catabolism and deposited in tissues [148]. Moreover, gene expression of fatty acid synthase [149], hormone-sensitive lipase, lipoprotein lipase and phosphoenolpyruvate carboxykinase in retroperitoneal fat is decreased with DHA and mixed EPA/DHA supplementation but not with EPA supplementation alone [41].</p><p>Portions of hepatic TG are secreted via very low density lipoprotein (VLDL), which delivers TG to peripheral tissues, such as WAT. Hepatic VLDL secretion is enhanced in obese individuals [150] possibly due to increased fatty acid delivery, elevated glucose and insulin concentrations, as well as impaired fat oxidation, which increases fatty acid esterification into TG [151]. Omega-3 PUFAs reduce lipogenesis and reduce hepatic VLDL secretion [51]. In vitro, n-3-PUFA-treated HepG2 cells have decreased hepatic VLDL secretion [152] and reduced apolipoprotein B100 production [153]. This has been validated in both DHA- and n-3-PUFA-supplemented animals [154]. Hence, through inhibition of VLDL formation, n-3 PUFAs could limit the supply of fatty acids to adipocytes and thereby limit adipocyte size and mass. In a deregulated system, n-3 PUFAs would also limit the amount of fatty acids delivered to muscle and liver. Additionally, in animal models, n-3 PUFAs modulate cholesteryl ester transfer protein mediated exchanges, resulting in increased blood HDL cholesterol and possibly apolipoprotein A-1 concentrations [155,156].</p><p>Omega-3 PUFAs alter expression and nuclear localization of both the transcription factor sterol-regulatory element-binding protein-1 (SREBP-1) and the carbohydrate response element binding protein (ChREBP), which control several lipogeneic genes, including those regulating cholesterol and fatty acid synthesis [154,157]. Nuclear translocation of ChREBP is inhibited by n-3 PUFAs and thus results in reduced expression of lipogenic and glycolytic genes, including FAS and pyruvate kinase respectively [158]. Furthermore, n-3 PUFAs suppress hepatic lipogenesis by reducing both messenger RNA (mRNA) and active protein expression of SREBP-1c, which results in reduced expression of many genes involved in lipogenesis, including FAS and acetyl-coA carboxylase [159–161]. Reduced SREBP-1c expression, via n-3 PUFA, has been attributed to inhibited transcription of nascent precursor SREBP-1c, accelerated transcript decay and reduced levels of the mature cleaved form of SREBP-1c [159,160], possibly through inhibition of proteolytic processing and reduced feed-forward activation of the Srebf1 gene. This inhibition could be due to interference with insulin signaling pathways, which promotes the proteolytic processing of SREBP-1c, potentially via an AKT-dependent mechanism [162,163].</p><p>The role of liver X receptor (LXR) is controversial. In vivo, EPA suppression of SREBP-1c promoter activity is dependent upon an intact SRE but not LXR response elements, suggesting that decreased transcription of nascent SREBP-1c with n-3 PUFA treatment results from decreased availability and thus reduced feed-forward activation [163]. In contrast, others indicate a role in the inhibition of LXRα in reduced SREBP-1c expression with n-3 PUFA, but this may be dependent upon cell types [164]. Accelerated degradation of SREBP-1c mRNA has also been proposed as a mechanism for reduced SREBP-1c expression [165]. Omega-3 PUFAs inhibit SREBP-1c cleavage processing, but the cleavage sites are unknown [166].</p><p>Activation of AMP-activated protein kinase by n-3 PUFA can also suppress SREBP-1c cleavage and nuclear translocation, perhaps via serine phosphorylation and/or by blocking activation of the insulin-responsive mechanistic target of rapamycin complex 1 (mTORC1)/S6K-signaling pathway [167]. SREBP-1c synthesis, transport and maturation are increased with insulin [162].</p><p>PUFAs, prostaglandins and leukotrienes can all act as ligands for PPARs. PPARs are transcription factors that form heterodimers with retinoid X receptors in the promoter regions of several genes involved in lipid and glucose metabolism [168,169]. For example, n-3 PUFA activation of PPARα decreases lipogenesis by suppressing FAS activity [161,170]. However, lipogenesis suppression by n-3 PUFA does not require PPARα activation [171].</p><p>PPARγ acts as a master regulator of adipogenesis and controls several genes and adipokines in lipid and glucose metabolism. Omega-3 PUFAs act as ligands for PPARγ and modulate several PPARγ target genes in mice [172] and 3T3-L1 adipocytes [97]. Omega-3 PUFAs enhance PPARγ binding to PPAR-response element in the promoter region of vascular endothelial growth factor-A, which promotes adipogenesis and alleviates hypoxia-induced adipocyte inflammation and insulin resistance [173]. It has been suggested that PPARγ plays a significant role in the ability of n-3 PUFA, specifically DHA, to stimulate M2 macrophage polarization and thereby reduce inflammation since these results are not seen in PPARγ knockdown RAW264.7 cells [174].</p><p>Omega-3 PUFAs have been shown to increase mitochondrial biogenesis and fatty acid oxidation in the liver [175,176], adipose tissue [43] and small intestine [177] of rodents, possibly through PPARα and Cox3 induction [175,178,179]. PUFA-controlled genes involved in lipid oxidation and thermogenesis include mitochondrial HMG-CoA synthase [180], peroxisomal acyl-CoA oxidase [64,181], hepatic CPT-1 [154], FABP [127] and fatty acid transporter proteins [182].</p><p>Activation of PPARα can also increase fatty acid oxidation. Increases in fatty acid oxidation by n-3 PUFA may also be mediated by AMPK, a known regulator of cellular energy metabolism. AMPK up-regulation by n-3 PUFA has been demonstrated in both adipose tissue and cultured adipocytes [55].</p><p>Taken together, n-3 PUFAs regulate lipid metabolism, favoring fatty acid oxidation and suppression of lipogenesis and leading to a favorable lipid profile and adipocyte metabolism.</p><!><p>Many have examined cold- and diet-induced thermogenesis mediated by mitochondrial uncoupling proteins (UCPs) in the presence of n-3 PUFA [43,48,54]. UCPs are inner mitochondrial proteins that function to transport hydrogen ions across the mitochondrial inner membrane. We have recently shown that BAT from EPA-supplemented mice expresses higher levels of thermogenic genes, such as PRDM16, peroxisome proliferator-activated receptor-gamma coactivator-1alpha and UCP1 [183].</p><p>Omega-3 PUFAs increase mitochondrial oxidative capacity in WAT [43] and skeletal muscle, possibly through UCP-3 up-regulation [48], but not in BAT or liver [43]. However, because most studies were carried out at 20°C, it is unclear whether increases in mitochondrial oxidative capacity are n-3 PUFA mediated or cold induced. Janovska et al. reported no differences in body weight but decreased epididymal fat mass after feeding an HF diet supplemented with n-3 PUFA in mice kept at 30°C, indicating that n-3 PUFA could attenuate body fat accumulation even at thermoneutrality and independent of cold-induced thermogenesis [52]. Mechanisms underlying the role of n-3 PUFA in possible induction of energy expenditure and prevention of body fat accumulation should be investigated further at various temperatures since thermogenic markers are activated even at 22°C [183].</p><!><p>The mechanism by which n-3 PUFAs have the potential to increase lean mass is not fully understood but likely involves both catabolic and anabolic pathways. Increased lean mass would result in improved body composition and possibly improved metabolism. Even though increases in RMR have been demonstrated with increases in lean mass [184], post-n-3 PUFA supplementation increases in lean mass are not always accompanied by increases in RMR [66]. Future studies are needed to examine the relationship between n-3 PUFA changes in lean mass in relation to RMR since metabolic rate significantly influences body weight.</p><p>Omega-3 PUFAs have been shown to attenuate muscle protein breakdown in isolated muscles of mice [185]. Increases in protein synthesis may be mediated by n-3 PUFA activation of the mTOR-p70s6k signaling pathway [186], a key pathway in muscle cell growth. Similarly, Clark et al. reported increased whole-body protein turnover under insulin stimulation but did not see significant increases in lean mass following 9 months of n-3 PUFA supplementation [187]. Certainly, changes in protein dynamics may not translate to increases in protein mass.</p><p>One possible mechanism of n-3 PUFA in increasing lean mass is the alteration of protein dynamics related to n-3 PUFA anti-inflammatory properties [66] since proinflammatory cytokines like TNF-α can increase protein degradation via ATP-ubiquitin-dependent pathways [188].</p><p>Another such potential mechanism relates to the ability of n-3 PUFA to lower cortisol levels [189]. Noreen et al. reported decreases in cortisol with n-3 PUFA supplementation but noted a significant correlation between cortisol level and changes in body composition [66]. Others have shown that a reduction in fat mass does not lower cortisol production [190]. Hence, n-3 PUFA supplementation may modulate cortisol levels so as to improve body composition [66]. Furthermore, proinflammatory cytokines such as 1L-6 have been shown to increase blood levels of cortisol [191], which increases protein catabolism [192]. Hence, anti-inflammatory properties of n-3 PUFA could aid in disrupting this pathway. However, increased muscle protein synthesis with n-3 PUFA supplementation is not likely mediated by changes in inflammation in a healthy population [186]. Further investigations are needed to elucidate the mechanisms by which n-3 PUFAs alter protein dynamics to increase lean mass.</p><!><p>Epigenetics may be an important contributor to many chronic diseases, including obesity [193,194]. Limited studies have examined n-3 PUFA and epigenetic modifications even though the expression of several genes involved in metabolic homeostasis is regulated by DNA methylation. The few that have been conducted report conflicting evidence for DNA methylation and n-3 PUFA. In a population-based study, n-3 PUFA intake was associated with DNA methylation in Alaska Yup'ik people [195]. A few studies have reported that fish oil supplementation did not alter the methylation pattern of genes [196,197], including leptin and the leptin receptor, in mouse epididymal fat [196]. It may be that n-3 PUFAs work through epigenetic mechanisms other than methylation [197]. In diet-induced obese mice, leptin expression may be regulated by n-3 PUFA via changes in methyl-CpG-binding domain protein 2 and histone modifications [198]. Since an HF diet has been shown to cause changes in the methylation of gene-specific promoter regions in the liver [199] and WAT [200], the influence of n-3 PUFAs on epigenetic modifications warrants further investigation.</p><p>MicroRNAs (miRNAs) are short noncoding RNAs that act as posttranscriptional regulators of genes by acting as sequence-specific inhibitors of mRNA. These miRNAs target transcription factors to indirectly affect entire signaling pathways. It has been documented that miRNAs act as key regulators in the pathogenesis of metabolic disease by affecting inflammation [201] and lipid metabolism [202].</p><p>Recent studies have shown n-3 PUFA to modulate miRNA expression [201,203,204]. A diet enriched in PUFA correlated to changes in circulating miRNAs, specifically miR-106a, along with changes in other miRNAs related to lipid metabolism and adipokine secretion in healthy women [203]. In animal models, n-3 PUFAs suppress inflammation through down-regulation of miR-19b-3p, −146b-5p and −183–5p by targeting toll-like receptor, NOD-like receptor, RIG-l-like receptor, mitogen-activated protein kinase and transforming growth factor-β pathways [201]. In obese rats, DHA has been shown to counteract obesity-related increases in hepatic miR-33a and miR-122, thus improving lipid metabolism via decreased SREBP2 and FAS expression, respectively [204]. Therefore, fully characterizing n-3 PUFA modulation of miRNA involved in key pathways, such as lipid metabolism and inflammation, is warranted and could play a key role in targeting MetS and obesity-related therapies.</p><!><p>Both animal and human studies have examined the beneficial effects of combining n-3 PUFA with other dietary supplements and pharmaceuticals including antidiabetic drugs, L-alanyl-L-glutamine [205], as well as α-lipoic acid [118] and krill oil [206]. Animal studies using the combination of n-3 PUFA and rosiglitazone have reported significantly greater reductions in body weight [63,143], enhanced oxidation of fatty acids [207] and counteraction of lipogenesis than with rosiglitazone therapy alone [63]. Omega-3 PUFA supplementation in addition to antidiabetic pharmaceuticals could attenuate body weight gain caused by pharmaceuticals and should be further investigated [208].</p><!><p>The management of obesity has shifted from a narrow focus on BMI to the wider field that includes the complications of obesity, with the goal to reduce obesity-associated comorbidities [209]. While n-3 PUFAs have not yet shown consistent benefits in terms of weight loss in humans, improvements in the metabolic profile of obese individuals have been demonstrated. Therefore, n-3 PUFAs may be an important adjunct to obesity management along with lifestyle modification and pharmacotherapy. Further study of the genetic and epigenetic molecular targets related to metabolism, appetite and energetics could aid the discovery of novel therapeutic targets for obesity-associated metabolic disorders.</p>

PubMed Author Manuscript

Nickel/Photoredox-Catalyzed Methylation of (Hetero)aryl Chlorides Using Trimethyl Orthoformate as a Methyl Radical Source

Methylation of organohalides represents a valuable transformation, but typically requires harsh reaction conditions or reagents. We report a radical approach for the methylation of (hetero)aryl chlorides using nickel/photoredox catalysis wherein trimethyl orthoformate, a common laboratory solvent, serves as a methyl source. This method permits methylation of (hetero)aryl chlorides and acyl chlorides at an early and late-stage with broad functional group compatibility. Mechanistic investigations indicate that trimethyl orthoformate serves as a source of methyl radical via \xce\xb2-scission from a tertiary radical generated upon chlorine-mediated hydrogen atom transfer.

nickel/photoredox-catalyzed_methylation_of_(hetero)aryl_chlorides_using_trimethyl_orthoformate_as_a_

INTRODUCTION<!>Reaction Optimization.<!>Substrate Scope.<!>Mechanistic Investigations.<!>Expansion of the Methodology.<!>CONCLUSION

<p>The methyl group is one of the most commonly occurring structural motifs in medicinal compounds, appearing in 80% of top-selling small-molecule pharmaceuticals in 2018.1 In drug development, installation of a methyl substituent - on an aromatic ring, for example - is a common strategy for rendering compounds with improved binding affinity, bioavailability, and metabolic stability (Figure 1A).2 The impact of methylation on the biological and physical properties of a molecule has been so pervasive that it has been named the "magic methyl effect".3 As such, synthetic reactions that enable the installation of methyl groups site-selectively and under mild conditions are of broad value. Transition metal-catalyzed cross-coupling is one of the most robust and modular methods for site-selective carbon–carbon bond formation.4 However, most traditional cross-coupling methods for methylation rely upon acutely toxic alkylating reagents or highly reactive organometallic reagents such that compatibility with common functional groups found in bioactive small molecules is problematic.5 Accordingly, there remains a great demand for cross-coupling methods that enable methylation at a late-stage for medicinal and/or process chemistry applications.3</p><p>Recently, researchers have turned to radical-based methods to overcome challenging C(sp3)–C cross coupling. For example, Minisci-type methylation reactions of heteroarenes have been reported with a variety of methyl radical sources, including acetic acid,6a tert-butylperacetate,6b dicumyl peroxide,6c methane,6d and methanol.6e-h Nevertheless, these methods are only effective for electron-deficient heteroarenes and their site selectivity can limit broad applicability. To address these challenges, a few radical-based methylation reactions of aryl halides have been described (Figure 1B). The Weix group demonstrated that N-hydroxy-phthalimide esters can serve as competent sources of methyl, primary, and secondary alkyl radicals for the alkylation of aryl iodides with nickel catalysis.7 Additionally, using nickel/photoredox catalysis in combination with a supersilane reagent, the MacMillan group achieved cross-coupling between aryl and alkyl bromides to furnish new C(sp3)–C(sp2) bonds, including one example of methylation using methyl tosylate.8 Methyl tosylate has also recently found application in nickel-catalyzed methylation of aryl bromides and tosylates, alkyl halides, and acid chlorides, as reported by Gong and coworkers.9 While highly enabling, these methods still require preparation of the methyl radical source or employment of electrophilic methylating reagents. Furthermore, a radical-based method that permits methylation of aryl chlorides, the most abundant and inexpensive aryl halide coupling partner, has yet to be reported. In addition to the benefits that use of aryl chlorides would permit for early-stage methylation, late-stage conversion of aryl chlorides to toluenes has demonstrable value in medicinal chemistry as well: for example, methylation of the aryl chloride in a Celecoxib precursor shortened its half-life such that it could be administered as the first selective COX-2 non-steroidal anti-inflammatory drag.10</p><p>Our group has recently reported an approach to the cross-coupling of chloride-containing electrophiles with C(sp3)–H bonds via nickel and photoredox catalysis.11 The chloride-containing electrophile serves as both the coupling partner and the source of chlorine radical for C(sp3)–H bond activation via hydrogen atom transfer (HAT). We questioned whether this reaction platform could be adapted to enable methylation of (hetero)aryl chlorides. While methane is the most analogous methyl radical source, initial reactions employing methane under our previously optimized conditions proved unfruitful. Instead, we sought an alternative methyl radical source that could be accessed via HAT and that would have similar attributes to methane, including its abundance, cost, and functional group compatibility.</p><p>In this context, we considered that trimethyl orthoformate, a common laboratory solvent, could serve as a methyl radical source (Figure 1C). Owing to its weak tertiary C(sp3)–H bond (88.7 kcal/mol), preferential HAT at the methine over the methyl positions was computed to be favorable (ΔBDFE = −1.2 kcal/mol) (Figure 2A). If addition of the resultant tertiary radical were slow to rebound into a nickel catalyst, we posited that unimolecular β-scission could occur to generate methyl radical and dimethyl carbonate (ΔG = −25.1 kcal/mol and ΔG‡ = 11.4 kcal/mol).12 Coupling the generation of high energy alkyl radicals with the formation of a stable carbonyl byproduct via β-scission has been well-studied,13 but only recently emerged as a strategy in transition metal-catalyzed cross coupling. Reported examples of β-scission from carbon-centered radicals in cross coupling are limited to xanthate esters14 that must be pre-synthesized and afford only stabilized radical species, thus precluding access to methyl radical.15 In contrast, trimethyl orthoformate would permit access to methyl radical from a commercial and abundant reagent.</p><!><p>To evaluate the feasibility of using trimethyl orthoformate as a source of methyl radical, we investigated the coupling of 4′-chloroacetophenone with trimethyl orthoformate (Figure 2B).11a Using Ni(cod)2 (10 mol%), 4,4'-di-tert-butylbipyridine (dtbbpy) (15 mol%), Ir[dF(CF3)ppy]2(dtbbpy)PF6 (1 mol%), and K3PO4 (2 equiv) in a 1:1 mixture of trimethyl orthoformate and benzene afforded the desired toluene 3 in 38% yield. Two other products were formed alongside the methylated product: ester product 2 (8% yield), arising from coupling at the tertiary position of trimethyl orthoformate (B in Figure 2), and benzylic ether product 1 (27% yield), arising from coupling at the primary C–H bonds (A in Figure 2).</p><p>To improve the yield and selectivity of the reaction, we undertook an optimization of the reaction conditions. Use of NiCl2•glyme as the Ni source resulted in a slightly improved yield of 43% (Table 1, Entry 1). Both yield and selectivity between 3 and 1 could be further improved by employing tert-butylbenzene as the reaction cosolvent, likely a result of greater stabilization of chlorine radical by the more electron-rich solvent (Table 1, Entry 2). Previous studies have shown that chlorine radical, an electrophilic radical species, can be stabilized by arenes to deliver more selective C(sp3)–H abstraction according to the Hammond postulate.16 Reducing the loading of trimethyl orthoformate to ten equivalents in benzene led to 26% yield of the desired toluene in 24 hours (Table 1, Entry 3). Conversely, performing reactions in trimethyl orthoformate without cosolvent led to the largest increase in yield and selectivity, providing the desired toluene in 61% yield (Table 1, Entry 4). Selectivity between 3 and 1 reached a ratio of 3.9:1, while ester formation was minimized (< 5% yield) relative to reactions run with aromatic cosolvents. Upon omission of the base, 3 was obtained in 9% yield, suggesting that HCl formation without sequestration may be deleterious to reactivity (Table 1, Entry 5). All other components of the reaction were required for productive methylation, as individual omission of NiCl2•glyme, dtbbpy, photocatalyst, and light source resulted in no product formation and complete recovery of the aryl chloride (Table 1, Entries 6-9).</p><!><p>With optimized conditions, we examined the reaction scope (Figure 3A). Generally, electron-deficient aryl chlorides underwent methylation in higher yields than electron-rich aryl substrates, consistent with their relative reactivity to Ni(0) oxidative addition. Unlike methods that employ reactive nucleophilic or electrophilic methylating reagents, a variety of sensitive functionality was well tolerated, including ketones 3 and 4, nitriles 5 and 6, aldehyde 7, and ester 8. Ortho-substituted aryl chlorides (5, 11) also delivered methylated product in moderate to high yield. Methylation of substrates containing heteroaryl functionality distal to the site of cross-coupling, including pyridines 13, 16, and 17, furan 14, and pyrrole 15, could also be achieved. Biologically relevant aryl chlorides loratadine 19, fenofibrate 20, and zomepirac 21 provided the corresponding methylated product in high yields, indicating that this method is amenable to late-stage functionalization of bioactive compounds.10 In the methylation of perphenazine (22), exclusive methylation of the aryl chloride was observed, in contrast to methods employing electrophilic methylating reagents that would be expected to methylate the primary alcohol. Finally, procymidone, which contains two chemically equivalent aryl chlorides, underwent selective mono-methylation to produce 23 in 55% yield, likely a result of the sensitivity of the catalytic system to electronic effects.</p><p>Employment of electron-rich aryl chlorides in the nickel/photoredox cross-coupling reaction delivered low yields of methylated products, likely due to sluggish oxidative addition (Table 2, Entries 1 and 4). To overcome this challenge, we turned to aryl bromides as substrates, but productive methylation was not observed from these substrates either, presumably because the weak H–Br bond (BDE = 88 kcal/mol) renders HAT from bromine radical less favorable. However, reactivity could be restored by using aryl bromides in conjunction with an exogenous chloride additive for halide exchange (Table 2, Entries 3 and 6), delivering 24 and 25 in 56% and 52% yield, respectively.</p><p>Next, we sought to explore the scope of heteroaryl chloride coupling partners (Figure 3B). A variety of nitrogen, oxygen, and sulfur-containing heteroaryl chlorides underwent methylation in moderate to high yields, including pyridines 26–29, quinolines 30–34 and 42, quinoxaline 35, quinazoline 41, pyrimidines 36–37, thiophenes 38 and 39, and thiazole 40. Importantly, this method for radical methylation enables functionalization at sites that are not accessible via Minisci-type reactivity; for example, the 3-, 6-, and 7-positions of quinolines (32, 33, and 34, respectively) and positions meta to nitrogen atoms in pyridines (27–29) underwent site-selective methylation. Biologically-relevant heteroaryl chlorides, such as etoricoxib (43), also underwent methylation in good yield.</p><p>The primary byproduct in this methodology is derived from alkoxymethylation of the aryl chloride. Since this byproduct is a benzylic ether, a solution that we pursued was subjecting the reaction to Pd/C hydrogenolysis to convert the byproduct into methylated product. Select examples of the improvement in yields afforded by this workup protocol, including for high-value targets 19 and 20, are shown in Figure 3.</p><p>Furthermore, our group has previously demonstrated that acid chlorides can be used as coupling partners in Ni/photo-redox-catalyzed C(sp3)–H functionalization of alkanes.11c We recognized that application of the methylation conditions to acid chloride coupling partners could potentially deliver a mild synthesis of aliphatic and aromatic methyl ketones9b as compared to traditional protocols which rely on harsh organometallic reagents or strong Lewis acids.17 Gratifyingly, we found that application of the optimized methylation conditions to this substrate class afforded access to methyl ketones from primary (44), secondary (45), tertiary (46), and aryl (47) acid chlorides (Figure 3C). The method was also applicable to acid chlorides prepared from biologically relevant ketoprofen (45) and probenecid (47). Additionally, ester 48 and tertiary amide 49 were prepared from the corresponding chloroformate and carbamoyl chloride.18</p><!><p>Having explored the scope of this transformation, we next sought to evaluate the mechanism of methyl incorporation from trimethyl orthoformate. According to our prior mechanistic work,11a we propose that photoelimination from Ni(III) intermediate 52 affords a chlorine radical which mediates HAT with trimethyl orthoformate (Figure 4A).19 Observation of methylated product 3, ester product 2, and benzylic ether product 1 lends initial support to a HAT-initiated process in the Ni-catalyzed cross-coupling. As further evidence for the intermediacy of organic radical intermediates, when the methylation of 4′-chloroacetophenone was conducted under standard conditions with one equivalent of TEMPO, none of these three products were observed.20</p><p>To further evaluate the mechanism, we monitored the reaction course via ReactIR. The reaction progress showed that as 4′-chloroacetophenone is consumed, dimethyl carbonate and 4′-methylacetophenone (3) are generated in a 1:1 ratio (Figure 4B, right). The formation of dimethyl carbonate and 3 could also be traced in a 1:1 ratio via quantitative 13C NMR experiments and both experiments suggest overall non-zeroth order kinetics (see SI, Figures S19-S20). While this observation is consistent with a β-scission mechanism which requires stoichiometric formation of dimethyl carbonate at a comparable rate to product formation, we also considered that methylated product 3 and dimethyl carbonate could arise from Ni-catalyzed hydrodealkoxylation of byproduct 1. However, a control reaction subjecting 1 to the photocatalytic Ni conditions did not generate methylated product 3 (Figure 4C).</p><p>Another pathway that is consistent with these results is the one shown in Figure 4D, wherein Ni(II)(Ar) species 53 mediates oxidation of the tertiary radical of trimethyl orthoformate (B), delivering Ni(I)–X that undergoes addition to the oxocarbenium intermediate (56).21 Such a process would produce Ni(III) intermediate 54 and dimethyl carbonate, which are both invoked in the proposed catalytic cycle (Figure 4A). A key difference between these proposals is that, according to the β-scission mechanism, methylation with trimethyl orthoformate should be possible in the absence of nickel. Indeed, when diethyl vinylphosphonate was reacted with benzoyl peroxide at 80 °C in trimethyl orthoformate, the methylated Giese product 57 was observed in 22% yield (Figure 4E). Taken together, these experiments provide support for the proposed β-scission mechanism for methyl radical generation from trimethyl orthoformate.</p><!><p>More generally, the results gathered in interrogating the β-scission mechanism suggest that trialkyl orthoformates can serve as broadly useful and practical aliphatic radical sources for the design of new synthetic methods. For example, in this Ni/photoredox cross-coupling, simple modification of the solvent to triethyl or triisopropyl orthoformate allowed access to the respective alkylated product 58 or 59 (Figure 5A).22</p><p>This finding prompted us to consider using acetals as sources of aliphatic radicals in this cross-coupling reaction. Dialkyl acetals have been shown to undergo β-scission with liberation of aliphatic radicals;13d-e however, the reactions are initiated under harsh conditions (homolysis from peroxides or high temperatures) and, to the best of our knowledge, acetals have not been used as a source of radicals in metallaphotoredox cross-coupling. For the transfer of more complex aliphatic radicals, we envisioned that these reagents could be attractive alternatives to orthoformates given the facile synthesis of acetals from aldehydes and a diverse array of alcohols and because acetals feature only two equivalents of the alkyl cross-coupling partner per molecule rather than three in the related orthoformate derivative. In a preliminary study, we were delighted to find that replacing the orthoformate cosolvent with a 1:1 mixture of benzaldehyde dimethyl acetal:benzene provided methylated product 4 in 50% yield (Figure 5B). Furthermore, reaction of 4-chlorobenzophenone with 12 equivalents of benzaldehyde dineopentyl acetal afforded the resulting alkylated product 60 in 41% yield. These preliminary results offer promise for utilizing β-scission of acetals for installing aliphatic groups selectively and under mild reaction conditions.</p><!><p>In conclusion, we have developed a Ni/photoredox approach to the site-selective methylation of chloride-containing electrophiles using trimethyl orthoformate as an abundant, nontoxic, and functional group-compatible methylating reagent. Methylation of feedstock, as well as chemically complex, (hetero)aryl and acyl chlorides is possible such that we anticipate that this method could find application in the pharmaceutical industry. Mechanistic investigations indicate that trimethyl orthoformate serves as a source for methyl radical via β-scission from a tertiary radical generated upon chlorine-mediated hydrogen atom transfer. As such, this approach offers an opportunity to circumvent traditional protocols for accessing low molecular weight aliphatic radicals from toxic or high-energy reagents.</p>

PubMed Author Manuscript

A Supramolecular Approach for Enhanced Antibacterial Activity and Extended Shelf-life of Fluoroquinolone Drugs with Cucurbit[7]uril

The host-guest interactions of a third-generation fluoroquinone, danofloxacin (DOFL), with the macrocyclic host cucurbit[7]uril (CB7) have been investigated at different pH values (~3.5, 7.5, and 10). The photophysical properties have been positively affected, that is, the fluorescence yield and lifetime increased, as well as the photostability of DOFL improved in the presence of CB7. The antibacterial activity of DOFL is enhanced in the presence of CB7, as tested against four pathogenic bacteria; highest activity has been found towards B. cereus and E. coli, and lower activity towards S. aureus and S. typhi. The antibacterial activity of two additional second-generation fluoroquinones, i.e., norfloxacin and ofloxacin, has also been investigated in the absence as well as the presence of CB7 and compared with that of DOFL. In case of all drugs, the minimum inhibitory concentration (MIC) was reduced 3-5 fold in the presence of CB7. The extended shelf-life (antibacterial activity over time) of the fluoroquinone drugs in the presence of CB7, irrespective of four types of bacteria, can be attributed to the enhanced photostability of their CB7 complexes, which can act as better antibiotics with a longer expiry date than uncomplexed DOFL.Fluoroquinolones (FQs), derived from nalidixic acid, constitute one of the most successful classes of antibiotic drugs in therapeutic applications that are used in the treatment of a variety of bacterial infections 1,2 . The 'ideal' fluoroquinolone combines good clinical efficacy with low minimal inhibitory concentration (MIC) without any cytotoxicity. To achieve low MIC values, different generations of fluoroquinolones have been developed over the years as their substitution pattern modulates their antimicrobial activity 1,2 , such as the insertion of a fluorine atom at position 6 and a piperazine ring 3 . FQs bear both an acidic group (carboxylic acid) and a basic tertiary amino one, inferring amphoteric properties. Depending on pH, FQs prevail in their protonated (below pH 6), neutral or zwitter ionic (pH 6-8), or anionic forms (above pH 8) 4,5 . It has been established that the zwitter ionic form of FQ is responsible for antimicrobial activity as this form is sufficiently lipid-soluble to be able to penetrate tissues 6 . However, over the past 20 years, fluoroquinolone research has aimed at improving activity against Gram-positive microorganisms, whilst retaining its bioactivity against Gram-negative organisms. Thus, first-generation quinolones are active against Gram-negative microorganisms, second-generation ones such as norfloxacin and ofloxacin are active against Gram-negative and some Gram-positive microorganisms, while third-and fourth-generation quinolones have an expanded activity against Gram-positive microorganisms 3 . Besides the covalent/synthetic modification of the basic chemical structure, noncovalently linked, externally controlled supramolecular

a_supramolecular_approach_for_enhanced_antibacterial_activity_and_extended_shelf-life_of_fluoroquino

<!>Results<!>+<!>Absorption and Fluorescence Behavior of Danofloxacin with CB7.<!>Determination of pK a<!>Antibacterial Activity and Photostability of CB7•DOFL Systems.<!>Discussion<!>Methods<!>ITC measurements. Isothermal titration calorimetry experiments were carried out on a VP-ITC from<!>Antibacterial activity measurements.<!>DFT-calculations.

<p>host-guest interactions using macrocyclic hosts such as cyclodextrins, cucurbiturils, etc. 7,8 are promising to improve and control the antibacterial activity of drugs. In a recent study, Henriques et al. reported that gallic acid shows good antibacterial activity upon complexation with β-cyclodextrin and its derivatives 9 . Zhang et al. 10,11 reported an enhanced antibacterial activity of a porphyrin photosensitizer, and Bai et al. described a supramolecular antibiotic switch to regulate the antibacterial activity using cucurbit [7]uril (CB7) as macrocyclic host 12 . Zhang et al. have also designed bacterial responsive supramolecular complex of perylene diimide derivative for photothermal therapy with high selectivity towards facultative anaerobic bacteria 13 . Recently, cucurbit[n]urils (CBn) have received attention due to the strong interactions of their negatively charged carbonyl portals with positively charged organic dye molecules, metal cations, nanoparticles, proteins, and surfactants or through peripheral binding with polyanions such as polyoxometalates [14][15][16][17][18][19][20][21][22][23][24][25] . Among the CBn homologues, CB7 is the most prominent one; it is highly water soluble, forms the strongest inclusion complexes with organic guests, and improves drastically their molecular properties such as photostability, aqueous solubility, fluorescence behavior, etc. 7,8,18,[26][27][28][29][30] . In recent years, cucurbituril-based host-guest complexes are found to have potential applications in nanosensors to discriminate cancer cells 31 , supramolecular assays for drug detection 32 , storage/delivery of polysulfide in lithium-sulfur batteries 33 , white light emitting materials 34 , molecular switching 35 , supramolecular catalysis to modulate the activity of reaction intermediates 22,36 and stimuli-responsive supramolecular assemblies 25 / vesicles for targeted drug delivery 37 etc.</p><p>Danofloxacin (DOFL) is a third-generation fluoroquinolone antimicrobial drug with a rapid bactericidal activity against a broad range of pathogens responsible for a number of disease syndromes of economic importance in the commercial rearing of livestock 38 . Though DOFL is used in veterinary medicine as the mesylate salt for the treatment of respiratory diseases in cattle, swine, and chicken, there is a major concern to human health for causing toxicity, allergy, and bacterial resistance problems 6 . Hence, there is a need either to remove DOFL from cow milk or to lower drug intake through more effective antibacterial activity in the treatment of animal disease. In a recent study, Valcárcel et al. have shown that β-cyclodextrin-modified nanocellulose can be used for the selective fluorimetric determination and extraction of danofloxacin from milk samples 39 . In this article, we establish a supramolecular approach to enhance antibacterial activity and shelf-life of CB7-encapsulated DOFL against two Gram positive (Staphylococcus aureus: S. aures; Bacillus cereus: B. cerus) and two Gram negative (Escherichia coli: E. coli; Salmonella typhi: S. typhi) pathogenic bacteria with 3-5 fold reduced MIC. The extended shelf-life is attributed to an increased photochemical and thermal stability of the drug in the presence of CB7.</p><!><p>Depending on the pH of the solution, danofloxacin exists in three different forms (cationic (I), DOFLH 2</p><!><p>; zwitter ion (II), DOFLH, and anionic (III), DOFL − ; Fig. 1a); the zwitter ionic form can also tautomerize into a putative neutral form in which the carboxylic group is stabilized by forming a six-membered cyclic structure through intramolecular hydrogen bonding 40 . The changes in the absorbance by varying pH of the drug at a particular wavelength showed two inflection points, corresponding to two different pK a values (inset of Fig. S1, Supporting Information). The ground-state pK a values evaluated from the fitted data (cf. solid curve in the inset of Fig. S1) were 6.3 ± 0.1 and 8.6 ± 0.1, which corresponded to the prototropic equilibria of the deprotonation of the carboxylic OH (pK a1 ) 41 and the tertiary 1,4-diazabicyclo[2.2.1]heptyl methylamino group (pK a2 ), respectively 42 .</p><!><p>Considering the multiple pK a values of DOFL, we have investigated the binding interaction of the different prototropic forms of DOFL with CB7 at three pH values, 3.5, 7.5, and ~10. At pH 7.5, DOFL exists as DOFLH, which showed an absorption within the range of 250-425 nm with an intense narrow band at ~275 nm and a weaker broad band at 340 nm 40 . Upon addition of CB7, there was a bathochromic shift in the 275 nm band and a slight increase in the absorbance of the broad band along with three isosbestic points (Fig. 2a). This major change in the absorption spectra of DOFL clearly indicated a strong interaction with CB7. However, there was no significant change in the absorption spectra of DOFLH 2 + and DOFL − , prevalent at lower and higher pH values, upon the addition of CB7. On the emission front, DOFLH at pH 7.5 showed a broad emission with spectral maximum at 425 nm. Upon gradual addition of CB7, the fluorescence intensity increased along with a bathochromic shift of ~17 nm and an isoemissive point at ~410 nm, indicating the formation of a host-guest complex (Fig. 2b). At pH ~3.5, DOFLH 2 + showed a broad fluorescence band with peak position at ~442 nm, whereas at pH ~10.2, DOFL − showed an emission maximum at ~433 nm. In contrast to the titrations of DOFLH, the fluorescence intensity of DOFLH 2 + and DOFL − decreased upon addition of CB7 to the solution. DOFLH 2 + showed a bathochromic shift (~3 nm) at the peak position and DOFL − displayed a hypsochromic shift of ~8 nm with an isoemissive point at 408 nm (Fig. S2). The characteristic changes in the fluorescence quantum yield (Φ f ) and lifetime (τ f ) of all three forms suggested host-guest complex formation with CB7. Coumarin 1 in water was used as the standard to measure the fluorescence quantum yields of all forms of the drug with and without CB7 in water 43 . The combined photophysical parameters of all three forms of danofloxacin in water, with and without CB7, are listed in Table 1.</p><p>Binding Constants of CB7•DOFL Systems from Fluorescence and ITC Measurements. The binding constants for the three different forms with CB7 were determined based on the fluorescence titrations by using a 1:1 binding model [44][45][46] . The binding constants obtained from the fluorescence titration curves (Insets of Fig. 2 and Insets of Fig. S2) reached a maximum for the CB7•DOFLH complex (1.6 × 10 5 M −1 ) and were lower for both, CB7•DOFLH 2 + (2.1 × 10 4 M −1 ) and CB7•DOFL − (6.5 × 10 3 M −1 ), see Table 1. Due to the small changes in the photophysical properties of DOFLH 2 + and DOFL − upon addition of CB7, the binding constants for all three complexes were more reliably determined by using isothermal titration calorimetry (ITC), which confirmed the ability of CB7 to complex DOFL at different pH values (Fig. 3a,b). The binding constants of DOFL to CB7 at pH 7.5 and 10.2 agree well with the ones obtained by using fluorescence titrations. However, the ITC results at in the presence of CB7 due to deshielding of the nuclei, which suggested that these protons reside near to the carbonyl portals of CB7. In contrast, the aliphatic protons (4 and 7-12) displayed upfield shifts (Δδ ranges from 0.2 to 0.8 ppm), which is attributed to the encapsulation of the diazabicyclo[2.2.1]heptyl group into the hydrophobic cavity of CB7. An upfield shift of the F atom by 1.2 ppm in the 19 F-NMR signal also confirms the inclusion of the fluorine atom in the cavity 47 . Energy Optimized Structures of CB7•DOFL Systems. Dispersion-corrected quantum-chemical calculations (DFT, wB97XD/6-31G*) were performed to determine the structure of the host-guest complexes and their stability. The optimized structure for all complexes showed that the diazabicyclo[2.2.1]heptyl group is encapsulated inside the CB7 cavity. For CB7•DOFLH 2 + and CB7•DOFLH, the complexes were stabilized over the CB7•DOFL − complex through ion-dipole interactions (see SI, Fig. S3). The calculated binding energies followed the order: CB7</p><!><p>Values of CB7•DOFL Systems. Others and we have established that there is a substantial variation in the pK a values of the guest molecules upon inclusion complex formation with macrocycles 28,44,[48][49][50][51][52] . Whenever the conjugate acid shows stronger binding interaction with the macrocyclic host than its corresponding base, the pK a value usually exhibits a large upward shift 52 . In the current complex system, due to the existence of similar kind of binding interactions, an upward shift of both the pK a values of DOFL in the presence of CB7 is observed. The consequence of complexation on the prototropic equilibria of DOFL was studied by monitoring the variations in the absorption spectra of the drug at different pH values, with 1 mM of CB7 and the results are shown in Fig. 4. In this system, assuming the complexation of all the three forms of drug with CB7, the acid-base equilibria of the drug should follow a six-state thermodynamic cycle 48 , as shown in Scheme S1, SI. Excess concentration of CB7 was used to confirm virtually the complete binding of the three different prototropic forms of the drug with CB7. As becomes evident in Fig. 4, the spectral changes are significant at ~285 nm in the investigated pH range from 3.5-12. The pH titration curve corresponding to the first deprotonation equilibrium for the carboxylic OH of the fluoroquinolone unit in the CB7 complex presented a value of pK′ a1 = 7.22 ± 0.05, about one unit higher than the pK a1 value of the uncomplexed drug. Such a shift in the pK a values specifies that the CB7-complexed drug converts into a stronger base, by ~12 times, than its uncomplexed form. The pH titration curve at higher pH values corresponds to the dissociation constant of the tertiary diazabicyclo[2.2.1]heptyl methylamino group in its CB7 complex; its pK′ a2 value was found to be 9.75 ± 0.1, considerably higher than the pK a2 value for the uncomplexed drug (pK a2 = 8.6 ± 0.1).</p><!><p>It is reported that the zwitterionic form of fluoroquinolones is mainly responsible for their antibacterial activity 6 . Since CB7 showed a high binding affinity towards this most active form, we studied the effect of CB7 complexation on the antibacterial activity of DOFL. In detail, we measured the reduction of bacterial growth of two Gram positive (Staphylococcus aureus and 2 and S1). Though the antibacterial activity of DOFLH 2 + (at pH 3.5) was lower than that of the zwitterionic form (at pH 7.5), the activity of CB7-complexed cationic form of DOFLH 2 + followed a similar trend as in the complexed zwitterionic form -irrespective of the individual micro-organism (Tables 2 and S1). At pH 8.1, where almost all the complexed DOFLH is in the zwitterionic form, the antibacterial activity also followed a similar trend and was found to be maximum for CB7-DOFLH system as compared to those at pH 3.5 and 7.5 (Table 2 and Fig S4 , SI). It may be noted here that even though the upward pK a shift effectively brings down the zwitterionic concentration by ~30% at pH 7.5, the increased contribution of the CB7-complexed protonated form cumulatively enhances the antibacterial activity. Since our experiments are related to bacterial growth in the physiological condition, we have carried out all the other experiments at pH 7.5. Furthermore, the minimum inhibitory concentration (MIC) of DOFLH with and without CB7 towards these four pathogenic bacteria was determined which matches the reported value 37 , see Table S2. Expectedly, the higher inhibitory effect of the drug in the presence of CB7 translated into a (medicinally desirable) decrease of the minimal inhibitory concentration (MIC), e.g. from ~0.261 μg/ml against B. cereus in the absence of CB7 to 0.052 μg/ml in the presence of 10 μM CB7 (Table S2) and is pictorially represented in Fig. 1b. Antibacterial activity of a drug molecule generally follows two mechanisms, either bacteriostatic or bactericidal. Bacteriostatic drugs only prevent the growth of bacteria while bactericidal ones permanently kill the bacteria 38 . In our system, it has been observed that CB7-complexed DOFL showed a higher antibacterial activity, but the activity remained bacteriostatic in nature.</p><p>Furthermore, we investigated the shelf-life (activity/quality over a specified period of time) of pre-dissolved danofloxacin solutions with and without CB7 at pH 7.5 stored under ambient conditions. Figure 5c presents the inhibition zone/bacterial activity of uncomplexed and CB7-complexed DOFL in the four studied bacteria with time. For CB7-complexed DOFL, the antibacterial activity remained constant for more than two months, irrespective of bacteria type. In contrast, uncomplexed DOFL showed a 30-40% decrease in antibacterial activity over the same period in all types of bacteria used in this study. We attribute the increased shelf life in the presence of CB7 to a reduced photochemical or thermal degradation of the drug. Indeed, when we followed sample integrity by absorbance and fluorescence, we observed a strong degradation for DOFL solutions kept at ambient conditions for a period of 10 days, which could be virtually suppressed in the presence of CB7 (Figs 6a and S5a, SI). Similarly, thermal degradation at elevated temperature was effectively suppressed in the presence of CB7 as shown in the absorbance changes monitored at 60 °C for about 2 hrs (Fig. S6a, SI). Table 2. Antibacterial activity (in terms of inhibition zone) of DOFL (~14 μM) with and without CB7 (1 mM) towards four pathogenic micro-organisms at two different pH values.</p><p>To generalize, we also investigated the host-guest interactions of two second-generation fluoroquinolones (ofloxacin: OFL and norfloxacin: NRFL) with CB7 by monitoring their absorption and fluorescence changes (Figs S7 and S8, SI) and chemical shift in the NMR signals (upfield shift for the aliphatic protons and downfield shift for the aromatic protons as shown in Fig. S9, SI) as well as their antibacterial activity. Unlike DOFL, both NRFL and OFL contain a piperazine instead of the diazabicyclo[2.2.1]heptyl group. As observed in DOFL, CB7-complexed OFL and NRFL showed higher antibacterial activity than the uncomplexed drug at both selected pH conditions, 3.5 and 7.5 (Fig. S10 and Table S1, SI). However, NRFL remained inactive against B. cereus and E. coli, regardless of the presence of CB7. The MIC value of the free and CB7-complexed NRFL and OFL drugs towards these pathogens are listed in Table S2. The extended shelf-life of the two drugs against the above bacteria in the absence and presence of CB7 at pH 7.5 was also determined, see Fig. S10, SI. In line with the observed increase in biological activity in the presence of CB7, the photo/thermal degradation of NRFL and OFL samples, directly followed by absorbance or fluorescence, was again reduced in both cases (Figs 6b,c and S5b,c and S6b,c), although the effect was less pronounced than for DOFL (see above). The combined studies suggest that the stabilizing and efficacy-enhancing properties of CB7 formulations on fluoroquinolone drugs are not specific for a single compound. From a chemical reactivity point of view, the extended shelf-life of the fluoroquinolone drugs in the presence of CB7 is presumably due to a stabilization of the aromatic or bicyclic amino substituents within the CB7 cavity. Norfloxacin and ofloxacin, for example, are known to undergo photochemical or thermal degradation of the piperazine side chain, either by degradation to an amino group or oxidation 40 .</p><!><p>CB7 forms host-guest complexes with the three prototropic forms of a third generation fluoroquinolone derivative, danofloxacin (DOFL). The fluorescence quantum yield, excited-state lifetime, and photostability of DOFL, in their distinct prototropic forms, are modulated significantly in the presence of CB7. Similarly, complexation behaviour is observed for two additional second generation fluoroquinolone drugs, namely, norfloxacin (NRFL) and ofloxacin (OFL). The observed shifts towards higher pK a values for the drugs encapsulated by CB7 is consistent with the better stabilization of the protonated forms of the guests by more attractive (for deprotonation of cationic guests to their neutral forms) or less repulsive (for deprotonation of neutral guests to their anionic forms) interactions with the carbonyl portals of the host 48,52 .</p><p>In regard to potential applications, the antibacterial activity of all three drugs is considerably enhanced in the presence of CB7, as we explored against four pathogenic bacteria at both, pH 3.5 and 7.5. Among the different variants studied, DOFL displays highest activity towards B. cereus and E. coli and lowest activity towards S. aureus and S. typhi at both pH values. Furthermore, the substantial reduction in MIC value (3-5 fold) and extended shelf-life along with increased antibacterial efficacy, generalized for all three drugs, are highly encouraging for the use of CB7 for the design and development of new long-acting antibiotic formulations.</p><!><p>Absorption and fluorescence measurements. Absorption spectra of the samples were recorded from solutions in 1-cm quartz cuvettes on a Varian Cary 4000 UV/Vis spectrophotometer or a Jasco UV-Vis spectrophotometer (model V-650). Steady-state fluorescence spectra were recorded by using a Varian Eclipse or a Hitachi F-4500 fluorometer. The fluorescence lifetime measurements at room temperature were performed by time-correlated single-photon-counting (TCSPC) on an FLS-920 fluorometer (Edinburgh instrument) incorporating a pulsed diode laser (PDL 800-B from Picoquant, λ ex = 373 nm, FWHM ca. 50 ps) suitable for lifetime measurements down to 300 ps. The fluorescence decays could be satisfactorily fitted (χ 2 < 1.05) by using mono/ bi-exponential decay functions.</p><!><p>Microcal, Inc., at 25 °C. The binding equilibria were studied using a cellular guest (danofloxacin) concentration of 0.12 mM, to which a 20-30 times more concentrated host (CB7) solution was titrated. Typically, 27 consecutive injections of 10 μL were used. All solutions were degassed prior to titration. Heats of dilution were determined by titration of host (CB7) solution into water. The first data point was removed from the data set prior to curve fitting with Origin 7.0 software according to a one-set-of-sites model. The knowledge of the complex stability constant (K a ) and molar reaction enthalpy (ΔH°) enabled the calculation of the standard free energy (ΔG°) and entropy changes (ΔS°) according to ΔG° = −RT ln K a = ΔH° − TΔS°. Photo-and thermal stability measurements. The photobleaching of FQ drugs with and without CB7 was monitored by measuring the absorbance/fluorescence at different times with daylight irradiation (ambient conditions) at pH 7.5. Thermal stability of these drugs was also observed by monitoring the absorbance changes of the drugs at 60 °C at different times.</p><!><p>Antibacterial assay was carried out by using the agar well diffusion method. Bacterial strains S. aureus (MU-50), B. cerus (NCIM-2156), E. coli (M-16), and S. typhi (SL1344), were used as indicator for this analysis. The antibacterial activity and the minimum inhibitory concentration (MIC) in terms of inhibition zone of these bacterial strains were determined by the standard agar diffusion method as per NCCLS; in short, cultures were grown overnight in nutrient broth. The cultures were diluted with saline to obtain an inocula of 10 7 CFU/ml using 0.5 McFarland's Standard and immediately plated. Wells were bored on the plate by using a cork borer (8 mm) and different dilutions of the drug samples were added to the wells and subsequently incubated at 37 °C for 24 hrs. The zone of inhibition values were estimated by measuring the diameter of the inhibition zone against test micro-organisms. The lowest dilution showing a significant zone of inhibition was considered as MIC. All the experiments were carried out in quadruplicate.</p><!><p>Quantum-chemical calculations were performed in gas-phase with the Gaussian 09 package, utilizing dispersion-corrected density functional theory (wB97XD) in combination with a 6-31G* basis set.</p>

Scientific Reports - Nature

Learning Protein-Ligand Binding Affinity with Atomic Environment Vectors

Scoring functions for the prediction of protein-ligand binding affinity have seen renewed interest in recent years when novel machine learning and deep learning methods started to consistently outperform classical scoring functions. Here we explore the use of atomic environment vectors (AEVs) and feed-forward neural networks, the building blocks of several neural network potentials, for the prediction of protein-ligand binding affinity.The AEV-based scoring function, which we term AEScore, is shown to perform as well or better than other state-of-the-art scoring functions on binding affinity prediction, with an RMSE of 1.22 pK units and a Pearson's correlation coefficient of 0.83 for the CASF-2016 benchmark. However, AEScore does not perform as well in docking and virtual screening tasks. We therefore show that the model can be combined with the classical scoring function AutoDock Vina in the context of ∆-learning, where corrections to the AutoDock Vina scoring function are learned instead of the protein-ligand binding affinity itself. Combined with AutoDock Vina, ∆-AEScore has an RMSE of 1.32 pK units and a Pearson's correlation coefficient of 0.80 on the CASF-2016 benchmark, while retaining the good docking and screening power of the underlying classical scoring function.

learning_protein-ligand_binding_affinity_with_atomic_environment_vectors

Introduction<!>Atomic Environment Vectors<!>Angular<!>Neural Network<!>Training and Test Datasets<!>∆-learning<!>Consensus Scoring<!>Software<!>AEScore Hyper Parameters Optimization<!>Scoring Power<!>Consensus Scoring<!>Implicit Hydrogen Atoms<!>Similarity between training and test sets<!>Ranking Power<!>Docking Power<!>∆-AEScore ∆-learning with AutoDock Vina<!>S RT<!>Model<!>Screening Power<!>Ligand-Only Affinity Prediction<!>Visualization<!>Conclusions

<p>Structure-based drug discovery exploits knowledge of protein structures in order to design novel and potent compounds for a specific target. Protein-ligand docking is one of the main computational tools employed in the early stages of structure-based drug discovery-where more accurate methods, such as free energy calculations, 1,2 are too time consuming-to predict the binding mode and binding affinity of different ligands in a binding site. 3 The binding mode search is usually guided by a scoring function. Sometimes the scoring function has the dual purposes of finding the binding poses (docking) and predicting the proteinligand binding affinity (scoring), 4 whilst at other times different scoring functions are used for different purposes (scoring, ranking, docking, or screening).</p><p>Scoring functions can be loosely assigned to four classes: physics-based, regression-based, knowledge-based or machine-learning based. 5 Many scoring functions belonging to the first three categories have been developed over the past decades. [6][7][8][9][10] Despite their successes in reproducing the binding pose, a rapid and accurate prediction of the protein-ligand binding affinity remains a very challenging task. 11 In recent years, machine learning and deep learning scoring functions have consistently improved protein-ligand binding affinity predictions. 12 These improvements build on decades of quantitative structure-activity relationship (QSAR) modelling, where simpler representations and regressors were used. 13,14 Deep learning architectures-which are outperforming standard algorithms in image recognition and natural language processing [15][16][17][18][19] -are under active research, as demonstrated by the large number of new scoring functions based on deep learning. [20][21][22][23][24][25] In this work we explore the use of a collection of feed-forward neural networks (NNs), each computing an atomic contribution to the protein-ligand binding affinity. We show this architecture, combined with atom-centred symmetry functions (ACSFs) to capture the local chemical environment of every atom in the protein-ligand binding site, performs as well as or better than current machine learning and deep learning architectures. This particular representation-commonly employed in the development of neural-network potentials (NNPs) 26,27 -has the advantage of being translationally and rotationally invariant, unlike NN-based or CNN-based scoring functions that use an order-dependent input vector or a grid-based representations as input.</p><!><p>In order to predict the binding affinity of a ligand to a target of interest we need a description of the protein-ligand binding site that allows the key protein-ligand interactions to be learned. Ideally, this representation should depend only on the relative positions of the ligand and the protein-the representation should be invariant under translation, rotation and mirror operations. Unfortunately, some machine learning and especially deep learning scoring functions employed in computational drug discovery, do not satisfy such conditions: grid-based methods are not translationally or rotationally invariant and need extensive data augmentation, 20 while vector-based representations are often order-dependent.</p><p>Local representations of the atomic environment satisfying the ideal properties outlined above have been employed with success in quantum machine learning. 26,[28][29][30] In particular, the ACSFs originally introduced by Behler and Parrinello and further developed to build the Accurate NeurAl networK engINe for Molecular Energies (ANAKIN-ME or "ANI" for short) family of NNPs have been successful in producing accurate molecular properties. 26,27,31,32 Here we employ the ACSFs defined for the ANI family of NNPs in order to represent the protein-ligand binding site, where protein residues with at least one atom within a distance, d, from the ligand are considered.</p><p>For each atom i of element X in the system, its chemical environment can be represented by combining radial (G R α,m ) and angular (G A α,β,m ) ACSFs in a one dimensional vector, G X i = {G R α 1 ,m 1 , . . . , G A α 1 ,β 1 ,m 1 , . . . }-called the atomic environment vector (AEV). X corresponds to the element of the atom for which the AEV is being computed, while α and β denote the elements of the neighbours within a cutoff radius, R c . The ACSFs capture the atom's radial and angular chemical environment, 27 and their locality is ensured by a cutoff function:</p><p>Radial symmetry functions are given by: 26,27</p><p>where the index m runs over the set of parameters {R s } and the summation over j runs over all the atoms of element α; η R controls the width of the radial Gaussian distributions, while R s controls their radial shift. The angular symmetry function is defined as: 27</p><p>where the index m runs over the set of parameters {{R s }, {θ s }} and the summation runs over pairs of atoms of elements α and β; η A and R s have the same role of η R and R s in the radial symmetry function described above, with θ s capturing different regions of the angular environment, while ζ controls the width of the peaks of the ACSF in the angular environment. 27 The AEV G X i of atom i of element X-composed of different ACSFs in a single vectorencodes the neighbour-dependent local atomic environment of atom i of element X. This corresponds essentially to a fine-grained and flexible atom typing, in contrast to the static and arbitrary atom types employed in standard scoring functions.</p><p>Figure 1 shows schematically the components of an AEV for an atom in a system composed only of the elements H, C and O. By construction, this vector is translationally and rotationally invariant as well as invariant under the exchange of two atoms of the same element.</p><!><p>...</p><!><p>The NN architecture is implemented using PyTorch 1.5, 34 loosely following the original work of Behler and Parrinello, the ANI family of NNPs, and the TorchANI implementation. 26,27,33 It consists of n e atomic neural networks, where n e is the number of elements in the dataset.</p><p>The atomic NNs are standard feed-forward NNs with rectified linear unit (ReLU) activation functions and dropout layers. The outputs of the atomic NNs are then summed together in order to obtain the final estimate of the binding affinity.</p><p>Figure 2 shows a schematic representation of the model for a hypothetical system com-posed of two hydrogen atoms, one carbon atom and one oxygen atom. The AEVs G X i corresponding to atoms of the same element X are propagated through the same atomic NNs (with the same weights). All atomic contributions are summed together in order to get the final prediction.</p><p>Figure 2: Propagation of AEVs, G X i , through atomic NNs for the four atoms of a hypothetical system composed of two hydrogen atoms, one carbon atom and one oxygen atom. The AEVs, G X i , are constructed for each atom i of element X as described in the main text and propagated through the atomic NN of the corresponding element. All atomic contributions are finally summed together to obtain the pK prediction. Loosely adapted from Smith et al. 27 .</p><p>The idea behind the decomposition of the binding affinity into atomic contributions is essentially the one that has been proven useful for short-range energy decomposition in NNPs. The negative logarithm of the binding affinity pK = − log 10 (K/c 0 ) is proportional to the Gibbs free energy of binding pK = − 1 ln(10) ∆G bind 0 RT and therefore decomposing pK into atomic contributions corresponds to a decomposition of the Gibbs free energy. As for the total energy in NNPs, this decomposition allows the description of local contributions only, 29 but it is very effective in practice-as demonstrated by the success of NNPs in fitting high-dimensional potential energy surfaces. 26,27,30,32,35 This decomposition also appears to be very effective in generalisation amd transferability, since it works for systems much larger than the ones included in the training set. 27</p><!><p>The PDBbind dataset provides protein-ligand complexes with associated experimentally determined inhibition constants, K i , dissociation constants, K d , and IC 50 measurements (in decreasing order of preference). 36,37 This dataset is divided in two parts: the PDBbind Refined set and the PDBbind General superset. The Refined set only contains high-quality structures with associated K i or K d values, while the General set also includes structures with associated IC 50 values. A curated subset of the PDBbind Refined set is provided for comparative assessment of scoring functions (CASF). 38,38 In this work the PDBbind 2016 Refined set is used for training and validation while the CASF-2013 and CASF-2016 data sets are used for testing and comparison with other machine learning and deep learning models, as well as classical scoring functions. [36][37][38][39] The PDBbind 2016 Refined set is randomly split into training and validation sets with a 90/10 ratio. Systems present in both PDBbind and CASF datasets are removed from the training and validation sets and used only for testing. This procedure ensures that there is no exact overlap ("hard overlap") of protein-ligand complexes between the PDBbind (training/validation) and CASF (test) datasets, although some overlap with similar targets and ligands remains. [38][39][40] In order to asses this remaining "soft overlap" between training and test sets-arising from similar proteins, similar binding sites, and similar ligands-we use the subset of the PDBbind 2016 dataset proposed by Su et al. 40 .</p><p>A detailed analysis of the CASF-2013 and CASF-2016 test sets-including the distribution of the protein-ligand binding constants and of some key properties of the protein-ligand complexes-is reported by Li et al. 38 (CASF-2013) and Su et al. 39 (CASF-2016). In partic-ular, the CASF-2016 dataset is composed of 57 protein classes-with at least 90% sequence similarity-each containing 5 protein-ligand complexes. 39 The CASF-2013 dataset is smaller is size, with 65 protein classes each containing 3 protein-ligand complexes. 38 Ligand SDF or MOL2 files from the datasets were either converted to PDB files using OpenBabel 41 and parsed using MDAnalysis (for scoring and ranking) or parsed directly with OpenBabel's Python bindings (docking and screening). [42][43][44] Protein PDB files were discarded when the element column was absent or could not be parsed correctly by MD-Analysis (this never occurred for the test set). All water molecules were removed from the dataset. All the systems in the PDBbind and CASF dataset were automatically protonated using OpenBabel, 41 and given the size of the dataset the protonation state was not further assessed.</p><p>The complexity of the NN model grows quickly with the number of atomic species present in the dataset since every element requires its own atomic NN. For this reason, we adopted two different strategies: selecting only protein and ligand atoms, or mapping metal centers to a dummy atom. Additionally, we removed the few selenoproteins present in the training or validation sets. When selecting only protein and ligand atoms, the following elements remained (in order of abundance for the ligands, see Fig. S1): H, C, O, N, S, P, F, Cl, Br, I. This resulted in a total 10 atomic NNs, one for each element. When metal centers were kept (see Fig. S2), all atoms outside of the previous list were mapped to a dummy element, X.</p><p>When "hard overlaps" with CASF-2016 were removed, the final training set consisted of 3377 complexes while the validation set consisted of 376 complexes. When "hard overlaps" with CASF-2013 were removed, the final training set consisted of 3464 complexes while the validation set consisted of 385 complexes. The CASF test sets are left unchanged.</p><p>Protein-ligand complexes 4O3C and 4IGT were removed from the PDBbind Refined Set since they contain lithium, which is not supported by AutoDock Vina, 10 the classical scoring function used as baseline in this work.</p><p>The advantage of mapping metal centers to a dummy atom is that metalloproteins, which are notoriously difficult to treat with docking and classical molecular dynamics, 45,46 are supported by our method. However, our treatment has the drawback of considering all metal atoms as equivalent, irrespective of their coordination number. As more experimental data on metalloproteins becomes available, more elements could be added into the model (with an increased computational cost).</p><!><p>∆-learning is a powerful machine learning approach where the model is trained to predict the corrections to a baseline towards the target value, instead of predicting the target value itself. 47 This approach has been applied successfully to the prediction of molecular properties from quantum mechanical calculations as well as for binding affinity predictions. [47][48][49] In the context of docking scoring functions, a ∆-learning approach has the advantage of retaining the good docking power of traditional methods while significantly improving the scoring function. 48 In this work we explored the use of a ∆-learning approach in combination with the AutoDock Vina scoring function. 10 The ∆-AEScore scoring function is therefore given by: ∆-AEScore = S + ∆ where S is the standard AutoDock Vina score (in pKunits) and ∆ is the learned correction.</p><!><p>In order to compensate for the variability introduced by random weights initialization and the stochastic optimization, we investigated the use of consensus scoring in order to evaluate our models. Consensus scoring has been shown, in some cases, to improve performance across targets in structure-based virtual screening. 50,51 During training, a total of five models were randomly initialized and independently trained. Final predictions were obtained as the average protein-ligand binding affinity of the models. This technique also allows the computation of the standard deviation associated with each prediction. The benefits of consensus scoring are analysed retrospectively below.</p><!><p>Our implementation is based on open source software from the Python 52 ecosystem. This includes: TorchANI 2.1, 33 PyTorch 1.5, 34 MDAnalysis 2.0-dev, 42,43</p><!><p>The hyper parameters of our model-the number and size of layers in the elemental NNs, dropout probability, batch size, and protein-ligand distance, d-were optimized with a gridbased method and manually fine-tuned in order to maximize the Pearson's correlation coefficient between the predicted and experimental binding affinities on the validation set.</p><p>We found that a protein-ligand distance d = 3.5 Å and 256-128-64-1 feed-forward NNs performed best, when combined with a batch size of 64 and a dropout probability of 25%.</p><p>Table S1 shows the performance of the model-with consensus scoring-on the validation test for different values of d. Using a distance of d = 4.0 Å does not change the performance, compared to d = 3.5 Å. However, the larger number of protein atoms causes the computational time to be increased. Visual inspection of a selection of systems showed that the d = 3.5 Å selects the important residues in the binding site. The model's weights are optimized using the ADAM optimizer with a learning rate of 1 × 10 −4 and using PyTorch's default parameters, β 1 = 0.9 and β 2 = 0.999. 34,61 Dropout layers are usually not employed in NNPs, but our hyperparameter search shows that they increase the performance of our model by decreasing overfitting on the training set, thus improving transferability.</p><!><p>The scoring power of a scoring function measures the linear correlation between predicted and experimental binding affinities and it is usually quantified by the Pearson's correlation coefficient:</p><p>where y denotes experimental values, ŷ denotes predicted values, and • denotes the average over all experimental or predicted values. Confidence intervals (CIs) for the correlation coefficient can be obtained by bootstrapping (with 10000 bootstrap replicates), as described in the CASF evaluation.</p><!><p>In the previous section we employed consensus scoring-with five independently trained models-since this has previously been shown to improve performance. 50,51 A small performance boost is also obtained in our case, as it can be verified retrospectively.</p><p>If we consider the CASF-2016 dataset, the average correlation coefficient of the five independent models is 0.77 (minimum 0.77, maximum 0.78) while consensus scoring reaches</p><!><p>In order to assess the impact of automatic protonation using OpenBabel 53 we also trained AEScore without hydrogen atoms for both the protein and the ligand. This results in the removal of one atomic NN, thus decreasing the number of parameters in the model.</p><!><p>As mentioned above, we removed the systems appearing in the CASF-2016 and CASF-2013 benchmark datasets from the training sets (removing the so-called "hard overlap"). However, some "soft overlap"-arising from similar proteins, similar binding sites, and similar ligands-between the training and test sets remains and could therefore artificially inflate the results. This is a known problem as shown by Boyles et al. 4 and, more recently, by Su et al. 40 We see that as the overlap threshold between the training and test sets increases, the performance of our model also increases. Interestingly, a similarity threshold of 95% does not negatively affect our scoring function, in contrast with other machine learning scoring functions. 40 This trend is similar to the RF model of Su et al. 40 , which is consistently outperformed by our model. Other machine learning scoring functions evaluated by Su et al. 40 are effectively negatively affected by removing structurally redundant samples already at high thresholds.</p><p>We also found that the model with a similarity threshold of 95% (denoted AEScore 95 hereafter) seems to perform slightly better than the model trained by only removing the "hard overlap". This could be attributed to the removal of some inconsistencies in the training 10)-although differences with other top-performing methods might not be statistically significant.</p><!><p>The ranking power of a scoring function measures its ability to rank different ligandsin a given binding pose-according to their binding affinity against a particular target.</p><p>The ranking power is usually measured by three quantities: Spearman's (rank-)correlation 40 . The raw data for the RF and DT scoring functions was kindly provided by Su et al. 40 upon request. RF and DT are respectively the best and worst performing models (at the 95% similarity threshold) presented in Su et al. 40 and are consistently outperformed by AEScore. coefficient, Kendall's (rank-)correlation coefficient and the predictive index (PI). 39,62 Our scoring function AEScore has an average Spearman's correlation coefficient of 0.64 ∈ [0.54, 0.71] CI 90% . This is similar to the best classical scoring function evaluated in the CASF-2016, 39</p><!><p>AEScore has been developed with the intent of predicting the binding affinity of a given protein-ligand complex. However, scoring functions can also be used to determine correct binding poses. Therefore we evaluate the docking power of AEScore using the docking decoys provided in CASF-2016 dataset. 39 If we consider a correct binding pose as one with a root mean squared deviation (RMSD) from the crystallographic binding mode that is smaller than 2 Å, we can define the docking success rate as the percentage of targets with a good pose ranked amongst the top one, top These results are not surprising, since AEScore has been trained in order to predict the experimental binding affinity given a protein-ligand complex and has therefore never been exposed to high-RMSD binding poses (decoys). In order to use the scoring function to determine low-RMSD poses one has to train for such task. One way to train a scoring function for docking is to train a pose classifier (distinguishing low RMSD poses from high RMSD poses), 20 but this requires a change in the model architecture. Another way to tailor a machine learning scoring function for docking is to train on docking scores as done for AGL-Score, 64 but this is not particularly interesting. A third way to improve binding affinity predictions while retaining the good docking and screening power of some classical scoring functions is to use ∆-learning. 48 In this work we explore the latter approach.</p><!><p>The use of AEVs combined with a collection of feed-forward NNs has proven successful to predict protein-ligand binding affinities on the CASF-2013 and CASF-2016 benchmark datasets using exclusively elements and atomic coordinates, as demonstrated above. Unfortunately, the results of the docking power test were unexpectedly deceiving. However, it has been previously demonstrated that a ∆-learning approach can retain the good screening power of a scoring function while improving the performance in the docking and screening power tests. 48 In the ∆-learning approach, a classical scoring function is used to obtain a crude prediction of the binding affinity, which is subsequently corrected with a machine learning or deep learning scoring function. If corrections to the AutoDock Vina scoring function can be learned by our model, combining such corrections with the docking power of AutoDock Vina would provide a scoring function with both good scoring and docking powers. 48 In order to combine AutoDock Vina and the experimental data of PDBbind, AutoDock Vina scores, S, are converted to pK values using pK = − log 10 e</p><!><p>, where T = 295 K and R is the ideal gas constant. Table 1 compares our ∆-learning results on the CASF-2013 and CASF-2016 data sets with the ∆ vina RF scoring function, arguably the most successful implementation of this approach. 48 Our model performs significantly better than ∆ vina RF on the CASF-2013 dataset and comparably on the CASF-2016. It is worth noting that ∆ vina RF is the best scoring function on the scoring and ranks power tests for the CASF-2016 benchmark, and is ranking consistently amongst the top scoring functions for the docking and screening power tests. However, ∆ vina RF is calibrated on protein-ligand complexes from the PDBbind, which overlaps with ∼50% of the CASF-2016 test set and its performance might therefore have been artificially enhanced by a large overlap between the training and test sets. 39 Both ∆ vina RF and ∆-AEScore outperform the classical scoring function AutoDock Vina in the scoring power test, by a large margin. 39 Table 1: Performance of ∆-AEScore compared to the ∆ vina RF for affinity prediction on the CASF-2013 and CASF-2016 benchmarks. For ∆-AEScore the "hard overlap" between the training and both test sets is removed while for ∆ vina RF only the "hard overlap" between the training set and CASF-2013 is removed. 48,65 The best performance for each test set is underlined. RMSE values are given in pK units.</p><!><p>Training 39 However, the much higher performance compared to AEScore indicates that the protein-ligand binding site representation and the model architecture used for AEScore are amenable to ∆-learning. We thus have good scoring power-significantly better than AutoDock Vina alone-while retaining the excellent docking power of Autodock Vina.</p><!><p>Given the good success rate of ∆-AEScore in the docking power test, we wanted to evaluate ∆-AEScore in the context of virtual screening as well. The screening power test assesses the ability of a scoring function to identify true binders among a large pool of decoys. There are two types of screening power tests provided in the CASF-2016 benchmark: in forward screening, the goal is to identify the true binders for a given target, while in reverse screening, the goal is to identify a potential target for for a given active compound. 39 For the forward screening power test, ∆-AEScore ranks the best ligand among the top 1%</p><!><p>In order to test the effect of protein information in the binding affinity prediction and to elucidate possible biases in the dataset, 66 we also trained a model with only the ligand atoms (d = 0 Å). The AEVs' parameters used to describe ligand atoms are left unchanged.</p><p>For the CASF-2013 dataset we obtained an RMSE of 1.65 pK units and a Pearson's correlation of 0.70, while for the CASF-2016 dataset we obtained an RMSE of 1.49 pK units and a Pearson's correlation of 0.74 (when only protein and ligand atoms are kept and systems are automatically protonated). Fig. 8 also reports the results when hydrogen atoms are removed and when the model is trained on a dataset with a protein/ligand/pocket similarity threshold of 95% similarity with the training set.</p><p>As shown in Fig. 8 (and, equivalently, Table S2), the performance of the model in absence of protein atoms (L) is always worse than that obtained when including both ligand and protein atoms (P + L). This indicates that the model is able to exploit the additional information about the binding site provided by the protein atoms in order to improve binding affinity predictions. However, the difference is not as striking as one might expect.</p><p>The same observations apply to the ∆-learning approach, although the difference between protein-ligand (P + L) and ligand-only (L) models is less pronounced. This suggests that corrections to the AutoDock Vina scoring function mainly stem from the information about the ligand and that information about the protein target plays a minor role.</p><p>RMSE (pK units) The fact that AEScore models using only information about the ligand already perform well is in line with recent work from Boyles et al. 4 who showed that ligand features alone are predictive of the mean protein-ligand binding affinity in PDBbind. 4 Additionally, ligand information plays a significant role in affinity prediction in deep learning models as well. 51,67,68 For ligand-only predictions, AEScore is essentially learning a conformation-dependent fingerprint of the active ligand and using such information to predict the mean binding affinity of said ligand; RDKit descriptors alone, combined with a random forest model, can already achieve a Pearson's correlation coefficient of 0.71 on CASF-2013 and of 0.76 on CASF-2016, as demonstrated by Boyles et al. 4 . Our results suggest that the AEScore model presented here can use AEVs as 3D ligand fingerprints and use such information to predict the average binding affinity of a ligand in the same way RDKit descriptors allow.</p><!><p>One advantage of working with atomic coordinates directly and using a end-to-end differentiable model, is that the gradient of the loss function can be computed with respect to the atomic coordinates. This technique has been previously used to interpret CNN-based scoring functions: 69 the negative of the gradient of the loss function with respect to the atomic coordinates indicates where the model would like the atoms to "move" in order to minimise the loss function. Leveraging PyTorch's automatic differentiation tools, such gradients are readily available.</p><p>Figure 9 shows the magnitude of the gradients for ligand and protein atoms for the complexes of the CASF-2016 test set with lowest and highest absolute error. For the 3ZT2 complex-the one with lowest absolute error-the maximum gradient magnitude is 1.91 and the gradients are small everywhere, with the exception of a particular functional group of the ligand. For the 2XDL complex-the one with highest absolute error-the maximum gradient magnitude is 21.31.</p><p>The gradients of the loss function with respect to the atomic coordinates can also be employed as fictitious forces for a local geometry optimisation: atoms can be displaced along the negative of the gradient with standard optimisation techniques in order to obtain is taken directly from the references reported. The same results are also reported in Table S3, together with RMSEs and additional information about models and training datasets.</p><p>In the literature there is some confusion about the CASF benchmark and the PDBbind Core set, as indicated on the PDBbind website. 75 In Table S3 However, a quantitative and statistically sound comparison with other methods is somewhat difficult because error bars and confidence intervals are often not reported. If confidence intervals were reported by all methods, one could compute the confidence interval of the difference between two methods with convenient approximations and use this information to test significance against the null hypothesis. 76 One of the main advantages of the AEV-based approach is that it is translationally and rotationally invariant, thus removing an additional source of variability. This is not the case for scoring functions based on standard CNNs, and we expect random translations and rotations of the input protein-ligand systems to give different results, while our results would remain unchanged. Figure S4 shows the variation in CNN-based predictions as a function of the angle of rotation for a particular complex. Data augmentation with random translations and rotations has proved to be essential to prevent overfitting and significantly improve training in CNN-based scoring functions, 20 but this is computationally expensive-another advantage of our approach.</p><p>In addition to being translationally and rotationally invariant, our model also requires minimal information about the system. Only elements and atomic coordinates are needed by the model. Other methods often require additional information such as force-field parameters or specific atom types and are therefore limited by these parameters and underlying assumptions.</p><p>Compared to "classical" machine learning scoring functions, our method performs similarly to RF Score and other RF-based scoring functions. 11,73 Despite recent advances in deep learning architectures, which consistently outperform "classical" machine learning algorithms in image recognition and natural language processing, [15][16][17][18][19] RFs remain very competitive for binding affinity predictions. All top-performing machine learning and deep learning methods considered here achieve similar performance on the CASF benchmarks-as measured by</p><p>Pearson's correlation coefficient. This is likely due to the fact that errors in the experimental measurements of the binding affinity and the X-ray crystallographic coordinates of the protein-ligand complex set a theoretical upper limit on the maximal performance of scoring functions trained on such noisy data. 37 It is instructive to also compare the performance of our model with standard docking scoring functions. Here we used the AutoDock Vina 10 scoring function as implemented in smina 77 as a baseline. We see that our model clearly outperforms the Vina scoring function for protein-ligand affinity predictions, as do other machine learning and deep learning ap-proaches. This is expected, since previous studies show that standard scoring functions do not perform very well in scoring and ranking power tests. 37 The removal of the systems in the CASF test set from the PDBbind Refined set used for training is a common practice with machine learning and deep learning scoring functions and therefore ensures a fair comparison with other methods. However, it has been previously noted that the performance on the CASF set is not necessarily very indicative of the a model's ability to generalize, since this dataset samples the same regions of the chemical and target spaces as the PDBbind dataset. 40,51 In order to better evaluate the ability of a model to generalize, we tested its performance when trained on a recently developed nonredundant training set. 40 We showed in Fig. 5 that the performance of AEScore deteriorates gradually when the similarity between the training set and the test set is reduced, in contrast with many other machine learning scoring functions that are severely inhibited by removing structurally redundant samples from the training set. 40 When we tested AEScore for docking power we obtained poor results. This is not surprising since the model was trained to predict binding affinities given the correct binding pose and it was not trained explicitly to distinguish low-from high-RMSD poses. However, we showed that by combining AEScore with the classical scoring function AutoDock Vina using a ∆-learning approach improves the performance in terms of docking and screening, while maintaining good scoring and ranking performance. As already demonstrated by ∆ vina RF, this is a good approach for developing a scoring function that works well on all four tasks: scoring, ranking, docking and screening. Usually, machine learning and deep learning scoring functions work very well for scoring but not as well for docking and virtual screening, while classical scoring functions have the opposite behaviour. Figure 11 collects most of the results of AEScore and ∆-AEScore on the CASF-2016 benchmark, together with the results for ∆ vina RF and AutoDock Vina (our baseline) as reported by Su et al. 39 . We also added the best-and worst-performing scoring functions for each of the CASF-2016 benchmarks reported in Su et al. 39 , whenever these scoring function were different from ∆ vina RF or AutoDock Vina. We see that both AEScore and ∆-AEScore perform well in scoring and ranking power tests, but AEScore performance for docking is low. However, the ∆-learning approach is able to recover a good docking power (similar to the AutoDock Vina baseline) while retaining a good performance in scoring and ranking. The performance of ∆-AEScore in forward screening is rather poor as measured by EF 1% or top 1% success rate, but greatly improves for EF 5% and the top 5% success rate.</p><p>Given the good performance of our ligand-only model-which was nonetheless consistently worse than that of the protein-ligand model-it is clear that the model is extracting a lot of information from the ligand. Finding strategies to force the model to rely more on protein information could further improve the model and make it more transferable. This is a known problem [66][67][68] and strategies to force the model to rely more on the protein structure are an active area of research. 78 The advantage of using an end-to-end differentiable model is that the gradient of the scoring function with respect to the input parameters can be readily obtained by backpropagation. Since the TorchANI AEVComputer is fully differentiable and its inputs are atomic coordinates, 33 the gradient of the scoring function with respect to atomic coordinates can be computed. This can be used for visualization, which could help to understand the behaviour of the scoring function. In future iterations of the model, such gradients could be employed in the context of a local geometry optimization of the binding pose. 70 Finally, it is worth noting that we exploited the representation and architecture commonly used to develop NNP to predict a different endpoint, namely the protein-ligand binding affinity, and corrections to classical scoring functions. However, given the success of NNPs 27,32 one could use them in a MM/PBSA-or MM/GBSA-style approach 79 to directly compute the free energy of binding on more physical grounds. In fact, approaches to combine NNP with molecular mechanics for drug discovery applications are already starting to appear. 39 ) is also added for comparison. The results include 90% confidence intervals (where they were available).</p><!><p>We demonstrated that AEVs are a promising representation of the protein-ligand binding site (and of the ligand alone, for ligand-based model) amenable to machine learning-based predictions of the protein-ligand binding affinity, and of corrections to classical scoring functions. This representation is rotationally and translationally invariant and, in contrast to CNN-based scoring functions, does not require data augmentation. The results reported here for AEScore show similar or better performance than other state-of-the-art machine learning and deep learning methods on the CASF-2013 and CASF-2016 benchmarks (as well as the Core 2016 set) in binding affinity prediction. However, AEScore still presents shortcomings also found in other machine learning and deep learning scoring functions: when only AEVs for the ligand are used, the performance is still surprisingly high. This suggests that information about the ligand provides a significant contribution to the protein-ligand binding affinity prediction, as discussed in the literature. 4,51,67,68 This is particularly evident in ∆-AEScore corrections, where the difference between ligand-only and protein-ligand models is small and corrections seems to stem mainly from ligand information.</p><p>Using training sets with decreasing similarity to the test set, first introduced by Boyles et al. 4 and later by Su et al. 40 , we showed that our model is not completely hindered by the removal of systems with high similarity, but that AEScore's performance deteriorates only gradually. This is in contrast with other machine learning and deep learning scoring functions, where a performance drop is observed as soon as a similarity threshold is introduced. 4,40 This property could be useful in real drug discovery applications, where data on similar or related systems (such as a congeneric series of ligands) is acquired gradually.</p><p>In this work we did not optimise the ANI parameters for radial symmetry functions, and we did not explore the full flexibility of the angular symmetry functions. A Bayesian optimisation of the symmetry functions' hyper parameter space could lead to further improvements of the scoring function.</p>

ChemRxiv

Cu(II), Ni(II), and Zn(II) Complexes of Salan-Type Ligand Containing Ester Groups: Synthesis, Characterization, Electrochemical Properties, and In Vitro Biological Activities

A salen ligand on reduction and N-alkylation affords a novel [N2O2] chelating ligand containing ester groups [L = diethyl-2,2′-(propane-1,3-diylbis((2-hydroxy-3-methoxy benzyl)azanediyl))diacetate]. The purity of the ligand was confirmed by NMR and HPLC chromatograms. Its Cu(II), Ni(II), and Zn(II) complexes were synthesized and characterized by a combination of elemental analysis, IR, NMR, UV-Vis, and mass spectral data, and thermogravimetric analysis (TG/DTA). The magnetic moments, UV-Vis, and EPR spectral studies support square planar geometry around the Cu(II) and Ni(II) ions. A tetrahedral geometry is observed in four-coordinate zinc with bulky N-alkylated salan ligand. The redox properties of the copper complex were examined in DMSO by cyclic voltammetry. The voltammograms show quasireversible process. The interaction of metal complexes with CT DNA was investigated by UV-Vis absorption titration, ethidium bromide displacement assay, cyclic voltammetry methods, and agarose gel electrophoresis. The apparent binding constant values suggest moderate intercalative binding modes between the complexes and DNA. The in vitro antioxidant and antimicrobial potentials of the synthesized compounds were also determined.

cu(ii),_ni(ii),_and_zn(ii)_complexes_of_salan-type_ligand_containing_ester_groups:_synthesis,_charac

1. Introduction<!>2.1. Materials and Methods<!>2.2. Physical Measurements<!>2.3. Synthesis of the Ligand (L)<!>2.4. Synthesis of the Metal Complexes<!>2.4.1. [CuL]·H2O<!>2.4.2. [NiL]·H2O<!>2.4.3. [ZnL]·H2O<!>2.5.1. UV-Vis Spectroscopic Studies<!>2.5.2. Fluorescence Studies<!>2.5.3. Electrochemical Studies<!>2.6. DNA Cleavage Experiment<!>2.7. Antioxidant Property<!>2.8. Antimicrobial Activity<!>3. Results and Discussion<!>3.1. Molar Conductance<!>3.2. IR Spectra<!>3.3. NMR Spectra<!>3.4. Electronic Absorption Spectra<!>3.5. Mass Spectra<!>3.6. EPR Spectra<!>3.7. Thermal Studies<!>3.8.1. Absorption Spectroscopic Studies<!>3.8.2. Fluorescence Spectral Studies<!>3.8.3. Electrochemical Studies<!>3.9. DNA Cleavage Activity<!>3.10. Antioxidant Property<!>3.11. Antibacterial Activity<!>4. Conclusion<!>

<p>Salen metal complexes are the interest of many workers because of their applications in food industry, in the treatment of cancer [1], as antibactericide agents [2, 3], as antivirus agents [4], as fungicide agents [5], and for other biological properties [6]. The antitumor activity of salen complex arises due to its DNA binding properties. The salen complexes are conformationally flexible and adopt a variety of geometries. Also, salen metal complexes have a unique flat electron-rich aromatic surface that may facilitate their interactions with nucleic acids. Hydroxyl groups in the salen complexes act as a quinone system which would cooperate to facilitate the formation of free radicals responsible for DNA cleavage [7]. The biological properties of salen complexes are enhanced by functionalization with a variety of substituents [8–11]. When salen compounds are reduced at the imine function, the more flexible, reduced salen derivatives (salan) are obtained.</p><p>Considerable attention has been devoted to the preparation and structural characterization of metal complexes containing salen-type ligands. However, little attention has been paid to systems in which functionalized salan is used as ligands. In the present investigation N-alkylated salan complexes are used for DNA binding and antimicrobial and antioxidant properties. In continuation of our earlier works on salen-type ligands [12–14], the present investigation reports on the synthesis and spectral characterization of Cu(II), Ni(II), and Zn(II) complexes with N-alkylated salan ligand. The interaction of the metal complexes with calf thymus (CT) DNA was studied by UV-Vis and fluorescence spectroscopy and cyclic voltammetric method. The DNA cleaving nature of the compounds was tested against pUC19 DNA in the absence and presence of hydrogen peroxide. The in vitro antimicrobial activity of the compounds was assessed against various microorganisms. The antioxidant activity of the metal complexes was investigated systematically.</p><!><p>All chemicals employed for the synthesis were of analytical reagent grade and of highest purity available. o-Vanillin and 1,3-diaminopropane were purchased from Sigma Aldrich and used as received. Solvents used for spectroscopic and electrochemical studies were purified and dried by standard procedures [15]. Metal acetates were purchased from Merck. CT DNA and pUC19 DNA were purchased from GeNei, Bangalore, and used without purification. Tris-(hydroxylmethyl)-aminomethane-HCl (Tris-HCl) and ethidium bromide (EB) were obtained from HiMedia. Tris-HCl-NaCl buffer solution was prepared with double-distilled water. Tetrabutylammonium perchlorate (TBAP) was used as a supporting electrolyte for recording cyclic voltammograms.</p><!><p>Elemental analyses were recorded on a Thermo Finnigan Flash EA 1112 elemental analyzer. Molar conductance values of the complexes in DMSO were obtained on a Systronics Model 611 digital conductivity meter. Magnetic susceptibility measurements on powder samples were carried out on a Gouy balance at room temperature using mercuric tetra(thiocyanato)cobaltate (II) as the calibrant. The infrared (IR), ultraviolet-visible (UV-Vis), and emission spectra were recorded on Shimadzu 8400 S, Shimadzu UV-2450, and Shimadzu RF-5301 PC spectrophotometers, respectively. The 1H and 13C NMR of ligand in CDCl3 and zinc complex in DMSO-d6 were recorded on Bruker AV 300 MHz spectrophotometer. Electrospray ionization (ESI) mass spectral measurements were recorded on Micromass Quattro II mass spectrometer. Electron paramagnetic resonance (EPR) spectrum of the Cu complex in DMSO solution was recorded on JES-FA200 spectrometer at 300 K and at 77 K using tetracyanoethylene (TCNE, g = 2.00277) as the g marker. The thermogravimetric analysis/differential thermal analysis (TGA/DTA) was carried out in dynamic nitrogen atmosphere with a heating rate of 20°C/min using NETZSCH STA 449F3 thermal analyzer. Cyclic voltammograms were recorded on a CHI 603 C electrochemical analyzer with a three-electrode compartment.</p><!><p>The synthetic procedure of the ligand was reported in our earlier work [14]. Color: yellow oil. Yield: 60%; Analytical data. Calculated data, for [C27H38N2O8] (%): C, 62.53; H, 7.38; N, 5.40. Found (%): C, 62.67; H, 7.25; N, 5.48. IR (cm−1): 1273 (phenolic –C–O), 1195 (–C–N), and 1747 ester (–C=O). 1H NMR (CDCl3): 4.16 (H9, 4H), 3.86 (H5, H11, 10H), 3.35 (H8, 4H), 6.58–6.87 (aromatic-H, 6H), 2.63 (H6, 4H), and 1.77 (H7, 2H) δ. λ max⁡ in CH3OH, 267, 333 nm.</p><!><p>The metal complexes were isolated as follows: metal acetate (2 mmol) dissolved in 10 mL of methanol was added dropwise to a 10 mL methanolic solution of the ligand (2 mmol). The mixture was kept under reflux for 1-2 hours. The solution thus obtained gave the compound on concentration and cooling. All the compounds are soluble in DMSO. The yields were around 70–80%. The synthesis is given in Scheme 1. For the crystallization, the compounds were dissolved in different solvent mixtures and evaporated slowly at room temperature. But our attempts to crystallize the compounds were unsuccessful.</p><!><p>Colour: dark brown; anal. calcd for [Cu(C27H36N2O8)]·H2O (%): C, 54.22; H, 6.40; N, 4.68; Cu, 10.62. Found (%): C, 54.35; H, 6.45; N, 4.76; Cu, 10.67. IR (cm−1): 1265 (phenolic –C–O), 1181 (–C–N), 1749 (–C=O in ester group), 584 (M–N), and 461 (M–O). m/z: 580. λ max⁡ in DMSO 285, 349, and 527 nm. μ eff: 1.85 BM. Λm: 12.72 mho cm2 mol−1.</p><!><p>Colour: red; anal. calcd for [Ni(C27H36N2O8)]·H2O (%): C, 54.66; H, 6.46; N, 4.72; Ni, 9.89. Found (%): C, 54.71; H, 6.50; N, 4.65; Ni, 9.97. IR (cm−1): 1261 (phenolic –C–O), 1187 (–C–N), 1747 (–C=O in ester group), 551 (M–N), and 470 (M–O). 1H NMR (DMSO-d6): 4.21 (H9, 4H), 3.91 (H11, 6H), 4.37 (H5, 4H), 3.70 (H8, 4H), 6.57–6.94 (aromatic-H, 6H), 3.20 (H6, 4H), and 1.85 (H7, 2H) δ. m/z: 575. λ max⁡ in DMSO 280, 345, and 553 nm. Λm: 11.56 mho cm2 mol−1.</p><!><p>Colour: pale yellow; anal. calcd for [Zn(C27H36N2O8)]·H2O (%): C, 54.05; H, 6.38; N, 4.67; Zn, 10.90. Found (%): C, 54.12; H, 6.46; N, 4.81; Zn, 10.97. IR (cm−1): 1267 (phenolic –C–O), 1183 (–C–N), 1749 (–C=O in ester group), 547 (M–N), and 451 (M–O). 1H NMR (DMSO-d6): 4.19 (H9, 4H), 3.92 (H11, 6H), 4.35 (H5, 4H), 3.72 (H8, 4H), 6.55–6.91 (aromatic-H, 6H), 3.16 (H6, 4H), and 1.79 (H7, 2H) δ. m/z: 582. λ max⁡ in DMSO 282 and 350 nm. Λm: 13.28 mho cm2 mol−1.</p><!><p>The DNA binding experiments were performed at room temperature. A solution of CT DNA in the buffer (5 mM Tris-HCl and 50 mM NaCl) gave a ratio of UV absorbance at 260 and 280 nm of about 1.8–1.9 : 1, indicating that the CT DNA was sufficiently free from protein [16]. The concentration of DNA was measured using its extinction coefficient at 260 nm (6600 mol L−1 cm−1) [17]. Concentrated stock solutions of the compounds in DMSO were prepared and diluted suitably with the buffer to the required concentrations for all the experiments. The absorption titrations of the compounds in buffer were performed using a fixed concentration (10 μM) to which increments of the DNA stock solution were added (R = [DNA]/[complex] = 0, 2, 4, 6, 8, and 10). Compound DNA solutions were allowed to incubate for 30 min before the spectra were recorded. From the absorption data, the intrinsic binding constant, K b, was determined using the following [18]: (1)[DNA](εa−εf)=[DNA](εb−εf)+1Kb(εb−εf), where ε a, ε f, and ε b are the apparent, free, and bound compound extinction coefficients, respectively. In the plots of [DNA]/(ε a − ε f) versus [DNA], K b is given by the ratio of slope to the intercept.</p><!><p>The interaction of the synthesized compounds with DNA was further studied by ethidium bromide (EB) displacement method. The excitation wavelength was fixed at 530 nm, and the emission range was adjusted before measurements. The changes in the fluorescence intensities at 595 nm of EB-bound CT DNA in Tris-HCl buffer (pH 7.2) were measured with respect to different concentrations of the compounds (0–120 μM). The magnitude of the binding strength of the compounds with CT DNA can be calculated using linear Stern-Volmer equation [19]: (2)I0I=1+Ksv[Q], where I 0 and I are the fluorescence intensities of EB-DNA in the absence and presence of quencher, respectively, Q is the concentration of metal complex, and K sv is linear Stern-Volmer quenching constant. The relative binding tendency of the complex to CT DNA was determined by the comparison of the slope of the line in the fluorescence intensity versus complex concentration plot. The apparent binding constant (K app) was calculated using the equation K app = K EB[EB]/[complex], where [complex] is the concentration of the complex at which there is 50% reduction in the fluorescence intensity of EB, K EB = 1.0 × 107 M−1, and [EB] = 5 μM [20].</p><!><p>The cyclic voltammetric studies of the copper complex were performed with a three-electrode system of glassy carbon as working electrode, Pt wire as auxiliary electrode, and Ag/AgCl as reference electrode. The supporting electrolyte is 0.05 M TBAP in DMSO solution. The cell was maintained oxygen-free by passing dry nitrogen through the solution. The interaction of the copper complex with CT DNA has been investigated by monitoring the changes observed in the cyclic voltammogram of CuL in buffer (5 mM Tris-HCl/50 mM NaCl) with increasing amount of DNA.</p><!><p>The DNA cleavage experiment was conducted by gel electrophoresis on pUC19 DNA. The reaction mixture was prepared as follows: 1 μL of pUC19 DNA, 5 μL of the compound in DMSO, and 1 μL of H2O2 followed by dilution with buffer (50 mM Tris-HCl and 50 mM NaCl) to a total volume of 25 μL. The reaction mixture was incubated at 37°C for 1 h. The 1% agarose gel was prepared and stained using ethidium bromide. The samples were then loaded on gel after mixing with 3 μL of loading dye (0.25% bromophenol and 40% sucrose). The gel was electrophoresed at 100 V using Tris-boric acid-EDTA buffer (pH = 8.0) until the bromophenol blue reached one-third of the gel. The bands were visualized and photographed under a UV transilluminator. The experiment was also carried out in the absence of H2O2.</p><!><p>2,2′-diphenyl-1-picrylhydrazyl (DPPH•) scavenging capacity (antioxidant activity) was measured according to the following procedure [21, 22]. The concentration of DPPH• used for antioxidant activity was 50 μM. Different concentration of the ligand and metal complexes in methanol was added to DPPH• in methanol solution and kept at room temperature for 30 min in dark. The reduction of the DPPH• was monitored by observing the decrease in absorbance at 517 nm using UV-Vis spectrophotometer. The radical scavenging capacity of the antioxidant was expressed in terms of % inhibition and IC50. The capability to scavenge the DPPH• was calculated using the following [23]: (3)%  Inhibition=(A0  −  AsampleA0)×100, where A 0 is the absorbance of DPPH• in methanol solution without an antioxidant and A sample is the absorbance of DPPH• in the presence of an antioxidant. The IC50 value is the concentration of the antioxidant required to scavenge 50% DPPH• and is calculated from the inhibition curve.</p><!><p>All the synthesized compounds were screened for their antibacterial activity against gram-positive bacteria: Streptococcus pyogenes and Staphylococcus aureus, and gram-negative bacteria: Escherichia coli, Klebsiella mobilis, Aeromonas aquariorum, and Serratia marcescens, by well diffusion method [24]. Standard antibiotics, ampicillin and amoxicillin, were used as controls. Stock solutions of tested compounds were prepared in DMSO to a final concentration of 10 mg mL−1. 20 mL of sterilized agar media was poured into each presterilized Petri dish and allowed to solidify by placing it in an incubator at 37°C for an hour. 24 h culture suspension was poured and neatly swabbed with the presterilized cotton swabs. Then holes of 5 mm diameter were punched carefully using a sterile cork borer, and these wells were completely filled with the prepared L or the metal complex solutions (50 μL). These dishes were transferred to an incubator maintained at 37°C for 24 h. During this period, the test solution diffused and the growth of the inoculated microorganism was affected. The inhibition zone was developed and measured at the end of the incubation period. Experiments were performed in triplicate, and standard deviation was calculated.</p><!><p>The ligand was synthesized by three steps. In the first step Schiff's base was obtained by the condensation of o-vanillin with 1,3-diaminopropane. The Schiff base was reduced using sodium borohydride in the second step. Finally ester compound is obtained by N-alkylation reaction using 2-bromoethylacetate in the presence of potassium carbonate. The ester compound obtained as yellow oily substance on complexation with metal ions forms powdered metal complexes.</p><!><p>The molar conductance of the synthesized metal complexes was measured in DMSO at 10−3 M solution. The values were found to be in the range of 12.72–16.56 mho cm2 mol−1 suggesting the nonelectrolytic nature of the complexes.</p><!><p>The ligand (L) has two characteristically strong bands (1747 and 1205 cm−1) arising from C=O and C(O)–O stretching vibrations of ester groups (Figure S1) (see Supplementary Material available online at http://dx.doi.org/10.1155/2013/439848). These vibrations are unchanged in the spectra of the complexes indicating that the ester group of L is not involved in complexation with metal ion. The ligand shows two bands at 1379 and 1280 cm−1 corresponding to O–H bending and C–O stretching vibrations of phenolic OH group. The disappearance of O–H bending and higher shifting of C–O stretching vibrations are observed in the spectra of metal complexes suggesting that the phenolic OH group of L is involved in coordination with metal ion after deprotonation. The band at 1193 cm−1 assigned for C–N stretching in the free ligand is shifted to lower wave number in complexes. This suggests that the tertiary nitrogen atom of the ligand is involved in coordination with the metal ions. In all the complexes, a broad band that appears in the region 3400–3500 cm−1 shows the existence of uncoordinated water molecule. New bands which are not present in the ligand appeared in the ranges 505–584 cm−1 and 450–470 cm−1 in the complexes attributed to ν (M–N) and ν (M–O) vibrations, respectively. From the spectral data, the ligand coordinates to the metal ion through phenolic –O and tertiary –N atoms.</p><!><p>Formation of nickel and zinc complex is confirmed by comparing the 1H NMR of ligand and its metal complex (Table S1). The N-methylene protons (H8) of ester part of L give the singlet at 3.35 δ. The methylene protons (H6) α to amino part of L show the signal at 2.63 δ. The sharp singlet at 3.86 δ corresponds to methylene protons (H5) α to phenyl ring (Figure S2). These methylene proton signals undergo higher deshielding up to 0.1 to 0.5 δ. This demonstrates that the tertiary amine nitrogen is involved in coordination with metal ion. The ligand shows that multiple signals at 4.16 and 1.43 δ are assignable to methylene and methyl protons of ester groups. These signals are not altered in the metal complex (Table S1). This suggests that the ester group of L is free from coordination with the metal ions. The aromatic protons of L show multiple signals in the region 6.58–6.87 δ. The sharp singlet at 3.86 δ corresponds to methoxy protons. These protons undergo smaller deshielding up to 0.06 δ.</p><p>The 13C NMR spectral data of L is compared with its ZnL complex (Figure S3 and Figure 1). The 13C NMR signal for ester group of L (C-9) is not altered in the zinc complex (Table S2). This suggests that the ester group of L is free from coordination with the metal ion. The signals for carbon atoms adjacent to nitrogen (C-7 and C-13) are observed at 54.20 and 45.37 δ, respectively. These signals are shifted to lower value in the metal complexes. Similarly, the carbon atom adjacent to phenolic oxygen (C-1) of L is shifted to higher value in the ZnL complex. The shifts in the positions of the carbon atoms adjacent to nitrogen and phenolic oxygen clearly demonstrate the bonding of the two nitrogen atoms of tertiary amine and two oxygen atoms of phenol to the Zn(II) ion forming tetrahedral geometry.</p><!><p>The UV-Vis absorption spectra of L and its metal complexes in DMSO were recorded at room temperature (Figure S4). The absorption spectrum of L shows bands at 267 and 335 nm, which are due to π → π* transitions of phenyl ring and H-bonding induced changes of OH proton-donor aromatic molecules and amine NH (intraligand charge transfer band), respectively. The spectrum of CuL (Figure S5) displaying the band at 527 nm is assigned to 2 B 1g →2 A 1g transition confirming the square planar geometry of the CuL (Table 1). The magnetic moment value for CuL (1.85 BM) is consistent with the square planar Cu(II) system. The NiL complex showed absorption at 553 nm ascribed to d-d transition (1 A 1g → 1 A 2g) which supports the square planar geometry around Ni(II) ion [25].</p><!><p>ESI mass spectra of all the metal complexes support the proposed structure of the complexes. The copper complex shows main peak at m/z 580 corresponding to the molecular weight of the complex (Figure S6). The fragmentation peaks of copper complex are observed at m/z 379 and 491. The molecular ion and fragmentation peaks have half intensity peaks due to isotopic distributions of copper (63Cu and 65Cu) [26, 27]. The spectral result shows that metal complexes are monomeric in nature and the metal to ligand ratio is 1 : 1. Nickel and zinc complexes are the same as copper, supported by analytical and spectral analysis.</p><!><p>The X-band EPR spectrum of the copper complex was recorded at 300 K and at 77 K using TCNE as the g marker (Figure S7). The absence of a half-field signal at 1600 G due to the m s = ±2 transitions ruling out any Cu-Cu interaction suggests the monomeric nature of the CuL complex. The observed g values are g II(2.29) > g ⊥(2.11) > g e (2.0027), suggesting the unpaired electron is in the d x2−y2   orbital (Table 2). The g II/A II value calculated for CuL (138 cm) lies between 90 and 140 cm indicating a square planar structure around the Cu (II) ion [28–30]. The g II value of 2.29 for the CuL complex indicates the covalent nature of the metal-ligand bond. The g values are related to exchange interaction coupling constant (G) by the expression G = (g II − 2.0027)/(g ⊥ − 2.0027). If G < 4, the ligand forming the copper complex is regarded as a strong-field ligand. For the present square planar complex, G = 2.67 indicates that the ligand is strong field and the metal-ligand bonding in the complex is covalent [31].</p><p>The bonding parameters α 2, β 2, and γ 2 which may be regarded as covalency of the in-plane σ bond, in-plane π bond, and out-of-plane π bond, respectively, were evaluated from the following expressions [32]: (4)α2=−AII0.036+(gII−2.0027)+37(g⊥−2.0027)+0.04, (5)β2  =  (gII−2.0027)E−8λα2, (6)γ2  =  (g⊥−2.0027)E−2λα2, where λ = −828 cm−1 for Cu(II) d9 system and E is the electronic transition energy. The α 2 value of 0.83 for CuL demonstrates that the complex has covalent character in the ligand environment. The observed β 2 and γ 2 values indicate that there is an interaction in the in-plane π bonding between the metal ion and ligand. This is also confirmed by orbital reduction factors, K II and K ⊥: K II = α 2 β 2 and K ⊥ = α 2 γ 2. In the present investigation the trend K II < K ⊥ for the copper complex implies a considerable in-plane π bonding is between the metal ion and the ligand [33].</p><!><p>The metal complexes show gradual loss in weight due to the decomposition with increasing temperature (Figure S8). The thermogram shows four decomposition steps within the temperature range of 25–1000°C. In the first step upto 100°C, the mass loss (3–4.2%) corresponds to loss of lattice water molecule. In the second step (100–275°C), the mass loss of 22–25% corresponds to removal of the ester and methoxy groups with evolution of CO2 gas. The third step of decomposition is noticed in the temperature range 275–450°C with loss of 28–30% due to the removal of the amino part of the ligand in the complexes. The fourth stage of decomposition occurs in the range 450–1000°C, which corresponds to the removal of the remaining part of the ligand leaving metal oxide as a residue.</p><!><p>All the metal complexes show intraligand (π → π*) transition in the region 270–280 nm. On addition of DNA, this band of the complexes was affected resulting in the tendency of hypochromism lying in the range 20–32% and a slight bathochromic shift in the range of 1.6-1.7 nm (Figure 2). These phenomena indicate that the complexes probably interact with CT DNA by intercalation binding mode. The extent of hypochromism is commonly consistent with the strength of intercalative interaction [34]. In order to study the binding ability of the compounds with CT DNA, the binding constant, K b, was determined. The K b values of the ligand (1.23 × 105 M−1) and its copper complex (1.00 × 105 M−1) are comparable. But the K b values are found to be lower than those reported for typical intercalators (for ethidium bromide and [Ru(Phen)2(dppz)]2+; the binding constants have been found to be of the order 1.4 × 106 and >106 M−1) [35]. The K b values indicate that the binding strength of the N-alkylated salan ligand (L) and CuL with DNA is stronger than that of salen, salan [14], NiL, and ZnL (Table 1).</p><!><p>The EB fluorescence displacement experiment has been widely used to investigate the interaction of metal complexes with DNA. The EB shows weak fluorescence in buffer solution. The fluorescence intensity of EB in presence of DNA can be greatly enhanced due to intercalation with DNA [36]. On addition of metal complexes (0–120 μM) to DNA-EB mixture, the metal complex competes with EB to bind with DNA. This leads to a decrease in the binding sites of DNA available for EB, and hence quenching of fluorescence intensity of EB-DNA mixture occurs (Figure 3). The quenching plot illustrates that the quenching of ethidium bromide bound to DNA by metal complexes is in agreement with the linear Stern-Volmer equation. The value of K sv for CuL, NiL, and ZnL is found to be 8.44 × 103, 6.58 × 103, and 7.89 × 103, respectively. The apparent binding constant (K app) values obtained for the CuL, NiL, and ZnL compounds are found to be 6.25 × 105, 5 × 105, and 4 × 105, respectively (Table 1). Furthermore, the quenching constants and binding constants calculated for the complexes suggest that the interaction of all the compounds with DNA occurs through intercalation. The DNA binding abilities of the complexes follow the order Cu(II) > Ni(II) > Zn(II), which is in conformity with the trend in DNA binding affinities obtained from absorption spectral studies.</p><p>Binding Analysis. The equilibrium binding constant and the number of binding sites can be analyzed according to the Scatchard equation [37, 38]: (7)log⁡I0−II=log⁡ Kbin+n log⁡[Q], where K bin is the binding constant of complex with DNA and n is the number of binding sites. From the plot of log (I 0 − I)/I versus log⁡[Q], the number of binding sites and binding constant have been obtained (Figure 4). The value of n is around one for all the compounds indicating the existence of only one independent class of binding sites for the metal complexes on DNA (Table 1). The values of K sv and K bin suggest that the complexes interact strongly with DNA.</p><!><p>The redox behaviour of the Cu(II) complex in DMSO was examined by means of cyclic voltammetry (potential range −1 V to +1 V) with different scan rates from 25 to 125 mVs−1 (Figure 5). The cyclic voltammogram shows a well-defined quasireversible peak for the redox couple Cu(III)/Cu(II) [Epa = 0.604 V and Epc = 0.260 V]. In the negative potential range copper shows irreversible cathodic peak at −0.675 V [Cu(II)/Cu(I)] with the scan rate of 100 mVs−1 (Table S3). The limiting peak to peak separation (ΔEp) for Cu(III)/Cu(II) process is greater than 59 mV which revealed that the couple is quasireversible. The ratio of anodic to cathodic peak current value is 1.1 demonstrating the simple one-electron process. Further, the anodic peak is shifted towards positive potential value, and the cathodic peak is shifted towards negative potential with a function of scan rate 25 to 125 mVs−1 which supports quasireversible process.</p><p>The cyclic voltammetric technique provides information about interaction between the metal complexes and DNA. DNA is denatured in DMSO medium, so we recorded the CV of copper complex in Tris-HCl buffer containing 10% DMSO. In Buffer medium, the copper complex shows a Cu(II)/Cu(I) couple with Epc at −0.438 V and Epa at 0.137 V (Figure S9). The separations of anodic and cathodic peaks (ΔEp) are found to be 0.575 V indicating quasireversible one-electron redox process (Table S4). In the presence of CT DNA with R = 10 (R = [DNA]/[complex]) both the anodic and cathodic peak currents are decreased with shifting of potential values indicating that there exist interactions between copper complex and CT DNA. The drop of the voltammetric current in the presence of CT DNA is due to slow diffusion of the copper complex bound to CT DNA. The formal potential, E 1/2, taken as the average of Epc and Epa shifts slightly towards the positive side on binding to DNA which suggests that copper complex binds intercalatively to CT DNA [39].</p><!><p>The cleavage of pUC19 DNA induced by the metal complexes in the presence of H2O2 is shown in Figure S10. In the absence of the complex (Lane 1), DNA remains in the supercoiled form. Incubation of DNA with copper complex (Lane 2) leads to its conversion to form II and form III. This indicates that copper complex has the ability to cleave pUC19 DNA in presence of oxidant H2O2. The probable reason may be the oxidation of deoxyribose moiety by hydroxyl free radicals followed by the hydrolytic cleavage of the sugar phosphate backbone. The cleavage efficiency was measured in terms of the ability of the complex to convert the supercoiled form to open circular form. The ability of nickel and zinc complexes (lane 3 and lane 4) in the interconversion of supercoiled form to open circular form (form I to form II) is less when compared to the copper complex. In the absence of H2O2, the synthesized complexes did not show any effect towards the cleavage of DNA.</p><!><p>The antioxidant activity of the ligand and the metal complexes was measured in terms of their hydrogen donating or radical scavenging capability by DPPH assay method. Upon addition of metal complexes, the reduction of DPPH radical is monitored by the decrease of the absorbance of its radical at 517 nm (Figure 6). The absorbance decreases as a result of color changes from purple to yellow as the radical is scavenged by antioxidants. The 50% inhibitory concentration (IC50) values of L, CuL, NiL, and ZnL are 103, 1019, 771, and 1429 μM, respectively. The higher free radical scavenging activity of the ligand may be due to the presence of free phenolic –OH groups. The IC50 values of synthesized compounds are much higher than the positive control like ascorbic acid, 11.55 μM [40].</p><!><p>The in vitro antibacterial activities of the synthesized compounds were tested against six human pathogenic microorganisms (gram-positive bacteria: Streptococcus pyogenes and Staphylococcus aureus; gram-negative bacteria: Escherichia Coli, Klebsiella mobilis, Aeromonas aquariorum, and Serratia marcescens) by well diffusion method using ampicillin and amoxicillin as standards. The susceptibility of the strains of bacteria towards the present compounds was judged by measuring the size of inhibition diameter (Figure S11). The comparison of the antimicrobial activity of the synthesized compounds and the known antibiotics showed that the metal complexes were more effective than the ligand or metal salts but less active than the controls against all the bacteria tested. The bulky N-alkylated ligand on chelation to the metal cation reduces the polarity of the metal ion due to the ligand orbital overlap with the metal orbitals, resulting in a delocalization of positive charge. This increases the lipophilic character of the metal chelate and favors its permeation through the lipoid layer of the bacterial membranes. Cu(II) complex has higher antibacterial activity against Streptococcus pyogenes and Escherichia Coli than the other metal complexes. Zn(II) complex has higher activity against Klebsiella mobilis, Aeromonas aquariorum. Ni(II) complex is found to have moderate activity towards the bacteria tested.</p><!><p>Salan-type ligand containing ester groups was synthesized. The N-alkylated salan was used to prepare Cu(II), Ni(II), and Zn(II) complexes. The synthesized compounds were characterized by spectral and analytical techniques. Spectral studies reveal that the ligand coordinates to the metal ion through the phenolic –O and tertiary –N atoms. The presence of ester groups in tertiary –N leads to distortion from the regular square planar geometry of the complexes. The cyclic voltammogram of copper complex reveals that the complex exhibits well-defined quasireversible Cu(III)/Cu(II) couple along with Cu(II) irreversible process. The synthesized compounds bind to CT DNA through intercalation mode. Copper complex has been found to promote cleavage of pUC19 DNA from the super coiled form to nicked form in presence of H2O2. The metal complexes show higher bacterial activity than the ligand. The ligand shows more effective free radical scavenger activity than the metal complexes.</p><!><p>Analytical and spectral data relevant to this article are given in the supplementary materials as Figures S1-S11 (IR, UV-Vis., 1H & 13C-NMR, ESI-Mass, EPR, TGA, CV spectra, DNA cleavage Pattern, antimicrobial activity) and Tables S1-S4 (NMR and CV data).</p><p>Click here for additional data file.</p>

PubMed Open Access